Clinical Biomarker For Alzheimer

Neuromethods 137
Robert Perneczky Editor
Biomarkers
for Preclinical
Alzheimer’s
Disease
Neuromethods
Series Editor
Wolfgang Walz
University of Saskatchewan
Saskatoon, SK, Canada
For further volumes:

http://www.springer.com/series/7657
Biomarkers for Preclinical
Alzheimer’s Disease
Edited by
Robert Perneczky
Department of Psychiatry and Psychotherapy,
Ludwig-Maximilians-Universität München,
Munich, Germany
Editor
Robert Perneczky
Department of Psychiatry and Psychotherapy
Ludwig-Maximilians-Universität München
Munich, Germany
ISSN 0893-2336 ISSN 1940-6045 (electronic)

Neuromethods
ISBN 978-1-4939-7673-7 ISBN 978-1-4939-7674-4 (eBook)
https://doi.org/10.1007/978-1-4939-7674-4
Library of Congress Control Number: 2017964263
© Springer Science+Business Media, LLC 2018

This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part of the material is
concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction
on microfilms or in any other physical way, and transmission or information storage and retrieval, electronic adaptation,
computer software, or by similar or dissimilar methodology now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication does not
imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and
regulations and therefore free for general use.
The publisher, the authors and the editors are safe to assume that the advice and information in this book are believed
to be true and accurate at the date of publication. Neither the publisher nor the authors or the editors give a warranty,
express or implied, with respect to the material contained herein or for any errors or omissions that may have been
made. The publisher remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Printed on acid-free paper
This Humana Press imprint is published by Springer Nature

The registered company is Springer Science+Business Media, LLC
The registered company address is: 233 Spring Street, New York, NY 10013, U.S.A.
Series Preface
Experimental life sciences have two basic foundations: concepts and tools. The Neuromethods
series focuses on the tools and techniques unique to the investigation of the nervous system
and excitable cells. It will not, however, shortchange the concept side of things as care has
been taken to integrate these tools within the context of the concepts and questions under
investigation. In this way, the series is unique in that it not only collects protocols but also
includes theoretical background information and critiques which led to the methods and
their development. Thus it gives the reader a better understanding of the origin of the
techniques and their potential future development. The Neuromethods publishing program
strikes a balance between recent and exciting developments like those concerning new ani-
mal models of disease, imaging, in vivo methods, and more established techniques, includ-
ing, for example, immunocytochemistry and electrophysiological technologies. New
trainees in neurosciences still need a sound footing in these older methods in order to apply
a critical approach to their results.
Under the guidance of its founders, Alan Boulton and Glen Baker, the Neuromethods
series has been a success since its first volume published through Humana Press in 1985.
The series continues to flourish through many changes over the years. It is now published
under the umbrella of Springer Protocols. While methods involving brain research have
changed a lot since the series started, the publishing environment and technology have
changed even more radically. Neuromethods has the distinct layout and style of the Springer
Protocols program, designed specifically for readability and ease of reference in a laboratory
setting.
The careful application of methods is potentially the most important step in the process
of scientific inquiry. In the past, new methodologies led the way in developing new disci-
plines in the biological and medical sciences. For example, Physiology emerged out of
Anatomy in the nineteenth century by harnessing new methods based on the newly discov-
ered phenomenon of electricity. Nowadays, the relationships between disciplines and meth-
ods are more complex. Methods are now widely shared between disciplines and research
areas. New developments in electronic publishing make it possible for scientists that
encounter new methods to quickly find sources of information electronically. The design of
individual volumes and chapters in this series takes this new access technology into account.
Springer Protocols makes it possible to download single protocols separately. In addition,
Springer makes its print-on-demand technology available globally. A print copy can there-
fore be acquired quickly and for a competitive price anywhere in the world.
Saskatoon, Canada Wolfgang Walz
v
Preface
The constantly increasing number of individuals with dementia due to Alzheimer’s disease
(AD) poses a significant financial and emotional burden on the affected families and the
global society. Therapeutic strategies designed to treat symptoms or alter the disease course
have so far failed to make a positive impact, despite billions of US dollars of R&D invest-
ments and massive efforts of the industry and academic communities. These failures in
treatment have led many to believe that symptomatic AD, including its earliest clinical
stages, is resistant to drug interventions.
The recently increased focus on biomarkers to diagnose early disease is fuelled by the
hope of identifying a therapeutic window, in which the brain is still largely intact and there-
fore amenable to treatment effects. The ability of biomarkers to adequately define the pre-
clinical (i.e., at risk of AD dementia) disease stage may ultimately allow novel or repurposed
drugs to finally achieve clinically meaningful results for the affected individuals and to help
to prevent dementia and associated disability.
This book discusses the usefulness of established biomarkers (imaging, fluid, and genetic)
to detect preclinical AD, providing detailed protocols for state-of-the-art biomarker mea-
surement and analysis. It is also explained how cutting-edge technology is used to develop
novel improved biomarkers for the earliest, presymptomatic stages of AD. Our book aims to
provide a comprehensive overview specifically of biomarkers for the earliest detectable dis-
ease stages. All previous publications cover biomarkers for clinical, symptomatic AD.
This publication comprises five parts: In Part I (Chaps. 1 and 2) we explain why AD is
one of the major challenges for the global societies and healthcare systems, which highlights
the urgent need for improved approaches for early diagnosis enabling disease prevention. In
Part II (Chaps. 3–5) clinical and research concepts are presented, which are important to
improve early recognition of AD. Part III (Chaps. 6–10) provides a comprehensive over-
view of methods currently used in the AD diagnostic work-up, and it is explained how these
methods can be applied to preclinical disease. In Part IV (Chaps. 11–15) cutting-edge tech-
nology innovations and their value for early AD diagnosis are discussed. Finally, in Part V
(Chaps. 16–18), we consider important ethical considerations in relation to biomarker-based
early diagnosis, and we also discuss the meaningfulness of biomarker endpoints in clinical
AD research.
The book is targeted at individuals with an interest in the use of advanced biomarker
strategies to significantly improve the early diagnosis of AD, and thereby to accelerate the
development of effective, disease-modifying drugs. This includes researchers, clinicians,
and those interested in regulatory and medical affairs, both from academia and industry. We
wish to present biomarker development approaches as a strategy for the study of AD with
the hope and expectation that the results will translate into more effective diagnosis and
treatment and improved public health policies. We expect this book to complement other
excellent volumes and monographs on AD that cover basic science or clinical aspects of the
disease.
London, UK Robert Perneczky
vii
Contents
Series Preface. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . v
Preface. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vii
Contributors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xi
Part I Alzheimer’s Disease as a Societal Challenge
1 Worldwide Economic Costs and Societal Burden of Dementia . . . . . . . . . . . . . . . . 3

Junfang Xu and Chengxuan Qiu
2 Epidemiology and Risk Factors of Alzheimer’s Disease: A Focus on Diet . . . . . . . . 15
Cécilia Samieri
Part II Diagnostic Concepts
3 The Dimensional Structure of Subjective Cognitive Decline . . . . . . . . . . . . . . . . . . 45

Miguel A. Fernández-Blázquez, Marina Ávila-Villanueva,
and Miguel Medina
4 Deriving and Testing the Validity of Cognitive Reserve Candidates . . . . . . . . . . . . 63
Yaakov Stern and Christian Habeck
5 Methods for Pathological Classification of Alzheimer’s Disease . . . . . . . . . . . . . . . 71
Johannes Attems, Kirsty E. McAleese, and Lauren Walker
Part III Established Markers of Pathophysiology and Risk
6 Cerebrospinal Fluid Biomarkers of Preclinical Alzheimer’s Disease . . . . . . . . . . . . . 93

Panagiotis Alexopoulos and Chaido Sirinian
7 Brain Structural Imaging in Alzheimer’s Disease . . . . . . . . . . . . . . . . . . . . . . . . . . 107
Sven Haller, Davide Zanchi, Cristelle Rodriguez,
and Panteleimon Giannakopoulos
8 Brain Functional Imaging in Preclinical Alzheimer’s Disease . . . . . . . . . . . . . . . . . . 119
Peter Häussermann, Thorsten Bartsch, and Oliver Granert
9 Imaging of Brain Amyloid Load in Early Alzheimer’s Disease . . . . . . . . . . . . . . . . . 149
Timo Grimmer and Jennifer Grace Perryman
10 Functional EEG Connectivity Alterations in Alzheimer’s Disease . . . . . . . . . . . . . . 157
Florian Hatz and Peter Fuhr
Part IV Biomarker Discovery
11 A Selected Reaction Monitoring Protocol for the Measurement

of sTREM2 in Cerebrospinal Fluid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 169
Amanda J. Heslegrave, Wendy E. Heywood, Kevin M. Mills,
and Henrik Zetterberg
ix
x Contents
12 Soluble Amyloid Precursor Proteins in Blood: Methods and Challenges . . . . . . . . . 179

Robert Perneczky and Panagiotis Alexopoulos
13 Tau Imaging in Preclinical Alzheimer’s Disease . . . . . . . . . . . . . . . . . . . . . . . . . . . 189
Paul Edison
14 Retinal Imaging in Early Alzheimer’s Disease . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 199
Tom MacGillivray, Sarah McGrory, Tom Pearson, and James Cameron
15 In Vivo Volumetry of the Cholinergic Basal Forebrain . . . . . . . . . . . . . . . . . . . . . . 213
Michel J. Grothe, Ingo Kilimann, Lea Grinberg, Helmut Heinsen,
and Stefan Teipel
Part V Patient Benefit and Ethical Considerations

16 Clinical Meaningfulness of Biomarker Endpoints in Alzheimer’s
Disease Research . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 235
Kok Pin Ng, Tharick A. Pascoal, Xiaofeng Li, Pedro Rosa-Neto,
and Serge Gauthier
17 The Ethics of Biomarker-Based Preclinical Diagnosis of Alzheimer’s Disease . . . . . 249
Alexander F. Kurz and Nicola T. Lautenschlager
18 Shared Decision-Making and Important Medical and Social Decisions
in the Context of Early Diagnosis of Alzheimer’s Disease . . . . . . . . . . . . . . . . . . . . 259
Katharina Bronner and Johannes Hamann
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 271
Contributors
Marina Ávila-Villanueva • Alzheimer Disease Research Unit, CIEN Foundation, Carlos

III Institute of Health, Queen Sofía Foundation Alzheimer Center, Madrid, Spain
Panagiotis Alexopoulos • Department of Psychiatry, University Hospital of Patras,
University of Patras, Patras, Greece; Department of Psychiatry and Psychotherapy,
Klinikum rechts der Isar, Technical University of Munich, Munich, Germany
Johannes Attems • Institute for Neuroscience, Newcastle University, Newcastle upon
Tyne, UK
Thorsten Bartsch • Department of Neurology, Memory Disorders and Plasticity Group,
Dementia and Alzheimer’s Clinic, University Hospital Schleswig-Holstein, Kiel,
Germany
Katharina Bronner • Department of Psychiatry and Psychotherapy, Technische
Universität München, Munich, Germany
James Cameron • Centre for Clinical Brain Sciences, University of Edinburgh, Edinburgh,
UK
Paul Edison • Division of Brain Sciences, Department of Medicine, Neurology Imaging
Unit, Imperial College London, Hammersmith Hospital, London, UK; Department of
Psychological Medicine, Cardiff University School of Medicine, Cardiff, UK
Miguel A. Fernández-Blázquez • Alzheimer Disease Research Unit, CIEN Foundation,
Carlos III Institute of Health, Queen Sofía Foundation Alzheimer Center, Madrid, Spain
Peter Fuhr • Section of Clinical Neurophysiology, Department of Neurology,
Hospital of the University of Basel, Basel, Switzerland
Serge Gauthier • The McGill University Research Centre for Studies in Aging,
McGill University, Montreal, QC, Canada
Panteleimon Giannakopoulos • Department of Psychiatry, Faculty of Medicine,
University of Geneva, Geneva, Switzerland; Division of Institutional Measures,
Medical Direction, University Hospitals of Geneva, Belle-Idée, Chêne-Bourg, Switzerland
Oliver Granert • Department of Neurology, Memory Disorders and Plasticity Group,
Dementia and Alzheimer’s Clinic, University Hospital Schleswig-Holstein, Kiel,
Germany
Timo Grimmer • Centre for Cognitive Disorders, Department of Psychiatry
and Psychotherapy, Klinikum rechts der Isar, Technical University of Munich, Munich,
Germany
Lea Grinberg • Department of Neurology, University of California, San Francisco, CA,
USA
Michel J. Grothe • German Center for Neurodegenerative Diseases (DZNE), Rostock,
Germany
Christian Habeck • Cognitive Neuroscience Division, Department of Neurology,
College of Physicians and Surgeons, Columbia University, New York, NY, USA
Sven Haller • Affidea Carouge Radiologic Diagnostic Center, Geneva, Switzerland;
Department of Surgical Sciences, Radiology, Uppsala University, Uppsala, Sweden;
Department of Neuroradiology, University Hospital Freiburg, Freiburg, Germany;
Department of Neuroradiology, Faculty of Medicine, University of Geneva, Geneva,
Switzerland
xi
xii Contributors
Johannes Hamann • Department of Psychiatry and Psychotherapy, Technische Universität

München, Munich, Germany
Florian Hatz • Section of Clinical Neurophysiology, Department of Neurology,
Hospital of the University of Basel, Basel, Switzerland
Peter Häussermann • Department of Geriatric Psychiatry, LVR Klinik Köln, Academic
Teaching Hospital, University of Cologne, Köln, Germany
Helmut Heinsen • Universidade de Sao Paulo Faculdade de Medicina, Sao Paulo, Brazil;
Department of Psychiatry, Psychosomatics and Psychotherapy, Julius-Maximilians-
University, Würzburg, Germany
Amanda J. Heslegrave • Department of Molecular Neuroscience, UCL Institute
of Neurology, London, UK
Wendy E. Heywood • Centre for Translational Omics, University College London
Institute of ChildHealth, London, UK
Ingo Kilimann • German Center for Neurodegenerative Diseases (DZNE), Rostock,
Germany; Department of Psychosomatic Medicine, University of Rostock, Rostock,
Germany
Alexander F. Kurz • Department of Psychiatry and Psychotherapy, Klinikum rechts der
Isar, Technische Universität München, Munich, Germany
Nicola T. Lautenschlager • Academic Unit for Psychiatry of Old Age, Department
of Psychiatry, The University of Melbourne and NorthWestern Mental Health,
Mental Health, Melbourne, Australia
Xiaofeng Li • The McGill University Research Centre for Studies in Aging,
McGill University, Montreal, QC, Canada; Department of Neurology,
The Second Affiliated Hospital of Chongqing Medical University, Chongqing,
People’s Republic of China
Tom MacGillivray • Centre for Clinical Brain Sciences, University of Edinburgh,
Edinburgh, UK
Kirsty E. McAleese • Institute for Neuroscience, Newcastle University, Newcastle upon
Tyne, UK
Sarah McGrory • Centre for Cognitive Ageing and Cognitive Epidemiology,
University of Edinburgh, Edinburgh, UK
Miguel Medina • Alzheimer Disease Research Unit, CIEN Foundation, Carlos III
Institute of Health, Queen Sofía Foundation Alzheimer Center, Madrid, Spain;
CIBERNED (Network Center for Biomedical Research in Neurodegenerative Diseases),
Universidad Autónoma de Madrid, Madrid, Spain
Kevin M. Mills • Centre for Translational Omics, University College London Institute
of ChildHealth, London, UK
Kok Pin Ng • The McGill University Research Centre for Studies in Aging,
McGill University, Montreal, QC, Canada; Department of Neurology,
National Neuroscience Institute, Singapore, Singapore
Tharick A. Pascoal • The McGill University Research Centre for Studies in Aging,
Tom Pearson • Centre for Clinical Brain Sciences, University of Edinburgh,
Edinburgh, UK
Robert Perneczky • Department of Psychiatry and Psychotherapy, Ludwig-Maximilians-
Universität München, Munich, Germany
Jennifer Grace Perryman • Ludwig-Maximilians-University Munich, Munich, Germany
Contributors xiii
Chengxuan Qiu • Aging Research Center, Department of Neurobiology, Care Sciences

and Society, Karolinska Institutet-Stockholm University, Stockholm, Sweden
Cristelle Rodriguez • Division of Institutional Measures, Medical Direction,
University Hospitals of Geneva, Belle-Idée, Chêne-Bourg, Switzerland
Pedro Rosa-Neto • The McGill University Research Centre for Studies in Aging,
Cécilia Samieri • INSERM, Bordeaux Population Health Research Center, UMR 1219,
University of Bordeaux, Bordeaux, France
Chaido Sirinian • Clinical and Molecular Oncology Laboratory, Division of Oncology,
Medical Faculty, School of Health Sciences, University of Patras, Patras, Greece
Yaakov Stern • Cognitive Neuroscience Division, Department of Neurology,
College of Physicians and Surgeons, Columbia University, New York, NY, USA
Stefan Teipel • German Center for Neurodegenerative Diseases (DZNE), Rostock,
Germany; Department of Psychosomatic Medicine, University of Rostock, Rostock, Germany
Lauren Walker • Institute for Neuroscience, Newcastle University, Newcastle upon Tyne, UK
Junfang Xu • Research Center for Public Health, School of Medicine, Tsinghua University,
Beijing, People’s Republic of China
Davide Zanchi • Department of Psychiatry (UPK), University of Basel, Basel, Switzerland
Henrik Zetterberg • Department of Molecular Neuroscience, UCL Institute
of Neurology, London, UK; Clinical Neurochemistry Laboratory, Department
of Psychiatry and Neurochemistry, Institute of Neuroscience and Physiology,
The Sahlgrenska Academy, University of Gothenburg, Mölndal, Sweden
Part I
Alzheimer’s Disease as a Societal Challenge

Chapter 1
Worldwide Economic Costs and Societal Burden

of Dementia
Junfang Xu and Chengxuan Qiu
Abstract
Dementia is the principal cause of functional dependence and institutionalization in the elderly and has already
posed tremendous economic burden on the aging society. Thus, dementia has been recognized by the World
Health Organization as a global public health priority. Here, we briefly introduce the conceptual and method-
ological issues of the cost-of-illness studies (COI) of dementia, summarize the major literature regarding the
economic costs of dementia, and identify some key issues that need be addressed in future COI studies of
dementia. We identified 17 COI studies that estimated the costs of dementia in different countries, in which the
annual total costs per patient with dementia varied from US$2935 to US$64168. Differences in methodology,
data sources, cost categories assessed, and severity of patients employed in the COI studies are likely to contrib-
ute to the substantial variations in cost estimates of dementia. Thus, to increase the comparability of findings
from COI studies, the methodological issues need to be harmonized and standardized, and consensus on the
methodological principles in the COI studies would be essential. Further well-designed COI studies of demen-
tia could provide critical evidence to help develop health policy and improve allocation of healthcare resources.
Key words Dementia, Aging, Care Costs, Economic Burden, Societal Burden, Cost-of-Illness Studies
1 Introduction
Dementia is defined as a clinical syndrome characterized by pro-

gressive deteriorations in multiple cognitive domains (e.g., memory
and executive function) that are severe enough to interfere with
daily social and professional functioning [1]. Clinically, Alzheimer’s
disease is the most common form of dementia, followed by vascular
dementia, accounting for 50–70% and 15–20% of all dementia
cases, respectively. However, population-based neuroimaging and
autopsy-verified studies have revealed that pure Alzheimer’s demen-
tia or pure vascular dementia cases are virtually rare, whereas a large
majority of dementia cases are attributable to mixed vascular and
neurodegenerative pathology in the brain. Dementia, as a clinical
syndrome in general, is a major cause of functional dependence and
institutionalization in older people, which has enormous economic
and societal impact on the aging society.
Robert Perneczky (ed.), Biomarkers for Preclinical Alzheimer’s Disease, Neuromethods, vol. 137,
https://doi.org/10.1007/978-1-4939-7674-4_1, © Springer Science+Business Media, LLC 2018
3
4 Junfang Xu and Chengxuan Qiu
2 Dementia: A Priority of Global Public Health
Population aging is a worldwide universal phenomenon. Thus,

dementia or Alzheimer’s disease as a strongly age-dependent dis-
order has posed huge impact on public health, healthcare, and
social service systems in all countries across the world. The World
Alzheimer Report 2015 estimated that worldwide approximately
47.5 million people were living with dementia, and the number is
projected to reach 75 million in 2030 and 132 million in 2050
[2]. The global costs of dementia care were estimated to increase
from US$604 billion in 2010 (~1.0% of global GDP) to US$818
billion in 2015 (~1.1% of global GDP). The growth of economic
costs is faster than that of prevalence of the disease. Thus, demen-
tia has been recognized by WHO and Alzheimer’s Disease
International (ADI) as a global public health priority [3]. The
2013 G8 Dementia Summit in London brought together world
leaders from major industrial countries, research community,
industry, and charities to discuss what can be done to stimulate
greater investment in innovative dementia research, to accelerate
the development of preventive and treatment approaches for
dementia, and to improve quality of life for people with dementia.
In March 2015, WHO hosted the first ministerial conference on
global action against dementia aiming to raise awareness of the
social and economic burden caused by dementia and to highlight
the potential that this huge burden can be reduced only if the
world collectively commits to placing dementia high on the global
public health agenda.
3 Estimation of Dementia Costs: Conceptual and Methodological Issues
The economic burden of a disease is often defined by cost-of-illness

(COI) studies, in which the overall economic impact of the disease
on individuals, caregivers, and the society can be estimated [4].
Quantifying dementia costs is highly relevant for policy making
and resource allocation as well as for determining the cost-
effectiveness of early diagnosis, treatment, and intervention
approaches for dementia [5]. Furthermore, it can motivate policy
makers to develop strategic action plan and to prioritize policy in
social care and proper management of dementia. The economic
burden of an illness largely depends on how the illness is defined
[5]. In most of the COI studies of dementia, dementia has been
clinically diagnosed according to the Diagnostic and Statistical
Manual of Mental Disorders criteria, fourth edition (DSM-IV) [1],
and the National Institute of Neurological and Communicative
Disorders and Stroke and the Alzheimer's Disease and Related
Disorders Association criteria (NINCDS-ADRDA) are used for
Worldwide Costs of Dementia 5
defining Alzheimer’s disease [6]. The severity of dementia and the

living conditions should also be specified as economic costs of
dementia are varied considerably with these factors [5, 7–10].
Moreover, to estimate the economic costs of a disease, it is neces-
sary to specify the study approaches, perspectives, and cost
components.
3.1 The Cost-of- The COI approaches are usually used to estimate dementia costs
Illness Approaches where epidemiological (e.g., number of patients with dementia)
and economic or cost data are integrated. The prevalence-based
and incidence-based approaches are the two main types of COI
studies [11]. The prevalence-based approach is based on the preva-
lence of a disease at a given year to estimate economic burden of
the disease. The total costs are estimated by multiplying the total
number of patients with dementia in a given year with the mean
cost per patient in the same year [11]. In the incidence-based
approach, the total costs are calculated by multiplying the total
number of patients with newly diagnosed (incident) dementia in a
given year with the lifetime costs of these patients [12]. The
prevalence-based approach has been frequently used to estimate
dementia costs for two reasons. First, most cost data of dementia
available in the real world cover costs of a short period of time
rather than the lifetime. Second, dementia, and Alzheimer’s dis-
ease in particular, is an irreversible chronic health condition with
an insidious onset, and prevalence could measure burden of the
disease in the society [13].
3.2 Estimation The economic burden of dementia can be analyzed from three
of Dementia Costs perspectives: family, healthcare provider, and the society [5]. The
from Different economic burden from a societal perspective refers to the costs of
Perspectives all resources consumed or lost due to dementia, irrespective of
who the payer is [14]. In this regard, costs from the societal per-
spective are most relevant for decision makers whose main interest
is the welfare of the society as a whole. A healthcare provider may
be concerned with only the costs of a hospital or nursing home in
order to optimize healthcare delivery of the hospital or nursing
home within a given budget [14]. The costs from a family perspec-
tive include those paid by the family due to the treatment or care
for dementia [14]. The cost components of different perspectives
can be varied, but it is important to include each cost item only
once regardless of the perspectives to avoid double counting.
3.3 Cost There are four cost components in COI studies of dementia [14]:
Components in Cost- (a) direct medical costs, (b) direct nonmedical costs, (c) indirect
of-Illness Studies costs, and (d) intangible costs (Fig. 1). Direct medical costs include
of Dementia all healthcare expenditures used for dementia, including goods and
service costs that are related to diagnosis and treatments of dementia
such as hospitalization and outpatient services. Direct nonmedical
Fig. 1 Main cost components in the cost-of-illness studies of dementia
costs are incurred by outpatient visits or inpatient hospitalizations,

including costs of transportation and accommodation for patient
and family members during these visits, as well as costs of formal
care at nursing home and informal care. Informal care cost often
refers to the opportunity costs (i.e., working time or leisure time
loss) of unpaid care provided by family members, relatives, or
friends [15]. Although it is sometimes considered to be part of
indirect costs, informal care cost is usually included in the direct
nonmedical costs because it replaces formal care services that have
fallen into this category [16]. Indirect costs included the costs of
productivity loss due to premature death and disability. Intangible
costs mean the monetary value of suffering from depression or
upset caused by dementia.
In most of the COI studies of dementia, costs of certain items
or components were not included in the estimates of total costs
due to lack of consensus on how to value these costs. For example,
direct nonmedical costs for home safety modifications, indirect
costs due to premature mortality, intangible costs, and costs of
crime committed by patients with dementia are usually not included
in cost estimates [17].
3.4 Estimation The bottom-up approach and top-down approach are the two
of Costs of the Main methods that are commonly used to estimate the direct costs of
Cost Components dementia [11]. In the bottom-up approach, cost data by individual
cost items are collected through questionnaire survey or review of
3.4.1 Direct Costs
medical records. Then, estimates of the total costs can be obtained
by multiplying the mean cost per patient with the number of
patients derived from prevalence studies. The top-down method
uses national or regional statistics on the total costs of diseases to
separate out the costs of a certain disease (e.g., dementia) [11].
However, it is often difficult to tease out the cost of a certain illness

from the national or regional databases because information on
different diseases is usually sparse in the statistics. Thus, the bot-
tom-up approach is more frequently used in the COI studies of
dementia. However, a potential disadvantage is that costs for
dementia may be overestimated due to double-counting costs
because patients with dementia are likely to have multiple coexist-
ing chronic health conditions (e.g., heart disease, stroke, and
depression). Thus, adjusting for the presence of these coexisting
conditions is important in estimating the costs due to dementia
alone.
3.4.2 Informal Care Informal care provided by family members, relatives, or friends has
Costs a significant impact on the societal costs of dementia [18]. Informal
caregiving, in the form of assistance with basic activities of daily
living (ADLs), is an important component of the support required
by patients with dementia, yet it is unclear how to attribute the
monetary cost to informal caregiver’s time. Indeed, it is compli-
cated to quantify the informal care costs owing to lack of consensus
in methodological issues. First, quantifying caregiver’s time can be
problematic. The time spending on informal care often depends on
the needs of care in self-care ADLs and instrumental ADLs. The
cost is much lower when only considering the cares in basic ADLs
compared with cares in both basic and instrumental ADLs, as well
as time spending in supervision [19]. Moreover, dementia patient
and the caregiver may both involve the same activities such as
shopping together [20]. Second, estimating informal care cost is
also complicated. Most of the previous studies have applied two
different methods to assessing informal care costs [20–22]: the
opportunity cost approach and the replacement cost approach. To
estimate the opportunity cost, the best alternative use of the care-
giver’s time is taken to indicate the value of that time spent in care.
If a caregiver gives up paid work to care, the lost earnings are esti-
mated. Quantifying the costs of caregiver’s time for retired people
can be challenging because there are no obvious market prices [16].
The replacement cost approach assumes that the informal caregiver’s
inputs should be calculated according to the cost of replacing their
care inputs with those of a professional caregiver [16].
3.4.3 Indirect Costs The appropriate method for estimating indirect costs remains
debatable [5]. The common approaches are the human capital, the
willingness-to-pay, and the friction cost methods [23]. In the
human capital approach, the average earnings by age and gender
are combined with time lost from work or years of working life lost
due to premature death to estimate unrealized earnings [24].
However, human capital approach with average wage may be sub-
ject to capital market imperfections. The willingness-to-pay
approach, in contrast, incorporates both lost productivity and the
intrinsic value of life to estimate the average costs that an individual

would be willing to pay for an additional year of life. The friction
cost method assumes that production losses are confined to the
period needed to replace the “sick” worker [25].
Of note, the national and international guidelines for COI
studies advocate that the past or future costs of a disease should be
discounted to the present costs, usually by applying a discount rate
of 3–5%. Most COI studies in the USA and Europe employ 3% or
3.5% as the discount rate [26, 27]. The following equation is used
to estimate costs:
n =1
C a = C t ∑ (1 + r ) ,
−n
t
where Ca denotes the cost in present year a, Ct represents the
cost in year t, r refers to the discount rate, and n implies the time
interval between year a and year t [5].
In addition, there are always uncertainties when estimating
costs of a disease because COI studies are often based on a set of
assumptions and data sources [28]. Sensitivity analysis can be car-
ried out to assess the impact of variations of certain factors on the
cost estimates. In the sensitivity analysis of dementia costs, we
could assess the impact on cost estimates of variations in factors
such as prevalence, medical costs, average wages, productivity
weight, time spending on informal care, and the proportion of
patients with dementia who receive informal and formal care [24,
29–31].
3.5 Sources The sources of cost data in COI studies include the following.
of Cost Data First, we can collect cost data through questionnaire survey that
includes costs by individual items, where direct and indirect costs
can be estimated. When designing the questionnaire, cost indica-
tors can be relatively comprehensive. However, data collected
through the survey may be subject to recall bias. Accurate and
precise data may be obtained through follow-up survey, but it can
be expensive and time consuming [23]. Moreover, the questions
and guides in the questionnaire may vary across studies, thus
resulting in inaccuracies of the data [23]. Second, cost data can be
derived from electronic medical records, which usually include
accurate and reliable information on medical expenses. However,
medical records often include data of direct medical costs, and the
costs may vary depending on available facilities of hospitals or medical
institutes. Thus, cost data from certain medical institutes may not
be representative of the target population [23]. Finally, relevant
data needed for cost estimates can also be obtained from literature
or government statistics. This is usually convenient and timesaving,
but it may not capture all the utilizations, and the representativeness
should also be kept in mind [32].
4 Economic and Societal Burden of Dementia: A Brief Summary of Literature
We searched PubMed, Google Scholar, Web of Science, and China

National Knowledge Infrastructure databases for studies estimat-
ing the economic costs of Alzheimer’s disease and other forms of
dementia that were published since the 2000s. Table 1 provides a
summary of major literature regarding economic burden of demen-
tia in 2000–2017. A total of 17 studies were identified. There are
substantial variations in the estimates of total costs per dementia
patient, ranging from US$2935 (China) to US$64168 (USA) [24,
33–48], largely depending on methodologies, perspectives of cost
estimation, data sources, cost categories, and setting and severity
of patients with dementia. Moreover, there are regional differences
in economic costs of dementia, such that dementia causes higher
economic costs in high-income countries (HICs) compared with
low- and middle-income countries (LMICs). To reduce future
burden of dementia, national and global strategic action plan to
promote primary and secondary prevention of dementia as well as
to improve social care of dementia is critical [24].
5 Future Perspectives
First, methodological issues need to be reported in detail in future

COI studies of dementia, because an insufficient description of the
methodology may lead to misunderstandings or misinterpretations
of the results. The methodological limitations should also be fully
discussed. These issues are particularly relevant given that consen-
sus on the methodology of the COI studies forms a basis for esti-
mating and comparing the economic burden of a disease across
studies from different regions. Moreover, the indicators selected in
the study should be combined with local conditions. For example,
the disability-adjusted life years (DALYs) index needs to be
adjusted according to the local economic index, age weight, dis-
ability weight, and life expectancy. In addition, the clinical stages
and severity of dementia should be taken into consideration when
estimating the costs of dementia.
Second, cost data from large samples with long-term follow-
ups are preferred for estimating dementia costs. Current COI
studies have been limited by small sample size [15] and a relatively
short period of follow-ups [49]. Moreover, survival benefits from
treatment, dynamic changes in prevalence of risk factors for demen-
tia (e.g., hypertension and diabetes), increases in care costs (e.g.,
use of medical devices), and hiring professional caregivers are likely
to affect the economic burden of dementia. However, these data
are generally missing, which limits the use of advanced approaches
(e.g., microsimulation approach).
Table 1
A brief summary of literature regarding economic costs of dementia, 2000–2017
Costs by categories (US$)
Perspectives of Direct Direct Indirect cost Total

Study, country cost estimation medical care nonmedical care informal care costs
Rigaud et al. [33], Societal 6663 5632 18,858 31,153
France
Scuvee-Moreau et al. NA 8218 5222 1233 14,673
[34], Belgium
Beeri et al. [35], Israel Societal 3974 9326 6593 19,893
Jönsson et al. [36], Societal 3113 7167 3757 14,038
Nordic region
Francois et al. [37], Societal NA NA NA 33,333
Finland
Wolstenholme et al. NA 1818 30,650 NA 32,468
[38], UK
Livingston et al. [39], Societal NA NA NA 35,287
UK
Lopez-Bastida et al. Societal 4848 2306 29,884 37,881
[40], Spain
Coduras et al. [41], Societal 4744 5798 12,016 22,558
Spain
Mesterton et al. [42], Societal 3155 39,373 4428 46,956
Sweden
Kiencke et al. [43], Healthcare payer 11,786 NA NA 11,786
Germany
Zencir et al. [44], NA 2128 NA 1364 3492
Turkey
Wang et al. [45], China NA 863 431 163 2935a
Xu et al. [24], China Societal 1152 4779b 88 6020d
4612c 5852e
Hurd et al. [46], USA Societal 33,329 NA 30,839f 64,168f
14,591g 47,920g
Suh et al. [47], Korea Societal 4394 445 655 11,389
Allegri et al. [48], Societal 3389 2488 1832 7709
Argentina
a
Net costs. NA: not available
b
Informal care
c
Nursing home care
d
Cost for patients living at home
e
Cost for patients living at nursing home
f
Valued according to replacement cost
g
Valued according to cost of forgone wages
Finally, the potential impact of secular trends in dementia

occurrence on future societal and economic burden of the disease
needs to be clarified [50]. Currently, all estimates of dementia costs
have been based on the assumption of a stable or increased preva-
lence over time. However, recent studies have suggested a declin-
ing trend in occurrence of dementia in some HICs (e.g., UK,
Sweden, and USA) [51–54]. In contrast, an increase in the age-
specific prevalence of dementia has been reported in Asia (e.g.,
Japan and China) or LMICs. As prevalence of dementia is the sec-
ond most important driver of societal costs of dementia after popu-
lation aging [2], changes in dementia occurrence are likely to have
significant impact on the total number of people with dementia in
the community and thus have a huge impact on the societal burden.
The impacts of time trends in dementia occurrence on global soci-
etal costs are also complicated by the fact that the number of
dementia cases would increase to a greater extent in LMICs
(~250%) than in HICs (~116%) [29]. Therefore, further research
to clarify the impact of time trends in dementia occurrence on soci-
etal costs is urgently needed, in which up-to-date epidemiological
data and cost data should be systematically reviewed and analyzed
using the standardized COI approaches. This will help develop
health policy and allocate resources.
Acknowledgment
Chengxuan Qiu received research grants from the Swedish

Research Council (VR), the Swedish Research Council for Health,
Working Life and Welfare (FORTE), and Karolinska Institutet,
Stockholm, Sweden.
References
1. American Psychiatric Association (2000) 4. Quentin W, Riedel-Heller S, Luppa M,
Diagnostic and statistical manual of mental Rudolph A, König HH (2010) Cost of illness
disorders, 4th edn. American Psychiatric studies of dementia: a systematic review focus-
Association, Washington DC ing on stage dependency of costs. Acta Psyciatr
2. Alzheimer’s Disease International (ADI) Scand 121:243–259
(2015) World Alzheimer report 2015, the 5. Costa N, Derumeaux H, Rapp T et al (2012)
global impact of dementia: an analysis of preva- Methodological considerations in cost of illness
lence, incidence, cost and trends. Alzheimer’s studies on Alzheimer disease. Health Econ Rev
Disease International (ADI), London. http:// 15:18
www.worldalzreport2015.org/downloads/ 6. McKhann G, Drachman D, Folstein M,
world-alzheimer-report-2015.pdf. Accessed 10 Katzman R, Price D, Stadlan EM (1984)
Jun 2017 Clinical diagnosis of Alzheimer’s disease: report
3. World Health Organization and Alzheimer’s of the NINCDS-ADRDA work group under
Disease International (2012) Dementia: a pub- the auspices of Department of Health and
lic health priority. World Health Organization, Human Services Task Force on Alzheimer's
Geneva. http://apps.who.int/iris/bitstr disease. Neurology 34:939–944
eam/10665/75263/1/9789241564458_ 7. Hope T, Keene J, Gedling K, Fairburn CG,
eng.pdf?ua=1. Accessed 10 June 2017 Jacoby R (1998) Predictors of institutionaliza-
tion for people with dementia living at home 21. McDaid D (2001) Estimating the costs of
with a carer. Int J Geriatr Psychiatry 13: informal care for people with Alzheimer’s dis-
682–690 ease: methodological and practical challenges.
8. Small GW, McDonnell DD, Brooks RL, Int J Geriatr Psychiatry 16:400–405
Papadopoulos G (2002) The impact of symp- 22. Nordberg G, Wimo A, Jönsson L et al (2007)
tom severity on the cost of Alzheimer disease. Time use and costs of institutionalised elderly
J Am Geriatric Soc 50:321–327 persons with or without dementia: results from
9. Mauskopf J, Racketa J, Sherrill E (2010) the Nordanstig cohort in the Kungsholmen
Alzheimer’s disease: the strength of association project: a population based study in Sweden.
of costs with different measures of disease Int J Geriatr Psychiatry 22:639–648
severity. J Nutr Health Aging 14:655–663 23. Xu JF (2016) Study on the economic burden
10. Herman N, Tam DY, Balshaw R, Sambrook R, of mental disorders and the influence of health
Lesnikova N, Lanctôt KL (2010) The relation insurance-based on patients in Shandong prov-
between disease severity and cost of caring for ince. Shandong University, Jinan
patients with Alzheimer disease in Canada. Can 24. Xu J, Wang J, Wimo A, Fratiglioni L, Qiu C
J Psychiatr 55:768–775 (2017) The economic burden of dementia in
11. Wimo A, Jönsson L, Fratiglioni L et al (2016) China, 1990–2030: implications for health
The societal costs of dementia in Sweden policy. Bull World Health Organ 95:18–26
2012 – relevance and methodological chal- 25. Oostdam N, Bosmans J, Wouters MG, Eekhoff
lenges in valuing informal care. Alzheimers Res EM, Mechelen WV, Poppel MNV (2012)
Ther 8:59 Cost-effectiveness of an exercise program dur-
12. Girmens JF, Koerber C, Miadi-Fargier H, ing pregnancy to prevent gestational diabetes:
Wittrup-Jensen KU (2013) Costs of incidence- results of an economic evaluation alongside a
based central retinal vein occlusion in France. randomised controlled trial. BMC Pregnancy
Value Health 16:A504–A504 Childbirth 12:64
13. Andlin-Sobocki P, Jönsson B, Wittchen HU, 26. Gold MR, Russel LB, Siegel JE, Daniels N,
Olesen J (2005) Cost of disorders of the brain Weinstein MC (1996) Cost effectiveness in
in Europe. Eur J Neurol 12(Suppl 1):1–27 health and medicine. JAMA 276:1172–1177
14. Leung GM, Yeung RY, Chi I, Chu LW (2003) 27. Luengo-Fernandez R, Leal J, Gray A, Sullivan
The economics of Alzheimer disease. Dement R (2013) Economic burden of cancer across
Geriatr Cogn Disord 15:34–43 the European Union: a population-based cost
15. Olesen J, Gustavsson A, Svensson M, Wittchen analysis. Lancet Oncol 14:1165–1174
HU, Jönsson B (2012) The economic cost of 28. Xu J, Wang J, Wimo A, Qiu C (2016) The eco-
brain disorders in Europe. Eur J Neurol nomic burden of mental disorders in China,
19:155–162 2005–2013: implications for health policy.
16. Wimo A, Jönsson L, Bond J, Prince M, Winblad BMC Psychiatry 16:137
B, Alzheimer Disease International (2013) The 29. Wimo A, Guerchet M, Ali GC et al (2017) The
worldwide economic impact of dementia 2010. worldwide costs of dementia 2015 and compari-
Alzheimers Dement 9:1–11.e3 sons with 2010. Alzheimers Dement 13:1–7
17. Deb A, Thornton JD, Sambamoorthi U, Innes 30. Wimo A, Winblad B, Jönsson L (2010) The
K (2017) Direct and indirect cost of managing worldwide societal costs of dementia: estimates
Alzheimer’s disease and related dementias in for 2009. Alzheimers Dement 6:98–103
the United States. Expert Rev Pharmacoecon 31. Wang H, Gao T, Wimo A, Yu X (2010)
Outcomes Res 17:189–202 Caregiver time and cost of home care for
18. Schneider J, Hallam A, Islam MK et al (2003) Alzheimer’s disease: a clinic-based observa-
Formal and informal care for people with tional study in Beijing, China. Ageing Int
dementia: variations in costs over time. Ageing 35:153–165
Soc 23:303–326 32. Zhang YY (2012) Study on the economic bur-
19. Prince M, Bryce R, Albanese E et al (2013) den of inpatients with hepatitis B owning basic
The global prevalence of dementia: a systematic medical insurance for urban workers in Tianjin.
review and metaanalysis. Alzheimers Dement Tianjin University, Tianjin
9:63–75.e2 33. Rigaud AS, Fagnani F, Bayle C, Latour F,
20. Jönsson L, Wimo A (2009) The cost of dementia Traykov L, Forette F (2003) Patients with
in Europe: a review of the evidence, and method- Alzheimer’s disease living at home in France:
ological considerations. PharmacoEconomics costs and consequences of the disease. J Geriatr
27:391–403 Psychiatry Neurol 16:140–145
34. Scuvee-Moreau J, Kurz X, Dresse A (2002) 44. Zencir M, Kuzu N, Beşer NG, Ergin A, Catak
The economic impact of dementia in Belgium: B, Sahiner T (2005) Cost of Alzheimer’s dis-
results of the National Dementia Economic ease in a developing country setting. Int
Study (NADES). Acta Neurol Belg J Geriatr Psychiatry 20:616–622
102:104–113 45. Wang G, Cheng Q, Zhang S et al (2008)
35. Beeri MS, Werner P, Adar Z, Davidson M, Noy Economic impact of dementia in developing
S (2002) Economic cost of Alzheimer disease countries: an evaluation of Alzheimer-type
in Israel. Alzheimers Dis Assoc Disord dementia in shanghai, China. J Alzheimers Dis
16:73–80 15:109–115
36. Jönsson L, Eriksdotter Jö nhagen M, Kilander 46. Hurd MD, Martorell P, Langa KM (2013)
L et al (2006) Determinants of costs of care for Monetary costs of dementia in the United
patients with Alzheimer’s disease. Int J Geriatr States. N Engl J Med 369:489–490
Psychistry 21:449–459 47. Suh GH, Knapp M, Kang CJ (2006) The eco-
37. Francois C, Sintonen H, Sulkava R, Rive DB nomic costs of dementia in Korea, 2002. Int
(2004) Cost effectiveness of memantine in J Geriatr Psychiatry 21:722–728
moderately severe to severe Alzheimer’s dis- 48. Allegri RF, Butman J, Arizaga RL et al (2007)
ease: a Markov model in Finland. Clin Drug Economic impact of dementia in developing
Invest 24:373–384 countries: an evaluation of costs of Alzheimer-
38. Wolstenholme J, Fenn P, Gray A, Keene J, type dementia in Argentina. Int Psychogeriatr
Jacoby R, Hope T (2002) Estimating the rela- 19:705–718
tionship between disease progression and cost 49. Cavallo MC, Fattore G (1997) The economic
of care in dementia. Br J Psychiatry and social burden of Alzheimer disease on fam-
181:36–42 ilies in the Lombardy region of Italy. Alzheimer
39. Livingston G, Katona C, Roch B, Guilhaume Dis Assoc Disord 11:184–190
C, Rive B (2004) A dependency model for 50. Hurd MD, Martorell P, Langa K (2015) Future
patients with Alzheimer’s disease: its validation monetary costs of dementia in the United
and relationship to the costs of care. The States under alternative dementia prevalence
LASER-AD study. Curr Med Res Opin scenarios. J Popul Ageing 8:101–112
20:1007–1016 51. Matthews FE, Arthur A, Barnes LE et al (2013)
40. Lopez-Bastida J, Serrano-Aguilar P, Perestelo- A two-decade comparison of prevalence of
Perez L, Olivamoreno J (2006) Social- dementia in individuals aged 65 years and older
economic costs and quality of life of Alzheimer from three geographical areas of England:
disease in the Canary Islands, Spain. Neurology results of the cognitive function and ageing
67:2186–2191 study I and II. Lancet 382:1405–1412
41. Coduras A, Rabasa I, Frank A et al (2010) 52. Matthews FE, Stephan BC, Robinson L et al
Prospective one-year cost-of-illness study in a (2016) A two decade dementia incidence com-
cohort of patients with dementia of Alzheimer’s parison from the cognitive function and ageing
disease type in Spain: the ECO study. studies I and II. Nat Commun 7:11398
J Alzheimers Dis 19:601–615 53. Satizabal CL, Beiser AS, Chouraki V, Chêne G,
42. Mesterton J, Wimo A, By A, Langworth S, Winblad Dufouil C, Seshadri S (2016) Incidence of
B, Jönsson L (2010) Cross sectional observational dementia over three decades in the Framingham
study on the societal costs of Alzheimer’s disease. heart study. N Engl J Med 374:523–532
Curr Alzheimer Res 7:358–367 54. Langa KM, Larson EB, Crimmins EM et al
43. Kiencke P, Dietmar D, Grimm C, Rychlik R (2017) A comparison of the prevalence of
(2011) Direct costs of Alzheimer disease in dementia in the United States in 2000 and
Germany. Eur J Health Econ 12:533–539 2012. JAMA Intern Med 177:51–58
Chapter 2
Epidemiology and Risk Factors of Alzheimer’s Disease:

A Focus on Diet
Cécilia Samieri
Abstract
Alzheimer’s disease (AD) appears to result from a highly dynamical, time-dependent cascade of neuro-
pathological processes, which involve the balance between two antagonist pathways: an accumulation of
lesions in the brain (evolving progressively over decades), balanced by a variety of cellular and tissular
compensatory mechanisms. Preclinical research has indicated that several nutritional factors have the
potential to modulate many components of this complex system; dietary interventions may thus constitute
promising strategies for the prevention of AD. Extensive epidemiological research has been conducted on
nutrition, diet, and AD dementia over the past two decades. This chapter reviews existing epidemiological
literature on nutrition and cognitive aging and AD dementia, with a specific focus on observational, pro-
spective studies, including both clinical outcomes and brain imaging biomarkers. Most studies investigated
candidate foods and nutrients, including “good” fats (e.g., marine long-chain omega-3 polyunsaturated
fatty acids provided by fish intake), antioxidant nutrients (i.e., vitamins C and E, carotenoids, and polyphe-
nols, found in plant products), and B vitamins (e.g., folate provided by green leafy vegetables). Overall,
associations between individual nutrients and both clinical outcomes and brain imaging biomarkers have
been relatively inconsistent. Healthy dietary patterns such as the Mediterranean diet, which combine sev-
eral healthy foods and nutrients, have been more consistently related to AD outcomes. Still, most random-
ized controlled trials on nutritional supplements for the prevention of AD have remained inconclusive so
far. Future research should leverage innovative approaches such as “omics” methods to explore dietary
exposures, novel imaging markers to capture prodromal AD, and longitudinal statistical approaches to
model long-term trajectories of risk factors and AD markers, to better understand how and when diet may
shape the brain and prevent AD.
Key words Alzheimer’s disease, Dementia, Epidemiology, Population-based, Diet, Nutrition, Vitamins,
Fatty acids, Clinical trial
1 Introduction
Dementia is the most frequent disease associated with brain aging.

Approximately 36 million of persons were affected by dementia in
2010 worldwide, and with the gradual aging of populations, this
prevalence is expected to dramatically increase in the next decades,
to reach possibly around 131 million persons by 2050 [1].
Alzheimer’s disease (AD) is the most frequent form of dementia,
15
16 Cécilia Samieri
accounting for approximately two thirds of cases. No curative

treatment for AD has been identified yet, and therapeutic research
has been extremely disappointing so far. Sporadic AD (the form
which occurs in older age and represents 99% of all cases) appears
to result from a complex interplay between genetic and environ-
mental risk factors. The strongest genetic risk factor for sporadic
AD is the ε4 allele of the apolipoprotein E gene (APOEε4).
Furthermore, about 20 single nucleotide polymorphisms have
been moderately associated with disease risk in genome-wide asso-
ciation studies [2]. Environmental risk factors for AD and demen-
tia include cardio-metabolic risk factors, psychosocial factors, and
lifestyle [3]. Recent epidemiological findings from several large
prospective studies across Europe and the USA have established a
decrease in the incidence of dementia in the last two decades [4, 5].
Improvements of educational level and in the management of car-
diovascular risk factors have been proposed as a plausible explana-
tion for such a decrease of incidence, suggesting that lifestyle-based
preventive interventions aimed at improving cardiovascular health
may be a promising strategy to lower the burden of dementia.
However, there is now strong evidence that pathophysiologi-
cal mechanisms evolve silently over years, if not decades, before
clinical diagnosis of dementia in the late life; it is thus likely that
prevention may be effective only through a long-course perspec-
tive, targeting individuals in the preclinical phase of the disease [3].
The mechanisms underlying AD dementia appear complex and are
still not fully understood. They involve a progressive accumulation
in the brain of several pathophysiological alterations, including (1)
Aβ proteins which aggregate extracellularly in the form of plaques,
(2) intracellular production of hyper-phosphorylated Tau proteins
in the form of tangles, and (3) neuro-inflammation which occurs
through activation of microglia. The interaction of these patho-
logical events will eventually lead to neurodegeneration and related
cognitive decline; however, the nature of such interactions and
their chronology are still under debate [3]. The primary theory of
AD focuses on the accumulation of amyloid plaques as the main
neuropathological culprit of the disease. According to the “amy-
loid cascade hypothesis,” brain Aβ accumulation, which results
from increased production of Aβ, decreased clearance, or both,
may further precipitate the production of Tau (also found in nor-
mal aging) and may increase neuro-inflammation caused by
microglial cells, which leads to neurodegeneration. An alternative
theory views accumulation of Aβ and Tau as two parallel processes,
causing AD and enhancing each other’s toxic effects [6]. Overall,
Aβ appears necessary but not sufficient to the development of clini-
cal dementia. Still, the reasons for brain Aβ accumulation remain
partly unknown, and the factors which dictate how brain health
will respond to amyloid/Tau pathology have still to be elucidated.
A key additional player may be vascular brain health. Indeed, age-
Diet and Alzheimer’s Disease 17
related alterations of the neurovasculature which constitutes the

blood-brain barrier (also known as the “blood-brain barrier break-
down”) may both increase the permeability of the brain to various
neurotoxic substances, cells, and pathogens (bacteria and viruses,
e.g., herpes virus, which can trigger amyloid deposition [7]) and
strongly hinder the clearance of Aβ from the brain [8]. Early dys-
function of the vascular system may thus be the first step that pre-
dispose to the accumulation of amyloid, along with reducing the
oxygen and metabolite supply (sugar, nutrients, etc.) to the brain.
Another important player, which adds to the complexity of the
disease, relies on compensatory mechanisms, generally described as
cognitive and brain reserve. Cognitive and brain reserve enable the
brain, thanks to a more developed tissue (i.e., more numerous neu-
rons and/or higher synaptic density) and alternative neuronal cir-
cuits, to cope with the presence of lesions for a long time before
the expression of clinical symptoms. Hence, higher educational
level (a proxy of cognitive reserve) has been associated with delayed
cognitive decline and AD dementia in epidemiological studies [9].
In summary, AD dementia appears to result from a highly
dynamical neuropathological system based on several processes
(amyloid production and clearance, Tau production, neuro-inflam-
matory processes, neurovascular health, and cognitive/brain
reserve), of which interactions over 10–20 years will determine
cognitive trajectories and the risk to develop dementia in the late
life. Furthermore, the neurovascular unit (made of the microvascu-
lar endothelium defining the blood-brain barrier and astrocytes,
pericytes, neurons, and microglial cells) is a key player of this
dynamic network, and each of the elements forming the neurovas-
cular unit has a strong plasticity (e.g., neurovascular plasticity, neu-
ronal/synaptic plasticity, microglia plasticity)—thus allowing the
system to adapt for a long time before falling down. Interestingly,
lifestyle factors, in particular nutrition, have the potential to modu-
late the plasticity of each of these elements over the life course.
Furthermore, selected nutrients may have a specific role against the
accumulation of amyloid, Tau, and vascular lesions in AD.
In the context of a disease evolving over years with no curative
treatment, prevention is considered one of the most promising
strategies to face dementia and AD [10]. Understanding the early
influence and interactions of non-modifiable risk factors (e.g., gen-
der, genetics, age) and modifiable risk factors (e.g., physical activ-
ity, diet, and cognitive stimulation) for dementia in population-based
samples is considered as a research priority to reduce the global
burden of dementia by 2025 [10]. In this chapter, we will provide
an overview of current knowledge on nutrition and diet for the
prevention of dementia and AD through an epidemiological per-
spective. We will focus on prospective observational studies, in
which exposure precedes outcome, and we will put less emphasis
18 Cécilia Samieri
on cross-sectional studies which assess exposure and outcome con-

comitantly, because prospective studies address temporality of
associations (a pillar of causality). We will start with studies on
clinical outcomes (i.e., dementia/AD diagnosis and cognitive
decline) and continue with most recent studies on AD biomarkers
(i.e., assessment of brain atrophy, vascular health, and amyloid load
using brain imaging) which give access to subtle, preclinical impair-
ments in early phases of the neuropathological continuum underly-
ing AD. Nutritional exposures will be examined through a
bottom-up approach, addressing first the most important individ-
ual candidates for brain health and second a global approach of
nutrition through dietary patterns, which take into account poten-
tial additive and/or synergistic effects of the food matrix on health.
Furthermore, we will examine both food/nutrient intake data
gained from dietary surveys and nutrient biomarkers, which are
less prone to measurement error and incorporate interindividual
variability in biodisponibility. The chapter will end with an over-
view on main findings from randomized controlled trials and brain
aging outcomes, discussing potential reasons for inconsistencies
with observational studies and providing a few insights on poten-
tial future innovations in the field.
2 Nutrition, Diet, and Alzheimer’s Disease Clinical Outcome
2.1 Nutrient Several nutrients and foods have neuroprotective effects which have
and Food Candidates been largely established in preclinical research. Overall, these nutri-
for Brain Health ent/food candidates fall into three main categories: (1) “good”
fats, i.e., unsaturated fats, in particular polyunsaturated fats from
the omega-3 family (n-3 PUFAs), mainly provided by vegetable oils
(for precursors, e.g., alpha linolenic acid [ALA]) and by fish (for the
long-chain derivatives, e.g., eicosapentaenoic acid [EPA] and doco-
sahexaenoic acid [DHA] which are directly useful for brain struc-
ture and function); (2) antioxidant nutrients, including vitamins E
and C, carotenoids, and polyphenols, provided by fruits, vegetables,
vegetable oils (e.g., olive oil), and plant-derived beverages (coffee,
tea, wine); (3) other vitamins involved in various neuroprotective
mechanisms, including B vitamins (folate or vitamin B9, particu-
larly abundant in green leafy vegetables, and vitamin B12 provided
by animal products) and vitamin D, found in fatty fish. Each of
these nutrients has been individually related to the risk of AD
clinical outcomes in prospective observational studies.
2.1.1 “Good” Fats: Lipids represent 50–60% of total brain weight, mostly under the
Omega-3 Polyunsaturated form of PUFAs in phospholipids of neuronal membranes or in the
Fats and Fish gain of myelin, and the long-chain (LC) n-3 PUFA DHA repre-
sents up to 60% of esterified fatty acids in neuronal membranes.
DHA is essential to the development of the mammalian brain—in
particular for the hippocampus, a key structure for memory which

is altered in both aging and AD neuropathology. Rodent studies
have demonstrated that LC n-3 PUFA preferentially accumulate in
the hippocampus and are involved in neuronal transmission, neu-
rogenesis, synaptic plasticity, learning, and memory [11].
Furthermore, hippocampal DHA levels decrease with age and in
AD, and this decrease is associated with reduced hippocampal-
dependent spatial learning memory ability [12]. EPA and DHA
have anti-inflammatory properties which may help lower neuro-
inflammation associated with normal brain aging and AD. DHA
may also exert anti-amyloid effects [13]. Hence, an adequate supply
of essential LC n-3 PUFA in the diet (mainly from fish) may be
critical for the maintenance of cognition and memory with aging.
Some observational studies have found that intakes of fish, or
higher concentrations of LC n-3 PUFA in blood, may be related to
a lower risk of dementia and to slower cognitive decline [14–18] in
older subjects. However, inconsistent findings have been reported,
especially with cognitive decline [19]. In 2016, a meta-analysis of
21 prospective cohort studies including 181,580 participants and
4438 dementia cases followed for 2–21 years concluded that each
1-serving/week increment of dietary fish was associated with a 5%
lower risk of dementia (95% confidence interval [CI], 1–10%) and
a 7% lower risk of AD (95% CI, 5–10%). Dietary DHA was also
significantly associated with lower risks of dementia and AD, while
there was no significant association with dietary EPA or blood EPA
and DHA biomarkers [20]. Likewise, blood ALA, the precursor of
the LC n-3 PUFA found in vegetable oils (colza, soybean, walnut),
was not associated with dementia risk—which may not be surpris-
ing given that the conversion of ALA to derivatives EPA and DHA
is considered very limited in humans [21].
Recently, the Rush Memory and Aging cohort reported,
among 286 older participants (aged 90 years old on average), a
cross-sectional association between higher intake of fish and lower
AD neuropathology after brain autopsy at death (an average 8 years
later); yet the association was limited to carriers of the APOEε4
allele [22, 23]. There is a biological rationale for a vulnerability of
APOEε4 carriers to lower DHA status, especially in early AD
stages, before marked neurodegeneration [24]. First, APOEε4
may decrease circulating DHA by increasing DHA hepatic catabo-
lism (through increased affinity to LDL and VLDL receptors).
Second, APOEε4 is associated with a breakdown of the blood-
brain barrier and may decrease the transfer of DHA to the brain
(by altering the efficacy of the Mfsd2a receptor). Third, APOEε4
is associated with hypolipidated lipoproteins and may thus com-
promise the transport of DHA in the brain. However, interactions
between fish and APOEε4 carrier status in relation to AD clinical
outcomes have been inconsistent in epidemiological studies so far,
with studies reporting association limited to APOEε4 carriers
20 Cécilia Samieri
[22, 23, 25, 26], others reporting association in APOEε4 noncar-

riers [27–30], and most studies reporting no interaction [20].
Moreover, an inverse association between low serum DHA and
cerebral amyloidosis was found independent of the APOEε4
genotype in a small study of older non-demented persons [31].
2.1.2 Antioxidant As n-3 PUFA, antioxidant nutrients (vitamins E and C, carot-

Nutrients and Fruits, enoids, and polyphenols) and their main food sources (fruits, veg-
Vegetables, and Specific etables, and specific polyphenol-rich foods and beverages, e.g.,
Antioxidant-Rich olive oil, wine, coffee, tea, cocoa) have been investigated for more
Compounds (Olive Oil, Red than 15 years in relation to brain aging outcomes in epidemiologi-
Wine, Coffee, Tea, Cocoa) cal research. Antioxidant nutrients may help lower oxidative stress
associated with AD neuropathology, thereby counteracting oxida-
tion of easily peroxidable brain LC PUFA (e.g., DHA). In particu-
lar, vitamin E, provided by seed oils and nuts, represents the major
lipid-soluble antioxidant found in cells; the main function of vita-
min E is to prevent the peroxidation of membrane phospholipids
[32]. Carotenoids, found in colored fruits and vegetables, are pre-
cursors of retinoids, key signaling molecules for synaptic plasticity
[33]. As all liposoluble vitamins (i.e., vitamins A, D, E, and K),
vitamins A and E may affect the generation and clearance of Aβ
both by direct effects and indirectly by altering the cellular lipid
homeostasis [32]. Aside from the “classical” antioxidant vitamins,
polyphenols have been increasingly investigated over the last 5
years in relation to brain health, since their contribution to the
total antioxidant capacity of the human diet appears much larger
than that of vitamins [34] and some polyphenols have powerful
biological effects for brain function and brain health. For example,
selected flavonoids have strong vascular beneficial properties and
improve vasoactivity both in the periphery and brain (e.g., flavan-
3-ols [35], flavanols [36]). A human clinical study demonstrated
that 3-month supplementation with flavanols from cocoa (cate-
chins and epicatechins) in healthy adults improved cognition and
enhanced hippocampal vascular plasticity at magnetic resonance
imaging (MRI) [36]. Beyond vascular beneficial effects, certain
polyphenols are found in brain areas involved in learning and
memory and improve neuronal signaling and cognitive behavior
(i.e., anthocyanins [37]), and some species appear to exert specific
anti-amyloid effects. For example, resveratrol, found in grape and
red wine, increases Aβ clearance in vitro [38]; however, the very
high doses necessary to reach such biological effects may be impos-
sible to reach from diet. Non-resveratrol polyphenols such as
anthocyanin and quercetin (a flavonol) metabolites [e.g., querce-
tin-3-O-glucuronide] appear to cross the blood-brain barrier and
may decrease Aβ aggregation and improve learning memory in
mice [39, 40].
Epidemiological studies have primarily investigated the rela-
tion of dietary intakes of antioxidant foods to the risk of dementia
or cognitive decline. Most studies on the overall consumption of

fruit and vegetables found an association between higher intakes of
vegetables and lower risk of AD or less cognitive decline, while
association with fruit intake was less consistent [41]. Furthermore,
a low-to-moderate consumption of alcoholic beverages was associ-
ated with a lower risk of dementia in several large prospective stud-
ies [42]; although all alcoholic beverages may be protective
(consistently with a neuroprotective role of ethanol at very moder-
ate doses), a stronger association was found with red wine, which
provides, in addition to ethanol, a large variety of polyphenols. As
red wine, olive oil (a specific feature of the Mediterranean diet)
contains many polyphenol species (e.g., tyrosol) and may contrib-
ute to protect the brain from age-related diseases. In the Three-
City (3C) study, intensive use of olive oil for cooking and seasoning
was related to lower risk of stroke and less decline in visual memory
[43, 44]. However, epidemiological research on olive oil and brain
health has been very limited so far and deserves further investiga-
tion. Likewise, coffee and tea intakes were related to lower cogni-
tive decline or dementia risk in a few studies; yet inconsistent
findings were also reported, especially from prospective studies
[45], and overall, these associations need to be further explored in
large-scale studies.
Intakes of individual antioxidant nutrients were also examined
in relation to dementia outcomes. However, studies on carotenoids
and vitamins were hampered by the inclusion of supplement users,
who have considerably higher intakes of antioxidants, and high-
dose supplements may be more harmful than beneficial [46]. A
meta-analysis of seven prospective studies on dietary intakes
(excluding supplements) of vitamins E and C and β-carotene and
AD risk found that moderate to high intakes (defined as being in
the second to last quantile of exposure distribution) of vitamins E
and C were related to a 24% (95% CI 16–33%) and a 17% (95% CI
6–28%) lower risk of AD, respectively, whereas the risk reduction
associated with β-carotene (−12%) was only borderline significant
(95% IC −27%, +3%) [47]. With antioxidant nutrients examined
through blood biomarkers, findings have been relatively inconsis-
tent so far. Lower plasma vitamin E was associated with higher
cognitive impairment/dementia/AD risk in several prospective
cohorts [48–52], but not all [53], and there is evidence that the
association between vitamin E and cognitive decline/dementia is
influenced by concurrent cholesterol levels [49, 51, 52]. Among
carotenoids, plasma lutein was the single specie modestly inversely
associated with dementia risk over 10 years in the 3C study [54],
whereas in the MacArthur Studies of Successful Aging, serum
β-carotene was associated with lower cognitive decline but only
among APOEε4 carriers [55]. In contrast, in the Nurses’ Health
Study, none of the individual carotenoid or vitamin E species measured
in plasma was associated with cognitive decline over 4 years [56].
22 Cécilia Samieri
Thus overall, epidemiological evidence for an association between

lower intakes of carotenoids and antioxidant vitamins in dementia
and AD has been mixed in existing literature.
Furthermore, promising, yet limited epidemiological research
has been conducted on polyphenols and brain aging outcomes.
Inverse associations were found between higher intakes of total
flavonoids [57–60], flavonols [61], and anthocyanidins found in
berries (blueberry, strawberry, raspberry) [60] and lower risk of
dementia or lower cognitive decline. An Italian study found an
association between total urinary polyphenols and lower cognitive
decline over 3 years [62]. However, more research is needed both
at the preclinical stage to better understand the complex bioavail-
ability and metabolism of polyphenols and at the clinical stage to
identify most promising targets for prevention.
2.1.3 Vitamins B and D B vitamins, including folate (vitamin B9), vitamin B6, and vitamin
B12, have a fundamental role in brain development and function-
ing and have been extensively examined in relation to age-related
brain diseases [63]. B vitamins regulate homocysteine (Hcy) levels,
and hyperhomocysteinemia is an established risk factor for cardio-
vascular diseases and dementia [64]. Hyperhomocysteinemia may
affect brain function through both vascular mediation and neuro-
toxicity [65]. B vitamins may also have a direct effect on the brain,
independently of Hcy: a lower status in B vitamins decreases
S-adenosylmethionine, a major intermediate of methylation reac-
tions which are involved in synaptic transmission and epigenetic
regulation [66]. Moreover, tetrahydrofolate, the active form of
folate, is involved in DNA replication which is critical for adult hip-
pocampal neurogenesis [67]; direct mechanistic effects of B vita-
mins on the brain independent of Hcy may therefore be particularly
relevant to folate. Prospective epidemiological studies have found
associations between higher intakes of B vitamins and a lower risk
of dementia and AD; four studies have reported associations with
dietary folate (but not with vitamins B6 and B12) [68–71],
although null [72–74] and even harmful associations (when folate
was combined with low vitamin B12 status) [74] were also
reported. Findings were also inconsistent in studies using blood
status in B vitamins (see for review [75]). Importantly, most stud-
ies were conducted in the USA, where fortification of all flour and
uncooked cereal grain product in folic acid became mandatory in
1998, leading to a subsequent dramatic increase in folate status
which may have biased observational findings. Moreover, it is pos-
sible that folate is protective for the brain in lower intake ranges
and become inefficient—and even detrimental for those with low
B12 status—at higher ranges (as those found in North American
population under folic acid fortification policy). In the French 3C
study in France (a country where folic acid fortification is not
mandatory), an inverse association was found between higher

folate intake and lower dementia risk over 10 years [71].
As B vitamins, vitamin D has a major role in brain development
and functioning; yet, beyond its well-documented role in calcium
homeostasis and bone health, vitamin D has gained interest only
quite recently for the prevention of age-related brain diseases [76].
Vitamin D results from endogenous synthesis after sun exposure
and to a lower extent (about 20%) from intake of animal products
including fatty fish, eggs, and dairies. The prevalence of vitamin D
deficiency is high worldwide; in a systematic review of studies
including 168,000 individuals from 44 countries, 37% of the
participants had 25(OH)D (the active form of vitamin D) seric
concentrations below 20 ng/mL (50 nmol/L; the threshold for
hypovitaminosis D [i.e., insufficiency or deficiency] according to
the definition of the World Health Organization and the US
Institute of Medicine) [77], with older persons particularly at risk.
Vitamin D is critical for brain function as it participates in neuro-
transmission (vitamin D receptors which are present in several
brain areas, including regions involved in learning and memory).
Furthermore, various neuroprotective properties have been attrib-
uted to vitamin D in preclinical studies, including modulation of
immune response and of inflammatory and oxidative pathways;
promotion of neurogenesis, synaptogenesis, and Aβ clearance; and
prevention of neuronal death [76]. Several prospective observa-
tional studies have reported an association between higher vitamin
D and lower cognitive decline, specifically of executive functions,
whereas associations with episodic memory have been less consis-
tent [78]. Prospective research on the risk of dementia has been
limited to date. A meta-analysis of five observational studies includ-
ing a total of 18,639 participants that compared to individuals with
deficient vitamin D status (25(OH)D < 20 nmol/L) to older per-
sons with sufficient vitamin D (25(OH)D > 50 nmol/L) found a
54% higher risk of dementia associated with vitamin D deficiency
[79]. However, this meta-analysis did not include the most recent
strong association of low vitamin D to cognitive decline and the
risk of dementia and AD in the 3C study [80] and the null findings
from two large cohorts, the Framingham Heart Study [81] and the
Uppsala Longitudinal Study of Adult Men [82].
In summary, 20 years of preclinical and epidemiological
research have established a strong scientific rationale for a role of a
least seven families of nutrients (n-3 PUFA, vitamins E and C,
carotenoids, polyphenols, vitamins B and D) in normal brain func-
tion and for the prevention of age-related brain diseases. Still, all
these nutrients are contained in foods and consumed within a diet;
moreover, they may have synergistic—or at least additive—effects
on metabolic pathways underlying normal brain function and
pathological brain aging. Thus, a higher level of data integration
must consider dietary patterns in relation to health and disease.
24 Cécilia Samieri
2.2 Nutrient Dietary patterns incorporate all dietary habits in a global appraisal
and Dietary Patterns of diet and may be more relevant to brain health and diseases than
nutrients taken in isolation. Indeed, many nutrients exert biologi-
cal effects on the brain (as developed above); yet the combination
of these nutrients in the diet may have a stronger effect than any
nutrient considered individually. Alternatively, individual nutrients
may be not sufficiently powerful to protect the brain from age-
related diseases, and global approaches of diet may be more prom-
ising. Moreover, nutrients may act in synergy (although the concept
of synergy of the food matrix has remained largely hypothetical to
date). For example, antioxidant nutrients (e.g., carotenoids, vita-
min D) may contribute to decrease the peroxidation of PUFAs in
neuronal membranes of the aging brain [83, 84]. Accordingly,
regular intake of fruits and vegetables (the main sources of antioxi-
dant nutrients) appeared necessary to observe a cognitive benefit
of fish consumption (a source of n-3 PUFAs and vitamin D) in the
3C study [28]. As done earlier with cardiovascular diseases, many
cohort studies have thus examined the associations between dietary
patterns and brain aging outcomes.
2.2.1 The Mediterranean Most convincing research on overall diet quality and brain health
Diet: A Landmark Model has certainly been obtained with the Mediterranean diet (MeDi).
of Healthy Diet and Brain The MeDi, now considered as a landmark model for a healthy diet,
Health has been defined as the traditional diet from Mediterranean coun-
tries; it is characterized by high intakes of plant foods (fruits, veg-
etables, cereals, and legumes), low intakes of meats and dairies,
moderate intakes of fish and wine during meals, and olive oil as a
main source of added fats [85]. Primary findings for a protective
role of the MeDi on AD [86] and cognitive aging [87] were pro-
vided by the US cohort WHICAP and the French 3C study. These
initial findings were replicated in several populations (see for review
[85]); a meta-analysis combining five cohort studies found that
subjects with higher MeDi adherence (i.e., in the last tertile of
distribution of a score of adherence to the MeDi) had a 33% lower
risk (95% CI 19–45%) of mild cognitive impairment or AD [88].
However, inconsistent results were also reported, especially with
cognitive decline in the non-Mediterranean North American pop-
ulation (e.g., the MeDi was associated with higher average cogni-
tive status, but not cognitive decline, in two large US cohorts [89,
90]). A Mediterranean diet score adapted to the Australian diet
was associated with less decline in executive function over 3 years,
but only in carriers of the APOEε4 allele [91].
2.2.2 Other Approaches Other indices of diet quality have been investigated in relation to
of Food Quality and Brain dementia outcomes. For example, among 27,860 subjects from 40
Health countries, a modified Alternative Healthy Eating Index, comprising
seven components (including vegetables, fruits, nuts and soybean
proteins, whole grain, deep-fried foods, ratio of fish to meat and
egg, and alcohol), was associated with risk of a ≥3-point decline in
Mini-Mental State Examination score over 5 years [92]. The Dietary

Approaches to Stop Hypertension (DASH) diet, which differs from
the MeDi in that it specifies low consumption of saturated fat and
commercial pastries and sweets and higher consumption of low-fat
dairies, was associated with lower cognitive decline in the Memory
and Aging Project [93], but not in the Nurses’ Health Study [94].
With the objective to conceptualize a diet pattern score more spe-
cific to brain health, Morris et al. recently proposed the MIND
(Mediterranean-DASH Intervention for Neurodegenerative Delay)
diet, a hybrid of the Mediterranean-DASH diets. Similar to the
Mediterranean and DASH diets, the MIND diet score emphasizes
plant-based foods and limited intakes of animal and high saturated
fat foods but uniquely specifies the consumption of berries and
green leafy vegetables and does not specify high fruit, high dairy,
or high potato consumption. The MIND diet score has 15 dietary
components including ten brain healthy food groups (green leafy
vegetables, other vegetables, nuts, berries, beans, whole grains,
fish, poultry, olive oil, and wine) and five unhealthy food groups
(red meats, butter and stick margarine, cheese, pastries and sweets,
and fried/fast food). Higher adherence to the MIND diet was asso-
ciated with lower cognitive decline and a lower risk of AD in the
Memory and Aging Project [95, 96]; yet replications in other
cohorts are warranted to assess the external validity of this diet for
the protection of the brain in non-US populations.
2.2.3 Exploratory Exploratory approaches have also been utilized to identify nutri-
Nutrient/Dietary Patterns ent/dietary patterns related to brain health without formulating
and Brain Health any a priori hypothesis on diet-to-brain relationships. Cluster analy-
sis, principal component analysis, and partial least squares analyses
(and a variant, the reduced rank regression) have been employed to
characterize dietary patterns in various populations (see for review
[97, 98]). Overall, findings from these studies are difficult to com-
pare because the patterns obtained with exploratory approaches are
constitutively specific to the studied population, and limited
research has explored longitudinal associations with dementia out-
comes so far. Studies have often interpreted their patterns as
“healthy” or “prudent” versus “unhealthy” or “western” diets, yet
definitions have varied largely across populations and associations
with cognitive decline/dementia were relatively inconsistent. For
example, in the Whitehall II study, the “prudent” diet, character-
ized by higher intakes of vegetables, fruits, dried legume, and fish,
was related to lower odds of cognitive impairment, while the “west-
ern” pattern (sweetened desserts, chocolates, fried food, processed
meat, pies, refined grains, high-fat dairy products, margarine) was
related to higher odds of cognitive deficit; however, adjustment for
education attenuated these cross-sectional associations [99]. In the
Australian Imaging, Biomarkers and Lifestyle (AIBL) study, the
“prudent” pattern obtained by factor analysis (characterized by high
intakes of vegetables, fruits, nuts, whole grains) was not associated
26 Cécilia Samieri
with cognitive decline over 3 years, while the “western” pattern

(red and processed meats, refined grains, poultry, potatoes, sweets,
margarine, high-fat dairies) was associated with increased decline in
visuospatial performances, but only among APOEε4 allele noncar-
riers [91]. In the Newcastle 85+ Study, three patterns were identi-
fied by cluster analysis; two of them, i.e., pattern 1 associated with
high intakes of red meat, potatoes, and gravy and pattern 2 associ-
ated with high intake of butter, were associated with lower cogni-
tive performances at baseline. However, no association was found
with cognitive decline over 5 years [100].
More convincing findings were obtained in the Swedish National
study on Aging and Care-Kungsholmen (SNAC-K) [101] and in
the US cohort Washington Heights-Inwood Columbia Aging
Project (WHICAP) [102]. For example, in WHICAP, a “healthy”
pattern, identified with reduced rank regression to reflect a diet rich
in n-3 PUFA, n-6 PUFA, vitamin E, and folate, but poor in SFA and
vitamin B12, was characterized by higher intakes of nuts, fish, toma-
toes, poultry, cruciferous vegetables, fruits, and dark and green leafy
vegetables and lower intakes of high-fat dairy, red meat, organ meat,
and butter; in multivariate models, a higher score to this dietary pat-
tern was strongly associated with a decreased risk of AD over 4 years
[102]. Finally, we recently reported in a subsample from the French
3C study (N = 666) a strong association between a nutrient bio-
marker pattern combining several nutrient deficiencies, including
low blood status vitamin D, low carotenoids, and low PUFAs (ver-
sus high SFAs) and the risk of dementia over 12 years. Compared
with individuals in the first quintile of the pattern score, participants
in the highest quintile of score had an approximately fourfold
increased risk of dementia in multivariate models—an association
that was stronger in magnitude than any of the associations between
the individual nutrient biomarkers and dementia [103].
Overall, although the exploratory approaches developed above
are uniquely adapted to identify novel food and nutrient combina-
tions predictive of dementia, their lack of reproducibility across
populations has limited their utility for etiological research so far.
3 Nutrition, Diet, and Alzheimer’s Disease Brain Imaging Biomarkers
With a disease evolving over decades before clinical symptoms

become apparent, biomarkers have been a core feature of research
on dementia and AD [104, 105]. Biomarkers are not only useful in
clinical research to capture early signs of disease in preclinical stages,
but they are also highly relevant to etiological research, as they may
give important insights on potential mechanisms underlying associa-
tions between risk/prevention factors and cognitive aging and
dementia. For example, when applied to research on diet and
dementia, the use of biomarkers may inform on whether associations
of certain dietary factors to dementia risk may be mediated by
pathways acting on lowering brain amyloid specifically, on preserving

brain structure by avoiding neurodegeneration more generally, and/or
on protecting the brain vasculature from age-related damage.
AD biomarkers have been divided into (1) pathophysiological
markers, reflecting AD pathology at any point on the disease con-
tinuum, and (2) topographical and/or prognostic markers, reflecting
downstream damage (e.g., neuronal loss associated with neurode-
generation, white matter microstructural alterations associated with
both neurodegeneration and brain vascular injury) [106].
Pathophysiological biomarkers assess the presence of amyloid and
Tau pathology and include both cerebrospinal fluid biomarkers and
imaging biomarkers obtained with PET (positron emission tomogra-
phy) scans (i.e., PET amyloid and PET Tau). Although not specific
to AD pathology, biomarkers of cerebral small vessel disease may also
be considered as part of pathophysiological biomarkers of AD [106].
Moreover, topographical biomarkers are exclusively assessed through
brain imaging and include markers of macro- and microstructural
changes in the brain at MRI (whole brain and hippocampal atrophy,
cortical thickness obtained with volumetric MRI, alterations of white
matter and gray matter microstructure obtained with diffusion tensor
imaging [DTI]) and markers of hypometabolism of neocortical
regions using fluorodeoxyglucose (FDG)-PET [106]. We will focus
here on brain imaging biomarkers (including markers of cerebral
small vessel disease), which have been studied in relation to nutrition
and diet only recently (very limited research has been conducted on
diet and cerebrospinal fluid biomarkers, and there has been no valid
plasma biomarker to date). We will address studies by order of impor-
tance in the literature in relation to nutrition and diet, by describing,
first, studies on MRI topographical biomarkers reflecting neuronal
loss; second, studies on pathophysiological markers of brain vascular
health; and, last, the very few studies on nutrition and brain amyloid.
Because most of imaging-based research has been cross-sectional,
cross-sectional studies will be included in this paragraph. Relevant
studies have been reviewed and summarized in Table 1.
3.1 Nutrition, Diet, Among individual dietary candidates for brain health, fish intake/
and Topographical LC n-3 PUFA exposures have certainly been among the most
Imaging Biomarkers studied in relation to brain volumes in previous research (Table 1).
(Reflecting Macro-/ We have reviewed six large epidemiological studies, mostly cross-
Microstructural and/or sectional, on the relationships between fish intake or blood status
Metabolic Changes in EPA and/or DHA and volumes of the whole brain, the hippo-
Associated campus, and/or the medial temporal lobe (which includes hippo-
with Neuronal Loss campus and other structures of the limbic system, in particular the
and/or Brain Vascular amygdala). Overall, studies have been inconsistent though. Two
Injury) cross-sectional analyses from the Cardiovascular Health Study
reported null associations [107, 108] (Table 1 lines 1–2); a longi-
3.1.1 Brain Atrophy tudinal retrospective analysis from the ADNI found relationships
limited to APOE4 carriers (Table 1 line 3); whereas three other
cohorts [109, 110] (lines 5–6) (including a single longitudinal
28 Cécilia Samieri
Table 1
Epidemiological studies on the relationship between nutrition, diet, and brain imaging biomarkers of AD dementia
(longitudinal studies are underlined in gray)
Line Type of imaging biomarker Study Study (1st and last Design (cross-sectional Nutritional Main findings (based on
author, year, country, [CS] versus longitudinal Exposure multivariate models)
cohort name) [L]; N; age at baseline
[minimum, mean or
range, whatever
available])
Topographical imaging biomarkers
(reflecting macro/microstructural and/or
metabolic changes associated with
neuronal loss and/or brain vascular injury)
Brain atrophy
1 Atrophy of the wholebrain/hippocampus/ Virtanen and CS; N=4,128 Fish intake Fish intake not associated with
enthorinal cortex/medial temporal lobe Mozaffarian; 2008; participants sulcal or ventricular grades
USA; Cardiovascular without known (visual grading), markers of
Health Study [107] cerebrovascular brain atrophy
disease; ≥65 years
2 Virtanen and CS; N=2,293 Plasma Higher plasma phospholipid
Mozaffarian; 2013; participants Phospholipid n-3 long-chain n-3 PUFA
USA; Cardiovascular without known PUFA (EPA+DPA+DHA) not associated
Health Study cerebrovascular with sulcal or ventricular grades
[107,108] disease; ≥65 years (visual grading), markers of
brain atrophy
3 Daiello and Ott for L (retrospective cohort; Fish oil use (mostly Fish oil use associated with
the Alzheimer’s exposure an outcomes of moderate to lower of both the whole brain
Disease modeled concomitantly long duration, ie, (for normal controls and MCI
Neuroimaging over 4 years); N=819; 70% had used fish participants) and the
Initative (ADNI); 55-90 years oil during > 1 year hippocampus (for normal
2015; USA; ADNI [30] before baseline) controls and AD patients), and
less cognitive decline but only
among APOE4 non-carriers
4 Samieri and L (volume change over Plasma long-chain Plasma EPA associated with
Barberger-Gateau; 4 years); N=281; 72.3 n-3 PUFA at lower atrophy in part of the
2012; France; Three- years baseline right medial-temporal lobe;
City Study [111] anatomical changes linked with
slower cognitive decline
5 Tan and Seshadri; CS; N=1,575 without EPA and DHA in red Lower DHA level (Q1 versus Q2-
2012; USA; the dementia; 67 years blood cells Q4) associated with lower total
Framingham brain volume
Offspring Study [109]
6 Pottala and Harris; CS (with brain imaging EPA and DHA in red Higher EPA+DHA in red blood
2014; USA; Women’s 8 years after blood blood cells cells (ie, omega-3 index)
Health Initiative draw); N=1,111 associated with larger brain
Memory Study women; 65 to 80 years volume and larger hippocampal
(WHIMS) [110] volume
7 Bowman and Quinn; CS; N=42 (with brain Nutrient biomarker A nutrient biomarker pattern
2012; USA; the imaging) with CDR<0.5; patterns (based on “BCDE” characterized by higher
Oregon Brain Aging 87 years a set of 30 plasma vitamins B (B1, B2, B6,
Study cohort [125] biomarkers) at folate, B12), C, D and E
baseline associated with higher total
brain volume
8 Titova and Benedict; CS (with brain imaging Mediterranean diet Higher Mediterranean diet
2013; Sweden; 5 years after diet score at baseline score not associated total brain
Prospective assessment); N=194; volume 5 years later (higher
Investigation of the 70.1 years meat intake significantly
Vasculature in associated with worse volume
Uppsala Seniors [114] and worse cognition).
9 Gu and Scarmeas; CS; N=674; 80.1 years Mediterranean diet Higher Mediterranean diet
2015; USA; score score associated with larger
Washington total brain volumes (especially
Heights/Hamilton high fish and low meat);
Heights Inwood associations were equivalent to
Columbia Aging 5 years of aging.
Project (WHICAP)
[113]
10 Pelletier and Samieri, CS (with brain imaging Mediterranean diet Higher Mediterranean diet
2015, Three-City 9 years after diet score at baseline score not associated with grey
Study [115] assessment); N=146; matter or white matter brain
73 years volumes in any region
11 Luciano and Deary; L (volume and corticial Mediterranean diet Higher Mediterranean diet
2017; Scotland; thickness change over score score not associated with total
Lothian Birth Cohort 3 years, with baseline brain volumes 3 years and 6
of 1936 [116] brain imaging 3 years years after diet assessment, but
after diet assessment); significantly associated with
N=401 (346 for cortical lower atrophy over 3 years (no
thickness) specific association with fish or
meat found)
(continued)
12 Cortical thickness Walhovd and Fjell; L (change over 3.6 Several exposures Higher DHA associated with
2014; Norway; years, with blood including plasma less cortical thinning (preceding
Cognition and biomarkers assessed at long-chain n-3 biomarker assessment) in the
Plasticity through the the second MRI visit); PUFA and vitamin left middle and superior
Lifespan project [112] N=92; 53.8 years D temporal cortex; higher vitamin
(range 23.5 to 87.8 D associated with less cortical
years) thinning in right lateral
prefrontal cortex
13 Mosconi and de Leon; CS; N=52; 54 years Mediterranean diet Higher Mediterranean diet
2014; USA; study at score score associated with larger
New York University cortical thickness in the
school of Medicine entorhinal, orbitofrontal and
[117] posterior cingulate regions
14 Gu and Scarmeas; CS; N=674; 80.1 years Mediterranean diet Higher Mediterranean diet
2015; USA; score score not associated with
Washington cortical thickness (association
Heights/Hamilton of fish to greater cortical
Heights Inwood thickness only).
Columbia Aging
Project (WHICAP)
[113]
15 Staubo and Roberts, CS; N=672; 79.8 years Mediterranean diet Higher Mediterranean diet
2017, USA, Mayo score score associated with larger
Clinic Study of Aging cortical thickness in many
[118] regions (especially high legume
and fish and low carbohydrate
and sugar), but not significantly
with an AD signature cortical
thickness
16 Luciano and Deary; L (change over 3 years, Mediterranean diet Higher Mediterranean diet
2017; Scotland; with baseline brain score score neither associated with
Lothian Birth Cohort imaging 3 years after cortical thickness 3 years and 6
of 1936 [116] diet assessment); years after diet assessment nor
N=401 (346 for cortical with cortical thickness change
thickness) over 3 years
Brain microstructure
17 White matter microstructure (DTI) Gu and Brickman, CS (with brain imaging Nutrient intake A pattern characterized by
2016, USA, WHICAP 5.1 years after diet patterns (based on higher intakes of PUFA (n-3 and
[119] assessment); N=239; a set of 24 n-6) and vitamin E associated
84.1 years nutrients) with higher average fractional
anisotropy; this association
with white matter
microstructure mediated the
relations of the pattern to
memory and cognition
18 Pelletier and Samieri, CS (with brain imaging Mediterranean diet Higher Mediterranean diet
2015, Three-City 9 years after diet score at baseline score associated with higher
Study [115] assessment); N=146 fractional anisotropy and lower
not demented at mean diffusivity in many white
baseline; 73 years matter regions; no significant
association with brain volumes
Brain hypometabolism
19 Mosconi and de Leon; CS; N=49 cognitively Intakes of 10 Higher intake of β carotene and
2014; USA; study at normal individuals candidate nutrients folate from food sources
New York University (70% with family associated with higher brain
school of Medicine history of late-onset AD metabolism in AD regions of
[121] and 39% carriers of the interest at 18F-
APOE4 allele); 54 years Fluorodeoxyglucose (FDG)-PET
(range 25 to 72 years)
20 Berti and Mosconi; CS; N=52 cognitively Nutrient intake Several nutrient intake patterns
2015; USA; study at normal individuals patterns (based on associated with higher brain
New York University (69% with family 35 nutrients) metabolism in AD regions:
school of Medicine history of late-onset AD - A pattern with higher intakes
[122] and 39% carriers of the of vitamin B12, vitamin D and
APOE4 allele); 54 years Zinc (also associated with
(range 25 to 72 years) lower amyloid load and higher
grey matter volume)
- A pattern with lower intakes
of saturated and trans fats,
cholesterol and sodium (also
(continued)
30 Cécilia Samieri
associated with higher grey

matter volume)
- A pattern with higher MUFA
and PUFA intakes
- A pattern with higher
carotenoids, vitamins A and C,
fibers
Pathophysiological imaging biomarkers
Markers of cerebral small vessel disease

21 Brain infarcts Scarmeas and CS (with brain imaging Mediterranean diet Higher Mediterranean diet
Brickman; 2011; 5.8 years after diet score score associated with reduced
WHICAP [126] assessment); N=707; burden of brain infarcts
80.3 years
22 Subclinical infarcts Virtanen and CS and L (2 MRI 5 years Fish intake Consuming tuna/other fish ≥3
Mozaffarian; 2008; apart); N=4,128 for CS times/week associated with
USA; Cardiovascular and N=1,124 for L lower prevalence of subclinical
Health Study [107] analyses, without known infarcts (ischemic lesions
cerebrovascular ≥3mm) compared to
disease; ≥65 years <once/month; non-significant
tend for association with
incident subclinical infarcts
23 Virtanen and CS and L (2 MRI 5 years Plasma Higher plasma phospholipid
Mozaffarian; 2013; apart); N=2,293 for CS Phospholipid n-3 long-chain n-3 PUFA
USA; Cardiovascular and N=1,056 for L PUFA (EPA+DPA+DHA) associated
Health Study analyses, without known with lower prevalence of
[107,108] cerebrovascular subclinical infarcts; no
disease; ≥65 years association with incident
subclinical infarcts
24 White matter hyperintensities Tan and Seshadri; CS; N=1,575 without EPA and DHA in red Higher DHA level associated
2012; USA; the dementia; 67 years blood cells with lower white matter hyper-
Framingham intensities volume but not after
Offspring Study [109] adjustment for the full set of
potential confounders; no
association with silent cerebral
brain infarcts
25 Pottala and Harris; CS (with brain imaging EPA and DHA in red EPA+DHA in red blood cells (ie,
2014; USA; Women’s 8 years after blood blood cells omega-3 index) not associated
Health Initiative draw); N=1,111 with ischemic lesion volume
Memory Study women; 65 to 80 years (represented by both diffuse
(WHIMS) [110] small-vessel disease and white
matter hyperintensities)
26 Bowman and Quinn; CS; N=42 (with brain Nutrient biomarker A nutrient biomarker pattern
2012; USA; the imaging) with CDR<0.5; patterns (based on characterized by higher plasma
Oregon Brain Aging 87 years a set of 30 long-chain n-3 PUFA associated
Study cohort [125] biomarkers) at with lower white matter hyper-
baseline intensities volume, but only
among participants without
depression
Biomarkers of brain amyloid

27 Mosconi and de Leon; CS; N=49 cognitively Intakes of 10 Higher intake of vitamin B12,
2014; USA; study at normal individuals candidate nutrients vitamin D, n-3 PUFA from food
New York University (70% with family sources associated with lower
school of Medicine history of late-onset AD amyloid load in AD regions at
[121] and 39% carriers of the 11C-Pittsburgh Compound B
APOE4 allele); 54 years (PIB)-PET

28 Berti and Mosconi; CS; N=52 cognitively Nutrient intake A nutrient intake pattern
2015; USA; study at normal individuals patterns (based on characterized by higher intakes
New York University (69% with family 35 nutrients) of vitamin B12, vitamin D and
school of Medicine history of late-onset AD Zinc associated with lower
[122] and 39% carriers of the amyloid load
APOE4 allele); 54 years
29 Yassine and Chui; CS; N=61 cognitively Blood DHA Serum DHA inversely related to
USA; Aging Brain normal individuals; 77 amyloid load using 11C PIB-PET
Study [31] years (range 67 to 88
years)
report from the Three-City (3C) study [111] (line 4)) found
significant associations of higher blood LC n-3 PUFA status to
higher brain volumes (/lower brain atrophy in 3C). In comple-
ment to these studies on brain volumes, two studies reported inter-
esting associations between fish/DHA and cortical thickness. The
Cognition and Plasticity through the Lifespan project found a lon-
gitudinal association between higher blood DHA and less cortical
thinning in part of the left temporal cortex [112] (line 12).
However, the study was strongly limited by a biomarker assess-
ment performed after ascertainment of cortical thinning.
Furthermore, in a cross-sectional analysis from the large MRI sam-
ple of the WHICAP, higher fish intake was the single component
of the MeDi score associated with greater cortical thickness [113]
(line 14). Thus, the relation of fish/LC n-3 PUFAs to cortical
thickness (a subtle marker of neuronal loss) certainly deserves fur-
ther research.
Aside from studies on fish/n-3 PUFAs, a number of cohorts
investigated overall diet through adherence to the MeDi; however,
as with fish, inconsistent results were found in relation to brain
atrophy. For example, both the Swedish Prospective Investigation
of the Vasculature in Uppsala Seniors and the 3C study found no
association of the MeDi to brain volumes [114, 115] (lines 8 and
10). Likewise, in the Lothian Birth Cohort, there was no associa-
tion between the MeDi and brain volumes assessed 3 years apart;
yet adherence to the MeDi was related to lower brain atrophy
between the two time points [116] (line 10). Only the WHICAP
reported a significant association between the MeDi and total brain
volumes [113] (line 9). More recent studies on cortical thickness
have not been more consistent than those which targeted volumes.
Two US cohorts, including the large sample (N > 500) from the
Mayo Clinic Study of Aging, reported a cross-sectional relation
between the MeDi and larger cortical thickness [117, 118] (lines
13 and 15). In contrast, none of the two large WHICAP and
Lothian Birth Cohort studies found any significant association of
the MeDi to cortical thickness [113, 116] (lines 14 and 16).
3.1.2 Alterations of Brain Studies on diet and brain microstructure, which have been
Microstructure extremely limited to date, yielded promising findings. Both the
WHICAP and the 3C study found associations of (1) a nutrient
pattern with higher PUFAs and vitamin E and (2) the MeDi,
respectively, to preserved white matter structure [115, 119]
(Table 1 lines 17 and 18). Interestingly, in both studies, the
reported associations were accompanied by strong relations with
cognitive function, suggesting that alterations in brain connectivity
may serve as a mediator in the relationship between low diet qual-
ity and cognitive aging. Moreover, in a small clinical trial (N = 65),
26 weeks of supplementation with EPA + DHA (2.2 g/day) in
older persons led to an improvement of white matter microstruc-
32 Cécilia Samieri
ture compared to placebo [120] (intervention study not shown in

Table 1). Together, these findings suggest that brain connectivity
may be a key neuroanatomical substrate for the effect of foods and
nutrients on brain health. Larger epidemiological studies are war-
ranted to better understand how diet may influence brain
connectomics.
3.1.3 Brain As with cerebral microstructure, studies on diet and brain hypome-
Hypometabolism tabolism are at their infancy. Two small studies from the NY
University School of Medicine reported associations between
nutrient intake patterns and higher brain metabolism in AD regions
of interest [121, 122] (Table 1, lines 19 and 20). However, in both
studies the number of nutrients investigated was almost superior to
the number of subjects included, and study samples were highly
selected. These pilot studies certainly deserve validation in larger
samples.
3.2 Nutrition, Diet, The term cerebral small vessel disease refers to a group of patho-
and Pathophysiological logical processes with various etiologies that affect the small arter-
Imaging Biomarkers ies, arterioles, venules, and capillaries of the brain. The causes are
3.2.1 Markers multiple and include cerebral amyloid angiopathy, hypertension,
of Cerebral Small and vascular risk factors in general. As a consequence of small ves-
Vessel Disease sel disease, various subcortical lesions develop in brain parenchyma,
including lacunar infarcts, white matter lesions, large hemorrhages,
and microbleeds [123]. These lesions reflecting small vessel disease
can be detected at MRI; as their prevalence dramatically increases
with age, they are commonly considered part of the spectrum of
vascular-related brain injury associated with brain aging. For exam-
ple, white matter hyperintensities (WMH), defined as white matter
lesions appearing as hyperintensities on T2-weighted images at
MRI, range from 11–21% around 64 years old to 94% at age 82
[124].
Assuming that diet, which has a primary influence on vascular
risk factors, may exert a beneficial role on the brain vasculature,
several of the largest cohorts on diet and dementia have examined
the relations between nutritional factors and MRI markers of cere-
bral small vessel disease. Primary studies were published a decade
ago; still, existing literature has remained scarce (mostly limited to
fish/n-3 PUFA) and, as with brain atrophy, generally inconsistent
(Table 1). The Cardiovascular Health Study was among the first to
examine the relation of fish/LC n-3 PUFA status to subclinical
infarcts; despite significant cross-sectional associations and a very
large sample size (N > 1000), findings were null for the prospective
relationships with incident lesions [107, 108] (Table 1 lines 22 and
23). In addition, two large US cohorts (N= > 1000) failed to evi-
dence any relation of blood LC n-3 PUFA to WMH volume [109,
110] (lines 24 and 25). A relation between a nutrient biomarker
pattern reflecting higher plasma LC n-3 PUFA and lower WMH
volume was reported in a small sample from the Oregon Brain
Aging Study Cohort [125]; however, the association was limited

to participants without depression, with a high probability of
chance finding in such a small sample (line 26). One of the rare
positive findings pertains to the WHICAP, in which a higher MeDi
score was associated with reduced burden of brain infarcts in a
cross-sectional analysis [126] (line 21). Despite this positive find-
ing, overall, one may conclude that no consistent relationship
between diet and conventional MRI markers of cerebral small ves-
sel disease has been found to date. This is not to say that nutrition
and diet do not influence brain health through preservation of
brain vasculature; more convincing findings may be obtained with
more subtle markers of neurodegeneration and brain vascular
injury, such as markers of white matter microstructure alterations,
as developed above.
3.2.2 Markers of Brain PET amyloid reflects the neuropathological hallmark of AD and is
Amyloid thus considered the “king” biomarker in AD research. However,
given its elevated cost, virtually no research has been made on diet
and PET amyloid so far. Recently, the Atherosclerosis Risk in
Communities (ARIC) study found a relationship between midlife
obesity and brain amyloid deposition 20 years later, suggesting a
strong, lifelong influence of diet-related risk factors on AD neuro-
pathology [127]. However, the dietary factors likely to influence
amyloid load have remained understudied to date. One of the sin-
gle studies has been conducted by the NY University School of
Medicine, which recently published two pilot studies on nutrients
and brain amyloid [121, 122] (Table 1, lines 27 and 28). In one of
these studies, which included 49 cognitively normal individuals
with high familial and genetic risk for AD, higher intakes of vita-
min B12, vitamin D, and n-3 PUFA were related to lower amyloid
load in regions with typical amyloid accumulation in AD. Moreover,
a small cross-sectional study from the Aging Brain Study reported
a correlation between higher blood DHA and lower amyloid load
among cognitively normal older adults (line 29). Added to the
important findings on midlife obesity and brain amyloid load, these
promising preliminary findings with nutrients call for additional
large-scale studies on diet and brain amyloid in future research.
4 Conclusion: Toward Innovative Approaches for Future Epidemiological

Research on Nutrition, Diet, and Alzheimer’s Disease
Epidemiological research on diet and dementia conducted over the

past 15–20 years has provided a considerable insight into the
impact of lifestyle on brain health. Many large-scale observational
studies have confirmed the importance of several nutrients and
foods for maintaining optimal brain structure and function with
aging, as suggested by preclinical studies. However, despite the
34 Cécilia Samieri
amount of positive findings gained from preclinical and observa-

tional epidemiological research, surprisingly, randomized con-
trolled trials which have tested the efficacy of supplementation in
individual nutrients (mostly n-3 PUFAs, vitamin E, and B vita-
mins) generally failed to demonstrate any clinical efficacy on cogni-
tion in humans (see a number of meta-analyses published recently
[128–132]). Several reasons may explain the discordance between
observational and interventional studies in diet and brain aging.
Observational studies have well-recognized limitations (e.g., mea-
surement error and residual confounding); however, randomized
trials in AD dementia have also a number of limitations, mainly due
to the complexity of a disease which evolves silently over decades,
and there may be several methodological explanations for the fail-
ure of most trials to demonstrate the clinical efficacy of nutrients in
brain aging. Methodological limitations of existing trials may
include (1) a too late time window for intervention (i.e., most tri-
als have included mild-to-moderate AD patients, while those most
likely to benefit from a nutritional intervention may need to be
relatively preserved from extended and irreversible neurodegenera-
tion), (2) a too short duration of intervention (i.e., many trials
tested less than 1 year of supplementation in AD dementia, while
long-term preventive interventions are certainly more relevant for
a disease which potentially evolves over decades), and (3) failure to
take into account baseline nutritional exposures (i.e., those most
likely to benefit from supplementation of a given nutrient may
have low intakes of that nutrient in their habitual diet).
It is also possible that single nutrients, which have been gener-
ally modestly associated with cognitive decline or the risk of
dementia in observational studies, are not sufficiently active to
modulate brain aging and that a combination of nutrients with
complementary biological properties (e.g., LC n-3 PUFA and
antioxidant nutrients) may be more effective for the prevention of
AD dementia. Accordingly, promising observational findings were
obtained with overall healthy diets (e.g., the MeDi and the MIND
diets [96, 133]) or with nutrient patterns [103]. Recently, a very
large US trial (N > 5000) failed to evidence the benefit of long-
term multivitamin supplementation (vitamins E, C, and β-carotene)
on cognitive aging [134]; however, the authors acknowledged that
the population of male health professional included in the trial may
be too well-nourished to benefit from a multivitamin. It is also pos-
sible that the combination of antioxidant vitamins tested in this
large trial may not be sufficient and that a broader set of nutrients
may be more relevant. For example, Nutricia developed the
Souvenaid (Fortasyn Connect) supplement, including 11 nutrients
involved in synaptic synthesis (uridine monophosphate; choline;
phospholipids; EPA; DHA; vitamins E, C, B12, and B6; folic acid;
and selenium), for the prevention of AD dementia. However, after
promising results of a 2-year phase II trial in mild-to-moderate AD
[135], Souvenaid failed to lower cognitive decline as an add-on of

usual AD medication in a phase III study (N = 527) [136] and did
not improve decline in global cognition in prodromal dementia
after 24- month supplementation (N=382); still, benefit was
observed in important secondary outcomes including hippocampal
atrophy [137]. In contrast with the mitigated findings obtained so
far with the two nutrient combinations cited above, the Spanish
PREDIMED (Prevención con Dieta Mediterránea) study evi-
denced the efficacy, for individuals with elevated cardiovascular
risk, of adopting a Mediterranean diet supplemented with virgin
olive oil or nuts for 5 years for age-related cognitive decline [138].
These significant interventional findings may certainly constitute a
strong positive signal for the relevance of healthy diet/nutrient
combinations in the prevention of AD dementia.
Thus, there is a large room for nutritional epidemiology in
future research on AD prevention. The challenges posed by
research on behavioral and biological risk factors of AD are multi-
ple and may be successfully addressed using innovative approaches
such as “omics” methods, cutting-edge imaging biomarkers, and
advanced statistical methods for the modeling of big data and bio-
marker trajectories. The first challenge to face with will be to deter-
mine the optimal combination of foods/nutrients for prevention.
It may include some of the candidate nutrients with established
associations with brain aging (e.g., vitamin D, a very promising
candidate yet understudied in combination with other nutrients).
The optimal food/nutrient combination may also incorporate
some novel dietary bioactives. For example, interesting molecules
for prevention could be identified in the food metabolome—
defined as the thousands of nutrients/non-nutrient metabolites
present in biofluids and tissues that directly derive from food diges-
tion. The food metabolome is a novel research area which promises
to give important insights on the role of diet in brain health.
Beyond the composition of the active compound(s), another
important challenge for future research will be to fully elucidate
the optimal target population for a preventive intervention. There
is suggestion that early time windows of exposure in primary pre-
vention may be more relevant than secondary preventive interven-
tions targeting demented persons in older age; yet the exact age
range at which nutrition mostly impacts AD neuropathology is still
unknown. Longitudinal epidemiological research, with repeated
measurements of nutritional exposures and AD clinical and imag-
ing markers, will be key to better understand the dynamic relation-
ships between trajectories of behavioral factors and AD biomarker
trajectories.
Finally, mechanistic epidemiological research will need to be
expanded to fully elucidate the multiple pathways by which nutri-
tion/diet shapes the brain and prevents AD. Indeed, a better
understanding of pathways will help refine the composition of
36 Cécilia Samieri
effective interventions and potentially identify vulnerable populations

(e.g., those carrying specific polymorphisms for certain genes).
For example, extensive research has been developed on the gut
microbiome as a potential mediator of the relationship between
diet and brain health—the gut-brain axis being certainly one of the
most exciting fields of investigation open for the next decades.
References
1. World Alzheimer report 2016 (2016) 12. Su HM (2010) Mechanisms of n-3 fatty
https://www.alz.co.uk/research/world- acid-mediated development and maintenance
report-2016. Accessed 16 May 2017 of learning memory performance. J Nutr
2. Lambert JC, Ibrahim-Verbaas CA, Harold Biochem 21(5):364–373. https://doi.
D et al (2013) Meta-analysis of 74,046 indi- org/10.1016/j.jnutbio.2009.11.003
viduals identifies 11 new susceptibility loci for 13. Lim GP, Calon F, Morihara T et al (2005) A
Alzheimer’s disease. Nat Genet 45(12):1452– diet enriched with the omega-3 fatty acid doc-
1458. https://doi.org/10.1038/ng.2802 osahexaenoic acid reduces amyloid burden in
3. Scheltens P, Blennow K, Breteler MM et al an aged Alzheimer mouse model. J Neurosci
(2016) Alzheimer’s disease. Lancet. https:// 25(12):3032–3040
doi.org/10.1016/S0140-6736(15)01124-1 14. Barberger-Gateau P, Letenneur L, Deschamps
4. Grasset L, Brayne C, Joly P et al (2016) V et al (2002) Fish, meat, and risk of demen-
Trends in dementia incidence: evolution tia: cohort study. BMJ 325(7370):932–933
over a 10-year period in France. Alzheimers 15. Morris MC, Evans DA, Tangney CC et al
Dement 12(3):272–280. https://doi. (2005) Fish consumption and cognitive
org/10.1016/j.jalz.2015.11.001 decline with age in a large community study.
5. Satizabal CL, Beiser AS, Chouraki V et al Arch Neurol 62(12):1849–1853
(2016) Incidence of dementia over three 16. Schaefer EJ, Bongard V, Beiser AS et al (2006)
decades in the Framingham Heart Study. N Plasma phosphatidylcholine docosahexae-
Engl J Med 374(6):523–532. https://doi. noic acid content and risk of dementia and
org/10.1056/NEJMoa1504327 Alzheimer disease: the Framingham Heart
6. Small SA, Duff K (2008) Linking abeta and Study. Arch Neurol 63(11):1545–1550
tau in late-onset Alzheimer’s disease: a dual 17. Samieri C, Feart C, Letenneur L et al (2008)
pathway hypothesis. Neuron 60(4):534– Low plasma eicosapentaenoic acid and
542. https://doi.org/10.1016/j. depressive symptomatology are independent
neuron.2008.11.007 predictors of dementia risk. Am J Clin Nutr
7. Kumar DK, Choi SH, Washicosky KJ et al 88:714–721
(2016) Amyloid-beta peptide protects 18. Lopez LB, Kritz-Silverstein D, Barrett Connor
against microbial infection in mouse and E (2011) High dietary and plasma levels of
worm models of Alzheimer’s disease. Sci the omega-3 fatty acid docosahexaenoic acid
Transl Med 8(340):340ra372. https://doi. are associated with decreased dementia risk:
org/10.1126/scitranslmed.aaf1059 the Rancho Bernardo study. J Nutr Health
8. Zhao Z, Nelson AR, Betsholtz C et al Aging 15(1):25–31
(2015) Establishment and dysfunction of 19. Huang TL (2010) Omega-3 fatty acids,
the blood-brain barrier. Cell 163(5):1064– cognitive decline, and Alzheimer’s disease: a
1078. https://doi.org/10.1016/j. critical review and evaluation of the literature.
cell.2015.10.067 J Alzheimers Dis 21(3):673–690. https://
9. Stern Y (2009) Cognitive reserve. doi.org/10.3233/JAD-2010-090934
Neuropsychologia 47(10):2015–2028 20. Zhang Y, Chen J, Qiu J et al (2016) Intakes of
10. Shah H, Albanese E, Duggan C et al fish and polyunsaturated fatty acids and mild-
(2016) Research priorities to reduce the to-severe cognitive impairment risks: a dose-
global burden of dementia by 2025. Lancet response meta-analysis of 21 cohort studies.
Neurol 15(12):1285–1294. https://doi. Am J Clin Nutr 103(2):330–340. https://
org/10.1016/s1474-4422(16)30235-6 doi.org/10.3945/ajcn.115.124081
11. Cunnane SC, Plourde M, Pifferi F et al (2009) 21. Burdge GC, Calder PC (2005) Conversion of
Fish, docosahexaenoic acid and Alzheimer’s alpha-linolenic acid to longer-chain polyun-
disease. Prog Lipid Res 48(5):239–256. saturated fatty acids in human adults. Reprod
doi:S0163-7827(09)00018-6 [pii]10.1016/j. Nutr Dev 45(5):581–597. https://doi.
plipres.2009.04.001 org/10.1051/rnd:2005047
22. Morris MC, Brockman J, Schneider JA et al 34. Geleijnse JM, Hollman P (2008) Flavonoids
(2016) Association of seafood consumption, and cardiovascular health: which com-
brain mercury level, and APOE epsilon4 pounds, what mechanisms? Am J Clin Nutr
status with brain neuropathology in older 88(1):12–13
adults. JAMA 315(5):489–497. https://doi. 35. Corder R, Mullen W, Khan NQ et al (2006)
org/10.1001/jama.2015.19451 Oenology: red wine procyanidins and vascular
23. van de Rest O, Wang Y, Barnes LL et al health. Nature 444(7119):566. https://doi.
(2016) APOE epsilon4 and the associa- org/10.1038/444566a
tions of seafood and long-chain omega-3 36. Brickman AM, Khan UA, Provenzano FA
fatty acids with cognitive decline. Neurology et al (2014) Enhancing dentate gyrus func-
86(22):2063–2070. https://doi. tion with dietary flavanols improves cognition
org/10.1212/WNL.0000000000002719 in older adults. Nat Neurosci 17(12):1798–
24. Yassine HN, Braskie MN, Mack WJ et al 1803. https://doi.org/10.1038/nn.3850
(2017) Association of docosahexaenoic acid 37. Andres-Lacueva C, Shukitt-Hale B,
supplementation with Alzheimer disease stage Galli RL et al (2005) Anthocyanins in
in apolipoprotein E epsilon4 carriers: a review. aged blueberry-fed rats are found cen-
JAMA Neurol. https://doi.org/10.1001/ trally and may enhance memory. Nutr
jamaneurol.2016.4899 Neurosci 8(2):111–120. https://doi.
25. Laitinen MH, Ngandu T, Rovio S et al (2006) org/10.1080/10284150500078117
Fat intake at midlife and risk of dementia and 38. Marambaud P, Zhao H, Davies P (2005)
Alzheimer’s disease: a population-based study. Resveratrol promotes clearance of Alzheimer’s
Dement Geriatr Cogn Disord 22(1):99–107 disease amyloid-beta peptides. J Biol Chem
26. Samieri C, Feart C, Proust-Lima C et al (2011) 280(45):37377–37382. https://doi.
Omega-3 fatty acids and cognitive decline: org/10.1074/jbc.M508246200
modulation by ApoEepsilon4 allele and 39. Ho L, Ferruzzi MG, Janle EM et al (2013)
depression. Neurobiol Aging 32(12):2313– Identification of brain-targeted bioactive
2322. https://doi.org/10.1016/j. dietary quercetin-3-O-glucuronide as a novel
neurobiolaging.2010.03.020 intervention for Alzheimer’s disease. FASEB
27. Huang TL, Zandi PP, Tucker KL et al (2005) J 27(2):769–781. https://doi.org/10.1096/
Benefits of fatty fish on dementia risk are fj.12-212118
stronger for those without APOE epsilon4. 40. Wang J, Ho L, Zhao Z et al (2006)
Neurology 65(9):1409–1414 Moderate consumption of Cabernet
28. Barberger-Gateau P, Raffaitin C, Letenneur Sauvignon attenuates Abeta neuropathology
L et al (2007) Dietary patterns and risk in a mouse model of Alzheimer’s disease.
of dementia: the Three-City cohort study. FASEB J 20(13):2313–2320. https://doi.
Neurology 69(20):1921–1930 org/10.1096/fj.06-6281com
29. Whalley LJ, Deary IJ, Starr JM et al (2008) 41. Boeing H, Bechthold A, Bub A et al (2012)
n-3 Fatty acid erythrocyte membrane con- Critical review: vegetables and fruit in the
tent, APOE {varepsilon}4, and cognitive vari- prevention of chronic diseases. Eur J Nutr
ation: an observational follow-up study in late 51(6):637–663. https://doi.org/10.1007/
adulthood. Am J Clin Nutr 87(2):449–454 s00394-012-0380-y
30. Daiello LA, Gongvatana A, Dunsiger S 42. Xu W, Wang H, Wan Y et al (2017) Alcohol
et al (2015) Association of fish oil supple- consumption and dementia risk: a dose-
ment use with preservation of brain volume response meta-analysis of prospective studies.
and cognitive function. Alzheimers Dement Eur J Epidemiol. https://doi.org/10.1007/
11(2):226–235. https://doi.org/10.1016/j. s10654-017-0225-3
jalz.2014.02.005 43. Samieri C, Feart C, Proust-Lima C et al
31. Yassine HN, Feng Q, Azizkhanian I et al (2011) Olive oil consumption, plasma oleic
(2016) Association of serum docosa- acid, and stroke incidence: the Three-City
hexaenoic acid with cerebral amyloidosis. Study. Neurology 77(5):418–425. https://
JAMA Neurol. https://doi.org/10.1001/ doi.org/10.1212/WNL.0b013e318220abeb
jamaneurol.2016.1924 44. Berr C, Portet F, Carriere I et al (2009)
32. Grimm MO, Mett J, Hartmann T (2016) Olive oil and cognition: results from the
The impact of vitamin E and other fat-soluble three-city study. Dement Geriatr Cogn
vitamins on Alzheimer’s disease. Int J Mol Disord 28(4):357–364. https://doi.
Sci 17(11). https://doi.org/10.3390/ org/10.1159/000253483
ijms17111785 45. Kim YS, Kwak SM, Myung SK (2015)
33. Ransom J, Morgan PJ, McCaffery PJ et al Caffeine intake from coffee or tea and cogni-
(2014) The rhythm of retinoids in the brain. tive disorders: a meta-analysis of observational
J Neurochem 129(3):366–376. https://doi. studies. Neuroepidemiology 44(1):51–63.
org/10.1111/jnc.12620 https://doi.org/10.1159/000371710
38 Cécilia Samieri
46. Bjelakovic G, Nikolova D, Gluud LL et al Aging 29(9):1394–1403. https://doi.

(2007) Mortality in randomized trials of org/10.1016/j.neurobiolaging.2007.03.006
antioxidant supplements for primary and 57. Commenges D, Scotet V, Renaud S et al
secondary prevention: systematic review and (2000) Intake of flavonoids and risk of
meta-analysis. JAMA 297(8):842–857 dementia. Eur J Epidemiol 16(4):357–363
47. Li FJ, Shen L, Ji HF (2012) Dietary intakes of 58. Letenneur L, Proust-Lima C, Le Gouge A
vitamin E, vitamin C, and beta-carotene and et al (2007) Flavonoid intake and cognitive
risk of Alzheimer’s disease: a meta-analysis. decline over a 10-year period. Am J Epidemiol
J Alzheimers Dis 31(2):253–258. https:// 165(12):1364–1371
doi.org/10.3233/JAD-2012-120349 59. Kesse-Guyot E, Fezeu L, Andreeva VA
48. Helmer C, Peuchant E, Letenneur L et al et al (2012) Total and specific polyphenol
(2003) Association between antioxidant intakes in midlife are associated with cogni-
nutritional indicators and the incidence of tive function measured 13 years later. J Nutr
dementia: results from the PAQUID pro- 142(1):76–83. https://doi.org/10.3945/
spective cohort study. Eur J Clin Nutr jn.111.144428
57(12):1555–1561 60. Devore EE, Kang JH, Breteler MM et al
49. Mangialasche F, Kivipelto M, Mecocci P (2012) Dietary intakes of berries and fla-
et al (2010) High plasma levels of vitamin E vonoids in relation to cognitive decline.
forms and reduced Alzheimer’s disease risk in Ann Neurol 72(1):135–143. https://doi.
advanced age. J Alzheimers Dis 20(4):1029– org/10.1002/ana.23594
1037. https://doi.org/10.3233/ 61. Root M, Ravine E, Harper A (2015) Flavonol
JAD-2010-091450 intake and cognitive decline in middle-aged
50. Cherubini A, Martin A, Andres-Lacueva adults. J Med Food 18(12):1327–1332.
C et al (2005) Vitamin E levels, cognitive https://doi.org/10.1089/jmf.2015.0010
impairment and dementia in older persons: 62. Rabassa M, Cherubini A, Zamora-Ros R et al
the InCHIANTI study. Neurobiol Aging (2015) Low levels of a urinary biomarker of
26(7):987–994 dietary polyphenol are associated with sub-
51. Ravaglia G, Forti P, Lucicesare A et al (2008) stantial cognitive decline over a 3-year period
Plasma tocopherols and risk of cognitive in older adults: the Invecchiare in Chianti
impairment in an elderly Italian cohort. Am study. J Am Geriatr Soc 63(5):938–946.
J Clin Nutr 87(5):1306–1313 https://doi.org/10.1111/jgs.13379
52. Mangialasche F, Solomon A, Kareholt 63. Reynolds E (2006) Vitamin B12, folic acid,
I et al (2013) Serum levels of vitamin E and the nervous system. Lancet Neurol
forms and risk of cognitive impairment 5(11):949–960
in a Finnish cohort of older adults. Exp 64. Seshadri S, Beiser A, Selhub J et al (2002)
Gerontol 48(12):1428–1435. https://doi. Plasma homocysteine as a risk factor for
org/10.1016/j.exger.2013.09.006 dementia and Alzheimer’s disease. N Engl
53. Sundelof J, Kilander L, Helmersson J Med 346(7):476–483
J et al (2009) Systemic tocopherols and 65. Morris MS (2003) Homocysteine and
F2-isoprostanes and the risk of Alzheimer's Alzheimer’s disease. Lancet Neurol
disease and dementia: a prospective pop- 2(7):425–428
ulation-based study. J Alzheimers Dis 66. Selhub J, Troen A, Rosenberg IH (2010)
18(1):71–78. https://doi.org/10.3233/ B vitamins and the aging brain. Nutr Rev
JAD-2009-1125 68(Suppl 2):S112–S118. https://doi.
54. Feart C, Letenneur L, Helmer C et al (2015) org/10.1111/j.1753-4887.2010.00346.x
Plasma carotenoids are inversely associated 67. Morris MS (2012) The role of B vitamins
with dementia risk in an elderly French cohort. in preventing and treating cognitive impair-
J Gerontol A Biol Sci Med Sci. https://doi. ment and decline. Adv Nutr 3(6):801–812.
org/10.1093/gerona/glv135 https://doi.org/10.3945/an.112.002535
55. Hu P, Bretsky P, Crimmins EM et al (2006) 68. Agnew-Blais JC, Wassertheil-Smoller S, Kang
Association between serum beta-carotene lev- JH et al (2015) Folate, vitamin B-6, and vita-
els and decline of cognitive function in high- min B-12 intake and mild cognitive impair-
functioning older persons with or without ment and probable dementia in the Women’s
apolipoprotein E 4 alleles: MacArthur studies Health Initiative Memory Study. J Acad
of successful aging. J Gerontol A Biol Sci Med Nutr Diet 115(2):231–241. https://doi.
Sci 61(6):616–620 org/10.1016/j.jand.2014.07.006
56. Kang JH, Grodstein F (2008) Plasma 69. Corrada MM, Kawas CH, Hallfrisch J et al
carotenoids and tocopherols and cognitive (2005) Reduced risk of Alzheimer’s dis-
function: a prospective study. Neurobiol ease with high folate intake: the Baltimore
Longitudinal Study of Aging. Alzheimers of incident dementia and subclinical indices

Dement 1(1):11–18. https://doi. of brain aging: the Framingham Heart Study.
org/10.1016/j.jalz.2005.06.001 J Alzheimers Dis 51(2):451–461. https://
70. Luchsinger JA, Tang MX, Miller J et al doi.org/10.3233/JAD-150991
(2007) Relation of higher folate intake to 82. Olsson E, Byberg L, Karlstrom B et al (2017)
lower risk of Alzheimer disease in the elderly. Vitamin D is not associated with incident
Arch Neurol 64(1):86–92 dementia or cognitive impairment: an 18-y
71. Lefevre-Arbogast S, Feart C, Dartigues JF follow-up study in community-living old men.
et al (2016) Dietary B vitamins and a 10-year Am J Clin Nutr. https://doi.org/10.3945/
risk of dementia in older persons. Forum ajcn.116.141531
Nutr 8(12). https://doi.org/10.3390/ 83. Nakagawa K, Kiko T, Hatade K et al (2009)
nu8120761 Antioxidant effect of lutein towards phospho-
72. Nelson C, Wengreen HJ, Munger RG et al lipid hydroperoxidation in human erythro-
(2009) Dietary folate, vitamin B-12, vita- cytes. Br J Nutr 102(9):1280–1284. https://
min B-6 and incident Alzheimer’s disease: doi.org/10.1017/S0007114509990316
the cache county memory, health and aging 84. Chen SJ, Huang LY, Hu CH (2015)
study. J Nutr Health Aging 13(10):899–905 Antioxidative reaction of carotenes against
73. Morris MC, Evans DA, Schneider JA et al peroxidation of fatty acids initiated by nitro-
(2006) Dietary folate and vitamins B-12 and gen dioxide: a theoretical study. J Phys
B-6 not associated with incident Alzheimer’s Chem B 119(30):9640–9650. https://doi.
disease. J Alzheimers Dis 9(4):435–443 org/10.1021/acs.jpcb.5b04142
74. Morris MC, Evans DA, Bienias JL et al 85. Feart C, Samieri C, Barberger-Gateau P
(2005) Dietary folate and vitamin B12 (2015) Mediterranean diet and cogni-
intake and cognitive decline among com- tive health: an update of available knowl-
munity-dwelling older persons. Arch Neurol edge. Curr Opin Clin Nutr Metab Care
62(4):641–645. https://doi.org/10.1001/ 18(1):51–62. https://doi.org/10.1097/
archneur.62.4.641 MCO.0000000000000131
75. Hinterberger M, Fischer P (2013) Folate 86. Scarmeas N, Stern Y, Tang MX et al (2006)
and Alzheimer: when time matters. J Neural Mediterranean diet and risk for Alzheimer’s
Transm 120(1):211–224. https://doi. disease. Ann Neurol 59(6):912–921
org/10.1007/s00702-012-0822-y 87. Feart C, Samieri C, Rondeau V et al (2009)
76. Landel V, Annweiler C, Millet P et al (2016) Adherence to a Mediterranean diet, cogni-
Vitamin D, cognition, and Alzheimer’s dis- tive decline, and risk of dementia. JAMA
ease: the therapeutic benefit is in the D-tails. 302(6):638–648
J Alzheimers Dis. https://doi.org/10.3233/ 88. Singh B, Parsaik AK, Mielke MM et al (2014)
jad-150943 Association of mediterranean diet with mild
77. Hilger J, Friedel A, Herr R et al (2014) cognitive impairment and Alzheimer’s dis-
A systematic review of vitamin D sta- ease: a systematic review and meta-analysis.
tus in populations worldwide. Br J Nutr J Alzheimers Dis 39(2):271–282. https://
111(1):23–45. https://doi.org/10.1017/ doi.org/10.3233/JAD-130830
S0007114513001840 89. Samieri C, Grodstein F, Rosner BA et al
78. Annweiler C, Montero-Odasso M, Llewellyn (2013) Mediterranean diet and cogni-
DJ et al (2013) Meta-analysis of memory and tive function in older age. Epidemiology
executive dysfunctions in relation to vitamin 24(4):490–499. https://doi.org/10.1097/
D. J Alzheimers Dis 37(1):147–171. https:// EDE.0b013e318294a065
doi.org/10.3233/JAD-130452 90. Samieri C, Okereke OI, ED E et al (2013)
79. Sommer I, Griebler U, Kien C et al (2017) Long-term adherence to the Mediterranean
Vitamin D deficiency as a risk factor for diet is associated with overall cognitive status,
dementia: a systematic review and meta- but not cognitive decline, in women. J Nutr
analysis. BMC Geriatr 17(1):16. https://doi. 143(4):493–499. https://doi.org/10.3945/
org/10.1186/s12877-016-0405-0 jn.112.169896
80. Feart C, Helmer C, Merle B et al (2017) 91. Gardener SL, Rainey-Smith SR, Barnes MB
Associations of lower vitamin D concentra- et al (2014) Dietary patterns and cogni-
tions with cognitive decline and long-term risk tive decline in an Australian study of ageing.
of dementia and Alzheimer’s disease in older Mol Psychiatry. https://doi.org/10.1038/
adults. Alzheimers Dement 13(11):1207–1216. mp.2014.79
https://doi.org/10.1016/j.jalz.2017.03.003. 92. Smyth A, Dehghan M, O'Donnell M et al
81. Karakis I, Pase MP, Beiser A et al (2016) (2015) Healthy eating and reduced risk of
Association of serum vitamin D with the risk cognitive decline: a cohort from 40 coun-
40 Cécilia Samieri
tries. Neurology. https://doi.org/10.1212/ for Alzheimer’s disease: the IWG-2 criteria.

WNL.0000000000001638 Lancet Neurol 13(6):614–629. https://doi.
93. Tangney CC, Li H, Wang Y et al (2014) org/10.1016/S1474-4422(14)70090-0
Relation of DASH- and Mediterranean- 105. Sperling RA, Aisen PS, Beckett LA et al

like dietary patterns to cognitive decline in (2011) Toward defining the preclinical
older persons. Neurology 83(16):1410– stages of Alzheimer’s disease: recommenda-
1416. https://doi.org/10.1212/ tions from the National Institute on Aging-
WNL.0000000000000884 Alzheimer’s Association workgroups on
94. Berendsen AA, Kang JH, van de Rest O et al diagnostic guidelines for Alzheimer’s disease.
(2017) The dietary approaches to stop hyper- Alzheimers Dement 7(3):280–292. https://
tension diet, cognitive function, and cogni- doi.org/10.1016/j.jalz.2011.03.003
tive decline in American older women. J Am 106. Dubois B, Hampel H, Feldman HH et al

Med Dir Assoc. https://doi.org/10.1016/j. (2016) Preclinical Alzheimer’s disease: defini-
jamda.2016.11.026 tion, natural history, and diagnostic criteria.
95. Morris MC, Tangney CC, Wang Y et al Alzheimers Dement 12(3):292–323. https://
(2015) MIND diet slows cognitive decline doi.org/10.1016/j.jalz.2016.02.002
with aging. Alzheimers Dement. https://doi. 107. Virtanen JK, Siscovick DS, Longstreth WT
org/10.1016/j.jalz.2015.04.011 Jr et al (2008) Fish consumption and risk
96. Morris MC, Tangney CC, Wang Y et al of subclinical brain abnormalities on MRI in
(2015) MIND diet associated with reduced older adults. Neurology 71(6):439–446
incidence of Alzheimer’s disease. Alzheimers 108. Virtanen JK, Siscovick DS, Lemaitre RN et al
Dement. https://doi.org/10.1016/j. (2013) Circulating omega-3 polyunsaturated
jalz.2014.11.009 fatty acids and subclinical brain abnormalities
97. Gu Y, Scarmeas N (2011) Dietary patterns in on MRI in older adults: the Cardiovascular
Alzheimer’s disease and cognitive aging. Curr Health Study. J Am Heart Assoc
Alzheimer Res 8(5):510–519 2(5):e000305. https://doi.org/10.1161/
98. Alles B, Samieri C, Feart C et al (2012) JAHA.113.000305
Dietary patterns: a novel approach to exam- 109. Tan ZS, Harris WS, Beiser AS et al (2012)
ine the link between nutrition and cognitive Red blood cell omega-3 fatty acid lev-
function in older individuals. Nutr Res Rev els and markers of accelerated brain aging.
25(2):207–222. https://doi.org/10.1017/ Neurology 78(9):658–664. doi:78/9/658
S0954422412000133 [pii] 10.1212/WNL.0b013e318249f6a9
99. Akbaraly TN, Singh-Manoux A, Marmot MG 110. Pottala JV, Yaffe K, Robinson JG et al

et al (2009) Education attenuates the associa- (2014) Higher RBC EPA + DHA cor-
tion between dietary patterns and cognition. responds with larger total brain and hip-
Dement Geriatr Cogn Disord 27(2):147–154 pocampal volumes: WHIMS-MRI Study.
100. Granic A, Davies K, Adamson A et al (2016) Neurology. https://doi.org/10.1212/
Dietary patterns high in red meat, potato, WNL.0000000000000080
gravy, and butter are associated with poor 111. Samieri C, Maillard P, Crivello F et al (2012)
cognitive functioning but not with rate of Plasma long-chain omega-3 fatty acids
cognitive decline in very old adults. J Nutr. and atrophy of the medial temporal lobe.
https://doi.org/10.3945/jn.115.216952 Neurology 79(7):642–650. https://doi.
101. Shakersain B, Santoni G, Larsson SC et al
org/10.1212/WNL.0b013e318264e394
(2016) Prudent diet may attenuate the adverse 112. Walhovd KB, Storsve AB, Westlye LT et al
effects of Western diet on cognitive decline. (2014) Blood markers of fatty acids and vitamin
Alzheimers Dement 12(2):100–109. https:// D, cardiovascular measures, body mass index,
doi.org/10.1016/j.jalz.2015.08.002 and physical activity relate to longitudinal
102. Gu Y, Nieves JW, Stern Y et al (2010)
cortical thinning in normal aging. Neurobiol
Food combination and Alzheimer dis- Aging 35(5):1055–1064. https://doi.
ease risk: a protective diet. Arch Neurol org/10.1016/j.neurobiolaging.2013.11.011
67(6):699–706. https://doi.org/10.1001/ 113. Gu Y, Brickman AM, Stern Y et al (2015)
archneurol.2010.84 Mediterranean diet and brain struc-
103. Amadieu C, Lefevre-Arbogast S, Delcourt
ture in a multiethnic elderly cohort.
C et al (2017) Nutrient biomarker patterns Neurology. https://doi.org/10.1212/
and long-term risk of dementia in older wnl.0000000000002121
adults. Alzheimers Dement. https://doi. 114. Titova OE, Ax E, Brooks SJ et al (2013)

org/10.1016/j.jalz.2017.01.025 Mediterranean diet habits in older individu-
104. Dubois B, Feldman HH, Jacova C et al
als: associations with cognitive functioning
(2014) Advancing research diagnostic criteria and brain volumes. Exp Gerontol 48(12):
1443–1448. https://doi.org/10.1016/j. Neurology 78(4):241–249. https://doi.

exger.2013.10.002 org/10.1212/WNL.0b013e3182436598
115. Pelletier A, Barul C, Feart C et al (2015)
126. Scarmeas N, Luchsinger JA, Stern Y et al

Mediterranean diet and preserved brain struc- (2011) Mediterranean diet and magnetic res-
tural connectivity in older subjects. Alzheimers onance imaging-assessed cerebrovascular dis-
Dement 11(9):1023–1031. https://doi. ease. Ann Neurol 69(2):257–268. https://
org/10.1016/j.jalz.2015.06.1888 doi.org/10.1002/ana.22317
116. Luciano M, Corley J, Cox SR et al (2017) 127. Gottesman RF, Schneider AL, Zhou Y et al
Mediterranean-type diet and brain struc- (2017) Association between midlife vascular
tural change from 73 to 76 years in a risk factors and estimated brain amyloid depo-
Scottish cohort. Neurology. https://doi. sition. JAMA 317(14):1443–1450. https://
org/10.1212/WNL.0000000000003559 doi.org/10.1001/jama.2017.3090
117. Mosconi L, Murray J, Tsui WH et al (2014) 128. Mazereeuw G, Lanctot KL, Chau SA et al
Mediterranean diet and magnetic resonance (2012) Effects of omega-3 fatty acids on
imaging-assessed brain atrophy in cognitively cognitive performance: a meta-analysis.
normal individuals at risk for Alzheimer’s dis- Neurobiol Aging 33(7):1482 e1417–1482
ease. J Prev Alzheimers Dis 1(1):23–32 e1429. https://doi.org/10.1016/j.
118. Staubo SC, Aakre JA, Vemuri P et al (2017) neurobiolaging.2011.12.014
Mediterranean diet, micronutrients and mac- 129. Jiao J, Li Q, Chu J et al (2014) Effect of n-3
ronutrients, and MRI measures of cortical PUFA supplementation on cognitive func-
thickness. Alzheimers Dement 13(2):168– tion throughout the life span from infancy
177. https://doi.org/10.1016/j. to old age: a systematic review and meta-
jalz.2016.06.2359 analysis of randomized controlled trials. Am
119. Gu Y, Vorburger RS, Gazes Y et al (2016) White J Clin Nutr 100(6):1422–1436. https://doi.
matter integrity as a mediator in the relation- org/10.3945/ajcn.114.095315
ship between dietary nutrients and cognition 130. Farina N, Isaac MG, Clark AR et al (2012)
in the elderly. Ann Neurol 79(6):1014–1025. Vitamin E for Alzheimer’s dementia and mild
https://doi.org/10.1002/ana.24674 cognitive impairment. Cochrane Database
120. Witte AV, Kerti L, Hermannstadter HM
Syst Rev 11:CD002854. https://doi.
et al (2014) Long-chain omega-3 fatty acids org/10.1002/14651858.CD002854.pub3
improve brain function and structure in older 131. Clarke R, Bennett D, Parish S et al (2014)
adults. Cereb Cortex 24(11):3059–3068. Effects of homocysteine lowering with B
https://doi.org/10.1093/cercor/bht163 vitamins on cognitive aging: meta-analysis
121. Mosconi L, Murray J, Davies M et al
of 11 trials with cognitive data on 22,000
(2014) Nutrient intake and brain biomark- individuals. Am J Clin Nutr. https://doi.
ers of Alzheimer's disease in at-risk cogni- org/10.3945/ajcn.113.076349
tively normal individuals: a cross-sectional 132. Li MM, Yu JT, Wang HF et al (2014) Efficacy
neuroimaging pilot study. BMJ Open of vitamins B supplementation on mild cog-
4(6):e004850. https://doi.org/10.1136/ nitive impairment and Alzheimer’s disease:
bmjopen-2014-004850 a systematic review and meta-analysis. Curr
122. Berti V, Murray J, Davies M et al (2015)
Alzheimer Res 11(9):844–852
Nutrient patterns and brain biomark- 133. Feart C, Samieri C, Alles B et al (2013)

ers of Alzheimer’s disease in cognitively Potential benefits of adherence to the
normal individuals. J Nutr Health Aging Mediterranean diet on cognitive health.
19(4):413–423. https://doi.org/10.1007/ Proc Nutr Soc 72(1):140–152. https://doi.
s12603-014-0534-0 org/10.1017/S0029665112002959
123. Pantoni L (2010) Cerebral small vessel dis- 134. Grodstein F, O'Brien J, Kang JH et al (2013)
ease: from pathogenesis and clinical character- Long-term multivitamin supplementation
istics to therapeutic challenges. Lancet Neurol and cognitive function in men: a randomized
9(7):689–701. https://doi.org/10.1016/ trial. Ann Intern Med 159(12):806–814
S1474-4422(10)70104-6 135. Scheltens P, Twisk JW, Blesa R et al

124. Debette S, Markus HS (2010) The clinical (2012) Efficacy of Souvenaid in mild
importance of white matter hyperintensi- Alzheimer’s disease: results from a ran-
ties on brain magnetic resonance imaging: domized, controlled trial. J Alzheimers Dis
systematic review and meta-analysis. BMJ 31(1):225–236. https://doi.org/10.3233/
341:c3666 JAD-2012-121189
125. Bowman GL, Silbert LC, Howieson D et al 136. Shah RC, Kamphuis PJ, Leurgans S et al

(2012) Nutrient biomarker patterns, cognitive (2013) The S-Connect study: results from
function, and MRI measures of brain aging. a randomized, controlled trial of Souvenaid
42 Cécilia Samieri
in mild-to-moderate Alzheimer’s disease. 138.

Valls-Pedret C, Sala-Vila A, Serra-Mir
Alzheimers Res Ther 5(6):59. https://doi. M et al (2015) Mediterranean diet and
org/10.1186/alzrt224 age-related cognitive decline: a randomized
137. Soininen H, et al. Lancet Neurol. 2017 Dec; clinical trial. JAMA Intern Med. https://
16(12):965–975. doi: 10.1016/S1474-4422 doi.org/10.1001/jamainternmed.2015.
(17)30332-0. 1668
Part II
Diagnostic Concepts
Chapter 3
The Dimensional Structure of Subjective Cognitive Decline

Miguel A. Fernández-Blázquez, Marina Ávila-Villanueva,
and Miguel Medina
Abstract
The self-experienced persistent decline in one’s cognitive abilities in comparison with a previously normal
status is referred to as subjective cognitive decline (SCD). During the last decades, evidence has emerged
about the close relationship between SCD and incident cognitive impairment to such an extent that SCD
is currently considered as a very early marker of future dementia. Here, we first discuss the strengths and
the weaknesses of this concept, and then we describe a procedure to measure SCD accurately. Our goal is
to provide the reader with specific guidelines on how to assess and classify individuals in terms of SCD in
order to identify individuals at high risk of developing cognitive impairment.
Key words Aging, Alzheimer’s disease, Cognitive symptoms, Dementia, Mild cognitive impairment,
Subjective cognitive decline
1 Introduction
Alzheimer’s disease (AD) is a multifactorial neurodegenerative

dementia that begins affecting the brain many years before cogni-
tive impairment is even noticeable. The National Institute on
Aging-Alzheimer Association (NIA-AA) has recently established
three different stages of AD as progression occurs over time. First,
there is a preclinical phase which is defined by the incipient pres-
ence of amyloid plaques, but objective cognitive function remains
normal. At the end of this preclinical phase, the individual might
subjectively experience some kind of cognitive decline [1]. A sec-
ond stage called prodromal AD or mild cognitive impairment
(MCI) that implies subtle objective cognitive deficits that are not
severe enough to significantly affect everyday activities [2]. Finally,
there is a third phase in which the extent of the cognitive deteriora-
tion leads to a functional impairment that defines a dementia syn-
drome [3]. Currently, there are no interventions capable of
modifying the progression of AD. Clinical trials in both prodromal
45
46 Miguel A. Fernández-Blázquez et al.
and dementia stages have not produced significantly positive

results. This scenario has led to the growing consensus that thera-
peutic interventions are expected to be more effective when they
are putting into practice as early as possible [4]. Thus, the focus of
research in the field is increasingly placing greater emphasis on the
search of early markers of preclinical AD since disease-modifying
therapeutic approaches will most likely be developed for future use
in at-risk populations [5].
The construct subjective cognitive decline (SCD) refers to a
self-experienced persistent decline in cognitive abilities in compari-
son with a previously normal status and independently of the
objective performance on neuropsychological tests. SCD has been
a focus of debate within the research literature during the past two
decades because of its potential clinical relevance in predicting the
onset of future dementia in older adults. In fact, it has been
described that subjects might experience some type of cognitive
decline up to 15 years before they develop MCI and AD [6].
Furthermore, in the absence of objective cognitive impairment,
evidence has been reported about the relationship between SCD
and some AD biomarkers such as brain amyloid deposition and
cerebral hippocampal hypometabolism [7].
Despite cognitive complaints have been traditionally treated as
equivalent to memory failures, SCD is in fact heterogeneous and
might affect all cognitive domains from a neuropsychological point
of view. For instance, there are symptoms that may be specifically
related to memory (“forgetting recent events”), while others are
associated with attention (“being unable to follow the thread of a
story”), language (“the tip of the tongue phenomenon”), visual
perception (“failing to recognize, by sight, close friends or rela-
tives”), or executive functions (“perseverations in old daily rou-
tines”). Although episodic memory impairment is usually the first
cognitive manifestation of MCI due to AD, perception of memory
failures might not be the primary complaint of cases with preclini-
cal AD. For instance, attentional symptoms like “following the
thread of a conversation” have been proved to be more associated
with risk of dementia than forgetfulness itself such as “problems to
recall recent information” [8]. In any event, research on SCD has
almost exclusively focused on the overall construct rather than in
examining the role of specific cognitive domains upon the progres-
sion of AD.
Since the expression of cognitive complaints is affected by vari-
ous factors (e.g., aging, personality, mood, drug side effects, neu-
rological disorders, etc.), SCD is not necessarily present in all AD
patients. Nevertheless, both cross-sectional [9–11] and longitudi-
nal studies [12–14] have provided strong evidence of SCD occur-
ring at preclinical AD. Perhaps, all these evidences along with the
ease and brevity of measuring complaints are the reasons why
The Dimensional Structure of Subjective Cognitive Decline 47
uring the last decades there has been an important increase in the
d
study of SCD as a very early sign of cognitive deterioration.
There are overwhelming epidemiological data in favor of the
relationship between SCD and incident cognitive impairment.
Thus, a recent meta-analysis focused on the longitudinal value of
SCD for detecting later MCI and dementia has shown that, inde-
pendently of the objective memory performance, 6.6% of older
adults with SCD develop MCI per year [15]. In addition, the rate
of progression to dementia among those individuals who report
complaints about their own cognitive performance is also twofold
during a 5-year follow-up period.
Despite increasing evidence indicating that SCD may represent
a very early manifestation of AD [16], little is known about the
clinical role of specific complaints on the transition from normal
aging to cognitive impairment. Three types of cognitive concerns
have demonstrated their usefulness to discriminate between cogni-
tively healthy older adults and MCI [17]; particularly higher scores
in specific complaints on recalling immediate events, executive
functioning, and prospective memory are related to prodromal
stages of dementia. The fact that other types of complaints such as
forgetfulness of objects or spatial orientation did not show differ-
ences between controls and MCI could be explained because the
first of them refers to a high prevalent oversight in the elderly pop-
ulation (“forgetting where you have put something” and “forget-
ting where things are normally kept or looking for them in the
wrong place”) and the second one is an idiosyncratic sign of very
mild dementia (“getting lost or turning in the wrong direction on
a journey, on a walk, or in a building that one really knows because
he or she has been there before”). These findings emphasize that
not all cognitive complaints have the same clinical significance for
predicting cognitive impairment.
However, despite its emerging role as a marker of preclinical
AD, the concept of SCD is not free from some limitations which
are necessary to address. For instance, many terms such as subjec-
tive memory complaints, subjective cognitive decline, or subjective
memory impairment have been used interchangeably to refer to the
same concept. This lack of consensus on a single definition of SCD
affects to the comparison of findings from different investigations
and epidemiological studies. Moreover, there is no accepted
approach among researchers about the assessment of SCD, includ-
ing the mode of administration (structured interview versus ques-
tionnaires), the cognitive domains to be examined (memory versus
non-memory domains), the number of items to be used (one or
two questions versus scales with a large number of items), and the
optimal way to respond the items (opened questions versus multi-
ple choice). Finally, it becomes difficult to determine which com-
plaints underlie AD because there is a close relationship between
SCD and subjective variables such as depression [18], anxiety [19],

perceived health [20], personality [21], and quality of life [22]. In
sum, the heterogeneity in definitions and the different approaches
for measuring SCD emphasize the necessity of searching for shared
terminology and common standards of evaluation.
To assess the potential usefulness of SCD for epidemiological
studies and clinical trials, an international working group, the
Subjective Cognitive Decline Initiative (SCD-I), agreed to a com-
mon terminology and research procedures to identify individuals
with SCD at risk of preclinical AD [23]. The SCD-I aimed at
knowing whether the self-experience of decline in cognition could
actually represent the first manifestation of AD. In order to dem-
onstrate that, some common specific features are required to estab-
lish a complete profile of SCD and to characterize two distinct
groups of individuals in accordance with such profile.
The SCD-I recommended to collect information regarding
features such as settings in which cognitive complaints are
expressed, association of SCD with medical help seeking, duration
and age at onset of SCD, subjective decline in memory and non-
memory domains, and association of SCD with experience of
impairment. In addition, the SCD-I proposed a set of particular
features which could be helpful to identify individuals at risk of
clinical progression. Those features include a more acute subjective
memory decline than any other cognitive domain, onset of com-
plaints within the last 5 years, age at onset over 60 years, worries
about SCD, and feeling of worse performance than other people
from the same age group. Fulfilling this set of features would give
rise to a more severe form of SCD referred to as subjective cogni-
tive decline plus (SCD plus). Thus, this new category could allow
us to explain the transition from a nonsymptomatic stage to the
first manifestation of AD.
2 Materials
In order to measure SCD as reliably as possible, some recommen-

dations can be extracted from a recent review of the SCD-I [24].
First, it is very important to select the most appropriate measures
according to the characteristics of the target population. Complaints
may have different implications depending on the research context
where they are collected. For instance, clinical samples are specifi-
cally recruited in medical settings, and therefore we can assume
that concerns on SCD may be higher in these samples compared to
population-based studies. Moreover, it would be desirable to rely
on measures previously published with adequate psychometric
properties for the reference population. Second, the SCD’s mea-
sures must have adequate content coverage with regard to the
target population. This recommendation means that all items

should be understandable; they should only inquire for a unique

domain; and they should be as specific as possible and related to
difficulties often found in daily life. Third, measures should explore
different cognitive and non-cognitive domains because the earliest
symptoms of AD may include a great variety of complaints beyond
memory. Fourth, it should be taken into account the response
options for the measures depending on the aim of our study. When
the purpose is to distinguish between groups, dichotomous items
may be enough. However, measures with ordinal response options
should be preferred when changing of SCD over time is pursued.
Finally, it is very important to specify the reference period of time
in which we want to examine the SCD. As a rule of thumb, it
should be preferred to inquire on self-experience over short peri-
ods of time (no longer than 1 year) to avoid problems with retro-
spective recall or estimation of complaints. Nevertheless, this does
not mean that we cannot ask for longer periods if we intend to
analyze the longitudinal change of SCD through the lifetime.
There is a decisive debate on how to determine whether or not
an individual has SCD. Indeed, the measurement of SCD may lead
to different results depending on the approach followed. In prac-
tice, SCD might be assessed by means of either individual ques-
tions or structured questionnaires. Although these two methods
tend to correlate, there are some differences among them. In many
studies the assessment of SCD only involves a single question simi-
lar to “Do you have any memory problems?” which is typically
coded as yes/no. Instead, other investigations include a set of SCD
questions in the context of a more detailed clinical interview.
Finally, it may be possible to use specific questionnaires to assess
SCD. Basically, the questionnaires consist of a checklist of com-
mon cognitive failures that must be rated according to the fre-
quency in which they are experienced by individuals. These
questionnaires provide information on the presence or not of any
specific complaint, the frequency and severity of them, and even
the cognitive strategies of an individual to deal with complaints. In
addition, since questionnaires use a relative large number of items,
psychometrical properties such as internal consistency and validity
tend to be higher than those for specific questions of SCD [25].
For all these reasons, when available, structured questionnaires are
usually considered the best approach of gaining insight into older
adults’ SCD [20]. Individual questions seem to be particularly use-
ful in population-based and epidemiological studies, when admin-
istering a questionnaire is very difficult because of time constraints.
However, a recent study specifically comparing open-ended ques-
tions and comprehensive questionnaire has highlighted that spon-
taneously reported complaints may be preferred over questionnaires
because questions better reflect the distress associated with SCD
[26]. In Table 1 a list of structured questionnaires on SCD that are
frequently used in research settings is showed.
Table 1
Structured SCD questionnaires frequently used in research
Questionnaires Number of items Response options

Cognitive Failures Questionnaire 25 Ordinal scale
(CFQ) [27] (5 points)
Everyday Cognition (ECog) [28] 39 Ordinal scale
(4 points)
Everyday Memory Questionnaire 35 Ordinal scale (7, 5
(EMQ) [29] or 3 points)
Informant Questionnaire on 26 Ordinal scale
Cognitive Decline (IQCODE) (5 points)
[30]
Memory complaint questionnaire 6 Ordinal scale
(MAC-Q) [31] (5 points)
Memory Functioning 28 Ordinal scale
Questionnaire (MFQ) [32] (7 points)
Subjective cognitive decline 27 Dichotomic scale
questionnaire (SCD-Q) [33] (yes/no)
Subjective memory complaints 14 Dichotomic scale
questionnaire (SMCQ) [34] (yes/no)
Subjective memory questionnaire 18 Ordinal scale
(SMQ) [35] (9 points)
3 Methods
According to the evidence and the recommendations discussed

above, epidemiological studies focused on examining the role of
SCD as preclinical marker of AD should combine different
approaches to measure SCD accurately. Thus, there would be four
specific points which should be bear in mind when a longitudinal
study is methodologically designed:
1. Use both open-ended questions and structured questionnaires
to measure different aspects of SCD. Specific questions should
inquire about clinical details of the self-experienced cognitive
decline (e.g., age at onset, seeking medical help, memory
performance compared to other people, etc.) as well as concerns
and frequency of particular cognitive complaints (e.g., forget-
ting recent events, being unable to follow the thread of a story,
difficulties to retrieve the adequate word, etc.). Finally, the
multiple- choice approach should vary from dichotomic to
ordinal Likert-type scales to grasp the dimensionality of SCD in
the best way possible.
2. Collect information in different ways to ensure a greater inter-

nal consistency. As far as possible, an interesting approach would
be to gather self-perceived data by means of a face-to-face inter-
view with a health-care professional along with self-administered
questionnaires. Since it is relatively frequent that an individual
reports qualitatively different features of SCD after a short
period of time, this procedure would allow to evaluate the sta-
bility of those complaints.
3. Inquire about SCD with regard to different time frames (e.g.,
last months, last years, youth, etc.). In this way the self-
experienced cognitive change over time might be examined.
4. Include items to cover all cognitive and non-cognitive domains,
not only episodic memory. Some cases of atypical AD or even
non-AD dementias may begin with symptoms distinct from
memory loss (e.g., problems to inhibit behavior may be indica-
tive of frontotemporal dementias rather than AD). For this rea-
son, it is important to request subjective information about the
whole spectrum of cognition and other neuropsychiatric
variables.
3.1 Assessment Since 2011 CIEN Foundation-Queen Sofía Foundation is carrying
of SCD During out a multidisciplinary study to identify early markers of AD,
Neurological and namely, the Vallecas Project [36]. In the context of this epidemio-
Neuropsychological logical research, we record detailed demographic and clinical infor-
Interviews mation from every participant, including self-perception of
cognitive decline. Precisely, SCD has grown a great relevance in
our research as a key factor which could be able to detect individuals
at high risk of future cognitive impairment.
Responses to every question of SCD are directly provided by
the participants since family members are not available in all cases.
Most importantly, to ensure the reliability and internal consistency,
SCD is assessed twice and independently within the same visit.
First, during the neurological examination, participants are asked
the following nine questions regarding specific cognitive domains:
1. Attention: “Are you easily distracted?”
2. Spatial orientation: “Do you get lost in familiar surroundings or
have trouble finding your way when driving?”
3. Episodic memory: “Do you often forget recent information or
events?”
4. Autobiographical memory: “Do you often forget autobio-

graphical information?”
5. Visual recognition: “Do you have trouble recognizing objects
or faces?”
6. Speech: “Do you have word-finding difficulties for people’s
names or common words?”
7. Language comprehension: “Do you understand simple verbal

and written instructions?”
8. Executive functions: “Do you have difficulty driving, managing
finances, or planning daily activities?”
9. Praxis: “Do you have difficulty sequencing movements (e.g.,
taking the necessary steps to prepare a bath)?”
It is important to note that all these previous questions are
merely tentative; thus, they are opened and spontaneously reported
in such a way that alternative questions within the same cognitive
domain could arise during the interview. Ultimately, subjective
experience of complaints for every cognitive domain is coded in a
dichotomic way (yes/no) based on the global impression of the
neurologist. Additionally, there are other questions regarding psy-
chiatric and behavioral symptoms which are a complement to the
nine questions on cognitive complaints.
Second, during the neuropsychological assessment, individuals
also complete an ordinal scale of cognitive complaints composed of
four items with four points each (ranged 0–3). This scale addresses
the following questions to be responded:
1. “How do you perceive your memory in comparison with that of
others of your age?”
(“3, bad”; “2, somewhat worse”; “1, somewhat better”; “0,
excellent”)
2. “How do you perceive your memory today compared with your
young adulthood?”
(“0, better”; “1, equal”; “2, somewhat worse”; “3, much
worse”)
3. “Do you perceive your memory today is worse than compared
with ten years ago?”
(“0, no”; “1, a little worse”; “2, somewhat worse”; “3, much
worse”)
4. “Do you perceive your memory today is worse than compared
with one year ago?”
(“0, no”; “1, a little worse”; “2, somewhat worse”; “3, much
worse”)
The sum of these items resulted in a total score of cognitive
concerns ranging from 0 (no complaints at all) to 12 (maxi-
mum complaints). Furthermore, five more open-ended ques-
tions are also collected during the neuropsychological
interview:
5. Age at onset of cognitive complaints: “How old were you when
your cognitive performance began to decline?”
6. Years of SCD’s progression: “How long do you believe you are
experiencing cognitive complaints?”
7. Worries associated with self-perceived complaints: “Are you

worried about your cognitive decline?”
8. Type of onset of cognitive complaints: “How did you perceive
the beginning of the cognitive decline?” (e.g., suddenly, pro-
gressive, etc.)
9. Self-experienced functional impairment associated with SCD:
“Do you believe your cognitive failures are impeding your daily
life activities?”
3.2 Assessment In addition to the whole information collected in both clinical
of SCD by Means interviews, individuals must accomplish a SCD scale, namely, the
of Self-Administered Everyday Memory Questionnaire (EMQ) [29]. This questionnaire
Questionnaire is selected because it has been previously validated in our country
and showed adequate psychometrical properties for older adults
[37]. In our study, EMQ is self-administered by following the
instructions provided in the validation study and always in the
presence of a member of the research team; individuals are required
to ask any doubt may arise.
The EMQ comprised 28 items about cognitive failures that
occur in everyday life. The items must be responded according to
the frequency in which they are experienced by a subject. All items
are scored pursuant to a three-point Likert-type scale, with 0 indi-
cating “never, rarely”, 1 “occasionally, sometimes”, and 2 “fre-
quently, almost always.” The total score ranges from 0 to 56, with
lower scores indicating fewer SCD. Furthermore, responses may
be scored in accordance to three different factors which have been
proved to distinguish between cognitively healthy individuals and
MCI [17]. Due to the content of their items, the three factors are
related to different cognitive domains, namely, episodic memory,
executive functions, and prospective memory.
3.3 Assessment As discussed earlier, non-cognitive symptoms may represent an ini-

of Non-cognitive tial manifestation of cognitive decline due to different etiologies.
Complaints Therefore, open-ended questions should be included in the con-
During Medical text of the medical interview in order to assess behavioral,
Interview functional, and psychiatric self-reported complaints. Specifically,
the aspects that should be covered may be the following:
1. Depression: “Do you feel sad, lonely, and depressed most of the
time?”
2. Anxiety: “Do you feel worried and anxious most of the time?”
3. Apathy: “Do you feel a lack of emotion, motivation, or interest
in hobbies and previously activities enjoyed?”
4. Disinhibition: “Do you feel difficulties to control your own
behavior in some situations?”
5. Irritability: “Do you feel an irritable mood most of the time?”
6. Hallucinations: “Do you feel you see or hear strange things

which maybe are unreal?”
7. Sleep: “Do you feel your sleep routine has changed?”
8. Falls: “Have you had falls recently?”
9. Gait: “Do you feel you have problems to walk?”
3.4 Classification Besides analyzing the implication of specific SCD features as early
of Individuals in SCD signs of AD, information about complaints should be also exam-
Groups ined according to the guidelines proposed by the SCD-I [23].
Following these guidelines, individuals might be grouped in three
different categories pursuant to the extent of SCD reported: (1) no
complaints (NCg); (2) subjective cognitive decline group (SCDg),
when subjects report some kind of cognitive complaint; and (3)
subjective cognitive decline plus (SCD-Pg), when individuals show
complaints in memory plus another cognitive domain and addi-
tionally they fulfill the rest of the criteria for SCD plus.
The classification in any of the three groups of SCD may be
carried out in two steps (a full description of this procedure is
shown in Fig. 1). Initially, based on the overall information gath-
ered both in clinical interviews and in self-administered EMQ,
SCD may be operationally defined as the self-rated presence of
cognitive deterioration using two criteria: (1) at least a positive
response to any yes/no-type question regarding complaints in any
cognitive domain from the neurological interview and (2) scores
above 1 on the SCD scale administered in the neuropsychological
assessment and above 8 on the self-administered EMQ. To be clas-
sified as SCDg, individuals have to mandatorily accomplish both
conditions. Thus, subjects who only fulfill one criterion or neither
of them will be considered as no complaints.
The second step of classification is only applied for those cases
categorized as SCDg. For these individuals some specific features
must be considered such as age at onset of SCD beyond 60 years,
turning up of complaints within the last 5 years, worry associated
with SCD, and feeling of worse performance than others of the
same age group. When all these conditions accompany the self-
experience of decline, then an individual is classified as SCD-Pg.
3.5 Assessment The assessment of demographic, clinical, and cognitive variables

of Demographic, should be inseparable from SCD, whether one’s goal is to deter-
Clinical, and Cognitive mine the properties of cognitive concerns for detecting preclinical
Variables AD and individuals at high risk of later MCI. To this end, all sub-
jects should undergo a detailed survey and assessment protocol to
gather information on demographics (age, gender, level of education,
marital status, living situation, socioeconomic status, occupation,
etc.), lifestyle (physical activity, social support, eating and sleeping
habits, etc.), quality of life (well-being, perceived health, etc.),
medical history (vital signs, physical symptoms, clinical anamnesis,
Fig. 1 Flow diagram for SCD classification. NCg no complaints group, SCDg subjective cognitive decline group,
and SCD-Pg subjective cognitive decline plus group
medication, neurological examination, etc.), family history of

dementia, and neuropsychological assessment.
Cognitive diagnosis should be always agreed between health-
care professionals at clinical consensus meetings. Every individual
must be independently diagnosed according to his/her age, gen-
der, cognitive reserve, functional information, and cognitive scores.
Nevertheless, rather than psychometrically invariable cutoffs, diag-
nosis must be based on clinical impression. NIA-AA’s criteria [2]
can be applied since they are very useful to diagnose core MCI and
mild dementia. Cognitively healthy subjects are given a score of
0 in the global clinical dementia rating (CDR) [38], while MCI
and mild dementia must score 0.5 and 1, respectively.
The comprehensive neuropsychological battery must be
administered by trained neuropsychologists. This battery should
include complete information about all cognitive domains such as
visual perception, attention, memory, language, praxis, and execu-
tive functions. As an example, in Table 2, a total of 12 cognitive
tests which comprise the neuropsychological battery of the
Table 2
Example of a comprehensive neuropsychological battery and cognitive domains covered
Global Visual Executive

cognition perception Attention Memory Language Praxis functions
Mini mental state X
examination (MMSE)
[39]
Clock-drawing test X X X
Rey-Osterreith complex X X X X
Fig [40, 41].
Free and Cued Selective X
Reminding Test
(FCSRT) [42]
Lexical and semantic X X
verbal fluency [43]
Forward and backward X X
digit span [44]
Five-point test [45] X
Rule Card Shifting Test X
[46]
Boston Naming Test X
(15-item version) [47]
Imitation of Bilateral X
Postures and Symbolic
Gesture [48]
Digit symbol coding [44] X
Incomplete letters X
(VOSP) [49]
Vallecas Project is shown. All these tests cover the whole spectrum
of cognition.
In addition, scales for measuring functional activities and
neuropsychiatric variables are very important. For instance, to this
end, the functional activities questionnaire (FAQ) [50] can be
administered to collect data with regard to instrumental activities
of daily living; the Geriatric Depression Scale (GDS) [51] and the
State-Trait Anxiety Inventory (STAI) [52] might be also adminis-
tered as part of the neuropsychological battery to quickly estimate
mood and anxiety symptoms.
4 Conclusions
Self-report of subtle cognitive complaints have been proposed to

appear at the end of the preclinical phase of AD even in the absence
of significant objective impairment detectable on standardized
neuropsychological assessment [1]. Indeed, there are overwhelm-
ing epidemiological data that support the role of SCD, a risk factor
for subsequent development of MCI and dementia in older adults
[15]. That explains why SCD is gaining an increasing prominence
in neurodegenerative research. Nevertheless, this construct must
face up to some limitations such as the absence of an operational
definition as well as the lack of harmonized criteria and assessment
protocols for measuring SCD. The insight and self-experience of
SCD is phenomenologically complex and may differ among indi-
viduals. Moreover, there is a great variability in SCD assessment
procedures which is reflected on the different questions, response
options, nature of items, and mode of administration across stud-
ies. This heterogeneity is likely affecting the outcomes obtained in
each study and also determines the comparison of results. Because
of this scenario, in 2014, an international working group, the
SCD-I, published a position paper agreeing to a common termi-
nology and research procedures to investigate SCD [23]. Since
then, we rely on a homogeneous framework to better study the
implication of cognitive complaints in dementia due to AD.
Currently, SCD could be defined as a self-experienced persistent
and progressive decline in the own cognitive capacity in comparison
with a previously normal performance and unrelated to any acute
event [53]. This means that complaints due to AD cannot appear
suddenly but gradually, and thus decline cannot be transient and be
associated with any particular medical/psychological condition. In
addition, from a methodologically point of view, some consider-
ations may be made with regard to assessment procedures and pro-
tocols of SCD. First, investigations should combine different
approaches to measure complaints as accurate as possible. It is highly
recommended to use both open-ended questions and items belong-
ing to structured questionnaires to measure different aspects of
SCD. Specific questions should measure clinical details of the self-
experienced cognitive decline (e.g., age at onset, seeking medical
help, memory performance compared to other people, etc.) as well
as concerns and frequency of particular cognitive complaints.
Likewise, depending on the feature of SCD which is under investiga-
tion, the response option may adopt the form of dichotomic (yes/
no-type question) or ordinal scales. Second, different modes of
administration may be carried out to ensure a higher internal consis-
tency of SCD. The collection of complaints data would benefit if it
is made by combining face-to-face clinical interview along with self-
administered questionnaires instead of one unique mode. Third, the
introduction of retrospective time frames allows for the study of the
course of SCD over time. It would be desirable to include items that

inquire about SCD in different periods of time (e.g., from adult-
hood to the last year) in order to examine the role of complaints in
AD. The use of questionnaires such as the EMQ is highly recom-
mended to quantify SCD and to monitor the longitudinal progres-
sion of individuals who report those cognitive complaints. Fourth,
since most instruments used to assess SCD focused on episodic
memory, it would be necessary to include items that cover other
cognitive and non-cognitive domains to further evaluate cases of
atypical AD or cases of non-AD neurodegenerative dementia.
Besides, not all cognitive complaints are effective in distinguishing
healthy elderly individuals from those with MCI. Specific complaints
related to episodic memory, executive functions, and prospective
memory seem to discriminate between controls and cognitive
impaired subjects [17].
Finally, because SCD should not be examined in isolation
when one is interested in determining the effect of complaints
upon AD spectrum, some other information must be obtained.
Demographic variables such as age, gender, and education, as well
as medical and lifestyle variables, can be gathered very easily by
means of a survey. These variables have the greatest interest due to
their possible implication in the expression or not of
SCD. Additionally, objective cognitive performance and diagnosis
are critical to establish the current stage of an individual in the
continuum of AD and the relationship between SCD and risk of
developing MCI and AD. Finally, neuropsychiatric variables seem
to mediate between SCD and cognitive decline, and thus they
should be collected as well. Symptoms such as depression, anxiety,
and apathy are usually more associated with SCD than real cogni-
tive performance, what likely indicates the mediator role of com-
plaints between mood and cognitive status [54]. Individuals who
accomplish all conditions of SCD-Pg have been proved to have a
four times higher risk for developing MCI compared to those sub-
jects without complaints (see Fig. 2); and more surprisingly, the
inclusion of gender, cognitive performance, and ApoE genotyping
does not seem to decrease the predictive power of the SCD-Pg
[55]. Thus, particular features associated with SCD (i.e., onset of
complaints within the last 5 years, age at onset over 60 years, wor-
ries associated with complaints, and feeling of worse performance
than other people from the same age group) seem to help to iden-
tify individuals at high risk of fast conversion to MCI. We strongly
encourage researchers to use the recommendations of SCD-I
because of their implications for clinical settings. For instance,
cognitive training programs should be implemented in subjects
meeting all features of SCD plus proposed by SCD-I [22]. Subjects
who report SCD plus might need special attention in terms of
close clinical follow-up of an early cognitive or pharmacological
intervention.
Fig. 2 Distribution of SCD groups and association with conversion rate to MCI. NCg no complaints group, SCDg
subjective cognitive decline group, and SCD-Pg subjective cognitive decline plus group. This graph is based on
the results published in a previous paper entitled “Specific Features of Subjective Cognitive Decline Predict
Faster Conversion to Mild Cognitive Impairment” (Fernández-Blázquez et al. [55]). No metric multidimensional
scaling was applied in order to analyze the relative position of SCD groups in terms of euclidean distance. The
analysis was based on all significant variables found in nonparametric ANOVAs. Sizes of bubbles indicate
the risk of annual conversion to MCI (the larger is the area, the greater is the risk). As is apparent, SCD-Pg is
the group with the highest rate of conversion (18.9%) followed by SCDg (5.6%) and NCg (4.9%). In addition,
the separation in perceptual map among the three groups indicates that there are three latent profiles underlying
psychiatric and cognitive variables
References
1. Sperling R, Aisen PS, Beckett L et al (2011) Alzheimers Dement 7:263–269. https://doi.
Toward defining the preclinical stages of org/10.1016/j.jalz.2011.03.005
Alzheimer’s disease: recommendations 4. Selkoe DJ (2011) Resolving controversies on
from the National Institute on Aging and the path to Alzheimer’s therapeutics. Nat Med
the Alzheimer’s Association workgroup. 17:1060–1065. https://doi.org/10.1038/
Alzheimers Dement 7:280–292. https://doi. nm.2460
org/10.1016/j.jalz.2011.03.003 5. Hampel H, Lista S, Teipel SJ et al (2014)
2. Albert MS, DeKosky ST, Dickson D et al Perspective on future role of biological mark-
(2011) The diagnosis of mild cognitive impair- ers in clinical therapy trials of Alzheimer’s dis-
ment due to Alzheimer’s disease: recommen- ease: a long-range point of view beyond 2020.
dations from the National Institute on Aging Biochem Pharmacol 88:426–449. https://
and Alzheimer’s Association workgroup. doi.org/10.1016/j.bcp.2013.11.009
Alzheimers Dement 7:270–279. https://doi. 6. Reisberg B, Prichep L, Mosconi L et al (2008)
org/10.1016/j.jalz.2011.03.008 The pre-mild cognitive impairment, subjec-
3. McKhann GM, Knopman DS, Chertkow tive cognitive impairment stage of Alzheimer’s
H et al (2011) The diagnosis of dementia disease. Alzheimers Dement 4:S98–S108.
due to Alzheimer’s disease: recommendations https://doi.org/10.1016/j.jalz.2007.11.017
from the National Institute on Aging- 7. Vannini P, Hanseeuw B, Munro CE et al (2017)
Alzheimer’s Association workgroups on diag- Hippocampal hypometabolism in older adults
nostic guidelines for Alzheimer’s disease. with memory complaints and increased amyloid
burden. Neurology 88:1759–1767. https:// cognitively normal states in a study of healthy

doi.org/10.1212/WNL.0000000000003889 elderly controls. Front Aging Neurosci 8:233.
8. Amariglio RE, Townsend MK, Grodstein https://doi.org/10.3389/fnagi.2016.00233
F et al (2011) Specific subjective mem- 18. Crane MK, Bogner HR, Brown GK,
ory complaints in older persons may Gallo JJ (2007) The link between depres-
indicate poor cognitive function. J Am sive symptoms, negative cognitive bias and
Geriatr Soc 59:1612–1617. https://doi. memory complaints in older adults. Aging
org/10.1111/j.1532-5415.2011.03543.x Ment Health 11:708–715. https://doi.
9. Rami L, Fortea J, Bosch B et al (2011) org/10.1080/13607860701368497
Cerebrospinal fluid biomarkers and memory 19. Comijs HC, Deeg DJ, Dik MG et al (2002)
present distinct associations along the con- Memory complaints; the association with
tinuum from healthy subjects to AD patients. psycho-affective and health problems and the
J Alzheimers Dis 23:319–326. https://doi. role of personality characteristics. A 6-year
org/10.3233/JAD-2010-101422 follow-up study. J Affect Disord 72:157–165.
10. Scheef L, Spottke A, Daerr M et al (2012) doi: S0165032701004530 [pii]
Glucose metabolism, gray matter structure, and 20. Montejo P, Montenegro M, Fernández-
memory decline in subjective memory impair- blázquez M et al (2014) Association of per-
ment. Neurology 79:1332–1339. https://doi. ceived health and depression for older adults’
org/10.1212/WNL.0b013e31826c1a8d subjective memory complaints: contrasting a
11. Wang Y, Risacher SL, West JD et al (2013) specific questionnaire with general complaints
Altered default mode network connectiv- questions. Eur J Ageing 11:77–87
ity in older adults with cognitive complaints 21. Pearman A, Storandt M (2004) Predictors of
and amnestic mild cognitive impairment. subjective memory in older adults. J Gerontol
J Alzheimers Dis 35:751–760. https://doi. B Psychol Sci Soc Sci 59:P4–P6
org/10.3233/JAD-130080 22. Montejo P, Montenegro M, Fernandez MA,
12. Dufouil C, Fuhrer R, Alpérovitch A (2005) Maestu F (2011) Subjective memory complaints
Subjective cognitive complaints and cog- in the elderly: prevalence and influence of tem-
nitive decline: consequence or predictor? poral orientation, depression and quality of life in
The epidemiology of vascular aging study. a population-based study in the city of Madrid.
J Am Geriatr Soc 53:616–621. https://doi. Aging Ment Health 15:85–96. https://doi.org/
org/10.1111/j.1532-5415.2005.53209.x 10.1080/13607863.2010.501062
13. Glodzik-Sobanska L, Reisberg B, De Santi 23. Jessen F, Amariglio RE, van Boxtel M et al
S et al (2007) Subjective memory com- (2014) A conceptual framework for research
plaints: presence, severity and future outcome on subjective cognitive decline in preclini-
in normal older subjects. Dement Geriatr cal Alzheimer’s disease. Alzheimers Dement
Cogn Disord 24:177–184. https://doi. 10:844–852. https://doi.org/10.1016/j.
org/10.1159/000105604 jalz.2014.01.001
14. Reisberg B, Shulman MB, Torossian C et al 24. Rabin LA, Smart CM, Crane PK et al (2015)
(2010) Outcome over seven years of healthy Subjective cognitive decline in older adults: an
adults with and without subjective cognitive overview of self-report measures used across
impairment. Alzheimers Dement 6:11–24. 19 international research studies. J Alzheimers
https://doi.org/10.1016/j.jalz.2009.10.002 Dis 48(Suppl 1):S63–S86. https://doi.
15. Mitchell AJ, Beaumont H, Ferguson D et al org/10.3233/JAD-150154
(2014) Risk of dementia and mild cogni- 25. Embretson SE (1996) The new rules of mea-
tive impairment in older people with subjec- surement. Psychol Assess 8:341–349. https://
tive memory complaints: meta-analysis. Acta doi.org/10.1037/1040-3590.8.4.341
Psychiatr Scand 130:439–451. https://doi. 26. Burmester B, Leathem J, Merrick P (2015)
org/10.1111/acps.12336 Assessing subjective memory complaints: a
16. Jessen F (2014) Subjective and objective cog- comparison of spontaneous reports and struc-
nitive decline at the pre-dementia stage of tured questionnaire methods. Int Psychogeriatr
Alzheimer’s disease. Eur Arch Psychiatry Clin 27:61–77. https://doi.org/10.1017/
Neurosci 264:3–7. https://doi.org/10.1007/ S1041610214001161
s00406-014-0539-z 27. Broadbent DE, Cooper PF, FitzGerald P,
17. Ávila-Villanueva M, Rebollo-Vázquez A, Ruiz- Parkes KR (1982) The Cognitive Failures
Sánchez de León JM et al (2016) Clinical Questionnaire (CFQ) and its correlates. Br
relevance of specific cognitive complaints in J Clin Psychol 21(Pt 1):1–16. https://doi.
determining mild cognitive impairment from org/10.1111/j.2044-8260.1982.tb01421.x
28. Farias ST, Mungas D, Reed BR et al (2008) The clinician. J Psychiatr Res 12:189–198. https://
measurement of everyday cognition (ECog): doi.org/10.1016/0022-3956(75)90026-6
scale development and psychometric proper- 40. Rey A (1941) L’examen psychologique dans
ties. Neuropsychology 22:531–544. https:// les cas d’encéphalopathie traumatique. Arch
doi.org/10.1037/0894-4105.22.4.531 Psychol (Geneve) 28:215–285. http://psyc-
29. Sunderland A, Harris JE, Gleave net.apa.org/psycinfo/1943-03814-001
J (1984) Memory failures in everyday 41. Peña-Casanova J, Gramunt-Fombuena N,
life following severe head injury. J Clin Quiñones-Úbeda S et al (2009) Spanish mul-
Neuropsychol 6:127–142. https://doi. ticenter normative studies (NEURONORMA
org/10.1080/01688638408401204 project): norms for the rey-osterrieth com-
30. Jorm AF, Jacomb PA (1989) The Informant plex figure (copy and memory), and free
Questionnaire on Cognitive Decline in the and cued selective reminding test. Arch Clin
Elderly (IQCODE): socio-demographic cor- Neuropsychol 24:371–393
relates, reliability, validity and some norms. 42. Buschke H (1984) Cued recall in amnesia.
Psychol Med 19:1015–1022 J Clin Neuropsychol 6:433–440. https://doi.
31. Crook TH, Feher EP, Larrabee GJ (1992) org/10.1080/01688638408401233
Assessment of memory complaint in age- 43. Peña-Casanova J, Quiñones-Ubeda S,
associated memory impairment: the MAC- Gramunt-Fombuena N et al (2009)
Q. Int Psychogeriatr 4:165–176. https://doi. Spanish Multicenter Normative Studies
org/10.1017/S1041610292000991 (NEURONORMA Project): norms for verbal
32. Gilewski MJ, Zelinski EM, Schaie KW (1990) fluency tests. Arch Clin Neuropsychol 24:395–
The Memory Functioning Questionnaire for 411. https://doi.org/10.1093/arclin/
assessment of memory complaints in adult- acp042
hood and old age. Psychol Aging 5:482–490 44. Wechsler D (1997) Wechsler Adult Intelligence
33. Rami L, Mollica MA, García-Sanchez C et al Scale-III. The Psychological Corporation, San
(2014) The Subjective Cognitive Decline Antonio, TX
Questionnaire (SCD-Q): a validation study. 45. Lee GP, Loring DW, Newell J, McCloskey L
J Alzheimers Dis 41:453–466. https://doi. (1994) Figural fluency on the Five-Point Test:
org/10.3233/JAD-132027 preliminary normative and validity data. Int
34. Youn JC, Kim KW, Lee DY et al (2009) Neuropsychol Soc Progr Abstr, p 51
Development of the Subjective Memory 46. Wilson BA, Alderman N, Burgess PW et al
Complaints Questionnaire. Dement Geriatr (1996) The behavioural assessment of the dys-
Cogn Disord 27:310–317. https://doi. executive syndrome. Thames Valley Company,
org/10.1159/000205512 Bury St Edmunds
35. Squire LR, Wetzel CD, Slater PC (1979) 47. Fernández-Blázquez MA, Ruiz-Sánchez de
Memory complaint after electroconvulsive León JM, López-Pina JA et al (2012) A new
therapy: assessment with a new self-rating shortened version of the Boston Naming
instrument. Biol Psychiatry 14:791–801 Test for those aged over 65: an approach
36. Olazarán J, Valentí M, Frades B et al (2015) from item response theory. Rev Neurol
The Vallecas Project: a cohort to identify early 55:399–407
markers and mechanisms of Alzheimer’s dis- 48. Peña-Casanova J (1990) Programa inte-
ease. Front Aging Neurosci 7:181. https:// grado de exploración neuropsicológica-Test
doi.org/10.3389/fnagi.2015.00181 Barcelona. Masson, Barcelona
37. Montejo Carrasco P, Montenegro Peña M, 49. Warrington EK, James M (1991) The visual
Sueiro MJ (2012) The Memory Failures object and space perception battery. Thames
of Everyday (MFE) test: normative data Valley Test Company, Bury St Edmunds
in adults. Span J Psychol 15:1424–1431. 50. Pfeffer RI, Kurosaki TT, Harrah CH et al
https://doi.org/10.5209/rev_SJOP.2012. (1982) Measurement of functional activities
v15.n3.39426 in older adults in the community. J Gerontol
38. Hughes CP, Berg L, Danziger WL (1982) A 37:323–329. https://doi.org/10.1093/
new clinical scale for the staging of dementia. geronj/37.3.323
Br J Psychiatry 140:566–572. https://doi. 51. Yesavage JA, Brink TL, Rose TL et al (1982)
org/10.1192/bjp.140.6.566 Development and validation of a geriat-
39. Folstein MF, Folstein SE, McHugh PR (1975) ric depression screening scale: a preliminary
“Mini-mental state”: a practical method for report. J Psychiatr Res 17:37–49. https://doi.
grading the cognitive state of patients for the org/10.1016/0022-3956(82)90033-4
52. Spielberger C, Gorsuch R, Leshene R (1970) involved in the relationship between mood
Manual for the State-Trait Anxiety Inventory. and mild cognitive impairment. J Alzheimers
Consulting Psychologists Press, Palo Alto, CA Dis 48(Suppl 1):S115–S123. https://doi.
53. Molinuevo JL, Rabin LA, Amariglio R et al org/10.3233/JAD-150371
(2017) Implementation of subjective cog- 55. Fernández-Blázquez MA, Ávila-Villanueva M,
nitive decline criteria in research studies. Maestú F, Medina M (2016) Specific features
Alzheimers Dement 13:296–311. https://doi. of subjective cognitive decline predict faster
org/10.1016/j.jalz.2016.09.012 conversion to mild cognitive impairment.
54. Yates JA, Clare L, Woods RT, Matthews FE J Alzheimers Dis 52:271–281. https://doi.
(2015) Subjective memory complaints are org/10.3233/JAD-150956
Chapter 4
Deriving and Testing the Validity of Cognitive Reserve

Candidates
Yaakov Stern and Christian Habeck
Abstract
Empirical support has established cognitive reserve (CR) well as a concept since its inception three decades
ago, and most brain researchers subscribe to some version of CR as a collection of subject factors that influence
cognitive performance beyond simple brain structural health. We give a simple but precise analytic recipe to
test requirements for any plausible cognitive reserve candidate based on brain imaging. Gradations of partial
fulfillment of some but not all of these requirements are possible.
Key words Cognitive reserve, Brain aging, Resting-state fMRI, Cognitive aging
1 Introduction
Cognitive reserve refers to subject factors that determine cognitive

performance beyond obvious indicators of brain structural health.
Cognitive reserve was first formulated as concept in response to sev-
eral seemingly puzzling findings: (1) autopsy data of person appearing
cognitively normal at death showed varying degrees of Alzheimer-like
pathology; (2) when matched for disease severity of clinical presenta-
tions, PET scans revealed worse metabolic deficits in parietotemporal
areas for patients with higher premorbid educational and occupational
attainment. Formulated differently, matching people for their brain
structural and metabolic (in short, brain) health leaves unexplained
incommensurate differences in cognitive performance and clinical
presentations. Thus, the cognitive performance of two people with
relatively similar brain health is not necessarily more similar than the
cognitive performance of two people with appreciably different levels
of brain health (even though brain health is conducive to good cogni-
tion). It became clear that a third variable was needed to adequately
account for the relationship between brain structure/metabolism and
cognitive performance. This variable is cognitive reserve.
There have been many empirical accounts of cognitive reserve
candidates: leisure activity [1–6], lifetime mental activity [7, 8],
63
64 Yaakov Stern and Christian Habeck
and educational and occupational attainment [9–17], to name a

selective and nonrepresentative sample of a much richer literature.
In the current report, we would like to present a formal frame-
work for rigorously deriving and testing cognitive reserve candidates.
First, we list three axioms regarding cognitive reserve. These axioms
operate as principles in our own research program and partly originate
from historical empirical findings. It is useful to make the axioms very
explicit. While they may sound self-evident and somewhat trivial to
some readers, others might disagree, possibly strongly. Algorithmically,
these axioms are quite rigorous, but they leave open many possibilities
about the implementation and mechanisms of cognitive reserve.
1.1 Cognitive 1. Cognitive reserve has well-established proxies like education, intel-
Reserve Axioms ligence, occupational attainment, leisure activities, and other sub-
ject variables that express cognitively stimulating engagement.
2. Cognitive reserve is either an independent factor that influences
cognitive performance in addition to brain structure or a moder-
ating factor that influences the relationship between brain struc-
ture and cognitive performance; these two manifestations are not
exclusive and can conveniently be incorporated in linear regres-
sion models that contain (1) brain structure and (2) the putative
CR measure as direct effects and (3) an interaction term.
3. Although some of the cognitive reserve proxies might corre-
late with brain structural measures, cognitive reserve itself is
not based on brain structure; this entails in particular that:
●● It is not a mediating influence of brain structure on cogni-
tive performance.
●● It should be uncorrelated with brain structural measures.
These three axioms can be used to constrain the derivation of plau-
sible cognitive reserve candidates. Specifically, (1) cognitive reserve
should display correlations with some, or all, of the preexisting proxies
and (2) should show direct effects on cognitive performance in addi-
tion to brain structure or moderate the effects of brain structure on
performance while (3) being uncorrelated to brain structure. This last
point is contentious with some researchers who might refer to obvious
correlations between brain structure and CR proxies like intelligence
or education. Recently, we have laid out how such correlations can be
conceptualized as brain maintenance [18] which should be considered
as separate from, and in correlational terms orthogonal to, cognitive
reserve. CR proxies thus might correlate with cognitive reserve and
brain structure, but the latter correlation strictly separates them from
being a viable candidate for cognitive reserve directly. In other words,
while CR proxies might indicate the state of somebody’s brain struc-
tural health and how well the brain has been preserved in the course of
aging, cognitive reserve strictly speaks to how well somebody utilizes their
brain regardless of how well it has been preserved. CR proxies thus consist
Deriving and Testing Cognitive Reserve 65
Fig. 1 Venn diagram of cognitive reserve proxies in relation to brain structure
of two parts: (1) a part that correlates with brain structure and points
to possible formative influences to build and maintain a healthy brain
and (2) cognitive reserve, which in our rigorous definition is indepen-
dent of brain structure. The Venn diagram in Fig. 1 gives a simple
schematic explanation.
2 Materials
The demonstrations of this paper were run in 451 participants, aged

20–80, using volumetric data as the indicator of brain integrity and
attempting to derive a measure of CR from resting BOLD data. The
methods described here could be extended to any structural and
“brain activity” measures. All MR images were acquired on a 3.0T
Philips Achieva magnet that has been used since the start of this
study. A structural MPRAGE scan was obtained (TE/TR = 3/6.5 ms;
flip angle = 8; resolution = 256 × 256; FOV = 25.6 cm × 25.6 cm;
slice thickness/gap = 1/0 mm; 180 slices), as well as a resting BOLD
scan (TE/TR = 20/2000 ms; flip angle = 72; resolution = 112 × 112;
FOV = 22.4 cm × 22.4 cm; slice thickness/gap = 3/0 mm; 41
slices). The resting BOLD scan was either 5 minutes or 9.5 min
long. In terms of demographic variables, we have education, verbal
intelligence, and three measures of vocabulary, which were averaged
to produce an average vocabulary domain z-score.
3 Methods
Using each individual’s T1-weighted MPRAGE image, structural

imaging measures of both global and regional brain volume were
derived using the FreeSurfer software package (http://surfer.nmr.
mgh.harvard.edu/). Brain volume calculations automatically assign a
neuroanatomic label to each voxel; this has previously been shown to
have accuracy comparable to manual labeling [19, 20]. From this label-
ing, a set of volumetric ROIs was defined [21]. Although the estima-
tion procedure is automated, we manually check the accuracy of the
spatial registration and the white matter and gray matter segmentations
following the analytic procedures outlined in Fjell et al. [22] . The
calculated volume within each region was adjusted for variations in
individual’s intracranial brain volume which is measured using
BrainWash [23]. Using FreeSurfer software, cortical thickness was
measured by first reconstructing the gray/white matter boundary [24]
and the cortical surface and then calculating the distances between
these surfaces at each point across the cortical mantle. The maps pro-
duced are capable of detecting submillimeter differences between
groups [25]. Although the thickness estimation procedure is auto-
mated, we manually check the accuracy of the spatial registration and
the white matter and gray matter segmentations following the analytic
procedures outlined in Fjell et al. [22]. Using a validated automated
labeling system [20], FreeSurfer divided the cortex into 33 different
gyral-based areas in each hemisphere and calculates the mean thickness
in each area.
Furthermore, resting-state fMRI was collected for either 5 or
9.5 min, while the participant is instructed to lie still with their eyes
closed, to not think of any one thing in particular, and to not fall asleep.
MCFLIRT is used to register all the volumes to a reference image [26]
followed by scrubbing and motion correction as described in Power
et al. [27] and incorporating the recommendation of replacing deleted
volumes with zero volumes prior to filtering [28]. Using the FSL com-
mand “flsmaths –bptf t,” the motion-corrected signals are passed
through a band-pass filter with the cutoff frequencies of 0.01 and
0.08 Hz. After filtering, the first three volumes are discarded due to the
lag of the digital filter. Finally, we residualized the motion-corrected,
scrubbed, and temporally filtered volumes by regressing out the FD,
RMSD, left- and right-hemisphere white matter, and lateral-ventricular
signals [29]. We followed the taxonomy of functional units presented
by Power et al. [30] and produced residualized and motion-corrected
resting BOLD time series at 264 ROI locations. This data reduction
has the advantage that it enables easy computation of global functional
connectivity between all 264 × 263/2 = 34,716 node pairs.
The demonstrations that follow used omnibus measures that col-
lapse across all ROIs and all ROI pairs. This was done for the sake of
simplicity; an ecologically valid research program most likely would con-
sider structural variables and CR candidates with a topographic specific-
ity. In particular functional connectivity did not respect the intricate
taxonomy of different resting-state networks [30] and did not break
down connectivity into intra- and inter-network connectivity, which
would be advisable for real basic and diagnostic research questions.
3.1 Consideration Deriving a candidate measure for cognitive reserve should ideally fulfill
of a Connectivity- our three above postulates. Further, ideally it should be motivated with
Based Measure a convincing mechanistic insight or hypothesis, which suggests
plausible plasticity on the basis of lifetime experience of cognitively

enriched environments. This is a demanding program which, for the
sake of demonstration here, we cannot do total justice.
The measure we will investigate came out of a purely empirical
data exploration of the subject-level Fisher-Z matrices. This was
done for didactic expediency of demonstrating the workings of the
CR postulates; we are not claiming that there are not more plausible
connectivity-based CR measures.
The measure we considered is the intraindividual variability of
functional connectivity. This was computed on a subject level by tak-
ing the standard deviation across all 34,716 ROI-pair Fisher-Z values.
Large values thus imply a lot of regional variation in functional con-
nectivity, while lower values suggest relatively more uniformity in
functional connectivity across the brain.
3.2 Test Let us now turn to our three postulates, which represent the linchpin
of Postulates of this report. The variability measure is denoted as stdZ and is a scaler
for Cognitive Reserve value, i.e., it gives one number for each of the 451 participants.
1. The first postulate concerns a relationship with some well-
known CR proxies. We checked the correlation with NART
and years of education. The correlation with education years
was nonsignificant (R = −0.0376, p = 0.4264), while it was
modestly significant with NART (R = −0.1076, p = 0.0267).
Thus, people with higher verbal intelligence had slightly less
regional variability of functional connectivity in their brains.
2. The test of the second postulate is computationally somewhat
more demanding and involves many more, potentially arbi-
trary, decisions on account of the analyst. The second postulate
embodies a key requirement: Our CR candidate should
account for a particular cognitive performance above and
beyond brain structural independent variables and possibly
moderate the structure-cognition relationship. Thus CR could
be an independent factor that determines cognitive perfor-
mance in addition to brain structure, or a moderator, or both.
We can write down the regression equation
Dep = sIVβ + CRγ + sIV × CRδ + ε .
where Dep denotes our cognitive outcome measure, vocabu-
lary, sIV denotes a structural independent variable, CR stands for
the cognitive reserve candidate, and sIV × CR denotes the interac-
tion term. Greek letters βγδ denote regression weights, and ɛ stands
for the unexplained residual of Dep. It is easily conceivable that
there are more than just one structural independent variable and
correspondingly more than one interaction term. Concretely, we
will use mean cortical volume and mean cortical thickness, which
results in direct effects of two structural independent variables and
CR, and two interaction terms. More fundamental is the depen-
dency of the postulate check on the choice of dependent and inde-
pendent variables. The test of the second postulate for any CR

candidate obviously is always relative to a particular choice of cogni-
tive outcome and structural independent variables. Ideally, the best
CR measure captures similarities with regard to different cognitive
outcomes and structural variables, but this obviously does not need
to be so. If, for instance, a plethora of cognitive outcomes are used,
it is not guaranteed a priori that the particular CR measure in ques-
tion will show consistent behavior with regard to all of them. In
fact, if the cognitive outcomes that are considered do not capture
one underlying construct, it is unlikely that the CR measure will
display consistent behavior.
Anyways, our current report is not concerned with multiple
cognitive outcomes and only considered vocabulary where three
indicator variables were averaged prior to the test of the second
postulate to produce one cognitive outcome measure. We ran our
linear regression, and z-scored all three independent variables,
such that they were mean-centered, to avoid collinearities between
direct effects and interaction terms. (This is very important in gen-
eral.) The results are displayed in Table 1.
Our CR measure, stdZ, thus significantly accounts for vocabulary
performance with consistent directionality to the correlation with
NART: higher stdZ, i.e., greater regional variability in functional con-
nectivity, is associated with worse vocabulary performance. The inter-
action terms, however, are not significant. Of the brain structural
independent variables, only cortical volume has an expected positive
association with performance.
3. We now come to our last postulate. We explained in our initial
remarks that this postulate is at times contentious: a suitable
CR candidate should not display any associations with brain
structure. We laid out elsewhere and only restate that amply
documented correlations between lifestyle factors, enrichment
variables, and brain structure should be conceptualized as
brain maintenance, rather than cognitive reserve. The check
Table 1
Linear regression to explain vocabulary performance. Listed are T-
and p-values for each independent variable
Independent variable T P
Mean cortical volume 3.0931 0.0021
Mean cortical thickness −1.7053 0.0889
stdZ −2.6077 0.0094
stdZ × volume 0.2702 0.7871
stdZ × thickness 0.6915 0.4896
for the third postulate is therefore quite easy and involves a

correlation between stdZ and our brain structural variables.
Both mean cortical volume and mean cortical thickness showed
no association with stdZ (R = −0.0394, p = 0.4044 and
R = −0.0604, p = 0.208, respectively). Our CR candidate is
also independent of brain structure and independent of how
well participants’ cortical volume and thickness have been
maintained throughout the course of their adult life.
All three postulates are therefore fulfilled, although postulate
no. 2 did not show any moderation effects by our CR candidate.
4 Conclusions
This brief presentation describes a quantitative approach for deriving

direct measures of CR from brain-based measures. More conven-
tional research may make use of proxies for CR, but these proxies do
not directly measure CR, rather they represent experiences that might
impart CR. Deriving direct brain measures of CR will eventually
allow us to enhance our understanding of the neural implementation
of this concept.
References
1. Verghese J, Lipton RB, Katz MJ et al (2003) rate of hippocampal atrophy. PLoS One
Leisure activities and the risk of dementia in 3(7):e2598. https://doi.org/10.1371/jour-
the elderly. N Engl J Med 348(25):2508–2516 nal.pone.0002598
2. Hu G, Sarti C, Jousilahti P et al (2005) 8. Gates N, Valenzuela M (2010) Cognitive exercise
Leisure time, occupational, and commuting and its role in cognitive function in older adults.
physical activity and the risk of stroke. Stroke Curr Psychiatry Rep 12(1):20–27. https://doi.
36(9):1994–1999 org/10.1007/s11920-009-0085-y
3. Rovio S, Kareholt I, Helkala EL et al (2005) 9. Bonaiuto S, Rocca W, Lippi A (1990) Impact
Leisure-time physical activity at midlife and of education and occupation on prevalence
the risk of dementia and Alzheimer's disease. of Alzheimer’s disease (AD) and multi-infarct
Lancet Neurol 4(11):705–711. https://doi. dementia (MID) in Appignano, Macerata
org/10.1016/S1474-4422(05)70198-8 Province, Italy. Neurology 40(suppl 1):346
4. Helzner EP, Scarmeas N, Cosentino S et al 10. Cohen CI (1994) Education, occupation, and
(2007) Leisure activity and cognitive decline Alzheimer’s disease. JAMA 272(18):1405.
in incident Alzheimer disease. Arch Neurol Author reply 1406
64(12):1749–1754 11. Gun RT, Korten AE, Jorm AF et al (1997)
5. Akbaraly TN, Portet F, Fustinoni S et al (2009) Occupational risk factors for Alzheimer disease:
Leisure activities and the risk of dementia in a case-control study. Alzheimers Dis Assoc
the elderly: results from the Three-City Study. Disord 11(1):21–27
Neurology 73(11):854–861. https://doi. 12. Helmer C, Letenneur L, Rouch I et al (2001)
org/10.1212/WNL.0b013e3181b7849b Occupation during life and risk of dementia in
6. Scarmeas N, Levy G, Tang MX et al (2001) French elderly community residents. J Neurol
Influence of leisure activity on the inci- Neurosurg Psychiatry 71(3):303–309
dence of Alzheimer’s disease. Neurology 13. Anttila T, Helkala EL, Kivipelto M et al (2002)
57(12):2236–2242 Midlife income, occupation, APOE status, and
7. Valenzuela MJ, Sachdev P, Wen W et al (2008) dementia: a population-based study. Neurology
Lifespan mental activity predicts diminished 59(6):887–893
14. Ravaglia G, Forti P, Maioli F et al (2002) Cortex 19(9):2001–2012. https://doi.

Education, occupation, and prevalence of org/10.1093/cercor/bhn232
dementia: findings from the Conselice study. 23. Ardekani BA, Guckemus S, Bachman A et al
Dement Geriatr Cogn Disord 14(2):90–100 (2005) Quantitative comparison of algorithms
15. Finkel D, Andel R, Gatz M et al (2009) The for inter-subject registration of 3D volumet-
role of occupational complexity in trajectories ric brain MRI scans. J Neurosci Methods
of cognitive aging before and after retirement. 142(1):67–76. https://doi.org/10.1016/j.
Psychol Aging 24(3):563–573. https://doi. jneumeth.2004.07.014
org/10.1037/a0015511 24. Dale AM, Fischl B, Sereno MI (1999) Cortical
16. Garibotto V, Borroni B, Sorbi S et al (2012) surface-based analysis. I Segmentation and sur-
Education and occupation provide reserve in face reconstruction. Neuroimage 9(2):179–194
both ApoE epsilon4 carrier and noncarrier 25. Fischl B, Dale AM (2000) Measuring the
patients with probable Alzheimer’s disease. thickness of the human cerebral cortex from
Neurol Sci 33(5):1037–1042. https://doi. magnetic resonance images. Proc Natl Acad
org/10.1007/s10072-011-0889-5 Sci U S A 97(20):11050–11055. https://doi.
17. Stern Y, Gurland B, Tatemichi TK et al (1994) org/10.1073/pnas.200033797
Influence of education and occupation on the 26. Jenkinson M, Bannister P, Brady M et al
incidence of Alzheimer’s disease. J Am Med (2002) Improved optimization for the robust
Assoc 271:1004–1010 and accurate linear registration and motion
18. Habeck C, Razlighi Q, Gazes Y et al (2016) correction of brain images. NeuroImage
Cognitive reserve and brain maintenance: 17(2):825–841
orthogonal concepts in theory and practice. 27. Power JD, Barnes KA, Snyder AZ et al
Cereb Cortex. https://doi.org/10.1093/ (2012) Spurious but systematic correla-
cercor/bhw208 tions in functional connectivity MRI net-
19. Fischl B, Salat DH, Busa E et al (2002) Whole works arise from subject motion. NeuroImage
brain segmentation: automated labeling of 59(3):2142–2154. https://doi.org/10.1016/
neuroanatomical structures in the human j.neuroimage.2011.10.018
brain. Neuron 33(3):341–355 28. Carp J (2013) Optimizing the order of opera-
20. Fischl B, van der Kouwe A, Destrieux C et al tions for movement scrubbing: comment on
(2004) Automatically parcellating the human Power et al. NeuroImage 76:436–438. https://
cerebral cortex. Cereb Cortex 14(1):11–22. doi.org/10.1016/j.neuroimage.2011.12.061
https://doi.org/10.1093/cercor/bhg087 29. Birn RM, Diamond JB, Smith MA et al (2006)
21. Kennedy KM, Erickson KI, Rodrigue KM et al Separating respiratory-variation-related fluc-
(2009) Age-related differences in regional brain tuations from neuronal-activity-related fluctua-
volumes: a comparison of optimized voxel-based tions in fMRI. NeuroImage 31(4):1536–1548.
morphometry to manual volumetry. Neurobiol https://doi.org/10.1016/j.neuroimage.2006.
Aging 30(10):1657–1676. https://doi. 02.048
org/10.1016/j.neurobiolaging.2007.12.020 30. Power JD, Cohen AL, Nelson SM et al (2011)
22. Fjell AM, Westlye LT, Amlien I et al (2009) Functional network organization of the human
High consistency of regional cortical thin- brain. Neuron 72(4):665–678. https://doi.
ning in aging across multiple samples. Cereb org/10.1016/j.neuron.2011.09.006
Chapter 5
Methods for Pathological Classification of Alzheimer’s

Disease
Johannes Attems, Kirsty E. McAleese, and Lauren Walker
Abstract
The hallmark protein aggregates required for a neuropathological diagnosis of Alzheimer’s disease are
intracellular neurofibrillary tangles and neuropil threads consisting of hyperphosphorylated tau, extracellular
parenchymal amyloid β plaques, and neuritic plaques. In this chapter, we describe the neuropathological
criteria for the neuropathological diagnosis of Alzheimer’s disease, as well as the fixation, processing, and
histology protocols required for the assessment of post-mortem human brains.
Key words Neuropathology, Alzheimer’s disease, Hyperphosphorylated tau, Amyloid β, Neuritic

plaques, Cerebral amyloid angiopathy, Post-mortem tissue, Histology
1 Introduction: Neurodegeneration (General Considerations)
The unifying feature of neurodegenerative diseases is the accumu-

lation of insoluble aggregates of physiologically soluble proteins of
the central nervous system. The native function of these proteins is
lost resulting in protein misfolding, oligomerization into fibrils,
and subsequent accumulation and deposition as intracellular (in
neurons and/or neuroglia) or extracellular aggregates. It is generally
assumed that misfolded proteins exert neurotoxicity leading to
progressive neuronal dysfunction with consecutive neuronal loss of
specific populations of neurons. However, the underlying mecha-
nisms are still poorly understood. These insoluble aggregates can
be visualized in human post-mortem brain tissue for microscopic
assessment using appropriate histochemical techniques and immu-
nohistochemical antibodies, and it is the molecular nature of the
respective protein deposit and its topographical localization that is
the basis of the classification of neurodegenerative disease.
71
72 Johannes Attems et al.
1.1 Alzheimer’s The hallmark protein aggregates required for a neuropathological

Disease: diagnosis of Alzheimer’s disease (AD) are intracellular neurofibril-
Neuropathology lary tangles/neuropil threads (NFTs/NTs) consisting of hyper-
phosphorylated tau (HP-T), extracellular amyloid β (Aβ) plaques,
and neuritic plaques (NP). In addition, cerebral amyloid angiopa-
thy (CAA) is typically present, but its assessment is not required for
stating a neuropathological diagnosis of AD according to current
diagnostic criteria.
1.1.1 Macroscopic On gross examination, post-mortem brains of AD patients often

Characteristics appear relatively normal without striking pathological changes, and
reduction in brain weight may be minimal. However, the charac-
teristic atrophy, i.e., thinning of the gyri and deepening of the
sulci, in AD involves medial temporal lobe structures including the
entorhinal cortex, hippocampus, amygdala, olfactory bulb, as well
as inferior temporal, superior frontal, and middle frontal gyri [1].
Cerebral atrophy leads to enlargement of the ventricles (i.e., hydro-
cephalus internus e vacuo), in particular the inferior horn of the
lateral ventricles. The substantia nigra (midbrain) is well pig-
mented, while the locus coeruleus (pons) might be pale, and the
cerebellum is normal (Fig. 1).
1.1.2 Neurofibrillary Neurofibrillary tangles (NFTs) and neuropil threads (NTs) are
Tangles and Neuropil aggregates of hyperphosphorylated microtubule-associated protein
Threads (MAP) tau (HP-T): a microtubule-stabilizing protein that is abun-
dant in axons and dendrites. Six isoforms of MAP tau exist in the
human brain, and aggregates of 3-repeat (3R) and 4-repeat (4R)
isoforms are seen in AD. Under pathological conditions, MAP tau
is hyperphosphorylated resulting in the disintegration of microtu-
bules and aggregation of insoluble filaments/fibrils composed of
HP-T [2]. NFTs/NTs are composed of aggregates of HP-T, while
so called pre-tangles represent non-aggregated HP-T and are con-
sidered to be precursors of NFT/NT. Using appropriate immuno-
histochemical antibodies, i.e., AT8 antibody, both NFT and NT
can be visualized for histological assessment. NFTs are located in
neuronal cell bodies and NTs in axons and dendrites, the latter
appearing in immunohistochemically stained sections as immu-
nopositive threads in the neuropil (Fig. 2). In AD, NFT and NT
primarily manifest in trans-entorhinal and entorhinal cortices and
then gradually spread toward the hippocampus and isocortex [3],
and this pattern is the basis for diagnostic Braak NFT stages [4]
(see Sect. 3.6.1). Autopsy studies have revealed pre-tangle HP-T
was found in the locus coeruleus in the majority of individuals
under 30 years of age who were devoid of cortical HP-T pathology,
suggesting that HP-T pathology begins in the locus coeruleus
rather than the trans-/entorhinal cortex [5, 6], but this is
controversial [7]. Importantly, the extent of neocortical NFT/NT
has been shown to correlate with cognitive deficits [8].
Pathological Classification of Alzheimer’s Disease 73
Fig. 1 Comparison between formalin-fixed tissue slices taken at the level of the
anterior hippocampus of an individual without dementia (A) and a patient with
severe AD. ((B) and (C)) There are marked enlargement of the lateral ventricle
(blue asterisk), considerable atrophy of the medial temporal lobe (red asterisk),
as well as thinning of the gyri and deepening of the sulci (black arrowheads).
However, the substantial nigra of the midbrain remains fairly well pigmented (c).
Photographs courtesy of Dr. Christopher Morris and the Newcastle Brain Tissue
Resource. Both scale bars represent 1 cm
1.1.3 Amyloid β Plaques Amyloid β (Aβ) peptides (4kDA) are generated by β- and γ-secretase
cleavage of the transmembraneous amyloid precursor protein
(APP) and comprise peptides terminating at carboxyl terminus 40
(Aβ40) and 42 (Aβ42) [9]. While Aβ40 constitutes the majority of
cerebral Aβ (over 95%), Aβ42 aggregates more readily and, there-
fore, is believed to initiate the formation of oligomers, fibrils, and
plaques [10, 11]. Aβ aggregates extracellularly, and using appro-
priate antibodies, i.e., 4G8 antibody, various morphological forms
of Aβ aggregates can be detected. The main parenchymal Aβ
deposits are diffuse or focal: diffuse Aβ deposits are usually large
(up to several 100 μm) and show ill-defined borders and are often
referred to as “fleecy” and “subpial band-like,” while focal Aβ
plaques incorporate a dense, spherical-like accumulation of Aβ
peptides (Fig. 3). The topographical locations of Aβ deposits
follow a distinct sequence that can be staged diagnostically using
Thal Aβ phases [12]. Cerebral Aβ depositions are a hallmark of
AD, but in the majority of clinicopathological correlative studies,
Fig. 2 Photomicrographs of neurofibrillary tangles and neuropil threads. (A) Distribution of HP-T immunopositiv-
ity in the hippocampus, subiculum, transentorhinal, and occipital-temporal cortex of an Alzheimer’s disease
case. (Ai) Magnified image of neurofibrillary tangles and neuropil threads in the CA1 region of the hippocam-
pus. (Aii) Neurofibrillary tangles encompass immunopositivity within the neuronal soma (black arrow); neuropil
thread comprises of immunopositivity in the neuronal process (red arrow head). CA1 and CA4, hippocampal
cornu ammonis (Ammon’s horn) sectors 1 and 4; GR, granule cell layer of the dentate gyrus; Sub, subiculum;
Trans Ent, transentorhinal cortex; OTC, occipital-temporal cortex. Immunohistochemistry with antibody against
hyperphosphorylated tau, AT8. Scale bars: A1, 50 μm; Aii, 10 μm
Fig. 3 Photomicrographs of Aβ plaques. (A) Distribution of Aβ immunopositivity in the prefrontal cortex of an

Alzheimer’s disease case. (Ai) Subpial Aβ deposits; (Aii) focal Aβ deposits; (Aiii) fleece Aβ deposits. WM, white
matter; Aβ, amyloid beta. Immunohistochemistry with antibody against amyloid precursor protein, 4G8. Scale
bars: Bi and Biii, 50 μm; Bii, 20 μm
cerebral Aβ load does not correlate with clinical dementia, and

parenchymal Aβ deposits are frequently seen to a considerable
extent in post-mortem brains of non-demented individuals (see
Sect. 3.6.2).
1.1.4 Neuritic Plaques Neuritic plaques (NP) consist of a focal deposit of Aβ42 [13], with
an external halo (the corona) of HP-T-positive processes forming
dystrophic neurites (Fig. 4). NPs are strongly associated with AD,
and assessment of the presence and density of NP is performed
using the age-adjusted criteria of the Consortium to Establish a
Registry for Alzheimer’s Disease (CERAD) [14] (see Sect. 3.6.3).
1.1.5 Cerebral Amyloid Cerebral amyloid angiopathy (CAA) is defined as the deposition of
Angiopathy a congophilic material (i.e., positive staining with a Congo red dye)
in cerebral leptomeningeal and intracortical arteries, arterioles,
capillaries, and rarely veins. CAA is also referred to as congophilic
amyloid angiopathy [15, 16] and is strongly associated with
AD-related pathology, present in 80–100% of AD cases [17, 18].
CAA is characterized by the deposition of predominantly Aβ40 [9]
in the tunica media of vessel walls that eventually spreads into all
vessel wall layers causing vessel wall thickening, smooth muscle cell
migration, and overall disruption of vessel architecture, although
Fig. 4 Photomicrographs of neuritic plaques. (A) Distribution of neuritic plaques in the hippocampus, subiculum,
transentorhinal, and occipital-temporal cortex of an Alzheimer’s disease case. (Ai) Magnified image of neurotic
plaques in the CA1 region of the hippocampus. (Aii) Neuritic plaque consisting of a Aβ core (red arrow head)
and dystrophic neurites consisting of HPτ (blue arrow). CA1 and CA4, hippocampal cornu ammonis (Ammon’s
horn) sectors 1 and 4; GR, granule cell layer of the dentate gyrus; Sub, subiculum; Trans Ent, transentorhinal
cortex; OTC, occipital-temporal cortex; Aβ, amyloid beta; HPτ, hyperphosphorylated tau. Histological staining
with Gallyas silver stain. Scale bars: Ai, 50 μm; Aii, 10 μm
Fig. 5 Photomicrographs of cerebral amyloid antipathy (CAA) in the occipital

region of an Alzheimer’s disease brain (A) cortical CAA; (Ai) magnified image of
cortical CAA; (B) leptomeningeal CAA; (C) capillary CAA. Immunohistochemistry
with antibody against amyloid precursor protein, 4G8. Scale bars: A, 100 μm; Ai,
20 μm; B, 50 μm; C, 10 μm
endothelial cells are preserved (Fig. 5). CAA frequently has a patchy

distribution and primarily affects leptomeningeal and cortical ves-
sels of neocortical regions [19]. The occipital lobe has been reported
to be the site that is both most frequently and severely affected by
CAA, followed by either frontal, temporal, or parietal lobes [17,
20–22]. Aβ depositions in the walls of capillaries are referred to as
capillary CAA (capCAA) and usually present as strong staining
lining capillary walls in sections stained with Aβ antibodies, e.g.,
4G8. The presence of capCAA distinguishes two types of CAA:
CAA-type 1 is characterized by the presence of capCAA and usually
also shows Aβ depositions in leptomeningeal and cortical vessels,
whereas in CAA-type 2, Aβ depositions are restricted to leptomen-
ingeal and cortical vessels without capillary involvement [23].
Histological assessment of CAA is not required for the diagnosis of
AD; however, it is an important AD-associated pathology that
should be considered. A neuropathological consensus criterion for
the scoring of CAA has been recently established [24].
1.2 Neuro- Neuropathological assessment of the post-mortem brain is per-

pathological Criteria formed in a standardized manner, applying reproducible methods
and following internationally recognized criteria. Neuropathological
consensus criteria are available for the neuropathological hallmark
lesions of AD, i.e., NFT/NT [4] Aβ plaques [12] and neuritic
plaques [14], as well as the National Institute on Aging-Alzheimer’s

Association (NIA-AA) score for AD neuropathologic change [25]
(see Sect. 3.6.4).
2 Materials
2.1 Brain Dissection 1. Macroblade (CellPath).

(Both Fresh and Fixed) 2. Scalpel holder (Swann Morton).
3. Disposable scalpel blade, size 22 (Swann Morton).
2.2 Fixative 1. Formalin solution: 100 ml formaldehyde (40%); 900 ml distilled

and Tissue Processing water (dH2O) (1:10 dilution) (Genta Medical).
2. Small amount of marble chips.
3. Ethanol (100, 95, 70, 50% in dH2O).
4. Chloroform.
5. Histology wax (melting point of 60 °C).
2.3 Tissue 1. Pretreated slides (4% 3-aminopropyltriethoxysilane (APES)) or

Preparation charged slides, i.e., Superfrost Plus (Thermo Shandon).
2. Rotary microtome (HM325, Thermo Scientific).
3. Oven capable of reaching temperatures from 37 to 60 degrees.
4. Ethanol (100, 75, and 50% on dH2O).
5. Xylene.
2.4 Gallyas Staining 1. Xylene.

2. Ethanol (100, 75, and 50% in dH2O).
3. 4% formalin solution: 100 ml formaldehyde (40%) (Genta
Medical), 900 ml dH2O (1:10 dilution).
4. 1% potassium permanganate.
5. 5% oxalic acid.
6. 5% periodic acid.
7. Alkaline silver solution: 4 g NaOH in 50 ml dH2O; add 10 g
potassium iodide. Slowly add 3.5 ml silver nitrate drop by
drop; make up to 100 ml dH2O.
8. 0.25% acetic acid.
9. Physical developer: solution A—50 g sodium carbonate in 1 L
dH2O. Solution B—in 1 L dH2O, add in order 2 g ammo-
nium nitrate, 2 g silver nitrate (analar), and 10 g tungstosilicic
acid (analar). Solution C—as in solution B, with the addition
of 6.1 ml 40% formalin. Slowly add 30 ml solution B to 100 ml
solution A while shaking vigorously, and then add 70 ml solu-
tion C (Note 1).
10. 0.2% gold chloride.

11. 5% “hypo” (sodium thiosulfate).
12. Harris hematoxylin.
13. DPX hard-set mountant (CellPath).
14. Coverslips.
2.5 Immuno- 1. Xylene.

histochemistry 2. Ethanol (100, 75, and 50% in dH2O).
3. Citrate buffer: 8.4 g citric acid, 4 L dH2O, 52 ml 2 molar (M)
sodium hydroxide, pH to 6.
4. Concentrated formic acid.
5. Microwave.
6. 3% hydrogen peroxide: 10 ml hydrogen peroxide (30%) in
90 ml tap water (Note 2).
7. Tris-buffered saline (TBS): 6.95 g Tris base, 30.3 g Tris-HCL,
45 g sodium chloride made up in 5 L dH2O, pH to 7.6.
8. Hydrophobic barrier pen.
9. Primary antibodies made up in TBS (see Table 3 for optimum
dilutions).
10. TBS Tween: 4 L TBS as above, 2 ml Tween (shake well to mix).
11. MenaPath X-Cell Plus HRP Polymer Detection Kit (Menarini
Diagnostics) (or alternative streptavidin-biotin system).
12.
3,3-diaminobenzidine (DAB) substrate kit (Menarini
Diagnostics).
13. Harris hematoxylin.
14. DPX hard-set mountant (CellPath).
15. Coverslips.
2.6 Neuropatho- 1. Microscope capable of 100× magnification.

logical Assessment 2. Ethanol for cleaning slides.
Criteria for AD
3 Methods
3.1 Fresh Brain The cerebrum, and spinal cord if required, is removed from the
Dissection deceased patient following standard protocols. All tissue, blood,
and Fixation and CSF must be treated as a potentially infectious case. The
pathologist and assisting personnel must wear a gown, nitrile
gloves, and shoe covers, and dissection of neurologic tissues is
carried out in a fume hood. The cerebrum, cerebellum, and brain-
stem are inspected and weighed, and the cerebrum length is mea-
sured from the frontal pole to the occipital pole. The arteries of the
circle of Willis are examined for the presence of atherosclerotic

changes. The dissection is carried out as follows:
1. Separation of the cerebellum and brainstem from the cerebral
hemispheres at pons-midbrain junction.
2. Cerebral hemispheres are separated by a midsagittal incision
through the corpus callosum; cerebellar hemispheres are
separated at the point of the left dentate nucleus; and brain-
stem is separated down the midline.
3. The exposed “cut” surfaces are covered in 60gsm medical
grade paper to prevent tissue shrinkage, and the right cerebral
and right cerebellar hemispheres and the right brainstem are
fixed in formalin with marble chips. The right cerebrum is sus-
pended in formalin solution (via the junction of the middle
cerebral artery and the posterior communicating artery); the
left cerebral and cerebellar hemisphere and brainstem are dis-
sected into a series of 1-cm-thick slices according to a standard
protocol.
4. The serial coronal slices are placed into individual plastic enve-
lopes, labeled, and sealed and then snap frozen on copper
plates at −120 °C and stored at −80 °C.
3.2 Fixed Dissection Tissue is fixed for 4 weeks at room temperature. The right cere-
and Embedding brum is weighed, and length is measured from the frontal pole to
the occipital pole. The arteries of the circle of Willis are examined
for the presence of atherosclerotic changes. The midbrain is
removed from the cerebrum cutting in between the mammillary
body and superior colliculus. The cerebral hemisphere, cerebellum,
and brainstem are then serially dissected at 7 mm intervals (Fig. 6).
Regions (including specific Brodmann areas (BA)) required for
Fig. 6 Following fixation, the right hemisphere (A) is dissected at 7 mm intervals (B) as are the fixed semi-
brainstem and right cerebellum (C). Brain regions important of the neuropathological diagnosis of AD are then
sub-dissected and processed into paraffin wax. Photographs courtesy of Dr. Christopher Morris and the
Newcastle Brain Tissue Resource. Scale bar in A and B represents 2 cm and 1 cm in C
Table 1
Brain regions (including specific Brodmann areas, BA) required for
neuropathological diagnosis
1. Prefrontal superior half (BA 9/46/32)

2. Mid-frontal superior half (BA 9/8)
3. Cingulate (BA 24/32)
4. Corpus striatum and amygdala
5. Motor (BA 4) and sensory (BA 3/1/2/5) cortex
and deep white matter
6. Trans-/entorhinal cortex (anterior: BA 27/28),
occipital-temporal cortex (BA 36), and dentate fascia
7. Hippocampus and parahippocampal gyrus (BA 27/28),
occipital-temporal cortex (BA 36), and inferior temporal cortex
8. Temporal cortex (BA 20/21/22/41/42)
9. Thalamus
10. Parietal cortex and deep white matter (BA 40/22)
11. Occipital cortex inferior half (BA 17/18/19)
12. Cerebellar hemisphere and dentate
13. Upper midbrain
14. Lower midbrain
15. Upper mid pons
16. Mid-medulla
neuropathological diagnosis are sub-dissected as required (Table 1).

All dissected tissue is individually bagged in biopsy bags or plastic
cassettes (depending on size of tissue sample), labeled, and sealed
for tissue processing.
Tissue processing is carried out in an automatic processor
(Leica ASP 300S) according to the schedule in Table 2.
Processed samples are then placed in clean molten histology
wax (60 °C) in the embedding center (Leica Microsystems,
EG1150H). Molten wax is poured into appropriate-sized plastic
mold, and the tissue sample and case label are added and then
placed on a cooling plate until solid (Leica Microsystems,
EG1150C). If required, excess wax is trimmed using a hot Para
Trimmer (Thermo Scientific).
3.3 Tissue To aid tissue adhesion, slides are pretreated with 4%

Preparations 3-
aminopropyltriethoxysilane (APES), or Superfrost slides are
and Sectioning used. To prepare APES-coated slides, clean glass slides are placed
into racks and immersed in a 4% solution of APES made up in
Table 2
Three-day CNS processing schedule
Step Time
4% formalin (no marble chips) 34 min
50% ethanol 2 h
70% ethanol 3 h 50 min
95% ethanol 5 h
100% ethanol 6 h
100% ethanol 6 h
100% ethanol 8 h
Chloroform 3 h
Chloroform 6 h
Chloroform 7 h
Histology wax 5 h 60 °C (pressure/
vacuum)
vacuum)
vacuum)
acetone for 5 min, followed by 5 min in acetone before a final

5 min immersed in 4% APES solution. Slides are drained in dH2O,
dried at 37 °C, and stored until used. The tissue is sectioned at
6 μm on a rotary microtome and dried for 2 days at 37 °C followed
by 1 h at 60 °C.
3.4 Gallyas Staining Silver staining such as Gallyas is used to visualize dystrophic neu-
rites associated with neuritic plaques; however, amyloid deposits
are not detected using this method. The protocol is as follows:
1. Sections placed in 60 °C oven to aid dewaxing.
2. Deparaffinize sections in two changes of xylene (5 min each).
3. Rehydrate through decreasing concentrations of alcohol
(2 × 100%, 1 × 75%, 1 × 50%, 2 min each) to water.
4. Postfix in 4% formalin (10 min).
5. Wash in dH2O.
6. Immerse in 1% potassium permanganate (5 min).
7. Wash in dH2O.
8. Place in 5% oxalic acid (3 min).

9. Wash well in dH2O.
10. Place in 5% periodic acid (5 min).
11. Wash in dH2O.
12. Apply alkaline silver iodide solution (1 min).
13. Wash in 0.25% acetic acid (three changes, 1 min each, with
gentle agitation).
14. Wash well in dH2O.
15. Place in physical developer (look for pale gray/brown color
change (5–20 min)).
16. Wash in 0.25% acetic acid (two changes, 1 minute each, with
gentle agitation).
17. Wash well in water dH2O.
18. Tone in 0.2% gold chloride (10–20 s).
19. Wash in water and fix in 5% “hypo” (sodium thiosulfate) (2 min).
20. Wash sections and counterstain with hematoxylin.
21. Dehydrate through increasing concentrations of alcohol (70%,
95%, and 2 × 100%), clear in two changes of xylene (5 min
each), and mount with DPX and coverslip (Note 3).
3.5 Immunohisto- Immunohistochemistry (IHC) of formalin-fixed paraffin-

chemistry embedded tissues allows the visualization of pathological protein
aggregates while preserving the morphological and cellular struc-
ture of the tissue. Enzyme precipitate IHC is preferred to fluores-
cent IHC for diagnostics as it is more cost-effective, and the
chromogen will not fade with time. Formalin fixation can cross-
link antigens and mask epitopes; therefore, to increase antigenicity
of the required epitope, antigen retrieval is performed prior to
immunohistochemistry. Information of antigen retrieval required
and manufacturer’s details of primary antibodies (including optimum
dilutions) are provided in Table 3.
Table 3
Optimum antibody dilutions and antigen retrieval protocol
Working
dilution
Antibody Target Host Manufacturer (in TBS) Antigen retrieval
AT8 Hyperphosphorylated Mouse Innogenetics, 1:4000 Immersion in 0.01 m.L-1
tau Ser202/205 Ghent, citrate buffer (pH 6) and
(HP-τ) Belgium microwaving for 10 min
4G8 β amyloid 17-24 Mouse Covance, 1:15,000 Immersion in concentrated
Alnwick, UK formic acid for 1 h
The protocol is as follows:

1. Sections placed in 60 °C oven to aid dewaxing.
2. Deparaffinize sections in two changes of xylene (5 min each).
3. Rehydrate through decreasing concentrations of alcohol
(2 × 100%, 1 × 75%, 1 × 50%, 2 min each) to water.
4. If using a peroxidase-based detection system, quench endoge-
nous peroxidase by immersing slides in 3% hydrogen peroxide
diluted in tap water (20 min).
5. Wash well in water, rinse in Tris-buffered saline (TBS), and (if
using) apply hydrophobic barrier pen around the tissue.
6. Incubate in primary antibody at optimum dilution (diluted in
TBS) for 1 h at room temperature.
7. Rinse in TBS, and final wash in TBS Tween.
We routinely use MenaPath X-Cell Plus HRP Polymer
Detection Kit (Menarini Diagnostics), which is a sensitive biotin-
free detection system capable of binding to mouse and rabbit anti-
bodies. Following manufacturer’s instructions, the primary antibody
is bounded by the universal probe (1), which is then detected by the
HRP-polymer (2). The increased number of enzymes bound to the
polymer backbone ensures a sensitive and clean end result. Sections
are rinsed well in TBS, and the chromogen is developed using
3,3-diaminobenzidine (DAB) substrate kit (also Menarini
Diagnostics). Once the desired color of brown has been achieved
(2–4 min), wash tissue well in tap water, and counterstain lightly in
hematoxylin (Note 4). Dehydrate through increasing concentra-
tions of alcohol, clear in xylene, and mount using DPX, which is a
colorless synthetic resin mountant that dries quickly and enables
long-term preservation of tissue samples (Note 5).
3.6 Neuropatho- Hallmark pathological lesions associated with AD (i.e., neurofibril-

logical Criteria for AD lary pathology, Aβ plaques, and neuritic plaques) follow a stepwise
topographical pattern of deposition through the brain, which cor-
responds to advancing clinical symptoms [4, 12, 14]. Current con-
sensus criteria for the degree of pathology attributed to AD (termed
AD neuropathologic change) evaluate the progression of each type
of pathology and yield a combined score to determine an overall
level of AD-related pathology in the brain [25, 26]. For the iden-
tification of pathological lesions, stained sections are viewed under
a light microscope at X100 magnification.
3.6.1 Assessment Braak stages describe the pattern of neurofibrillary pathology (NTs
of Hyperphosphorylated and NFTs) as outlined by Braak and colleagues [4]. Subsequently,
Tau (NTs and NFTs) the BrainNet Europe Consortium identified four distinct anatomi-
cal regions required to accurately assign Braak stage, the anterior
hippocampus at the level of the uncus, posterior hippocampus at
Fig. 7 Schematic illustrating semiquantitative scoring utilized in Braak staging for assessment of neurofibril-
lary tangles (NFTs) and neuropil threads (NTs) (A). Regions are classed as having mild pathology when immu-
noreactive structures positive for AT8 antibody are barely noted at X100 magnification. Moderate pathology is
assigned when immunoreactive structures are easily visible at ×100 magnification, while the severe score is
assigned when NFTs and NTs are visible even without a microscope. Specific brain regions are required to
assign Braak stages (B) including the anterior hippocampus at the level of the uncus (i), the posterior hippo-
campus at the level of the lateral geniculate (ii), and the visual cortex including the calcarine fissure (iii). The
semiquantitative stage is important when assessing Braak stages as Braak stage I can be assigned with mild
pathology in the transentorhinal cortex (green square, Bi); however, the remaining stages require pathology of
at least moderate density to qualify as the next Braak stage (orange or red squares, Bi, Bii, and Biii). For a full
description of regions affected by each stage, see Sect. 3.6.1. Scale bar represents 100 μm and is valid for all
photomicrographs
the level of the lateral geniculate nucleus, middle temporal gyrus,

and visual cortex including calcarine fissure [27]. Slides stained for
hyperphosphosphorylated tau (for protocol, see Sect. 3.5) are
examined under a light microscope. In addition to the presence of
NFTs and NTs in a given brain region, the amount of pathology is
taken into account when assigning Braak stages, which is based on
a four-tiered scale (absent, mild, moderate, and severe (Fig. 7A).
Brain regions involved in each Braak stage are illustrated in Fig. 7B
and described as follows:
1. Braak stage 0—All regions must be described as HP-T

negative.
2. Braak stage I—Trans-entorhinal region in the anterior hippo-
campus at the level of uncus is immunopositive for HP-T of at
least mild density.
3. Braak stage II—In the posterior hippocampus at the level of
lateral geniculate nucleus, outer layers of remnants of the ento-
rhinal region show at least moderate HP-T immunopositivity,
or outer and inner layers show at least mild HP-T
immunopositivity.
4. Braak stage III—At least moderate HP-T immunopositivity
extends to the occipitotemporal gyrus.
5. Braak stage IV—At least moderate HP-T immunopositivity up
to the middle temporal gyrus.
6. Braak stage V—At least moderate HP-T immunopositivity in
the peristriate area of the occipital cortex.
7. Braak stage VI—At least moderate HP-T immunopositivity in
layer V of the striate area in the occipital cortex (BA17—
primary visual cortex). In Braak stages V and VI, HP-T immu-
nopositivity is seen in other neocortical areas as well, but only
the occipital cortex needs to be assessed for staging purposes.
3.6.2 Assessment The distribution of Aβ plaques follows a hierarchical pattern

of β-Amyloid Plaques throughout the brain parenchyma, and this is reflected by the Thal
phases of Aβ deposition [12]. Thal phases can be assigned using
Aβ-stained sections (for protocol, see Sect. 3.5, Note 6) of the
specified brain regions. Of note, unlike Braak stages, Thal phases
are based on a dichotomous scoring system (i.e., present or absent),
and the presence of any single parenchymal Aβ deposit is sufficient
to score this area positive for the respective Thal phase (Fig. 8):
1. Phase 1—Deposits are visible in the neocortex (e.g., tempo-
ral/frontal cortex).
2. Phase 2— Hippocampus (in particular cornu ammonis (CA)
sector 2), entorhinal cortex, and cingulate are involved.
3. Phase 3—Subcortical structures including striatum, thalamus,
and hypothalamus.
4. Phase 4—Brainstem structures including substantia nigra and
medulla are affected.
5. Phase 5—Aβ deposits are visible in the cerebellum.
3.6.3 Assessment The Consortium to Establish a Registry for Alzheimer’s Disease

of Neuritic Plaques (CERAD) guidelines is followed to assess neuritic plaque pathol-
ogy [14]. Tissue sections are included containing middle frontal
gyrus, superior/middle temporal gyri, and inferior parietal lobe
and are stained with Gallyas silver stain (for protocol, see Sect. 3.4).
Fig. 8 Diagram illustrating progression of amyloid β (Aβ) through the brain. Aβ plaques are first seen in phase
1 in the neocortex, in particular the inferior parietal and superior temporal gyrus (A). In phase 2, Aβ extends to
the entorhinal region, cornu ammonis (CA) sector 2 of the hippocampus and the cingulate cortex (B). Subcortical
structures are affected in phase 3 including striatum, thalamus, hypothalamus, and claustrum (C). Phase 4 is
characterized by Aβ deposits in the brainstem including the substantia nigra and medulla oblongata (D), and
finally Aβ extends to the molecular layer of the cerebellum in phase 5 (E)
Fig. 9 Photomicrographs stained with Gallyas silver stain, demonstrating semiquantitative scoring of neuritic
plaque pathology. Density of plaques per square millimeter is assessed at X100 magnification and described
as mild (scored A) (A), moderate (scored B) (B), or severe (scored C) (C). Arrows point to individual neuritic
plaques. Scale bar represents 100 μm and is also representative of B and C
Semiquantitative assessment of plaque density per square milli-

meter is determined and scored as sparse (A), moderate (B), or
frequent (C) (Fig. 9).
3.6.4 Determining Level Individual “scores” taken from Braak stages, Thal phases, and
of AD Neuropathologic CERAD assessment are combined to determine the level of AD
Change neuropathologic change using guidelines set out by the National
Fig. 10 The level of AD neuropathologic change is determined by considering all three pathologies. Amyloid β
Thal phase (A score), Braak stage (B score), and CERAD (C score) are combined to determine the level of AD
neuropathologic change as no, low, intermediate, or high. CERAD, Consortium to Establish a Registry for
Alzheimer’s disease
Table 4
Protocol to assess CAA severity
Type of CAA
assessed 0 1 2 3
Parenchymal CAA Absent Scant Aβ deposition Some circumferential Widespread circumferential
CAA CAA
Meningeal CAA Absent Scant Aβ deposition Some circumferential Widespread circumferential
CAA CAA
Capillary CAA Absent Present NA NA
Vasculopathy Absent Occasional vessel Many vessels affected NA
affected
Institute on Aging-Alzheimer’s Association (NIA-AA) [25, 26].

Amyloid β Thal phase, Braak stage, and CERAD score are com-
bined to yield an “ABC” score which is reported (regardless of
clinical diagnosis) to determine the level of AD neuropathologic
change on a four-tiered scale that is either no, low, intermediate, or
high. A summary of ABC scoring is summarized in Fig. 10.
3.7 Cerebral Amyloid Although not required for the neuropathological diagnosis of AD,
Angiopathy it is important to note the level of CAA when considering Aβ
depositions in post-mortem tissue. A recent consensus protocol has
provided a semiquantitative scoring system to assess parenchymal
CAA, meningeal CAA, capillary CAA, and vasculopathy [24].
Using immunohistochemically sections for Aβ from frontal, tem-
poral, parietal, and occipital cortices (for protocol, see Sect. 3.5),
the consensus-based protocol for assigning CAA severity is out-
lined in Table 4.
4 Conclusions
Neuropathological post-mortem assessment is the gold standard in

the diagnosis of neurodegenerative disease but has to be performed
according to standardized methodologies. By following the pub-
lished guidelines for the post-mortem diagnosis of AD neuropatho-
logic change [25], an accurate neuropathological diagnosis can be
made, and this is crucial to provide reliable information regarding
the most frequent age-associated neurodegenerative changes (i.e.,
HP-T and Aβ) seen in human brains.
5 Notes
1. Physical developer solution to be made immediately before use.

If precipitate forms during mixing, this will spoil the developer.
2. Hydrogen peroxide blocking solution should be prepared
immediately before use.
3. The use of a solvent-based hard-set mounting media will prevent
the sample from drying out and aids sample preservation.
4. DAB is a mutagenic and potential carcinogen; therefore, correct
personal protective equipment must be used and safety proce-
dures adhered to.
5. Samples to be produced in as few batches as possible to mini-
mize variability in staining.
6. In addition to extracellular Aβ plaques, 4G8 antibody also
reacts with physiological intracellular APP, and this must not
be taken into consideration when assessing Aβ pathology.
Acknowledgments
We thank Lynne Ramsay and Ros Hall for their technical expertise
and assistance in writing this chapter and Dr. Christopher Morris
and the Newcastle Brain Tissue Resource for providing the photo-
graphic images.
References
1. Halliday GM, Double KL, Macdonald V et al 3. Braak H, Braak E (1991) Neuropathological
(2003) Identifying severely atrophic cortical stageing of Alzheimer-related changes. Acta
subregions in Alzheimer's disease. Neurobiol Neuropathol 82(4):239–259
Aging 24(6):797–806 4. Braak H, Alafuzoff I, Arzberger T et al (2006)
2. Ballatore C, Lee VM, Trojanowski JQ (2007) Staging of Alzheimer disease-associated neuro-
Tau-mediated neurodegeneration in fibrillary pathology using paraffin sections and
Alzheimer's disease and related disorders. Nat immunocytochemistry. Acta Neuropathol
Rev Neurosci 8(9):663–672. https://doi. 112(4):389–404. https://doi.org/10.1007/
org/10.1038/nrn2194 s00401-006-0127-z
5. Braak H, Del Tredici K (2011) The pathologi- 17. Attems J (2005) Sporadic cerebral amyloid
cal process underlying Alzheimer's disease in angiopathy: pathology, clinical implications,
individuals under thirty. Acta Neuropathol and possible pathomechanisms. Acta
121(2):171–181. https://doi.org/10.1007/ Neuropathol 110(4):345–359. https://doi.
s00401-010-0789-4 org/10.1007/s00401-005-1074-9
6. Braak H, Thal DR, Ghebremedhin E et al 18. Attems J, Jellinger KA (2004) Only cerebral
(2011) Stages of the pathologic process in capillary amyloid angiopathy correlates with
Alzheimer disease: age categories from 1 to Alzheimer pathology--a pilot study. Acta
100 years. J Neuropathol Exp Neurol Neuropathol 107(2):83–90. https://doi.
70(11):960–969. https://doi.org/10.1097/ org/10.1007/s00401-003-0796-9
NEN.0b013e318232a379 19. Thal DR, Ghebremedhin E, Orantes M et al
7. Attems J, Thomas A, Jellinger K (2012) (2003) Vascular pathology in Alzheimer dis-
Correlations between cortical and subcortical ease: correlation of cerebral amyloid angiopa-
tau pathology. Neuropathol Appl Neurobiol thy and arteriosclerosis/lipohyalinosis with
38(6):582–590. https://doi. cognitive decline. J Neuropathol Exp Neurol
org/10.1111/j.1365-2990.2011.01244.x 62(12):1287–1301
8. Giannakopoulos P, Herrmann FR, Bussiere T 20. Pfeifer LA, White LR, Ross GW et al (2002)
et al (2003) Tangle and neuron numbers, but Cerebral amyloid angiopathy and cognitive
not amyloid load, predict cognitive status in function: the HAAS autopsy study. Neurology
Alzheimer's disease. Neurology 58(11):1629–1634
60(9):1495–1500 21. Vinters HV, Gilbert JJ (1983) Cerebral amy-
9. Glenner GG, Wong CW (1984) Alzheimer's loid angiopathy: incidence and complications
disease: initial report of the purification and in the aging brain. II. The distribution of amy-
characterization of a novel cerebrovascular loid vascular changes. Stroke 14(6):924–928
amyloid protein. Biochem Biophys Res 22. Tomonaga M (1981) Cerebral amyloid angi-
Commun 120(3):885–890 opathy in the elderly. J Am Geriatr Soc
10. Naslund J, Haroutunian V, Mohs R et al 29(4):151–157
(2000) Correlation between elevated levels of 23. Thal DR, Ghebremedhin E, Rub U et al
amyloid beta-peptide in the brain and cognitive (2002) Two types of sporadic cerebral amyloid
decline. JAMA 283(12):1571–1577 angiopathy. J Neuropathol Exp Neurol
11. Younkin SG (1995) Evidence that a beta 42 is 61(3):282–293
the real culprit in Alzheimer's disease. Ann 24. Love S, Chalmers K, Ince P et al (2014)
Neurol 37(3):287–288. https://doi. Development, appraisal, validation and imple-
org/10.1002/ana.410370303 mentation of a consensus protocol for the
12. Thal DR, Rub U, Orantes M et al (2002) assessment of cerebral amyloid angiopathy in
Phases of a beta-deposition in the human brain post-mortem brain tissue. Am J Neurodegener
and its relevance for the development of Dis 3(1):19–32
AD. Neurology 58(12):1791–1800 25. Montine T, Phelps C, Beach T et al (2012)
13. Guntert A, Dobeli H, Bohrmann B (2006) National Institute on Aging–Alzheimer’s
High sensitivity analysis of amyloid-beta pep- Association guidelines for the neuropatho-
tide composition in amyloid deposits from logic assessment of Alzheimer’s disease: a
human and PS2APP mouse brain. Neuroscience practical approach. Acta Neuropathologica
143(2):461–475. https://doi.org/10.1016/j. 123:1–11
neuroscience.2006.08.027 26. Hyman BT, Phelps CH, Beach TG et al (2012)
14. Mirra SS, Heyman A, McKeel D et al (1991) National Institute on Aging-Alzheimer's
The consortium to establish a registry for Association guidelines for the neuropathologic
Alzheimer's disease (CERAD). Part assessment of Alzheimer's disease. Alzheimer's
II. Standardization of the neuropathologic & dementia : the journal of the Alzheimer's
assessment of Alzheimer's disease. Neurology Association 8(1):1–13. https://doi.
41(4):479–486 org/10.1016/j.jalz.2011.10.007
15. Attems J, Jellinger K, Thal DR et al (2011) Review: 27. Alafuzoff I, Arzberger T, Al-Sarraj S et al
sporadic cerebral amyloid angiopathy. Neuropathol (2008) Staging of neurofibrillary pathology in
Appl Neurobiol 37(1):75–93. https://doi. Alzheimer's disease: a study of the BrainNet
org/10.1111/j.1365-2990.2010.01137.x Europe consortium. Brain pathol (Zurich,
16. Vinters HV (1987) Cerebral amyloid angiopa- Switzerland) 18(4):484–496. https://doi.
thy. A critical review. Stroke 18(2):311–324 org/10.1111/j.1750-3639.2008.00147.x
Part III
Established Markers of Pathophysiology and Risk

Chapter 6
Cerebrospinal Fluid Biomarkers of Preclinical

Alzheimer’s Disease
Panagiotis Alexopoulos and Chaido Sirinian
Abstract
The recent paradigm shift toward a more biologically oriented definition of Alzheimer’s disease (AD) in
clinical settings increases the importance of biomarkers of AD. The established cerebrospinal fluid (CSF)
biomarkers of AD, reflecting both amyloidopathy and tauopathy, are being increasingly incorporated into
both diagnostic guidelines, enabling the diagnosis of AD independently of clinical symptoms, and inclu-
sionary criteria for clinical trials. The present chapter provides an overview of the clinical utility of the three
established CSF AD biomarkers and covers both clinical and methodological issues. A brief summary is
given on relevant laboratory techniques to determine levels of AD CSF biomarkers; methodological and
clinical challenges in the field are also discussed.
Key words Cerebrospinal fluid amyloid β 1–42, Amyloid β 1–40, Total tau, Hyperphosphorylated tau,
Preclinical Alzheimer’s disease, Prognosis, Diagnosis, ELISA, Luminex xMAP, MSD multi-array
1 Introduction
Alzheimer’s disease, accounting for 60–70% of dementia cases, is

the most common cause of dementia [1, 2]. Dementia embodies a
global health challenge, the social, health care, and financial bur-
den of which will expand in the next decades. According to the
World Alzheimer Report 2016, approximately 46.8 million people
worldwide suffered from major dementia in 2015, and this num-
ber is estimated to reach 131.5 million in 2050.
In recent years a change in the way AD is defined and under-
stood has taken place [3, 4]. Traditionally the diagnosis of AD
antemortem was based on dementia symptoms and on the exclu-
sion of other potential causes of cognitive impairment, while a defi-
nite diagnosis of AD was neuropathologically established, based on
brain biopsy or autopsy findings [5, 6]. The pathology-driven AD
diagnosis is based on the detection of extracellular amyloid-β (Aβ)
plaques, mirroring amyloidopathy; of intracellular neurofibrillary
tangles, reflecting tauopathy; and of axonal degeneration, neuronal
93
94 Panagiotis Alexopoulos and Chaido Sirinian
death, and loss of synapses [2]. However, the tremendous progress

achieved in the last decades in unraveling the pathogenesis of AD
has resulted in a new, more biology underpinned AD conceptual-
ization [6, 7]. According to it, prior to its dementia phase, the
disease has a preclinical and an oligosymptomatic stage, commonly
termed in the current nomenclature mild cognitive impairment
(MCI) [8]. The oligosymptomatic stage is characterized by declin-
ing cognitive performance, mainly memory function, and mildly
affected complex activities of daily living, involving memory or
complex reasoning [9]. In the preclinical stage of AD, amyloidopa-
thy and tauopathy are present in the absence of cognitive deficits
[10]. It is of note that preclinical AD, which is present one or more
decades prior to the outburst of clinical symptoms and is supposed
to be the stage of the disease in which future causal therapies will
be most effective, can be detected in vivo through markers of AD
pathology [4, 11].
Biomarkers are diagnostic tools which detect a fundamental
feature of a pathobiological process, characterizing a specific dis-
ease, and increase diagnostic certainty [12]. Since cerebrospinal
fluid (CSF) is in direct contact with the brain and the flow of pep-
tides from and to it is restricted by the blood-CSF barrier, it is
likely that brain pathological alterations are mirrored in CSF [13].
According to the consensus report published by the Working
Group on Molecular and Biochemical Markers of Alzheimer’s
Disease, biomarkers of AD should have at least 80% sensitivity and
specificity in the detection of neuropathologically confirmed AD
cases [14]. A number of CSF peptides associated with AD pathol-
ogy have been shown to fulfill the aforementioned criteria [15].
The neurochemical signature of AD consists of the 42-amino
acid Aβ peptide (Aβ42; a marker of amyloid plaques); hyperphos-
phorylated tau (pTau), being a marker of neurofibrillary tangles;
and total tau (tTau), which is a marker of axonal damage. Aβ42 is
the most abundant Aβ isoform in amyloid plaques and is decreased
in the CSF of patients suffering from AD to approximately 50% of
control levels [16]. The marker tTau, which is not specific for AD,
embraces all isoforms of tau in CSF independently of phosphoryla-
tion state, and the increase of its concentration in patients with AD
is approximately 300% of the levels of controls. Protein tau has
numerous phosphorylation sites. The phosphorylation at either
threonine 181 or threonine 231 stands in focus of the measure-
ment methods of pTau in CSF, which unravel a 200% increase of
pTau in AD in comparison to cognitively healthy elderly individu-
als [7]. It is of note that pTau appears to be relative specific for
AD. In a number of laboratories, the 40-amino acid Aβ peptide
(Aβ40), being a common protein found in patients without the
disease, is measured, in order to build the Aβ42/Aβ40 ratio, while
in other centers this analysis is carried out only in the case of
Established CSF Markers of Alzheimer’s Disease 95
conflicting CSF biomarker findings such as increased tau proteins

and normal concentrations of Aβ42 [17–19].
CSF AD biomarkers become abnormal years prior to the devel-
opment of clinical symptoms. CSF Aβ42 levels have been shown to
begin to decline 25 years before the expected onset of symptom-
atic AD and to plateau down to fully decreased levels at least 9 years
prior to dementia onset, while tau increases approximately 15 years
prior to the expected onset of symptoms [20, 21]. Longitudinal
decreases in Aβ42 were observed in some individuals in early mid-
dle age, while the concentrations of tau proteins increased in mid
and late middle age [22]. Nonetheless, another study points that
alterations in pTau CSF levels may precede those of Aβ42 [23].
This finding is in line with postmortem evidence of the presence of
neurofibrillary tangles in the entorhinal cortex and in hippocampus-
related structures as the first neuropathologic event in AD [10].
The three established CSF biomarkers of AD are a useful
instrument for detecting AD. CSF AD biomarker abnormality
distinguishes patients suffering from dementia due to AD from
controls with a sensitivity and specificity reaching 85–90% [24].
The sensitivity and specificity of these markers in detecting
patients with MCI who will develop AD dementia range between
81–95% and 72–95%, respectively [25–27]. A recent meta-analy-
sis which included more than 28,000 elderly individuals unveiled
a high accuracy of established CSF AD biomarkers in differentiat-
ing AD dementia from controls (up to 2.5-fold changes between
patients and controls). The differentiation between cohorts with
MCI due to AD and those with stable MCI was very strong too
[15]. Nonetheless, the utility of CSF AD biomarkers to differen-
tiate AD from other dementia causes is relatively limited. Increases
in tTau are commonly observed also in other diseases, for instance
in traumatic and ischemic brain injury and in subtypes of fronto-
temporal lobar degeneration, while abnormal CSF Aβ42 levels
overlap between AD and dementia with Lewy bodies [24, 28].
Aβ42/Aβ40 ratio has been shown to have added value with
respect to Aβ42 alone having an accuracy of >80% in distinguish-
ing AD dementia from dementias caused by other neurodegen-
erative diseases [17, 29]. Longitudinal studies in cognitively
healthy individuals reported a correlation between baseline CSF
biomarkers and development of oligosymptomatic AD for Aβ42
alone or in combination with tTau or pTau [10]. Of note, in
individuals with abnormal Aβ42, the additional presence of
abnormal tau levels is associated with a faster progression to a
symptomatic stage of the disease [10]. Recently, neurochemical
biomarkers have been advocated as part of the clinical diagnostic
workup of memory clinics worldwide and are being increasingly
incorporated into both d iagnostic guidelines and inclusionary
criteria for clinical trials [4, 10, 30].
2 Materials
CSF samples are obtained by lumbar puncture in the L3/L4 or

L4/L5 intervertebral space (see Notes 1 and 2) and collected in
polypropylene tubes (see Note 3), without fasting (see Note 4). They
are centrifuged at 2000 × g for 10 min at room temperature (see
Note 5) and then split into 0.5 mL aliquots in cryotubes and imme-
diately frozen at −20 °C for a maximum of 2 months, until routine
analysis (see Note 6).
3 Methods
The majority of laboratories worldwide uses enzyme-linked immu-

nosorbent assays (ELISAs) or the bead-based xMAP platform to
determine CSF AD biomarkers. Meso Scale Discovery technology
is used by a smaller number of laboratories [31].
3.1 Enzyme-Linked Levels of Aβ40 and Αβ42 can be determined using high sensitivity
Immunosorbent Assay ELISA kits (RE59651 and RE59661) from IBL International
(ELISA) (Hamburg, Germany) [29]. Both assays are based on microtiter
plates, pre-coated with a monoclonal antibody (capture antibody)
3.1.1 Quantitative
raised against the C terminus of either the Aβ40 or Aβ42 peptides.
Determination of Aβ40
The N terminus of the Aβ40 and Αβ42 peptides is recognized by a
and Aβ42
second monoclonal antibody (detection antibody), which is conju-
gated to a horseradish peroxidase enzyme that is sequentially
detected by the addition of the chromogenic substrate tetrameth-
ylbenzidine (TMB). The concentration of both Aβ40 and Αβ42 is
proportional to the obtained optical density.
1. Add 100 μL of each Standard, Control and diluted patient
sample into the respective wells of microtiter plate in duplicate.
Cover plate with lid, and incubate for 120 min at room tem-
perature (RT) (18–25 °C) on an orbital shaker (500 rpm).
2. Aspirate each well and wash all unbound antigen five times.
3. Add 100 μL of diluted enzyme conjugate in each well. Cover
plate with lid, and incubate for 60 min at RT (18–25 °C) on an
orbital shaker (500 rpm).
4. Aspirate each well and wash five times.
5. Pipette 100 μL of TMB substrate solution into each well.
Briefly mix contents by gently shaking the plate.
6. Incubate microtiter plate for 30 min at RT (18–25 °C), and
stop the substrate reaction by adding 100 μL of TMB stop
solution into each well.
7. Measure optical density with a photometer at 450 nm.
Both assays present with high analytical specificity (low cross-

reactivity) toward their analytes, but the Aβ42 detection kit pres-
ents with a significant higher analytical sensitivity.
3.1.2 Quantitative For quantification of tTau and Tau phosphorylated at threonine

Determination of tTau 181, two commercially available ELISA kits INNOTEST™ hTAU
and pTau Ag (INX74378, Innogenetics, Gent, Belgium) and INNOTEST™
PHOSPHO-TAU(181P) (FRI30913, Fujirebio, Gent, Belgium) may
be employed. In both assays, target proteins, Tau and pTau, are
captured by a monoclonal human antibody (AT120 and HT7,
respectively) which is coated on a microtiter plate. Human CSF
samples are simultaneously combined with detection antibodies
conjugated to biotin (conjugate 1) and loaded to the appropriate
ELISA plate wells in duplicates. In the case of the tTau detection
ELISA, two human monoclonal detection abs (BT2 and HT7) are
employed, while for the detection of pTau, a phospho-specific Tau
threonine 181 monoclonal antibody is used. The antigen-antibody
complex is then recognized by peroxidase-labeled streptavidin
molecules (conjugate 2). In addition, run validation controls and
calibrators are used for the standardization of the quantification of
both tTau and pTau proteins. The procedure is briefly described in
the following.
1. Determine the number of wells (in duplicates) for controls,
calibrators, and CSF samples.
2. Add conjugate working solution 1 to each well of the antibody-
coated plate.
3. Add control, calibrator, and CSF samples to duplicate wells,
and cover the strips with an adhesive sealer, and incubate over-
night (see Note 7).
4. Aspirate each well and wash all unbound antigen five times.
5. Add conjugate working solution 2 to each well. Cover with a
new adhesive sealer, and incubate at 25 ± 2 °C (see Note 8).
6. Aspirate each well and wash five times.
7. Add substrate working solution to each well for 30 min in the
dark. Stop the reaction, and determine the optical density of
each well immediately, using a microplate reader set to 540 nm.
3.2 Multiplexed Today, single-analyte immunoassay is seeing competition from

Immunoassays newer bead-based immunoassays that employ Luminex’s xMAP
and Meso Scale Discovery’s Multi-Array technology. Current mul-
tiplexed immunoassays are based on multi-marker strategies, in a
single reaction vessel from a relatively smaller sample volume with
high efficiency.
Fig. 1 General overview of the xMAP immunoassay
3.2.1 Luminex Multi- The Luminex xMAP technology allows the measurement of mul-
Analyte Profiling (xMAP) tiple proteins simultaneously in a single sample. This technology
uses 5.6-μm microsphere beads, either polystyrene or paramag-
netic. The microsphere beads are internally dyed with precise
amounts of two or three spectrally distinct fluorescent dyes to pro-
duce up to 100–500 different sets of microspheres, depending on
the system design. Any molecule, such as proteins, antibodies,
ligands, or nucleic acids, interacting with the desired target, can be
coupled to the beads [32]. Initially, the coupled beads are incu-
bated with the sample and bind to the desire target. Subsequently
the bead-target complex is incubated with a mixture of biotinyl-
ated antibodies, which are specific against the target analyte, asso-
ciated with a fluorescent bead. A fluorescent dye (phycoerythrin,
PE) is pre-conjugated to streptavidin molecules and is used to indi-
cate the occurrence of positive binding and quantify the amount of
target analytes (Fig. 1). The multiplexing detection is performed
on a Luminex analyzer, where two lasers (a 635-nm, 10-mW red
diode laser and a 532-nm, 13-mW yttrium aluminum garnet laser)
are used to identify and quantify fluorescence from each individual
microsphere bead-target analyte-streptavidin complex. The ana-
lyzer, simultaneously, identifies the target analyte through the spe-
cific bead-fluorescence spectral address and determines its quantity
(PE-fluorescence), as the sample passes through the analyzer [33].
Innogenetics manufactures one such xMAP-based test, the
INNO-BIA AlzBio3, which determines the amount of pTau, tTau,
and Aβ42 proteins in CSF samples. According to the test, a mix of
the microsphere beads of a specific region number is captured
selectively by a first monoclonal antibody [AT270 for pTau, AT120
for tTau, 4D7A3 for Aβ42)], coupled covalently. A mix of the
microspheres is added in a specific volume to the filter plates. CSF
samples or standards are added on the filter plate, together with a
mix of biotinylated antibodies. Each biotinylated detector anti-
body detects one or several parameters [e.g., 3D6 for Aβ42, HT7
for detection of pTau and tTau]. The antigen-antibody complex is
then detected by a phycoerythrin-labeled streptavidin conjugate.

Each analyte is identified and quantified as described previously.
3.2.2 Meso Scale In the MSD assay main steps that are followed are highly similar
Discovery’s Multi-Array to ELISA assay. The plate is initially coated with a capture reagent
(MSD) before the addition of the samples and followed by the addition of
detection reagents. However, MSD platform uses a different type
of uncoated plates, which have a working electrode surface that
absorbs capture reagent within the wells. The plates, used for the
MSD assay, can be coated with a variety of capture molecules and
reagents, including antibodies, virus-like particles, cells, peptides,
and lysates. After coating, the sample and a solution containing
detection antibodies conjugated with electrochemiluminescent
labels are added to the wells. The analytes in the sample bind to
the capture reagent immobilized on the electrode surface, and this
complex recruits the detection antibodies forming a “sandwich.”
An MSD read buffer that provides the appropriate chemical envi-
ronment for electrochemiluminescence is loaded to the plate and
analyzed into an MSD imager. The imager applies a voltage to the
plate electrodes, causing the SULFO-TAG in close proximity to
the bottom of the plate to emit light through a series of reduction
and oxidation reaction. The imager measures the intensity of
emitted light to provide a quantitative measure of analytes in the
sample [34].
Based on the above, at least two possible MSD immunoassay
formats may be developed for amyloid peptides (Aβ40 and Αβ42)
and Tau protein (tTau and pTau) identification and quantification.
First, a typical format of an immunoassay that employs a specific
antibody as the capture reagent (capture antibody) that coats the
well of a plate can be developed. In a second format of the MSD
assay, instead of using a capture antibody to coat a well-plate posi-
tion, the sample (CSF sample) is directly coated in the plate well as
the capture material. Then specific antibodies are used to identify
and quantify the under-study analyte. According to recently pub-
lished data, based on correlation studies between all the above-
described immunoassays, the data obtained with the MSD platform
for AD CSF biomarkers are sparse at this time compared to the
ELISA and xMAP platforms [31].
3.3 Interpretation The interpretation of findings relies on cutoff values recommended

of CSF AD Biomarker by the manufacturers and is not affected by the patients’ age [17].
Values It is of note that the manufacturers recommend that each laboratory
establishes its own cutoff values (see Note 9), since the interlabora-
tory variability ranges between 20% and 30% even when the same
assay is used [35, 36].
If all three CSF biomarkers are abnormal, the constellation is
neurochemically compatible with AD. Constellations with all bio-
markers normal point to neurochemical incompatibility with a
diagnosis of AD. The use of “gray zones” is a common strategy for

treating results that are somehow indeterminate. The recently
published National Institute on Aging-Alzheimer’s Association
diagnostic guidelines refer to borderline biomarker results too,
even though they avoid defining the range of borderline values [6].
A proposed pragmatic approach to the range of borderline CSF
biomarker values is to define a gray zone as 10% increment of the
cutoff value in the case of tau proteins and a 10% decrement of the
cutoff value in the case of Aβ42 or Aβ42/Aβ40 ratio. The value
should be at least 10% above or below the corresponding cutoff
value in order to be treated as abnormal. In such a way the possibil-
ity of false-positive or false-negative interpretation of the findings
is minimized. In the light of the absence of causal therapies, it is
ethically acceptable to falsely negatively underdiagnose a number
of individuals suffering from preclinical AD than to falsely overdi-
agnose a number of healthy controls [17]. In cases with conflicting
biomarker findings, no diagnostic label can be given, and further
diagnostic workup is necessary (for instance, Aβ40 measurement
and/or brain imaging) (see Note 10). Additionally, the indication of
follow-up and in-depth clinical and neuropsychological character-
ization of the patient should be considered before final conclusions
are drawn. Isolated low CSF levels of Aβ42 or isolated tau pathol-
ogy indicates that the individual is at risk for AD but are not suffi-
cient for diagnosing preclinical AD [10].
3.4 Challenges The typical clinical phenotype of AD is not a straightforward con-

sequence of amyloidopathy and tauopathy, and the presence of AD
3.4.1 Clinical Challenges
pathology does not deterministically lead to symptomatic AD. As
autopsy reports underscore, a plethora of pathologies accompany
AD pathological alterations in the aging brain (for instance, cere-
brovascular alterations or Lewy body pathology) [31, 37]. Such
concurrent pathologies can synergistically lower the threshold for
the development of clinical symptoms. Such co-pathologies poten-
tiate the clinical expression of AD-associated brain alterations
which would have remained clinically silent in the absence of co-
pathologies, because they are still not sufficiently advanced to
become clinically recognizable [38]. The synergies of co-
pathologies in the aging brain are modified by the influence of
both harmful and protective environmental factors such as e xposure
to occupational hazards or cognitive, physical, and social activity
levels which could superimpose or counterbalance the effects of
the presence of brain pathologies [39]. This complex interplay fac-
tors undermine, in effect, the prognostic utility of established AD
biomarkers at the individual level, and as a result biomarkers turn
out to be less efficient in routine care in comparison to research
settings [40]. Hence, the individualized estimation of prognosis of
preclinical AD presupposes the development of novel strategies
approaching the disease course as a multivariate system, so that the
patients can benefit at the individual level from advances in bio-

marker research [3].
The limited accuracy of CSF AD biomarkers in estimating the
individual prognosis and the absence of causal therapeutic strate-
gies raise ethical issues and issues of cost-effectiveness of CSF AD
biomarker analyses. The diagnosis of an asymptomatic neurode-
generative disease with uncertain prognosis and no available ther-
apy may cause significant psychological distress. Therefore, the
diagnosis of preclinical AD has to be accompanied by appropriate
psychosocial counseling, so that the risk of unnecessary psycho-
logical distress is minimized [41]. Furthermore, in the light of the
lacking disease-modifying strategies, health payers are in a number
of countries reluctant to reimburse the full cost of CSF biomarker
analyses due to their insufficient effect on patient management,
despite their relatively low cost [40]. On the other hand, it should
be underscored that an accurate diagnosis improves the quality of
treatment and health care at the individual level. For instance, rul-
ing out AD pathology shifts the focus of the diagnostic workup on
alternative, and possible curable diseases or the diagnosis of pre-
clinical or oligosymptomatic AD may not only foster protective
behaviors (e.g., reduction of stress, changes of dietary patterns),
which could positively affect the impact of known factors potenti-
ating neurodegeneration (inflammation, oxidative stress) [3], but
also enable the initiation of the necessary psychosocial interven-
tions early enough to maximize their effect.
CSF biomarkers, especially Aβ42, have been employed in clini-
cal trials of drug candidates with disease-modifying potential. In
the light of the disappointing results of several clinical trials which
have included patients at the dementia stage of AD, it is reckoned
that the emphasis should be moved to oligosymptomatic or even
preclinical stages of the disease. Individuals with preclinical AD
would probably benefit most from interventions. In such trials
CSF biomarkers can be employed as pharmacodynamic markers
and/or as surrogate markers. For instance, changes in CSF Aβ42
concentrations in participants can be instructive in selecting dosage
for development, and a go or no-go decision can be based on
them. Nonetheless, CSF Aβ42 levels are decreased and reach a pla-
teau already a decade prior to onset of dementia symptoms [21].
Thus, it seems that the assessment of the potential curative effect
of a drug candidate cannot be based on CSF concentration changes
of a biomarker that has already reached a plateau. As a consequence
new surrogate and pharmacodynamic markers are urgently needed.
Despite having an invasive character, lumbar puncture is a
straightforward procedure with few complications. The list of con-
tradictions to lumbar puncture includes hypertension with brady-
cardia, bleeding diathesis, vertebral deformities, and infection at
the site of the lumbar puncture. The frequency of complications
such as postlumbar puncture headache, back pain, headache, and
dizziness is low especially in the elderly [42–44], so that CSF col-

lection by means of lumbar puncture is being increasingly imple-
mented in diagnostic routine of memory clinics worldwide.
3.4.2 Analytical The measured CSF biomarker concentrations differ between stud-
Challenges ies which is the consequence of a number of preanalytical, analyti-
cal, and/or assay-related factors. The source of overall variability
can be attributed to within-assay run variability (between duplicate
samples), intralaboratory longitudinal variability, between-
laboratory variability, and within- and between-assay kit lot vari-
ability [36]. The overall variability is reported to be around
20%–30%, with lower numbers for ELISA than for xMAP and
MSD measurements. Interlaboratory variability is a major contrib-
utor, while intralaboratory longitudinal variability has a smaller and
the within-run variability an even smaller effect. For some of the
analytes, a significant impact from between-lot-dependent variabil-
ity has been observed. Hence, it is important that quality of kits is
improved, in order to minimize lot-to-lot variation, being caused
by matrix effects, variation in production of different kit compo-
nents, etc. Furthermore, the development of detailed experimental
protocols is an essential step for decreasing variability [31, 35, 36].
Interestingly, the development of a novel fully automated electro-
chemiluminescence assay for the quantitation of CSF Aβ42 has
been recently reported [45]. It has a markedly improved precision
in comparison to other available assays and possibly opens the road
to overcoming the shortcoming of the high variability. In addition,
several, already underway quality control initiatives aim to contrib-
ute to standardization efforts (harmonizing laboratory practices,
defining procedures on CSF collection and handling, establishing
reference measurement procedures) [31, 35]. The development of
the first reference procedure for CSF Aβ42 was an additional
important step toward standardization [46].
The differences in mean concentrations of CSF markers
between the analytical techniques ELISA, xMAP, and MSD are a
major limitation for establishing multicentric cooperation, gold-
standard protocols, and reference values to be shared by distinct
laboratories. This difference is attributable to the lack of certified
reference materials (CRMs), which are also called standard refer-
ence materials, and calibrators for CSF biomarkers. Metrology
institutes develop CRMs. Establishing a protein CRM is a complex
task, since heterogeneities engendered by post-transnational modi-
fications or contaminations make it difficult to establish the purity
of proteins. The development of CRMs for CSF AD biomarkers
embodies a major challenge. In order to achieve it, the coordinated
efforts of researchers, industry, and metrology are necessary. If suc-
cessful, full global traceability and comparability of biomarker
results yielded by different analytical techniques and at different
laboratories will be enabled [31, 36].
4 Notes
1. CSF biomarker concentrations from nonhemorrhagic samples

are not influenced by fractionated sampling. Thus, there is no
need to recommend that a certain fraction is to be taken when
sampling CSF for AD biomarker analyses [47].
2. There is no need for obtaining CSF samples at a particular
time of day, since there is no diurnal variation with AD bio-
markers [48].
3. Glass or polystyrene tube should not be used, since Aβ pep-
tides tend to bind nonspecifically to them, leading to lower
measured concentrations of Aβ42. The degree of this phe-
nomenon is related to the tube surface area, to the duration of
contact with the tube surface, as well as to the temperature,
sample, or tube type. Tau levels are also affected but to a more
limited extent. Recent reports indicate, however, that tubes
made of pure polypropylene may cause adsorption of Aβ pep-
tides. The adsorption of Aβ onto the tube may be prevented
by pretreating CSF with a nonionic detergent such as
Tween-20 [47].
4. There is no evidence that food intake or glucose levels affect
CSF biomarker levels [47].
5. Centrifugation is only necessary in the case of visually hemor-
rhagic samples, which could correspond to a red cell count
between 500 and 1000/μL. In the case of traumatic lumbar
puncture, it is recommended that the first 1–2 mL is discarded.
The centrifugation should take place as soon as possible and
not later than 2 h after lumbar puncture, in order to avoid cell
lysis, influencing CSF AD biomarker levels [47].
6. It is of note that there is no statistically significant change in
Aβ42, pTau, and tTau concentrations irrespective of whether
the sample is frozen within 2 h or kept at room temperature for
up to 5 days before freezing [49]. Moreover, since the method
of freezing (− 20 °C or −80 °C) seems to influence tTau and
pTau levels, it is recommended that freezing at −80 °C should
be preferably carried out for storage. CSF AD biomarker con-
centrations seem to be very stable for an extended period, at
least 10 years, at −80 °C [50]. Though CSF AD biomarkers
are stable during freeze/thawing when measured with enzyme-
linked immunosorbent assay, it is recommended that the num-
ber of freeze/thaw cycles is limited to one or two, so that any
possible effects are minimized [47].
7. Overnight incubation temperatures differ between the two
assays. For the detection of tTau protein, samples are incu-
bated at 25 ± 2 °C, whereas for the detection of pTau, samples
are incubated at 2–8 °C.
8. Incubation time for tTau protein is 30 min and for pTau is

60 min.
9. The validation for clinical routine of cutoff values designed for
a particular laboratory presupposes that the chosen cutoff is
the best compromise in terms of sensitivity and specificity for
the laboratory’s specific clinical objectives, that the accuracy of
the cutoff value is periodically reassessed in larger samples
recruited during clinical routine, and that any implemented
preanalytical or analytical modifications within the framework
of standardization strategies make a new validation of previous
cutoffs necessary [17].
10. It should be borne in mind when interpreting conflicting bio-
marker findings that AD biological markers mirror a continuum
between healthy aging and dementia with typical CSF biomark-
ers constellations and that they change over time within the
framework of a continuous process lasting many years.
References
1. De-Paula VJ, Radanovic M, Diniz BS et al 7. Forlenza OV, Radanovic M, Talib LL et al
(2012) Alzheimer's disease. Subcell Biochem (2015) Cerebrospinal fluid biomarkers in
65:329–352. https://doi.org/10.1007/ Alzheimer's disease: diagnostic accuracy and
978-94-007-5416-4_14 prediction of dementia. Alzheimers Dement
2. Kocahan S, Dogan Z (2017) Mechanisms of (Amst) 1(4):455–463. https://doi.org/10.
Alzheimer's disease pathogenesis and preven- 1016/j.dadm.2015.09.003
tion: the brain, neural pathology, N-methyl-D- 8. Albert MS, DeKosky ST, Dickson D et al
aspartate receptors, tau protein and other risk (2011) The diagnosis of mild cognitive impair-
factors. Clin Psychopharmacol Neurosci ment due to Alzheimer's disease: recommen-
15(1):1–8. https://doi.org/10.9758/cpn. dations from the National Institute on
2017.15.1.1 Aging-Alzheimer's Association workgroups on
3. Alexopoulos P, Kurz A (2015) The new con- diagnostic guidelines for Alzheimer's disease.
ceptualization of Alzheimer's disease under the Alzheimers Dement 7(3):270–279. https://
microscope of influential definitions of disease. doi.org/10.1016/j.jalz.2011.03.008
Psychopathology 48(6):359–367. https:// 9. Perneczky R, Pohl C, Sorg C et al (2006)
doi.org/10.1159/000441327 Complex activities of daily living in mild cogni-
4. McKhann GM (2011) Changing concepts of tive impairment: conceptual and diagnostic
Alzheimer disease. JAMA 305(23):2458–2459. issues. Age Ageing 35(3):240–245. https://
https://doi.org/10.1001/jama.2011.810 doi.org/10.1093/ageing/afj054
5. McKhann G, Drachman D, Folstein M et al 10. Dubois B, Hampel H, Feldman HH et al
(1984) Clinical diagnosis of Alzheimer's disease: (2016) Preclinical Alzheimer's disease: defini-
report of the NINCDS-ADRDA work group tion, natural history, and diagnostic criteria.
under the auspices of Department of Health and Alzheimers Dement 12(3):292–323. https://
Human Services Task Force on Alzheimer's dis- doi.org/10.1016/j.jalz.2016.02.002
ease. Neurology 34(7):939–944 11. Jack CR, Knopman DS, Jagust WJ et al (2013)
6. McKhann GM, Knopman DS, Chertkow H Tracking pathophysiological processes in
et al (2011) The diagnosis of dementia due to Alzheimer's disease: an updated hypothetical
Alzheimer's disease: recommendations from model of dynamic biomarkers. The Lancet
the National Institute on Aging- Alzheimer's Neurology 12(2):207–216. https://doi.
Association workgroups on diagnostic guide- org/10.1016/S1474-4422(12)70291-0
lines for Alzheimer's disease. Alzheimers 12. Martins-de-Souza D (2010) Is the word 'bio-
Dement 7(3):263–269. https://doi. marker' being properly used by proteomics
org/10.1016/j.jalz.2011.03.005 research in neuroscience? Eur Arch Psychiatry
Clin Neurosci 260(7):561–562. https://doi. https://doi.org/10.1001/jamaneurol.2015.

org/10.1007/s00406-010-0105-2 1285
13. Engelborghs S (2013) Clinical indications for 23. Braak H, Zetterberg H, Del Tredici K et al
analysis of Alzheimer's disease CSF biomarkers. (2013) Intraneuronal tau aggregation precedes
Rev Neurol (Paris) 169(10):709–714. https:// diffuse plaque deposition, but amyloid- beta
doi.org/10.1016/j.neurol.2013.07.024 changes occur before increases of tau in cere-
14. The Ronald and Nancy Reagan Research brospinal fluid. Acta Neuropathol 126(5):631–
Institute of the Alzheimer’s Association and 641. https://doi.org/10.1007/s00401-013-
the National Institute on Aging Working 1139-0
Group (1998) Consensus report of the 24. Teunissen CE, Willemse E (2013)
Working Group on: “Molecular and Cerebrospinal fluid biomarkers for Alzheimer's
Biochemical Markers of Alzheimer’s Disease”. disease: emergence of the solution to an impor-
Neurobiol Aging 19(2):109–116 tant unmet need. EJIFCC 24(3):97–104
15. Olsson B, Lautner R, Andreasson U et al 25. Hansson O, Zetterberg H, Buchhave P et al
(2016) CSF and blood biomarkers for the (2006) Association between CSF biomarkers
diagnosis of Alzheimer's disease: a systematic and incipient Alzheimer's disease in patients
review and meta-analysis. The Lancet with mild cognitive impairment: a follow-up
Neurology 15(7):673–684. https://doi. study. Lancet Neurol 5(3):228–234. https://
org/10.1016/S1474-4422(16)00070-3 doi.org/10.1016/S1474-4422(06)70355-6
16. Blennow K, Zetterberg H, Fagan AM (2012) 26. Parnetti L, Chiasserini D, Eusebi P et al (2012)
Fluid biomarkers in Alzheimer disease. Cold Performance of abeta1-40, abeta1-42, total
Spring Harb Perspect Med 2(9):a006221. tau, and phosphorylated tau as predictors of
https://doi.org/10.1101/cshperspect. dementia in a cohort of patients with mild cog-
a006221 nitive impairment. J Alzheimers Dis 29(1):229–
17. Molinuevo JL, Blennow K, Dubois B et al 238. https://doi.org/10.3233/JAD-2011-
(2014) The clinical use of cerebrospinal fluid 111349
biomarker testing for Alzheimer's disease diag- 27. Mattsson N, Zetterberg H, Hansson O et al
nosis: a consensus paper from the Alzheimer's (2009) CSF biomarkers and incipient
biomarkers standardization initiative. Alzheimer disease in patients with mild cogni-
Alzheimers Dement 10(6):808–817. https:// tive impairment. JAMA 302(4):385–393.
doi.org/10.1016/j.jalz.2014.03.003 https://doi.org/10.1001/jama.2009.1064
18. Lewczuk P, Esselmann H, Otto M et al (2004) 28. Riedl L, Mackenzie IR, Forstl H et al (2014)
Neurochemical diagnosis of Alzheimer's Frontotemporal lobar degeneration: current
dementia by CSF Abeta42, Abeta42/Abeta40 perspectives. Neuropsychiatr Dis Treat
ratio and total tau. Neurobiol Aging 10:297–310. https://doi.org/10.2147/
25(3):273–281. https://doi.org/10.1016/ NDT.S38706
S0197-4580(03)00086-1 29. Alexopoulos P, Eisele T (2014) Efficiency of
19. Bibl M, Mollenhauer B, Esselmann H et al the ratio of cerebrospinal fluid Aβ42/Aβ40
(2006) CSF diagnosis of Alzheimer's disease concentrations in detecting Alzheimer's dis-
and dementia with Lewy bodies. J Neural ease: a step forwards. Alzheimers Dement
Transm (Vienna) 113(11):1771–1778. https: 10(4):P799
//doi.org/10.1007/s00702-006-0537-z 30. Dubois B, Feldman HH, Jacova C et al (2014)
20. Bateman RJ, Xiong C, Benzinger TLS et al Advancing research diagnostic criteria for
(2012) Clinical and biomarker changes in Alzheimer's disease: the IWG-2 criteria. Lancet
dominantly inherited Alzheimer's disease. N Neurol 13(6):614–629. https://doi.
Engl J Med 367(9):795–804. https://doi. org/10.1016/S1474-4422(14)70090-0
org/10.1056/NEJMoa1202753 31. Kang J-H, Korecka M, Toledo JB et al (2013)
21. Stomrud E, Minthon L, Zetterberg H et al Clinical utility and analytical challenges in mea-
(2015) Longitudinal cerebrospinal fluid bio- surement of cerebrospinal fluid amyloid-β1–42
marker measurements in preclinical sporadic and τ proteins as Alzheimer disease biomark-
Alzheimer's disease: a prospective 9-year study. ers. Clin Chem 59(6):903–916. https://doi.
Alzheimers Dement (Amst) 1(4):403–411. org/10.1373/clinchem.2013.202937
https://doi.org/10.1016/j.dadm.2015.09.002 32. Vignali DA (2000) Multiplexed particle-based
22. Sutphen CL, Jasielec MS, Shah AR et al (2015) flow cytometric assays. J Immunol Methods
Longitudinal cerebrospinal fluid biomarker 243(1–2):243–255
changes in preclinical Alzheimer disease during 33. Dunbar SA and Li D (2010) Introduction to
middle age. JAMA Neurol 72(9):1029–1042. Luminex® xMAP® Technology and Applications
for Biological Analysis in China. Asia Pac Biotech 43. Blennow K, Wallin A, Hager O (1993) Low
14(10):26–30 frequency of post-lumbar puncture headache
34. Meso Scale Discovery®. MSD sector and in demented patients. Acta Neurol Scand
quickplex plates, 18075-v2-2014 Jul. www. 88(3):221–223
mesoscale.com 44. Zetterberg H, Tullhog K, Hansson O et al
35. Mattsson N, Andreasson U, Persson S et al (2010) Low incidence of post-lumbar punc-
(2013) CSF biomarker variability in the ture headache in 1,089 consecutive memory
Alzheimer's Association quality control pro- clinic patients. Eur Neurol 63(6):326–330.
gram. Alzheimers Dement 9(3):251–261. https://doi.org/10.1159/000311703
https://doi.org/10.1016/j.jalz.2013.01.010 45. Bittner T, Zetterberg H, Teunissen CE et al
36. Mattsson N, Andreasson U, Persson S et al (2016) Technical performance of a novel, fully
(2011) The Alzheimer’s Association external automated electrochemiluminescence immu-
quality control program for cerebrospinal fluid noassay for the quantitation of beta-amyloid
biomarkers. Alzheimers Dement 7(4):386– (1-42) in human cerebrospinal fluid.
395.e6. https://doi.org/10.1016/j.jalz. Alzheimers Dement 12(5):517–526. https://
2011.05.2243 doi.org/10.1016/j.jalz.2015.09.009
37. Lam B, Masellis M, Freedman M et al (2013) 46. Mattsson N, Zegers I, Andreasson U et al
Clinical, imaging, and pathological heteroge- (2012) Reference measurement procedures for
neity of the Alzheimer's disease syndrome. Alzheimer's disease cerebrospinal fluid bio-
Alzheimers Res Ther 5(1):1. https://doi. markers: definitions and approaches with focus
org/10.1186/alzrt155 on amyloid beta42. Biomark Med 6(4):409–
38. Alexopoulos P, Roesler J, Thierjung N et al 417. https://doi.org/10.2217/bmm.12.39
(2015) Mapping CSF biomarker profiles onto 47. Vanderstichele H, Bibl M, Engelborghs S et al
NIA-AA guidelines for Alzheimer's disease. Eur (2012) Standardization of preanalytical aspects
Arch Psychiatry Clin Neurosci 266(7):587–597. of cerebrospinal fluid biomarker testing for
https://doi.org/10.1007/s00406-015-0628-7 Alzheimer's disease diagnosis: a consensus
39. Perneczky R, Alexopoulos P, Kurz A (2014) paper from the Alzheimer's biomarkers stan-
Soluble amyloid precursor proteins and secre- dardization initiative. Alzheimers Dement
tases as Alzheimer's disease biomarkers. Trends 8(1):65–73. https://doi.org/10.1016/j.jalz.
Mol Med 20(1):8–15 2011.07.004
40. Maier W (2016) Searching biomarkers for 48. Cicognola C, Chiasserini D, Eusebi P et al
mental disorders-lessons from Alzheimer’s dis- (2016) No diurnal variation of classical and
ease. Eur Arch Psychiatry Clin Neurosci candidate biomarkers of Alzheimer's disease in
266(7):583–585. https://doi.org/10.1007/ CSF. Mol Neurodegener 11(1):65. https://
s00406-016-0732-3 doi.org/10.1186/s13024-016-0130-3
41. Perneczky R, Guo L-H (2016) Plasma pro- 49. Zimmermann R, Lelental N, Ganslandt O et al
teomics biomarkers in Alzheimer’s disease: lat- (2011) Preanalytical sample handling and sam-
est advances and challenges. Methods Mol Biol ple stability testing for the neurochemical
1303:521–529. https://doi.org/10.1007/ dementia diagnostics. J Alzheimers Dis
978-1-4939-2627-5_32 25(4):739–745. https://doi.org/10.3233/
JAD-2011-110212
42. Duits FH, Martinez-Lage P, Paquet C et al
(2016) Performance and complications of lum- 50. Bjerke M, Portelius E, Minthon L et al (2010)
bar puncture in memory clinics: results of the Confounding factors influencing amyloid Beta
multicenter lumbar puncture feasibility study. concentration in cerebrospinal fluid. Int
Alzheimers Dement 12(2):154–163. https:// J Alzheimers Dis 2010. https://doi.
doi.org/10.1016/j.jalz.2015.08.003 org/10.4061/2010/986310
Chapter 7
Brain Structural Imaging in Alzheimer’s Disease

Sven Haller, Davide Zanchi, Cristelle Rodriguez,
and Panteleimon Giannakopoulos
Abstract
Central nervous system diseases are usually associated with significant modifications in brain morphometry
and function. These alterations may be subtle, in particular at early stages of the disease progress, and thus
not evident by visual inspection alone in magnetic resonance imaging. Group-level statistical comparisons
have dominated neuroimaging studies for many years, leading to better insight into the patterns of regional
vulnerability in brain neurodegenerative pathologies. However, such group-level results have no diagnostic
value at the individual level. Recently, pattern recognition approaches using multivariate analyses have led
to a fundamental shift in paradigm, aiming to predict the cognitive fate of each individual on the basis of
MRI-based algorithms of structural parameters. We review here the state-of-the-art fundamentals of pat-
tern recognition including feature selection, cross validation, and classification techniques and discuss limi-
tations including interindividual variation in normal brain anatomy and neurocognitive reserve. We
conclude with a special reference to future trends including multimodal pattern recognition and multi-
center approaches with data sharing and cloud computing.
Key words MRI, Pattern recognition, Support vector machines (SVMs), Multivariate, Predictive
modeling
1 Introduction
Most central nervous system (CNS) disorders cause systematic

modifications in brain structure that can be imaged using MRI.
These modifications are usually subtle in early stages of the disease
and cannot be identified by visual inspection alone. Initially, whole-
brain morphometry from structural MRI has been used to train
models that can discriminate between healthy controls and patients,
such as AD and frontotemporal dementia [1–4]. Later on, a series
of structural and functional MRI parameters such as cortical thin-
ning and arterial spin labeling have been proposed as possible pre-
dictors of cognitive decline in preclinical AD [5, 6]. While such
group-level results are relevant from a research viewpoint, they do
not automatically translate into diagnostic procedures at the
individual level. In fact, many CNS disorders are characterized by
107
108 Sven Haller et al.
diffuse rather than focal changes [7], increasing the number of

MRI variables to consider in models of prediction. Taking into
account the interindividual responsibility and number of structural
imaging variables, univariate techniques are frequently insufficient
in this respect. Tools from pattern recognition—commonly
referred to as multivariate pattern analysis (MVPA)—have the abil-
ity to integrate information across multiple variables. Exploiting
multivariate data structure can significantly improve sensitivity, in
particular when only subtle alterations occur.
The interest of machine learning techniques in the processing
of neuroimaging data has been first cited by Haxby and colleagues
[8] where they explicitly recognized the distributed nature of acti-
vation patterns from fMRI in the visual cortex. Until then, most
fMRI analyses were performed using mass univariate techniques
(i.e., voxel by voxel), which did not exploit inter-voxel dependen-
cies. While regional patterns can be identified by multivariate sta-
tistical techniques (e.g., partial least squares [9]), there was a
growing interest to employ them using tools from machine learn-
ing in order to predict mental states [10, 11]. During the late
2000, several contributions demonstrated that it is possible to train
data-driven models that can subsequently decode information
from brain images including semantic meaning of words [12],
emotional prosody pronounced by actors [13], or more recently
visual imagery [14] and even dreams [15].
The following chapter will summarize the way to proceed
focusing on the step-by-step approach that can be applied in struc-
tural MRI data.
2 Methods
2.1 Data Processing Conventional confirmatory analysis is based on a (predefined) gen-

Pipeline erative model that is fitted to the data. Statistical hypothesis testing
then provides forward inference on how well the model explains
the data. Many group studies use such schemes to identify signifi-
cant differences between populations. Pattern recognition tools
reverse the direction of inference; i.e., the model is constructed by
the data during a training phase in order to predict the explanatory
variable. The model performance is then validated during a valida-
tion phase. The models depend on the type of classifier, but they
are usually flexible. The task of a classifier refers to the predicted
variable; e.g., healthy versus pre-MCI. The data available to the
classifier, the definition of the task, and the performance of the
classifiers are three essential ingredients for the correct interpreta-
tion of the results (see Fig. 1).
As the first step, feature extraction converts the raw data into
the best possible form that maximizes the amount of information
and minimizes the effect of confounders due to various sources of
Structural Imaging in AD 109
Fig. 1 Schematic illustration of the data processing pipeline. Several types of data input such as T1, diffusion-
weighted imaging, or fMRI are in a first step preprocessed depending on the specific demands of spatial nor-
malization into a standard space, including field inhomogeneity correction, spatial (and temporal) smoothing,
atlasing, independent component decomposition, structural or functional connectivity analyses, and others. In
the next step, the preprocessed data is used for feature selection and classification
noise. Structural features can be extracted from imaging data after

a time-consuming procedure (e.g., spatial normalization to bring
the data in the same “brain space”). For instance, a typical struc-
tural MRI dataset can easily contain more than 100,000 voxels. In
most cases, the features are related (similarity of adjacent voxels or
voxels in homologous regions), and only a limited number of brain
regions (and consequently features) will carry true discriminative
information. Other possibilities of features include those extracted
from functional imaging (e.g., contrast of an experimental condi-
tion in a conventional activation study or functional connectivity
from resting-state fMRI), diffusion-weighted imaging (e.g., mea-
sures of structural connectivity), and neuropsychological and clini-
cal measures. These features can be combined to improve the
performance of the models of prediction.
2.2 Feature Selection In most cases, the number of features (e.g., voxels in the case of
structural MRI) is much larger than the number of individuals.
The selection of the best features is a long-standing problem in
machine learning that can be addressed either explicitly (by a sepa-
rate feature selection step) or implicitly (by regularization in the
classification method). Conventional feature selection reduces the

dimensionality of the feature space by ranking them according to
their univariate statistical significance in case-control comparisons.
An arbitrary number of features can be retained in multivariate
models. Another approach is to apply multivariate techniques that
project the feature space onto a (linear) subspace; e.g., principal
component analysis can be used to identify those feature dimen-
sions that explain most of the observed variance. Overall, these
features are usually not optimal for prediction (i.e., since only the
best predictive feature would be selected) [16]. For instance, it is
better to select a whole-brain region as predictive for a given task,
instead of a few voxels that might not be very stable due to noise
or slight variations in brain morphometry or the data. Therefore,
features are also sometimes transformed. The spatial wavelet trans-
form leads to a representation that is more compact for piecewise
smooth signals, a property that is often used to improve detect-
ability in MRI studies [17].
2.3 Classification Support vector machines (SVMs) have been widely applied to neu-
roimaging data, mostly because of their robustness against outliers,
but they are by far not the only choice available [18]. Typically,
SVMs are used for binary classification such as discriminating
patients from healthy controls. Recently, more advanced classifiers
have been developed (e.g., decision forests [19]) and should be
considered for multi-class classification. However, there is some
trade-off between increasing the complexity of the classifier and
improving the feature extraction. Usually, good features allow for
better classification with almost any classifier.
Another issue with classifiers arises when features are multimodal;
i.e., combining different types of imaging data with other measures
such as scores from neuropsychological tests or clinical parameters.
As the model of a single classifier will not allow the necessary flexibil-
ity, “ensemble learning” is a rich field in machine learning that deals
with combining multiple classifiers (and thus multiple models) to
integrate the richness of the data structure. One promising approach
in the future is the use of multilevel classifier algorithms to aggregate
information in a hierarchical way (see Fig. 2). This approach makes it
possible to take into account previously established knowledge about
the relative weight of each classifier.
Finally it is also important to mention the recent breakthrough
in artificial intelligence applied to neuroimaging classification: deep
learning [20]. This technique has been successfully applied to mul-
tiple fields including visual recognition, mostly with full supervi-
sion. A typical deep-learning architecture for visual recognition
builds upon convolutional neural network. Given large-scale
training data and the power of high-performance computational
infrastructure, deep learning has achieved tremendous improve-
ment in visual recognition with thousands of categories [21].
Fig. 2 Simplified processing pipeline for image classification. Processing pipeline for a single domain classification
(a) from data input to feature selection/classification to output. In multimodal classification (b–d), all input data can
directly enter one single feature selection/classification (b). Alternatively, it is possible to have one feature selection/
classification per input data, followed by one single super-feature selection/classification at the second level (c).
Moreover, it is possible to introduce an additional feature selection/classification levels, for example, regrouping all
imaging modalities, followed by a super-feature selection/classification at the third level (d)
Its applications nowadays are moving toward feature representa-

tion for AD/MCI classification [22].
2.4 Cross Validation A key step to consider in multimodal prediction is the distinction
between training and validation of each classifier. This later phase
concerns separate parts of the data according to a cross validation
methodology. As already mentioned, the model of classification is
built using part of the data where both the features and the pre-
dicted variable are given. Then, the classifier is evaluated on the
remaining part of the data by comparing its outcome against the
ground-truth prediction. The procedure is repeated after removing
different parts of the data to produce an average performance across
the validation folds (e.g., in terms of specificity and sensitivity).
Ideally, the classifier should predict the main data structure and
ignore the noise, and its model should be applicable to unseen data.
3 Challenges, Limitations, and Future Directions
We review here the fundamental challenges, potential pitfalls and

limitations as well as future trends that deserve consideration, in
particular with respect to clinical applications of pattern recogni-
tion analyses.
3.1 Normal Some structural MRI variables display a substantial interindividual

Anatomical variability within the normal range. As an example, a 15% variabil-
Interindividual ity of cortical thickness has been reported for the first time 20 years
Variability and Spatial ago in healthy controls [23]. At-risk mental state subjects could
Neighborhood thus be discriminated from healthy controls only when considering
Relationships the within-subject cortical asymmetry and not only the direct
assessment of cortical thickness between subjects [24]. Moreover,
adjacent voxels are more likely to carry similar information than
distant voxels in non-related areas. It is also notable that spatial
information can be integrated at several scales; e.g., across neigh-
boring voxels, subareas of predefined anatomical structures, or
even areas distributed over larger distances based on prior knowl-
edge (e.g., [25]).
Current classification analyses typically consider each voxel as
an independent feature ignoring the highly ordered structure of
the brain. In other words, SVMs are invariant to permutation of
the feature dimensions. To acknowledge spatial neighborhood
relationships, one must include spatial transformations in the fea-
ture extraction step (e.g., spatial averaging or wavelet transforms as
mentioned above). This could potentially improve classification
accuracy and robustness.
3.2 Incorporating There is a substantial body of evidences accumulated over the past
Prior Knowledge 30 years on the structure/function relationship in a wide range of
different diseases. This extensive prior knowledge is largely ignored
in recent individual level pattern recognition analyses, although it
might potentially improve classification accuracy and robustness.
Similar to the discussion about the anatomic structure of the
brain above, it is likely beneficial to inject any available domain
knowledge to improve the information content of the features
despite the fact that feature selection and state-of-the-art classifiers
are designed to deal with high-dimensional learning. From this
viewpoint, well-documented case-control studies with a priori
hypotheses are needed to guide and enrich the technical perfor-
mance of multilevel classifiers. Purely data-driven methods could
lead to biologically irrelevant observations in dementia prediction.
In addition, the interpretation of the results could often become
easier when features represent domain-relevant information.
Modern methods have recently been proposed to exploit spatial
structure; e.g., based on hierarchical clustering to regroup similar
voxels and reinforce the robustness [26]. For some applications,
features can also be extracted from specific regions of interest
instead of the whole brain, or the “locality” of information can be
probed by a “searchlight” approach where classification perfor-
mance is reported in an information map for features extracted
from local neighborhoods [27].
3.3 The Cognitive Pattern recognition analyses at the individual level (as well as uni-
Reserve variate group-level analyses) are based on the assumption that
there is a direct relationship between brain pathology and symp-
tomatology. In the example of cognitive decline, the assumption is
that decreasing cognitive functions are paralleled by progressive
brain atrophy. This assumption is, however, only rarely true. Due
to sociodemographic factors such as education and social integra-
tion, some individuals are able to compensate clinically the struc-
tural MRI changes even for one or two decades. The example of
AD is illustrative in this respect. The appearance of various biologi-
cal changes (including hippocampal volume loss or tau-related cor-
tical thinning) suggestive of AD pathology in controls has been
extensively documented. They may precede clinically overt demen-
tia by many years, or even decades, thus defining a temporally wide
preclinical phase of the disease [28]. This interindividual variation
in the neurocognitive reserve was described already in 1968 [29].
In the domain of pattern recognition, the same amount of struc-
tural brain alterations could be associated with clinically overt cog-
nitive decline only if the neurocognitive reserve is exhausted. This
interindividual variability in the neurocognitive reserve represents
a fundamental limitation for pattern recognition approaches espe-
cially in a clinical setting.
3.4 Disbalance As already mentioned, a typical feature set extracted from MRI
Between Number data can easily contain more than 100,000 features, which clearly
of Features exceeds the number of individuals in single-center studies. In most
and Number cases, the features are related (similarity of adjacent or homologous
of Subjects: The Risk voxels), and only a limited number will carry discriminative
of Over-Fitting information.
While the cross validation methodology is essential to train and
evaluate classifiers, it still has the risk of over-fitting the data as the
parameters can be tuned. Consequently, several nested levels of
cross validation should be used, as well as separate independent
test datasets to guarantee the quality of the multivariate recogni-
tion patterns. Often the sample size for single-center studies is
insufficient to estimate such real-world performance. In addition,
parameter tuning for both feature selection and classifiers requires
an additional inner cross validation loop that decreases the amount
of available data for learning and increase the risk of over-fitting.
In summary, selecting and determining the importance of fea-
tures is an essential processing step in classification analyses, yet
identifying the related parameters such as optimal number of fea-
tures or regularization tuning is not a trivial practice. This is still an
active field of research in machine learning; e.g., stability selection,
which refers to the consistency of features when subsampling the
feature space, including applications to neuroimaging [30].
3.5 Variability Related Additional potentially confounding factors include scanner hetero-
to Patient Selection, geneity [31, 32], variability in data preprocessing, and patient
Inter-Scanner selection.
Variability, and Data
Preprocessing
4 Conclusions
4.1 Toward New The simple classification of healthy subjects versus clinically overt
Biomarkers AD or even MCI is of marginal importance for clinical practice.
Modern AD research focuses on the prediction of cognitive decline
both in MCI and preclinical AD cases. The identification of predic-
tive biomarkers in AD research is one of the more promised and
challenging domains in human neurobiology. Although there is a
wide agreement about the need to identify in situ AD at the pre-
clinical stages and if possible long before the emergence of MCI,
passing from group differences to individual prediction is quite
difficult in the light of the impressive interindividual variability of
most biomarkers in cognitively intact cases that imply the need for
including a substantial number of cases in research cohorts, the
uncertainty regarding the cognitive fate of these cases, the diffi-
culty to propose curative treatments (at the level of primary pre-
vention) in the absence of any alert sign.
The definitions of MCI have substantially evolved and changed
over the past years, which goes beyond the scope of this review.
Depending on the MCI subtype, only about half of MCI subjects
will progress to clinically overt dementia, whereas the other half
may remain stable or evolve to other forms of dementia [33–35].
Assuming that only about half of unselected MCI individuals will
progress to clinically overt dementia, the prediction of individuals
at risk for consecutive cognitive decline is of paramount impor-
tance for early individual treatment as well as for clinical trials. In a
typical placebo-controlled pharmaceutical trial, 25% of unselected
MCI will remain stable despite being in the placebo group, while
only 25% of individuals will progress and obtain the active com-
pound. Therefore, preselection of at-risk individuals for future cog-
nitive decline would substantially improve the design of clinical
trials. As MRI is routinely performed in the clinical workup of cog-
nitive decline, advanced data analysis techniques as those by pat-
tern recognition tools make use of already existing data, which is
thus cost effective and without additional discomfort for the
patient. It is possible to predict future cognitive decline in MCI
using baseline MRI based on gray matter voxel-based morphome-
try (VBM) [1, 36, 37], white matter DTI [38, 39], or iron deposi-
tion [38]. It is further possible to combine diffusion tensor imaging
(DTI) and resting-state fMRI to identify MCI individuals [40] and
to predict MCI to AD conversion using multimodal measures also
in combination with neuropsychological scores or cerebrospinal

fluid biomarkers [26, 41] or by the combination of structural MRI
and FDG-PET [42–44]. Furthermore, it is possible to classify MCI
subtypes, who have different risk of disease progression and who
might benefit from different types of treatment based on DTI
[45]. One should take into account that while Alzheimer disease
lesions and vascular pathology are commonly co-occurring, they
are generally considered to be separate phenomena [46]. Recently,
neuroimaging findings have increasingly shown a possible interac-
tion between pathophysiological mechanisms in AD and vascular
dementia [47–49]. Along this line of research, the detection on
MR images of white matter hyperintensities, which reflects vascular
damage, can be used to predict future cognitive decline and diag-
nosis of AD [50] and represents an early and independent predic-
tor of AD risk [46].
In summary, the application of techniques from the field of
pattern recognition to neuroimaging data is an emerging field.
These methods have a number of attractive features, including the
use of multivariate information and the possibility to predict for
previously unseen data. Ongoing research is still needed to over-
come a number of limitations, including optimal feature selection
that incorporates better domain knowledge and integration of
multimodal measurements. In addition, future methodological
developments should be increasingly based on large datasets and
multicentric studies to increase both reproducibility and predict-
ability. Recent data sharing initiatives such as ADNI [51], in
combination with cloud-computing power, will provide the neces-
sary prerequisites for these developments.
References
1. Fan Y, Batmanghelich N, Clark CM et al via high-dimensional pattern classification of
(2008) Spatial patterns of brain atrophy in MRI. Neuroimage 41:1220–1227. https://
MCI patients, identified via high-dimensional doi.org/10.1016/j.neuroimage.2008.03.050
pattern classification, predict subsequent cog- 5. Xekardaki A, Rodriguez C, Montandon M-L
nitive decline. Neuroimage 39:1731–1743. et al (2015) Arterial spin labeling may contrib-
https://doi.org/10.1016/j. ute to the prediction of cognitive deterioration
neuroimage.2007.10.031 in healthy elderly individuals. Radiology
2. Fan Y, Resnick SM, Wu X, Davatzikos C 274:490–499. https://doi.org/10.1148/
(2008) Structural and functional biomarkers of radiol.14140680
prodromal Alzheimer’s disease: a high- 6. Wang L, Benzinger TL, Hassenstab J et al
dimensional pattern classification study. (2015) Spatially distinct atrophy is linked to
Neuroimage 41:277–285. https://doi. β-amyloid and tau in preclinical Alzheimer dis-
org/10.1016/j.neuroimage.2008.02.043 ease. Neurology 84:1254–1260. https://doi.
3. Klöppel S, Stonnington CM, Chu C et al (2008) org/10.1212/WNL.0000000000001401
Automatic classification of MR scans in 7. Seeley WW, Crawford RK, Zhou J et al (2009)
Alzheimer’s disease. Brain J Neurol 131:681– Neurodegenerative diseases target large-scale
689. https://doi.org/10.1093/brain/awm319 human brain networks. Neuron 62:42–52.
4. Davatzikos C, Resnick SM, Wu X et al (2008) https://doi.org/10.1016/j.neuron.2009.
Individual patient diagnosis of AD and FTD 03.024
8. Haxby JV, Gobbini MI, Furey ML et al (2001) 20. Wu J, Yu Y, Huang C, Yu K (2015) Deep multiple
Distributed and overlapping representations of instance learning for image classification and
faces and objects in ventral temporal cortex. auto-annotation. IEEE, pp 3460–3469
Science 293:2425–2430. https://doi. 21. van der Burgh HK, Schmidt R, Westeneng H-J
org/10.1126/science.1063736 et al (2017) Deep learning predictions of sur-
9. Krishnan A, Williams LJ, McIntosh AR, Abdi vival based on MRI in amyotrophic lateral
H (2011) Partial Least Squares (PLS) methods sclerosis. Neuroimage Clin 13:361–369.

for neuroimaging: a tutorial and review. https://doi.org/10.1016/j.nicl.2016.10.008
Neuroimage 56:455–475. https://doi. 22. Suk H-I, Shen D (2013) Deep learning-based
org/10.1016/j.neuroimage.2010.07.034 feature representation for AD/MCI classifica-
10. Cox DD, Savoy RL (2003) Functional mag- tion. Med Image Comput Comput Assist
netic resonance imaging (fMRI) “brain read- Interv 16:583–590
ing”: detecting and classifying distributed 23. Haug H (1987) Brain sizes, surfaces, and neu-
patterns of fMRI activity in human visual cor- ronal sizes of the cortex cerebri: a stereological
tex. Neuroimage 19:261–270. https://doi. investigation of man and his variability and a
org/10.1016/S1053-8119(03)00049-1 comparison with some mammals (primates,
11. Pereira F, Mitchell T, Botvinick M (2009) whales, marsupials, insectivores, and one ele-
Machine learning classifiers and fMRI: a tuto- phant). Am J Anat 180:126–142. https://doi.
rial overview. Neuroimage 45:S199–S209. org/10.1002/aja.1001800203
https://doi.org/10.1016/j. 24. Haller S, Borgwardt SJ, Schindler C et al
neuroimage.2008.11.007 (2009) Can cortical thickness asymmetry anal-
12. Mitchell TM, Shinkareva SV, Carlson A et al ysis contribute to detection of at-risk mental
(2008) Predicting human brain activity associ- state and first-episode psychosis? A pilot study.
ated with the meanings of nouns. Science Radiology 250:212–221. https://doi.
320:1191–1195. https://doi.org/10.1126/ org/10.1148/radiol.2501072153
science.1152876 25. Hackmack K, Weygandt M, Wuerfel J et al
13. Ethofer T, Van De Ville D, Scherer K, (2012) Can we overcome the “clinico-
Vuilleumier P (2009) Decoding of emotional radiological paradox” in multiple sclerosis?
information in voice-sensitive cortices. Curr J Neurol 259:2151–2160. https://doi.
Biol 19:1028–1033. https://doi. org/10.1007/s00415-012-6475-9
org/10.1016/j.cub.2009.04.054 26. Cui Y, Sachdev PS, Lipnicki DM et al (2012)
14. Nishimoto S, Vu AT, Naselaris T et al (2011) Predicting the development of mild cognitive
Reconstructing visual experiences from brain impairment: a new use of pattern recognition.
activity evoked by natural movies. Curr Biol Neuroimage 60:894–901. https://doi.
21:1641–1646. https://doi.org/10.1016/j. org/10.1016/j.neuroimage.2012.01.084
cub.2011.08.031 27. Kriegeskorte N, Goebel R, Bandettini P (2006)
15. Horikawa T, Tamaki M, Miyawaki Y, Kamitani Information-based functional brain mapping.
Y (2013) Neural decoding of visual imagery Proc Natl Acad Sci U S A 103:3863–3868.
during sleep. Science 340:639–642. https:// https://doi.org/10.1073/pnas.0600244103
doi.org/10.1126/science.1234330 28. Lazarczyk MJ, Hof PR, Bouras C,
16. Tolosi L, Lengauer T (2011) Classification Giannakopoulos P (2012) Preclinical
with correlated features: unreliability of feature Alzheimer disease: identification of cases at risk
ranking and solutions. Bioinformatics among cognitively intact older individuals.
27:1986–1994. https://doi.org/10.1093/ BMC Med 10:127. https://doi.org/10.
bioinformatics/btr300 1186/1741-7015-10-127
17. Van De Ville D, Blu T, Unser M (2006) Surfing 29. Tomlinson BE, Blessed G, Roth M (1968)
the brain. IEEE Eng Med Biol Mag Q Mag Observations on the brains of non-demented
Eng Med Biol Soc 25:65–78 old people. J Neurol Sci 7:331–356
18. Cristianini N, Shawe-Taylor J (2000) An intro- 30. Langs G, Menze BH, Lashkari D, Golland P
duction to support vector machines: and other (2011) Detecting stable distributed patterns of
kernel-based learning methods. Cambridge brain activation using Gini contrast.
University Press, New York Neuroimage 56:497–507. https://doi.
19. Criminisi A, Shotton J (2013) Decision forests org/10.1016/j.neuroimage.2010.07.074
for computer vision and medical image analysis. 31. Kruggel F, Turner J, Muftuler LT, Alzheimer’s
Springer Publishing Company, Incorporated, Disease Neuroimaging Initiative (2010)
Berlin Impact of scanner hardware and imaging pro-
tocol on image quality and compartment
volume precision in the ADNI cohort. and mild cognitive impairment. Neuroimage
Neuroimage 49:2123–2133. https://doi. 55:856–867. https://doi.org/10.1016/j.
org/10.1016/j.neuroimage.2009.11.006 neuroimage.2011.01.008
32. Abdulkadir A, Mortamet B, Vemuri P et al 43. Zhang D, Shen D, Alzheimer’s Disease
(2011) Effects of hardware heterogeneity on Neuroimaging Initiative (2012) Multi-modal
the performance of SVM Alzheimer’s disease multi-task learning for joint prediction of mul-
classifier. Neuroimage 58:785–792. https:// tiple regression and classification variables in
doi.org/10.1016/j.neuroimage.2011.06.029 Alzheimer’s disease. Neuroimage 59:895–907.
33. Petersen RC (2004) Mild cognitive impair- https://doi.org/10.1016/j.
ment as a diagnostic entity. J Intern Med neuroimage.2011.09.069
256:183–194. https://doi. 4 4. Zhang D, Shen D, Initiative ADN (2012)
org/10.1111/j.1365-2796.2004.01388.x Predicting future clinical changes of MCI
34. Mariani E, Monastero R, Mecocci P (2007) patients using longitudinal and multi-
Mild cognitive impairment: a systematic review. modal biomarkers. PLoS One 7:e33182.
J Alzheimers Dis 12:23–35 https://doi.org/10.1371/journal.pone.
35. Forlenza OV, Diniz BS, Nunes PV et al (2009) 0033182
Diagnostic transitions in mild cognitive impair- 45. Haller S, Missonnier P, Herrmann FR et al
ment subtypes. Int Psychogeriatr 21:1088– (2013) Individual classification of mild cogni-
1095. https://doi.org/10.1017/S10416102 tive impairment subtypes by support vector
09990792 machine analysis of white matter DTI. AJNR
36. Misra C, Fan Y, Davatzikos C (2009) Baseline Am J Neuroradiol 34:283–291. https://doi.
and longitudinal patterns of brain atrophy in org/10.3174/ajnr.A3223
MCI patients, and their use in prediction of 46. Haller S, Barkhof F (2017) Interaction of vas-
short-term conversion to AD: results from cular damage and Alzheimer dementia: focal
ADNI. Neuroimage 44:1415–1422. https:// damage and disconnection. Radiology
doi.org/10.1016/j.neuroimage.2008.10.031 282:311–313. https://doi.org/10.1148/
37. Plant C, Teipel SJ, Oswald A et al (2010) radiol.2016161564
Automated detection of brain atrophy patterns 47. Saito S, Ihara M (2016) Interaction between
based on MRI for the prediction of Alzheimer’s cerebrovascular disease and Alzheimer pathol-
disease. Neuroimage 50:162–174. https:// ogy. Curr Opin Psychiatry 29:168–173.
doi.org/10.1016/j.neuroimage.2009.11.046 https://doi.org/10.1097/YCO.00000000
38. Haller S, Bartsch A, Nguyen D et al (2010) 00000239
Cerebral microhemorrhage and iron deposi- 48. Kapasi A, Schneider JA (2016) Vascular contri-
tion in mild cognitive impairment: butions to cognitive impairment, clinical
susceptibility-
weighted MR imaging assess- Alzheimer’s disease, and dementia in older per-
ment. Radiology 257:764–773. https://doi. sons. Biochim Biophys Acta 1862:878–886.
org/10.1148/radiol.10100612 https://doi.org/10.1016/j.bbadis.2015.
39. O’Dwyer L, Lamberton F, Bokde ALW et al 12.023
(2012) Using support vector machines with 49. Madigan JB, Wilcock DM, Hainsworth AH
multiple indices of diffusion for automated (2016) Vascular contributions to cognitive
classification of mild cognitive impairment. impairment and dementia: topical review of
PLoS One 7:e32441. https://doi. animal models. Stroke 47:1953–1959.
org/10.1371/journal.pone.0032441 https://doi.org/10.1161/STROKEAHA.
40. Wee C-Y, Yap P-T, Zhang D et al (2012) 116.012066
Identification of MCI individuals using struc- 50. Brickman AM (2013) Contemplating
tural and functional connectivity networks. Alzheimer’s disease and the contribution of
Neuroimage 59:2045–2056. https://doi. white matter hyperintensities. Curr Neurol
org/10.1016/j.neuroimage.2011.10.015 Neurosci Rep 13:415. https://doi.
41. Cui Y, Liu B, Luo S et al (2011) Identification org/10.1007/s11910-013-0415-7
of conversion from mild cognitive impairment 51. Mueller SG, Weiner MW, Thal LJ et al (2005)
to Alzheimer’s disease using multivariate pre- Ways toward an early diagnosis in Alzheimer’s
dictors. PLoS One 6:e21896. https://doi. disease: the Alzheimer’s Disease Neuroimaging
org/10.1371/journal.pone.0021896 Initiative (ADNI). Alzheimers Dement 1:55–
42. Zhang D, Wang Y, Zhou L et al (2011) 66. https://doi.org/10.1016/j.
Multimodal classification of Alzheimer’s disease jalz.2005.06.003
Chapter 8
Brain Functional Imaging in Preclinical Alzheimer’s

Disease
Peter Häussermann, Thorsten Bartsch, and Oliver Granert
Abstract
In this chapter, we summarize the most important methods and techniques of functional brain imaging
(FBI) in the context of biomarkers in preclinical Alzheimer’s disease (pAD). We analyze putative models of
pAD, covering both genetic and clinical aspects of pAD. Studies on cognitively healthy individuals carrying
the APOE-ε4 allele and patients with mild cognitive impairment have been included. To address method-
ologies, studies on metabolism and cerebral perfusion are analyzed. Although perfusion is not part of the
criteria for pAD, there is increasing evidence that perfusion imaging techniques, including PET/SPECT
methods, and magnetic resonance imaging (MRI) techniques, such as functional MRI (fMRI) and arterial
spin labeling (ASL), can be useful for elucidating the underlying pathophysiological changes in pAD.
However, metabolic imaging with glucose PET still remains the most established FBI biomarker of pAD.
Key words Functional brain imaging, Neurodegeneration, Age-associated cognitive decline,

Subjective memory dysfunction, Mild cognitive impairment, Alzheimer’s disease, PET, MRI, Default
mode network
1 Introduction
1.1 Definition Actually, there is only sparse knowledge about changes occurring
and Potential Models in the brain during the years preceding the cognitive and func-
of Preclinical AD tional impairment associated with Alzheimer’s disease (AD).
Normal cognitive status in the elderly (i.e., incidental forgetfulness
of aging) can be described as aging-associated cognitive decline
without relevant memory complaints.
In 2011, the National Institute on Aging-Alzheimer’s
Association (NIA-AA) published criteria to stage preclinical
Alzheimer’s disease (pAD), mainly based on the amyloid cascade
hypothesis of AD. These criteria [1] define three consecutive stages:
●● Stage 1: Asymptomatic cerebral amyloidosis
●● Stage 2: Cerebral amyloid deposition + downstream neurode-
generation with pathological CSF/PET or MRI markers
119
120 Peter Häussermann et al.
(tau/phospho-tau; 18
F-FDG-PET, atrophy as visualized by
structural MRI)
●● Stage 3: Cerebral amyloidosis + neurodegeneration + subtle
cognitive or behavioral decline, not meeting the criteria for
MCI
Clinically, elderly patients reporting cognitive problems have
recently been classified as suffering from subjective cognitive
decline (SCD). SCD is defined as a subjectively experienced persis-
tent continuous cognitive decline, or decline in cognitive capacity,
compared with a previously normal state, not related to an acute
event. Performance on standard tests used to define mild cognitive
impairment (MCI) is normal. Patients with SCD expressing con-
cern about these deficits have a threefold increased risk of later
developing dementia and a sixfold increased risk of later develop-
ing AD. In models of the very early stages of AD, SCD can be
considered to occur in late stages of pAD, characterized by increas-
ing compensatory cognitive efforts and subtle cognitive changes
before the threshold of MCI or prodromal AD is reached [2].
There are several human models of pAD. The clinically most
important human model is the mild cognitive impairment (MCI)
stage of cognitive decline in the elderly. In MCI, patients are par-
tially symptomatic with substantiated cognitive impairment that
does not significantly interfere with normal daily functioning. The
amnestic MCI subgroup (aMCI) has the highest risk to develop
AD in the future [3]. The annual conversion rate from MCI to
dementia is approximately between 5–25%, depending on MCI
criteria and subtype [3, 4]. MCI can, therefore, be considered as a
transitional period between normal aging and dementia.
Another approach to better understand presymptomatic AD is
to study cognitively healthy individuals with known genetic risk
factors. This includes nondemented individuals who are at risk of
later developing AD, such as subjects with recognized genetic
traits. Individuals carrying the ε4 allele of the apolipoprotein E
(APOE-ε4) gene have an increased risk of later developing AD. The
APOE gene is thought to play a role in cerebrovascular lesions in
dementia [5] and is considered to be the single most important
genetic factor in late onset AD [6]. Carriage of the APOE-ε4 allele
significantly increases AD risk in a gene-dose-dependent manner
and lowers the age of AD onset [7, 8].
1.2 Biomarkers Generally, a biomarker is a characteristic that is objectively mea-

and Functional sured and evaluated as an indicator of pathologic processes [9].
Imaging Effective biomarkers in AD depend on the knowledge of the
in Preclinical AD underlying pathophysiology. The most prominent etiological mod-
els of AD are related to cerebral Aß deposition and tau-
hyperphosphorylation, whereas cerebrovascular changes and
perfusional abnormalities are increasingly being recognized as
important cofactors in the pathogenesis of AD [10, 11]. To date,
Functional Imaging in AD 121
these models only partially mirror underlying pathophysiological

changes in AD; however, the cause of AD still remains unclear.
In AD, an ideal biomarker detects important features of AD
pathophysiology and neuropathology, is neuropathologically vali-
dated, and shows a high diagnostic sensitivity and specificity.
Furthermore, a good biomarker should also be trustworthy, con-
sistent, noninvasive, easy to perform, and inexpensive [12].
Currently deployed imaging biomarkers of preclinical AD
(pAD) include those related to amyloid, tau, cerebral (glucose)
metabolism, and brain atrophy. Alterations of cerebral blood flow
(CBF) have been studied more extensively in the recent years in
pAD and its models. However, perfusion changes are not (yet)
included in the definition of pAD [1].
Functional imaging is currently used to explore the sequence
of pathophysiological changes occurring along the continuum
from normal aging to AD. Functional brain imaging (FBI) bio-
markers of pAD are mainly based on imaging techniques that
explore the metabolic changes of the aging brain. In our review, we
have also included FBI methods exploring CBF changes as there is
evidence that CBF alterations occur early in the course of pAD,
even before amyloid deposition becomes detectable.
According to a vascular AD model, neurodegeneration primar-
ily results from cerebral hypoperfusion, leading to glial and neuro-
nal damage and producing Aß and tau deposition with cognitive
decline and later AD [11]. There is recent evidence supporting an
inverse relationship between cerebral perfusion and AD progres-
sion; therefore, CBF may represent a valuable biomarker of pAD
[10, 13, 14]. A recent study linked the amyloid cascade and the
vascular hypotheses of AD. Using two different mouse models of
AD, Maier et al. [15] showed that only in the presence of cerebral
amyloid angiopathy (CAA), Aß deposition is accompanied by a
decline of regional CBF. Hypoperfusion as seen in AD may, there-
fore, be associated with CAA.
FBI methods can be subdivided into two methodological
approaches, electrophysiological techniques and hemodynamic
methods. Electrophysiological techniques, such as electroencepha-
lography (EEG), event-related potentials (ERP), and magnetenceph-
alography (MEG), directly measure neuronal activity (i.e., action
potentials) as the physiological correlates of neuronal activity.
Positron emission tomography (PET), single photon emission
computed tomography (SPECT), and different magnetic resonance
imaging (MRI) techniques measure secondary effects of neuronal
activation (e.g., changes in CBF linked to neuronal metabolism).
Recently, optical imaging (OI) and near-infrared spectroscopy
(NIRS) have been added to the list of functional imaging tools in
neurodegenerative disorders. Furthermore, other FBI techniques,
such as MR spectroscopy, can detect the regional chemical compo-
sition of the brain. MR spectroscopy, optical imaging, and NIRS
are not routinely used, neither in the clinical nor in the research
setting. Therefore, these techniques are not described in this

chapter.
Electrophysiological tools have a high time resolution with a
relatively low spatial resolution. Hemodynamic tools have a good
spatial resolution with lower time resolution. Recent developments
show promising results combining electrophysiological and hemo-
dynamic techniques (e.g., EEG and fMRI).
In this chapter, we will focus on metabolic as well as hemody-
namic functional imaging methods as they have the potential to
mirror underlying pathophysiological changes in pAD [10, 16, 17].
2 Functional Brain Imaging Methods in Preclinical Alzheimer’s Disease
2.1 Spectrum Single photon emission computed tomography (SPECT) is a

of Techniques nuclear medicine FBI technique using gamma rays. SPECT uses a
and Methods gamma-emitting radioisotope attached to a specific ligand to cre-
ate a radioligand, which then binds to cerebral structures. This
2.1.1 SPECT
radioligand can be visualized by a gamma camera. SPECT has a
lower spatial resolution (about 10 mm) than positron emission
tomography (PET). In FBI, 99mTc-HMPAO (technetium-99m-
hexamethyl propylene amine oxime) is usually used to measure
CBF, which is coupled to regional brain metabolism and energy
use. As compared to 18F-FDG-PET, 99mTc-HMPAO-SPECT is
relatively cheap and simple to generate. Therefore, 99mTc-HMPAO-
SPECT is more widely available in the hospital setting [18]. From
a technical point of view, correction for partial volume effects
using high-resolution MRI is recommended to control for effects
of brain atrophy on PET/SPECT (perfusion) imaging [19].
2.1.2 PET Positron emission tomography (PET) is an in vivo molecular imag-

ing method detecting gamma rays emitted by a positron-emitting
tracer. An injected radioisotope undergoes positron emission decay,
emitting a positron that interacts with an electron to produce
gamma photons detected by the PET scanner. For brain metabolic
imaging, 18F-FDG is by far the most widely used tracer, visualizing
and quantifying mainly the presynaptic regional cerebral metabolic
rate of glucose uptake (rCMRglc) with a half-life of 110 min.
18
F-FDG-PET represents a proxy for neuronal activity and
constitutes an established biomarker for neurodegeneration. There
is a slight but significant decline in rCMRglc in many cortical
regions as a function of age [20]. CBF can be assessed by PET using
the tracer oxygen-15 which has a short half-life (about 2 min).
Several PET tracers are ligands for specific neuroreceptor subtypes
such as the D2/D3 receptors, opioid receptors, serotonin receptors
and transporters, nicotinic acetylcholine receptors, as well as enzyme
substrates. PET imaging can also be used to visualize protein aggre-
gates (e.g., tau and amyloid) within the brain [21].
2.1.3 Functional MRI Functional magnetic resonance imaging (fMRI) is a FBI method
(fMRI) that uses the phenomenon of nuclear magnetic resonance to indi-
rectly measure and map brain function (not anatomy) by detecting
cerebral blood flow (CBF) changes following alterations in neural
activity. fMRI is based on the so-called blood-oxygen-level-
dependent (BOLD) contrast. This method quantifies circum-
scribed changes in deoxyhemoglobin level, which exhibits an
interwoven dependency on CBF, cerebral metabolic rate of O2, as
well as on cerebral blood volume [22]. To understand the fMRI
signal, two effects need to be considered. First, there is a difference
in magnetic properties between oxygen-rich arterial (oxygenated)
and oxygen-poor venous (deoxygenated) blood. Second, neural
activity paradoxically leads to a much greater increase in CBF than
oxygen consumption, which in turn leads to the blood being more
oxygenated with increasing neural activity. An increase in neural
activity in certain brain areas thereby leads to a small but detectable
change in the magnetic resonance signal. The fMRI signal, there-
fore, depends on physiological changes associated with CBF and
oxygen metabolism [22].
fMRI is widely available and relatively cheap and does not use
ionizing radiation, being able to localize cerebral activity in a milli-
meter range. fMRI can be combined with electrophysiological tech-
niques such as EEG and transcranial magnetic stimulation [22].
fMRI acquisition is usually coupled to a carefully designed
experiment, carried out by the subject during imaging. Experimental
paradigms trigger an activation of functional brain networks or
regions coupled. Limitations concern the reliance on cooperation
of the patient; children and dementia patients often cannot per-
form complex fMRI tasks. These experiments are time-consuming
and require a complex experimental setup with MRI-compatible
equipment. In contrast to the very controlled task-based fMRI
experiments, the so-called resting-state fMRI (rs-fMRI) is much
easier to perform and independent from a specific task. By analyz-
ing correlations in the BOLD signal over time between different
regions or voxels, rs-fMRI tries to assess the functional connectivity
(fc) patterns of the brain during rest. The term functional connec-
tivity is used to describe the specific connectivity between brain
areas sharing functional properties (i.e., the temporal correlation
between spatially remote neurophysiological events in distributed
neuronal networks). Slow fluctuations (<0.08 Hz) of the BOLD
signal with a high degree of temporal correlation between func-
tionally connected brain regions during rest and intrinsic process-
ing represent the neuronal baseline activity. These fluctuations
correspond to inter- and intraindividually stable resting-state net-
works (RSNs). Resting-state fc analyses measure task-independent
changes of brain function, thereby providing important insights
into brain plasticity [23].
Several RSNs have been described, linking regions with func-

tional dependencies. RSNs include brain areas involved in sensory
and motor control, visual processing, executive functioning, audi-
tory processing, memory, and the default mode network (DMN).
Using rs-fMRI, brain activity in patients (e.g., children and patients
suffering from AD) that would not be able to complete long exper-
iments or perform complex fMRI tasks can be examined. Unlike
task-based imaging, typically highlighting one single brain network
associated with the specific paradigm, rs-fMRI allows the observa-
tion of several networks simultaneously. The simplicity and short
duration (5–10 min as compared to usually 30 min or more in
task-based fMRI studies) make rs-fMRI an attractive tool in the
research setting [22–26].
The default mode network (DMN) represents an association
of brain regions active during daydreaming, retrieving memories
or planning for the future. Therefore, the DMN represents the
resting state of our brain, showing a direct correspondence to
structural connections. Neuroanatomically, the posterior cingulate
cortex (PCC), inferior parietal, medial temporal lobe (MTL), and
medial frontal gyri belong to the DMN, exhibiting a high level of
correlated BOLD signal activity. These spatially separated but
functionally coupled brain regions interact even if the brain is not
engaged in focused mental activity. Neurophysiologically, there are
slow, synchronized oscillations within the DMN, which are quite
robust and even persist during sleep [27]. The DMN and other
RSNs can be visualized by various methods, including fMRI,
18
F-FDG-PET, and EEG, but rs-fMRI provides the best spatial
imaging resolution and does not utilize radioactivity. There is some
evidence for altered processing of the DMN in AD [28].
2.1.4 Arterial Spin Arterial spin labeling (ASL) is a noninvasive MRI technique, mag-
Labeling: MRI netically labeling arterial water and using it as an endogenous tracer
to measure CBF changes associated with neuronal activity [10, 29].
The BOLD signal, which is the usual basis for the fMRI signal, is
linked to a signal change between two or more conditions. In
fMRI, there is no information about the absolute level of blood
flow. In contrast to this, the ASL technique has the ability to quan-
tify the absolute level of CBF by subtracting the signal of magneti-
cally tagged and untagged arterial blood. Thereby, the static signal
from all hydrogen nuclei subtracts out, ideally leaving only the sig-
nal arising from arterial blood flow [29, 30]. The ASL signal is
therefore independent from changes in neuronal activity. However,
the disadvantage is an increase in acquisition time, since the ASL
technique requires the acquisition of two images (one with the
labeled and one with non-labeled blood). ASL MRI has the poten-
tial to precisely estimate both extent and site of neural function
[29]. Changes in CBF are more localized to the brain parenchyma
(capillary bed), whereas the BOLD signal is more localized to
cerebral veins and venules [29, 30]. As ASL MRI is a noninvasive,

easily repeatable method with good reliability and reproducibility
[29, 31], it is currently one of the most promising functional imag-
ing biomarkers of pAD.
2.2 Imaging Besson et al. [32] characterized a group of cognitively healthy sub-
of Metabolism jects by structural MRI, 18F-FDG-PET, and amyloid PET (florbeta-
in Preclinical AD pir) and dichotomized these individuals into positive or negative for
each neuroimaging marker. The 18F-FDG-positive group showed
2.2.1 F-FDG-PET
18
an AD typical hypometabolic pattern. Furthermore, in the “neuro-
degeneration” group with both structural MRI and FDG-PET
AD-like changes, the authors found a trend for an inverse relation-
ship with Aß deposition (i.e., subjects with neurodegeneration
exhibited less Aß and vice versa). Therefore, Besson et al. suggested
to combine structural MRI and amyloid/glucose PET biomarkers.
These methods offer unique insight into the time course of Aß
deposition and neurodegeneration in pAD. Furthermore, neurode-
generative changes seen in this study were largely independent from
cerebral Aß deposition.
Teipel et al. [33] analyzed hippocampal and posterior cingu-
late cortex (PCC) volume, PCC glucose metabolism, and Aß load
in a cohort of 667 subjects participating in the Alzheimer’s Disease
Neuroimaging Initiative (ADNI). Controls and early MCI patients
exhibited hypometabolism in PCC which was associated with hip-
pocampal atrophy. In late stages of MCI, hypometabolism in PCC
was associated both with PCC and hippocampal atrophy as well as
PCC Aß deposition. In patients suffering from AD, PCC hypome-
tabolism was associated with PCC atrophy.
In another study, Ewers et al. [34] showed that temporopari-
etal and prefrontal hypometabolism as well as decline in executive
function predicted conversion of cognitively healthy elderly to
MCI or dementia during a follow-up period of 3–4 years. Best
predicting results for conversion to MCI/dementia were obtained
when combining executive function assessment with rCMRglc in
the above-mentioned areas [34]. The same group examined the
effects of cognitive reserve capacity (CRC) on glucose metabolism
in cognitively healthy individuals with (Aß+) and without (Aß−)
CSF biomarkers of preclinical AD [35]. When using a ROI
approach, higher CRC was associated with lower glucose metabo-
lism in the PCC and angular gyrus in the Aß+ group, while higher
CRC was associated with PCC hypermetabolism in the Aß− group.
CRC may, therefore, play an important compensatory role in sus-
taining cognition in the presence of cerebral Aß deposition and
altered glucose metabolism.
Another study [36] analyzed the longitudinal changes of Aß
deposition, glucose metabolism, and medial temporal lobe (MTL)
atrophy in three different cohorts (pAD, MCI and dementia).
Patients with initial pAD stages II and III showed greater MTL
volume loss and hypometabolism during follow-up than the stage

I pAD group. Apparently, in this study, higher rates of MTL neu-
rodegenerative changes occurred during follow-up in individuals
with pAD who initially exhibited Aß and neurodegeneration bio-
markers, as objectified with hypometabolism and MTL atrophy.
In another longitudinal study, Wirth et al. [37] showed accel-
erated cognitive decline in pAD with both positive Aß and neuro-
degeneration imaging markers in AD vulnerable regions (i.e.,
temporoparietal cortices and hippocampus) as compared to healthy
controls. In this study, non-memory decline was best predicted by
Aß and temporoparietal neurodegenerative alterations, while
memory decline was best predicted by Aß and hippocampal neuro-
degenerative changes. These results suggest regional specificity of
neurodegeneration in pAD. Ossenkoppele et al. [38] showed that
initial glucose hypometabolism and amyloid deposition were related
to longitudinal worsening in memory, attention and executive
function in pAD, MCI and AD over a mean follow-up period of
2.2 years. Gray matter volume loss was mainly associated with lon-
gitudinal memory decline.
Knopman et al. [39] applied imaging biomarkers of neurode-
generation and Aß in three groups: one containing subjects with
different stages of pAD, one with subjects that had negative imag-
ing biomarkers, and one with subjects with positive biomarkers for
neurodegeneration without Aß deposition (sNAP, suspected non-
Alzheimer pathophysiology). The sNAP group did not differ from
stages II and III pAD concerning clinical symptoms of parkinson-
ism, glucose metabolism, MR volumetry, cerebrovascular risk fac-
tors, or ischemic brain lesions, indicating that neurodegenerative
changes are largely unrelated to Aß deposition in cognitively
healthy individuals.
2.2.2 18F-FDG-PET Seo et al. [40] analyzed APOE-ɛ4, rCMRglc, and cerebral amyloid
in APOE-ɛ4 Carriers load (18F-florbetapir PET) in cognitively healthy controls, MCI
patients, and AD patients. In controls, carriage of the APOE-ɛ4
allele was associated with hypometabolism in the bilateral frontal,
temporal, and left parietal regions. In MCI, the same comparison
showed hypometabolism in the bilateral posterior parietal, temporal,
and left frontal regions. In AD, APOE-ɛ4 carriers exhibited hypo-
metabolism in the left hippocampus, right insular, and right tempo-
ral gyrus. After adjustment for Aß, the significant differences within
temporal areas disappeared in controls and MCI, whereas in AD, all
significant metabolic differences disappeared. The authors deducted
that Aß-independent APOE-ɛ4 influence on glucose metabolism is
restricted to frontal and parietal brain regions, mainly in the early
stages of cognitive decline. The same group [41] also showed that
temporoparietal hypometabolism disappears when controlling for
Aß in APOE-ɛ4 carriers without cognitive symptoms as compared to
healthy noncarriers [41], while medial frontal and anterior temporal
hypermetabolism seen in the abovementioned comparison persists
when controlling for Aß [41]. Global CMRglc declines with age,

and in particular rCMRglc decline within the PCC/precuneus
region and the lateral parietal region might be caused by the com-
bined effects of aging and genetic (i.e., APOE-ɛ4 carriage) traits
[20]. Depression and anxiety may also have an impact on rCMRglc.
Therefore, these neuropsychiatric symptoms should always be deter-
mined and taken into consideration when interpreting metabolic
patterns in models of pAD [42].
To better understand the influence of genetic factors on
CMRglc, Didic et al. [43] explored glucose metabolism in healthy
APOE-ɛ4 carriers and noncarriers, both at a local and a cerebral
network level. Carriers showed hypometabolism in the left anterior
MTL, including the entorhinal and perirhinal cortices. MTL
metabolism was related to memory performance, and there was a
stronger metabolic connectivity of the MTL with frontoparietal
regions in carriers than in noncarriers, proposing compensatory
metabolic network alterations in carriers.
Higher education as proxy for CRC mediates rCMRglc in cog-
nitively healthy controls carrying the APOE-ɛ4 allele. Higher CRC
was positively correlated to frontotemporal metabolism and epi-
sodic memory performance in these individuals as compared to
noncarriers, independent from Aß deposition [44]. Furthermore,
subjects with lower CRC carrying the APOE-ε4 allele exhibited
medial temporal and prefrontal hypometabolism relative to non-
carriers. Higher educated carriers were comparable to noncarriers
in these areas and showed temporal lobe hypermetabolism. These
results indicate that CRC helps to counteract the effects of
APOE-ε4 on metabolism, independent from cerebral Aß load,
thereby representing a protective factor which helps to delay cog-
nitive deterioration in APOE-ε4 carriers [44].
Using 18F-FDG-PET to characterize metabolic networks in a
large sample of cognitively healthy elderly, patients with MCI and
AD, Yao et al. [45] dichotomized these individuals into two
groups: one with the APOE-ε4 allele carriage status and one with-
out. The authors demonstrated increased local short distance inter-
regional metabolic correlations and disrupted long distance
interregional metabolic correlations in the APOE-ε4 carrier group.
The pattern observed in APOE-ε4 allele-positive subjects resem-
bled the pattern observed in MCI and AD.
Furthermore, there seems to be an APOE-ε4 gene-dose effect
in cognitively normal, late-middle-aged subjects (ε4 homozygotes,
ε4 heterozygotes, and noncarriers) on regional gray matter volume
and rCMRglc. Gray matter atrophy and hypometabolism corre-
lated with the APOE-ε4 gene dose [46].
2.2.3 F-FDG-PET
18
MCI is clinically the most important model of pAD. The amnestic
in MCI subtype has the highest risk of conversion to AD. The metabolic
pattern of aMCI resembles that seen in early stages of AD.
PCC and precuneus hypometabolism constitute predictors for

cognitive decline in MCI, whereas temporoparietal hypometabo-
lism may either be associated with memory deficits or with conver-
sion to dementia [47, 48]. Metabolic changes within the left lateral
frontal cortex were described in another study to anticipate con-
version to dementia in aMCI [49].
Coutinho et al. [50] studied glucose metabolism and CSF bio-
markers of AD in non-amnestic MCI, aMCI, and healthy controls.
As compared to controls, both MCI groups exhibited reduced
CMRglc in the precuneus. In relation to aMCI, the non-amnestic
MCI patients also showed right prefrontal hypometabolism and
higher Aß CSF levels. In this study, the CSF biomarker profile in
the aMCI group was more closely related to manifest AD than in
the non-amnestic MCI group.
Bailly et al. [51] compared glucose metabolism and atrophy of
the hippocampus, amygdala, precuneus, and ACC/PCC of cogni-
tively healthy controls, MCI patients, and AD patients. In relation
to controls, mean PCC volume was reduced in MCI and AD. The
precuneus and PCC regions also exhibited hypometabolism in
MCI and AD, as compared to controls.
Another study also found the hypometabolic pattern of MCI
to AD converters to cover the precuneus and PCC, with further
hypometabolism in temporoparietal and frontal cortices [52]. In
aMCI, declarative memory storage impairments correlated with
hippocampal atrophy, glucose hypometabolism, and cerebral Aß
deposition. Delayed recall deficits correlate with neuroimaging
biomarkers of MCI to AD converters and mirror the conversion
from aMCI to AD [53].
Franzmeier et al. [54] demonstrated precuneus hypometabo-
lism in aMCI patients with Aß deposition as compared to healthy
controls, with more pronounced rCMRglc reductions in patients
with higher CRC. Global left frontal cortex (LFC) connectivity
was linked to higher CRC. The authors concluded that higher
global LFC connectivity represents a putative substrate of CRC in
aMCI patients which may help to maintain memory performance
despite glucose hypometabolism in this model of pAD. Förster
et al. [55] demonstrated that aMCI patients have significantly
improved cognition and attenuated metabolic decline in the left
anterior temporal pole and ACC after a 6-month multicomponent
cognitive intervention as compared to an active control group.
Chen et al. [56] examined 1-year rCMRglc declines in AD,
aMCI, and cognitively healthy controls. Both the AD and the
aMCI group showed significant metabolic decline in the PCC,
medial and lateral parietal, medial and lateral temporal, frontal, and
occipital cortex as compared to controls within 1 year.
The MTL is of particular interest in MCI and helps to distin-
guish aMCI from other MCI syndromes. Patients with aMCI have
decreased MTL glucose metabolism as compared to patients with
non-amnestic MCI [57]. Also, cognitively deteriorating patients

with amnestic MCI exhibit MTL hypometabolism as compared to
controls and MCI patients that do not show cognitive decline [58].
2.2.4 A Methodological In a recent paper, Granert et al. [17] analyzed 18

F-FDG-PET
Example: Metabolic images in a large cohort of patients with:
Topology in Early and Late
●● Early stages of neurodegeneration (i.e., aMCI, early AD, and
Stages
patients with Parkinson’s disease)
of Neurodegeneration
●● More advanced stages of neurodegeneration (i.e., patients with
Parkinson’s disease dementia and Lewy body dementia)
●● Cognitively normal age-matched controls (Co)
There is abundant clinical, pathophysiological, and imaging
overlap between the abovementioned diseases. Therefore, one can
question the concept of distinct disease entities and suggest a
somehow continuous spectrum of early neurodegenerative changes
between normal aging, Parkinson’s disease (PD), and AD. The
same hypothesis might also hold true in more advanced stages of
neurodegeneration. According to this concept, the authors [17]
built a topological map based on rCMRglc to rank and localize
single subjects’ disease status according to parkinsonian and
dementia typical pattern expression in patients clinically diagnosed
to suffer from PD, Parkinson’s disease dementia (PDD), Lewy
body dementia (DLB), aMCI, or AD.
The authors describe a metabolic topological map which con-
firmed an inseparable spectrum of disease manifestation according
to two different and dissociable expression patterns. The expres-
sion values, which were extracted from the AD versus controls pat-
tern, were highly correlated with cognitive deficits, but not motor
deficits. The opposite was found for the corresponding expression
values of the PD versus controls pattern, which was highly corre-
lated with motor deficits, but not cognitive deficits.
This metabolic imaging approach supports the notion that
there is a continuous spectrum of neurodegeneration between
neurodegenerative syndromes affecting predominantly motor skills
(PD, PDD) and neurodegenerative syndromes primarily leading to
cognitive deficits (aMCI and AD, DLB).
The voxelwise SPM group comparisons with the controls and
PD as well as AD exhibited typical and distinct metabolic profiles for
both the PD as well as the AD group (Fig. 1). The AD group showed
a decline of rCMRglc bilaterally in the PCC, the lateral temporal
lobe, and the inferior parietal cortex (Fig. 1a). Temporoparietal and
PCC reductions of rCMRglc in the group of early AD patients are in
accordance with previous imaging studies focusing on rCMRglc
alterations in AD [59–63]. These rCMRglc reductions in early AD
were localized within cortical areas closely linked to the DMN, a
phenomenon that has been described before [64]. The PD group
Fig. 1 Visualization of the statistical maps used to calculate the expression values corresponding to the group
t-contrasts. (Panel a) AD versus CON. (Panel b) PD versus CON. Statistical maps were shown with a threshold
of punc < 0.001 and overlaid onto the single-subject MR T1 template as delivered with the SPM software.
exhibited frontal and parieto-occipital hypometabolism as well as

temporal and pontocerebellar hypermetabolism (Fig. 1b). This pat-
tern was closely related to previous neuroimaging findings in PD
[59]. Finally, there were neither clinical nor metabolic differences
between the PDD and DLB groups [17].
2.3 Imaging Perfusion describes the process by which oxygenated arterial blood
of Perfusion is provided to the capillary bed in a predefined mass of brain tissue.
in Preclinical AD It is measured in milliliters of blood per 100 g of tissue per min
[22]. CBF values in human being approach 50 mL/100 g per min,
and gray matter CBF is approximately three times higher than
white matter CBF [22]. Regional CBF can be assessed using PET/
SPECT methods, the more ancient 133Xe inhalation technique, and
MRI-based techniques such as fMRI and ASL.
In normal aging, cognitively healthy adults exhibit grey matter
perfusion reductions of about 0.45% per year [31]. Hypoperfusion
in aging tends to be widespread, affecting frontal, medial temporal,
parietal, and subcortical regions [13, 31, 65, 66]. In the stage of
clinically manifest dementia, perfusion in patients suffering from
AD correlates negatively with disease severity [67]. A p
henotypically
normal cognitive status in the elderly despite perfusion deficits may
reflect cognitive reserve capacity (CRC) of the aging brain [68].
As neither oxygen nor glucose can be stored within the brain,
neuronal energy supply strongly depends on a constant delivery of
oxygen and glucose by CBF. Therefore, there is a strong link
between neural activity, energy metabolism, and CBF, which rep-
resents the rationale of FBI methods using perfusion, not only in
pAD [10]. Several factors mediate the relationship between CBF
and cognition, including CRC, age, sex, genotype, and vascular
risk factors.
In this chapter, we mainly present exemplary fMRI and rs-fMRI

studies in pAD and its models as there is abundant and quite het-
erogeneous literature on this topic, exceeding the volume of this
chapter.
2.3.1 PET/SPECT To date, there are no recent studies applying SPECT or PET for
Perfusion Imaging perfusion imaging in pAD, as defined by Sperling et al. [1]. However,
one study [69] investigated the relation between rCBF (as measured
with 99mTc-exametazime-SPECT) and CSF biomarkers in cogni-
tively healthy elderly. The authors found that high CSF p-tau and
total tau levels are correlated with decreased rCBF in the right medial
frontal lobe and that high CSF p-tau levels are correlated with
increased rCBF in the left frontotemporal border zone. There were
no significant associations between rCBF and CSF Aß1–42 levels.
These results indicate a correlation between tau pathology rather
than Aß deposition and CBF abnormalities in pAD.
Another study from the Baltimore Longitudinal Study of
Aging using repeated resting-state 15O-H2O PET scans found
rCBF differences between cognitively normal elderly with high and
low cerebral amyloid load [70]. Greater longitudinal reductions of
CBF in the group exhibiting high amyloid values were found in the
anterior as well as middle cingulate cortex, left thalamus, right
supramarginal gyrus, and midbrain. Greater longitudinal gains in
rCBF in the group exhibiting high amyloid values were encoun-
tered in left medial and inferior frontal gyri, left inferior parietal
gyrus, right precuneus, and left postcentral gyrus.
2.3.2 fMRI, Resting-State Edelman et al. [71] combined fMRI and Pittsburgh compound
fMRI and Functional B-PET (PIB-PET) in a face-name memory task. In their study,
Connectivity Aß-positive healthy elderly displayed an increased (possibly com-
in Preclinical AD pensatory) hippocampal activation during the task. In another
study, Gordon et al. [72] combined fMRI and CSF, respectively,
PET biomarkers of AD in two attentional control tasks to examine
changes related to Aß and tau pathology in cognitively healthy
subjects. They found that increased levels of tau and p-tau are
linked to overactivations of attentional control areas, indicating
that alterations of attentional networks are associated with AD
neuropathology.
Sheline et al. [73] applied rs-fMRI and PIB-PET in AD patients
and two groups of cognitively healthy individuals, one with evi-
dence of brain Aß (Aß+) and one without Aß (Aß−). The Aß+
group as well as the AD group showed significant differences in
functional connectivity (fc) within the DMN. The authors con-
cluded that brain Aß deposition leads to alterations in fc of the
DMN in pAD, in the same anatomical regions and in the same
direction as seen in AD.
Schultz et al. [74] examined fc in several cortical association
networks, including the DMN, in cognitively healthy individuals.
All subjects underwent Aß and tau PET imaging. In Aß-positive

individuals, there was increased connectivity in the DMN with low
tau levels, whereas elevated tau levels were associated with decreased
connectivity in Aß-positive individuals. These results propose a
hyperconnectivity phase followed by a hypoconnectivity phase in
the course of pAD.
Brier et al. [75] used rs-fMRI in two groups of cognitively
healthy subjects: one group with evidence of pAD as assessed by
CSF markers of AD and the other group with normal CSF bio-
markers. The authors found significant interactions between age
and biomarker status in the DMN and salience and dorsal atten-
tion (DAN) resting-state networks. The presence of AD biomark-
ers increased the size of the aging effect within these networks.
The authors observed alterations of fc in pAD with age, not exis-
tent in the group without evidence of pAD. These results suggest
that underlying AD pathology in pAD may be responsible for a
considerable amount of the effects of aging on fc in cognitively
healthy subjects.
In another study, the same group [76] examined cognitively
normal elderly with CSF biomarkers of AD and fc-fMRI. Resting-
state fMRI imaging has shown that brain networks deteriorate in
manifest AD. In their study, the authors found large-scale discon-
nections in CSF-defined pAD, similar to that seen in manifest AD.
With progressive deterioration of cognition, this study identified
three important hub regions (brain regions highly connected with
other regions) on the medial prefrontal cortices and ACC that lose
their functions gradually from normal aging to manifest AD.
The same group [77] also showed that both Aß and tau pathol-
ogy affect DMN integrity before AD becomes clinically manifest.
In their study, the authors found CSF biomarkers of AD to be
independently associated with reduced DMN integrity. The stron-
gest decrease in fc was seen between the PCC and MTL regions.
These fc reductions could not be attributed to age or atrophy.
Drzezga et al. [78] analyzed fc in cortical hub regions in non-
demented cognitively healthy individuals (CH-Aß+ and CH-Aß−)
and MCI patients (MCI-Aß+ and MCI-Aß−) with and without
cerebral amyloidosis. All subjects underwent amyloid PET and
18
F-FDG-PET imaging, as well as structural MRI and rs-fMRI.
Disruptions of whole-brain connectivity were seen in the MCI-Aß+
group in the PCC/precuneus region (an important hub region),
strongly overlapping with glucose hypometabolism. There were
also minor connectivity disruptions and hypometabolism in the
CH-Aß+ group. These findings suggest that a disruption of fc and
glucose hypometabolism represent early functional consequences
of amyloid burden (i.e., AD pathology), even before dementia
becomes clinically apparent. The PCC/precuneus region has a
particular susceptibility to amyloid deposition and glucose hypo-
metabolism, thereby putatively reflecting a link between synaptic
dysfunction and functional disconnection.
2.3.3 Arterial Spin As described above, reductions of CBF mirror the pathophysiologi-
Labeling MRI cal changes seen in pAD. Hays et al. [10] summarized ASL MRI
in Preclinical AD studies in several models of pAD. In the early preclinical stage, there
is evidence of hyperperfusion, whereas in later preclinical stages,
there exist both hyper- and hypoperfusion. Later, in manifest AD,
there is evidence for widespread reductions of CBF. Cerebral amy-
loid angiopathy may be an important contributing factor.
A recent study [14] found continuous CBF decreases in pAD,
mainly in temporal and parietal regions, along the stages as defined
by Sperling et al. [1]. Other studies have found CBF decreases
related to Aß deposition in the inferior temporal and parietal
cortex, In control subjects, high Aß levels were associated with
greater CBF reductions [70, 79].
2.4 APOE-ɛ4 Carriers Generally, there have been few PET/SPECT perfusion studies
focusing on cognitively normal APOE-ɛ4 carriers. Thambisetty
2.4.1 PET/SPECT
et al. [80] studied longitudinal CBF differences between two
Perfusion Imaging
groups of nondemented elderly from the Baltimore Longitudinal
Study of Aging using 15O-PET. One group contained APOE-ɛ4
carriers and the other contained noncarriers. APOE-ɛ4 carriers
exhibited widespread and greater regional CBF decline in frontal,
temporal, and parietal cortices as compared to noncarriers.
In patients with manifest AD, there was a dose-dependent
effect of the APOE-ɛ4 allele on rCBF, with more hypoperfusion
seen in frontal, temporoparietal, MTL, and occipital regions in the
APOE-ɛ4 carrier group. AD patients homozygous for the ɛ4 allele
also exhibit more severe MTL atrophy than noncarriers or AD
patients heterozygous for the ɛ4 allele [19, 81–83]. Furthermore,
there is a more rapid decline of CBF in the APOE-ɛ4 group com-
pared to noncarriers [19].
MCI patients carrying the APOE-ɛ4 allele have a greater
chance to convert to AD. The converters show hypoperfusion in
the MTL and parietal regions [84] as well as within the postcentral
region, as compared to nonconverters [85].
2.4.2 fMRI, Resting-State Fleisher et al. [86] compared resting CBF (DMN signal differ-
fMRI, and Functional ences) and fMRI BOLD response during an associative memory-
Connectivity encoding task in APOE-ɛ4 carriers with a positive familial history
of AD versus healthy controls without these risk factors. BOLD
activations during encoding did not differ between both groups.
Deactivations during encoding were more pronounced in the low
risk group within the parietal cortex. However, resting state DMN
analysis better discriminated both groups than encoding associated
fMRI. The authors conclude that resting state imaging is a more
effective and promising tool in assessing AD risk as compared to
activation fMRI.
Harrison et al. [87] analyzed fc of the hippocampus in cogni-
tively normal elderly, subdivided into two groups, one consisting
of APOE-ɛ4 carriers and one consisting of noncarriers. The para-

digm used was a paired-associates memory task. During encoding,
APOE-ɛ4 carriers showed lower fc change compared to baseline
between the anterior hippocampus and right precuneus, anterior
insula, and cingulate cortex. During retrieval, the bilateral supra-
marginal gyrus and right precuneus showed lower fc change with
anterior hippocampus in the APOE-ɛ4 carriers. The APOE-ɛ4
group, in contrast to the noncarriers, showed strong negative fc
changes compared to noncarriers where positive fc changes were
seen. These differences may represent prodromal functional
changes mediated, at least in part, by the APOE-ɛ4 allele. These
results were considered to be consistent with the anterior- to-
posterior theory of AD progression in the hippocampus. Altered fc
of the hippocampus can, therefore, be regarded as an early FBI
biomarker for pAD.
In another study, Sheline et al. [88] applied rs-fMRI and PIB-
PET in two groups of cognitively healthy individuals, one with
APOE-ε4 allele carriers (ε4+) and one group without APOE-ε4
allele carriers (ε4−). Amyloid-negative ε4+ carriers differed signifi-
cantly from the amyloid-negative ε4− group in fc of the DMN.
The observed alterations in the ε4+ carriers were similar to the
changes described in AD. Disruptions of fc within brain regions of
the DMN in the ε4+ group were seen in the absence of cerebral Aß
deposition. These results suggest that genetic risk factors for AD
may produce changes of resting-state networks (i.e., the DMN)
even in the absence of cerebral amyloidosis.
2.4.3 Arterial Spin Michels et al. [89] assessed APOE-ε4 genotype, Aß load, and CBF
Labeling MRI in cognitively healthy (CH) individuals and aMCI patients. The
global CBF was lower in the group of APOE-ε4-positive subjects
as compared to noncarriers. In aMCI, the global CBF was lower
compared to CH, and participants with increased Aß showed a
trend for lower global CBF. Therefore, the authors concluded that
the APOE-ε4 allele has an impact on CBF, which is at least par-
tially independent from Aß.
Bangen et al. [90] used ASL MRI and fMRI to assess resting
CBF as well as task-dependent CBF and BOLD response, in a
memory-encoding paradigm in cognitively healthy elderly (with
and without the APOE-ε4 allele) and in MCI patients (also with
and without the APOE-ε4 allele). In cognitively healthy APOE-ε4
carriers, there is evidence of hyperperfusion with increased resting-
state CBF in the MTL compared to ε4 noncarriers, while MCI
patients have decreased resting-state CBF within MTL structures
relative to controls. There were no significant group differences for
task-dependent BOLD response (MCI versus controls). These
findings suggest that alterations in resting-state CBF and CBF
response to memory encoding may serve as early markers of brain
dysfunction in different models of pAD.
Wierenga et al. [91] used ASL MRI to investigate the influ-

ence of the APOE genotype and age on CBF in two groups of
cognitively healthy younger and older adults, one containing
APOE-ε4 carriers and one containing noncarriers. Older adults
exhibited decreased gray matter CBF, as compared to younger
adults, in widespread brain areas. The ACC region exhibited a
putatively compensatory increased CBF in young APOE-ε4 carri-
ers and reduced CBF in older APOE-ε4 carriers.
In conclusion, studies on CBF in clinically nondemented
APOE-ε4 carriers demonstrated both increases and decreases of
CBF relative to cognitively healthy elderly [13, 80]. Studies also
support the notion that effects of the APOE genotype on CBF are
liaised by age, with elderly subjects displaying more areas of hypo-
perfusion and younger subjects exhibiting more areas of hyperper-
fusion [91].
In young APOE-ε4 carriers, increased CBF seems to be associ-
ated with better executive functioning, suggesting that hyperper-
fusion is a compensatory mechanism [91]. Elderly APOE-ε4
carriers have abnormally low CBF, which may indicate a break-
down of this initial compensatory mechanism.
Disease severity may also mediate the relationship between
APOE genotype and CBF: nondemented APOE-ε4 carriers and
AD patients with this genotype show widespread hypoperfusion as
compared to noncarriers. In MCI, APOE-ε4 carriers exhibited
regional hyperperfusion in the right parahippocampal, bilateral
cingulate gyri, and the right PCC [92].
2.5 MCI Using serial 15O-water PET imaging data from the Baltimore
Longitudinal Study of Aging, one study [93] found alterations of
2.5.1 PET/SPECT rCBF in elderly individuals later converting to MCI. Compared to
Perfusion Imaging cognitively stable elderly, MCI converters displayed greater longi-
tudinal rCBF decreases in temporal, parietal, and thalamic regions
and rCBF increases in orbitofrontal, MTL, and ACC regions.
These rCBF changes were seen within brain regions that show
early accumulation of pathology of AD, suggesting a link between
early neuropathological changes and rCBF alterations.
2.5.2 fMRI, Resting-State Adriaanse et al. [94] analyzed fc using rs-fMRI in the DMN in
fMRI, and Functional cognitively healthy elderly and patients with MCI and AD. Amyloid
Connectivity deposition was assessed with PIB-PET. The authors found no asso-
ciation between fc of the DMN and Aß load in DMN areas in any
of the groups or between groups. The same group described
decreased fc within the DMN in AD in the precuneus and PCC
compared to controls. MCI patients exhibited fc changes that were
situated between patients suffering from AD and controls [95].
Liang et al. [96] examined fc in MCI patients in three different
areas of the inferior parietal cortex, notably the intraparietal sulcus
(IPS), the angular gyrus (AG), and the supramarginal gyrus (SG).
With these anatomical hallmarks, the authors were able to trace

three different approved RSNs within this region: the DMN, the
salience network (SN), and the executive control network (ECN).
The AG displayed reduced fc within the DMN. The IPS also exhib-
ited decreased fc with the right inferior frontal gyrus and increased
fc with the left frontal lobes in the ECN. The SG showed decreased
fc with the frontal and parietal cortex and increased fc with some
subcortical areas of the salience network. Connectivity within these
networks correlated with cognitive impairment in the MCI patients.
The authors concluded that MCI is associated with changes of large-
scale functional brain networks.
Another group [78] analyzed fc in cortical hub regions (PCC/
precuneus) in MCI patients with and without cerebral amyloidosis.
Disruptions of whole-brain connectivity were seen in the MCI-Aß+
group in the PCC/precuneus region, strongly overlapping with
glucose hypometabolism in these regions.
2.5.3 Arterial Spin Perfusion changes seen in MCI are suggestive of neurodegeneration-
Labeling MRI induced CBF dysregulation, which may result from altered metabolic
requests to preserve cognitive function [97]. Mild cognitive impair-
ment is linked to increases and also decreases in resting-state CBF.
Kim et al. [92] used ASL MRI to study CBF in patients suffer-
ing from AD and MCI as well as in cognitively normal controls.
Each group was further divided into carriers or noncarriers of the
APOE-ɛ4 allele. In the MCI APOE-ɛ4 carriers, hyperperfusion
was detected in the right parahippocampal gyrus, bilateral cingu-
late gyri, and right PCC as compared to noncarriers.
Michels et al. [89] assessed Aß, CBF, and APOE genotype in
cognitively healthy individuals and aMCI patients. They found
widespread CBF reductions, largely independent from Aß load, in
elderly APOE-ε4 carriers. Global CBF was lower in patients with
aMCI compared with cognitively healthy elderly. Interestingly, car-
riage of the APOE-ε4 allele had a more deleterious effect on both
local and global CBF than cerebral Aß deposition.
However, another study found that cerebral amyloid pathol-
ogy affected CBF across the span from cognitively healthy elderly
(with cerebral amyloidosis) to clinically manifest AD. In their
study, the authors [79] showed Aß pathology to be associated with
hypoperfusion in pAD, whereas in later stages of neurodegenera-
tion, atrophy becomes more prominent.
Other studies found that CBF increases within the temporal
lobe, ACC, insula, putamen, left hippocampus, right amygdala,
and basal ganglia. Decreases of CBF have been described in frontal
and temporoparietal areas (extending to the MTL), PCC, precu-
neus, and left middle occipital lobe [90, 98–101].
The PCC seems to play an important role as indicator of cog-
nitive decline, even in cognitively normal elderly [68]. This brain
region, known to be involved in early AD changes, exhibits
hypoperfusion in MCI and also in cognitively healthy individuals

later showing cognitive decline as compared to the cognitively sta-
ble elderly.
Increases of CBF often correlate with better memory perfor-
mance, not only in MCI, whereas decreases of CBF often correlate
with cognitive dysfunction. Studies of functional CBF changes
showed hypoperfusion in the right precuneus/cuneus in the
aMCI-group as compared to age-matched controls during the
control condition. During a memory encoding task condition,
hypoperfusion extended to the PCC. The perfusion rates corre-
lated with the cognitive abilities in aMCI [101]. Another study
[102] demonstrated that perfusion changes induced by a memory-
encoding task can also help to distinguish aMCI and cognitively
healthy elderly. Task-enhanced ASL MRI may therefore represent
a potential FBI biomarker of pAD.
3 Discussion
The rationale for defining pAD is that the best time for preventing
neurodegeneration in elderly subjects is early in the predementia
stage. Innovative pharmacological and non-pharmacological pre-
ventive interventions can best be assessed in pAD [1, 103]. Hence,
there is a need for personalized, biomarker-based approaches to
identify future dementia patients.
In recent years, functional brain imaging (FBI) has increas-
ingly been recognized as having the potential to represent such
biomarkers of pAD. Table 1 summarizes the most important FBI
findings in pAD, MCI, and nondemented APOE-ε4 allele carriers.
Metabolic imaging (i.e., 18F-FDG-PET) has significantly con-
tributed to our understanding of pAD, showing hypometabolic
changes within the PCC, precuneus, MTL, and temporoparietal
regions that resemble metabolic alterations observed in AD.
Concerning perfusion imaging, more heterogeneous results
have been obtained. In AD, there is evidence for widespread reduc-
tions of CBF, with a negative correlation between CBF and disease
severity [67]. In pAD, both hypo- and hyperperfusion have been
reported in various brain regions in different stages of cognitive
decline [14]. Hays et al. [10] summarized ASL MRI studies in
several models of pAD. In the preclinical phase, hyperperfusion
prevails, mainly in MTL structures, while later in the subclinical
phase, both hyper- and hypoperfusion coexist, with some evidence
of initial hyperperfusion and later hypoperfusion.
CBF dysregulation in pAD and its models suggests the pres-
ence of vascular regulative mechanisms, putatively in response to
an increased need for oxygen and glucose, or changes in cerebral
metabolism to maintain cognitive function [97]. However, the
direction of CBF alterations is not specified. Hyper- as well as
hypoperfusion may indicate underlying pathological or compensa-

tory changes in different predementia stages. Cerebral amyloid
angiopathy may constitute an important contributing factor link-
ing the classical amyloid cascade hypothesis with the vascular
hypothesis of AD [15].
Small vessel disease and cardiovascular risk factors have an
important impact on CBF; thus, CBF changes in the elderly do not
necessarily mirror underlying pAD pathophysiology. Hypoperfusion,
therefore, reflects the combined disease burden of microangiopa-
thy, aging, and neurodegeneration. Normal aging processes have an
impact on CBF, Aß− and tau− load. Aging is associated with a
slight but continuous reduction of gray matter CBF (about 0.45%
per year) [31], and hypoperfusion affects frontal, temporo-parietal,
and subcortical regions [13, 31, 65, 66, 104].
Only markers of cerebral Aß deposition and later neurodegen-
erative changes have been included to define pAD [1]. However,
there is growing evidence that perfusional alterations are also
involved in the very early pathogenesis of AD, even before deposi-
tion of proteinaceous substances becomes manifest [10]. FBI has
significantly contributed to this so-called vascular hypothesis of
AD. Novel, easily applicable methods to quantify CBF changes in
cognitively healthy individuals at risk for AD have been imple-
mented in the clinical and research setting. ASL MRI and fMRI
have largely replaced SPECT and PET in evaluating resting perfu-
sion and task-dependent perfusion changes in pAD and AD.
There is another important point questioning the existing
model of pAD [1]. Both neuroimaging studies focusing on CBF
and studies focusing on metabolism in pAD have found alterations
of CBF and metabolism to be at largely independent from cerebral
Aß load (Table 1). Studies analyzing different biomarker profiles of
neurodegeneration (brain atrophy as measured with structural
MRI, 18F-FDG-PET, CSF-tau levels) and amyloid deposition
(PIB-/18F-florbetapir PET and CSF-Aß1–42) suggest that both, a
neurodegeneration-first and an amyloid-first pathophysiological
pathway, may later lead to AD [105]. It is difficult to distinguish
individuals with positive biomarkers for neurodegeneration, but
without Aß deposition (sNAP, suspected non-Alzheimer patho-
physiology) from individuals with advanced stages of pAD, as
defined by Sperling et al. [1] concerning their clinical and neuro-
imaging characteristics [39].
There are methodological limitations that may explain some of
the apparently contradictory findings in pAD, including inconsis-
tencies regarding increases or decreases of CBF and rCMRglc as
well as regional differences in CBF/rCMRglc alterations between
studies. Patient demographics with age and CRC, diagnostic crite-
ria, comorbid depression and anxiety, disease severity, genetic
traits, and differences in the cardiovascular risk profile may account
for some of the disparate findings in pAD [65, 90].
Table 1
FBI findings in models of preclinical AD
pADa MCI APOE-ɛ4 carriersb

Imaging of metabolism
• 18F-FDG-PET • Temporoparietal and PCC hypometabolism as • Comparable metabolic changes as seen in Hypometabolism in typical AD regions
•
typically seen in AD early AD • Rather independent from Aß
• Hypometabolism rather independent from Aß • Hypometabolism of PCC and precuneus
deposition • Temporoparietal and frontal
hypometabolism
Imaging of perfusion
• PET/SPECT rCBF decreases within the cingulate cortex
• • rCBF alterations in brain regions exhibiting • Frontal, temporal, and parietal rCBF
• rCBF changes rather associated with tau, not early pathological changes in AD reductions
Aß pathology • Compensatory hyperperfusion in different • Gene-dose effects on CBF
brain areas
• fMRI and • Large scale disconnections as seen in manifest • Alterations of large-scale functional brain • Altered functional connectivity of the
rs-fMRI AD networks hippocampus
• R educed DMN integrity • Decreased functional connectivity within • Disruptions of fc within the DMN
• Hyperconnectivity phase followed by several RSNs (i.e., the DMN)
hypoconnectivity phase in the course of pAD • Disruptions of whole-brain connectivity
within the PCC/precuneus region
• ASL MRI • Decrease of CBF along the continuum of Global CBF reductions in MCI
• • Alterations of rs CBF and CBF response to
pADa • MTL, PCC, and precuneus hypoperfusion memory encoding in MTL
• Both hypo- and hyperperfusion may reflect • Relation between rCBF changes and Aß • Age-dependent reductions of CBF in MTL
pathology in different stages of pAD depositions remains unclear • CBF reductions, independent from cerebral
Aß
Abbreviations: FBI functional brain imaging, pAD preclinical Alzheimer’s disease, MCI mild cognitive impairment, Aß amyloid-ß1–42, rCBF regional cerebral blood flow, RSNs resting-state
networks, DMN default mode network, MTL medial temporal lobe, PCC posterior cingulate cortex, SPECT single photon emission computer tomography, 18F-FDG-PET 18F-fluoro-
desoxy-glucose positron emission tomography, MRI magnetic resonance imaging, fMRI functional magnetic resonance imaging, rs-fMRI resting-state fMRI, ASL arterial spin labeling
a
Preclinical AD as defined by Sperling et al. [1]
b
Only results of nondemented patients are summarized
Concerning the methods presented here, there are also some

limitations for each technique. PET and SPECT have the disadvan-
tage of being expensive and time-consuming, often inaccessible to
clinicians, and associated with exposure to radiation. Studies using
ASL MRI and fMRI are in line with previous PET and SPECT
studies on brain perfusion, with MRI imaging being noninvasive,
cheaper, and more easily accessible in clinical practice. However,
ASL MRI has the disadvantage of showing low sensitivity, low
temporal resolution, and also contamination from the BOLD sig-
nal. Individuals with early pAD exhibit increased MTL resting
CBF, which has direct impact on the fMRI BOLD response.
BOLD changes largely depend on resting-state CBF, which can be
either increased or decreased in pAD. BOLD activations should
therefore be interpreted with caution and do not always reflect dif-
ferences in neuronal activation.
Neuroanatomically, CBF changes and metabolic alterations
largely affect the similar regions in AD, thereby indicating that
18
F-FDG-PET might in the future be replaced by MRI techniques
quantifying CBF changes [106].
Resting-state fc-fMRI imaging has shown that brain networks
and their connectivity degrade during the course of AD. There is
widespread loss of intranetwork as well as internetwork connectiv-
ity in AD. In pAD, studies on fc show heterogeneous results. There
is some evidence for loss of long distance internetwork connectiv-
ity with increased local short distance connectivity. Resting-state
fc-fMRI may also have the potential to evaluate therapeutic effects
of drugs in AD as well as in pAD [107].
In AD as well as in pAD, disruptions of the DMN have been
observed [76, 77, 108]. Interestingly, there is a significant degree
of overlap between typical hypometabolic brain regions in AD and
the DMN [17].
Concerning the models of pAD presented here, there are also
several limitations for each model. As described above, the defini-
tion of pAD requires a reappraisal concerning measures of CBF
and its sole dependency on the amyloid cascade hypothesis. There
is some evidence that the proneness for AD in APOE-ε4 allele car-
riers and in pAD is mediated independently from Aß. Studies in
nondemented carriers of the APOE-ε4 allele have, however, shown
fc alterations that were quite similar to nondemented subjects with
cerebral Aß, and both groups also resembled AD patients [88].
Furthermore, in a memory clinic setting, genetic testing is cur-
rently not recommended in the vast majority of patients, except for
very early dementia cases with a clear-cut family history of
dementia.
To date, there is insufficient knowledge about the role of SCD
as a model of pAD. Therefore, MCI still is the most relevant clini-
cal model. The aMCI subgroup has the highest risk to develop
clinically manifest dementia, but more than 50% of these patients

stay stable or even improve within 30 months [3]. Further bio-
marker support is required to identify MCI patients that later
develop AD. In clinical practice, CSF, not FBI biomarkers, is actu-
ally state of the art to detect those subjects prone to AD.
Memory performances, mainly hippocampus-dependent
declarative learning and delayed recall impairments, are hallmarks
of the neuropsychological profile of AD and can also be detected
in the pAD stage. These neuropsychological deficits are associated
with early metabolic and CBF changes within the MTL and tem-
poroparietal cortical areas [65]. Compensatory increases of CBF
and metabolism in the MTL are associated with better memory
performance in models of pAD, whereas decreases of MTL CBF
and metabolism are linked to cognitive dysfunction. Normal cog-
nitive status in the elderly despite biomarkers indicative of pAD
may reflect cognitive reserve capacity (CRC) of the aging brain
[68]. Years of education as a proxy for CRC should be taken into
account when applying FBI biomarkers in the elderly. Strategies
increasing the CRC are probably a fruitful approach to slow down
cognitive decline.
In the future, fully automated and validated brain-mapping
algorithms (e.g., based on machine learning techniques) will grad-
ually replace visual reading by clinicians. With ever-improving
technology, this development has yet reached FBI and also struc-
tural imaging methods. Morbelli et al. [109] described that visual
analyses of 18F-FDG-PET images by experts are the most accurate
method; however, an automated image analyzing system also
showed highly specific results, superior to moderately skilled
readers.
In conclusion, there is now an urgent need to improve the
diagnosis of pAD. The preexisting criteria are strongly relying on
the amyloid cascade hypothesis. As described in this chapter, CBF
and its measures need to be added to the existing criteria of pAD.
Furthermore, there is a need to quantify fc changes, and methods
to analyze fc need to be improved.
Finally, even when considering the abovementioned disadvan-
tages (i.e., exposure to radiation), 18F-FDG-PET still remains to
date the most established FBI biomarker in pAD.
Clinicians may best rely on a multimodal approach combining
clinical, neuropsychological, CSF, and neuroimaging-derived assess-
ments of biomarkers to detect AD, MCI, and pAD. Although there
is no causal, disease-modifying treatment for AD available to date,
individuals at risk of later developing AD can be advised to control
for cardiovascular risk factors, sedentary lifestyle, and exposure to
contact sports and leisure activities with a relevant risk of traumatic
brain injury [103].
References
1. Sperling RA, Aisen PS, Beckett LA et al lar disorder? Am J Cardiovasc Dis 3(4):
(2011) Toward defining the preclinical stages 197–226
of Alzheimer’s disease: recommendations 12. Trojanowski JQ, Growdon JH (1998) A new
from the National Institute on Aging- consensus report on biomarkers for the early
Alzheimer’s Association workgroups on diag- antemortem diagnosis of Alzheimer disease:
nostic guidelines for Alzheimer’s disease. current status, relevance to drug discovery,
Alzheimers Dement 7(3):280–292. https:// and recommendations for future research. J
doi.org/10.1016/j.jalz.2011.03.003 Neuropathol Exp Neurol 57(6):643–644
2. Jessen F, Amariglio RE, van Boxtel M et al 13. Wierenga CE, Hays CC, Zlatar ZZ (2014)
(2014) A conceptual framework for research Cerebral blood flow measured by arterial spin
on subjective cognitive decline in preclinical labeling MRI as a preclinical marker of
Alzheimer’s disease. Alzheimers Dement Alzheimer’s disease. J Alzheimers Dis
10(6):844–852. https://doi.org/10.1016/j. 42(Suppl 4):S411–S419. https://doi.
jalz.2014.01.001 org/10.3233/JAD-141467
3. Fischer P, Jungwirth S, Zehetmayer S et al 14. Binnewijzend MAA, Benedictus MR, Kuijer
(2007) Conversion from subtypes of mild cog- JPA et al (2016) Cerebral perfusion in the
nitive impairment to Alzheimer dementia. predementia stages of Alzheimer’s disease.
Neurology 68(4):288–291. https://doi. Eur Radiol 26(2):506–514. https://doi.
org/10.1212/01.wnl.0000252358.03285.9d org/10.1007/s00330-015-3834-9
4. Mitchell AJ, Shiri-Feshki M (2009) Rate of 15. Maier FC, Wehrl HF, Schmid AM et al (2014)
progression of mild cognitive impairment to Longitudinal PET-MRI reveals beta-amyloid
dementia—meta-analysis of 41 robust incep- deposition and rCBF dynamics and connects
tion cohort studies. Acta Psychiatr Scand vascular amyloidosis to quantitative loss of
119(4):252–265. https://doi. perfusion. Nat Med 20(12):1485–1492.
org/10.1111/j.1600-0447.2008.01326.x https://doi.org/10.1038/nm.3734
5. Yip AG, McKee AC, Green RC et al (2005) 16. Kato T, Inui Y, Nakamura A et al (2016)
APOE, vascular pathology, and the AD brain. Brain fluorodeoxyglucose (FDG) PET in
Neurology 65(2):259–265. https://doi. dementia. Ageing Res Rev 30:73–84.
o r g / 1 0 . 1 2 1 2 / 0 1 . https://doi.org/10.1016/j.arr.2016.02.003
wnl.0000168863.49053.4d 17. Granert O, Drzezga AE, Boecker H et al
6. Coon KD, Myers AJ, Craig DW et al (2007) (2015) Metabolic topology of neurodegen-
A high-density whole-genome association erative disorders: influence of cognitive and
study reveals that APOE is the major suscep- motor deficits. J Nucl Med 56(12):1916–
tibility gene for sporadic late-onset 1921. https://doi.org/10.2967/
Alzheimer’s disease. J Clin Psychiatry jnumed.115.156067
68(4):613–618 18. Matsuda H (2007) Role of neuroimaging in
7. Saunders AM, Strittmatter WJ, Schmechel D Alzheimer’s disease, with emphasis on brain
et al (1993) Association of apolipoprotein E perfusion SPECT. J Nucl Med 48(8):1289–
allele epsilon 4 with late-onset familial and 1300. https://doi.org/10.2967/
sporadic Alzheimer’s disease. Neurology jnumed.106.037218
43(8):1467–1472 19. Sakamoto S, Matsuda H, Asada T et al (2003)
8. Corder EH, Saunders AM, Strittmatter WJ et Apolipoprotein E genotype and early
al (1993) Gene dose of apolipoprotein E type Alzheimer’s disease: a longitudinal SPECT
4 allele and the risk of Alzheimer’s disease in study. J Neuroimaging 13(2):113–123
late onset families. Science 20. Knopman DS, Jack CR, Wiste HJ et al (2014)
261(5123):921–923 18F-fluorodeoxyglucose positron emission
9. Biomarkers Definitions Working Group tomography, aging, and apolipoprotein E
(2001) Biomarkers and surrogate endpoints: genotype in cognitively normal persons.
preferred definitions and conceptual frame- Neurobiol Aging 35(9):2096–2106. https://
work. Clin Pharmacol Ther 69(3):89–95. d o i . o r g / 1 0 . 1 0 1 6 / j .
https://doi.org/10.1067/mcp.2001.113989 neurobiolaging.2014.03.006
10. Hays CC, Zlatar ZZ, Wierenga CE (2016) 21. Nasrallah I, Dubroff J (2013) An overview of
The utility of cerebral blood flow as a bio- PET neuroimaging. Semin Nucl Med
marker of preclinical Alzheimer’s disease. Cell 43(6):449–461. https://doi.org/10.1053/j.
Mol Neurobiol 36(2):167–179. https://doi. semnuclmed.2013.06.003
org/10.1007/s10571-015-0261-z 22. Buxton RB (2009) Introduction to func-
11. Kelleher RJ, Soiza RL (2013) Evidence of tional magnetic resonance imaging: principles
endothelial dysfunction in the development and techniques, 2nd edn. Cambridge
of Alzheimer’s disease: is Alzheimer’s a vascu- University Press, Cambridge
23. Damoiseaux JS, Rombouts SARB, Barkhof F 34. Ewers M, Brendel M, Rizk-Jackson A et al
et al (2006) Consistent resting-state networks (2014) Reduced FDG-PET brain metabolism
across healthy subjects. Proc Natl Acad Sci U and executive function predict clinical pro-
S A 103(37):13848–13853. https://doi. gression in elderly healthy subjects.
org/10.1073/pnas.0601417103 Neuroimage Clin 4:45–52. https://doi.
24. Guerra-Carrillo B, Mackey AP, Bunge SA org/10.1016/j.nicl.2013.10.018
(2014) Resting-state fMRI: a window into 35. Ewers M, Insel PS, Stern Y et al (2013)
human brain plasticity. Neuroscientist Cognitive reserve associated with FDG-PET
20(5):522–533. https://doi.org/10.1177/ in preclinical Alzheimer disease. Neurology
1073858414524442 80(13):1194–1201. https://doi.
25. Chuang K-H, van Gelderen P, Merkle H et al org/10.1212/WNL.0b013e31828970c2
(2008) Mapping resting-state functional con- 36. Knopman DS, Jack CR, Wiste HJ et al (2013)
nectivity using perfusion MRI. Neuroimage Selective worsening of brain injury biomarker
40(4):1595–1605. https://doi. abnormalities in cognitively normal elderly
org/10.1016/j.neuroimage.2008.01.006 persons with β-amyloidosis. JAMA Neurol
26. Biswal BB, van Kylen J, Hyde JS (1997) 70(8):1030–1038. https://doi.
Simultaneous assessment of flow and BOLD org/10.1001/jamaneurol.2013.182
signals in resting-state functional connectivity 37. Wirth M, Oh H, Mormino EC et al (2013)
maps. NMR Biomed 10(4–5):165–170 The effect of amyloid β on cognitive decline is
27. Watanabe T, Kan S, Koike T et al (2014) modulated by neural integrity in cognitively
Network-dependent modulation of brain normal elderly. Alzheimers Dement 9(6):687–
activity during sleep. Neuroimage 98:1–10. 698.e1. https://doi.org/10.1016/j.
https://doi.org/10.1016/j.neuroimage. jalz.2012.10.012
2014.04.079 38. Ossenkoppele R, van der Flier WM, Verfaillie
28. Greicius MD, Srivastava G, Reiss AL et al SCJ et al (2014) Long-term effects of amyloid,
(2004) Default-mode network activity distin- hypometabolism, and atrophy on neuropsy-
guishes Alzheimer’s disease from healthy chological functions. Neurology 82(20):
aging: evidence from functional MRI. Proc 1768–1775. https://doi.org/10.1212/
Natl Acad Sci U S A 101(13):4637–4642. WNL.0000000000000432
https://doi.org/10.1073/pnas.0308627101 39. Knopman DS, Jack CR, Wiste HJ et al (2013)
29. Liu TT, Brown GG (2007) Measurement of Brain injury biomarkers are not dependent on
cerebral perfusion with arterial spin labeling: β-amyloid in normal elderly. Ann Neurol
Part 1. Methods. J Int Neuropsychol Soc 73(4):472–480. https://doi.org/10.1002/
13(3):517–525. https://doi.org/10.1017/ ana.23816
S1355617707070646 40. Seo EH, Kim SH, Park SH et al (2016)
30. Lee SP, Duong TQ, Yang G et al (2001) Topographical APOE ɛ4 genotype influence
Relative changes of cerebral arterial and on cerebral metabolism in the continuum of
venous blood volumes during increased cere- Alzheimer’s disease: amyloid burden adjusted
bral blood flow: implications for BOLD analysis. J Alzheimers Dis 54(2):559–568.
fMRI. Magn Reson Med 45(5):791–800 https://doi.org/10.3233/JAD-160395
31. Parkes LM, Rashid W, Chard DT et al (2004) 41. Yi D, Lee DY, Sohn BK et al (2014) Beta-
Normal cerebral perfusion measurements amyloid associated differential effects of
using arterial spin labeling: reproducibility, APOE ε4 on brain metabolism in cognitively
stability, and age and gender effects. Magn normal elderly. Am J Geriatr Psychiatr
Reson Med 51(4):736–743. https://doi. 22(10):961–970. https://doi.
org/10.1002/mrm.20023 org/10.1016/j.jagp.2013.12.173
32. Besson FL, La Joie R, Doeuvre L et al (2015) 42. Krell-Roesch J, Ruider H, Lowe VJ et al
Cognitive and brain profiles associated with (2016) FDG-PET and neuropsychiatric
current neuroimaging biomarkers of preclini- symptoms among cognitively normal elderly
cal Alzheimer’s disease. J Neurosci persons: the Mayo Clinic study of aging. J
35(29):10402–10411. https://doi. Alzheimers Dis 53(4):1609–1616. https://
org/10.1523/JNEUROSCI.0150-15.2015 doi.org/10.3233/JAD-160326
33. Teipel S, Grothe MJ (2016) Does posterior 43. Didic M, Felician O, Gour N et al (2015)
cingulate hypometabolism result from dis- Rhinal hypometabolism on FDG PET in
connection or local pathology across preclini- healthy APO-E4 carriers: impact on memory
cal and clinical stages of Alzheimer’s disease? function and metabolic networks. Eur J Nucl
Eur J Nucl Med Mol Imaging 43(3): Med Mol Imaging 42(10):1512–1521.
526–536. https://doi.org/10.1007/ https://doi.org/10.1007/s00259-015-
s00259-015-3222-3 3057-y
44. Arenaza-Urquijo EM, Gonneaud J, Fouquet controls. A European Alzheimer’s Disease

M et al (2015) Interaction between years of Consortium (EADC) study. Neuroimage
education and APOE ε4 status on frontal and Clin 7:34–42. https://doi.org/10.1016/j.
temporal metabolism. Neurology nicl.2014.11.007
85(16):1392–1399. https://doi. 53. Espinosa A, Alegret M, Pesini P et al (2017)
org/10.1212/WNL.0000000000002034 Cognitive composites domain scores related
45. Yao Z, Hu B, Zheng J et al (2015) A FDG- to neuroimaging biomarkers within probable-
PET study of metabolic networks in apolipo- amnestic mild cognitive impairment-storage
protein E ε4 allele carriers. PLoS One subtype. J Alzheimers Dis 57(2):447–459.
10(7):e0132300. https://doi.org/10.1371/ https://doi.org/10.3233/JAD-161223
journal.pone.0132300 54. Franzmeier N, Duering M, Weiner M et al
46. Chen K, Ayutyanont N, Langbaum JBS et al (2017) Left frontal cortex connectivity under-
(2012) Correlations between FDG PET glu- lies cognitive reserve in prodromal Alzheimer
cose uptake-MRI gray matter volume scores disease. Neurology 88(11):1054–1061.
and apolipoprotein E ε4 gene dose in cogni- https://doi.org/10.1212/WNL.
tively normal adults: a cross-validation study 0000000000003711
using voxel-based multi-modal partial least 55. Förster S, Buschert VC, Teipel SJ et al (2011)
squares. Neuroimage 60(4):2316–2322. Effects of a 6-month cognitive intervention
https://doi.org/10.1016/j.neuroimage. on brain metabolism in patients with amnestic
2012.02.005 MCI and mild Alzheimer’s disease. J
47. Chételat G, Eustache F, Viader F et al (2005) Alzheimers Dis 26(Suppl 3):337–348.
FDG-PET measurement is more accurate https://doi.org/10.3233/JAD-2011-0025
than neuropsychological assessments to pre- 56. Chen K, Langbaum JBS, Fleisher AS et al
dict global cognitive deterioration in patients (2010) Twelve-month metabolic declines in
with mild cognitive impairment. Neurocase probable Alzheimer’s disease and amnestic
11(1):14–25. https://doi. mild cognitive impairment assessed using an
org/10.1080/13554790490896938 empirically pre-defined statistical region-of-
48. Morbelli S, Piccardo A, Villavecchia G et al interest: findings from the Alzheimer’s disease
(2010) Mapping brain morphological and neuroimaging initiative. NeuroImage
functional conversion patterns in amnestic 51(2):654–664. https://doi.org/10.1016/j.
MCI: a voxel-based MRI and FDG-PET neuroimage.2010.02.064
study. Eur J Nucl Med Mol Imaging 57. Clerici F, Del Sole A, Chiti A et al (2009)
37(1):36–45. https://doi.org/10.1007/ Differences in hippocampal metabolism
s00259-009-1218-6 between amnestic and non-amnestic MCI
49. Nobili F, Salmaso D, Morbelli S et al (2008) subjects: automated FDG-PET image analysis.
Principal component analysis of FDG PET in Q J Nucl Med Mol Imaging 53(6):646–657
amnestic MCI. Eur J Nucl Med Mol Imaging 58. Pagani M, Dessi B, Morbelli S et al (2010)
35(12):2191–2202. https://doi. MCI patients declining and not-declining at
org/10.1007/s00259-008-0869-z mid-term follow-up: FDG-PET findings.
50. Coutinho AMN, Porto FHG, Duran FLS et Curr Alzheimer Res 7(4):287–294
al (2015) Brain metabolism and cerebrospinal 59. Teune LK, Bartels AL, Jong BM d et al
fluid biomarkers profile of non-amnestic mild (2010) Typical cerebral metabolic patterns in
cognitive impairment in comparison to neurodegenerative brain diseases. Mov Disord
amnestic mild cognitive impairment and nor- 25(14):2395–2404. https://doi.
mal older subjects. Alzheimers Res Ther org/10.1002/mds.23291
7(1):58. https://doi.org/10.1186/ 60. Donnemiller E, Heilmann J, Wenning GK et
s13195-015-0143-0 al (1997) Brain perfusion scintigraphy with
51. Bailly M, Destrieux C, Hommet C et al 99mTc-HMPAO or 99mTc-ECD and
(2015) Precuneus and cingulate cortex atro- 123I-beta-CIT single-photon emission
phy and hypometabolism in patients with tomography in dementia of the Alzheimer-
Alzheimer’s disease and mild cognitive type and diffuse Lewy body disease. Eur J
impairment: MRI and (18)F-FDG PET Nucl Med 24(3):320–325
quantitative analysis using FreeSurfer. Biomed 61. Herholz K, Salmon E, Perani D et al (2002)
Res Int 2015:583931. https://doi. Discrimination between Alzheimer dementia
org/10.1155/2015/583931 and controls by automated analysis of multi-
52. Pagani M, Carli F d, Morbelli S et al (2015) center FDG PET. Neuroimage 17(1):
Volume of interest-based 18Ffluorode 302–316
oxyglucose PET discriminates MCI convert- 62. Meltzer CC, Zubieta JK, Brandt J et al (1996)
ing to Alzheimer’s disease from healthy Regional hypometabolism in Alzheimer’s dis-
ease as measured by positron emission tomog- Alzheimer’s disease biomarkers. Neurobiol

raphy after correction for effects of partial Aging 36(5):1771–1779. https://doi.
volume averaging. Neurology org/10.1016/j.neurobiolaging.2015.01.019
47(2):454–461 73. Sheline YI, Raichle ME, Snyder AZ et al
63. Tam CWC, Burton EJ, McKeith IG et al (2010) Amyloid plaques disrupt resting state
(2005) Temporal lobe atrophy on MRI in default mode network connectivity in cogni-
Parkinson disease with dementia: a compari- tively normal elderly. Biol Psychiatry
son with Alzheimer disease and dementia 67(6):584–587. https://doi.org/10.1016/j.
with Lewy bodies. Neurology 64(5):861– biopsych.2009.08.024
865. https://doi.org/10.1212/01. 74. Schultz AP, Chhatwal JP, Hedden T et al
WNL.0000153070.82309.D4 (2017) Phases of hyperconnectivity and hypo-
64. Buckner RL, Andrews-Hanna JR, Schacter connectivity in the default mode and salience
DL (2008) The brain’s default network: anat- networks track with amyloid and tau in clini-
omy, function, and relevance to disease. Ann cally normal individuals. J Neurosci
N Y Acad Sci 1124:1–38. https://doi. 37(16):4323–4331. https://doi.
org/10.1196/annals.1440.011 org/10.1523/JNEUROSCI.3263-16.2017
65. Bangen KJ, Restom K, Liu TT et al (2009) 75. Brier MR, Thomas JB, Snyder AZ et al (2014)
Differential age effects on cerebral blood flow Unrecognized preclinical Alzheimer disease
and BOLD response to encoding: associations confounds rs-fcMRI studies of normal aging.
with cognition and stroke risk. Neurobiol Neurology 83(18):1613–1619. https://doi.
Aging 30(8):1276–1287. https://doi. org/10.1212/WNL.0000000000000939
org/10.1016/j.neurobiolaging.2007.11.012 76. Brier MR, Thomas JB, Fagan AM et al (2014)
66. Lee C, Lopez OL, Becker JT et al (2009) Functional connectivity and graph theory in
Imaging cerebral blood flow in the cogni- preclinical Alzheimer’s disease. Neurobiol
tively normal aging brain with arterial spin Aging 35(4):757–768. https://doi.
labeling: implications for imaging of neurode- org/10.1016/j.neurobiolaging.2013.10.081
generative disease. J Neuroimaging 77. Wang L, Brier MR, Snyder AZ et al (2013)
19(4):344–352. https://doi.org/10.1111/ Cerebrospinal fluid Aβ42, phosphorylated
j.1552-6569.2008.00277.x Tau181, and resting-state functional
67. Sandson TA, O’Connor M, Sperling RA et al connectivity. JAMA Neurol 70(10):1242–

(1996) Noninvasive perfusion MRI in 1248. https://doi.org/10.1001/jamaneurol.
Alzheimer’s disease: a preliminary report. 2013.3253
Neurology 47(5):1339–1342 78. Drzezga A, Becker JA, van Dijk KRA et al
68. Xekardaki A, Rodriguez C, Montandon M-L (2011) Neuronal dysfunction and disconnec-
et al (2015) Arterial spin labeling may con- tion of cortical hubs in non-demented sub-
tribute to the prediction of cognitive deterio- jects with elevated amyloid burden. Brain
ration in healthy elderly individuals. Radiology 134(Pt 6):1635–1646. https://doi.
274(2):490–499. https://doi.org/10.1148/ org/10.1093/brain/awr066
radiol.14140680 79. Mattsson N, Tosun D, Insel PS et al (2014)
69. Stomrud E, Forsberg A, Hägerström D et al Association of brain amyloid-beta with cere-
(2012) CSF biomarkers correlate with cerebral perfusion and structure in Alzheimer’s
bral blood flow on SPECT in healthy elderly. disease and mild cognitive impairment. Brain
Dement Geriatr Cogn Disord 33(2–3):156– 137(Pt 5):1550–1561. https://doi.
163. https://doi.org/10.1159/000338185 org/10.1093/brain/awu043
70. Sojkova J, Beason-Held L, Zhou Y et al 80. Thambisetty M, Beason-Held L, An Y et al
(2008) Longitudinal cerebral blood flow and (2010) APOE epsilon4 genotype and longi-
amyloid deposition: an emerging pattern? J tudinal changes in cerebral blood flow in nor-
Nucl Med 49(9):1465–1471. https://doi. mal aging. Arch Neurol 67(1):93–98.
org/10.2967/jnumed.108.051946 https://doi.org/10.1001/archneurol.
71. Edelman K, Tudorascu D, Agudelo C et al 2009.913
(2017) Amyloid-beta deposition is associated 81. Suwa A, Nishida K, Utsunomiya K et al
with increased medial temporal lobe activa- (2015) Neuropsychological evaluation and
tion during memory encoding in the cogni- cerebral blood flow effects of apolipoprotein
tively normal elderly. Am J Geriatr Psychiatry E4 in Alzheimer’s disease patients after one
25(5):551–560. https://doi.org/10.1016/j. year of treatment: an exploratory study.
jagp.2016.12.021 Dement Geriatr Cogn Dis Extra 5(3):414–
72. Gordon BA, Zacks JM, Blazey T et al (2015) 423. https://doi.org/10.1159/000440714
Task-evoked fMRI changes in attention 82. Lehtovirta M, Kuikka J, Helisalmi S et al
networks are associated with preclinical
(1998) Longitudinal SPECT study in
Alzheimer’s disease: relation to apolipopro- 92. Kim SM, Kim MJ, Rhee HY et al (2013)
tein E polymorphism. J Neurol Neurosurg Regional cerebral perfusion in patients with
Psychiatry 64(6):742–746 Alzheimer’s disease and mild cognitive
83. Lehtovirta M, Soininen H, Laakso MP et al impairment: effect of APOE epsilon4 allele.
(1996) SPECT and MRI analysis in Neuroradiology 55(1):25–34. https://doi.
Alzheimer’s disease: relation to apolipopro- org/10.1007/s00234-012-1077-x
tein E epsilon 4 allele. J Neurol Neurosurg 93. Beason-Held LL, Goh JO, An Y et al (2013)
Psychiatry 60(6):644–649 Changes in brain function occur years before
84. Devanand DP, van Heertum RL, Kegeles LS the onset of cognitive impairment. J
et al (2010) (99m)Tc hexamethyl-propylene- Neurosci 33(46):18008–18014. https://
aminoxime single-photon emission computed doi.org/10.1523/JNEUROSCI.
tomography prediction of conversion from 1402-13.2013
mild cognitive impairment to Alzheimer dis- 94. Adriaanse SM, Sanz-Arigita EJ, Binnewijzend
ease. Am J Geriatr Psychiatr 18(11):959–972. MAA et al (2014) Amyloid and its association
https://doi.org/10.1097/ with default network integrity in Alzheimer’s
JGP.0b013e3181ec8696 disease. Hum Brain Mapp 35(3):779–791.
85. Alegret M, Cuberas-Borrós G, Espinosa A et https://doi.org/10.1002/hbm.22213
al (2014) Cognitive, genetic, and brain perfu- 95. Binnewijzend MAA, Schoonheim MM, Sanz-
sion factors associated with four year inci- Arigita E et al (2012) Resting-state fMRI
dence of Alzheimer’s disease from mild changes in Alzheimer’s disease and mild cog-
cognitive impairment. J Alzheimers Dis nitive impairment. Neurobiol Aging
41(3):739–748. https://doi.org/10.3233/ 33(9):2018–2028. https://doi.
JAD-132516 org/10.1016/j.neurobiolaging.2011.07.003
86. Fleisher AS, Sherzai A, Taylor C et al (2009) 96. Liang P, Wang Z, Yang Y et al (2012) Three
Resting-state BOLD networks versus task- subsystems of the inferior parietal cortex are
associated functional MRI for distinguishing differently affected in mild cognitive impair-
Alzheimer’s disease risk groups. Neuroimage ment. J Alzheimers Dis 30(3):475–487.
47(4):1678–1690. https://doi. https://doi.org/10.3233/
org/10.1016/j.neuroimage.2009.06.021 JAD-2012-111721
87. Harrison TM, Burggren AC, Small GW et al 97. Ostergaard L, Aamand R, Gutierrez-Jimenez
(2016) Altered memory-related functional E et al (2013) The capillary dysfunction
connectivity of the anterior and posterior hip- hypothesis of Alzheimer’s disease. Neurobiol
pocampus in older adults at increased genetic Aging 34(4):1018–1031. https://doi.
risk for Alzheimer’s disease. Hum Brain Mapp org/10.1016/j.neurobiolaging.2012.09.011
37(1):366–380. https://doi.org/10.1002/ 98. Alexopoulos P, Sorg C, Forschler A et al
hbm.23036 (2012) Perfusion abnormalities in mild cog-
88. Sheline YI, Morris JC, Snyder AZ et al (2010) nitive impairment and mild dementia in
APOE4 allele disrupts resting state fMRI con- Alzheimer’s disease measured by pulsed arte-
nectivity in the absence of amyloid plaques or rial spin labeling MRI. Eur Arch Psychiatry
decreased CSF Aβ42. J Neurosci Clin Neurosci 262(1):69–77. https://doi.
30(50):17035–17040. https://doi.org/ org/10.1007/s00406-011-0226-2
10.1523/JNEUROSCI.3987-10.2010 99. Dai W, Lopez OL, Carmichael OT et al
89. Michels L, Warnock G, Buck A et al (2016) (2009) Mild cognitive impairment and
Arterial spin labeling imaging reveals wide- alzheimer disease: patterns of altered cerebral
spread and Abeta-independent reductions in blood flow at MR imaging. Radiology
cerebral blood flow in elderly apolipoprotein 250(3):856–866. https://doi.org/10.1148/
epsilon-4 carriers. J Cereb Blood Flow Metab radiol.2503080751
36(3):581–595. https://doi.org/10.1177/ 100. Johnson NA, Jahng G-H, Weiner MW et al
0271678X15605847 (2005) Pattern of cerebral hypoperfusion in
90. Bangen KJ, Restom K, Liu TT et al (2012) Alzheimer disease and mild cognitive impair-
Assessment of Alzheimer’s disease risk with ment measured with arterial spin-labeling MR
functional magnetic resonance imaging: an imaging: initial experience. Radiology
arterial spin labeling study. J Alzheimers Dis 234(3):851–859. https://doi.org/10.1148/
31(Suppl 3):S59–S74. https://doi.org/ radiol.2343040197
10.3233/JAD-2012-120292 101. Xu G, Antuono PG, Jones J et al (2007)

91. Wierenga CE, Clark LR, Dev SI et al (2013) Perfusion fMRI detects deficits in regional
Interaction of age and APOE genotype on CBF during memory-encoding tasks in MCI
cerebral blood flow at rest. J Alzheimers Dis subjects. Neurology 69(17):1650–1656.
34(4):921–935. https://doi.org/10.3233/ https://doi.org/10.1212/01.
JAD-121897 wnl.0000296941.06685.22
102. Xie L, Dolui S, Das SR et al (2016) A brain and frontotemporal dementia: two sides of
stress test: cerebral perfusion during memory the same coin? Eur Radiol 25(10):
encoding in mild cognitive impairment. 3050–3059. https://doi.org/10.1007/
Neuroimage Clin 11:388–397. https://doi. s00330-015-3696-1
org/10.1016/j.nicl.2016.03.002 107. Wang L, Day J, Roe CM et al (2014) The
103. Vanitallie TB (2013) Preclinical sporadic
effect of APOE ε4 allele on cholinesterase
Alzheimer’s disease: target for personalized inhibitors in patients with Alzheimer disease:
diagnosis and preventive intervention. Metab evaluation of the feasibility of resting state
Clin Exp 62(Suppl 1):S30–S33. https://doi. functional connectivity magnetic resonance
org/10.1016/j.metabol.2012.08.024 imaging. Alzheimer Dis Assoc Disord
104. Lee IH, Kim ST, Kim H-J et al (2010)
28(2):122–127. https://doi.org/10.1097/
Analysis of perfusion weighted image of CNS WAD.0b013e318299d096
lymphoma. Eur J Radiol 76(1):48–51. 108. Sheline YI, Raichle ME (2013) Resting state
https://doi.org/10.1016/j. functional connectivity in preclinical
ejrad.2009.05.013 Alzheimer’s disease. Biol Psychiatry
105. Jack CR, Wiste HJ, Weigand SD et al (2013) 74(5):340–347. https://doi.org/10.1016/j.
Amyloid-first and neurodegeneration-first biopsych.2012.11.028
profiles characterize incident amyloid PET 109. Morbelli S, Brugnolo A, Bossert I et al (2015)
positivity. Neurology 81(20):1732–1740. Visual versus semi-quantitative analysis of
https://doi.org/10.1212/01. 18F-FDG-PET in amnestic MCI: an
wnl.0000435556.21319.e4 European Alzheimer’s Disease Consortium
106. Verfaillie SCJ, Adriaanse SM, Binnewijzend (EADC) project. J Alzheimers Dis 44(3):
MAA et al (2015) Cerebral perfusion and 815–826. https://doi.org/10.3233/
glucose metabolism in Alzheimer’s disease JAD-142229
Chapter 9
Imaging of Brain Amyloid Load in Early Alzheimer’s

Disease
Timo Grimmer and Jennifer Grace Perryman
Abstract
The content of this chapter deals with the usefulness and indications of amyloid PET for the diagnosis of
patients with Alzheimer’s disease (AD). In addition to a description of amyloid PET itself, topics such as
early and differential diagnosis of AD are discussed. Measuring amyloid in CSF as an alternative method to
amyloid PET and the associations between cerebral amyloid load and the apolipoprotein E genotype are
also covered in this section.
Key words Alzheimer’s disease, Positron-emission tomography, Cerebrospinal fluid, Biomarker,

Early diagnosis, Prognosis, Apolipoprotein E
1 Introduction
As Alzheimer’s disease (AD) is the primary cause of dementia [1],

it seems intuitive that a method which is able to recognize, predict,
and differentiate this disease should be offered to patients world-
wide. The measuring of cerebral amyloid plaque load by use of
positron-emission tomography (PET) can achieve these aims.
Senile plaques and neurofibrillary tangles together with a loss
of neurons and synapses are the characteristic histopathological
features of AD [2, 3]. Amyloid β-protein (Aβ) is the main compo-
nent of these senile plaques. As known from cross-sectional histo-
pathological studies, cerebral amyloid load slowly increases with
the progression of AD, making its way from neocortical across
allocortical areas all the way to the cerebellum [3, 4]. In amyloid
PET, tracers like [11C]PiB [5] attach themselves to the amyloid
load, thereby allowing for its imaging. A more detailed description
of the process will be given in the methods section below. The
signal intensity measured by amyloid PET closely corresponds to
the extent of amyloid plaques as measured by histopathology [6];
149
150 Timo Grimmer and Jennifer Grace Perryman
hence, amyloid PET appears to be a valid method of measuring

amyloid plaque load in vivo.
In addition to supporting the diagnosis of AD, amyloid PET
could also be useful in three other major clinical scenarios: (1) to
differentiate between causes of dementia; (2) to diagnose AD at
pre-dementia stages, i.e., at early symptomatic or even asymptom-
atic stages; and (3) to rule out AD as an underlying cause in patients
and in persons with subjective and/or objective memory impair-
ment [7].
2 Implications and Utility of Amyloid Depiction via Positron-Emission

Tomography
2.1 Differential The value of amyloid PET for differential diagnosis varies between
Diagnosis the particular causes of dementia, depending on the occurrence of
cerebral amyloid load [7]. The first study that successfully used
amyloid PET to differentiate between semantic dementia, a sub-
type of the frontotemporal lobar degenerations (FTLD) [8] usu-
ally caused by non-amyloid pathologies, and AD yielded a perfect
group discrimination [9]. A similar result was achieved differenti-
ating AD from the behavioral variant of frontotemporal dementia
(bvFTD) [10], another FTLD subtype. However, distinguishing
AD from dementias associated with Parkinson’s disease, such as
Lewy body disease (LBD), may be difficult as heightened cerebral
amyloid plaque load can be found in both [11]. The usefulness of
amyloid PET for differential diagnosis is supported by a recent
meta-analysis confirming the associations between amyloid positiv-
ity as shown by amyloid imaging and clinical diagnosis [12]:
Patients with the clinical diagnosis of dementia due to AD were
93% amyloid positive at 50 years of age and 79% at 90 years of age.
Although this finding does not seem intuitive at first glance, there
are several explanations as to why a smaller percentage of amyloid
positivity prevalence can be found in older patients. The primary
cause may be the occurrence of rivaling causes for dementia in
older patients, resembling the clinical picture of AD. Although less
likely, it may be possible that either AD in elderly patients is caused
less by cerebral amyloid load or that the tracer binds less efficiently
to amyloid of older patients. The mean prevalence estimate for
patients with the clinical diagnosis of dementia due to AD was
88%. Patients with dementia due to Lewy bodies (DLB) were amy-
loid positive in 51% of cases, a figure compatible with the histo-
pathological finding that around 50% of patients with DLB show
amyloid depositions as discussed above. Patients with FTLD were
amyloid positive only in 12% of cases, which is compatible with the
histopathological finding that FTLD are caused by non-amyloid
pathology only.
Imaging of Brain Amyloid Load in Early Alzheimer’s Disease 151
Despite these convincing results, clinicians are advised to care-

fully evaluate the results of an amyloid PET. They must judge
whether the symptomatology is adequately explained by the brain
disease revealed through amyloid PET or perhaps an additional
cause.
2.2 Early Diagnosis Amyloid PET may also be used as a means to diagnose AD at a very
early stage as amyloid deposition precedes initial clinical symptoms
by years if not decades [13], with estimates giving a time period of
20–30 years between the first discovery of an abnormal amyloid
load and the beginning of dementia. In patients with mild cogni-
tive impairment (MCI), a predisposing factor and frequent precur-
sor of AD, a positive amyloid PET, for example, using the [11C]
PiB compound, can be found in 50% [1]. In accordance with the
assumption that amyloid precedes clinical symptoms, it was shown
that amyloid positivity in clinically asymptomatic persons increases
with age [12]. Usage of amyloid PET could therefore help dis-
cover the ailment and warn patients many decades prior to the
outbreak of symptoms.
As there is currently no treatment available for AD at an asymp-
tomatic stage, however, the advantages of early diagnosis are debat-
able. Nevertheless, it should be mentioned that early diagnosis
would allow patients to plan better for their future circumstances
regarding financial matters and caregiving and also inform them-
selves and relevant people about matters such as safety in everyday
life, driving, and counseling options [14].
2.3 Excluding As well as indicating a progression to AD, amyloid PET may also
Alzheimer’s Disease be used to rule out the disease as the cause of dementia: According
to the guidance documents of the German Society of Nuclear
Medicine and the American Society of Nuclear Medicine and
Molecular Imaging [15, 16], it is highly unlikely to suffer from
Alzheimer’s disease without cerebral amyloid load in the form of
plaques. Should the amyloid PET therefore be negative, one may
assume with a high probability that AD is not the cause for the
patient’s dementia.
It may also help reassure patients who are concerned about
developing the illness in the future. A survey by the Harvard School
of Public Health showed that around 89% of patients stated that
they would like to be informed whether or not AD was the cause
for their impairments [14].
2.4 Limitations Negative amyloid PET findings may be due to misdiagnosis; unsuc-
cessful tracer production may also be possible in some cases [7].
Another potential limitation is the availability of tracer produc-
tion/delivery and PET cameras. Furthermore, one should keep in
mind that the examination involves radiation exposure
(5.8–7.0 mSv).
2.5 Appropriate Use Amyloid PET is already commonly clinically used and recom-
Criteria mended in several guidelines (German Society of Nuclear
Medicine (DGN) [16], European Association of Nuclear
Medicine (EANM) [17], the American Society of Nuclear
Medicine and Molecular Imaging (SNMMI) [15], German S3
Guideline Dementia [18]). In order to provide a framework for
guiding clinicians who intend to apply amyloid PET, appropriate
use criteria for this particular method were published in 2013
[15]. For patients suffering from a subjectively and objectively
identified cognitive impairment, with AD as a likely, but as yet
undecided by experts, diagnosis, and assuming that a positive or
negative amyloid PET would provide assurance for the clinical
diagnosis, the authors concluded that amyloid PET is a useful
tool to draw on in the following situations:
1. The patient suffers from MCI in a persistent or progressively
unexplained form.
2. Patients with the clinical criteria for possible AD but with
unclear clinical presentation, e.g., an atypical clinical course or
etiologically mixed.
3. Patient with early-onset dementia.
The following scenarios are deemed as inappropriate for the
use of amyloid PET:
1. Patient fulfilling the clinical criteria for probable AD including
typical age of onset.
2. To determine the severity of the dementing syndrome.
3. Asymptomatic persons with a positive family history of demen-
tia or presence of apolipoprotein E (APOE) ε4 only.
4. The patient reports cognitive complaints which are, however,
without substantiation on clinical examination (memory
complainer).
5. In asymptomatic individuals.
6. For nonmedical use (e.g., legal insurance coverage or employ-
ment screening).
Unfortunately, the authors missed one important indication:
identifying participants for anti-amyloid therapy. In persons or
patients interested in participating in an anti-amyloid drug trial,
amyloid positivity needs to be confirmed in advance. This may be
achieved in vivo and noninvasively by amyloid PET.
Association of cerebral amyloid load and the apolipoprotein
E 4 genotype and its implications, as well as an alternative method
for assessing amyloid in vivo, shall be discussed in the notes sec-
tion below.
3 Materials
Authorized radiopharmaceuticals include florbetapir (Amyvid®)

[19], flutemetamol (Vizamyl®) [20], and florbetaben (Neuraceq®)
[21]. All show a decreased signal-to-noise ratio compared to [11C]
PiB but are comparable with regard to accuracy: 97% concordance
comparing PiB to florbetapir [22], 100% concordance comparing
PiB to flutemetamol [23], and 100% concordance comparing PiB
to florbetaben [24], respectively.
The comparatively long 110-min half-life of F-18-labeled PET
compounds allows for the use of amyloid PET in regular clinical
practice [15]. Production controls the quality of each product
before distribution. Refer to individual products for storage infor-
mation such as temperature and length. Information concerning
their toxic attributes and consequent safety measures may also be
found there. The compound [11C]PiB is only available at special
centers due to its short half-life of 20 min, which necessitates an
on-site cyclotron.
4 Method
In the following, the method of amyloid PET shall be described

using the [11C]PiB compound as an example (Fig. 1). The proce-
dure takes place as follows [9, 25]:
1. Examinations are performed on the Siemens ECAT HR PET
scanner (CTI, Knoxville, TN) and follow a simplified image
acquisition procedure [26] in order to avoid lengthy scanning
procedures and to minimize the risk of movement artifacts in
the cognitively impaired patients.
2. All patients are injected with 370 MBq [11C]PiB at rest outside
the scanner. Thirty minutes later, patients are placed in the
scanner.
3. At 40 min postinjection, three 10-min frames are started and
later summed into a single frame (40–70 min).
4. Acquisition is carried out in 3-D mode, and a transmission scan
is carried out to allow for later attenuation correction.
5. Images are reconstructed, corrected of dead time, scatter, and
attenuation, and standard uptake value ratio (SUVR) 40–70-
min images are generated.
6. Images are spatially normalized using SPM2 [27]. The [11C]PiB
data are first coregistered to each individual’s volumetric MRI
and then automatically spatially normalized to the T1 MRI
MNI template in SPM2 using warping parameters derived from
the individual MRI normalization performed previously [28].
Fig. 1 Amyloid PET using 11C-PiB in a patient with dementia due to Alzheimer’s
disease overlaid on an MRI scan, sagittal view
7. For the assessment of an amyloid PET using F18 tracers, the

FDA-approved evaluation algorithm should be used. Please
compare the instructions accompanying the packaged tracers
for further information on certified evaluation programs.
8. Concerning the evaluation of the PET scan, two different
methods may be drawn upon: visual and fully automated anal-
yses. The visual analysis may be aided by surface projection of
tracer uptake pattern [29]. The automated analysis is based on
thresholds of target (e.g., cerebral cortex) to reference tissue
(e.g., cerebellar vermis) ratios.
The effectiveness of both methods to predict the progression
of patients with MCI to manifest AD dementia within a 2-year
time frame was compared by Grimmer et al. [30]. Both methods
resulted in perfect negative predictive values (NPV) indicating that
a negative amyloid PET rules out the development of manifest AD
dementia within a manageable time frame. However, an MCI
patient with a positive amyloid PET does progress to AD dementia
in approximately 50% of cases, regardless of the method.
5 Associations with Other Biomarkers
As amyloid PET is not available everywhere, there is an alternative

way of measuring cerebral amyloid load through analyzing the
cerebrospinal fluid (CSF) obtained by spinal tap. Compared to
markers in CSF, PET shows only one of the typical histopathologi-
cal changes in AD [7]. However, one needs to keep in mind that
β-amyloid measured in CSF is not interchangeable with its mea-
surement in PET. The latter is a marker of lifetime amyloid deposi-

tion, while β-amyloid in CSF reflects current efflux rates. While
both measurements may be used interchangeably with regard to
diagnosis of AD [31–33], levels of amyloid in CSF and in PET are
not associated throughout the brain but in brain areas adjacent to
CSF only [34].
As mentioned above, there is an association between both the
amount [35] and the progression of cerebral amyloid load [4] and
the apolipoprotein E ε4 genotype (APOE-ε4) in AD. APOE-ε4 is
a known risk factor for AD [36], the assumption being that
APOE-ε4 transports amyloid through the brain-blood barrier [37]
and is responsible for higher levels of cerebral hypometabolism in
patients with AD that carry the gene as opposed to those who are
ε4-negative [35]. A gene dose effect between the ApoE genotype
and the uptake ratio of [11C]PiB has been suggested [4], which lies
in accordance with observations that there is a more rapid deterio-
ration of clinical symptoms shown by ApoE ε4 carriers [36, 38].
The findings of such studies may be of help when designing future
anti-amyloid treatment strategies [4].
References
1. Jansen WJ, Ossenkoppele R, Knol DL et al totemporal dementia. The Lund and
(2015) Prevalence of cerebral amyloid pathol- Manchester Groups. J Neurol Neurosurg
ogy in persons without dementia: a meta- Psychiatry 57:416–418
analysis. JAMA 313:1924–1938 9. Drzezga A, Grimmer T, Henriksen G et al
2. Braak H, .Braak E (1991) Neuropathological (2008) Imaging of amyloid plaques and cere-
stageing of Alzheimer-related changes. Acta bral glucose metabolism in semantic dementia
Neuropathol 82:239–259 and Alzheimer’s disease. Neuroimage
3. Thal DR, Rüb U, Orantes M, Braak H (2002) 39:619–633
Phases of Aß-deposition in the human brain 10. Engler H, Santillo AF, Wang SX et al (2008) In
and its relevance for the development of vivo amyloid imaging with PET in frontotem-
AD. Neurology 58:1791–1800 poral dementia. Eur J Nucl Med Mol Imaging
4. Grimmer T, Tholen S, Yousefi BH et al (2010) 35:100–106
Progression of cerebral amyloid load is associ- 11. McKeith IG, Dickson DW, Lowe J et al (2005)
ated with the apolipoprotein E epsilon4 geno- Diagnosis and management of dementia with
type in Alzheimer’s disease. Biol Psychiatry Lewy bodies: third report of the DLB
68:879–884 Consortium. Neurology 65:1863–1872
5. Klunk WE, Engler H, Nordberg A et al (2004) 12. Ossenkoppele R, Jansen WJ, Rabinovici GD
Imaging brain amyloid in Alzheimer’s disease et al (2015) Prevalence of amyloid PET positiv-
with Pittsburgh Compound-B. Ann Neurol ity in dementia syndromes: a meta-analysis.
55:306–319 JAMA 313:1939–1949
6. Ikonomovic MD, Klunk WE, Abrahamson EE 13. Bateman RJ, Xiong C, Benzinger TL et al
et al (2008) Post-mortem correlates of in vivo (2012) Clinical and biomarker changes in
PiB-PET amyloid imaging in a typical case of dominantly inherited Alzheimer’s disease. N
Alzheimer’s disease. Brain 131:1630–1645 Engl J Med 367:795–804
7. Grimmer T, Drzezga A, Kurz A (2010) 14. Blendon RJ, Benson JM, Wikler EM et al
Visualization of amyloid with positron emis- (2012) The impact of experience with a family
sion tomography. Useful improvement in the member with Alzheimer’s disease on views
diagnosis of dementia? Nervenarzt about the disease across five countries. Int
81:602–606 J Alzheimers Dis 2012:903645
8. Brun A, Englund B, Gustafson L et al (1994) 15. Johnson KA, Minoshima S, Bohnen NI et al
Clinical and neuropathological criteria for fron- (2013) Appropriate use criteria for amyloid
PET: a report of the Amyloid Imaging Task Compound B amyloid imaging PET studies: a
Force, the Society of Nuclear Medicine and comparative analysis. J Nucl Med
Molecular Imaging, and the Alzheimer’s 46:1959–1972
Association. J Nucl Med 54:476–490 27. Friston KJ, Holmes AP, Worsley KJ (1994)
16. Barthel, H., Meyer, P. T., Drzezga, A., Statistical parametric mapping in functional
Bartenstein, P., Boecker, H., Brust, P., Buchert, imaging: a general linear approach. Hum Brain
R., Coenen, H. H., La Fourgère, C., Gründer, Mapp 2:189–210
G., Grünwald, F., Krause, B. J., Kuwert, T., 28. Ziolko SK, Weissfeld LA, Klunk WE et al
Schreckenberger, M., Tatsch, K., Langen, K. J., (2006) Evaluation of voxel-based methods for
and Sabri, O. (2015) Beta- amyloid- PET- the statistical analysis of PIB PET amyloid
Bildgebung des Gehirns – DGN- imaging studies in Alzheimer’s disease.
Handlungsempfehlung (S1-Leitlinie). www. Neuroimage 33:94–102
awmf.org/uploads/tx_szleitlinien/031-052l_ 29. Minoshima S, Frey KA, Koeppe RA, Foster
S 1 _ B e t a _ A m y l o i d - P E T- B i l d g e b u n g _ NL, Kuhl DE (1995) A diagnostic approach in
Gehirn_2016-02.pdf Alzheimer’s disease using three-dimensional
17. Minoshima S, Drzezga AE, Barthel H et al stereotactic surface projections of fluorine-18-
(2016) SNMMI procedure standard/EANM FDG PET. J Nucl Med 36:1238–1248
practice guideline for amyloid PET imaging of 30. Grimmer T, Wutz C, Alexopoulos P et al (2016)
the brain 1.0. J Nucl Med 57:1316–1322 Visual versus fully automated analyses of 18F-
18. Deutsche Gesellschaft für Psychiatrie und FDG and amyloid PET for prediction of
Psychotherapie, Psychosomatik und dementia due to Alzheimer disease in mild cog-
Nervenheilkunde DGPPN and Deutsche nitive impairment. J Nucl Med 57:204–207
Gesellschaft für Neurologie (DGN) (2016) 31. Palmqvist S, Mattsson N, Hansson O (2016)
S3-Leitlinie “Demenzen”. https://www. Cerebrospinal fluid analysis detects cerebral
dgppn.de/fileadmin/user_upload/_medien/ amyloid-beta accumulation earlier than positron
download/pdf/kurzversion-leitlinien/REV_ emission tomography. Brain 139:1226–1236
S3-leiltlinie-demenzen.pdf
32. Fagan AM, Mintun MA, Shah AR et al (2009)
19. https://www.accessdata.fda.gov/drugsatfda_ Cerebrospinal fluid tau and ptau(181) increase
docs/label/2012/202008s000lbl.pdf. 18 Apr with cortical amyloid deposition in cognitively
2017 normal individuals: implications for future clin-
20. https://www.accessdata.fda.gov/drugsatfda_ ical trials of Alzheimer’s disease. EMBO Mol
docs/label/2016/203137s005lbl.pdf. 18 Apr Med 1:371–380
2017 33. Mattsson N, Insel PS, Landau S et al (2014)
21. https://www.accessdata.fda.gov/drugsatfda_ Diagnostic accuracy of CSF Ab42 and florbeta-
docs/label/2014/204677s000lbl.pdf. 18 Apr pir PET for Alzheimer’s disease. Ann Clin
2017 Transl Neurol 1:534–543
22. Landau SM, Breault C, Joshi AD et al (2013) 34. Grimmer T, Riemenschneider M, Forstl H et al
Amyloid-beta imaging with Pittsburgh com- (2009) Beta amyloid in Alzheimer’s disease:
pound B and florbetapir: comparing radiotrac- increased deposition in brain is reflected in
ers and quantification methods. J Nucl Med reduced concentration in cerebrospinal fluid.
54:70–77 Biol Psychiatry 65:927–934
23. Vandenberghe R, Van Laere K, Ivanoiu A et al 35. Drzezga A, Grimmer T, Henriksen G et al
(2010) 18F-flutemetamol amyloid imaging in (2009) Effect of APOE genotype on amyloid
Alzheimer disease and mild cognitive impair- plaque load and gray matter volume in
ment: a phase 2 trial. Ann Neurol 68:319–329 Alzheimer disease. Neurology 72:1487–1494
24. Villemagne VL, Mulligan RS, Pejoska S et al 36. Cosentino S, Scarmeas N, Helzner E et al
(2012) Comparison of 11C-PiB and (2008) APOE epsilon 4 allele predicts faster
18F-florbetaben for Abeta imaging in ageing cognitive decline in mild Alzheimer disease.
and Alzheimer’s disease. Eur J Nucl Med Mol Neurology 70:1842–1849
Imaging 39:983–989 37. Deane R, Sagare A, Hamm K et al (2008) apoE
25. Grimmer T, Henriksen G, Wester HJ et al isoform-specific disruption of amyloid beta
(2009) Pittsburgh Compound B(PIB) uptake peptide clearance from mouse brain. J Clin
in positron emission tomography is associated Invest 118:4002–4013
with clinical severity of Alzheimer’s disease. 38. Craft S, Teri L, Edland SD et al (1998)
Neurobiol Aging 30(12):1902–1909 Accelerated decline in apolipoprotein
26. Lopresti BJ, Klunk WE, Mathis CA et al (2005) E-epsilon4 homozygotes with Alzheimer’s dis-
Simplified quantification of Pittsburgh ease. Neurology 51:149–153
Chapter 10
Functional EEG Connectivity Alterations in Alzheimer’s

Disease
Florian Hatz and Peter Fuhr
Abstract
Quantitative EEG analyses, especially connectivity and graph analyses, are promising biomarkers in
neurodegenerative disorders such as Alzheimer’s disease. To develop reliable and valid biomarkers,
many aspects in pre- and post-processing of EEG data are important.
Artifact detection and segment selection are done by visual inspection or automated routines, both
with specific advantages and difficulties. Inverse solution is an elegant way of solving the issue of common
source but may be somewhat imprecise in the absence of defined graphoelements such as epileptiform
discharges. Connectivity analyses allow a description of interactions between brain regions, which may
represent one of the earliest signs of cognitive decline.
Key words Alzheimer’s disease, Dementia, Quantitative electroencephalography, Functional connectivity,

Network analysis
1 Introduction
Quantitative EEG (qEEG) analyses, including functional connec-

tivity analyses, are candidate biomarkers for various neurodegen-
erative disorders, especially to characterize physiological alterations
related to cognitive deterioration [1–3]. Many different methods
for quantification of EEG connectivities have been applied, most
focusing on analyses in resting state. For resting state EEG record-
ing, subjects sit or lay calm with eyes closed, not concentrating on
anything yet without falling asleep. Analysis of these EEG record-
ings requires pre- and post-processing and is influenced by both
technical and physiological artifacts.
157
158 Florian Hatz and Peter Fuhr
2 Preprocessing of EEG Data
Especially the selection of segments for analysis highly influences the

results. The main technical artifacts of EEG arise from line noise (50
or 60 Hz) and from noisy or broken electrodes. Line noise can be
reduced by notch filters, while noisy or broken electrodes must be
detected and eliminated by replacing their signals using interpolation
from nearby intact electrodes. High-density EEG recording is best
for this procedure. The interpolation of signals is important, as the
simple exclusion of noisy electrodes would change the number of
evaluated electrodes per subject, thereby introducing difficulties for
the statistical analysis [4]. Physiological artifacts arise mainly from eye
movements, blinking, muscle activity, and h eartbeats. They can be
reduced or eliminated by selecting visually the least affected channels
and segments. Physiological artifacts can also be isolated and
eliminated by independent component analysis (ICA) or regression
to electrooculographic (EOC) or e lectrocardiographic (ECG) signals
[5]. These methods generally rely on a personal intervention involving
a visual search for artifacts. That process is inherently subjective and
hard to standardize across raters and studies. To eliminate the
subjectivity, automated p rocessing has been proposed. Nolan et al.
[6] introduced the toolbox FASTER for the automated preprocessing
of event-related potentials (ERP) (FASTER; Nolan et al. [6])
including an ICA for artifact reduction and an automated detection
of bad channels, epochs, and ICA activations. The detection routines
are based on median, kurtosis, correlation, and variance analyses of
the channels. Other toolboxes such as Cartool [7] and Brainstorm
[8] also include automated routines for bad-channel detection; no
validation of their underlying routines is currently available. The

widely known toolbox FieldTrip [9] uses frequency analysis for the
automatic selection of bad channels. To our knowledge, the validity
of this procedure is currently also not available, though it has been
recently suggested that MEG processing should include frequency
analysis of the recorded data for rating bad channels and segments
[4]. Still another possibility is using the toolbox TAPEEG [10],
which combines many of the previously mentioned methods and
allows a fully automated preprocessing (sites.google.com/site/-tap-
eeg; see Fig. 1). Some important aspects of preprocessing include:
Fig. 1 (continued) from the resulting values, and signals with a z-value >3 are labeled as “bad.” Results are
summarized in a matrix (lower part of the figure). All signal periods labeled as “bad” by one of the methods are
represented in red, the others in green color. For selection of bad channels or activations, the resulting matrix
is evaluated horizontally (lower right matrix in figure, threshold: >70% bad signals). For selection of segments,
the green squares (= good signals) in the resulting matrix are replaced by Peak2Min values (brighter green
colors represent higher Peak2Min values), the red squares by zero values. Maximizing the summed values of
the squares allows the automated selection of the optimal segments for further processing (lower left matrix
in figure; blue frame represents a selected segment)
Fig. 1 Automated selection of segments, bad channels, and ICA activations in TAPEEG. TAPEEG includes
routines to automatically select segments of EEG for processing and the detection of channels with high arti-
fact load (bad channels). For artifact reduction, the automated selection of bad ICA activations is possible. For
all three tasks, TAPEEG processes the EEG data similarly, using the workflow depicted in the figure. EEG data
is first splitted into periods (e.g. 4 s). In every period, the signals (= channels) are analyzed using a set of
predefined methods as listed in the central boxes of the figure. Every method evaluates the signals
independently. FFT = Fast Fourier transformation using an 80% Hanning window. Band power 0.5–2 Hz > 70%,
band power 15–30 Hz > 50%, and 50 Hz-power > 50% are labeled as bad. Hurst exponent = Hurst exponent
is a measure of long-range dependence within a signal, with a range of 0–1. Human EEG has values around 0.7
[41]. Topo correlation = Correlation to surrounding channels (high correlation values are expected for high density
EEG). Median gradient = Median value of all 1 timeframe differences in amplitude of every channel.
Kurtosis = Kurtosis measures the peakedness of data and is high for ICA activations loading on one single elec-
trode only. EOG correlation = Correlation to electrooculogram. ECG correlation = Correlation of peaks in signals to
peaks in electrocardiogram. Peak2Min = Ratio of amplitude of peak frequency and minimal amplitude in lower
frequency for every period in parieto-occipital electrodes (amplitude of IAF divided by amplitude of TF in ).
Peak2Min represents the amount of alpha activity and is positively correlated to alertness [10]. For Hurst expo-
nent, variance, topo correlation, median gradient, kurtosis, ECG, and EOG correlation, z-values are c alculated
1. Of a recording of several minutes of resting state EEG, usually

only a part can be included in further analysis, while time peri-
ods with signs of sleepiness, movement, or other less frequent
artifacts have to be eliminated. Visual rating has been the stan-
dard method but carries a problem of reliability, as raters are
influenced by many unpredictable circumstances such as
changes in concentration or interference with other rated EEG
in close temporal relationship. In contrast, an automated rou-
tine for selecting segments is not able to recognize as many
relevant EEG features as an experienced reader. The coarser
but highly standardized selection of segments as used in auto-
mated systems might be a fair compromise if identification of
specific graphoelements is less important. Regarding results of
a frequency analysis, reliability of automated preprocessing is
similar to preprocessing by trained raters, while it is less time-
consuming, completely standardized, and independent of rat-
ers and their training [6, 10].
2. For quantitative EEG analyses, it is important to differentiate
between task-related or evoked analyses and analyses in resting
state. In resting state EEG recording, laboratory conditions
are simple and well defined, while the physiological condition
is not, as patients may be tense or relaxed or may concentrate
on very different aspects of their life. In contrast, non-rest
EEG conditions are more difficult to standardize across differ-
ent laboratories, but the physiological condition is quite well
defined. Brain a ctivity can be characterized by looking at the
power of different frequency bands. Traditionally, the total
band power is subdivided in the delta (1–4 Hz), theta (4–8 Hz),
alpha1 (8–10 Hz), alpha2 ( 10–13 Hz), and beta (13–30 Hz)
bands [11]. Delta power is the dominant frequency of deep
non-REM sleep but is also associated with functional cortical
deafferentation during internal concentration when solving
mental tasks [12]. Alpha and theta power respond in different
and opposite ways. Relative theta power increases with increas-
ing task demands, whereas relative alpha power decreases [13].
Attentional tasks and semantic m emory demands lead to a
selective suppression of alpha power in different “sub-bands.”
For resting state, increase of alpha2 power has been linked to
an increased capacity to initiate new tasks [14]. Beta band
power may also represent suppression of cortical activity, as
shown for the motor cortex during motor activity [15].
Apart from separating the frequency according to the
standardized frequency bands, frequency bands can be defined

according to the individual alpha frequency (= peak frequency, IAF)
and the transition frequency between theta and alpha power [14]:
Band power results defined by IAF represent the brain activity of
a subject in a far more physiological way. However, results in groups
EEG Connectivity in AD 161
Fig. 2 Calculation of individual frequency bands according to Klimesch 1999. IAF = individual alpha frequency,
TF = transition frequency between theta and alpha
of subjects may differ by applying individual or predefined bands, and

an intraindividual shift in peak frequency may go undetected with the
IAF method. Furthermore, the IAF method is not applicable when
no peak is clearly identifiable.
3 Connectivity Analyses
In addition to the band powers of different brain regions, the

interaction of brain activity between regions is also of interest.
Connectivity can be characterized as “functional” or “effective”.
Functional connectivity denotes any kind of connection between
two locations independently from direction or causal relation. For
functional connectivity, the linkage can be direct, indirect, or due
to a common trigger. Effective connectivity is defined as a connec-
tion where location A influences location B and refers to the influ-
ence that one neural system exerts over another [16]. Connectivities
are best described separately for the different frequency bands, as
in theory only cell populations that are synchronized in phase are
able to c ommunicate and interact [17]. This fact raises the prob-
lem of v olume conduction and of the common source. Activity of
deeper brain regions is recorded simultaneously by many electrodes,
and activity at a single electrode is always measured against the

common reference electrode. Therefore, brain activity recorded at
the reference electrodes explains a large part of the recorded signal,
and this part is the same for all recording electrodes. Measuring
connectivity using simple coherence (= correlation of frequency
spectra) overestimates connectivities, and the strong common
source effect masks real connectivities.
Mainly two approaches have been used to solve this problem:
Different connectivity measures were developed for reducing or
eliminating effects of the common source. Nolte et al. [18] proposed
a connectivity measure based on the imaginary part of the coherence,
not being influenced by effects of common source [18]. However, as
the imaginary part depends on the amplitude of the signal and
magnitude of the phase delay between two signals, this measure is not
ideal [19]. Other proposed measures rely on the phase information
only, mainly using Hilbert transformation. Stam et al. [20] proposed
the phase lag index (PLI), measuring the amount of stable phase lags
between two signals. This method has the advantage of disregarding
zero-lag phase difference and, therefore, annihilating effects of
common source recording and volume conduction at the cost of
neglecting the possibility of synchronous activation of two brain
regions by a common pacemaker. Mutual information and the closely
related synchronization likelihood [19] are measuring the degree of
interaction between two or more time series, the former using entropy
and the latter dynamical interdependencies. However, SL is also
strongly influenced by the common source problem. Furthermore, all
previously mentioned methods are not detecting the direction of
information flow and thus are by definition measures of functional
connectivity. Measures of effective connectivity such as Granger
causality [21] and direct transfer function [22] are using multivariate
auto-regressive models to estimate connectivity strength and direction
between different brain regions.
4 Inverse Solutions
A second method to reduce the problems of common source and of

volume conduction is the application of inverse solution methods. In
theory, inverse solution methods are transposing the signals recorded
at the surface back to their origins. A forward solution model is used
to describe the electrical conductivity between different points in the
head and the electrodes on the surface. Based on such models,
different methods of inverse solution try to reconstruct the source of
the signals recorded at the surface. Inverse solution methods carry
some uncertainties. First, no perfect forward solution model exists.
Forward solutions are calculated based on a MRI scan of the head,
while conductivity values are defined for the different compartments.
The most important compartment is the skull, having the lowest

conductivity. Unfortunately, a good estimation of the skull conductivity
is difficult to obtain for single individuals. Therefore, all forward solu-
tions can only be seen as a good approximation. Second, the methods
to reconstruct sources based on signals recorded at the surface and a
forward solution model have no unique solution. Mainly two m ethods
are used today. One is the minimum norm estimation or its derivatives
(as Loreta or sLoreta), trying to minimize the amount of power to
explain all activity recorded at the surface while using a correction for
deeper sources [23]. The other commonly used method is beamform-
ing, using radar technology to locate the source of a s ignal recorded at
many surface electrodes [24]. This method has a higher spatial resolu-
tion than minimum norm estimation but needs a very good estima-
tion of the covariance matrix, which is often difficult to determine
from the recorded EEG data. In summary, inverse solution transfor-
mation is interesting, allows analysis of source signals, and reduces the
effect of common source and volume conduction. However, due to
the methodological issues, results have always to be taken with cau-
tion, especially for analysis in resting state EEG. In contrast, in epilep-
tology spikes simplify the source localization.
Newer methods of connectivity analysis are trying to take the
non-stationarity of EEG recordings during resting state into
account. Examples are the analysis of synchronization dynamics
using the SL [25], the microstate segmented PLI [26], or the
correlation of connectivities over time [27].
5 Graph Analysis
Based on results of connectivity analysis, graph analysis allows the

characterization of networks by descriptive single measures.

Common measures are gamma, lambda, and degree diversity.
Gamma represents the amount of local connectivity, given by the
number of connections between immediate neighbors of a location
or “node”; lambda the efficiency of information flow, given by the
shortest path between nodes; and degree diversity the scale-free
distribution, given by a h ierarchical organization of many nodes
with few connections and few hubs with many. Higher scale-free
distribution is linked to the amount of hubs in a network [28].
Stam et al. described [29] a model of p hysiological brain networks,
showing the “normal brain” as an intermediate network state
between a regular, a random, and a s cale-free network with a small-
world characteristic. According to Watts and Strogatz, the small-
world network is a network with many local c onnections and a few
long-distance connections, resulting in an enhanced signal propa-
gation speed, computational power, and synchronizability [30].
6 Connectivity Analyses in Alzheimer’s Disease
For Alzheimer’s disease (AD) and its precursor, the mild

neurocognitive disorder (DSM-5) due to AD, many studies ana-
lyzed alterations of connectivities and graph measures related to
cognition. Alpha band connectivity correlates negatively [31–33],
and theta and delta band connectivities correlate positively with
the cognitive state [34, 35]. The networks derived from alpha
band connectivities lose the small-world structure and evolve
toward a more random network structure [31]. Primarily affected
by this process are the hubs in the parieto-occipital and less the
temporal regions [31].
7 Conclusions
Quantitative EEG analyses, especially connectivity analysis, are

promising candidates for biomarkers. Test-retest reliability has

been shown for PLI and msPLI [26, 36], and criterion validity has
been documented by correlation between connectivity measures
and clinical measures of cognition in patients with neurodegenera-
tive diseases [34, 37–39]. The challenge will be to standardize ana-
lyzing methods and to document feasibility and validity of qEEG
measures as possible diagnostic, prognostic, predictive, or response
biomarkers [40].
References
1. Caviness JN, Hentz JG, Evidente VG, 5. Wallstrom GL, Kass RE, Miller A, Cohn JF,
Driver-Dunckley E, Samanta J, Mahant P,
Fox NA (2004) Automatic correction of
Connor DJ, Sabbagh MN, Shill HA, Adler CH ocular artifacts in the EEG: a comparison of

(2007) Both early and late cognitive dysfunction regression-based and component-based m ethods.
affects the electroencephalogram in Parkinson’s Int J Psychophysiol 53:105–119. https://doi.
disease. Parkinsonism Relat Disord 13:348–354 org/10.1016/j.ijpsycho.2004.03.007
2. Fonseca LC, Tedrus GMAS, Letro GH, Bossoni 6. Nolan H, Whelan R, Reilly RB (2010) FASTER:
AS (2009) Dementia, mild cognitive impairment fully automated statistical thresholding for
and quantitative EEG in patients with Parkinson’s EEG artifact rejection. J Neurosci Methods
disease. Clin EEG Neurosci 40:168–172 192:152–162. https://doi.org/10.1016/j.
3. Roh JH, Park MH, Ko D, Park K-W, Lee D-H, jneumeth.2010.07.015
Han C, Jo SA, Yang K-S, Jung K-Y (2011) 7. Brunet D, Murray MM, Michel CM (2011)
Region and frequency specific changes of spectral Spatiotemporal analysis of multichannel
power in Alzheimer’s disease and mild cognitive EEG: CARTOOL. Comput Intell
impairment. Clin Neurophysiol 122:2169–2176. Neurosci 2011:813870. https://doi.
https://doi.org/10.1016/j.clinph.2011.03.023 org/10.1155/2011/813870
4. Gross J, Baillet S, Barnes GR, Henson RN, 8. Tadel F, Baillet S, Mosher JC, Pantazis D,
Hillebrand A, Jensen O, Jerbi K, Litvak V, Leahy RM (2011) Brainstorm: a user-friendly
Maess B, Oostenveld R, Parkkonen L, Taylor application for MEG/EEG analysis. Comput
JR, van Wassenhove V, Wibral M, Schoffelen Intell Neurosci 2011:879716. https://doi.
J-M (2013) Good practice for conducting org/10.1155/2011/879716
and reporting MEG research. Neuroimage 9. Oostenveld R, Fries P, Maris E, Schoffelen JM
65:349–363. https://doi.org/10.1016/j. (2011) FieldTrip: open source software for
neuroimage.2012.10.001 advanced analysis of MEG, EEG, and invasive
electrophysiological data. Comput Intell ethods. Econometrica 37:424–438. https://

m
Neurosci 2011:1 doi.org/10.2307/1912791
10. Hatz F, Hardmeier M, Bousleiman H, Rüegg 22. Kamiński MJ, Blinowska KJ (1991) A new
S, Schindler C, Fuhr P (2014) Reliability of method of the description of the information flow
fully automated versus visually controlled pre- in the brain structures. Biol Cybern 65:203–210
and post-processing of resting-state EEG. Clin 23. Pascual-Marqui RD (2002) Standardized low-
Neurophysiol. https://doi.org/10.1016/j. resolution brain electromagnetic tomography
clinph.2014.05.014 (sLORETA): technical details. Methods Find
11. Deuschl G, Eisen A (1999) Recommendations Exp Clin Pharmacol 24(Suppl D):5–12
for the practice of clinical neurophysiology: 24. Hillebrand A, Barnes GR (2005)
guidelines of the International Federation of Beamformer analysis of MEG data. In:
Clinical Physiology. Clin Neurophysiol EEG Magnetoencephalography. Academic Press,
Suppl 52, Chapter 1.5 New York, pp 149–171
12. Harmony T (2013) The functional s ignificance 25. Betzel RF, Erickson MA, Abell M, O’Donnell BF,
of delta oscillations in cognitive processing. Hetrick WP, Sporns O (2012) Synchronization
Front Integr Neurosci. https://doi. dynamics and evidence for a repertoire of network
org/10.3389/fnint.2013.00083 states in resting EEG. Front Comput Neurosci.
13. Klimesch W (2012) Alpha-band oscillations, https://doi.org/10.3389/fncom.2012.00074
attention, and controlled access to stored infor-
26. Hatz F, Hardmeier M, Bousleiman H,
mation. Trends Cogn Sci 16:606–617. https:// Rüegg S, Schindler C, Fuhr P (2016)
doi.org/10.1016/j.tics.2012.10.007 Reliability of f unctional connectivity of
14. Klimesch W (1999) EEG alpha and theta electroencephalography applying microstate-

oscillations reflect cognitive and memory
segmented versus classical calculation of phase
performance: a review and analysis. Brain Res lag index. Brain Connect 6:461–469. https://
Rev 29:169–195. https://doi.org/10.1016/ doi.org/10.1089/brain.2015.0368
S0165-0173(98)00056-3 27. Allen EA, Damaraju E, Eichele T, Wu L,
15. Miller KJ, Hermes D, Honey CJ, Hebb AO, Calhoun VD (2017) EEG signatures of
Ramsey NF, Knight RT, Ojemann JG, Fetz dynamic functional network connectivity states.
EE (2012) Human motor cortical activity Brain Topogr. https://doi.org/10.1007/
is selectively phase-entrained on underlying s10548-017-0546-2
rhythms. PLoS Comput Biol. https://doi. 28. de Haan W, Mott K, van Straaten ECW,
org/10.1371/journal.pcbi.1002655 Scheltens P, Stam CJ (2012) Activity dependent
16. Friston KJ (2011) Functional and effective degeneration explains hub vulnerability in

connectivity: a review. Brain Connect 1:13–36. Alzheimer’s disease. PLoS Comput Biol
https://doi.org/10.1089/brain.2011.0008 8:e1002582. https://doi.org/10.1371/jour-
17. Fries P (2005) A mechanism for cognitive nal.pcbi.1002582
dynamics: neuronal communication through neu- 29. Stam CJ, van Straaten ECW (2012) The
ronal coherence. Trends Cogn Sci 9:474–480. organization of physiological brain networks.

https://doi.org/10.1016/j.tics.2005.08.011 Clin Neurophysiol 123:1067–1087. https://
18. Nolte G, Bai O, Wheaton L, Mari Z, Vorbach doi.org/10.1016/j.clinph.2012.01.011
S, Hallett M (2004) Identifying true brain
30. Watts DJ, Strogatz SH (1998) Collective
interaction from EEG data using the imaginary dynamics
of “small-world” networks.
part of coherency. Clin Neurophysiol Nature 393:440–442. https://doi.org/
115:2292–2307. https://doi.org/10.1016/j. 10.1038/30918
clinph.2004.04.029 31. Stam CJ, De Haan W, Daffertshofer A, Jones BF,
19. Stam CJ, van Dijk BW (2002) Synchronization Manshanden I, van Cappellen van Walsum AM,
likelihood: an unbiased measure of generalized Montez T, Verbunt JPA, de Munck JC, van Dijk
synchronization in multivariate data sets. Phys BW, Berendse HW, Scheltens P (2009) Graph
Nonlinear Phenom 163:236–251. https://doi. theoretical analysis of m agnetoencephalographic
org/10.1016/S0167-2789(01)00386-4 functional connectivity in Alzheimer’s disease.
20. Stam CJ, Nolte G, Daffertshofer A (2007) Brain 132:213
Phase lag index: assessment of functional
32. Stam CJ, van Cappellen van Walsum AM,
connectivity from multi channel EEG and MEG Pijnenburg YAL, Berendse HW, de Munck JC,
with diminished bias from common sources. Scheltens P, van Dijk BW (2002) Generalized
Hum Brain Mapp 28:1178–1193. https://doi. synchronization of MEG recordings in

org/10.1002/hbm.20346 Alzheimer’s disease: evidence for involve-
21. Granger CWJ (1969) Investigating causal ment of the gamma band. J Clin Neurophysiol
relations by econometric models and cross-spectral 19:562–574
33. Vecchio F, Miraglia F, Piludu F, Granata G, cortical sources of resting state EEG rhythms.
Romanello R, Caulo M, Onofrj V, Bramanti Front Neurosci. https://doi.org/10.3389/
P, Colosimo C, Rossini PM (2017) “Small fnins.2016.00047
World” architecture in brain connectivity and 38. Ranasinghe KG, Hinkley LB, Beagle AJ,
hippocampal volume in Alzheimer’s disease: a Mizuiri D, Dowling AF, Honma SM, Finucane
study via graph theory from EEG data. Brain MM, Scherling C, Miller BL, Nagarajan SS,
Imaging Behav 11:473–485. https://doi. Vossel KA (2014) Regional functional connec-
org/10.1007/s11682-016-9528-3 tivity predicts distinct cognitive impairments
34. Hatz F, Hardmeier M, Benz N, Ehrensperger M, in Alzheimer’s disease spectrum. Neuroimage
Gschwandtner U, Rüegg S, Schindler C, Monsch Clin 5:385–395. https://doi.org/10.1016/j.
AU, Fuhr P (2015) Microstate connectivity nicl.2014.07.006
alterations in patients with early Alzheimer’s 39. Triggiani AI, Bevilacqua V, Brunetti A, Lizio
disease. Alzheimers Res Ther 7:78. https://doi. R, Tattoli G, Cassano F, Soricelli A, Ferri R,
org/10.1186/s13195-015-0163-9 Nobili F, Gesualdo L, Barulli MR, Tortelli
35. Suárez-Revelo JX, Ochoa-Gómez JF, R, Cardinali V, Giannini A, Spagnolo P,
Duque-Grajales JE, Tobón-Quintero CA
Armenise S, Stocchi F, Buenza G, Scianatico G,
(2016) Biomarkers identification in Alzheimer’s Logroscino G, Lacidogna G, Orzi F, Buttinelli
disease using effective connectivity analysis
C, Giubilei F, Del Percio C, Frisoni GB,
from electroencephalography recordings. Ing Babiloni C (2017) Classification of healthy
E Investig 36:50–57. 10.15446/ing.investig. subjects and Alzheimer’s disease patients with
v36n3.54037 dementia from cortical sources of resting state
36. Hardmeier M, Hatz F, Bousleiman H, Schindler EEG rhythms: a study using artificial neu-
C, Stam CJ, Fuhr P (2014) Reproducibility of ral networks. Front Neurosci. https://doi.
functional connectivity and graph measures based org/10.3389/fnins.2016.00604
on the phase lag index (PLI) and weighted phase 40. Amur S, LaVange L, Zineh I, Buckman-Garner
lag index (wPLI) derived from high resolution S, Woodcock J (2015) Biomarker qualifica-
EEG. PLoS One 9:e108648. https://doi. tion: toward a multiple stakeholder framework
org/10.1371/journal.pone.0108648 for biomarker development, regulatory accep-
37. Babiloni C, Triggiani AI, Lizio R, Cordone tance, and utilization. Clin Pharmacol Ther
S, Tattoli G, Bevilacqua V, Soricelli A, Ferri 98:34–46. https://doi.org/10.1002/cpt.136
R, Nobili F, Gesualdo L, Millán-Calenti JC, 41. Bian N-Y, Wang B, Cao Y, Zhang L (2006)
Buján A, Tortelli R, Cardinali V, Barulli MR, Automatic removal of artifacts from EEG data
Giannini A, Spagnolo P, Armenise S, Buenza using ICA and exponential analysis. In: Proc.
G, Scianatico G, Logroscino G, Frisoni GB, third int. conf. advnaces neural netw.—vol
del Percio C (2016) Classification of single Part II. Springer-Verlag, Berlin, Heidelberg,
normal and Alzheimer’s disease individuals from pp 719–726
Part IV
Biomarker Discovery
Chapter 11
A Selected Reaction Monitoring Protocol

for the Measurement of sTREM2 in Cerebrospinal Fluid
Amanda J. Heslegrave, Wendy E. Heywood, Kevin M. Mills,
and Henrik Zetterberg
Abstract
Mass spectrometry plays an increasingly important role in the biomarker field with the advent of targeted
proteomics. Tryptic peptides from a protein of interest can be used to create a targeted assay to interrogate
cerebrospinal fluid (CSF) for biomarkers. Since heterozygous mutations in the TREM2 gene have been
associated with an increased risk of Alzheimer’s disease, measuring this soluble protein in CSF has become
a priority. This chapter demonstrates the development, optimization, and validation of a method to measure
soluble TREM2 using a single reaction monitoring (SRM) targeted mass spectrometry assay.
Key words Proteomics, Biomarker, Alzheimer’s disease, TREM2, Inflammation
1 Introduction
Since the mid-1990s, proteomics by mass spectrometry has been an

extremely useful tool in the discovery of new biomarkers and for
elucidating new disease mechanisms. Many papers have been
published using untargeted proteomic (more recently referred to as
hypothesis generating) MS workflows which can demonstrate and
detect differences between disease and control cases in neurodegen-
erative disorders [1–3]. Until recently validation of possible bio-
markers has mainly been possible by immunological means, such as
western blot or enzyme-linked immunosorbent assay (ELISA).
However, mass spectrometry is proving to be ever more useful in
providing accurate and specific quantitative data on target proteins
of interest in body fluids [4]. In particular this has proved of great
interest to those researchers studying the composition of cerebro-
spinal fluid (CSF) to find biomarkers for Alzheimer’s disease (AD)
[3] and in the development of new tests for this disease.
Targeted proteomics (single reaction monitoring (SRM)
detecting one analyte and multiple reaction monitoring (MRM),
detecting more than one analyte) is a tandem mass spectrometry
169
170 Amanda J. Heslegrave et al.
(MS/MS) technique usually performed on a triple quadrupole

MS-based instrument. Tandem mass spectrometry has been used
for more than 20 years [5, 6] but almost exclusively for the quan-
titation of small molecules, such as metabolites and drugs. However
it is only recently that tandem mass spectrometry has been used
more and more to quantify proteolytically generated peptides that
act as surrogates of the corresponding intact proteins. The use of
stable isotope-labelled peptide standards can make this method of
quantitation not only extremely accurate but also more specific
than conventional immunological-based assays.
This technique is generally used in conjunction with the
hypothesis generating or proteomics-based technology such as
label-free liquid chromatography-tandem mass spectrometry (LC-
MS/MS) or label-based proteomic methods such as stable isotope
labelling with amino acids in cell culture (SILAC) and tandem
mass tagging (TMT). These types of methods provide protein
identification and relative protein abundance using the spectral
count, stable isotope labelling, or the total number of MS/MS
spectra taken on peptides from a given protein in a given LC-MS/
MS analysis. This is then linearly correlated with protein abun-
dance but does not give absolute quantification [7–9]. For the
development of targeted proteomics or peptide-based MS quanti-
tation, the first step is proteolytic digestion of the proteins by a
protease, typically trypsin, which allows the creation of fragments
small enough to be analyzed within the mass range of triple
quadrupole-based instruments. Trypsin cleaves at the carboxylic
acid side of peptide bonds of lysine and arginine residues and is
generally used due to its robustness, specificity, and the creation of
peptides that are amenable to mass spectrometry. The peptides
produced are now of optimal size and molecular weight and in the
range of most quadrupole or ion trap based tandem mass spec-
trometers. Once the protein has been digested, it is possible to
choose a “marker” peptide for that protein. This marker can then
be weighed in the first mass analyzer (MS1) and its unique mass-
to-charge (m/z) ratio recorded. The peptide is then fragmented in
a collision cell by colliding it with argon gas to yield product ions
which have their m/z ratio recorded in the second part of the mass
spectrometer or MS2. This fragmentation is known as collision-
induced dissociation (CID), and from this, a unique product ion is
chosen as being the characteristic fingerprint of that peptide. Here
we have two levels of identification for the peptide, firstly its intact
mass—precursor ion—and secondly it is identifying peptide frag-
ment mass, product ion. In method development more than one
precursor and product ion will be chosen to optimize, and the
most specific/sensitive will be chosen. Finally, a third and final
level of identification is given by the elution time of the peptide
from the UPLC chromatography column. Because in a complex
biological sample digest, it is possible to get several peptides
with the same precursor/product masses as your target, this
sTREM2 in CSF 171
third identifying step is very important. It is the use of fast

chromatography-based systems such as UPLC with tandem mass
spectrometry that provides extra specificity for the analysis of
proteins that immunological-based assays cannot match. However,
it should be noted that some immunological-based assays can be
significantly more sensitive than MS-based analyses.
TREM2 is a receptor glycoprotein of 230 amino acids which
belongs to the immunoglobulin superfamily. In the brain, TREM2
is expressed exclusively by myeloid cells which include microglia
the brains resident immune cell [10]. It was known that homozy-
gous mutations in this protein caused polycystic lipomembranous
osteodysplasia with sclerosing leukoencephalopathy, also known as
Nasu-Hakola disease [11, 12]. However, in 2013, it was discovered
that heterozygous mutations in the TREM2 gene were noted as
being significant risk factors for AD [13, 14].
This discovery led to intense interest in TREM2 and a require-
ment for sensitive assays to measure the soluble protein. In 2013
there were very few immunoassays available, and those that were
available were not robust. Therefore, a different method of mea-
suring sTREM2 was required, and this led to the development of
an SRM technique that was published in early 2016 [15]. The
benefit of this method is the huge scope for multiplexing as shown
by Heywood et al. in [3]. Although this chapter is using sTREM2
as an example for the method, other interesting proteins can easily
be multiplexed and augmented this way.
2 Materials
2.1 Mass Targeted proteomics were performed on a Waters ultraperfor-

Spectrometer mance liquid chromatography system (Manchester, UK) coupled
to a Waters XEVO TQ-S triple quadrupole mass spectrometer. The
column used was a Waters CORTECS UPLC C18+ column, 90 Å
3 mm × 100 mm attached to a C18+ VanGuard pre-column.
2.2 Samples CSF samples were collected from memory clinics in Sweden and
the UK according to local ethics procedures. Samples were collected
in the morning into polypropylene tubes (Sarstedt, Numbrecht,
Germany) according to standard operating procedures; samples
were then further aliquoted into polypropylene tubes, frozen, and
stored at −80 °C within 2 h.
2.3 Reagents All reagents are of mass spectrometry grade and prepared and
stored at room temperature unless otherwise stated.
2.3.1 Digestion Reagents Digestion buffer—100 mM Tris pH 7.8 containing 6 M urea, 2 M
thiourea, and 2% amidosulfobetaine-14 (ASB-14). 2 M dithioth-
reitol (DTT) is made by dissolving 30 mg of DTT in 1 mL of
100 mM Tris pH 7.8. It is important that this solution is made up
fresh on the day and kept on ice until use. Iodoacetamide (IAA)
was prepared by dissolving 36 mg in 1 mL of 100 mM Tris pH 7.8;
similarly this solution was made up fresh on the day and protected
from light. Sequence grade trypsin (Promega, UK) is prepared by
resuspending lyophilized 1 mg aliquots of the protease with 11 μL
of 40 mM ammonium bicarbonate buffer. All reagents are from
Sigma-Aldrich, Dorset, UK, unless otherwise stated.
2.3.2 Mass Spectrometry MS mobile phases are A, LC-MS grade water with 0.1% formic
Reagents acid (FA), and B, LC-MS grade ACN with 0.1% FA.
2.3.3 Peptides TREM2 peptide VLVEVLADPLDHR was synthesized by Generon

and Proteins (Maidenhead, UK). Yeast enolase standard was from Sigma-Aldrich
(Dorset, UK).
3 Methods
Figure 1 shows a typical SRM development workflow, while Fig. 2

depicts an assay workflow.
3.1 SRM Set Up To set up the SRM assay for sTREM2, we first identified a unique
peptide from the protein sequence using Skyline, https://skyline.
ms/project/home/software/Skyline/. Skyline is a freely available
and open source Windows client application for building selected
reaction monitoring (SRM)/multiple reaction monitoring (MRM)
and parallel reaction monitoring (PRM—targeted MS/MS),
although, it would be preferable to identify peptides from previous
mass spec discovery experiments. Another useful resource is the
Global Proteome Machine http://www.thegpm.org/. We checked
that the peptide was completely unique to TREM2 and that none
of the amino acids would be subject to posttranslational modifica-
tion. This peptide was synthesized by Generon (Berkshire, UK)
and then used to develop the assay.
For tuning of the peptide, standards are reconstituted in 50%
ACN containing 0.1% FA before being infused into the mass
spectrometer. The amount of peptide will need to be optimized
but a suggested starting point is 100 μg/mL. A full scan can be
performed initially to optimize tuning concentration, and the
peptide is diluted accordingly. Once the concentration has been
optimized, optimal collision and cone energies are obtained
using the automatic “Intellistart” tuning software of the Xevo
TQS mass spectrometry system. After the optimal fragmentation
and MS parameters are obtained and entered into the SRM, the
retention time of the peptide is assessed. In this case a 10 min
chromatographic run time is used, which is shown in Table 1
(Notes 1 and 2).
sTREM2 in CSF 173
Fig. 1 Typical single reaction monitoring (SRM) development workflow
The advantage of using peptides generated from actual digests,

under typical laboratory conditions, is that it allows the user to
identify the peptides that are amenable to mass spectrometry, i.e.,
in silico determined tryptic fragments cannot predict chromato-
graph peak shape or fragmentation patterns. This often saves
considerable time in method development (Note 3).
Once the assay has been developed using synthetic peptides, it
is important to then ensure your peptide can be detected in the
Fig. 2 Typical assay workflow
Table 1
10-minute chromatographic run
Time (min) Flow rate (mL/min) %A %B Gradient

Initial 0.8 97 3 0
0.2 0.8 97 3 6
7.00 0.8 60 40 6
7.01 0.8 0.1 99.9 6
8.00 0.8 0.1 99.9 6
8.01 0.8 97 3 6
10.00 0.8 97 3 1
appropriate sample matrix. In this case, the synthetic peptide is

used to test this by spiking varying amounts in to CSF. It is also
necessary to optimize the amount of your sample to dry down and
the amount of digest to inject onto the column (Notes 4–6).
sTREM2 in CSF 175
3.2 Sample 100 μL of each CSF sample is aliquoted into a fresh siliconized
Preparation Eppendorf tube and freeze dried. The resulting proteins are then
resuspended in 20 μL of digestion buffer and shaken at RT for
20 min. 1.5 μL of 2 M DTT is added to reduce disulfide bridges
and the sample shaken for a further hour at RT. 3 μL of (2 M) IAA
is added in order to carboamidomethylate any cysteines contained
in the proteins, and the sample is then incubated at RT for a fur-
ther 45 min in the dark. The sample is then diluted with 165.5 μL
of LC-MS grade water and 10 μL of a solution containing 1 μg of
trypsin. Samples are then vortexed and left in a water bath at 37 °C
overnight. Before mass spectral analysis, centrifuge samples at
13,000 × g for 10 min to remove particulates before transfer into
vials suitable for mass spectral analysis.
3.3 Mass Using the chromatographic and mass spectral conditions devel-
Spectrometry oped and described above, the samples can be input into the
sample manager and the analyses completed. During any analyses
a standard curve of your peptide should also be included, and this
should cover the range that the endogenous levels of peptides are
expected to be present in your subject samples. Each standard
curve is made up in pooled CSF in order that any matrix ion sup-
pression is accounted for. It is also good practice to create quality
controls as well using pooled CSF spiked with your peptide(s) in
order that you can check the sensitivity of the assay at appropriate
intervals (should fall within ±10% of established quantities). It is
recommended to run QC samples at intervals of 10–15 subject
samples. Internal standard values should also be monitored, and
coefficients of variation (%CV) of these should also be used to
monitor assay stability.
3.4 Data Analysis Data analysis uses Waters MassLynx and TargetLynx V4.1 software
to process the raw data. Methods can be set up so that the inte-
grated peaks can be ratioed to the internal standard which can be
yeast enolase in this case or a stable isotope-labelled peptide.
Integration should be checked to confirm accuracy, and then
response values can be interpolated from your standard curve.
4 Conclusion
In summary, the SRM method for the quantification of sTREM2

described in this chapter has been used to investigate levels of
sTREM2 in a cohort of dementia subjects and controls and later
validated on a separate cohort. More recently this protein has been
found to be increased in early AD [16, 17] and highlights the
importance of inflammation as an early event in this disease. This is
a highly adaptable method, and multiplexing is a matter of adding
suitably optimized MRM files as shown above.
5 Notes and Troubleshooting
1. For one transition the method time can be shortened, say

5 min, when multiplexing; it is suggested that MRM files
should be grouped and split according to retention time, in
order to maximize the dwell time of each ion.
2. Transitions should be selected on the best intensity and absence
of interfering peaks.
3. Synthetic peptides that contain cysteine residues must be car-
boamidomethylated prior to infusion.
4. If matrix interference makes analyte detection a problem, then
the MS resolution can be reduced; this results in an increase of
sensitivity. Accurate mass will need to be determined for the
synthetic peptide at the new resolution.
5. The least amount of sample to enable the analyte to be detected
should be injected onto the column. Too much sample can
lead to ion suppression effects. In a multiplex situation, it is
worth grouping transactions according to the endogenous
abundance of the analyte in the sample, in order that you can
optimize the amount of sample to inject.
References
1. Conti A, Alessio M (2015) Comparative pro- 12(5):2013–2021. https://doi.org/10.1021/

teomics for the evaluation of protein expres- pr301200e
sion and modifications in neurodegenerative 5. Xu RN, Fan L, Rieser MJ et al (2007) Recent
diseases. Int Rev Neurobiol 121:117–152. advances in high-throughput quantitative bio-
https://doi.org/10.1016/bs. analysis by LC-MS/MS. J Pharm Biomed Anal
irn.2015.05.004 44(2):342–355. https://doi.org/10.1016/j.
2. Sultana R, Boyd-Kimball D, Cai J et al (2007) jpba.2007.02.006
Proteomics analysis of the Alzheimer’s disease 6. Kondrat RW, McClusky GA, Cooks RG (1978)
hippocampal proteome. J Alzheimer’s Dis Multiple reaction monitoring in mass sectrom-
11(2):153–164 etry for diresct analysis of complex mixtures.
3. Heywood WE, Galimberti D, Bliss E et al Anal Chem 50:2017–2021
(2015) Identification of novel CSF biomarkers 7. Bondarenko PV, Chelius D, Shaler TA (2002)
for neurodegeneration and their validation by a Identification and relative quantitation of pro-
high-throughput multiplexed targeted protein mixtures by enzymatic digestion followed
teomic assay. Mol Neurodegener 10:64. by capillary reversed-phase liquid
https://doi.org/10.1186/ chromatography- tandem mass spectrometry.
s13024-015-0059-y Anal Chem 74(18):4741–4749
4. Manwaring V, Heywood WE, Clayton R et al 8. Chelius D, Bondarenko PV (2002) Quantitative
(2013) The identification of new biomarkers profiling of proteins in complex mixtures using
for identifying and monitoring kidney disease liquid chromatography and mass spectrometry.
and their translation into a rapid mass J Proteome Res 1(4):317–323
spectrometry-based test: evidence of presymp- 9. Wang W, Zhou H, Lin H et al (2003)
tomatic kidney disease in pediatric Fabry and Quantification of proteins and metabolites by
type-I diabetic patients. J Proteome Res mass spectrometry without isotopic labeling
sTREM2 in CSF 177
or spiked standards. Anal Chem 75(18): risk of Alzheimer’s disease. N Engl J Med
4818–4826 368(2):107–116. https://doi.org/10.1056/
10. Colonna M (2003) TREMs in the immune NEJMoa1211103
system and beyond. Nat Rev Immunol 3(6): 15. Heslegrave A, Heywood W, Paterson R et al
445–453. https://doi.org/10.1038/nri1106 (2016) Increased cerebrospinal fluid soluble
11. Paloneva J, Manninen T, Christman G et al TREM2 concentration in Alzheimer’s disease.
(2002) Mutations in two genes encoding dif- Mol Neurodegener 11:3. https://doi.
ferent subunits of a receptor signaling complex org/10.1186/s13024-016-0071-x
result in an identical disease phenotype. Am 16. Suarez-Calvet M, Araque Caballero MA,
J Hum Genet 71(3):656–662. https://doi. Kleinberger G et al (2016) Early changes in
org/10.1086/342259 CSF sTREM2 in dominantly inherited
12. Kondo T, Takahashi K, Kohara N et al (2002) Alzheimer’s disease occur after amyloid deposi-
Heterogeneity of presenile dementia with bone tion and neuronal injury. Sci Transl Med
cysts (Nasu-Hakola disease): three genetic 8(369):369ra178. https://doi.org/10.1126/
forms. Neurology 59(7):1105–1107 scitranslmed.aag1767
13. Guerreiro R, Wojtas A, Bras J et al (2013) 17. Suarez-Calvet M, Kleinberger G, Araque
TREM2 variants in Alzheimer’s disease. N Caballero MA et al (2016) sTREM2 cerebro-
Engl J Med 368(2):117–127. https://doi. spinal fluid levels are a potential biomarker for
org/10.1056/NEJMoa1211851 microglia activity in early-stage Alzheimer’s
14. Jonsson T, Stefansson H, Steinberg S et al disease and associate with neuronal injury
(2013) Variant of TREM2 associated with the markers. EMBO Mol Med 8(5):466–476.
10.15252/emmm.201506123
Chapter 12
Soluble Amyloid Precursor Proteins in Blood:

Methods and Challenges
Robert Perneczky and Panagiotis Alexopoulos
Abstract
Soluble amyloid precursor proteins (sAPP) are under investigation as novel biomarkers of Alzheimer’s
disease (AD) pathophysiology, but protein levels in cerebrospinal fluid do not seem to follow a con-
sistent pattern, which limits their usefulness as diagnostic and prognostic tool. More recently, evi-
dence has been presented for sAPP blood plasma concentration differences between patients with AD
dementia and healthy control subjects. Blood can be easily accessed, which makes it an interesting
target for biomarker discovery. This chapter presents a brief overview of sAPP in blood as a new AD
biomarker candidate. Issues and challenges are discussed, both from a perspective of laboratory assess-
ment and clinical application. A relevant technique for sAPP ascertainment is also provided.
Key words Alzheimer’s disease, Dementia, Blood biomarker, Early diagnosis, Prognosis, Tau,
Amyloid, Mild cognitive impairment
1 Introduction
Alzheimer’s disease (AD) is characterized by a slowly progressive

neurodegeneration, which as the disease progresses increasingly
affects cognitive abilities and daily activities; dementia due to AD
severely impairs the affected individuals’ and their families’ quality
of life, and it ultimately leads to complete dependence upon care of
others [1]. The first stage of AD, which can last several decades, is
asymptomatic and preclinical. During the second stage, neurode-
generation becomes severe enough to affect cognitive perfor-
mance, but normal daily activities remain largely intact; this
pre-dementia clinical stage is usually referred to as mild cognitive
impairment (MCI) [2]. Only in the last stage of AD, neurodegen-
eration becomes advanced enough to result in the typical amnestic-
type dementia syndrome with significantly impaired everyday
function [3]. Until recently, an AD diagnosis could only be estab-
lished in clinical settings if evidence for dementia was available;
however, more recently updated diagnostic criteria were presented,
179
180 Robert Perneczky and Panagiotis Alexopoulos
which include a stronger focus on a biologically defined diagnosis,

with an emphasis on biomarkers, including imaging studies and
cerebrospinal fluid (CSF) proteins. The National Institute on
Aging-Alzheimer’s Association (NIA-AA) guidelines, for example,
conceptualize AD as a progressive disorder including all possible
stages from asymptomatic to severely impaired [1, 4]. According
to these guidelines, AD can be diagnosed in the MCI stage in case
of a typical pattern of symptoms and if the presence of AD patho-
physiology is evidenced by supporting biomarker findings. The
according stage is termed MCI due to AD. This new approach
implies that degenerative brain damage due to AD pathophysio-
logical changes foregoes the onset of clinical disease by many years
and that cerebral lesions may be present in individuals long before
they develop dementia [5].
The hope for disease modification, rather than only symptom-
atic treatment, by one of the many compounds currently scruti-
nized in clinical trials as well as technological advances in diagnostic
procedures and biomarker discovery, fuels the search for reliable
markers of AD pathophysiology and clinical disease progression.
Diagnostic and therapeutic progress depends on the ability to suc-
cessfully diagnose the disease early in its course, before significant
neural damage occurs, as well as to understand its natural evolution
over time. The published literature in early clinical AD provides
insight into the cognitive domains that are affected first (i.e., epi-
sodic memory and executive functioning) and associated biomarker
changes. However, there remains an unmet need for prospectively
collected real-world information regarding prognostic factors, dis-
ease course, functional decline, and disease burden in this early
stage of the disease; furthermore, biological predictors and indica-
tors of disease progression should be explored in unbiased,
population-based cohorts. The majority of publications have so far
focused on what may be termed “late” MCI, i.e., those most likely
to transition to dementia within a year or two. The information is
limited on what are the cognitive deficits and how these manifest
and change at times more distal to the onset of AD dementia.
Given these challenges, the need for current information on indi-
viduals at varying levels of risk for MCI due to AD would be useful
to improve our understanding of the natural history and identify
opportunities for intervention.
There is robust evidence that the so-called established CSF pro-
tein biomarkers amyloid-β (Aβ)1–42, total-Tau (tTau), and
phosphorylated-Tau (pTau)181 [6] as well as functional and struc-
tural neuroimaging studies, including fluorodeoxyglucose positron
emission tomography (PET) and magnetic resonance imaging of
the mediotemporal lobe [7], have good diagnostic and prognostic
accuracy. More recently developed CSF and imaging biomarker
candidates show promise, such as PET tracers of Aβ [8] and tau [9].
Novel CSF protein markers, which mirror upstream events of Aβ
Soluble APP in Blood 181
generation, such as soluble amyloid precursor proteins (sAPP) [10],

have also been studied, albeit with mixed results. Even if improved
imaging or CSF markers of AD pathophysiology were successfully
developed, their use in larger-scale studies would still be limited due
to the associated costs, exposure to radiation, and invasiveness.
Therefore, there is an urgent need to develop biomarkers which can
be obtained with relative ease from peripheral body fluids such as
blood to replace or assist the existing fluid and imaging markers.
2 Soluble Amyloid Precursor Protein in Blood as a New Biomarker Candidate
Blood sampling, in contrast to obtaining CSF, does not involve a

laborious and invasive procedure, which makes the execution of
repeated assessments and large-scale screening studies much easier.
Blood is in constant exchange with all organs and tissue including the
central nervous system; a wide range of physiological and physiopath-
ological processes are therefore reflected in it; hence, blood is an opti-
mal medium for biomarker discovery for a range of different health
conditions [11]. Signaling proteins used by the brain to exert control
over body functions can be detected in blood [12], and specific pat-
terns of signaling protein changes may be associated with AD [13].
The sAPPα and sAPPβ, which are products of physiological and
amyloidogenic cleavage of APP, respectively, are related to central
upstream pathophysiological events in AD [14] (Fig. 1). Some ear-
lier studies found increased sAPPβ CSF levels in AD vs. controls
[15, 16] and stable vs. progressive MCI [10]. Other published
reports do not support these results [17–20], including studies
which nevertheless provide evidence for the potential usefulness of
sAPPβ as a marker of target engagement in studies on inhibitors of
the rate-limiting enzyme in the physiopathological cascade of Aβ
generation, i.e., beta-site APP cleaving enzyme 1 (BACE1) [21].
More recently, results were published that indicate the possible
usefulness of sAPPβ in blood plasma to support AD diagnosis.
Significantly decreased plasma concentrations in patients with AD
dementia were found as compared to age-matched healthy control
subjects and patients with behavioral-variant frontotemporal demen-
tia, i.e., a neurodegenerative condition with a different pathomecha-
nism which does not typically involve APP-related changes; no
concentration differences between the groups were described for
sAPPα [22]. This study provides a first piece of evidence in support of
sAPPβ in blood as a promising new biomarker candidate of AD, which
may potentially improve the diagnostic accuracy of existing markers
and also enable a less invasive diagnostic work-up. Further research is
required to establish normal ranges and to replicate the results in inde-
pendent cohorts including larger numbers of participants covering a
wider spectrum of cognitive impairment. Our initial results also
indicate that sAPPβ plasma concentrations as compared to healthy
Fig. 1 Schematic diagram representing proteolytic cleavage of amyloid precursor protein (reproduced with
permission from: R. Perneczky in [27]). The membrane-bound amyloid precursor protein (APP) is alternatively
cleaved by either α-secretase, resulting in non-amyloidogenic fragments including soluble APP (sAPP)α or by
β-secretase, resulting in sAPPβ and the harmful Aβ fragment via further cleavage by γ-secretase
controls are lower in MCI patients with an AD-typical PET metabolic

pattern and comparable to plasma concentrations of AD dementia
patients (manuscript in preparation).
3 Methods
3.1 General Pre- See [23] for proposed consensus guidelines for the standardization
analytic Guidelines of pre-analytic variables for blood-based biomarker studies in AD
research.
1. Sample withdrawal should only be undertaken by qualified
medical personnel. The identity of the patient should be
checked, and the analysis order/study documentation should
be filled out completely before withdrawal of the sample.
Written informed consent by the patient or study subject must
be available before any procedures are performed. Tubes should
be clearly labelled beforehand and immediately closed after tak-
ing the sample to avoid contamination and evaporation.
2. Samples should optimally be taken after overnight fasting at a
standard time in the morning, before any morning medica-
tions are taken. Closed blood withdrawal systems (e.g.,
Monovette, Vacuette) reduce the preanalytical influence on
the results and the infection risk for medical personnel and are
therefore preferable to syringes.
3. To avoid contamination, tubes without additive should be col-
lected before tubes with additive. When withdrawing several
tubes, the thinning tube should never be at the beginning
(release of tissue factors on puncture).
4. Venous stasis should be around 20–30 mmHg under the sys-

tolic blood pressure so that arterial pulsation is maintained.
During longer blood withdrawals (several tubes), the stasis
should be relieved. Venous stasis should not last too long.
5. Blood should not be taken from horizontal veins or arterial
entry sites. If this cannot be avoided, the entry site must be
rinsed with 0.9% NaCl solution before withdrawal and a suffi-
cient amount of blood (for adults 10–20 mL) thrown away.
Directly before blood withdrawal, no medications should be
injected through the catheter.
6. Blood withdrawal should not be performed with too fine can-
nulae since hemolysis can occur if cannulae are too fine. The
blood should flow freely into the collection vessel. If the pres-
sure is too low, hemolysis can occur. Samples should not be
exposed to direct sunlight.
3.2 Collection 1. Whole blood samples are collected by venepuncture into com-
of Blood Samples mercially available anticoagulant-treated tubes (e.g., citrate or
for Biomarker Testing EDTA) and put into the freezer at 4 °C within 120 min after
collection.
2. Plasma is isolated from blood cells (see Note 1) by centrifuga-
tion at 2500 × g for 15 min.
3. If the sample is not analyzed immediately, it should be appor-
tioned into aliquots (e.g., 500 μL), and the aliquots should
immediately be transferred to −20 °C and should within
2 weeks be transferred to −80 °C for long-term storage.
3.3 Considerations 1. Biomarker development and testing frequently involve collab-

for Multicenter oration between different centers and laboratories, which is
Research associated with additional challenges and requires meticulous
adherence to sample collection protocols. Multicenter studies
require each individual laboratory to follow exactly the same
protocol for plasma collection.
2. Aliquots of the samples are kept at −20 °C for not longer than
2 weeks or at −80 °C until use. Plasma samples must be shipped
on dry ice to the central laboratory and are required to be
immediately stored at −20 °C for not longer than 2 weeks or
at −80 °C until further analysis (see Note 2).
3.4 Enzyme-Linked Different methods have been established to quantify protein con-
Immunosorbent Assay centrations in body fluids. Here, ELISA will be discussed as an
(ELISA) example, since it is a widely used method, which has been shown
to be able to detect sAPP in human blood samples. ELISA is usu-
ally performed in duplicate and according to the commercial man-
ufacturer’s instructions (see [22] for an example of a sAPP assay).
An ELISA includes coating, blocking, incubation, and detection
according to the following general guidelines (specific assays may

require adjustments to certain parameters):
1. Determine wells for reagent blank, test sample blank, test sam-
ple and diluted standard, and then coat the microwells with
100 μL appropriate diluted antigens (see Note 3).
2. Incubate the plate at room temperature (RT) for 2 h or at 4 °C
after covering it with plate lid overnight.
3. Vigorously wash all unbound antigen off the plate to prevent
false positive for at least three times with 300 μL wash buffer
(see Note 4). Then block nonspecific binding by adding
200 μL of blocking buffer to each well. Incubate at RT for 1 h
or at 4 °C overnight, and then wash plate as above.
4. Pipette 100 μL of diluted samples (see Notes 2 and 5) and
standards to appropriate wells, incubate for 1 h at RT or at
4 °C overnight, and then repeat washing step.
5. Add 100 μL of second step antibody to wells and incubate for
1 h at RT; then repeat washing step.
6. For color development, 100 μL of chromogenic substrate is
added to each well. Cover the plate and incubate for 15 min or
until a suitable color has developed. The plate should prefera-
bly be protected against light during this incubation.
7. Determine the optical density of each well within 30 min
through the bottom of the microwell plate using an auto-
mated or semiautomated photometer (ELISA reader) with
appropriate wavelength). Determine the concentration of the
samples from the standard curve using curve fitting software
(see Note 6).
4 Conclusions
Proteomic analysis of blood samples is a powerful tool to identify

novel, clinically more relevant biomarkers of AD. Several analytical
platforms are available, with antibody-based techniques being par-
ticularly valuable if many samples have to be analyzed quantita-
tively. ELISA technology is at present the gold standard in
quantifying protein markers. It combines the sensitivity of simple
enzyme assays with the specificity of antibodies, being only capable
of measuring the concentration of a single protein at a time, albeit
in a highly specific and high-throughput manner. ELISA is able to
detect proteins in complex matrices such as CSF and blood in the
pg/mL range; research is simplified by the commercial availability
of standardized off-the-shelf assays. Many researchers are compe-
tent in using ELISA, and they are a comparably low-cost solution.
There is also no need for specialist equipment to run a standard
96-well assay. Having said that, there are also several shortcom-
ings, including the relatively large sample volumes needed to run
an ELISA), and restrictions in automatization due to the required
wash steps, even though manual handling of samples may be prone
to failure. A recent study across 14 clinical neurochemistry labora-
tories in Europe (Germany, Austria, and Switzerland) reported
that the coefficient of variation CV (CV) of ELISAs was in the
20–30% range when using commercially available ELISA kits for
AD diagnosis [24].
From reviewing the available literature on sAPP as a novel AD
biomarker, two main conclusions arise. First, concentration
changes of sAPPα/sAPPβ in CSF do not seem to follow a consis-
tent pattern limiting their usefulness as diagnostic markers.
Secondly, initial results relating to sAPPβ in blood are encourag-
ing, but replication and further validation are required before firm
conclusions can be drawn. Some evidence exists that a combina-
tion of the new and established biomarkers may improve the over-
all diagnostic accuracy [10], but more research in this field is
needed. There also remains a need to standardize laboratory pro-
cedures to generate reliable results across laboratories in multi-
center studies. Several quality control initiates have been launched
worldwide to establish standardized protocols for biomarker assess-
ment, including the Alzheimer’s Association Global Biomarkers
Consortium [25].
5 Notes
1. Compared with serum, plasma sampling needs no clotting time,

and the isolation of cells and liquid phase is easily accomplished
and, thus, is less time-consuming. Furthermore, plasma volume
yield is about 10–20% higher compared to serum; the concen-
tration of proteins in plasma is also greater than in serum, which
contains clotting factors and related constituents [26].
2. Test samples should be measured soon after collection. For the
stored frozen samples, it is important to avoid additional
freeze-thaw cycles. When using frozen samples, it is recom-
mended to thaw the samples at a low temperature and mix
them completely by vortexing prior to use. Hemolyzed, icteric,
or lipemic samples might invalidate certain tests.
3. Wells reserved for chromogen blanks should be left empty.
4. Always remove the wash buffer completely by tapping the
pre-coated plate on paper towel. Do not wipe wells with
paper towel.
5. Plasma samples are recommended to be diluted 1:50 to 1:100.
6. The dose-response may be nonlinear beyond the standard

point and inaccurate; thus, do not extrapolate the standard
curve beyond the highest point; diluted samples that are
greater than the highest standard should be reanalyzed, and
the results should be multiplied by the appropriate dilution
factor.
References
1. Jack CR Jr, Albert MS, Knopman DS et al age.2007.09.020. S1053-8119(07)00806-3
(2011) Introduction to the recommenda- [pii]
tions from the National Institute on Aging- 9. Scholl M, Lockhart SN, Schonhaut DR
Alzheimer’s Association workgroups on et al (2016) PET imaging of Tau deposi-
diagnostic guidelines for Alzheimer’s disease. tion in the aging human brain. Neuron
Alzheimers Dement 7(3):257–262. https:// 89(5):971–982. https://doi.org/10.1016/j.
doi.org/10.1016/j.jalz.2011.03.004. S1552- neuron.2016.01.028
5260(11)00100-2 [pii] 10. Perneczky R, Tsolakidou A, Arnold A et al
2. Lopez OL (2013) Mild cognitive impair- (2011) CSF soluble amyloid precursor pro-
ment. Continuum (Minneap Minn) teins in the diagnosis of incipient Alzheimer
19(2 Dementia):411–424. https://doi. disease. Neurology 77(1):35–38. https://doi.
org/10.1212/01.CON.0000429175.29601.97 org/10.1212/WNL.0b013e318221ad47
3. McKhann GM, Knopman DS, Chertkow 11. Chan K, Lucas D, Hise D et al (2004)
H et al (2011) The diagnosis of dementia Analysis of the human serum proteome. Clin
due to Alzheimer’s disease: recommenda- Proteomics 1:101–226
tions from the National Institute on Aging- 12. Wyss-Coray T (2006) Inflammation in
Alzheimer’s Association workgroups on Alzheimer disease: driving force, bystander
diagnostic guidelines for Alzheimer’s disease. or beneficial response? Nat Med 12(9):1005–
Alzheimers Dement 7(3):263–269. https:// 1015. https://doi.org/10.1038/nm1484.
doi.org/10.1016/j.jalz.2011.03.005. S1552- nm1484 [pii]
5260(11)00101-4 [pii]
13. Ray S, Britschgi M, Herbert C et al (2007)
4. McKhann GM (2011) Changing concepts Classification and prediction of clinical
of Alzheimer disease. JAMA 305(23):2458– Alzheimer’s diagnosis based on plasma signal-
2459. https://doi.org/10.1001/jama.2011. ing proteins. Nat Med 13(11):1359–1362.
810.305/23/2458 [pii] https://doi.org/10.1038/nm1653
5. Giaccone G, Arzberger T, Alafuzoff I et al 14. Zhang H, Ma Q, Zhang YW et al (2012)
(2011) New lexicon and criteria for the Proteolytic processing of Alzheimer’s beta-
diagnosis of Alzheimer’s disease. Lancet amyloid precursor protein. J Neurochem
Neurol 10(4):298–299.; author reply 120(Suppl 1):9–21. https://doi.
300–301. https://doi.org/10.1016/ org/10.1111/j.1471-4159.2011.07519.x
S1474-4422(11)70055-2
15. Wu G, Sankaranarayanan S, Wong J et al (2012)
6. Blennow K, Hampel H, Weiner M et al (2010) Characterization of plasma beta-secretase
Cerebrospinal fluid and plasma biomark- (BACE1) activity and soluble amyloid precursor
ers in Alzheimer disease. Nat Rev Neurol proteins as potential biomarkers for Alzheimer’s
6(3):131–144. https://doi.org/10.1038/ disease. J Neurosci Res 90(12):2247–2258.
nrneurol.2010.4 https://doi.org/10.1002/jnr.23122
7. Frisoni GB, Fox NC, Jack CR Jr et al (2010) 16. Lewczuk P, Popp J, Lelental N et al (2012)
The clinical use of structural MRI in Alzheimer Cerebrospinal fluid soluble amyloid-beta pro-
disease. Nat Rev Neurol 6(2):67–77. https:// tein precursor as a potential novel biomark-
doi.org/10.1038/nrneurol.2009.215 ers of Alzheimer’s disease. J Alzheimers Dis
8. Drzezga A, Grimmer T, Henriksen G et al 28(1):119–125. https://doi.org/10.3233/
(2008) Imaging of amyloid plaques and cerebral JAD-2011-110857
glucose metabolism in semantic dementia and 17. Rosen C, Andreasson U, Mattsson N et al
Alzheimer’s disease. NeuroImage 39(2):619– (2012) Cerebrospinal fluid profiles of amyloid
633. https://doi.org/10.1016/j.neuroim-
beta-related biomarkers in Alzheimer’s disease. ease. Transl Psychiatry 3:e227. https://doi.

Neuromolecular Med 14(1):65–73. https:// org/10.1038/tp.2013.11
doi.org/10.1007/s12017-012-8171-4 23. O’Bryant SE, Gupta V, Henriksen K et al
18. Brinkmalm G, Brinkmalm A, Bourgeois P et al (2015) Guidelines for the standardization
(2013) Soluble amyloid precursor protein alpha of preanalytic variables for blood-based bio-
and beta in CSF in Alzheimer’s disease. Brain Res marker studies in Alzheimer’s disease research.
1513:117–126. https://doi.org/10.1016/j. Alzheimers Dement 11(5):549–560. https://
brainres.2013.03.019 doi.org/10.1016/j.jalz.2014.08.099
19. Olsson A, Hoglund K, Sjogren M et al
24. Lewczuk P, Beck G, Ganslandt O et al (2006)
(2003) Measurement of alpha- and beta- International quality control survey of neu-
secretase cleaved amyloid precursor pro- rochemical dementia diagnostics. Neurosci
tein in cerebrospinal fluid from Alzheimer Lett 409(1):1–4. https://doi.org/10.1016/j.
patients. Exp Neurol 183(1):74–80 neulet.2006.07.009
20. Zetterberg H, Andreasson U, Hansson O et al 25. Carrillo MC, Blennow K, Soares H et al (2013)
(2008) Elevated cerebrospinal fluid BACE1 Global standardization measurement of cerebral
activity in incipient Alzheimer disease. Arch spinal fluid for Alzheimer’s disease: an update from
Neurol 65(8):1102–1107. https://doi. the Alzheimer’s Association Global Biomarkers
org/10.1001/archneur.65.8.1102 Consortium. Alzheimers Dement 9(2):137–140.
21. Perneczky R, Alexopoulos P, Alzheimer’s
https://doi.org/10.1016/j.jalz.2012.11.003
Disease neuroimaging Initiative (2014) 26. Tammen H, Schulte I, Hess R et al (2005)
Cerebrospinal fluid BACE1 activity and Peptidomic analysis of human blood speci-
markers of amyloid precursor protein mens: comparison between plasma specimens
metabolism and axonal degeneration in and serum by differential peptide display.
Alzheimer’s disease. Alzheimers Dement Proteomics 5(13):3414–3422. https://doi.
10(5 Suppl):S425–S429.e1. https://doi. org/10.1002/pmic.200401219
org/10.1016/j.jalz.2013.09.006 27. Malpass K (2011) Alzheimer disease: a novel
22. Perneczky R, Guo LH, Kagerbauer SM et al biomarker to detect early-stage Alzheimer dis-
(2013) Soluble amyloid precursor protein beta ease. Nat Rev Neurol 7(8):420. https://doi.
as blood-based biomarker of Alzheimer’s dis- org/10.1038/nrneurol.2011.112
Chapter 13
Tau Imaging in Preclinical Alzheimer’s Disease

Paul Edison
Abstract
Aggregated tau is a major neuropathological protein implicated in the pathophysiology of common
neurodegenerative disorders, such as Alzheimer’s disease (AD), Parkinson’s disease without (PD) and
with later dementia (PDD), Lewy body dementia (LBD), frontotemporal dementia (FTD), and cortico-
basal degeneration (CBD). Aggregated tau tangles, the result of hyperphosphorylation, is a pathological
characteristic of a group of neurodegenerative conditions known as the tauopathies. In AD, it has been
shown that the density of tau tangles closely correlates with neuronal dysfunction and cell death, which
is not the case for β-amyloid. Until now, diagnostic and pathological demonstration of tau deposition
has only been possible by invasive techniques such as brain biopsy or postmortem examination. The
recent advances in the development of selective tau positron emission tomography (PET) tracers have
allowed in vivo investigation of the presence and extent of tau pathology in patients suspected of having
tauopathies and the role of tau in the early phases of neurodegenerative diseases. In this review, the role
of aggregated tau will be discussed, as well as the challenges posed by, and the current status of, the
development of selective tau tracers as biomarkers, and the new clinical information that has been
uncovered, in addition to the opportunities for refining the diagnosis of tauopathies in the future.
Key words Tau imaging, Dementia, Neurodegenerative diseases, PET, Alzheimer’s disease
1 Introduction
Alzheimer’s disease (AD), Parkinson’s disease without (PD) and

with later dementia (PDD), Lewy body dementia (LBD), fronto-
temporal dementia (FTD), and corticobasal degeneration (CBD)
are common neurodegenerative disorders. These neurodegenera-
tive conditions are characterized by the disturbance of protein
homeostasis, leading to the pathologic accumulation of protein,
especially tau and β-amyloid (Aβ). Tau is a microtubule-associated
protein necessary for the cytoarchitecture of the cells. However,
aggregated tau, due to hyperphosphorylation, is a pathological
characteristic of a group of neurodegenerative conditions known as
the tauopathies [1]. The neuropathological substrates of AD are
tau neurofibrillary tangles (NFTs) and Aβ plaques, while activated
microglia, astrocytes, and neuropil threads also play a significant
189
190 Paul Edison
role in disease pathogenesis. Recent advances in selective tau tracer

development for positron emission tomography (PET) imaging
have, for the first time, allowed in vivo exploration of the presence
and extent of tau pathology in patients suspected of having
tauopathies.
2 Tau Protein in Tauopathies
Tau is a protein that exists mainly in axons and binds to microtu-

bules to stabilize them. Microtubules are critical for maintaining
cellular structural integrity [1–3]. Tau protein can occur as three-
repeat (3R) or four-repeat (4R) forms. Equal numbers of 3R and
4R isoforms exist in the healthy adult human. In neurodegenerative
diseases, the hyperphosphorylated form of tau causes its aggrega-
tion, resulting in neuronal dysfunction and death by reducing their
stability. Aggregated tau in AD exists as paired helical filaments
which further coalesce into neurofibrillary tangles (NFTs) [3–5].
These aggregates spread transynaptically throughout the brain, and
it is shown that tau deposition correlates with the clinical progres-
sion of the disease. While Aβ and tau are the cardinal features of
AD, it is clear that a single aggregated protein can be the underlying
cause of several different clinical phenotypes or a single clinical phe-
notype can be caused by different aggregated proteins.
To date, research has focused primarily on the amyloid cascade
hypothesis and this amyloid Aβ-centric approach to AD [2, 3].
Amyloid Aβ therapeutic approaches, particularly monoclonal anti-
bodies against both soluble and fibrillary forms of amyloid Aβ beta,
have failed in preventing clinical progression of AD, despite a reduc-
tion in the amyloid Aβ plaque burden [4]. However, there is some
hope that these drugs modify disease progression if given early
enough in the disease course, but long-term efficacy data in patients
with prodromal AD are only starting to emerge [5, 6]. A recurring
argument against the amyloid hypothesis is that despite the promi-
nence of amyloid Aβ plaques, tau-associated pathologic changes are
more closely correlated with disease severity [7]. Thus, tau may be
an important protein in early detection and staging of AD. However,
tau pathology is nonspecific and is a marker of neurodegeneration
beyond AD in diseases such as FTD, traumatic brain injury, pro-
gressive supranuclear palsy (PSP), CBD, and chronic traumatic
encephalopathy.
While amyloid Aβ deposition starts in the inferior frontal cingu-
late areas and then spreads to the association cortex, tau deposition
takes place initially in the transentorhinal cortex (stages 1 and 2) in
the stage of AD before symptoms are apparent, then spreads to the
limbic, and symptoms then become evident [6]. As disease pro-
gresses, NFTs finally involve association cortical areas by Braak
Tau Imaging in AD 191
stages 5–6 areas at which time symptoms become severe. It is also

shown that, at postmortem, occasional NFTs can be found in the
brains of seemingly healthy 30-year-olds [7]. In vivo PET imaging
studies have demonstrated that Aβ deposition can occur 10–20 years
before the appearance of symptoms, and, as per the amyloid cascade
neuroinflammation hypothesis, this amyloid Aβ deposition leads to
a sequence of downstream events which result in tau hyperphos-
phorylation and aggregation leading to neuronal dysfunction.
While the imaging studies support the theory that amyloid Aβ
deposition takes place decades before symptoms appear, it is now
recognized that Aβ deposition alone cannot account for the cogni-
tive dysfunction present in AD. It has been shown that cognitively
normal subjects with increased amyloid Aβ may have accelerated
neurodegeneration and with progressive cognitive impairment, but
the levels of amyloid Aβ do not correlate with cognitive function [8,
9]. However, studies have demonstrated that tau deposition corre-
lates with cognitive function [10] and continues to progress
throughout the course of disease, correlating with symptom sever-
ity [11]. It has also been demonstrated that injection of Aβ42 into
transgenic tau mouse models leads to accelerated tau formation,
and it spreads rapidly from the site of injection demonstrating that
tau can propagate in neuronal networks [12, 13].
Different ultrastructural forms of tau can cause different disease
phenotypes. Different diseases have different isoforms, and the
brains of patients with AD contain equal amounts of 3R and 4R
isoforms; Pick’s disease is characterized by 3R isoform aggregation
into Pick bodies, while CBD, PSP, and argyrophilic grain disorders
contain aggregated 4R isoforms [1]. In addition to the above con-
ditions, chronic traumatic encephalopathy (CTE), present in sub-
jects who have suffered serial mild concussive brain injuries due to,
for example, American football, rugby, boxing, or frequent falls due
to multiple etiology, is characterized histologically by tau deposi-
tion in areas of axonal injury [14]. Hence, it is essential to have a
biomarker which could detect tau deposition in vivo. Studies have
shown that the deposition of abnormal tau correlates closely with
cell dysfunction and symptoms in contrast to amyloid Aβ deposition
which occurs before symptoms appear. While amyloid Aβ may have
an early and significant role in AD, the disappointing results from
anti-amyloid therapies suggest that amyloid Aβ alone is not enough
to cause neuronal damage, which again highlights the importance
of imaging tau [3]. There are also suggestions that, similar to the
presence of amyloid Aβ in the cognitively rate over time [15, 16], it
is possible that tau deposition can occur early, and persistent tau
deposition could lead to progressive disease. Tau imaging will allow
us to evaluate whether tau can cause neuronal damage indepen-
dently of amyloid Aβ production, and indeed whether tau produc-
tion is associated with neuroinflammation.
192 Paul Edison
3 Tau Imaging Agents
For an effective tau imaging agent, it must cross the blood-brain

barrier, and should be taken up by neurofibrillary tangles, and
should not bind to other aggregated proteins like amyloid. They
should also be eliminated very quickly, and could be imaged using
11C and 18F radionuclides [2, 17], should have low nonspecific
binding, and should be able to quantify by using simple quantitative
methods. For clinical utility, these tracers should have the potential
to differentiate easily between the patients and the controls even by
visual inspection. Due to the predominantly intracellular nature of
the tau, these agents should be able to engage readily with the tar-
gets and should have specific binding which increases with the den-
sity of the neurofibrillary tangles, ideally able to engage with the
targets (tau tangles) within the cells which undergo predominant
neurodegeneration. Additionally, these tracers should bind with
different forms of posttranslational modifications. Whether differ-
ential binding to 3R or 4R is beneficial should still be evaluated.
Human tau is composed of six distinct splice variants that vary
with respect to inclusion or exclusion of three alternatively spliced
segments. In addition to phosphorylation and glycosylation, new
modifications, such as the acetylation of amino groups, are being
discovered. This could differentially influence the different binding
affinity of closely related molecules. Such heterogeneity may con-
tribute to the differential binding affinity for potential new tau
tracers. Successful radiotracers must bind with high selectivity to
tau and not be influenced by the amyloid or any other protein or
substrates. It is also important that binding potential should repre-
sent the density of the tau (ideally receptor binding), and Bmax be
proportional to the concentration of the tau protein.
Tau is an intracellular protein and is present in much lower
quantities than amyloid, and the intracellular nature of the tau
requires the tracer to cross the blood-brain barrier and the plasma
membrane which will require tracers with low molecular weight.
This is further compounded by the fact that tau is present in much
lower quantities than amyloid, which will mean that a tau tracer
should have high specific binding to tau and minimal binding to
amyloid [18–22]. As with all PET tracers, it should be able to clear
from the brain very rapidly and should not have active metabolites
that enter the brain [19–22]. Figure 1 shows a number of tau PET
imaging agents in current use.
3.1 [18F]T807 [18F]T807 (Flortaucipar) was originally tested in vivo and demon-
(AV-1451) strated high binding affinity and selectivity [23]. [18F]T807 dem-
onstrated high specificity, with 25-fold selectivity for PHF-tau over
Aβ, with rapid brain penetration and washout, and no active
metabolites entering the brain [24]. While [18F]T807 was found
(a) [18F]THK-523 (b) [18F]THK-5105 (c) [18F]THK-5117
O
18F OH OH
O 18F O
N 18F
H
N N N
H CH3 CH3
N N
CH3 CH3
(d) [18F]THK-5351
(e) [11C]PBB3
H OH N
F O
S H
HO
N N
11CH3
CH3
N N
(f) [18F]T808
(g) [18F]T807
H
N N
N N N
N
N H
N 18F
18F
Fig. 1 Chemical structures of tau tracers in current use. (a) [18F]THK-523, (b) [18F]THK-5105, (c) [18F]THK-
5117, (d) [18F]THK-5351, (e) [11C]PBB3, (f) [18F]T808, and (g) [18F]T807
promising, a closely related compound, [18F]T808, was thought to

be unstable due to the defluorination. Additionally, [18F]T807
uptake closely correlated with histology. Initial in vivo studies in
humans demonstrated good tracer kinetics, with a potential to dis-
criminate patients who may have significant tau uptake compared to
those who have not got any retention [25]. The preliminary work
demonstrated significant deposition in AD compared to controls,
and later studies further substantiated the original finding [26]
(Fig. 2). It was further demonstrated that, in MCI subjects as the
disease advances, there is progressive increase in tau deposition, and
this was associated with a worse performance in cognitive function
tests [27, 28]. It was also encouraging to see, in patients with poste-
rior cortical atrophy, that there was a regional increase in the poste-
rior cortex. However, consistent with the n europathology, there was
a regional predilection of the tau deposition. In PCA there was
occipitotemporal and occipitoparietal uptake. Tau was also increased
in FTD-tau and PSP, CBD, PD, dementia with Lewy bodies (DLB),
and cerebral Aβ angiopathy (CAA), along with logopenic variant
primary progressive aphasia, where the regional predilection differed
depending on the specific site of involvement in different diseases
194 Paul Edison
Fig. 2 [18F]AV-1451 in AD, MCI, and control subjects. Top panel demonstrates transverse section of tau images
using [18F]AV1451 in Alzheimer’s subjects (a), mild cognitive impairment subjects (b), and healthy controls (c).
Bottom panel demonstrates the coronal sections for the corresponding subjects
[29, 30]. Other conditions where tau [31] is primarily driving the
disease process like FTD, chronic traumatic encephalopathy (CTE),
presents with different pattern of deposition of tau [32]. In cogni-
tively normal elderly subjects, [18F]T807 binding has been shown
to correlate with levels of CSF tau [33].
3.2 THK Compounds Okamura and colleagues reported first [18F]THK compounds,
of which [18F]THK523 was evaluated at the Tohuku University
[34]. Pharmacokinetic studies showed excellent uptake and
clearance in the brains of mice [35], and in vivo studies demon-
strated significant uptake. While the uptake correlated well with
cognitive function, there was significant uptake in white matter
[36] making the tracer less favorable for clinical use. Two closely
related compounds, [18F]THK5105 and [18F]THK5117,
were then evaluated. In vivo studies showed that there was
higher cortical retention in AD patients compared with healthy
controls, and this correlated well with cognitive test results and
brain volumes [37]. [18F]THK5117 uptake in AD and MCI
subjects correlated well with cognitive performance, and there
was some correlation between tau uptake and amyloid load.

There was increased binding in the temporal region in all sub-
jects with some correlation between the uptake and cerebral
glucose metabolism [38].
In an attempt to improve the specificity of these compounds,
[18F]THK5351 was recently developed, and preliminary studies
in AD and healthy control subjects demonstrated very high affinity
in classical regions with tau deposition. In further comparisons
between [18F]THK5351 and [18F]THK5117, the former dem-
onstrated faster kinetics, potentially allowing for better visualiza-
tion [39]. However, recent studies have demonstrated that these
compounds could be blocked by selegiline, indicating the non-
specificity of the tracer and the interaction with MAOB.
3.3 [11C]PBB3 2-((1E,3E)-4-(6-((11)C-methylamino)pyridin-3-yl)buta-1,3-

dienyl)benzo[d]thiazol-6-ol ((11)C-PBB3) was another PET
probe developed for tau imaging. It has been shown that [11C]
PBB3 binds strongly to NFTs in AD brains. Fluorescence micros-
copy demonstrated labelling of non-ghost and ghost tangles along
with detection of dystrophic neurites. Interestingly, radioligand
binding to brain homogenates revealed multiple binding compo-
nents with differential affinities for [11C]PBB3 with higher avail-
ability of binding sites on PSP tau deposits for [11C]PBB3 than
[18F]AV-1451, and it is proposed that [11C]PBB3 can capture a
wide range of tauopathies. Preliminary studies have demonstrated
that there is significant binding in areas where there is presence of
tau [40, 41]. These preliminary reports suggested that this could
bind to other tauopathies, even with the absence of amyloid Aβ,
and potentially could have a wide range of uses [40].
3.4 Other Gobbi et al. have developed further high affinity binders,
Compounds RO6931643, RO6924963, and RO6958948. These compounds
bind with high affinity and specificity to the [3H]T808 binding
site on tau aggregates [42]. Importantly, they lack affinity for Aβ
plaques and showed low nonspecific binding to healthy brain tis-
sue. Additionally, the compounds demonstrate macro- and micro-
colocalization of radioligand binding, rapid brain entry, and
washout with well-tolerated metabolic patterns. These tracers are
currently in clinical testing.
4 Conclusions
Tau imaging will allow us to visualize and quantify the tau aggrega-
tion in neurodegenerative diseases and is able to provide us with
further information about the neuropathology of the disease pro-
cess. This could revolutionize the field, as we are now able to
understand more about the disease trajectory, and test different
196 Paul Edison
novel anti-tau therapeutic strategies. Tau imaging has the potential

of becoming part of the diagnostic algorithm in the clinical
practice.
References
1. Spillantini MG, Goedert M (2013) Tau pathol- 12. Gotz J, Chen F, van Dorpe J, Nitsch RM
ogy and neurodegeneration. Lancet Neurol (2001) Formation of neurofibrillary tangles in
12(6):609–622. https://doi.org/10.1016/ P301l tau transgenic mice induced by Abeta
s1474-4422(13)70090-5 42 fibrils. Science 293(5534):1491–1495.
2. Lee HG, Perry G, Moreira PI et al (2005) https://doi.org/10.1126/science.1062097
Tau phosphorylation in Alzheimer’s disease: 13. Lewis J, Dickson DW, Lin WL et al (2001)
pathogen or protector? Trends Mol Med Enhanced neurofibrillary degeneration in
11(4):164–169. https://doi.org/10.1016/j. transgenic mice expressing mutant tau and
molmed.2005.02.008 APP. Science 293(5534):1487–1491. https://
3. Giacobini E, Gold G (2013) Alzheimer disease doi.org/10.1126/science.1058189
therapy—moving from amyloid-beta to tau. 14. Gandy S, Ikonomovic M, Mitsis E, Elder G,
Nat Rev Neurol 9(12):677–686. https://doi. Ahlers S, Barth J, Stone J, DeKosky ST (2014)
org/10.1038/nrneurol.2013.223 Chronic traumatic encephalopathy: clinical-
4. Grundke-Iqbal I, Iqbal K, Tung Y-C, Quinlan biomarkers correlations and current concepts
M, Wisniewsk H, Binder L (1986) Abnormal in pathogenesis. Mol Degener 9:37
phosphorylation of the microtubule-asso- 15. Villemagne VL, Pike KE, Chetelat G et al
ciated protein tau in Alzheimer cytoskel- (2011) Longitudinal assessment of Abeta and
etal pathology. Proc Natl Acad Sci U S A cognition in aging and Alzheimer disease.
83:4913–4917 Ann Neurol 69(1):181–192. https://doi.
5. Kosik K, Joachim C, Selkoe D (1986) org/10.1002/ana.22248
Microtubule-associated protein tau is a major 16. Villemagne VL, Burnham S, Bourgeat P et al
antigenic component of paired helical filaments (2013) Amyloid B deposition, neurodegen-
in Alzheimer disease. Proc Natl Acad Sci U S A eration, and cognitive decline in sporadic
83:4044–4048 Alzheimer’s disease: a prospective cohort
6. Braak H, Braak E (1991) Neuropathological study. Lancet Neurol 12:357–367. https://
staging of Alzheimer related changes in doi.org/10.1016/S1474-4422(13)70044-9
Alzheimer’s disease. Acta Neuropathol 17. Laruelle M, Slifstein M, Huang Y (2003)
82:239–259 Relationships between radiotracer properties
7. Braak H, Braak E (1997) Frequency of stages and image quality in molecular imaging of the
of Alzheimer-related lesions in different age brain with positron emission tomography. Mol
categories. Neurobiol Aging 18(4):351–357 Imaging Biol 5(6):363–375
8. Becker JA, Hedden T, Carmasin J et al 18. Villemagne VL, Furomoto S, Fodero-Tavoletto
(2011) Amyloid-beta associated cortical thin- M et al (2012) The challenges of tau imaging.
ning in clinically normal elderly. Ann Neurol Future Neurol 7(4):409–421
69(6):1032–1042. https://doi.org/10.1002/ 19. Villemagne VL, Okamura N (2014) In vivo tau
ana.22333 imaging: obstacles and progress. Alzheimers
9. Engler H, Forsby A, Almqvist O et al (2006) Dement 10(3 Suppl):S254–S264. https://doi.
Two year follow up of amyloid deposition org/10.1016/j.jalz.2014.04.013
in patients with Alzheimer’s disease. Brain 20. Okamura N, Harada R, Furumoto S et al (2014)
129:2856–2866 Tau PET imaging in Alzheimer’s disease. Curr
10. Gómez-Isla T, Hollister R, West H et al (1997) Neurol Neurosci Rep 14(11):500. https://
Neuronal loss correlates with but exceeds neu- doi.org/10.1007/s11910-014-0500-6
rofibrillary tangles in Alzheimer’s disease. Ann 21. Shah M, Catafau AM (2014) Molecular
Neurol 41:17–24 imaging insights into neurodegeneration:
11. Ingelsson M, Fukumoto H, Newell KL et al focus on tau PET radiotracers. J Nucl Med
(2004) Early amyloid accumulation and pro- 55(6):871–874. https://doi.org/10.2967/
gressive synaptic loss, gliosis and tangle for- jnumed.113.136069
mation in the Alzheimer’s disease brain. 22. Villemagne V, Fodero-Tavoletti M, Masters
Neurology 62:925–931 CL, Rowe C (2015) Tau imaging: early
progress and future directions. Lancet Neurol 32. Dickerson B, McGinnis S, Gomperts S et al.
14(1):114–124 (2015) [18F]T807 of frontotemporal lobar
23. Gao M, Wang M, Zheng QH (2014) Concise degeneration. Human Amyloid Imaging
and high-yield synthesis of T808 and T808P Conference Book of Abstracts 2015 P15
for radiosynthesis of [(18)F]-T808, a PET tau Submission ID 61:37
tracer for Alzheimer’s disease. Bioorg Med 33. Chhatwal J, Schultz A, Marshall G et al. (2015)
Chem Lett 24(1):254–257. https://doi. Entorhinal, parahippocampal, and inferior tem-
org/10.1016/j.bmcl.2013.11.025 poral F18-T807 SUVR correlates with CSF
24. Xia CF, Arteaga J, Chen G et al (2013) [(18) total tau and tau T181P in cognitively normal
F]T807, a novel tau positron emission tomog- elderly. Human Amyloid Imaging Conference
raphy imaging agent for Alzheimer’s disease. Book of Abstracts 2015 Submission ID 115:88
Alzheimers Dement 9(6):666–676. https:// 34. Okamura N, Suemoto T, Furumoto S et al
doi.org/10.1016/j.jalz.2012.11.008 (2005) Quinoline and benzimidazole deriva-
25. Chien DT, Bahri S, Szardenings AK et al tives: candiate probes for in vivo imaging of tau
(2013) Early clinical PET imaging results with pathology in Alzheimer’s disease. J Neurosci
the novel PHF-tau radioligand [F-18]-T807. 25(47):10857–10862
J Alzheimers Dis 34(2):457–468. https://doi. 35. Harada R, Okamura N, Furumoto S et al
org/10.3233/JAD-122059 (2013) Comparison of the binding characteris-
26. Pontecorvo M, Devous Ml, Joshi A, Lu M, tics of [18F]THK523 and other amyloid imag-
Siderowf A, Arora A, Mintun M (2015) ing tracers to Alzheimer’s disease pathology.
Relationships between florbetapir PET amy- Eur J Nucl Med Mol Imaging 40(1):125–132
loid and 18F AV-1451 (aka 18F-T807) PET 36. Villemagne VL, Furumoto S, Fodero-Tavoletti
tau binding in cognitively normal subjects and MT et al (2014) In vivo evaluation of a novel
patients with cognitive impairments suspected tau imaging tracer for Alzheimer’s disease. Eur
of Alzheimer’s disease. Human Amyloid J Nucl Med Mol Imaging 41(5):816–826
Imaging Conference Book of Abstracts 2015 37. Okamura N, Furumoto S, Fodero-Tavoletti
ID submission 98:131 MT et al (2014) Non-invasive assessment of
27. Johnson K, Becker J, Sepulcre J et al. (2014) Alzheimer’s disease neurofibrillary pathol-
Tau PET: initial experience with F18 T807. ogy using 18F-THK5105 PET. Brain 137(Pt
Human Amyloid Imaging Conference Book of 6):1762–1771. https://doi.org/10.1093/
Abstracts 2014 ID submission 103:22 brain/awu064
28. Mintun M, Devous M, Pontecorvo M, Joshi 38. Saint-Aubert L, Lemoine L, Marutle A et al.
A, Siderowf A, Johnson K, Navitsky M, Lu (2015) Relationship between post-mortem
M (2015) Potential for PET imaging tau THK5117 binding and in-vivo PET biomarkers
tracer 18F-AV-1451 (also known as 18F- uptake in Alzheimer’s disease. Human Amyloid
T807) to detect neurodegenerative progres- Imaging Conference Book of Abstracts 2015
sion in Alzheimer’s disease. Human Amyloid ID submission 94:94
Imaging Conference Book of Abstracts 2015 39. Harada R, Okamura N, Furumoto S et al.
Submission ID 95:87 (2015) First-in-human PET study of a novel
29. Ossenkoppele R, Schonhaut D, Baker S et al. tau tracer [18F]THK5351. Human Amyloid
(2015) Distinct [18F]AV1451 retention pat- Imaging Conference Book of Abstracts 2015
terns in clinical variants of Alzheimer’s disease. Submission ID 24
Human Amyloid Imaging Conference Book of 40. Maruyama M, Shimada H, Suhara T et al
Abstracts 2015 P55 Submission 113:62 (2013) Imaging of tau pathology in a
30. Marquie M, Normandin M, Vanderburg C tauopathy mouse model and in Alzheimer
et al. (2015) Towards the validation of novel patients compared to normal controls.
PET tracer T807 on postmortem human Neuron 79(6):1094–1108. https://doi.
brain tissue samples. Abstract Online: Human org/10.1016/j.neuron.2013.07.037
Amyloid Imaging Book of Abstracts 2015 41. Shimada H, Higuchi M, Ikoma Y et al
Submission ID 89:28 (2013) In vivo visualization of tau pathol-
31. Mitsis EM, Riggio S, Kostakoglu L et al (2014) ogy. Alzheimers Dement 9(4):P101–P102.
Tauopathy PET and amyloid PET in the diag- https://doi.org/10.1016/j.jalz.2013.05.172
nosis of chronic traumatic encephalopathies: 42. Gobbi LC, Knust H, Körner M et al (2017)
studies of a retired NFL player and of a man Identification of three novel radiotracers for
with FTD and a severe head injury. Transl imaging aggregated tau in Alzheimer’s disease
Psychiatry 4:e441. https://doi.org/10.1038/ with positron emission tomography. J Med
tp.2014.91 Chem 60(17):7350–7370
Chapter 14
Retinal Imaging in Early Alzheimer’s Disease

Tom MacGillivray, Sarah McGrory, Tom Pearson, and James Cameron
Abstract
Changes in the brain that lead to Alzheimer’s disease are thought to start decades before cognitive symp-
toms emerge. If biomarkers for these early stages could be identified, it would contribute to a more accu-
rate estimation of an individual’s risk of developing disease and enable the monitoring of high-risk
(presymptomatic) persons as well as providing the means for assessing the efficacy of new interventions.
The retina links to the visual processing and cognitive centers of the brain, but it is also an extension of the
brain sharing embryological origins as well as a blood supply and nerve tissue. It therefore has huge poten-
tial as a site for biomarker investigation through easy, noninvasive imaging and computational image analy-
sis to reveal valuable information about microvascular health, deposition, and neurodegenerative damage.
Capturing reliable longitudinal data pertaining to the onset of Alzheimer’s disease is a key target, but a
high degree of standardization is necessary if the potential of the retina is to be fully realized. Our goal is
to provide the reader with guidelines on how to execute robust retinal imaging and analysis for neuroreti-
nal biomarker discovery and to highlight advantages and limitations of the techniques.
Key words Alzheimer’s disease, Dementia, Retinal imaging, Non-invasive, Biomarker, Fundus,
Blood vessel, Neurodegeneration
1 Introduction
The retina is the light-sensing inner surface of the human eye and is
comprised of layers of tissue consisting of neurons and supporting
cells, interconnected by synapses (see Fig. 1). The retinal ganglion
cell (RGC) layer, the innermost cellular layer, projects its axons
across the inner retina, called the retinal nerve fiber layer (RNFL),
through the optic nerve to the visual processing and cognitive cen-
ters of the brain. Developmentally and anatomically, the retina is an
extension of the brain sharing embryonic origins as well as features,
besides nerve tissue, such as small blood vessels and a blood-tissue
barrier [1]. It also has the advantage of being easier to image than
the brain and with far superior resolution—resolvable detail in
images of the eye being at a micron level compared to millimeters
with conventional magnetic resonance (MR) imaging. The retina is
199
200 Tom MacGillivray et al.
Fig. 1 Anatomy of the human eye [26]. (a) Diagram showing the main structures. (b) Illustration of the organi-
zation of the retinal layers and cells
therefore an appealing research site with the potential to reveal

important aspects of brain health (see Fig. 1).
Cerebral small vessel disease is part of Alzheimer’s disease
(AD) etiology [2, 3]. Damage to the lining of the small vessels in
the brain sees failure of this barrier and leakage of material into tis-
sue, which causes further injury. However, these vessels are diffi-
cult to visualize directly in vivo. Due to the homology between the
retinal and cerebral microvasculature, the condition of the retinal
vasculature is a proxy for cerebral microvascular health [4].
Measuring geometric features of the vascular pattern visible in reti-
nal imaging (such as vessel widths, tortuosity, and network branch-
ing complexity) infers the health of the retinal microvascular system
and therefore the state of the cerebral microvasculature. Thus, the
retina offers opportunity for changes or abnormalities in its micro-
vasculature to act as surrogate markers of microvascular pathology
in AD. Indeed, such changes have previously been reported in
patients with AD (the most consistent finding being a sparser reti-
nal vascular network), but the results across studies have been vari-
able with inconsistencies most likely due to the application of
assorted methodologies [5]. Standardizing methods and measure-
ments will likely advance biomarker discovery in this area.
Deposition of amyloid β (Aβ) protein in the brain is a key path-
ological feature of AD. Aβ also accumulates in the inner layers of
the retina [6]. Novel methods are emerging for detection in the
back of the eye such as fluorescence imaging of labeled Aβ plaques
[7]. However, these remain in their infancy with continuing devel-
opments in the technology and initial studies largely confined to
nonhuman subjects. Aβ is also found in drusen, a marker of age-
related macular degeneration (AMD) predominantly occurring in
the macula at the center of the retina [8]. Drusen are abnormal
Retinal Imaging in Early Alzheimer’s Disease 201
accumulations of extracellular material that build up in the retinal

pigmented epithelium (RPE) layer and appear as white or yellow
spots when imaging the fundus or surface of the retina. The pro-
gression of the formation of drusen in the peripheral retina was
more prevalent in patients with AD in comparison to age-matched
controls [9]. Preliminary findings, albeit on a small cohort (n = 29),
suggest a nascent imaging-derived biomarker for similar plaque
formation in the brain that merits further investigation.
Optical coherence tomography (OCT) yields cross-sectional
images of the internal retinal structure, and the high resolution of
newer systems feature automated measurement of the different
cellular layers. Quantification of RNFL thickness (and diminution
over time) is a marker of axonal loss, and thinning occurs in AD
and its preceding stage of mild cognitive impairment [10].
Neuronal loss can be accessed indirectly from macular thickness
measurements (as 30–35% of the retina thickness in the macular is
comprised by the RGCs and their fibers) and, with recent advance-
ments in the technology, thickness mapping of the ganglion cell
complex (GCC) layer which comprises RNFL, RGC, and the inner
plexiform layer [11, 12]. Hence, OCT presents additional oppor-
tunity for neuroretinal biomarker discovery.
OCT angiography (OCT-A) is a recent advancement in retinal
imaging that enables visualization of the microvasculature within the
retina using a fast capture of OCT over a small volume and then deci-
phering the component of the returning signal related to blood flow
[13]. This modality images very small vessels that are otherwise only
visible with more invasive fluorescein angiography imaging. With small
vessel disease linked to AD, OCT-A may reveal valuable information
pertaining to the earliest signs of changes in the retina and brain.
We now present an overview of established retinal imaging
modalities (see Sect. 2) and their analysis methodologies (see Sect. 3)
that are currently viable for biomarker discovery in AD.
2 Materials
2.1 Setup Prior Table 1 gives a summary of the principal retinal imaging modalities we
to Imaging use for biomarker discovery in AD. Retinal imaging is a patient-friendly
tool that is noninvasive, quick, and easy to perform. Most people are
familiar with the typical imaging procedure, as many of them will have
attended the high street optometrist for routine eye health checks. For
each instrument (see Note 1), the person being imaged is asked to sit
with his or her chin on a rest and with their forehead placed firmly
against a stabilizing bar (see Note 2). Before proceeding, the instru-
ment operator checks that the person is comfortable, adjusting the
height of the instrument table if necessary. The operator also adjusts
the headrest frame to ensure that the individual’s eye is in line with the
height adjustment mark (usually a horizontal white line). The person is
Table 1
A summary of the principal retinal imaging modalities
Light source Field of Resolution

Modality (nm) view (°) (μm to pixel) Description
Optical 780 – – Interferometry measures the eye’s axial
biometrya length, distance from front to back
Fundus white flash 45 7 (a) Color photograph showing blood vessels,
camerab optic nerve head, and macula
SLOc 820 30 5 (a) Infrared laser scans retina to produce an image
similar to a fundus camera
UWF-SLOd 532 and 633 135 10–15 (a) Two lasers scan a larger field of view showing
areas of peripheral retina
OCTc 870 10–30 3.9–7 (a), 14(t) Infrared light penetrates retina and
interferometry resolves tissue layers
OCTe 840 10–30 5a, 15 (t)
OCT-Ae 840 3–10 5 (a) Differences in backscattered OCT between
sequential scans at same cross-section used
to construct a map of blood flow
Included is resolution, with axial (a) and transverse (t) values where appropriate
a
IOL Master (Carl Zeiss Meditec, Germany)
b
Canon CR-DGi (Canon USA Inc., Lake Success, NY)
c
Heidelberg SPECTRALIS (Heidelberg Engineering, Dossenheim, Germany)
d
Optos Daytona (Optos Plc, Dunfermline, UK)
e
RTVue XR 100 Avanti (Optovue, Inc., Fremont, CA)
then asked to look at a fixation light (see Note 3) while the operator
focuses and aligns the system before instigating the image acquisition,
which involves the instrument illuminating the retina with its light
source and then capturing the reflected light.
We image both eyes (see Note 4) with each of the devices and
without administering drops for dilation (see Note 5). Conducting
several imaging protocols one after the other can see an individual
become fatigued resulting in a decline of image quality. While pupil
dilation might help mitigate this effect, the ability to conduct
imaging without it amounts to a real advantage as administering
eye drops adds considerable complexity to the protocols. For
instance, individuals must have their intraocular pressure and ante-
rior chamber (the fluid-filled space inside the eye between the iris
and the cornea) assessed as a precaution for acute angle-closure
glaucoma, which is more common in older people. There is also a
restriction on driving and a sensitivity to bright light, which extends
for several hours after administering drops.
Where possible, we perform all the imaging of an individual for

a particular study visit on the same day. In addition to the modali-
ties described below, we use optical biometry to measure axial
length, i.e., the distance from the front of the eye to the back. Axial
length may influence the measurement of some retinal parameters,
such as vessel widths, and thus needs adjusted for in data analysis.
However, the significance of this measurement error is uncertain,
and along with other variables such as instrument optics/magnifi-
cation, it may be that future focus will be on retinal metrics more
independent of optical and refractive variables like global vascular
branching complexity.
We highlight retinal imaging’s susceptibility to the presence of
media opacities, such as cataract, which are more common in the
elderly population. The presence of opacities in the eye can obscure the
retinal detail, although experienced operators may be able to adjust the
instrument positioning and operation to achieve useful images.
2.2 Fundus Imaging Fundus imaging generates a 2D image of the retina directly behind
the pupil. A fundus camera achieves this with a specialized low-
power microscope, an attached digital camera, and a flash of white
light for illumination, which creates a color photograph of the ret-
ina. We use a Canon CR-DGi (Canon USA Inc., Lake Success, NY)
to image an area showing the optic disk (OD) or optic nerve head,
macula, and blood vessels (see Fig. 2a). We standardize by center-
ing the image acquisition midway between the centers of the OD
and macula to give a digital fundus photograph suitable for quanti-
tative analysis of key vascular and structural features. Repeating the
same acquisition when capturing further images within or between
individuals reduces uncertainties when comparing measurements.
A scanning laser ophthalmoscope (SLO) such as in the
Heidelberg SPECTRALIS (Heidelberg Engineering, Dossenheim,
Germany) uses an infrared (IR) laser beam and confocal micro-
scope arrangement to scan across the surface of the retina and form
a fundus picture. This particular SLO images at a slightly higher
resolution than the fundus camera but captures a smaller area (see
Table 1; Fig. 2b). The main purpose of the SLO imaging in the
SPECTRALIS is to guide the operator to the desired location for
OCT scanning and to enable accurate visit-to-visit registration for
repeat imaging sessions. Yet, the clarity and resolution of the SLO
images lend themselves to additional computational analysis of the
retinal vasculature, especially the larger vessels where small changes
or abnormalities might be more feasible to detect than with the
fundus camera. We center imaging on the optic nerve head.
To image more of the retina surface in a single capture, an ultra-
widefield SLO (UWF-SLO) employs scanning mirrors to direct laser
light to regions of the back of the eye that are inaccessible with a
fundus camera or standard SLO. We use an Optos Daytona (Optos
Fig. 2 Analyzing the fundus. (a) Color fundus camera image of a right eye showing the region containing the
OD, macula, and blood vessels. (b) SLO imaging of the same eye centered on the OD and acquired with an IR
light source, which generates a grayscale image. Green circle denotes standard positioning of peripapillary
OCT acquisition around the optic nerve head. (c) UWF-SLO view centered on the macula, imaging a larger
region of the retina than the fundus camera (dashed yellow circle) and SLO (dashed yellow square). The stan-
dard positioning of OCT-A acquisitions centered on the OD and macula is also denoted (solid magenta squares).
(d–f) Retinal blood vessels automatically detected by computerized image processing techniques in each of
the three fundus imaging modalities. (g) SLO image overlaid with the standard measurement zone B (dashed
cyan circles) used in the analysis of blood vessel widths. The six largest arterioles (red) and venules (blue) and
their individual widths (yellow) are used to calculate CRAE, CRVE, and AVR. (h) Generalized dimension spec-
trum (Dq) calculated for the vasculature detected in the fundus camera image and used to quantify the com-
plexity/sparsity of the vascular branching pattern. For each value of q, the calculation of Dq from the detected
vessels is repeated a number of times to give a mean value and associated error. (i) Width measurements
(black dots) performed along the path of a vessel extracted from the UWF-SLO image with straight line fit
(dashed black line) to quantify tapering as the vessel heads out toward the periphery
Plc, Dunfermline, UK) to facilitate the study of features and changes

in the periphery that we cannot see with the other modalities,
though at the sacrifice of a lower resolution for an increase in image
field (see Table 1; Fig. 2c). Image acquisition is centered on the
macula.
2.3 Optical OCT enables cross-sectional imaging of the internal retinal structures
Coherence (with a limited tissue depth <2 mm). An optical beam directed at the
Tomography retina and interferometry to resolve the back-scattered light signals
and Angiography and determine the depth of reflecting structures is an axial scan or
A-scan. Scanning the beam across the retina to acquire numerous
A-scans builds up the 2D image. The Heidelberg SPECTRALIS uses
spectral domain OCT (SD-OCT) in which the depth information is
computed by analyzing the interference signal based on the wave-
length of returning light. We capture a circular peripapillary OCT
scan in order to measure changes in structure around the OD, par-
ticularly in RNFL thickness (see Fig. 3a–d). In addition, we capture a
volume scan centered on the macula to measure retinal and RGCL
layer thickness. The RTVue XR Avanti (Optovue, Inc., Fremont,
CA), also a SD-OCT, maps RNFL and GCC layer thickness with an
OD-centered OCT acquisition (see Fig. 3e, f).
The RTVue XR Avanti has an angiographic mode that reveals
the very small vessels at different depths in the retina. We perform
Fig. 3 Analyzing the retinal layers. (a) SLO image focused on OD, showing the location of the peripapillary OCT
scan acquisition (green circle). (b) OCT image showing the retinal layers and delineation of the ILM (red line) and
RNFL border (cyan line), which gives RNFL thickness. (c) Evaluating RNFL thickness using the standard TSNIT
map which subdivides the measurements and color codes them compared to normative values. (d) RNFL thick-
ness profile (black line) with color coding emphasizing statistical significance in comparison to normal limits—
green indicates within normal limits; blue, borderline above; yellow, borderline below; and red, below. The person
imaged here shows mostly green, whereas we might expect more yellow and red for a person with AD progres-
sion. (e) Map of RNFL thickness around the OD and (f) GCC thickness, also color coded to emphasize the presence
of any abnormally thin areas. Again, in AD progression, thinner or more red areas might be anticipated
Fig. 4 Analyzing the retina using OCT-A. (a) Nerve head segment and (b) radial peripapillary capillary (RPC)
segment and (c) superficial vascular plexus and (d) deep vascular plexus region in the macula. Measuring
retinal thickness from OCT and vessels density from OCT-A using a TSNIT map (blue zones) in the superficial
vascular plexus of (e) a normal eye and (f) an eye of individual who has suffered a stroke (and so thought to
have deteriorated vascular brain health). (g–h) Color coding of vessel density measurements performed using
a 3 × 3 grid (white dashed lines) to subdivide the superficial vascular plexus into squares. The individual
believed to have poor vascular brain health exhibits a lower vessel density
2 OCT-A captures, one centered on the OD and one centered on

the macula (see Fig. 4a–d). This procedure is the most susceptible
to patient movement (see Note 6).
3 Image Analysis
3.1 Overview We utilize expert human grading to detect and classify retinopathy
(i.e., type, number, size, and location), including drusen, in our
retinal imaging. Computerized software tools that exploit image
processing and analysis techniques enable further quantitative and
reproducible measurements of structural features such as the vas-
culature visible in fundus images and the different retinal layers in
OCT [14, 15]. Such computer-assisted methods measure varia-
tions or abnormalities in the retina that might be imperceptible to
or missed by a human grader, in an objective and repeatable
fashion. In this section, we focus on the key quantitative analysis

pertaining to the modalities previously described in Sect. 2.
3.2 Fundus Imaging Our group has developed software (VAMPIRE; Vascular Assessment
(See Fig. 2) and Measurement Platform for Images of the REtina, Universities of
Edinburgh and Dundee, UK) to extract parametric information
from fundus imaging [16–18]. Retinal vessel segmentation is the
precursor to the measurement of features such as vessel width, tor-
tuosity, and complexity of the vascular network [19]. We perform
this for fundus camera, SLO, and UWF-SLO images (see Fig. 2d–f).
Additional semiautomatic classification distinguishes venules from
arterioles, removes non-vessel artifacts, and detects the center and
boundary of the OD along with center of the macula. The operator
can correct any of these labels through a user interface that allows
them to interact with an image being analyzed.
VAMPIRE software places the standard set of circular
measurement zones on a fundus camera or SLO image—Zone B is
a ring 0.5–1 OD diameters away from the center, and Zone C is a
ring extending from OD boundary to 2 OD diameters away.
Measurements of the widths of the six widest arterioles and venules
crossing Zone B (see Fig. 2g) are used to calculate the central retinal
arteriolar equivalent (CRAE) and central retinal venular e quivalent
(CRVE) and give the arteriole to venule ratio (AVR; CRAE/
CRVE) [20]. Similarly, the tortuosity (i.e., how much a vessel
twists and turns) of the six widest arterioles and venules crossing
Zone C is measured [21] and then further summarized by calculat-
ing median arteriolar and venular values along with the standard
deviations and ranges.
The complexity (or sparsity) of the vascular branching pattern
visible with fundus imaging can be quantified using fractal analysis to
yield a fractal dimension that combines contributions of individual
vessel parameters into a single global value [14]. Variations in the
fractal dimension are an indicator of deviation away from normative
branching arrangements and so a potential marker of disease [22].
We use a particular approach called multifractal analysis, which gen-
erates several fractal dimensions at different “scales” (see Fig. 2h),
and we record dimensions D0, D1, and D2 as these have previously
been reported as likely sensitive markers of small vascular changes
[23]. Application of the analysis is to vessels appearing in Zone C
and separated into arteriolar and venular measurements.
With UWF-SLO, it is not only possible to assess retinopathy and
the vasculature in the peripheral retina but to explore novel param-
eters for biomarker discovery in addition to conventional Zone B–C
measurements (see Note 7). For instance, we can analyze the main
vessels over a longer path distance than is achievable with other fun-
dus modalities and look at how the width changes as the vessel
extends out toward the periphery (see Fig. 2i). An abnormal decline
in the vessel’s width (quantified by fitting a model such as a straight
line) may indicate an unhealthy vessel.
Computational analysis of features appearing in fundus imaging

such as the retinal vasculature has the ability to return large num-
bers of measurements. However, further endeavor is required to
develop these into effective summary parameters that characterize
the state or health of the retina and then to investigate which are
relevant to brain changes and AD.
3.3 Optical The major task in analyzing OCT is accurate segmentation of reti-
Coherence nal layers as thickness changes are an important indicator of disease
Tomography status. Modern OCT instruments such as the Heidelberg
and Angiography SPECTRALIS and RTVue XR Avanti are equipped with software
(See Fig. 3) to automatically delineate the borders of the internal limiting
membrane (ILM) and the RNFL, as the distance between these
two boundaries gives the measurement of RNFL thickness (see
Fig. 3b). Thickness of the RNFL is evaluated using the standard
TSNIT (temporal, superior, nasal, inferior, temporal) mapping
approach which subdivides or regionalizes the measurement and
color codes statistical significance compared to a database of nor-
mative values (see Fig. 3c). RNFL thickness also presented as a
profile plot with color coding again underlining statistical signifi-
cance in comparison to normal limits (see Fig. 3d) or as a thickness
map (see Fig. 3e). Further delineation of boundaries in OCT
enables mapping of other layers such as the GCC (see Fig. 3f).
While these quantitative OCT mapping options present exciting
opportunities for biomarker discovery, the pathological processes
behind retinal thinning in AD are incompletely understood, and fur-
ther investigations are required to determine the use of retinal OCT
as a potential surrogate marker of neurodegeneration (Fig. 4).
OCT-A works by comparing consecutive scans at the same loca-
tion to detect flow using motion contrast and enable demarcation of
blood vessels. As a means of quantifying OCT-A, the instrument calcu-
lates vessel density parameters, i.e., the percentage area occupied by the
larger appearing vessels and the microvasculature in particular regions.
This can be the entire scan area as well as standard regions such as
TSNIT sectors or a grid-based division of the image (see Fig. 4e–g).
For each scanned region, the software calculates the vessel densities
and thicknesses in various layers of the retina and optic nerve head.
OCT-A offers potentially exciting avenues to explore in the
context of neuroretinal biomarker discovery, particularly around
small vessels and AD. However, caution is urged in the interpreta-
tion and analysis of OCT-A imaging. The modality detects flow
above a threshold level, and so not all vessels may appear in the
images such as blocked or leaky ones. In addition, it is more suscep-
tible to image artifact caused by patient motion or reflections from
internal retinal structures that may not always be obvious to the
untrained eye or computer algorithm.
4 Conclusions
Monitoring for the development and progression of changes in the

retina via imaging has the potential to yield biomarkers for the
onset of brain changes and deterioration linked to AD. The retinal
imaging protocols described here help support the conduct of
experimental medicine studies to understand and investigate the
in vivo pathophysiology of small disease, deposition and neurode-
generation in AD, and to identify indicative retinal changes.
We highlight that the ease-of-use of modern retinal imaging
devices does not remove the importance of detailed and systematic
review of the captured images, for quality, registration, segmentation
errors, and other important processing steps that are essential before
data extraction. Trained and experienced operators are vital to both
the capture of suitable images and the subsequent measurements.
Future challenges for researchers in this field include distin-
guishing normal retinal changes due to aging from those linked to
failing brain health and investigating whether changes in the retina
relate to other recognized brain changes such as abnormalities on
MR imaging, declining cognitive scores, and measures from bio-
logical samples such as blood and spinal fluid. Thus, an essential
target is reliable longitudinal data, not just in the eye but other
non-retinal markers, in appropriately powered prospective studies.
We envisage that, with concerted investigative effort, over the
next few years, retinal imaging and analysis will contribute biomark-
ers for AD and that this could include presymptomatic screening.
5 Notes
1. Set up and positioning for the Optos device is different to the

other machines. The person sits in front of it, holds the hand-
grips on both sides of the instrument, and places their head
against the faceplate to look through a small eyehole. They
move themselves into the correct position for imaging by
adjusting themselves against the faceplate until a target light
they can see turns from blue or red to green.
2. Having the person’s forehead placed firmly against the stabiliz-
ing bar of the headrest frame helps keep their head very still
and minimizes movement that can result in images that are
unsuitable for analysis.
3. A fixation light is a visual cue inside the device for a person to
look at. This directs imaging to particular parts of the retina. It
can be a different colored light for different instruments but is
most commonly blue, green, or orange.
4. Increasing evidence suggest some elements of the visual system

are asymmetric [24]. We therefore image both eyes in order to
determine the impact that this might have on associations of
retinal parameters with other measurements related to brain
changes and AD.
5. Pupil dilation with eye drops is not necessary with the imaging
protocols described here. A darkened room is usually sufficient
for a person’s pupils to enlarge and to take a picture with the
fundus camera. The other instruments generally work well in
low levels of background lighting. With the exceptions of the
fundus camera and UWF-SLO, the illuminating light source is
not strong and causes only small amounts of pupil constric-
tion. Between acquisitions, we ask a person to replenish their
tear film by either closing or blinking their eyes. Asking them
to fixate with the other eye (i.e., the one not under examina-
tion) on a distant target (and not the internal fixation light)
can also help to enlarge the pupil.
6. The procedure involves two 3–4 s bursts of data capture per
imaging location on the retina during which the person must
stay extremely still and stare at the fixation light. Any slight
movement introduces noticeable “streaking” artifacts into the
resulting OCT-A image.
7. In the analysis of UWF-SLO to measure features, given the
larger surface of imaged retina, it is crucial to account for
the curvature of the eye; otherwise measurements will be
invalid [25].
Acknowledgments
Support from the Engineering and Physical Sciences Research

Council (EPSRC) (grant number EP/M005976/1), the Medical
Research Council (MRC) (grant number MR/L015994/1), the
Alzheimer’s Research UK Scotland Network Centre, the
University of Edinburgh Innovation Initiative Grants scheme,
the Edinburgh and Lothians Health Foundation, Optos, and
SINAPSE (Scottish Imaging Network: A Platform for Scientific
Excellence) is gratefully acknowledged. We also thank the
Computer Vision and Image Processing Group at the University
of Dundee, NHS Lothian R&D, the Edinburgh Clinical Research
Facility, Edinburgh Imaging, and the Anne Rowling Regenerative
Neurology Clinic.
References
1. London A, Benhar I, Schwartz M (2013) The 2:24. https://doi.org/10.1186/
retina as a window to the brain-from eye research s40942-016-0049-4
to CNS disorders. Nat Rev Neurol 9(1):44–53. 12. Lisboa R, Paranhos A Jr, Weinreb RN et al
https://doi.org/10.1038/nrneurol.2012.227 (2013) Comparison of different spectral
2. Knopman DS (2006) Dementia and cerebro- domain OCT scanning protocols for diagnos-
vascular disease. Mayo Clin Proc 81(2):223– ing preperimetric glaucoma. Invest Ophthalmol
230. https://doi.org/10.4065/81.2.223 Vis Sci 54(5):3417–3425. https://doi.
3. Reitz C, Brayne C, Mayeux R (2011) org/10.1167/iovs.13-11676
Epidemiology of Alzheimer disease. Nat Rev 13. de Carlo TE, Romano A, Waheed NK et al (2015)
Neurol 7(3):137–152. https://doi. A review of optical coherence tomography angi-
org/10.1038/nrneurol.2011.2 ography (OCTA). Int J Retina Vitreous 1:5.
4. Patton N, Aslam T, Macgillivray T et al (2005) https://doi.org/10.1186/s40942-015-0005-8
Retinal vascular image analysis as a potential 14. Patton N, Aslam TM, MacGillivray T et al
screening tool for cerebrovascular disease: a ratio- (2006) Retinal image analysis: concepts, appli-
nale based on homology between cerebral and cations and potential. Prog Retin Eye Res
retinal microvasculatures. J Anat 206(4):319– 25(1):99–127. https://doi.org/10.1016/j.
348. https://doi. preteyeres.2005.07.001
org/10.1111/j.1469-7580.2005.00395.x 15. Abramoff MD, Garvin MK, Sonka M (2010)
5. McGrory S, Cameron JR, Pellegrini E et al Retinal imaging and image analysis. IEEE
(2017) The application of retinal fundus cam- Trans Med Imaging 3:169–208. https://doi.
era imaging in dementia: a systematic review. org/10.1109/RBME.2010.2084567
Alzheimers Dement (Amst) 6:91–107. https:// 16. MacGillivray T (2012) VAMPIRE: vessel
doi.org/10.1016/j.dadm.2016.11.001 assessment and measurement platform for
6. Ratnayaka JA, Serpell LC, Lotery AJ (2015) images of the retina. In: Tan EYKNJH, Acharya
Dementia of the eye: the role of amyloid beta in UR, Suri JS (eds) Human eye imaging and
retinal degeneration. Eye (Lond) 29(8):1013– modeling. CRC Press, Boca Raton, FL
1026. https://doi.org/10.1038/eye.2015.100 17. Cameron JR, Ballerini L, Langan C et al (2016)
7. Koronyo-Hamaoui M, Koronyo Y, Ljubimov Modulation of retinal image vasculature analysis
AV et al (2011) Identification of amyloid to extend utility and provide secondary value
plaques in retinas from Alzheimer’s patients and from optical coherence tomography imaging.
noninvasive in vivo optical imaging of retinal J Med Imaging (Bellingham) 3(2):020501.
plaques in a mouse model. Neuroimage https://doi.org/10.1117/1.JMI.3.2.020501
54(Suppl 1):S204–S217. https://doi. 18. Pellegrini E, Robertson G, Trucco E et al
org/10.1016/j.neuroimage.2010.06.020 (2014) Blood vessel segmentation and width
8. Anderson DH, Talaga KC, Rivest AJ et al estimation in ultra-wide field scanning laser
(2004) Characterization of beta amyloid ophthalmoscopy. Biomed Opt Express
assemblies in drusen: the deposits associated 5(12):4329–4337. https://doi.org/10.1364/
with aging and age-related macular degenera- BOE.5.004329
tion. Exp Eye Res 78(2):243–256 19. Trucco E, Giachetti A, Ballerini L et al. (2015)
9. Aslam A, Peto T, Barzegar-Befroei N et al Morphometric measurements of the retinal
(2014) Assessing peripheral retinal drusen pro- vasculature in fundus images with vampire. In:
gression in Alzheimer’s dementia: a pilot study Biomedical image understanding. John Wiley
using ultra-wide field imaging. Invest & Sons, Inc., pp 91–111. doi:https://doi.
Ophthalmol Vis Sci 55(13):659–659 org/10.1002/9781118715321.ch3
10. Thomson KL, Yeo JM, Waddell B et al (2015) 20. Knudtson MD, Lee KE, Hubbard LD et al
A systematic review and meta-analysis of retinal (2003) Revised formulas for summarizing retinal
nerve fiber layer change in dementia, using vessel diameters. Curr Eye Res 27(3):143–149
optical coherence tomography. Alzheimers 21. Lisowska A, Annunziata R, Loh GK et al
Dement (Amst) 1(2):136–143. https://doi. (2014) An experimental assessment of five
org/10.1016/j.dadm.2015.03.001 indices of retinal vessel tortuosity with the
11. Cunha LP, Almeida AL, Costa-Cunha LV et al RET-TORT public dataset. Conf Proc IEEE
(2016) The role of optical coherence tomogra- Eng Med Biol Soc 2014:5414–5417. https://
phy in Alzheimer’s disease. Int J Retina Vitreous doi.org/10.1109/EMBC.2014.6944850
22. Doubal FN, MacGillivray TJ, Patton N et al Eye Res. https://doi.org/10.1016/j.
(2010) Fractal analysis of retinal vessels sug- preteyeres.2017.04.003
gests that a distinct vasculopathy causes 25. Croft DE, van Hemert J, Wykoff CC et al
lacunar stroke. Neurology 74(14):1102– (2014) Precise montaging and metric quantifi-
1107. https://doi.org/10.1212/ cation of retinal surface area from ultra-wide-
WNL.0b013e3181d7d8b4 field fundus photography and fluorescein
23. Stosic T, Stosic BD (2006) Multifractal analysis angiography. Ophthal Surg Lasers Imaging
of human retinal vessels. IEEE Trans Med Retina 45(4):312–317. https://doi.
Imaging 25(8):1101–1107 org/10.3928/23258160-20140709-07
24. Cameron JR, Megaw RD, Tatham AJ et al 26. Kolb H (1995) Simple anatomy of the retina. In:
(2017) Lateral thinking – interocular sym- Kolb H, Fernandez E, Nelson R (eds) Webvision:
metry and asymmetry in neurovascular pat- the organization of the retina and visual system.
terning, in health and disease. Prog Retin University of Utah Health, Salt Lake City, UT
Chapter 15
In Vivo Volumetry of the Cholinergic Basal Forebrain

Michel J. Grothe, Ingo Kilimann, Lea Grinberg,
Helmut Heinsen, and Stefan Teipel
Abstract
Degeneration of cortically projecting acetylcholine-containing neuronal populations within the basal forebrain
cholinergic system (BFCS) is a central pathogenetic aspect of Alzheimer’s disease (AD) and forms the rationale
for the use of cholinomimetics as antidementive treatment. The role of the cholinergic deficit in AD patho-
physiology has mainly been studied in experimental animal models and in neuropathologic examinations of
human autopsy data of advanced disease stages. Interactions between cholinergic deficits and accumulation of
cortical amyloid pathology point to a relevant role of BFCS degeneration for the preclinical stage of AD. The
advent of novel computational morphometry techniques for the automated analysis of high-resolution MRI
data allows studying AD-related atrophy in the living human brain with ever-increasing temporal and regional
detail. Combining these morphometry techniques with recently developed stereotactic mappings of the BFCS
provides a method for automated MRI-based volumetry that can sensitively assess degenerative changes of the
BFCS in vivo. Here, we outline the general methodological approach for MRI-based BFCS volumetry and
describe the specifics of different image processing choices and analysis strategies. We further discuss possibili-
ties and limitations of this method for studying BFCS degeneration in the course of AD, with a special empha-
sis on using MRI-based BFCS volumetry as an imaging biomarker for defining the preclinical disease stage.
Key words Cholinergic basal forebrain, Nucleus basalis Meynert, Substantia innominata, MRI,
Morphometry, Amyloid pathology
1 Introduction
Decades of neuropathologic and pharmacologic research implicate

a cholinergic neurotransmission deficit in the pathophysiology of
Alzheimer’s disease (AD) that is caused by severe neurofibrillary
degeneration of cortically projecting cholinergic neurons of the
basal forebrain cholinergic system (BFCS), most notably in the
nucleus basalis Meynert (NBM) [1–4]. Consistent observations of
a correlation between the extent of cholinergic loss at autopsy and
antemortem dementia severity led to the “cholinergic hypothesis”
of AD [5] and form the rationale for the use of cholinomimetic
drugs as antidementive treatment in AD [6]. Moreover, cholinergic
system degeneration was found to be coupled to the accumulation
213
214 Michel J. Grothe et al.
of cortical amyloid pathology [7–10], particularly in the preclinical

stage of AD as indicated by cortical amyloid pathology in the
absence of cognitive impairments at time of death [11–13]. This
contrasts with the well-described degenerative changes in medial
temporal and neocortical brain regions, which typically follow
accumulation of amyloid pathology with a temporal delay and
demarcate the transition to clinical stages of the disease [14, 15].
Together these data point to a great potential of using MRI-based
in vivo measurements of BFCS degeneration as AD biomarker with
relevance to the preclinical, at risk of AD dementia, stage. However,
the complex anatomy of the BFCS and the indistinguishability of cho-
linergic cell clusters on high-resolution structural MRI scans place
important constraints on the in vivo structural analysis of this brain
region [16, 17]. A first approach to in vivo measurement of choliner-
gic degeneration in AD focused on manual measurement of a circum-
scribed part of the BFCS (usually referred to as “substantia innominata”
in this context) that can be consistently traced by the use of the ante-
rior commissure as external anatomic landmark [17]. This approach is
limited in that it only assesses a small and relatively unspecific subsec-
tion of the BFCS, which consists of several groups of distinct cholin-
ergic cell clusters that extend approximately 20 mm in anterior-posterior
direction [18–21] (Fig. 1). These subdivisions of the BFCS are char-
acterized by a differential cortical innervation preference [22, 23],
show differential associations with specific cognitive and behavioral
functions [24, 25], and vary in their vulnerability to AD-related neu-
rodegeneration [3]. Thus, although this imaging marker can readily
detect the severe BFCS degeneration present in AD dementia [17,
26–28], its sensitivity for subtle degenerative processes at prodromal
disease stages may be limited, and studies using this marker in early
disease stages have reported conflicting results [26, 29, 30].
As an extension to the manual volumetry approach, we have
developed automated techniques for BFCS morphometry that are
based on stereotactic mappings of the BFCS into MRI standard
space in combination with computational methods for image pro-
cessing and regional volume extraction [18, 21, 31–34]. While still
being indirect in vivo measures of cholinergic degeneration based
on volumetric changes on MRI, these techniques allow a much
more comprehensive assessment of total and subregional BFCS
changes in the course of AD. Over the last years, these techniques
have been used in several independent cohort studies by us and
other international research groups, providing novel insights into
the differential course of BFCS degeneration during aging and AD
pathogenesis and how these changes relate to the accumulation of
cortical amyloid pathology and the emergence of cognitive
impairments [35–44]. Thus, in these studies it could be shown
that although the BFCS is highly vulnerable to the normal aging
process, AD-associated degeneration of this region greatly exceeds
the age-related changes and can already be detected at prodromal
In Vivo Volumetry of the Cholinergic Basal Forebrain 215
Fig. 1 Overview of substantia innominata and cholinergic basal forebrain. (a) A coronal proton-weighted MRI scan at
the level of the anterior commissure illustrating the anatomic location of the substantia innominata and neighboring
structures (Adapted from [21] with permission to reprint granted by Oxford University Press). (b) Computer-assisted
3D reconstruction of the basal forebrain cholinergic nuclei based on histology of a postmortem brain. Cholinergic
nuclei stretch in anterior-posterior direction from the medial septum (Ch1) over the vertical and horizontal limb of the
diagonal band of Broca (Ch2 and Ch3) to the nucleus basalis (Ch4). ac anterior commissure, fo fornix
stages of the disease. Coinciding with h

istopathologic evidence for
a subregion-specific BFCS vulnerability to AD-related neurodege
neration [3], in vivo BFCS atrophy appears to manifest first in the
posterior subdivision of the NBM and spreads anteriorly to affect
Fig. 2 In vivo association between basal forebrain volume and cortical amyloid load. Scatter plots with linear
regression lines illustrate the association between cortical amyloid load measured in vivo by AV45-PET standard
uptake value ratio (SUVR) (x-axis) and cholinergic basal forebrain volume extracted using MRI-based volumetry
(y-axis) in cognitively normal (CN+; left) and early mild cognitive impairment (EMCI+; right) individuals with
detectable amyloid pathology. Figures in the bottom row show voxel-wise effects of a linear regression of corti-
cal AV45-PET SUVR on regional gray matter volume in the combined CN(+) and EMCI(+) subgroups. Effects are
superposed on coronal sections of the reference template and magnified to better depict the cholinergic basal
forebrain. Blue numbers indicate MNI space coordinates. Note the relative regional specificity of the amyloid
effect for the cholinergic basal forebrain nuclei. AC anterior commissure; AMG amygdala; DB diagonal band of
Broca; CA caudate; GP globus pallidus; (a, i, p) NBM, anterior, intermediate, posterior nucleus basalis Meynert;
PU putamen; TH thalamus (Figure adapted from [36] with permission to reprint granted by Elsevier)
the whole NBM and the rostral cholinergic nuclei as individuals

progress to clinically manifest AD. Importantly, by combining
BFCS volumetry with PET- or CSF-derived biomarkers of amyloid
pathology, the distinct association between cholinergic degenera-
tion and cortical amyloid deposition in presymptomatic and prede-
mentia stages of AD could be reproduced in vivo [36–38] (Fig. 2).
Accordingly, BFCS atrophy was found to be a significant predictor
of an underlying amyloid pathology among predemented individu-
als, whereas hippocampus atrophy was not, suggesting a distinct
potential of BFCS volumetry as an imaging biomarker for the char-
acterization of preclinical AD [35].
2 Materials
2.1 Stereotactic Generation of MRI-based stereotactic atlases of the BFCS has been
Maps of the BFCS hindered by the lack of imaging contrast for the distinct cholinergic
cell clusters on common high-resolution structural MRI scans.
However, their localization can be mapped indirectly to stereotactic
Fig. 3 Stereotactic mapping based on histology and postmortem MRI. The upper part of the figure shows the
postmortem processing with MRI scans of the brain in cranio (i.e., in situ) and after formalin fixation. Afterward
the subregions were identified by histological staining and manually transferred into the MRI of the dehydrated
brain in the alcohol space (color-coded mask at the lowest row of the figure). MRI scans were transferred first
into the MRI space of the in cranio scan and then into MNI standard space. The combined transformation
matrices were used to transfer the BFCS mask from alcohol space to MNI space via MRI in cranio space
(Reprinted from [34], with permission from IOS Press)
standard space using histologic delineation in autopsy data and

subsequent spatial normalization to stereotactic standard space
using a corresponding postmortem MRI scan (Fig. 3) [18, 21, 34].
In the following we will describe this process based on our latest
development of a stereotactic BFCS atlas [34].
2.1.1 Acquisition Stereotactic mapping of the BFCS nuclei was based on an autopsy
of Postmortem MRI brain specimen of a non-demented 56-year-old man who died from
and Histologic Preparation a myocardial infarction. Results from a standardized and validated
questionnaire interview with an informant of the brain donor were
reviewed by a behavioral neurologist and two geriatricians and con-
sidered to be without any evidence for cognitive decline or psychiat-
ric illnesses prior to death [45]. A cerebral MRI scan was performed
in situ 15 h after death. Afterward the brain was removed from the
skull and placed into formalin deposit for 3 months and subsequently
dehydrated by upscaling ethanol storage. Before and after the brain
dehydration procedure, further MRI scans were performed.
After dehydration, the brain was prepared for histological stain-
ing [46, 47]. BFCS subregions were identified and delineated on
digital pictures of the stained slices (400 μm) following the Mesulam
nomenclature [22, 48]. The basal forebrain refers to a brain region
whose central parts are formed by the anterior perforate substance in
humans. This region is included in the “unnamed medullary sub-
stance” (substantia innominata) described by Reil [49], which
extends parallel to the optic nerve on the basis of the forebrain. In
their comprehensive studies, Heimer et al. [50, 51] summarized that
a high number of fiber systems and nuclei, including the ventral
striatum with ncl. accumbens, ventral pallidum, and extended amyg-
dala, are crowded in the substantia innominata. The NBM is distin-
guished by predominantly big hyperchromic neurons. Individual
neurons can already be identified at low magnification as small blue
dots after staining with gallocyanin, a classical Nissl stain (Fig. 4).
The cholinergic nature of many of these big hyperchromic neurons
Fig. 4 (continued) level of the anterior compartment of the ncl. accumbens [52] and central parts of the parater-
minal gyrus. Bar 1 mm, valid for all figures of the plate. Left hemisphere, 58-year-old male, myocardial infarct. (b)
Plane of section 2 mm caudal to (a) passing ventral pallidum extending into caudal compartment of ncl. accum-
bens. (c) Plane of section 4 mm caudal to (a) comprising the caudal parts of the anterior perforate substance and
the bed nucleus of the stria terminalis. (d) Plane of section 11 mm caudal to (a) comprising the optic tract with
supraoptic nucleus. (e) Plane of sections 16 mm caudal to (a) comprising crura cerebri and central nucleus of the
amygdaloid complex. (f) Horizontal gallocyanin-stained section through the left hemisphere of a 60-year-old
female, myocardial infarct. Arrowheads = branches of the anterolateral central arteries (Aa. centrales anterolate-
rales, perforant arteries); white arrowhead in figure (c) = perforating branch of central artery subdividing Ch4am
from Ch4al; asterisks = artifacts caused by either vacuum embedding of tissue, leakage of the blood-brain bar-
rier, or condensed galactolipids during alcohol dehydration of brain tissue. a c anterior commissure, acc ncl.
accumbens, acj juxtacommissural cells of the anterior commissure [53, 54], a l ansa lenticularis, a p ansa pedun-
cularis, Ayala pp periputaminal part of Ayala’s nucleus, Ayala sp subputimale part of Ayala’s nucleus, B st bed
nucleus of the stria terminalis, caud caudate nucleus, Ce central nucleus of the amygdala, Ch1 medial septum,
Ch2 vertical limb of the diagonal band of Broca, Ch3 horizontal limb of the diagonal band of Broca, Ch4al antero-
lateral part of Ch4, Ch4am anteromedial part of Ch4, Ch4p posterior part of Ch4, cr crus cerebri, c t claustrum
temporale or endopiriform nucleus, d B diagonal band of Broca, g p globus pallidus, gpe external pallidum, gpi
internal pallidum, g pt paraterminal (subcallosal) gyrus, I C insula magna of Calleja, i ol olfactory islands, put puta-
men, s o supraoptic nucleus, s t stria terminalis, s th subthalamic nucleus, tr o optic tract, v p ventral pallidum
Fig. 4 Histology of the cholinergic basal forebrain nuclei. (a) Coronal gallocyanin (Nissl)-stained section at the
caused Mesulam to dub particular clusters and strands of NBM

neurons Ch4am, Ch4al, Ch4iv, Ch4id, and Ch4p. Similar hyper-
chromic/cholinergic neurons of the medial septum, the vertical, and
the horizontal limb of the diagonal band of Broca were designated
Ch1, Ch2, and Ch3, respectively [22, 48]. Ayala’s nucleus subpu-
taminalis and smaller hyperchromic neurons closely attached to cer-
tain regions of the anterior commissure (juxtacommissural) were not
included in Mesulam’s terminology [53–56]. Taken together, these
cholinergic neurons and clusters form the BFCS that arches from
the septum through Broca’s diagonal band and deep regions of the
anterior perforate substance (Fig. 1b). Topographically, anterior
commissure and BFCS show a more or less close relationship.
Therefore, the former can serve as a landmark for localization and
assessment of the rostrocaudal extent of the latter.
An interesting observation is that the different components of
the BFCS appear to be closely associated with fibers. Although
fibers remain unstained in Nissl sections, clear spaces in between
and overall arrangement of BFCS neurons reflect course, caliber,
and density of intranuclear fibers and fascicles. A few thin fibers are
associated with Ch1 neurons, whereas three to four fiber fascicles
appear to pierce Ch2 (Fig. 4a). A mixture of a few thick and numer-
ous thinner fibers guide neuronal arrangement in the most rostral
parts of Ch4am (Fig. 4b), whereas numerous small fibers and less
numerous small neurons are characteristic of Ch3. The cell-rich
compact Ch4al/am sector is pierced by thin fibers that apparently
enter or leave the cluster from its dorsomedial extreme (Fig. 4c).
The conspicuous fiber tract of the ansa peduncularis separates Ch4i
into a dorsal and ventral part (Fig. 4d). Horizontal sections through
the ansa peduncularis illustrate the conspicuous components of the
ansa peduncularis that connect thalamic, amygdaloid, and hypo-
thalamic regions and prove the close topographic relationship
between ansa lenticularis and Ch4i cell clusters (Fig. 4f). In addi-
tion to a close spatial relationship with fibers and tracts, NBM clus-
ters are arranged around perforating branches of the arteriae
striatales anterolaterales. They form conspicuous sleeves around
the latter (Fig. 4f). One more or less invariably penetrating artery
was described by Mesulam to separate Ch4al from Ch4am (Fig. 4c,
white arrowhead). In our experience, Ayala’s subputaminal nucleus
was always closely attached to a medium caliber perforating artery
that coursed in an oblique medio-lateral direction through the
basal forebrain (Fig. 4c, left arrowhead).
2.1.2 Mapping The BFCS subregions identified on the histological slices were
from Histologic Slices manually transferred from the digital pictures into the correspond-
to Stereotactic Standard ing MR slices of the dehydrated brain. Due to the small size of
Space Via Postmortem MRI some of the delineated cell clusters, clusters corresponding to Ch1
and Ch2 were merged into one subdivision (Ch1-2), and within the
NBM we only considered a distinction between an anterior-inter-
mediate (Ch4a-i) and a posterior subdivision (Ch4p). Transformation

of the delineations from the space of the dehydrated brain into the
space of the in situ brain scan was based on an initial 12-parameter
affine transformation followed by a high-dimensional nonlinear
registration between the two brain scans. The improved gray/white
matter contrast of the in situ brain scan enabled the use of the highly
accurate high-dimensional DARTEL (Diffeomorphic Anatomical
Registration Through Exponentiated Lie Algebra) registration
method [57, 58] for the final transformation from in situ space into
Montreal Neurological Institute (MNI) standard space.
The “in situ mask” described above is one of three currently
available stereotactic maps of the BFCS that have been developed
over the last years (Fig. 5). These maps primarily differ in the number
of analyzed postmortem brains, the delineation criteria and detail
used for histologic definition of the BFCS and its subdivisions, as well
as the methods and computational algorithms used for stereotactic
mapping into MNI standard space. A first map published in 2005
[21] was also based on a single (but separate) postmortem brain
specimen and used a delineation system based on Mesulam’s nomen-
clature, although the final map did not distinguish between distinct
subdivisions within the BFCS. Due to methodical limitations, the
histologic delineations were only performed within one brain hemi-
sphere and were mirrored to the contralateral hemisphere after ste-
reotactic normalization to MNI space based on a 12-parameter affine
transformation. Zaborszky and colleagues later refined the stereotac-
tic BFCS mapping by generating probabilistic maps of different
BFCS subdivisions based on the analysis of ten postmortem brains
[18]. Stereotactic mapping into MNI space was performed using a
low-dimensional nonlinear registration to the single-subject MNI
space template, which was applied directly to the postmortem MRI
of the fixed brain specimen. In contrast to the single-subject map-
pings, the probabilistic maps encode valuable information about the
interindividual variability of BFCS anatomy. However, due to inher-
ent limitations of the stereotactic mapping procedure, these probabil-
ities are also influenced by spatial registration inaccuracies in addition
to interindividual variability in regional anatomy. The BFCS atlas as
originally published includes probabilistic maps of different BFCS
subdivisions termed Ch1-2, Ch3, Ch4, and Ch4p, although the
delineation criteria may have slightly differed from those originally
proposed by Mesulam, particularly in the definition of Ch3 and its
distinction from Ch4am (see Sect. 2.1.1, Fig. 4). However, the maps
that have finally been made publicly available through the SPM
Anatomy toolbox [59] only include merged versions of Ch1-2/Ch3
and Ch4/Ch4p, and these also show some anatomic differences to
the original mappings [18] (Fig. 5 C + D).
2.2 Structural MRI Requirements for the structural MRI data used for automated
Data for Automated BFCS volumetry match the general recommendations for auto-
BFCS Morphometry mated morphometry approaches [60]. The structural MRI data
Fig. 5 Stereotactic maps of the cholinergic basal forebrain. Illustrations of the stereotactic maps of the cholin-
ergic basal forebrain in MNI space as published by [21] (a), [34] (b), and [18] (c). Panel (d) shows the stereo-
tactic maps from [18] that have finally been made publicly available through the SPM Anatomy toolbox [59].
Panel (c) is taken from the original publication in [18] (with permission to reprint granted by Elsevier), and
panels (a), (b), and (d) have been produced by projecting the stereotactic maps onto coronal sections with
identical MNI space coordinates as in (c) (running from y = 8 in the upper left corner to y = −7 in the lower
right corner). The probabilistic maps in (c) and (d) are represented as discrete maximum probability maps.
Note that these maps are designed to match the single-subject MNI space template, whereas maps in (a) and
(b) correspond to standard MNI space (MNI152 average template)
should be based on a high-resolution three-dimensional

T1-weighted imaging sequence with an approximate image resolu-
tion of 1 × 1 × 1 mm isotropic voxel size, such as provided by mag-
netization prepared (MP-RAGE) or spoiled gradient echo
(IR-SPGR) acquisitions. While acquisitions at 3 T magnetic field
strength are generally preferable due to their higher signal-to-noise
ratio and contrast-to-noise ratio between brain tissues, the method
also sensitively detects AD-related BFCS atrophy using MRI scans
acquired at 1.5 T [32, 33, 39].
3 Methods
3.1 Automated The availability of detailed stereotactic maps of the BFCS renders
Volumetry Approaches this brain region accessible to “atlas-based” automated volumetry
approaches. The overarching principle of these volumetry approaches
is to spatially transform (“normalize”) the individual native space
MRI scan into the standardized stereotactic space where the BFCS
atlas is defined in (typically MNI space) and use this transform to
calculate individual BFCS volumes. After matching the native space
scan to a stereotactic standard space template, the regional informa-
tion of an individual’s anatomy is encoded in the transformation
parameters that describe the deformations necessary for the nonlin-
ear matching (represented by vector fields, also called “deforma-
tion” fields). The Jacobian determinant map of these vector fields
specifically reflects the regional volumetric changes, i.e., the amount
of dilation or shrinkage that occurred at each voxel during matching
to the stereotactic standard space template.
In the most widely used “voxel-based morphometry” approach
to automated volumetry, the native space MRI scan is first seg-
mented into broad classes of tissue types (gray matter, white matte,
cerebrospinal fluid), and then the gray matter tissue of interest is
spatially transformed into stereotactic standard space, and voxel
values of the spatially transformed gray matter maps are finally
scaled (“modulated”) by the Jacobian determinant to reflect the
amount of regional gray matter tissue present before warping. The
sum of modulated gray matter voxel values within a given region
of interest (ROI) defined in template space (here the BFCS atlas)
represents the gray matter volume of this region in the individual
MRI scan (Fig. 6) [60].
In an alternative approach, sometimes referred to as “Jacobian
integration” or “tensor-based morphometry” [60, 61], the voxel
values are directly summed up on the Jacobian determinant maps
instead of the modulated gray matter maps. This approach produces
individual native space volumes of a given ROI independent of tissue
type and is equivalent to the volume calculation of a ROI mask that
has been transformed from template space to the individual’s native
space based on the inverse spatial transformation parameters [62].
Fig. 6 Automated analysis of cholinergic basal forebrain volume. Flowchart illustrating the main preprocessing and
analysis steps for automated volumetry of the cholinergic basal forebrain. For details, see descriptions in the text
3.2 Preprocessing According to the automated volumetry pipeline outlined above,

of Structural MRI Scan preprocessing of the structural MRI scans principally consists of
tissue-type segmentation and spatial normalization to the stereo-
tactic template space where the BFCS mask is defined in. Given the
relatively small spatial extent of the BFCS and its subdivisions, the
whole approach critically depends on the accuracy of the spatial
normalization procedure, aiming for a perfect fit of the native space
scan to the anatomy of the normalization template. Thus, best
results can be achieved using new-generation high-dimensional
nonlinear spatial registration algorithms, such as DARTEL [57] or

SyN [63], which have been demonstrated to outperform more tra-
ditional (and still more commonly used) affine or low-dimensional
nonlinear registration algorithms [58].
In addition to the choice of registration algorithm, accuracy of
the spatial normalization procedure may also depend on the target
template used for spatial normalization. While the most commonly
used templates for spatial normalization are MNI standard space
templates, these may not be the optimal targets for spatial normal-
ization of scans derived from individuals of advanced age and with
neurodegenerative disease. Given that the neuroanatomy of these
individuals may markedly deviate from the characteristics of MNI
space templates reflecting a healthy adult population, larger defor-
mations are necessary in the registration process, thereby increasing
the propensity for registration errors. Moreover, differences in the
amount of anatomic deviation from the MNI space templates
between patients and healthy age-matched controls may result in a
systematic group bias in registration accuracy [64, 65]. Thus, the
accuracy and homogeneity of the spatial normalization may be
improved for these populations by using group- or population-spe-
cific normalization templates [33, 39, 65]. However, this also comes
with the drawback of having to spatially transform the stereotactic
BFCS masks that are originally designed to match MNI standard
space, and the optimal trade-off between possible improvements in
inter-subject registration accuracy and potential anatomic misregis-
tration errors during spatial transformation of the BFCS ROI is not
clear a priori and likely depends on the studied population.
The routine used for automated tissue-type segmentation may
also have a marked influence on the automated assessment of BFCS
volume. Automated tissue-type segmentation routines of the most
widely used software packages are usually optimized for detecting
gray/white matter contrasts in cortical brain regions, whereas the
cholinergic nuclei of the BFCS are scattered throughout an anatomi-
cally complex subcortical brain region consisting of several other neu-
ronal populations as well as traversing white matter fiber bundles.
Moreover, some of the theoretical assumptions about tissue-type dis-
tribution and corresponding intensity contrasts markedly differ
between available segmentation routines, resulting in notable differ-
ences in segmentation outcomes [64, 66]. Thus, depending on the
routine used for tissue-type segmentation, more or less parts of the
BFCS will actually be segmented as gray matter. This potentially lim-
its the applicability of a standard “voxel-based morphometry”
approach to BFCS volumetry and renders “Jacobian integration/
tensor-based morphometry” a suitable, tissue-independent, alterna-
tive in this context [31, 32, 37]. However, at least the tissue segmen-
tation routine implemented in SPM’s VBM8 toolbox [67] typically
segments all or most parts of the BFCS map as gray matter, and in
previous work, we found that these BFCS gray matter volume mea-
surements appeared to be more sensitive for AD-related changes than

regional volumes extracted through Jacobian integration [33].
In summary, for most settings we now generally recommend a
processing pipeline for automated BFCS volumetry that is based on
a “voxel-based morphometry” approach as outlined above, in com-
bination with VBM8-type tissue segmentation [67] and high-dimen-
sional spatial normalization using SPM’s DARTEL algorithm [57].
3.3 Analysis Following the automated volumetry approach described above,

of Cholinergic Basal individual volumes of the BFCS or its subdivisions can be derived
Forebrain Volumes by summing up Jacobian determinant voxel values or modulated
gray matter voxel values within respective ROI masks derived from
stereotactic BFCS maps (see Sect. 2.1). In order for these measures
to reflect the extent of atrophic change rather than interindividual
differences in head size, the individual volumes further need to be
scaled to the total intracranial volume before subjecting them to
statistical analysis [62, 68]. Alternatively, interindividual differ-
ences in head size may be accounted for directly within the statisti-
cal model that is used for assessing the diagnostic or prognostic
potential of BFCS volume as imaging biomarker for AD [68].
Detailed neuropathologic studies have pointed to a differential
vulnerability of BFCS subregions to AD-related neurodegeneration,
being markedly more pronounced in the NBM and particularly its
posterior subdivision compared to anterior medial nuclei [3]. While
assessment of subregional BFCS volumes principally allows capturing
such subregional differences using the outlined in vivo volumetry
approach, inference will be limited to the idiosyncrasies of the delin-
eation system and detail used for defining BFCS subdivisions in the
available stereotactic maps [18, 21, 34, 69]. A complementary and
spatially unbiased approach for determining subregional specificity
for AD-related neurodegeneration within the BFCS is to assess volu-
metric differences on the voxel level. Thus, instead of extracting and
comparing individual volumes of predefined BFCS subdivisions,
diagnostic group differences may be assessed for every voxel within a
wider BFCS ROI using mass univariate statistical testing (Fig. 6).
Although this approach does not directly produce individual values
that could be used as an imaging biomarker, the resulting voxel-wise
map of AD-specific BFCS atrophy may be used in independent
cohorts for extracting subregional BFCS volume with a potentially
higher sensitivity for capturing AD-related changes [37].
4 Conclusions
The outlined method for automated MRI-based BFCS volumetry pro-

vides a means to study BFCS degeneration in the course of aging and
AD in vivo. Studies using this method could provide important in vivo
evidence for long-standing neuropathologic observations of a correla-
tion between the degree of BFCS degeneration and the severity of

antemortem cognitive impairments in AD as well as in other neuro-
logic dementing diseases, including Lewy body disorders and primary
progressive aphasias [20, 32, 40, 70–76]. In contrast to autopsy stud-
ies, which are usually confined to relatively advanced disease stages and
limited sample sizes, MRI-based BFCS volumetry may be particularly
useful for studying early disease stages and their distinction from the
normal aging process [33, 34, 38, 39, 77]. Most interestingly, using
combined in vivo BFCS volumetry and PET- or CSF-based biomarker
assessments of amyloid pathology in large-scale cohort studies could
consistently reproduce the initial neuropathological evidence for a dis-
tinct association between BFCS degeneration and cortical amyloid
pathology in the preclinical phase of AD [11–13, 35–38] (Fig. 2). By
contrast, established MRI-based measures of medial temporal lobe
atrophy did not show similar associations with amyloid pathology at
this early disease stage, indicating that BFCS volumetry may be a par-
ticularly useful MRI-based biomarker for characterizing preclinical
AD. Of course, in the existing cross-sectional autopsy and neuroimag-
ing studies, the direction of the amyloid-BFCS interaction remains
unknown. Preclinical studies in cultured cells as well as in experimental
animal models could provide evidence both for a pronounced neuro-
toxic effect of amyloid pathology on cholinergic basal forebrain neu-
rons [78–81], as well as for an acceleration of cortical amyloid
accumulation due to cholinergic denervation [82–84], possibly caused
by dysfunctional clearance mechanisms [85]. Assessment of the rela-
tionships between amyloid pathology and BFCS degeneration through
in vivo neuroimaging and biomarker data provides the unique possibil-
ity to study the temporal relation between these pathologic hallmarks
of AD longitudinally and may thus provide entirely novel insights into
the role of these interactions during natural disease progression.
An important limitation of MRI-based in vivo volumetry of the
BFCS is that the volumetric measurement is only an indirect
marker of cholinergic system integrity. Although the use of histol-
ogy-based stereotactic atlases of the BFCS increases the corre-
spondence of the analyzed brain region with the localization of
cholinergic cell clusters in the basal forebrain, one cannot exclude
that the volumetric measurement may also reflect change in other
non-cholinergic neuronal populations. For example, cholinergic
neurons in the basal forebrain are intermingled, and closely inter-
act, with cortically projecting GABAergic neurons, which are also
increasingly being recognized for their critical role in basal fore-
brain-mediated modulation of cortical activity and cognition
[86]. In addition, the arrangement of the different BFCS compo-
nents is closely associated with traversing fiber populations as well
as with perforating arteries (see Sect 2.1.1). Thus, age- or disease-
related decreases in fiber number and diameter as well as increases
in vessel diameter could have an impact on in vivo BFCS volum-
etry. Hence, while the consistently observed associations with
amyloid pathology and emerging cognitive deficits underline the

pathologic and clinical relevance of MRI-based BFCS volumetry
in AD, the implications of these volumetric alterations for cholin-
ergic neurotransmission need to be studied in more detail.
Pharmacologic challenge paradigms [87, 88] and multimodal
imaging studies employing PET-based imaging tracers of cholin-
ergic neurotransmission [89–91] may be suitable analytic
approaches to further investigate the neurotransmitter specificity
of volumetric BFCS changes on structural MRI.
With respect to the potential of MRI-based BFCS volumetry as
an imaging biomarker for preclinical AD, one has to keep in mind that
the outcome of the volumetry approach is sensitive to several method-
ological choices and there is currently no consensus on the optimal
analytic pipeline for this rather experimental technique. Variability
may arise through differences in the spatial accuracy and regional
detail of the stereotactic atlases used to map the anatomic location of
the BFCS, as well as differences in the automated algorithms used for
tissue-type segmentation and spatial normalization of the MRI images.
In addition to the diagnostic or prognostic potential under experi-
mental conditions, a widespread clinical use of a biomarker will ulti-
mately also depend on the harmonization and standardization of the
respective measurement protocol, such as is currently being done on
an international basis for hippocampus volumetry as the best estab-
lished MRI-based imaging biomarker for AD so far [92, 93].
References
1. Whitehouse PJ, Price DL, Struble RG et al senile plaques and mental test scores in senile
(1982) Alzheimer's disease and senile demen- dementia. Br Med J 2(6150):1457–1459
tia: loss of neurons in the basal forebrain. 8. Ikonomovic MD, Klunk WE, Abrahamson EE
Science 215(4537):1237–1239 et al (2011) Precuneus amyloid burden is associ-
2. McGeer PL, McGeer EG, Suzuki J et al (1984) ated with reduced cholinergic activity in Alzheimer
Aging, Alzheimer's disease, and the choliner- disease. Neurology 77(1):39–47. https://doi.
gic system of the basal forebrain. Neurology org/10.1212/WNL.0b013e3182231419
34(6):741–745 9. Beach TG, McGeer EG (1992) Senile plaques,
3. Vogels OJ, Broere CA, ter Laak HJ et al (1990) amyloid beta-protein, and acetylcholinesterase
Cell loss and shrinkage in the nucleus basa- fibres: laminar distributions in Alzheimer's disease
lis Meynert complex in Alzheimer's disease. striate cortex. Acta Neuropathol 83(3):292–299
Neurobiol Aging 11(1):3–13 10. Arendt T, Bigl V, Tennstedt A et al (1985)
4. Perry EK, Blessed G, Tomlinson BE et al Neuronal loss in different parts of the nucleus
(1981) Neurochemical activities in human basalis is related to neuritic plaque formation
temporal lobe related to aging and Alzheimer- in cortical target areas in Alzheimer's disease.
type changes. Neurobiol Aging 2(4):251–256 Neuroscience 14(1):1–14
5. Bartus RT, Dean RL 3rd, Beer B et al (1982) 11. Beach TG, Honer WG, Hughes LH (1997)
The cholinergic hypothesis of geriatric mem- Cholinergic fibre loss associated with diffuse
ory dysfunction. Science 217(4558):408–414 plaques in the non-demented elderly: the
6. Birks J (2006) Cholinesterase inhibitors for preclinical stage of Alzheimer's disease? Acta
Alzheimer's disease. Cochrane Database Neuropathol 93(2):146–153
Syst Rev 1:CD005593. https://doi. 12. Potter PE, Rauschkolb PK, Pandya Y et al (2011)
org/10.1002/14651858.cd005593 Pre- and post-synaptic cortical cholinergic defi-
7. Perry EK, Tomlinson BE, Blessed G et al (1978) cits are proportional to amyloid plaque presence
Correlation of cholinergic abnormalities with and density at preclinical stages of Alzheimer's
disease. Acta Neuropathol 122(1):49–60. 24. Muir JL (1997) Acetylcholine, aging, and
https://doi.org/10.1007/s00401-011-0831-1 Alzheimer’s disease. Pharmacol Biochem
13. Beach TG, Kuo YM, Spiegel K et al (2000) The Behav 56(4):687–696
cholinergic deficit coincides with Abeta depo- 25. Pappas BA, Bayley PJ, Bui BK et al (2000)
sition at the earliest histopathologic stages of Choline acetyltransferase activity and cogni-
Alzheimer disease. J Neuropathol Exp Neurol tive domain scores of Alzheimer's patients.
59(4):308–313 Neurobiol Aging 21(1):11–17
14. Young AL, Oxtoby NP, Daga P et al (2014) A 26. George S, Mufson EJ, Leurgans S et al
data-driven model of biomarker changes in spo- (2011) MRI-based volumetric measure-
radic Alzheimer's disease. Brain 137(Pt 9):2564– ment of the substantia innominata in amnes-
2577. https://doi.org/10.1093/brain/awu176 tic MCI and mild AD. Neurobiol Aging
15. Jack CR Jr, Knopman DS, Jagust WJ et al 32(10):1756–1764. https://doi.org/10.1016/
(2013) Tracking pathophysiological processes j.neurobiolaging.2009.11.006
in Alzheimer's disease: an updated hypothetical 27. Hanyu H, Tanaka Y, Shimizu S et al (2005)
model of dynamic biomarkers. Lancet Neurol Differences in MR features of the substan-
12(2):207–216. https://doi.org/10.1016/ tia innominata between dementia with Lewy
s1474-4422(12)70291-0 bodies and Alzheimer's disease. J Neurol
16. Huckman MS (1995) Where's the chicken? 252(4):482–484. https://doi.org/10.1007/
AJNR Am J Neuroradiol 16(10):2008–2009 s00415-005-0611-8
17. Hanyu H, Asano T, Sakurai H et al (2002) MR 28. Gao FQ, Pettersen JA, Bocti C et al (2013) Is
analysis of the substantia innominata in normal encroachment of the carotid termination into the
aging, Alzheimer disease, and other types of substantia innominata associated with its atrophy
dementia. AJNR Am J Neuroradiol 23(1):27–32 and cognition in Alzheimer's disease? Neurobiol
18. Zaborszky L, Hoemke L, Mohlberg H et al Aging 34(7):1807–1814. https://doi.
(2008) Stereotaxic probabilistic maps of the org/10.1016/j.neurobiolaging.2013.01.009
magnocellular cell groups in human basal fore- 29. Muth K, Schonmeyer R, Matura S et al (2010)
brain. NeuroImage 42(3):1127–1141. https:// Mild cognitive impairment in the elderly is
doi.org/10.1016/j.neuroimage.2008.05.055 associated with volume loss of the cholin-
19. Halliday GM, Cullen K, Cairns MJ (1993) ergic basal forebrain region. Biol Psychiatry
Quantitation and three-dimensional recon- 67(6):588–591. https://doi.org/10.1016/j.
struction of Ch4 nucleus in the human basal biopsych.2009.02.026
forebrain. Synapse 15(1):1–16. https://doi. 30. Moon WJ, Kim HJ, Roh HG et al (2008)
org/10.1002/syn.890150102 Atrophy measurement of the anterior com-
20. Teipel SJ, Flatz W, Ackl N et al (2014) Brain atro- missure and substantia innominata with 3T
phy in primary progressive aphasia involves the high-resolution MR imaging: does the mea-
cholinergic basal forebrain and Ayala's nucleus. surement differ for patients with fronto-
Psychiatry Res 221(3):187–194. https://doi. temporal lobar degeneration and Alzheimer
org/10.1016/j.pscychresns.2013.10.003 disease and for healthy subjects? AJNR Am
J Neuroradiol 29(7):1308–1313. https://doi.
21. Teipel SJ, Flatz WH, Heinsen H et al (2005) org/10.3174/ajnr.A1103
Measurement of basal forebrain atrophy in
Alzheimer's disease using MRI. Brain 128(Pt 31. Teipel SJ, Meindl T, Grinberg L et al (2011)
11):2626–2644. https://doi.org/10.1093/ The cholinergic system in mild cogni-
brain/awh589 tive impairment and Alzheimer's disease:
an in vivo MRI and DTI study. Hum Brain
22. Mesulam MM, Mufson EJ, Levey AI et al Mapp 32(9):1349–1362. https://doi.
(1983) Cholinergic innervation of cortex by org/10.1002/hbm.21111
the basal forebrain: cytochemistry and cortical
connections of the septal area, diagonal band 32. Grothe M, Zaborszky L, Atienza M et al (2010)
nuclei, nucleus basalis (substantia innomi- Reduction of basal forebrain cholinergic sys-
nata), and hypothalamus in the rhesus monkey. tem parallels cognitive impairment in patients
J Comp Neurol 214(2):170–197. https://doi. at high risk of developing Alzheimer's disease.
org/10.1002/cne.902140206 Cereb Cortex 20(7):1685–1695. https://doi.
org/10.1093/cercor/bhp232
23. Zaborszky L, Csordas A, Mosca K et al (2013)
Neurons in the basal forebrain project to the cor- 33. Grothe M, Heinsen H, Teipel SJ (2012)
tex in a complex topographic organization that Atrophy of the cholinergic basal forebrain
reflects Corticocortical connectivity patterns: an over the adult age range and in early stages
experimental study based on retrograde tracing of Alzheimer’s disease. Biol Psychiatry
and 3D reconstruction. Cereb Cortex. https:// 71(9):805–813. https://doi.org/10.1016/j.
doi.org/10.1093/cercor/bht210 biopsych.2011.06.019
34. Kilimann I, Grothe M, Heinsen H et al 45. Grinberg LT, Ferretti RE, Farfel JM et al (2007)
(2014) Subregional basal forebrain atrophy Brain bank of the Brazilian aging brain study
in Alzheimer's disease: a multicenter study. group - a milestone reached and more than 1,600
J Alzheimers Dis 40(3):687–700. https://doi. collected brains. Cell Tissue Bank 8(2):151–162.
org/10.3233/jad-132345 https://doi.org/10.1007/s10561-006-9022-z
35. Teipel S, Heinsen H, Amaro E Jr et al (2014) 46. Grinberg LT, Amaro E Jr, Teipel S et al (2008)
Cholinergic basal forebrain atrophy pre- Assessment of factors that confound MRI and
dicts amyloid burden in Alzheimer's disease. neuropathological correlation of human postmor-
Neurobiol Aging 35(3):482–491. https://doi. tem brain tissue. Cell Tissue Bank 9(3):195–203.
org/10.1016/j.neurobiolaging.2013.09.029 https://doi.org/10.1007/s10561-008-9080-5
36. Grothe MJ, Ewers M, Krause B et al (2014) 47. Grinberg LT, Amaro Junior E, da Silva AV et al
Basal forebrain atrophy and cortical amyloid (2009) Improved detection of incipient vascular
deposition in nondemented elderly subjects. changes by a biotechnological platform combin-
Alzheimers Dement 10(5 Suppl):S344–S353. ing post mortem MRI in situ with neuropathol-
https://doi.org/10.1016/j.jalz.2013.09.011 ogy. J Neurol Sci 283(1–2):2–8. https://doi.
37. Kerbler GM, Fripp J, Rowe CC et al (2015) org/10.1016/j.jns.2009.02.327
Basal forebrain atrophy correlates with amy- 48. Mesulam MM, Mufson EJ, Wainer BH et al
loid beta burden in Alzheimer's disease. (1983) Central cholinergic pathways in the rat: an
Neuroimage Clin 7:105–113. https://doi. overview based on an alternative nomenclature
org/10.1016/j.nicl.2014.11.015 (Ch1-Ch6). Neuroscience 10(4):1185–1201
38. Schmitz TW, Nathan Spreng R (2016) Basal 49. Reil JC (1809) Untersuchungen über den Bau
forebrain degeneration precedes and predicts des grossen Gehirns im Mensch. Arch Physiol
the cortical spread of Alzheimer's pathol- (Halle) 9:136–146
ogy. Nat Commun 7:13249. https://doi. 50. Alheid GF, Heimer L (1988) New perspectives in
org/10.1038/ncomms13249 basal forebrain organization of special relevance
39. Grothe M, Heinsen H, Teipel S (2013) for neuropsychiatric disorders: the striatopallidal,
Longitudinal measures of cholinergic fore- amygdaloid, and corticopetal components of
brain atrophy in the transition from healthy substantia innominata. Neuroscience 27:1–39
aging to Alzheimer's disease. Neurobiol Aging 51. Heimer L, De Olmos J, Alheid GF et al (1991)
34(4):1210–1220. https://doi.org/10.1016/j. “perestroika” in the basal forebrain: opening
neurobiolaging.2012.10.018 the border between neurology and psychiatry.
40. Grothe MJ, Heinsen H, Amaro E Jr et al (2016) Prog Brain Res 87:109–165
Cognitive correlates of basal forebrain atro- 52. Lauer M, Heinsen H (1996) Cytoarchitectonics
phy and associated cortical Hypometabolism of the human nucleus accumbens. Journal of.
in mild cognitive impairment. Cereb Cortex Brain Res 37:243–254
26(6):2411–2426. https://doi.org/10.1093/ 53. Grinberg LT, Heinsen H (2007) Computer-
cercor/bhv062 assisted 3D reconstruction of the human basal
41. Ray NJ, Metzler-Baddeley C, Khondoker MR forebrain. Dementia & Neuropsychologia
et al (2015) Cholinergic basal forebrain struc- 2:140–146
ture influences the reconfiguration of white 54. Heinsen H, Hampel H, Teipel SJ (2006)
matter connections to support residual mem- Nucleus subputaminalis: neglected part of the
ory in mild cognitive impairment. J Neurosci basal nucleus of Meynert - response to Boban
35(2):739–747. https://doi.org/10.1523/ et al.: computer-assisted 3D reconstruction
jneurosci.3617-14.2015 of the nucleus basalis complex, including the
42. Cantero JL, Zaborszky L, Atienza M (2016) nucleus subputaminalis (Ayala's nucleus). Brain
Volume loss of the nucleus Basalis of Meynert 129(4):U1–U4
is associated with atrophy of innervated regions 55. Ayala G (1915) A hitherto undifferentiated
in mild cognitive impairment. Cereb Cortex. nucleus in the basal forebrain (nucleus subpu-
https://doi.org/10.1093/cercor/bhw195 taminalis). Brain 37:433–438
43. Grothe MJ, Scheef L, Bauml J et al (2016) 56. Simic G, Mrzljak L, Fucic A et al (1999)
Reduced cholinergic basal forebrain integrity links Nucleus subputaminalis (Ayala): the still disre-
neonatal complications and adult cognitive deficits garded magnocellular component of the basal
after premature birth. Biol Psychiatry. https:// forebrain may be human specific and connected
doi.org/10.1016/j.biopsych.2016.12.008 with the cortical speech area. Neuroscience
44. Kilimann I, Hausner L, Fellgiebel A et al 89(1):73–89
(2017) Parallel atrophy of cortex and basal 57. Ashburner J (2007) A fast diffeomorphic
forebrain cholinergic system in mild cognitive image registration algorithm. NeuroImage
impairment. Cereb Cortex 27(3):1841–1848. 38(1):95–113. https://doi.org/10.1016/j.
https://doi.org/10.1093/cercor/bhw019 neuroimage.2007.07.007
58. Klein A, Andersson J, Ardekani BA et al (2009) disease. Front Aging Neurosci 6:264. https://
Evaluation of 14 nonlinear deformation algo- doi.org/10.3389/fnagi.2014.00264
rithms applied to human brain MRI registra- 69. Boban M, Kostovic I, Simic G (2006) Nucleus
tion. NeuroImage 46(3):786–802. https:// subputaminalis: neglected part of the basal nucleus
doi.org/10.1016/j.neuroimage.2008.12.037 of Meynert. Brain 129(Pt 4):E42.; author reply

59. Eickhoff SB, Stephan KE, Mohlberg H E43. https://doi.org/10.1093/brain/awl025
et al (2005) A new SPM toolbox for 70. Schliebs R, Arendt T (2006) The significance
combining probabilistic cytoarchitec- of the cholinergic system in the brain during
tonic maps and functional imaging data. aging and in Alzheimer’s disease. J Neural
NeuroImage 25(4):1325–1335. https://doi. Transm 113(11):1625–1644. https://doi.
org/10.1016/j.neuroimage.2004.12.034 org/10.1007/s00702-006-0579-2
60. Ashburner J (2009) Computational anatomy 71. Rogers JD, Brogan D, Mirra SS (1985) The
with the SPM software. Magn Reson Imaging nucleus basalis of Meynert in neurological disease:
27(8):1163–1174. https://doi.org/10.1016/j. a quantitative morphological study. Ann Neurol
mri.2009.01.006 17(2):163–170. https://doi.org/10.1002/
61. Boyes RG, Rueckert D, Aljabar P et al (2006) ana.410170210
Cerebral atrophy measurements using Jacobian 72. Samuel W, Terry RD, DeTeresa R et al (1994)
integration: comparison with the boundary shift Clinical correlates of cortical and nucleus basa-
integral. NeuroImage 32(1):159–169. https:// lis pathology in Alzheimer dementia. Arch
doi.org/10.1016/j.neuroimage.2006.02.052 Neurol 51(8):772–778
62. Malone IB, Leung KK, Clegg S et al (2015) 73. Iraizoz I, Guijarro JL, Gonzalo LM et al
Accurate automatic estimation of total intra- (1999) Neuropathological changes in the
cranial volume: a nuisance variable with less nucleus basalis correlate with clinical measures
nuisance. NeuroImage 104:366–372. https:// of dementia. Acta Neuropathol 98(2):186–196
doi.org/10.1016/j.neuroimage.2014.09.034 74. Grothe MJ, Schuster C, Bauer F et al (2014)
63. Avants BB, Epstein CL, Grossman M et al Atrophy of the cholinergic basal forebrain in
(2008) Symmetric diffeomorphic image regis- dementia with Lewy bodies and Alzheimer's dis-
tration with cross-correlation: evaluating auto- ease dementia. J Neurol 261(10):1939–1948.
mated labeling of elderly and neurodegenerative https://doi.org/10.1007/s00415-014-7439-z
brain. Med Image Anal 12(1):26–41. https:// 75. Arendt T, Bigl V, Arendt A et al (1983) Loss
doi.org/10.1016/j.media.2007.06.004 of neurons in the nucleus basalis of Meynert
64. Callaert DV, Ribbens A, Maes F et al (2014) in Alzheimer's disease, paralysis agitans
Assessing age-related gray matter decline with and Korsakoff's disease. Acta Neuropathol
voxel-based morphometry depends signifi- 61(2):101–108
cantly on segmentation and normalization pro- 76. Teipel S, Raiser T, Riedl L et al (2016)
cedures. Front Aging Neurosci 6:124. https:// Atrophy and structural covariance of the cho-
doi.org/10.3389/fnagi.2014.00124 linergic basal forebrain in primary progressive
65. Shen Q, Zhao W, Loewenstein DA et al (2012) aphasia. Cortex 83:124–135. https://doi.
Comparing new templates and atlas-based org/10.1016/j.cortex.2016.07.004
segmentations in the volumetric analysis of 77. Wolf D, Grothe M, Fischer FU et al (2014)
brain magnetic resonance images for diagnos- Association of basal forebrain volumes and
ing Alzheimer's disease. Alzheimers Dement cognition in normal aging. Neuropsychologia
8(5):399–406. https://doi.org/10.1016/ 53:54–63. https://doi.org/10.1016/j.
j.jalz.2011.07.002 neuropsychologia.2013.11.002
66. Eggert LD, Sommer J, Jansen A et al (2012) 78. Olesen OF, Dago L, Mikkelsen JD (1998)
Accuracy and reliability of automated gray mat- Amyloid beta neurotoxicity in the cholinergic
ter segmentation pathways on real and simulated but not in the serotonergic phenotype of RN46A
structural magnetic resonance images of the cells. Brain Res Mol Brain Res 57(2):266–274
human brain. PLoS One 7(9):e45081. https://
doi.org/10.1371/journal.pone.0045081 79. Zheng WH, Bastianetto S, Mennicken F et al
(2002) Amyloid beta peptide induces tau phos-
67. Gaser C (2009) Partial volume segmentation phorylation and loss of cholinergic neurons
with adaptive maximum a posteriori (MAP) in rat primary septal cultures. Neuroscience
approach. NeuroImage 47:S121. https://doi. 115(1):201–211
org/10.1016/s1053-8119(09)71151-6
80. Boncristiano S, Calhoun ME, Kelly PH et al
68. Voevodskaya O, Simmons A, Nordenskjold R (2002) Cholinergic changes in the APP23 trans-
et al (2014) The effects of intracranial volume genic mouse model of cerebral amyloidosis.
adjustment approaches on multiple regional J Neurosci 22(8):3234–3243. doi:20026314
MRI volumes in healthy aging and Alzheimer's
81. Bell KF, Ducatenzeiler A, Ribeiro-da-Silva Aging 36(10):2709–2715. https://doi.

A et al (2006) The amyloid pathology pro- org/10.1016/j.neurobiolaging.2015.07.009
gresses in a neurotransmitter-specific man- 88. Teipel SJ, Bruno D, Grothe MJ et al (2015)
ner. Neurobiol Aging 27(11):1644–1657. Hippocampus and basal forebrain vol-
https://doi.org/10.1016/j.neurobiolaging. umes modulate effects of anticholiner-
2005.09.034 gic treatment on delayed recall in healthy
82. Nitsch RM, Slack BE, Wurtman RJ et al older adults. Alzheimers Dement (Amst)
(1992) Release of Alzheimer amyloid pre- 1(2):216–219. https://doi.org/10.1016/j.
cursor derivatives stimulated by activation of dadm.2015.01.007
muscarinic acetylcholine receptors. Science 89. Kendziorra K, Wolf H, Meyer PM et al (2011)
258(5080):304–307 Decreased cerebral alpha4beta2* nicotinic ace-
83. Ramos-Rodriguez JJ, Pacheco-Herrero tylcholine receptor availability in patients with
M, Thyssen D et al (2013) Rapid beta- mild cognitive impairment and Alzheimer's dis-
amyloid deposition and cognitive impair- ease assessed with positron emission tomography.
ment after cholinergic denervation in APP/ Eur J Nucl Med Mol Imaging 38(3):515–525.
PS1 mice. J Neuropathol Exp Neurol https://doi.org/10.1007/s00259-010-1644-5
72(4):272–285. https://doi.org/10.1097/ 90. Bohnen NI, Muller ML, Kuwabara H
NEN.0b013e318288a8dd et al (2009) Age-associated leukoara-
84. Kolisnyk B, Al-Onaizi M, Soreq L et al (2016) iosis and cortical cholinergic deafferentation.
Cholinergic surveillance over hippocampal Neurology 72(16):1411–1416. https://doi.
RNA metabolism and Alzheimer's-like pathol- org/10.1212/WNL.0b013e3181a187c6
ogy. Cereb Cortex. https://doi.org/10.1093/ 91. Bohnen NI, Kaufer DI, Hendrickson R et al
cercor/bhw177 (2005) Cognitive correlates of alterations in
85. Ovsepian SV, Herms J (2013) Drain of the acetylcholinesterase in Alzheimer's disease.
brain: low-affinity p75 neurotrophin receptor Neurosci Lett 380(1–2):127–132. https://
affords a molecular sink for clearance of corti- doi.org/10.1016/j.neulet.2005.01.031
cal amyloid beta by the cholinergic modulator 92. Wolf D, Bocchetta M, Preboske GM et al
system. Neurobiol Aging 34(11):2517–2524. (2017) Reference standard space hippocam-
https://doi.org/10.1016/j.neurobiolaging. pus labels according to the EADC-ADNI
2013.05.005 harmonized protocol: utility in automated
86. Anaclet C, Pedersen NP, Ferrari LL et al (2015) volumetry. Alzheimers Dement. https://doi.
Basal forebrain control of wakefulness and cor- org/10.1016/j.jalz.2017.01.009
tical rhythms. Nat Commun 6:8744. https:// 93. Frisoni GB, Jack CR Jr, Bocchetta M et al
doi.org/10.1038/ncomms9744 (2015) The EADC-ADNI harmonized pro-
87. Lim YY, Maruff P, Schindler R et al (2015) tocol for manual hippocampal segmentation
Disruption of cholinergic neurotransmission on magnetic resonance: evidence of validity.
exacerbates Abeta-related cognitive impairment Alzheimers Dement 11(2):111–125. https://
in preclinical Alzheimer's disease. Neurobiol doi.org/10.1016/j.jalz.2014.05.1756
Part V
Patient Benefit and Ethical Considerations

Chapter 16
Clinical Meaningfulness of Biomarker Endpoints

in Alzheimer’s Disease Research
Kok Pin Ng, Tharick A. Pascoal, Xiaofeng Li, Pedro Rosa-Neto,
and Serge Gauthier
Abstract
Advancement in biomarker research has enabled the in vivo detection of Alzheimer’s disease (AD) patho-
physiology, including amyloid plaques, neurofibrillary tangles, and neuronal degeneration. AD biomarkers
play an important role in characterizing the trajectory of AD and have been incorporated in the research
criteria for AD diagnosis. The presence of abnormal biomarkers for AD pathology in cognitively normal
individuals has further led to the proposal of a preclinical stage in the AD spectrum. With the emerging
conceptual framework that intervention in the early stages of the disease offers the greatest chance of suc-
cess in preventing or delaying AD progression, recent clinical trials are now focusing on individuals with
preclinical AD. While there are clear benefits from the use of biomarkers in research settings such as the
enrichment of clinical trial population to confirm the presence of target brain pathology and target engage-
ment by the intervention, the role of biomarkers in the clinical setting is less clear, especially in asymptom-
atic individuals. Potential ethical issues also arise with the use of biomarkers due to the conflict between
the principles of benefits and not doing harm. In fact, a unique set of ethical issues arises in asymptomatic
individuals, such as the disclosure of genetic mutation status, and abnormal biomarker results when their
diagnostic validity is uncertain. In this chapter, we will discuss the issues and clinical meaningfulness of
biomarkers in AD research. Specifically, we will focus on the potential benefits and ethical considerations
when genetics and biomarkers for amyloid, tau, and neurodegeneration are used in the early stages of AD.
Key words Alzheimer’s disease, Dementia, Mild cognitive impairment, Biomarker, Cerebrospinal
fluid, Imaging, Tau, Amyloid-beta, Ethics, Genetics, Early diagnosis, Prognosis
1 Introduction
Amyloid plaques and neurofibrillary tangles are the core histopath-

ological features of Alzheimer’s disease (AD) [1]. Advancement in
biomarker research has enabled the characterization of the trajec-
tories of these pathologies in vivo across the AD spectrum [2] and
has deepened the understanding of AD pathophysiology. As such,
biomarkers have been incorporated in the research criteria for AD
diagnosis, which aimed to detect both the earliest stages of AD and
the full spectrum of the disease [3, 4].
235
236 Kok Pin Ng et al.
The identification of AD biomarkers crossing pathological

threshold in cognitively normal individuals and presymptomatic
autosomal-dominant mutation carriers has further led to the con-
ceptual framework which proposes a preclinical stage in both spo-
radic and familial AD [5–7]. In fact, recent therapeutic trials have
focused on preclinical AD individuals defined using abnormal AD
biomarkers [8], given that early intervention may offer the greatest
chance of treatment success. For example, solanezumab and gan-
tenerumab are being studied in asymptomatic and very mildly
symptomatic autosomal-dominant mutation carriers in the
Dominantly Inherited Alzheimer Network Trials Unit (DIAN-TU)
prevention trial [9], while solanezumab is being studied in asymp-
tomatic or mildly symptomatic older individuals who have bio-
marker evidence of brain amyloid deposition in the Anti-Amyloid
Treatment in Asymptomatic AD (A4) study [10].
There are benefits of AD biomarkers for both patients and
families suffering from AD. With genetic counseling, genetic test-
ing plays an important role in individuals with a positive family
history of dementia, such as alleviating uncertainty and anxiety and
allowing one to make plans for the future [11]. For the clinicians,
AD biomarkers may also improve clinical diagnostic certainty,
especially in patients with atypical or early-onset presentations
[12]. This may allow an earlier diagnosis and, hence, earlier access
to emerging research studies and clinical trials. AD biomarkers
such as amyloid positron emission tomography (PET) may also be
used in the selection of patients with the target pathology to enrich
clinical trial population with individuals with a higher probability
of AD-related clinical progression [13].
However, ethical issues arise with the use of AD biomarkers,
due to a conflict between the principles of beneficence and non-
maleficence [14]. As current treatment frameworks focus on the
earlier stages of the disease, which involves asymptomatic indi-
viduals, ethical issues such as the disclosure of biomarker result
and its social implications must be considered. For example,
APOE ε4 represents a higher risk for AD, and APOE ε4 carriers
may benefit from preventive or disease-modifying clinical trials
[15]. But given that APOE ε4 is neither necessary nor sufficient
for developing AD, there may be ethical concerns regarding the
disclosure of this genetic status, especially in asymptomatic indi-
viduals positive for APOE ε4 as they may face unnecessary social
discrimination and psychological distress when they may never
develop AD dementia.
In this chapter, we will discuss the issues and clinical meaning-
fulness of biomarkers in AD research. Specifically, we will focus on
the potential benefits and ethical considerations when genetics,
amyloid, tau, and neurodegeneration biomarkers are used to define
the preclinical stage of AD.
Patient Benefit and Ethical Considerations of Biomarkers 237
2 Definition of Preclinical AD
The National Institute on Aging–Alzheimer’s Association

(NIA-AA) proposed an operational three-stage research criteria in
2011 [16]. Stage 1 begins when asymptomatic cerebral amyloido-
sis is diagnosed using amyloid biomarkers (PET and cerebrospinal
fluid (CSF)). This is followed by stage 2, when neurodegeneration
markers diagnosed using CSF tau, [18F]flurodeoxyglucose (FDG)
PET, or magnetic resonance imaging (MRI) are present in addi-
tion to cerebral amyloidosis in asymptomatic individuals. Stage 3 is
defined when the above individuals develop subtle cognitive or
behavioral decline.
The international working group (IWG-2) defined two pre-
clinical states in 2014 [6]. The presymptomatic state is defined as
asymptomatic individuals who are destined to develop AD, because
they carry an autosomal-dominant mutation. The asymptomatic
at-risk state is defined as asymptomatic individuals who have either
decreased CSF Aβ42 together with increased CSF total tau (t-tau)
or phosphorylated tau (p-tau) or increased retention of amyloid
measured with amyloid PET.
Most recently, the definition of preclinical AD has been
updated based on the stratification of high or low risk for progres-
sion to clinical AD [7]. In preclinical AD when both amyloid and
tau biomarkers are beyond pathologic thresholds, the risk of clini-
cal progression to AD is particularly high. In asymptomatic at risk
for AD when either amyloid or tau biomarker is beyond pathologic
threshold, the risk of progression to clinical AD is less likely.
3 Ethical Principles
Ethics regulation has played an important role in AD research as it

protects the interest of AD patients, especially when their auton-
omy and mental capacity may be impaired in the later stages of the
disease. Ethical issues often arise due to the tension between the
principle of beneficence, such as the development of an effective
treatment, and non-maleficence, by not doing harm in this vulner-
able population [14]. These ethical issues have focused on the
decision to disclose the diagnosis and the potential social stigma
following the disclosure of the diagnosis. In addition, the benefits
of symptomatic treatment in the mild to moderate disease stages,
decision to stop treatment in the severe stage of the disease, and
the competence of the patients to give consent to research and
therapy are often debated.
However, as the AD research landscape currently shifts toward
preventive and disease-modifying treatments in the preclinical
stage of the disease, a different set of ethical issues has arisen.
Although there is no doubt regarding the autonomy and compe-

tency of the cognitively normal individuals participating in AD
research, researchers have to bear in mind the unique ethical
issues, such as the disclosure of AD diagnosis, the uncertainty of a
diagnosis based on biomarkers without full validation, and the
social stigma of having a preclinical AD, mild cognitive impair-
ment (MCI), and AD diagnosis [14]. These ethical issues arising
from the use of specific AD biomarkers will be elaborated in the
subsequent paragraphs.
4 Genetic Biomarkers
4.1 Autosomal- Familial AD represents <1% of AD, and patients often present
Dominant AD Genes with an early-onset (before 65 years old) of cognitive symptoms.
(APP, PS1, PS2) In familial AD, there is overproduction of Aβ, either due to APP
gene mutation or mutations in genes encoding presenilin 1
(PSEN1) or presenilin 2 (PSEN2), which are key components of
the γ-secretase complexes that cleave and release Aβ. Given the
autosomal-dominant inheritance of these genes, individuals with a
positive family history of first-degree relatives with these genetic
mutations have a 50% chance of having the mutation. Also, with a
penetrance of almost 100%, the chance of progression to AD in
these mutation carriers is almost 100%.
4.1.1 Benefits In individuals with a positive family history of autosomal-dom-

inant AD (ADAD), the knowledge of their carrier status of the
genetic mutation may alleviate the uncertainty and anxiety of
developing AD [17] and allow them to make important plans
for their future, such as financial and family arrangements and
when to stop working, as the likelihood of progression to AD in
mutation carriers is almost 100%.
After testing for the mutation status, these individuals with or
without ADAD genetic mutation may participate and contribute
to AD research, such as the Dominantly Inherited Alzheimer
Network (DIAN) study [18]. Given that AD pathophysiology is
postulated to develop many years before the clinical onset of
dementia [19], the DIAN aims to study early pathophysiological
changes in the asymptomatic mutation carriers, with the family
members without the mutation serving as a comparison group. In
addition, presymptomatic mutation carriers will also have access
to therapeutic clinical trials, such as the DIAN-TU [9]. DIAN-TU
is a preventive study that investigates interventions such as solan-
ezumab and gantenerumab in asymptomatic and mildly symp-
tomatic mutation carriers, as it is postulated that these individuals
may benefit from Aβ-targeting interventions. It is hypothesized
that treatment in the preclinical stage of AD will be more effective
and have a higher chance of slowing cognitive decline in these
individuals [7].
4.1.2 Ethical Issues The presymptomatic ADAD mutation carriers represent a vulner-
able group of individuals who has full autonomy and competency.
Hence, genetic counseling plays a key role in protecting their inter-
est, and ample time must be allowed for making a decision to
undergo genetic testing. In fact, a joint practice guideline for
genetic counseling and testing has been proposed to guide clini-
cians in the effective counseling on hereditary risk and genetic test-
ing [11]. In addition, a structured three-phase framework has also
been proposed for the disclosure of AD diagnosis, which is espe-
cially important in this group of asymptomatic individuals with full
insight [20]. In the “before phase,” the key objective is to deter-
mine whether one wants to know the diagnosis, their coping strat-
egies and their psychological profiles, the time and place where
disclosure will take place, and the information that will be con-
veyed. In the “during phase,” the individual’s understanding of the
disease should be established, and confidentiality has to be main-
tained unless consent is obtained to disclose the diagnosis to any
family member. The subsequent phase focuses on ensuring that the
individual understands the information that is delivered and fol-
low-up meeting and support are provided.
The association of MCI and AD diagnosis with social stigmati-
zation and discrimination is well established [21]. This social stig-
matization and discrimination will likely be faced by ADAD
mutation carriers, who are destined to develop AD. Given a recent
shift of research from clinical stages to presymptomatic disease
stages, the knowledge of the mutation status and its impact on
social stigmatization in the cognitively normal population are still
elusive. Hence, it is important that this sensitive information be
kept confidential as they may have serious implications on employ-
ment and insurance. This is especially important when a mutation
carrier has intact cognition and mental capacity and is perfectly
well to work and contribute to the society.
While the potential benefits of ADAD genetic testing in gain-
ing access to clinical trials have been discussed, the therapeutic risk-
benefit ratio of the clinical trial in presymptomatic ADAD carriers
has to be considered. For example, the use of a novel anti-amyloid
treatment, although desirable in individuals with higher risk of
AD, may be less acceptable in healthy asymptomatic individuals,
due to potential side effects such as amyloid-related imaging
abnormalities [22]. In this regard, it is important to ensure that
the interventions being studied must pass through rigorous clinical
trials that evaluate drug safety and effectiveness [9].
4.2 APOE Gene There are three polymorphic alleles in the APOE gene, ε2, ε3, and
ε4, with a frequency of 8.4%, 77.9%, and 13.7%, worldwide.
However, the frequency of ε4 allele is increased in AD at an esti-
mate of 40% [23]. The ε4 allele is the strongest genetic risk factor
for sporadic AD, leading to an earlier age of onset in a gene dose-
dependent manner [15]. The frequency of AD and the mean age

of cognitive decline are 91% and 68 years old in ε4 homozygotes,
47% and 76 years old in ε4 heterozygotes, and 20% and 84 years
old in ε4 non-carriers, respectively [15, 24]. APOE ε4 is associated
with impaired memory performance and increased risk of memory
decline in middle-aged (40–59 years) and elderly (60–85 years)
people with MCI, as well as increased risk of progression from
MCI to AD [15].
4.2.1 Benefits In a population-based study, physical inactivity, dietary fat intake,

alcohol drinking, and smoking at midlife are associated with the
risk of dementia and AD, especially among the APOE ɛ4 carriers
[25]. Type 2 diabetes is also a risk factor for AD, and the associa-
tion between diabetes and AD is particularly important among the
APOE ε4 carriers, which is supported by clinical and neuropatho-
logical data [26]. Hence, lifestyle modifications such as healthy
diet, physical activity, and optimization of cardiovascular risk fac-
tors may modify dementia risk, particularly among APOE ɛ4 carri-
ers. In this regard, a study which followed cognitively normal
individuals for 9 years to detect incident dementia cases found that
high education, active leisure activities, or maintaining vascular
health may lower dementia risk related to APOE ε4 [27].
APOE genotyping has also been included in clinical trials to
enrich the study population with individuals with a high risk of
developing AD. For example, the Alzheimer’s Prevention Initiative
(API) APOE ɛ4 trial, which aims to prevent or delay AD using
either the active amyloid immunotherapy CAD106 or the beta-
secretase inhibitor 1 (BACE1) inhibitor CNP520 versus placebo,
recruits cognitively normal individuals with homozygous ɛ4, who
have the highest risk of progression to AD.
4.2.2 Ethical Issues However, as APOE ε4 is neither necessary nor sufficient for
developing AD, genetic testing for APOE ε4 is controversial, and
genetic counseling for the individual is important to discuss the
benefits and risks prior to testing. Although there is no effective
prevention or treatment when one is tested positive for APOE ε4,
these individuals may benefit from participating in preventive tri-
als. It is also important to emphasize that the confidentiality of
APOE ε4 status should be maintained, given that APOE ε4 is
increasingly tested in the research setting. This will protect the
interest of the individual as APOE ε4 carriers may face social stig-
matization and discrimination.
There have been concerns that disclosure of the genetic
screening results may lead to negative emotions among the APOE
ε4 carriers. In the Risk Evaluation and Education for AD
(REVEAL) study which examines the effects of APOE genotype
status disclosure to asymptomatic individuals with a first-degree
relative with AD, no difference was found in terms of anxiety and
depression in individuals who are informed that they are APOE ε4

carriers, compared to non-carriers, and those not informed of
their APOE status [28]. However, given the limitations of the
study, the authors concluded that larger studies that follow sub-
jects for more than 1 year will be required to detect uncommon
and long-term effects, such as delayed emotional repercussions
and injudicious life decisions.
5 Amyloid Biomarkers
Low CSF Aβ42 and high uptake of amyloid PET tracers are bio-
markers of amyloid fibrillary deposition [2]. There is a close cor-
relation between in vivo amyloid biomarkers and brain amyloid
plaques assessed in postmortem studies [29–31]. Longitudinal
studies have suggested that AD biomarkers, beginning with
amyloid deposition, become abnormal sequentially in a pro-
longed preclinical period [19]. In fact, CSF Aβ42 levels decline
up to 20 years before symptom onset [32]. There is high con-
cordance between CSF Aβ42 and amyloid PET in MCI and AD
patients [33]. In addition, amyloid PET provides information
regarding regional distribution and longitudinal changes of amy-
loid deposition. However, 10–30% of cognitively normal people
have a positive amyloid PET scan [34], and pathological amyloid
deposition in the brain is also shown in non-demented individu-
als [35]. Hence, amyloid pathology is proposed to be necessary,
but not sufficient to cause progression to a symptomatic stage of
the disease [2].
5.1 Benefits Given a high degree of pathological heterogeneity underlying

the clinical diagnosis of MCI [1], AD biomarkers, including
amyloid, have been recently incorporated in the research criteria
to improve the certainty of making a diagnosis of MCI due to
AD [36]. An abnormal amyloid PET scan increases the likeli-
hood of AD, while a negative amyloid PET scan provides useful
information in ruling out AD etiology in atypical or early-onset
presentations [12]. Hence, amyloid PET is considered appropri-
ate in individuals with atypical presentation or dementia of
uncertain etiology [37]. Amyloid PET in MCI may also provide
useful prognostic i nformation. A study showed that PIB-positive
MCI patients are significantly more likely to convert to AD than
PIB-negative patients, and faster converters have higher PIB
retention levels at baseline [38].
Amyloid biomarkers have also been incorporated in the pre-
clinical AD research criteria, which, together with tau, identify indi-
viduals with the highest risk of progression to AD dementia [7]. In
this regard, amyloid biomarkers may guide study eligibility and tar-
get engagement in disease-modifying clinical trials, especially in
early asymptomatic stage, as these biomarkers reflect in vivo bio-

logical process that occurs many years before symptom onset. In the
phase 3 clinical trial with anti-Aβ monoclonal antibody bapineu-
zumab, up to 36.1% of study participants have negative amyloid
PET scans [39]. Since then, subsequent clinical trials such as the
EXPEDITION 3 with solanezumab and PRIME with aducanumab
have added amyloid biomarkers in the inclusion criteria to enrich
their study population. The A4 trial, which aims to delay progres-
sion from normal cognition to MCI or dementia using solane-
zumab, also recruits asymptomatic and mildly symptomatic
individuals with biomarker evidence of brain amyloid deposition
[10]. Amyloid biomarkers are used in outcome measures to dem-
onstrate target engagement. For example, following treatment with
aducanumab, brain Aβ plaques as measured by florbetapir PET
imaging are reduced in a dose- and time-dependent fashion [13].
5.2 Ethical Issues While it is expected that AD biomarkers, including amyloid, will be
able to identify preclinical individuals who will progress to AD,
current biomarker follow-up studies provide elusive information
regarding biomarker signatures of those who will progress to
dementia [40, 41]. The uncertainty of AD diagnosis based on bio-
markers is further demonstrated in MCI individuals, where 50%
have amyloid pathology, but only 40% of them develop dementia
in 3 years [38]. Hence, the clinical use of amyloid biomarkers is
currently not recommended in asymptomatic individuals [37].
The disclosure of amyloid status to study participants is also a
key ethical issue. Among investigators, there are debates as to
whether the results of amyloid biomarkers should be revealed to
the participants. The Alzheimer’s Disease Neuroimaging Initiative
(ADNI) study has maintained a position where biomarker results,
including amyloid, are not disclosed to study participants, due to
uncertainty of the prognostic value of the biomarker [42].
However, this stance is likely to change once clinical utility associ-
ated with biomarkers is better established. In fact, recent clinical
trials, such as the A4 which studies disease-modifying agents in
amyloid-positive individuals, will, by study design, result in
asymptomatic participants being informed of their amyloid imag-
ing results [10]. Given that the implications of amyloid status on
employment, health insurance, and social stigmatization are not
fully understood, there is a need to accelerate the development of
a best practice framework for disclosure of amyloid status and to
identify individuals at risk of potential devastating reactions.
While amyloid PET has a clear benefit in research, its clinical
cost-effectiveness is less clear [43]. There remains an uncertainty
regarding the validity of amyloid PET, such as the presence of false
positives and false negatives, the adjustment of threshold for amyloid
positivity due to age and genetic factors, and the optimal threshold
for determining scan positivity using a quantitative approach [44].
Similarly, interlaboratory variations of CSF analysis including Aβ42

using current immunoassays are of a concern [45]. In the absence of
disease-modifying therapy for AD at present, the important goal of
a diagnostic evaluation is to identify treatable conditions or modifi-
able comorbidities. In this regard, amyloid PET adds little to the
diagnostic workup of patients with a classic amnestic AD phenotype
[37] and does not offer additional information on treatable or modi-
fiable conditions such as cerebrovascular disease. As such, there
remains a strong debate regarding potential cost-savings versus
increasing cost when utilizing amyloid PET in dementia evaluation
[46]. However, this is likely to change once there are effective dis-
ease-modifying therapies and therefore the clinical necessity of
molecular imaging to better identify the therapeutic targets.
6 Tau Biomarkers
Neurofibrillary tangles (NFT) are intraneuronal aggregates of

hyperphosphorylated and misfolded tau proteins and become
extraneuronal when neurons with tangles die [1, 35]. There is a
strong correlation between cortical NFT and cognitive decline
[47, 48]. Elevated CSF p-tau and high uptakes of tau PET tracers
are biomarkers of NFT. CSF p-tau is specific for the burden of
AD-type tau pathology [49, 50], and its levels correlate with the
severity of tau pathology in postmortem studies [51]. Tau PET
imaging constitutes a new frontier in AD research [52], and cur-
rent evidence suggests that tau imaging agents are able to demon-
strate the topography of tau spread consistent with the description
at autopsy by Braak in the typical AD spectrum [53, 54].
6.1 Benefits Tau biomarkers have been incorporated in the research diagnostic
criteria which enables the detection of AD pathophysiology at the
early stages of the disease [4, 36]. Increased CSF p-tau is more
sensitive and specific than CSF t-tau and Aβ42 in discriminating AD
from normal aging and non-AD dementia [55], and a high CSF
p-tau/Aβ42 ratio also provides high diagnostic accuracy in differen-
tiating AD from normal controls and from subjects with non-AD
dementia [56]. High uptake of tau tracers in the temporal lobes is
able to distinguish AD from healthy individuals [57], and in
amyloid-positive individuals, increasing levels of tau binding are
associated with higher cognitive impairment [58].
Studies and clinical trials have tested interventions that target
pathological tau aggregation [59, 60], and similar to amyloid bio-
markers, it is likely that tau biomarkers will play a key role in future
clinical trials and observational studies. Although tau PET imaging
is still in the early stages of development, it is anticipated that tau
imaging agents will enable the staging of the disease, selection of
patients appropriate for a given tau therapy, confirmation of target
engagement, and monitoring treatment efficacy for AD [54].
6.2 Ethical Issues Given that tau PET imaging is still in the early stages of develop-
ment, there has been limited clinical and research data for its cur-
rent use compared to amyloid PET imaging. In fact, off target
binding of tau tracers has been reported [61, 62], and its implica-
tions on the interpretation of scan findings will need to be carefully
studied, given the potential use of tau PET imaging in future clini-
cal trials as a marker of target engagement. In this regard, the dis-
closure of tau PET imaging results is also a potential ethical issue,
given the uncertainty of the validity of this biomarker.
7 Neurodegeneration Biomarkers
Neurodegeneration represents the progressive loss of structure or

function of neurons [63]. High CSF t-tau, decreased FDG PET
uptake, and atrophy on structural MRI are considered biomark-
ers of neuronal degeneration. CSF t-tau levels correlate with neu-
ronal tissue damage and reflect the intensity of neuronal
degeneration in AD [49, 64]. FDG PET measures cerebral meta-
bolic rates of glucose, which is postulated to be a proxy for syn-
aptic activity. AD is characterized by a regional hypometabolism
in the parietotemporal cortex, medial temporal lobes, and poste-
rior cingulate cortex [63]. MRI measures brain atrophy, which is
associated with neuronal degeneration and tissue loss. Vulnerable
brain regions for atrophy in AD include the hippocampus and
entorhinal cortex and are able to predict progression from MCI
to AD dementia [65].
7.1 Benefits Neurodegeneration biomarkers play an important role in the clini-

cal evaluation of cognitive symptoms. From 36 studies with about
2500 AD patients and 1400 controls, the pooled sensitivity and
specificity for CSF t-tau in differentiating AD from controls are
81% and 90%, respectively [49]. In a longitudinal study of MCI
patients, a combination of CSF t-tau and Aβ42 at baseline gave a
sensitivity of 95% and a specificity of 83% for detection of incipient
AD [66], further supporting the diagnostic use of CSF t-tau in the
MCI stage. CSF t-tau levels also correlate with poor cognitive per-
formance in AD [67]. FDG PET has high sensitivity in distinguish-
ing AD from controls and individuals with higher versus lower AD
risk and has good quantitative and topographical correlation with
clinical progression [68]. FDG also improves accuracy in distin-
guishing frontotemporal dementia (FTD) from AD [69]. Atrophy
of medial temporal structures in MRI is a valid diagnostic marker
at the MCI stage, and hippocampal atrophy is able to distinguish
AD from individuals with no cognitive impairment using visual rat-
ing scales [70]. The rates of change in structural measures, such as

the entorhinal cortex, hippocampus, and temporal lobes, also cor-
relate closely with changes in cognitive performance. However, it
is important to note that medial temporal atrophy by itself lacks
specificity to exclude other dementia, and its combination with
other biomarker modalities such as CSF enhances the accuracy of
making a diagnosis of AD [65].
MRI, which is widely available and devoid of radiation risk, has
the added benefit of assessing comorbidities that increase AD risk,
such as white matter hyperintensities, a marker of small vessel cere-
brovascular disease, or other causes of cognitive symptoms, such as
a space-occupying lesion. This adds valuable information during
clinical evaluation, which aid in the treatment management of the
patient, such as optimization of cardiovascular risk factors.
7.2 Ethical Issues While neurodegeneration biomarkers support the diagnosis of AD,
it is important to note that these biomarkers are least specific for
AD compared to biomarkers for amyloid and tau pathologies. Brain
atrophy in MRI and hypometabolism in FDG PET scan involving
regions associated with AD can also occur in a variety of other dis-
orders [71]. For example, atrophy in the anterior, medial, and basal
temporal lobes can occur in non-AD tauopathies such as progres-
sive supranuclear palsy, as well as in cerebrovascular disease, epi-
lepsy, hippocampal sclerosis, primary age-related tauopathy, and
argyrophilic grain disease. Hypometabolism in the temporoparietal
region can also be found in corticobasal degeneration and cerebro-
vascular disease [50]. The non-specificity of neurodegeneration
biomarkers has to be taken into account when explaining the
biomarker results to patients as this can lead to inaccurate diagnosis,
wrong treatment, and unnecessary anxiety.
8 Summary
Biomarkers play a key role in AD research which advanced our

understanding of the AD pathophysiology. With the finding of
abnormal AD biomarkers in cognitively normal individuals, a
conceptual framework of a preclinical stage of AD has been pro-
posed. While there are definite benefits linked to the use of bio-
markers to define this early disease stage in research setting, a
unique set of ethical issues have arisen as a result of a conflict
between beneficence and non-maleficence. It is important to
keep in mind these ethical concerns so as to protect the interests
of research participants, especially those in the asymptomatic
stages of the disease.
Acknowledgments
Kok Pin Ng is supported by the National Medical Research

Council, Research Training Fellowship Grant (Singapore).
Pedro Rosa-Neto and Serge Gauthier are funded by the
Canadian Institutes for Health Research.
References
1. Serrano-Pozo A, Frosch MP, Masliah E, 11. Goldman JS, Hahn SE, Williamson Catania
Hyman BT (2011) Neuropathological altera- J et al (2011) Genetic counseling and testing
tions in Alzheimer disease. Cold Spring Harb for Alzheimer disease: joint practice guidelines
Perspect Med 1:a006189 of the American college of medical genetics
2. Jack CR, Knopman DS, Jagust WJ et al (2013) and the national society of genetic counselors.
Tracking pathophysiological processes in Genet Med 13:597–605
Alzheimer’s disease: an updated hypothetical 12. Morris E, Chalkidou A, Hammers A et al
model of dynamic biomarkers. Lancet Neurol (2016) Diagnostic accuracy of (18)F amyloid
12:207–216 PET tracers for the diagnosis of Alzheimer’s
3. Dubois B, Feldman HH, Jacova C et al disease: a systematic review and meta-analysis.
(2007) Research criteria for the diagnosis of Eur J Nucl Med Mol Imaging 43:374–385
Alzheimer’s disease: revising the NINCDS- 13. Sevigny J, Chiao P, Bussière T et al (2016) The
ADRDA criteria. Lancet Neurol 6:734–746 antibody aducanumab reduces Aβ plaques in
4. McKhann GM, Knopman DS, Chertkow H Alzheimer’s disease. Nature 537:50–56
et al (2011) The diagnosis of dementia due to 14. Gauthier S, Leuzy A, Racine E, Rosa-Neto
Alzheimer’s disease: recommendations from P (2013) Diagnosis and management of
the national institute on aging-Alzheimer’s Alzheimer’s disease: past, present and future
association workgroups on diagnostic guide- ethical issues. Prog Neurobiol 110:102–113
lines for Alzheimer’s disease. Alzheimers 15. Liu C, Kanekiyo T, Xu H, Bu G (2013)
Dement 7:263–269 Apolipoprotein E and Alzheimer disease: risk,
5. Sperling RA, Aisen PS, Beckett LA et al mechanisms and therapy. Nat Rev Neurol
(2011) Toward defining the preclinical stages 9:106–118
of Alzheimer’s disease: recommendations 16. Sperling RA, Aisen PS, Beckett LA et al (2011)
from the national institute on aging and the Toward defining the preclinical stages of
Alzheimer’s association workgroup. Alzheimers Alzheimer’s disease: recommendations from
Dement 7:1–13 the national institute on aging- Alzheimer’s
6. Dubois B, Feldman HH, Jacova C et al (2014) association workgroups on diagnostic guide-
Advancing research diagnostic criteria for lines for Alzheimer’s disease. Alzheimers
Alzheimer’s disease: the IWG-2 criteria. Lancet Dement 7:280–292
Neurol 13:614–629 17. Steinbart EJ, S D MZ et al (2001) Impact of
7. Dubois B, Hampel H, Feldman HH et al DNA testing for early-onset familial Alzheimer
(2016) Preclinical Alzheimer’s disease: defini- disease and frontotemporal dementia. Arch
tion, natural history, and diagnostic criteria. Neurol 58:1828
Alzheimers Dement 12:292–323 18. Bateman RJ, Xiong C, Benzinger TLS,

8. Folch J, Petrov D, Ettcheto M et al (2016) Fagan AM, Goate A, Fox NC, Marcus DS,
Current research therapeutic strategies for Cairns NJ, Xie X, Nlazey TM, Holtman DM,
Alzheimer’s disease treatment. Neural Plast Santacruz A, Buckles V, Oliver A, Moulder
2016:1–15 K, Aisen PS, Ghetti B, Klunk WE, McDade
9. Bateman RJ, Benzinger TL, Berry S et al E, Martins RN, Masters CL, Mayeux R,
(2017) The DIAN-TU next generation Ringman JM, Rossor MN, Schofield PR,
Alzheimer’s prevention trial: adaptive design Sperling RA, Salloway S, Morris JC (2013)
and disease progression model. Alzheimers Clinical and biomarker changes in dominantly
Dement 13:8–19 inherited Alzheimer’s disease. N Engl J Med
10. Sperling RA, Aisen PS (2016) Anti-amyloid 367:795–804.
treatment of asymptomatic AD: A4 and 19. Bateman RJ, Xiong C, Benzinger TLS et al
beyond. Alzheimers Dement 12:P326–P327 (2012) Clinical and biomarker changes in
dominantly inherited Alzheimer’s disease. N 33. Leuzy A, Chiotis K, Hasselbalch SG et al

Engl J Med 367:795–804 (2016) Pittsburgh compound B imaging
20. Pepersack T (2008) Disclosing a diagnosis of and cerebrospinal fluid amyloid-β in a multi-
Alzheimer’s disease. Rev Med Brux 29:89–93 centre European memory clinic study. Brain
21. Nicole L Batsch, Mary S Mittelman (2012) 139:2540–2553
World Alzheimer report 2012 overcoming the 34. Chételat G, La Joie R, Villain N et al (2013)
stigma of dementia. https://www.alz.org/ Amyloid imaging in cognitively normal individu-
documents_custom/world_report_2012_ als, at-risk populations and preclinical Alzheimer’s
final.pdf. Accessed 8 Apr 2017 disease. Neuroimage Clin 2:356–365
22. Peters KR, Lynn Beattie B, Feldman HH 35. Price JL, Morris JC (1999) Tangles and plaques
(2013) A conceptual framework and ethics in nondemented aging and “preclinical”
analysis for prevention trials of Alzheimer dis- Alzheimer’s disease. Ann Neurol 45:358–368
ease. Prog Neurobiol 110:114–123 36. Albert MS, DeKosky ST, Dickson D et al
23. Farrer LA, Cupples LA, Haines JL et al (1997) (2011) The diagnosis of mild cognitive
Effects of age, sex, and ethnicity on the associa- impairment due to Alzheimer’s disease: rec-
tion between apolipoprotein E genotype and ommendations from the national institute on
Alzheimer disease. A meta-analysis. APOE and aging-Alzheimer’s association workgroups on
Alzheimer disease meta analysis consortium. diagnostic guidelines for Alzheimer’s disease.
JAMA 278:1349–1356 Alzheimers Dement 7:270–279
24. Corder EH, Saunders AM, Strittmatter WJ 37. Johnson KA, Minoshima S, Bohnen NI et al
et al (1993) Gene dose of apolipoprotein E (2013) Appropriate use criteria for amyloid
type 4 allele and the risk of Alzheimer’s disease PET: a report of the amyloid imaging task
in late onset families. Science 261:921–923 force, the society of nuclear medicine and
25. Kivipelto M, Rovio S, Ngandu T et al (2008) molecular imaging, and the Alzheimer’s asso-
Apolipoprotein E ɛ4 magnifies lifestyle risks for ciation. Alzheimers Dement 9:E1–E16
dementia: a population-based study. J Cell Mol 38. Okello A, Koivunen J, Edison P et al (2009)
Med 12:2762–2771 Conversion of amyloid positive and negative
26. Peila R, Rodriguez BL, Launer LJ (2002) Type MCI to AD over 3 years: an 11C-PIB PET
2 diabetes, APOE gene, and the risk for demen- study. Neurology 73:754–760
tia and related pathologies the Honolulu-Asia 39. Salloway S, Sperling R, Fox NC et al (2014)
aging study. Diabetes 51:1256–1262 Two phase 3 trials of Bapineuzumab in mild-
27. Ferrari C, W-L X, Wang H-X et al (2013) to-moderate Alzheimer’s disease. N Engl
How can elderly apolipoprotein E ε4 carriers J Med 370:322–333
remain free from dementia? Neurobiol Aging 40. Hertze J, Minthon L, Zetterberg H et al (2010)
34:13–21 Evaluation of CSF biomarkers as predictors of
28. Green RC, Roberts JS, Cupples LA et al Alzheimer’s disease: a clinical follow-up study
(2009) Disclosure of APOE genotype for of 4.7 years. J Alzheimers Dis 21:1119–1128
risk of Alzheimer’s disease. N Engl J Med 41. Bertens D, Knol DL, Scheltens P, Visser PJ
361:245–254 (2015) Temporal evolution of biomarkers and
29. Strozyk D, Blennow K, White LR, Launer cognitive markers in the asymptomatic, MCI,
LJ (2003) CSF Abeta 42 levels correlate with and dementia stage of Alzheimer’s disease.
amyloid-neuropathology in a population-based Alzheimers Dement 11:511–522
autopsy study. Neurology 60:652–656 42. Shulman MB, Harkins K, Green RC, Karlawish
30. Ikonomovic MD, Klunk WE, Abrahamson EE J (2013) Using AD biomarker research results
et al (2008) Post-mortem correlates of in vivo for clinical care: a survey of ADNI investiga-
PiB-PET amyloid imaging in a typical case of tors. Neurology 81:1114–1121
Alzheimer’s disease. Brain 131:1630–1645 43. Leuzy A, Zimmer ER, Heurling K et al (2014)
31. Sojkova J, Driscoll I, Iacono D et al (2011) In Use of amyloid PET across the spectrum of
vivo fibrillar β-amyloid detected using [11C] Alzheimer’s disease: clinical utility and associ-
PiB positron emission tomography and neu- ated ethical issues. Amyloid 21:143–148
ropathologic assessment in older adults. Arch 44. Laforce R, Rabinovici GD (2011) Amyloid
Neurol 68:232–240 imaging in the differential diagnosis of demen-
32. Fagan AM, Xiong C, Jasielec MS et al (2014) tia: review and potential clinical applications.
Longitudinal change in CSF biomarkers in Alzheimers Res Ther 3:11
autosomal-dominant Alzheimer’s disease. Sci 45. Blennow K, Dubois B, Fagan AM et al (2015)
Transl Med 6:226ra30 Clinical utility of cerebrospinal fluid biomarkers
in the diagnosis of early Alzheimer’s disease. clinical diagnosis, age and cognition. Brain
Alzheimers Dement 11:58–69 140:748–763
46. Caselli R, Woodruff B (2016) Clinical impact 59. Gauthier S, Feldman HH, Schneider LS et al
of amyloid positron emission tomography—is it (2016) Efficacy and safety of tau-aggregation
worth the cost? JAMA Neurol 73:1396–1398 inhibitor therapy in patients with mild or
47. Sabbagh MN, Cooper K, DeLange J et al moderate Alzheimer’s disease: a randomised,
(2010) Functional, global and cognitive decline controlled, double-blind, parallel-arm, phase
correlates to accumulation of Alzheimer’s 3 trial. Lancet 388:2873–2884
pathology in MCI and AD. Curr Alzheimer 60. Yanamandra K, Jiang H, Mahan TE et al
Res 7:280–286 (2015) Anti-tau antibody reduces insoluble tau
48. Nelson PT, Alafuzoff I, Bigio EH et al (2012) and decreases brain atrophy. Ann Clin Transl
Correlation of Alzheimer disease neuropatho- Neurol 2:278–288
logic changes with cognitive status: a review 61. Ng KP, Pascoal TA, Mathotaarachchi S et al
of the literature. J Neuropathol Exp Neurol (2017) Monoamine oxidase B inhibitor, sele-
71:362–381 giline, reduces 18 F-THK5351 uptake in the
49. Blennow K, Hampel H (2003) CSF markers for human brain. Alzheimers Res Ther 9:25
incipient Alzheimer’s disease. Lancet Neurol 62. Vermeiren C, Mercier J, Viot D et al (2015)
2:605–613 T807, a reported selective tau tracer, binds
50. Jack CR, Hampel HJ, Universities S et al with nanomolar affinity to monoamine oxidase
(2016) A/T/N: an unbiased descriptive classi- a. Alzheimers Dement 11:P283
fication scheme for Alzheimer disease biomark- 63. Mosconi L (2013) Glucose metabolism in normal
ers. Neurology 87:539–547 aging and Alzheimer’s disease: m ethodological
51. Tapiola T, Alafuzoff I, Herukka S-K et al and physiological considerations for PET stud-
(2009) Cerebrospinal fluid β-amyloid 42 and ies. Clin Transl Imaging 1:217–233
tau proteins as biomarkers of Alzheimer-type 64. Ost M, Nylén K, Csajbok L et al (2006) Initial
pathologic changes in the brain. Arch Neurol CSF total tau correlates with 1-year outcome in
66:734–746 patients with traumatic brain injury. Neurology
52. Villemagne VL, Fodero-Tavoletti MT, Masters 67:1600–1604
CL, Rowe CC (2015) Tau imaging: early 65. Frisoni GB, Fox NC, Jack CR et al (2010) The
progress and future directions. Lancet Neurol clinical use of structural MRI in Alzheimer
14:114–124 disease. Nat Rev Neurol 6:67–77
53. Braak H, Braak E (1995) Staging of 66. Hansson O, Zetterberg H, Buchhave P et al
Alzheimer’s disease-related neurofibrillary (2006) Association between CSF biomarkers
changes. Neurobiol Aging 16:271–278. discus- and incipient Alzheimer’s disease in patients
sion 278–284 with mild cognitive impairment: a follow-up
54. Johnson KA, Schultz A, Betensky RA et al study. Lancet Neurol 5:228–234
(2016) Tau positron emission tomographic 67. Skillbäck T, Farahmand BY, Rosén C et al (2015)
imaging in aging and early Alzheimer disease. Cerebrospinal fluid tau and amyloid-β 1-42 in
Ann Neurol 79:110–119 patients with dementia. Brain 138:2716–2731
55. Koopman K, Le Bastard N, Martin J, Nagels 68. Mosconi L, Berti V, Glodzik L et al (2010)
G (2009) Improved discrimination of Pre-clinical detection of Alzheimer’s disease
autopsy-confirmed Alzheimer’s disease (AD) using FDG-PET, with or without amyloid
from non-AD dementias using CSF P-tau imaging. J Alzheimers Dis 20:843–854
181P. Neurochem Int 55:214–218 69. Foster NL, Heidebrink JL, Clark CM et al
56. Maddalena A, Papassotiropoulos A (2003) (2007) FDG-PET improves accuracy in dis-
Biochemical diagnosis of Alzheimer disease by tinguishing frontotemporal dementia and
measuring the cerebrospinal fluid ratio of phos- Alzheimer’s disease. Brain 130:2616–2635
phorylated tau protein to β-amyloid peptide42. 70. Scheltens P, Leys D, Barkhof F et al (1992)
Arch Neurol 60:1202–1206 Atrophy of medial temporal lobes on MRI in
57. Harada R, Okamura N, Furumoto S et al “probable” Alzheimer’s disease and normal
(2016) 18F-THK5351: a novel PET radio- ageing: diagnostic value and neuropsychologi-
tracer for imaging neurofibrillary pathology in cal correlates. J Neurol Neurosurg Psychiatry
Alzheimer’s disease. J Nucl Med 57:208–214 55:967–972
58. Pontecorvo MJ, Devous MD Sr, Navitsky 71. Fotuhi M, Do D, Jack C (2012) Modifiable
M et al (2017) Relationships between flor- factors that alter the size of the hippocampus
taucipir PET tau binding and amyloid burden, with ageing. Nat Rev Neurol 8:189–202
Chapter 17
The Ethics of Biomarker-Based Preclinical Diagnosis

of Alzheimer’s Disease
Alexander F. Kurz and Nicola T. Lautenschlager
Abstract
In older adults who are developing Alzheimer-typical pathology but do not experience any symptoms—a
condition termed “preclinical” Alzheimer’s disease—the individual prognostic value of current biomarkers
is low. Only a minority of those who have positive findings on biomarkers in the cerebrospinal fluid or
upon brain imaging experience cognitive decline within 3–5 years. On the other hand, the majority of
people who have negative biomarker results remain cognitively intact. This may be expected in research
settings where alternative causes of cognitive impairment are infrequent. The main reasons for the poor
predictive performance of biomarkers in asymptomatic individuals are the dynamic nature of the assess-
ments with abnormality evolving over time and the large interindividual variability of the threshold for
clinical manifestation. We do not recommend biomarker assessment and disclosure of biomarker results to
asymptomatic individuals outside of research protocols or clinical trials at this point in time. More needs
to be learned about the interplay between biomarkers, risk factors, and protective factors. Also, studies are
required with extended follow-up intervals on less selected participant groups to determine the long-term
predictive potential of biomarkers in presymptomatic Alzheimer’s disease.
Key words Alzheimer’s disease, Biomarkers, Preclinical, Ethics
1 Introduction
The clinical symptoms of Alzheimer’s disease (AD) including mild

cognitive impairment and dementia are late manifestations of the
neurodegeneration. It is believed that the pathological changes
begin decades before the diagnosis of AD can be established on
clinical grounds [1, 2]. The stage of the neurodegeneration when
pathological changes accumulate but are not clinically expressed is
termed “preclinical AD.” The subsequent stage of “prodromal
AD” is characterized by subtle deficits in cognitive areas such as
attention, information processing speed, working memory, epi-
sodic memory, and executive function, often labeled as “mild
cognitive impairment” (MCI) [3]. The prodromal stage precedes
the clinical diagnosis of dementia for several years [4]. In parallel,
249
250 Alexander F. Kurz and Nicola T. Lautenschlager
it has been shown that key features of the pathology including

deposition of β-amyloid, aggregation of tau, metabolic dysfunction
of n
eurons, and neuronal loss can be identified by measurements in
the cerebrospinal fluid (CSF), positron emission tomography (PET),
and, at a more advanced stage, magnetic resonance imaging (MRI)
independently of clinical symptoms. These assessments are collec-
tively referred to as “biomarkers.” An additional class of diagnostic
indicators is genetic tests for rare pathogenic mutations. Since the
latter reveal the predisposition for developing the disease rather
than the presence of brain changes, they are usually not included in
the group of biomarkers. A large number of drug trials conducted
during the past 20 years have uniformly shown that at the stage of
overt clinical symptoms, treatment effects are disappointing. One
of the explanations for this is that the damage to neuronal net-
works has advanced too far. Taken together, these observations
have given rise to the hope that intervening with the neurodegen-
erative process at the preclinical or prodromal stage might have a
greater impact on the evolution of pathology, on cognitive and
functional decline, and ultimately on the development of dementia
and disability. This novel scenario of AD management would
require the combination of excellent diagnostic services, precise
biomarkers, and effective disease-modifying treatments. In this
chapter we focus on with the ethical issues that are associated with
the assessment of biomarkers in asymptomatic individuals.
2 Methods
We conducted a selective literature search using the databases

CINAHL, EMBASE, MEDLINE, and PubMed and the search
terms Alzheimer’s disease, biomarker, early diagnosis, and ethics.
3 Results
3.1 What Kind Two types of biomarkers can be distinguished: one category indi-
of Information Do cating β-amyloid pathology (low β-amyloid concentration in the
Biomarkers Provide? CSF and evidence of β-amyloid deposits on imaging) and another
category demonstrating neuronal injury and neurodegeneration
(high total tau or phosphorylated tau in the CSF, glucose hypome-
tabolism in typical locations, accumulation of tau, and hippocam-
pal atrophy on imaging). When using and interpreting biomarkers,
particularly in asymptomatic people, their specific nature and asso-
ciated problems need to be observed.
–– Biomarkers of both categories capture different single ele-
ments of AD neurodegeneration [5, 6]. These features are not
specific of AD but also occur in other brain diseases such as
Ethics of Biomarker-Based Preclinical Diagnosis 251
Lewy body disease, frontotemporal degenerations,

Creutzfeldt-Jakob disease, and traumatic brain injury and even
in old age itself [7–9]. Therefore, biomarkers should not be
looked at in isolation.
–– Biomarker findings have been found to evolve in a temporal
sequence. In most cases, indicators of β-amyloid pathology are
the first to become abnormal [10–12]. However, neurodegen-
eration in terms of atrophy of the mediotemporal lobe may
occur in the absence of detectable β-amyloid accumulation [13,
14]. This implies that AD does not always present the complete
“signature,” and findings may be conflicting [15]. No algo-
rithm has been developed yet to deal with this situation [16].
–– Conflicting results may not only be encountered between but
also within biomarker categories. The concordance of assess-
ments is not perfect, and the biomarkers of one category are
not interchangeable. This is particularly true for the indicators
of neuronal injury and neurodegeneration [17–21].
–– There is no close association between the amount of patho-
logical changes in the brain and the severity of clinical symp-
toms. This especially applies to β-amyloid burden. The clinical
manifestation of pathology is counterbalanced by resilience or
“reserve” factors such as education, occupational attainment,
lifetime mental activity, and other factors [22, 23]. The rela-
tionship between these factors and the biomarkers is largely
unexplored [24].
For these reasons, information provided by biomarkers is par-
ticularly difficult to interpret in asymptomatic people. In individu-
als who already experience symptoms, e.g., are diagnosed with
MCI, normal biomarker findings make the presence of AD pathol-
ogy less likely and point to an alternative cause of cognitive dete-
rioration [25]. Conversely, abnormal biomarker results support
the diagnosis of prodromal AD [26, 27] and predict progression to
dementia with high accuracy [19, 28]. The scenario in asymptom-
atic individuals is different. Here, normal biomarker findings may
indicate that there is no brain disease and no increased risk of cog-
nitive decline and dementia. However, results will also be normal if
the pathological changes are still below detection threshold. The
biomarkers may become abnormal several years later, and clinical
deterioration will eventually occur. Abnormal biomarker results in
asymptomatic individuals are associated with an increased risk of
future cognitive decline [29], particularly if they refer to both cat-
egories [30]. However, they do not predict whether or when the
clinical manifestations of AD will develop during life. Thus, in
asymptomatic individuals, even if they harbor the disease, a much
wider interindividual variability can be expected regarding the
distance to clinical manifestation than in people who already have
symptoms, and therefore predictions of cognitive deterioration will

necessarily be uncertain.
Predictive testing (i.e., in asymptomatic individuals) for
pathogenic mutations in the amyloid precursor and presenilin
genes differs from CSF and imaging biomarkers in several respects.
It affects a small minority of individuals, since only 1–5% of AD
follows an autosomal dominant pattern of inheritance [31]. The
results of these tests are invariable over time and have a very high
predictive value. Moreover, they have implications not only for
the tested individuals but also for their blood-related family mem-
bers, as their disease risk will increase dramatically by positive find-
ings. The ε4 variant of the apolipoprotein E gene is a susceptibility
gene. Testing has neither diagnostic nor predictive relevance and
therefore is not recommended for clinical practice [32].
3.2 How Well Do Among cognitively healthy older adults, 25–45% have abnormal CSF
Biomarkers Predict or imaging biomarker values [33]. Amyloid positivity on imaging
Cognitive Decline? increases with age from 5% in those 50–60 years to 46% in people
aged 81 years and older [34]. Abnormal CSF biomarker results are
associated with an increased likelihood of cognitive decline within a
follow-up interval of 3 years. However, even in a research setting,
using the best biomarker combination and optimal cutoff values,
individual prediction only achieved 65% accuracy, with sensitivity and
specificity values being 74% and 57%, respectively. These figures indi-
cate that one third of cognitive decliners were not identified, while
almost half of the predictions were false positive [35]. In a longitudi-
nal study at Washington University, cognitively healthy adults over
the age of 65 years who were positive for CSF ß-amyloid showed
clinical deterioration over 5 years in only 13%. If they also had a posi-
tive biomarker for neuronal injury, the progression rate increased to
25% [1]. Thus, a considerable number of healthy older individuals
with abnormal CSF biomarker findings will not experience cognitive
deterioration within the next couple of years. Similarly, abnormal
findings on ß-amyloid imaging are generally associated with greater
decline in memory and other domains and indicate an increased risk
of dementia on a group level [36]. However, such results do not
predict with certainty if and when someone will develop clinical
symptoms [37–40]. In the Australian Imaging, Biomarkers and
Lifestyle (AIBL) study, only 26% of cognitively healthy individuals
who had an AD-like result on ß-amyloid imaging progressed to MCI
or dementia within 3 years. If a positive ß-amyloid scan was com-
bined with reduced hippocampal volume on MRI, the percentage of
cognitive decliners increased to 47%. Among participants with nor-
mal imaging findings in both modalities, 92% remained cognitively
intact [41]. These findings show that within a follow-up period of
several years, less than half of cognitively healthy older adults who
have abnormal biomarker findings will experience cognitive decline.
Most but not all of those who have normal biomarker results are
likely to remain cognitively stable. This may be expected in research

settings where the a priori probability of alternative causes of cogni-
tive decline is low. Studies on less selected participants with longer-
follow-up intervals are needed to determine the longer-term
predictive value of biomarkers in asymptomatic individuals.
3.3 What Are The greatest benefit of biomarker assessment in an asymptomatic

the Benefits individual would be if, in the case of abnormal findings, a treatment
of Biomarker could be initiated which delays the progression of disease and post-
Information? pones the onset of cognitive impairment. However, such interven-
tions are currently not available. A potential benefit may be seen in
the participation in a clinical trial which evaluates the efficacy of a
treatment for secondary prevention. Due to high costs and meth-
odological difficulties, we expect that only few studies of this kind,
e.g., the A4 trial, will be conducted and also will be an option only
for a small minority due to strict inclusion and exclusion criteria.
Moreover, potential participants need to be aware that treatments
in drug trials are experimental and may not be effective [42]. An
important benefit for the worried well, particularly for family mem-
bers of people who have or have had dementia, is the assurance they
might get from negative biomarker findings that they are not at
risk. However, due to the dynamic nature of the biomarkers, nor-
mal findings may not remain normal over time. Thus, in asymptom-
atic individuals, the benefit of normal CSF and imaging biomarker
results would only materialize upon repeated assessment, which
appears to be an unlikely scenario. In the absence of any preventive
intervention, the benefits of biomarker determination may refer to
future planning and lifestyle modification. Positive biomarker find-
ings might promote providing wills and advance directives or affect
financial planning and decisions whether to continue working or
when and where to retire [43]. They may also stimulate people to
pursue a healthier lifestyle or engage in risk-lowering behaviors such
as enhanced cognitive or physical activity [16]. However, one could
take the position that these steps should be part of everyone’s plan
for their later years whether or not there is an increased likelihood
of cognitive decline [44].
3.4 What Are There are two kinds of risks associated with biomarker assessment.
the Risks Associated The first kind are minor adverse effects that may be associated with
with Biomarker the assessment procedure, i.e., the invasiveness of lumbar puncture
Information? for CSF measurements, and the radiation exposure associated with
PET scanning. The second kind of risk is related to the informa-
tion provided by the biomarker results. In the absence of preven-
tive or treatment options, this information is a potentially harmful
knowledge. It may cause psychological distress including anxiety,
depression or even suicidal ideation, disruption of family dynamics,
and worries about loss of or failure to obtain insurance or other
social or economical benefits such as employment [45, 46] or lead
to social stigmatization. Out of fear, presymptomatic persons may

also be more vulnerable to therapeutic misconception and might
falsely expect a therapeutic benefit of biomarker research [47].
Higher education has been found to be associated with consider-
ation of suicide in response to (nongenetic) biomarker information
[48]. Negative psychological outcomes are particularly likely if
people have the misconception that biomarker findings provide a
diagnosis and not only risk information, without predicting if or
when cognitive deterioration will occur. Other risks of biomarker
information include stigmatization and misuse of the information
by third parties.
3.5 Whether Asymptomatic individuals who are interested in an assessment of

and How Should AD biomarkers need to know that CSF indicators and β-amyloid
Biomarker Results imaging provide information about the presence of amyloid depo-
Be Disclosed? sition in the brain which is associated with an increased risk of
cognitive deterioration and ultimate development of dementia, but
is frequent in old age and occurs in several brain disorders such as
Lewy body disease, and thus does not establish a diagnosis [29]. In
like manner, elevated CSF levels of tau and increased tau PET
tracer take-up indicate an ongoing neurodegeneration but do not
distinguish between the different causes [49]. Applicants for a bio-
marker evaluation should also consider the dynamic nature of these
indicators. A negative or indetermined result at one time might
turn into a positive result a few years later. Moreover, because of
the individually variable threshold of clinical manifestation which
may be related to brain reserve, these biomarkers cannot tell
whether and when cognitive impairment will eventually occur even
in people where evidence for AD pathology is found. In contrast,
the rare genetic mutations are associated with an almost 100% risk
of developing the pathology and clinical manifestations of AD, but
there is also considerably interindividual variation of onset and
symptomatology [31, 50]. In light of these limitations, asymptom-
atic subjects need to weigh potential benefits and risks of obtaining
biomarker information. In research settings, disclosure of the apo-
lipoprotein E genotype [51] or of amyloid PET results provoked
little change in levels of depressive, anxiety, and stress symptoms.
Reported risk of self-harm [52] and catastrophic reactions follow-
ing such disclosure are rare [53]. However, these findings may not
apply to biomarker findings which have a much greater impact on
individual prognosis. An international working group recommends
not to disclose biomarker information to asymptomatic individu-
als, except when well-informed subjects request the information,
in cases of high social responsibility (e.g., pilots, bus drivers) or in
case of inclusion in research protocols or clinical trials [2].
Regarding predictive genetic testing, most persons at risk for a
dominantly inherited mutation do not want to know their risk
marker status [54], and very few actually choose to be tested [55].
Predictive genetic testing should include pre- and posttest counsel-

ing as well as baseline neurologic and neuropsychological or psy-
chiatric evaluation [56, 57].
3.6 What Will Novel treatments are currently being evaluated which address the
Change amyloid and tau components of AD. Examples are passive or active
with the Advent immunization which inhibits the aggregation of the two proteins
of Disease-Modifying and/or removes the proteins from the brain or enzyme blockers
Treatment? that reduce the production of the proteins. Expected efficacy is a
slowing of neurodegeneration, and the hope is that early use of
these treatments will delay the onset of cognitive impairment and
disability in daily living. Biomarkers will then be the gatekeepers to
treatment and determine the access to treatment. However, these
novel treatments are not without side effects and most probably
will not be tolerated by everyone. In addition, the indication (when
to start treatment, when to stop) and reimbursement by health
insurances is unknown to date. Furthermore, a proportion of peo-
ple who might be eligible for treatment may not wish to be treated,
balancing the benefit of postponing the onset of symptoms in an
uncertain future with the procedural burden and cost of treatment
[58]. It is therefore likely that only a minority of people who are at
risk, i.e., who have no symptoms but ongoing neurodegeneration,
will access these novel treatments. The majority of those tested
positively for the presence of AD pathology will not gain access to
disease-slowing treatment but will have the information of a fatal
brain disease. For them and their families, practical coping strate-
gies and concrete problem solutions need to be delineated [47].
4 Discussion
In the field of AD research, biomarkers can be viewed as reflections

of brain pathology. In asymptomatic individuals their major role is
to predict or exclude future cognitive decline and development of
dementia. However, the prognostic potential of current assess-
ments is low, and they must not be interpreted mechanistically.
People with positive biomarker findings may remain free of
symptoms for an extended period of time, because there is a wide
interindividual variability regarding the threshold of clinical mani-
festation. Negative biomarker results may convey false certainty,
because they gradually evolve over time, and findings may change
upon repeated measurement. To achieve a better estimation of
cognitive decline in asymptomatic individuals who are interested in
learning about their prognosis, the role of co-pathologies includ-
ing vascular abnormalities and depression, the contribution of risk
factors, but also the impact of protective factors such as brain
reserve [59] need to be better understood [24, 60, 61]. Since
there are no pharmacological interventions to date that would
slow the neurodegeneration and postpone the onset of cognitive

and functional impairment, the potential benefits of biomarker
assessment are limited to lifestyle modification. However, most
changes regarding healthier living, financial planning, or advance
directives should be part of everybody’s life plan [44, 48]. On the
other hand, biomarker assessments are associated with significant
risks, primarily regarding the potentially negative psychological,
social, and economical repercussions of the information provided.
The potentially negative effects beyond the emotional burden are
related to the impact on such issues as insurance and employment.
We assume that the advent of novel disease-modifying treatments
will change the risk-benefit ratio of biomarker assessment only for
a minority of asymptomatic individuals. Most probably, access to
these treatments will be limited, costs will be high, and many will
decide against participation in view of a significant procedural bur-
den, long treatment duration, or significant side effects.
Furthermore, there will be a large group of individuals who cannot
be treated because of contraindications or poor tolerance. They
will not have the benefit of the intervention but only the burden of
knowing their risk.
In summary, at this point in time, we do not recommend bio-
marker assessment and disclosure of biomarker results to asymp-
tomatic individuals outside of research protocols or clinical trials. If
in research contexts biomarker assessments are employed, partici-
pants need to be comprehensively informed about the nature of
the information they might receive and about the consequences
this information possibly has for their future planning, and they
should be given the choice of knowing or not knowing the bio-
marker results.
References
1. Vos SJB, Xiong C, Visser PJ et al (2013) 6. Witte MM, Foster NL, Fleisher AS et al
Preclinical Alzheimer’s disease and its out- (2015) Clinical use of amyloid-positron emis-
come: a longitudinal cohort study. Lancet sion tomography neuroimaging: practical and
Neurol 12:957–965 bioethical considerations. Alzheimers Dement
2. Dubois B, Hampel H, Feldman HH et al 1:358–367
(2016) Preclinical Alzheimer’s disease: defini- 7. Jack CR, Albert M, Knopman DS et al (2011)
tion, natural history, and diagnostic criteria. Introduction to revised criteria for the diag-
Alzheimers Dement 12:292–323 nosis of Alzheimer’s disease: national institute
3. Petersen RC (2011) Mild cognitive impair- on aging and the Alzheimer association work-
ment. N Engl J Med 364:2227–2234 groups. Alzheimers Dement 7:257–262
4. Ritchie K, Carrière I, Berr C et al (2016) The 8. Arlt S (2013) Non-Alzheimer’s disease–related
clinical picture of Alzheimer’s disease in the memory impairment and dementia. Dialogues
decade before diagnosis: clinical and biomarker Clin Neurosci 15:465–473
trajectories. J Clin Psychiatry 77:e305–e311 9. Villemagne VL, Burnham S, Bourgeat P et al
5. Roberts JS, Dunn LB, Rabinovici GD (2013) Amyloid beta deposition, and cogni-
(2013) Amyloid imaging, risk disclosure and tive decline in sporadic Alzheimer’s disease:
Alzheimer’s disease: ethical and practical issues. a prospective cohort study. Lancet Neurol
Neurodegener Dis Manag 3:219–229 12:357–367
10. Han SD, Gruhl J, Beckett L et al (2012) 23. Murray AD, Staff RT, McNeil CJ et al (2011)
Beta amyloid, tau, neuroimaging, and cogni- The balance between cognitive reserve and
tion: sequence modeling of biomarkers for brain imaging biomarkers of cerebrovascular and
Alzheimer’s disease. Brain Imaging Behav Alzheimer’s diseases. Brain 134:3687–3696
6:610–620 24. Almeida RP, Schultz SA, Austin BP et al (2015)
11. Jack CR, Knopman DS, Jagust WJ et al (2013) Cognitive reserve modifies age-related altera-
Tracking pathophysiological processes in tions in CSF biomarkers of Alzheimer’s disease.
Alzheimer’s disease: an updated hypothetical JAMA Neurol 72:699–706
model of dynamic biomarkers. Lancet Neurol 25. Doraiswamy PM, Sperling RA, Johnson K
12:207–216 et al (2014) Florbetapir F 18 amyloid PET
12. Sutphen CL, Jasielec MS, Shah AR et al (2015) and 36-month cognitive decline: a pro-
Longitudinal cerebrospinal fluid biomarker spective multicenter study. Mol Psychiatry
changes in preclinical Alzheimer disease during 19:1044–1051
middle age. JAMA Neurol 72:1029–1042 26. McKhann GM, Knopman DS, Chertkow
13. Wirth M, Villeneuve S, Haase CM et al (2013) H et al (2011) The diagnosis of dementia
Association between Alzheimer disease bio- due to Alzheimer’s disease: recommenda-
markers, neurodegeneration, and cognition in tions from the national institute on aging
cognitively normal older people. JAMA Neurol and the Alzheimer’s association workgroup.
70:1512–1519 Alzheimers Dement 7(3):263–269
14. Knopman DS, Jack CR, Lundt ES et al (2015) 27. Janelidze S, Zetterberg H, Mattsson N
Role of beta-amyloidosis and neurodegenera- et al (2016) CSF Abeta 42/Abeta 40 and
tion in subsequent changes in mild cognitive Abeta42/−abeta38 ratios: better diagnos-
impairment. JAMA Neurol 72:1475–1483 tic markers of Alzheimer’s disease. Ann Clin
15. Alexopoulos P, Werle L, Roesler J et al (2016) Transl Neurol 3:154–165
Conflicting cerebrospinal fluid biomarkers and 28. Counts SE, Ikonomovic MD, Mercao N
progression to dementia due to Alzheimer’s et al (2017) Biomarkers for the early detec-
disease. Alzheimers Res Ther 8:51 tion and progression of Alzheimer’s disease.
16. Porteri C, Frisoni GB (2014) Biomarker-based Neurotherapeutics 14:35–53
diagnosis of mild cognitive impairment due to 29. Chételat G, La Joie R, Villain N et al (2013)
Alzheimer’s disease: how and what to tell. A Amyloid imaging in cognitively normal indi-
kickstart to an ethical discussion. Front Aging viduals, at-risk populations and preclini-
Neurosci 6:41 cal Alzheimer’s disease. NeuroImage Clin
17. Landau SM, Lu M, Joshi AD et al (2013) 2:356–365
Comparing PET imaging and CSF measure- 30. Soldan A, Pettigrew C, Cai Q et al (2016)
ments of A beta. Ann Neurol 74:826–836 Hypothetical preclinical Alzheimer disease
18. Toledo JB, Weiner MW, Wolk DA et al (2014) groups and longitudinal cognitive change.
Neuronal injury biomarkers and prognosis in JAMA Neurol 73:698–705
ADNI subjects with normal cognition. Acta 31. Goldman JS (2012) New approaches to genetic
Neuropathol Commun 2:26 counseling and testing for Alzheimer’s disease
19. Leuzy A, Carter SF, Chiotis K et al (2015) and frontotemporal degeneration. Curr Neurol
Concordance and diagnostic accuracy of 11C Neurosci Rep 12:502–510
PIB PET and cerebrospinal fluid biomarkers in 32. Hauser PS, Ryan RO (2013) Impact of apo-
a sample of patients with mild cognitive impair- lipoprotein E on Alzheimer’s disease. Curr
ment and Alzheimer’s disease. J Alzheimers Alzheimer Res 10:809–817
Dis 45:1077–1088 33. Randall C, Mosconi L, de Leon M et al (2014)
20. Leuzy A, Chiotis K, Hasselbalch SG et al Cerebrospinal fluid biomarkers of Alzheimer’s
(2016) Pittsburgh compound B imaging and disease in cognitively healthy elderly. Front
cerebrospinal fluid amyloid-beta in a multi- Biosci 18:1150–1173
centre European memory clinic study. Brain 34. Fleisher AS, Chen K, Liu X et al (2013)
139:2540–2553 Apolipoprotein E e4 effects on florbetapir
21. Vos SJB, Gordon BA, Su Y et al (2016) positron emission tomography in healthy
NIAA-AA staging of preclinical Alzheimer dis- aging and Alzheimer disease. Neurobiol Aging
ease: discordance and concordance of CSF and 34:1–12
imaging biomarkers. Neurobiol Aging 44:1–8 35. Steenland K, Zhao L, Goldstein F et al (2014)
22. Valenzuela MJ (2008) Brain reserve and the Biomarkers for predicting cognitive decline in
prevention of dementia. Curr Opin Psychiatry those with normal cognition. J Alzheimers Dis
21:296–302 40:587–594
36. Johnson KA, Minoshima S, Bohnen NI et al 49. Villemagne VL, Fodero-Tavoletti MT, Masters
(2013) Appropriate use criteria for amy- CL et al (2015) Tau imaging: early progress and
loid PET: a report of the amyloid imaging future directions. Lancet Neurol 14:114–124
task force, the society of nuclear medicine 50. Tanzi RE (2012) The genetics of Alzheimer
and molecular imaging, and the Alzheimer’s disease. Cold Spring Harb Perspect Med
association. Alzheimers Dement 9: 2:a006296
e-1–e-15 51. Bemelmans SASA, Tromp K, Bunnik EM et al
37. Morris JC, Roe CM, Grant EA et al (2009) PIB (2016) Psychological, behavioral and social
imaging predicts progression from cognitively effects of disclosing Alzheimer’s disease bio-
normal to symptomatic Alzheimer’s disease. markers to research participants: a systematic
Arch Neurol 66:1469–1475 review. Alzheimers Res Ther 8:46
38. Villemagne VL, Pike KE, Chételat G et al 52. Lim YY, Maruff P, Getter C et al (2016)
(2011) Longitudinal assessment of Abeta and Disclosure of positron emission tomography
cognition in aging and Alzheimer disease. Ann amyloid imaging results: a preliminary study
Neurol 69:181–192 of safety and tolerability. Alzheimers Dement
39. Harkins K, Sankar P, Sperling R et al (2015) 12:454–458
Development of a process to disclose amyloid 53. Paulsen JS, Nance M, Kim JI et al (2013) A
imaging results to cognitively normal older review of quality of life after predictive testing
adult research participants. Alzheimers Res for and earlier identification of neurodegenera-
Ther 7:26 tive diseases. Prog Neurobiol 110:2–28
40. Rabinovici GD, Karlawish J, Knopman D et al 54. Hooper M, Grill JD, Rodriguez-Agudelo Y
(2016) Testing and disclosure related to amy- et al (2013) The impact of the availability of
loid imaging and Alzheimer’s disease: common prevention studies on the desire to undergo
questions and fact sheet summary. Alzheimers predictive testing in persons at-risk for autoso-
Dement 12:510–515 mal dominant Alzheimer’s disease. Contemp
41. Rowe CC, Bourgeat P, Ellis KA et al (2013) Clin Trials 36:256–262
Predicting Alzheimer disease with beta-amyloid 55. Steinbart EJ, Smith CO, Poorkay P et al (2001)
imaging: results from the Australian imaging, Impact of DNA testing for early-onset familial
biomarkers, and lifestyle study of ageing. Ann Alzheimer disease and frontotemporal demen-
Neurol 74:905–913 tia. Arch Neurol 58:1828–1831
42. Howe E (2013) Clinical implications of the 56. Cassidy MR, Roberts JS, Bird TD et al (2008)
new diagnostic guidelines for dementia. Innov Comparing test-specific distress of suscep-
Clin Neurosci 10:32–38 tibility versus deterministic genetic testing
43. Baum ML (2016) Patient requests for off-label for Alzheimer’s disease. Alzheimers Dement
bioprediction of dementia. Camb Q Healthc 4:406–413
Ethics 25:686–690 57. Goldman JS, Rademakers R, Huey ED et al
44. Gordon M (2013) Identification of potential (2011) An algorithm for genetic testing of
or preclinical cognitive impairment and the frontotemporal lobar degeneration. Neurology
implications of sophisticated screening with 76:476–483
biomarkers and cognitive testing: does it really 58. Molinuevo JL, Cami J, Carné S et al (2016)
matter? Biomed Res Int 2013:976130 Ethical challenges in preclinical Alzheimer’s
45. Karlawish J (2011) Addressing the ethi-
disease observational studies and trials: results
cal, policy, and social challenges of pre- of the Barcelona summit. Alzheimers Dement
clinical Alzheimer disease. Neurology 77: 12:614–622
1487–1493 59. Soldan A, Pettigrew C, Li S et al (2013)
46. Kim SYH, Karlawish J, Berkman BE (2015) Relationship of cognitive reserve and CSF bio-
Ethics of genetic and biomarker test disclosures markers to emergence of clinical symptoms
in neurodegenerative disease prevention trials. in preclinical Alzheimer’s disease. Neurobiol
Neurology 84:1488–1494 Aging 34:2827–2834
47. Schicktanz S, Schweda M, Ballenger JF et al 60. Donovan NJ, Hsu DC, Dagley AS et al (2015)
(2014) Before it is too late: professional respon- Depressive symptoms and biomarkers of
sibilities in late-onset Alzheimer’s research Alzheimer’s disease in cognitively normal older
and pre-symptomatic prediction. Front Hum adults. J Alzheimers Dis 46:63–73
Neurosci 8:921 61. Garrett MD, Valle R (2016) A methodologi-
48. Caselli RJ, Langbaum J, Marchant GE et al cal critique of the national institute of aging
(2014) Public perceptions of presymptomatic and Alzheimer’s association guidelines for
testing for Alzheimer’s disease. Mayo Clin Proc Alzheimer’s disease, dementia, and mild cogni-
89:1389–1396 tive impairments. Dementia 15:239–254
Chapter 18
Shared Decision-Making and Important Medical

and Social Decisions in the Context of Early
Diagnosis of Alzheimer’s Disease
Katharina Bronner and Johannes Hamann
Abstract
Shared decision-making is a model of medical decision-making, which aims at reducing the informa-
tional and decisional power asymmetry between physician and patient to obtain jointly reached
agreements on medical treatment. SDM can also be applied in the case of early stage of Alzheimer’s
disease, as long as individuals’ with dementia decisional capacity still exists. Because of the loss of
cognitive functions in the course of the disease, upcoming important issues like medical treatment
and legal and social issues should be made as soon as possible after diagnosis, so that persons with
dementia can be involved in the process of decision-making.
Key words Alzheimer’s disease, Social aspects, Early diagnosis, Dementia, Shared decision-making
1 Introduction
In recent years many efforts have been undertaken to improve

early diagnosis of Alzheimer’s disease (AD). This issue of AD has
raised awareness due to its high prevalence and incidence. In
Germany, for example, there are 1.6 million individuals with
dementia, thereof approximately 70% affected by Alzheimer’s dis-
ease, and 300,000 new cases a year. In addition, demographic
development with an increase of the elderly population will lead to
an increase of persons affected, i.e., approximately 3 million in
2050 in Germany [1] and 131 [2] to 135 [3] million worldwide.
AD causes great expenditures that range from approximately
15,000 Euro per individual and year in the case of mild dementia
to approximately 42,000 Euro in the case of severe dementia.
About 75% of total cost origins from costs of care and nursing [4].
Thus, on the one side, dementia burdens persons with AD and
their caregivers and families and on the other side society and health-
care systems, because enormous financial resources must be made
available [2, 3], which are limited in a healthcare system based on
259
260 Katharina Bronner and Johannes Hamann
social principles [5]. As a consequence persons with AD are threat-

ened by a shortage of resources because of their higher age, non-
existing curative therapy, and lacking lobby. Additionally, dementia
is rather stigmatized in the population because of the lacking chances
of recovery. In general population AD is the best known, most fre-
quent, and most feared neurodegenerative disorder [3]. In a survey
of a German health insurance company, 51% of respondents were
afraid of dementia [6].
Despite the discrepancy between already made progresses
regarding molecular genetics, laboratory, neuropsychological and
instrument-based diagnostics, and the still stagnant therapeutic
possibilities, it is recommended to make the diagnosis of AD as
soon as possible. One rationale behind this recommendation is that
diagnosis is the basis for therapy and thorough examination of per-
sons with cognitive decline may help detecting treatable reasons of
dementia symptoms, for example, depression [7].
The other rationale is founded in the progressive cognitive
decline throughout the course of AD. Thus, in later stages of AD
persons with AD inevitably loose decisional capacity, making it rea-
sonable to make important medical and social decisions early, as
long as persons with AD themselves still have the ability to make
them. These decisions include medical treatment, legal issues, and
coping with illness [8]. “Shared decision-making” may be a feasi-
ble approach to facilitate joint decision-making between persons
with dementia, their carers, and physicians [9].
2 Models of Medical Decision-Making
2.1 Four Models In their seminal paper, Emanuel and Emanuel [10] describe four
of Physician-Patient models of the doctor-patient relationship. In the paternalistic model
Relationship patients get selected information and instructions from the physi-
cian, which they are expected to follow. The physician makes deci-
sions about medical procedures like clinical diagnostics and therapy
all alone. And there is no explicit patient involvement in decision-
making. In the informative model the physician provides the patient
with all relevant specialized information without giving an advice.
The patient then comes to a decision all alone, which the physician
is expected to implement. In the interpretative model, the physician
provides the relevant information, inquires moral values and objec-
tives, and proposes appropriate measures of therapy. The patient
decides alone; however, the physician is involved in decision-
making. The deliberative model assumes that the physician discusses
the best treatment option with the patient. He knows about the
living conditions and also the moral values and objectives of the
patient. He tries to convince the patient to accept the best option,
but if the patient disagrees, the physician doesn’t try to overrule
him. Decisions are made by the patient and physician together.
Shared Decision-Making and Important Medical and Social Decisions 261
2.2 Shared Shared decision-making (SDM) has been located right in the middle
Decision-Making of the paternalistic and the informative model [11]. Already in 1997
four key features were postulated: (1) At least two persons are partici-
pating, physician and patient; (2) both sides share information with
each other; (3) both sides take steps to actively engage in a process of
decision-making; (4) both sides accept the decision and actively con-
tribute to its implementation [9]. Since the 1990s enormous political
efforts have been undertaken to foster research on and implementa-
tion of SDM in various healthcare systems [12–14].
Research has shown that SDM has the potential to improve
patients’ treatment satisfaction as well as their adherence with
therapies [15, 16].
3 Decision-Making and Decision-Making Capacity
3.1 Decision-Making In mental health, patient participation and SDM have not yet been
in Mental Illness fully implemented, although the majority of individuals with men-
tal illnesses express a clear wish for information and participation in
medical decisions and although there is growing evidence that
SDM may positively influence treatment outcomes [17]. Meanwhile
several clinical guidelines explicitly recommend patient participa-
tion in all medical decisions for the majority of mental illnesses,
such as depression or schizophrenia (e.g., National Institute for
Health and Care Excellence (NICE) [18, 19]).
However, in mental health it may be a special challenge to
engage individuals into medical decision-making, especially due to
the cognitive limitations, lack of insight, and reduced decisional
capacity that all may be a result of the mental illness [20].
3.2 Decision-Making The challenge of reduced decisional capacity is of special concern in

in AD AD, since AD is per definition a continually progressive disease with
loss of cognitive functions, and as individuals’ decision- making
capacity decreases, the more the disease is progressing [21].
However, individual’s decisional capacity for medical as well as
social decisions can still be present in early stages of AD [22]. This
pattern is also reflected in the guidelines of the National Institute
for Health and Care Excellence (NICE) that on the one side
emphasize the individual’s autonomy as a major aim but on the
other side judge the reduced cognitive capacity as a limiting factor
for participating [23]. Therefore, clinical guidelines do not demand
patient participation in decision-making for individuals with AD as
explicitly as for other mental illnesses. Nevertheless self-determina-
tion and autonomy in general are important issues for persons with
AD. In 2006 the US Alzheimer’s Association founded an initiative
with the objective of promoting public relation and lobby work for
persons with early AD [24]. In Europe, Alzheimer Europe and its
member organizations support autonomy and self-determination of
persons with AD [25].
3.3 Decision-Making Individuals in early stages of dementia are in most cases still capable
in Early-Stage of making decisions [26] and actually want to have a say in decisions
Dementia concerning their own treatment [27]. Thereby, participation pref-
erences of persons with dementia were especially marked in deci-
sions about social issues, for example, future housing or restitution
of the driver’s license. On the contrary, individuals with dementia
prefer to hand over decisions about medical issues to their physi-
cians. From the persons’ with dementia point of view, relatives
should play a minor part in medical as well as in social decisions
[28]. It can be assumed, however, that individuals’ participation in
decision-making does not take place, because psychiatrists tend to
exclude cognitively impaired individuals in decision-making [29],
i.e., when they are worried about the decrease of decisional capacity
in the progression of dementia [26, 30]. Additionally many persons
with AD are prevented from the process of decision-making,
because their relatives tend to take over control [27].
In most studies on decision-making in early-stage dementia,
more emphasis was placed on the relationship between individuals
with dementia and relatives and relatives and physicians than on
the issue of shared decision between individuals with dementia and
physicians [21, 31, 32].
3.4 Decision-Making For individuals at the stage of MCI and even more for individuals
at the Stage of MCI in a preclinical stage of AD, decisional capacity is in most cases not
and at the Preclinical a problematic issue in decision-making. Here it is rather a compli-
Stage of AD cated matter regarding the often debatable decisions (e.g., for or
against excessive diagnostic testing) that can or should be made at
these stages of the disease.
Thus, before dementia can be diagnosed, many individuals
pass through a period of mild cognitive impairment (MCI) [33].
This diagnosis has actually no therapeutic consequences, because
there is no medical indication for the use of the licensed antide-
mentives for prevention at the stage of MCI and at the preclinical
stage of AD, respectively [34, 35]. Moreover, there is an ongoing
debate about whether persons without any symptom of dementia
should participate in a predictive examination of AD by use of
diagnostic methods of early detection [36]. Although studies have
finally identified relevant cardiovascular risk factors for develop-
ment of a dementia (smoking, hypertension, obesity, diabetes mel-
litus, and physical inactivity) [37, 38] and there are hints for
protecting dementia by special nutrition [39] and sport activity
[40], there are no specific recommendations for risk reduction of
AD but only generally effective measures, which have a preventive
effect on other diseases, too (“what is good for the heart, is good
for the brain”) [41]. Predictive diagnostics can also be adverse for
individuals. Experiences with cancer screening have shown that
early detection is in no way without side effects [42]. Using cur-
rent methods there is, except the rare heritable form of AD [43],
no individual prediction of 100%, but only statements regarding

one’s individual risk for developing AD [44]. So there may be
false-positive or false-negative results [45] and therefore, because
of the lack of therapeutic options, a risk of a worsening of the qual-
ity of life by developing anxieties and depression.
The German Alzheimer Society therefore does not recom-
mend early detection in persons without any symptoms of demen-
tia. By explicit request of individuals, there is a need for a competent
medical consultation and a comprehensive education about the
consequences by well-informed specialists. The wish for participa-
tion in clinical trials to support dementia research is an additional
reason for an early detection [46].
4 Important Medical and Social Decisions After Early Diagnosis of Alzheimer’s

Disease
The diagnosis of early AD is a major turning point in the indi-

vidual’s and relative’s life and often challenges the future life of
whole families. Because of the progressive course with an inevi-
table loss of cognitive abilities including decisional capacity,
important decisions should be made as soon as possible after
diagnosis to ensure that persons with dementia can fully partici-
pate in decision-making, if they want to do so. For individuals
with AD, spouses, and physicians, a wide range of medical and
social decisions has been reported to be of importance [8]. Some
of these decisions refer to the current situation of the individual
with AD, while others rather deal with issues of the future. As
relatives have a disposition to take over more and more respon-
sibility and decisions, even if persons with AD still have deci-
sional capacity [21], the decisions mentioned below should
ideally be made within a joint decision process including persons
with AD, their carers, and physicians.
4.1 Medical Although there isn’t an effective therapy yet, which can cure or at least
Treatment stop the disease process, there are drugs with disease-delaying proper-
ties and drugs for accompanying symptoms such as depression, anxiety,
and aggression [47]. It is also possible for persons with AD to partici-
pate in clinical trials. In addition, psychosocial interventions are avail-
able, which can increase individuals’ with dementia and relatives’
well-being and quality of life.
4.1.1 Drug Treatment Currently there is no causal medical therapy for AD. There are
hints that early drug treatment may have positive effects on the
disease process [48]. Licensed antidementives are able to delay
dementia symptoms for all stages of AD [49] and also have an
effect on mental and behavioral symptoms [50] in some cases.
However, over the course of time, all individuals with AD will
develop severe dementia with the need for care, and, like all drugs,
antidementives may have side effects, especially on the gastrointes-
tinal system. On the part of the physicians, comprehensive infor-
mation should take place about the potential benefits and side
effects using a shared decision-making approach. In addition, the
benefit-risk profile should be observed carefully, especially when
prescribed for early stages of AD [7].
In addition to treatment with licensed drugs, persons with
dementia may participate in clinical trials, which are performed
at specialized and certificated centers. The European parent
organization Alzheimer Europe has published information and a
register of current clinical trials [51].
4.1.2 Psychosocial Besides drug treatment psychosocial interventions are part of persons’
Interventions with dementia and relatives’ support in dementia treatment. This
includes, for example, cognitive training, occupational therapy, physi-
cal activity, and art and sensory therapy. When making the decision
whether or not to accept support with psychosocial therapies, provid-
ers and individuals with dementia are faced with conflicting evidence
on these therapies because of the heterogeneous studies and partially
poor study designs [52]. The effects measured are often small and not
constant over several studies. Nevertheless, various psychosocial thera-
pies were recommended, because they have shown beneficial effects
[7, 23], for example, cognitive training for depression, cognition, and
quality of life [53]; occupational therapy for apathy, quality of life,
cognition [54], and functional decline [55]; and physical activity for
basic life skills [56]. In addition, support for relatives with the aim of
improving the individuals’ with dementia situation yields effects
regarding behavioral symptoms and depression [51]. Since mental
and physical health of the caregivers is often affected [57], structured
offers can help to prevent illnesses and reduce burden [58]. Persons
with dementia and relatives, which are integrated in dementia net-
works, are more frequently in contact with physicians and get more
frequently drug-free therapy and psychosocial interventions [59].
4.2 Legal Issues Over the course of the disease, individuals with dementia lose their
capacity to conclude legal transactions or to provide a declaration of
will. Therefore, all legal issues that could be of importance in the
early or later stages of AD should be discussed with persons with
dementia as long as they are definitely capable to do so. Written
powers and advance directives help persons of authority to make
arrangements in lieu of the individual with dementia in later stages
of the disease and avoid lengthy legal disputes. However, it is advis-
able to prove and certify the individual’s with dementia ability to
make a will even in early stages of the disease by a medical specialist,
mostly a psychiatrist.
4.2.1 Healthcare Proxy, After diagnosis of AD, it is the right time, according to experts, to
Last Will, Advance settle legal issues [8], for example, to arrange the legacy; the
Healthcare Directive, healthcare proxy, which designates a person for arranging personal
and Advance Care Planning and economic affairs, if AD is progressing; and the advance health-
care directive, which among others defines the extent of medical
procedures acceptable to the person with dementia. Because of the
increased accident risk of dementia, it may be advisable to contract
homeowner’s insurance or to request a disabled person’s pass,
which in some countries includes financial relief [60] and advanced
worker protection [61].
Again, last will, healthcare proxy, and advance healthcare directives
can only be done by persons who are still capable to consent and who
can sign personally.
In recent years the concept of advance care planning (ACP)
[62] has gained more and more attention. It addresses persons in
critical stages of illness and may remedy shortcomings of advance
healthcare directives, which persons often feel overburdened to
complete. Moreover, an advance healthcare directive is often
vaguely formulated; it may not be available at the critical moment
and therefore will be disregarded. ACP provides personal assis-
tance and support in filling in individual forms. In addition an
emergency guide is linked, which specifies how to act in a case of
emergency [63].
4.2.2 Car Driving Another important “legal” issue is car driving. Here the persons’
with dementia wish for autonomy may conflict with his or her abil-
ity to keep driving motor vehicles. Although AD at early stages is
not necessarily associated with the loss of capacity to drive [64],
reaction capacity and speed estimation can be affected. In some
European countries [65] there are obligatory regular medical
examinations at a certain age, which confirm or reject the ability to
drive. In other countries the responsibility is with persons con-
cerned, their relatives, or their physicians. Here individuals with
dementia should be informed by their physicians at the time of
diagnosis that AD leads to loss of the driving ability during the
course of the disease. In case of doubt, a medical-psychological
assessment can give some guidance in the decision-making process.
If AD progresses and individuals’ ability to drive is impaired, it may
be required that another person informs authorities to revoke the
persons’ with dementia driving license.
4.3 Social Issues and The progression of AD symptoms raises a lot of social issues.
Coping with Illness After diagnosis the question arises, how individuals with AD
and their relatives will deal with the illness. Many persons have
4.3.1 Coping with Illness
the desire to “carry on as always” [66] and not to further look
into AD. This also touches the issue of informing the social
environment about the disease, which individuals with demen-
tia want to impart or not. Many individuals with dementia and
relatives feel to have no reason to be concerned with future

planning, as long as dementia symptoms aren’t well pronounced
[8], because they aren’t fully aware of the importance of timely
decisions [67]. Even GPs have problems to communicate to
individuals with dementia the importance of future planning
because of the absence of decisional capacity in the later course
of the disease [68].
4.3.2 Assistance During the course of the disease, individuals with dementia lose
by Family Members more and more abilities and become dependent on support. In
and External Help most cases, relatives take over the additional assistance needed,
which is both expected and claimed from persons with dementia
and is taken for granted from relatives, who adapt their living con-
ditions for that. Both sides rather disapprove the possibility of
external help. In the course of AD, many caring relatives become
overburdened with this situation [8], have a worse quality of life,
risk their own health, and cannot ensure adequate care with the
best will in the world. Therefore professionals in dementia recom-
mend to plan external help and to extend the social network as
early as possible [69].
4.3.3 Living Conditions The issue of keeping up living conditions (especially housing condi-
tions) or changing them is also a difficult one, which from the point
of view of professionals in dementia is often repressed and ignored.
Most individuals with AD express the wish to stay in their familiar
environment as long as possible. Their relatives confirm this and
often promise that individuals shall stay at home even in the later
stages of AD [8]. During the course of AD, living at home may,
however, become more and more difficult, especially for persons
with dementia living alone without families or lack of assistance by
family. Therefore, persons with dementia and their relatives should
timely be informed about alternative living conditions such as
assisted community living or care homes for dementia.
5 Implementation of SDM in Early AD
To implement SDM for these decisions in the early course of AD,

several challenges have to be overcome: First, the model of SDM
has been shown to increase patient participation in decision-
making, but it does not clearly address decision-making of triads
(of patients, carers, and physicians) [70]. Second, the decisions to
be made are complex, and—in contrast to “classical” decisions
often described in decision aids—scientific evidence might neither
be available nor helpful in case of most decisions (e.g., housing
conditions). Third, some individuals with early AD might already
exhibit cognitive deficits, making it necessary to account for this
limitation. Finally, many individuals with dementia and their carers
tend to postpone decisions. Any implementation strategy will

therefore have to motivate them not to procrastinate.
We therefore suggest that a complex intervention consisting of
elements of SDM (e.g., physicians’ communicative competencies
for SDM [71], decision aids [72] (mainly as an aid to memory),
and advance care planning (i.e., a guide who helps patient-carer
dyads through the decision-making process)) might be suited to
the persons’ with AD interests. Due to the complexity and quan-
tity of decisions, a multi-session approach will be necessary. Thus,
with the help of a decision guide and a decision aid containing the
most important decision to be made, the different options and the
potential pros and cons patient-carer dyads may be prepared for
decisions or decisional talks with their physicians in charge.
6 Conclusions
Alzheimer’s disease is a progressive disease which is related to high

burden for the individuals with dementia, their relatives, and the
society. In the course of the disease, persons with AD lose their
decisional capacity and consequently the possibility to make deci-
sions. In the early stages of AD, decisional capacity is, however, still
present. Thus, shortly after diagnosis important decisions should
be made to allow participation of the persons affected as long as
they still have decisional capacity. However, many individuals with
dementia postpone these decisions not to burden themselves or
others. This often leads to decision-making in later stages of the
disease without the individual with dementia, and decisions might
therefore be not in line with the individuals’ initial preferences.
Shared decision-making (SDM) and advance care planning might
be suitable approaches to foster early and joint decision-making of
individuals with dementia, their carers, and physicians.
References
1. Deutsche Alzheimer Gesellschaft (2017) Die 5. BMBF (2016) Rahmenprogramm der

Häufigkeit der Demenzerkrankungen. https:// Gesundheitsforschung der Bundesregierung
www.deutsche-alzheimer.de/fileadmin/alz/pdf/ 6. Forsa (2013) .Angst vor Krankheiten. https://
factsheets/infoblatt1_haeufigkeit_demenzer- www.dak.de/dak/download/forsa-umfrage-
krankungen_dalzg.pdf. Accessed 25 April 2017 demenz-1331362.pdf. Accessed 26 April 2017
2. Prince MJ et al (2015) The world Alzheimer 7. Deutsche Gesellschaft für Psychiatrie,
report 2015, The global impact of dementia: Psychotherapie und Nervenheilkunde
an analysis of prevalence, incidence, cost and (DGPPN); Deutsche Gesellschaft für Neurologie
trends. https://www.alz.co.uk/research/ (Hrsg.) (2016) S3-Leitlinie “Demenzen”.
world-report-2015. Accessed 26 April 2017 http://www.dgppn.de/fileadmin/user_
3. WHO (2016) Dementia fact sheet April 2016. upload/_medien/download/pdf/kurzversion-
http://www.who.int/mediacentre/fact- leitlinien/S3-LL-D emenzen-240116-1.pdf.
sheets/fs362/en/. Accessed 26 April 2017 Accessed 27 April 2017
4. Leicht H et al (2011) Net costs of demen- 8. Bronner K et al (2016) Which medical and
tia by disease stage. Acta Psychiatr Scand social decision topics are important after early
124:384–395 diagnosis of Alzheimer’s disease from perspec-
tives of people with Alzheimer’s disease, spouses 22. Okonkwo OC et al (2008) Medical decision-
and professionals? BMC Res Notes 9:149 making capacity in mild cognitive impair-
9. Charles C et al (1997) Shared decision-making ment: a 3-year longitudinal study. Neurology
in the medical encounter: what does it mean? 71:1474–1480
(or it takes at least two to tango). Soc Sci 23. NICE (2006) Dementia: supporting people with
44:681–692 dementia and their carers in health and social
10. Emanuel EJ, Emanuel LL (1992) Four mod- care: NICE Clinical Guideline [CG42]. https://
els of the physician-patient relationship. JAMA www.nice.org.uk/guidance/cg42/resources/
267:2221–2226 dementia-supporting-people-with-dementia-
11. Charles C et al (1999) What do we mean by and-their-carers-in-health-and-social-care-
partnership in making decisions about treat- pdf-975443665093. Accessed 27 April 2017
ment? BMJ 319:780–782 24. Alzheimer’s Association Early-Stage Initiative
12. Coulter A et al (2017) Shared decision making (2006) History of the Alzheimer’s associa-
in the UK: moving towards wider uptake. Z Evid tion early-stage initiative. http://www.alz.
Fortbild Qual Gesundhwes 123- 124:99–103. org/documents/national/EDInitiative.pdf.
https://doi.org/10.1016/j.zefq.2017.05.010 Accessed 27 April 2017
13. Spatz ES et al (2017) Shared decision making 25. Alzheimer Europe (2006) Paris declaration
as part of value based care: new U.S. policies 2006. http://www.alzheimer-europe.org/
challenge our readiness. Z Evid Fortbild Qual Policy-in-Practice2/Paris-Declaration-2006.
Gesundhwes 123-124:104–108. https://doi. Accessed 27 April 2017
org/10.1016/j.zefq.2017.05.012 26. Moye J et al (2006) Neuropsychological pre-
14. Martin H et al (2017) The long way of imple- dictors of decision-making capacity over 9
menting patient-centered care and shared months in mild-to-moderate dementia. J Gen
decision-making in Germany. Z Evid Fortbild Intern Med 21:78–83
Qual Gesundhwes 123-124:46–51. https:// 27. Hirschman KB et al (2005) Do Alzheimer’s
doi.org/10.1016/j.zefq.2017.05.006 disease patients want to participate in a treat-
15. Stacey D et al (2017) Decision aids for people ment decision, and would their caregivers let
facing health treatment or screening decisions. them? Gerontologist 45:381–388
Cochrane Database Syst Rev 4:CD001431 28. Hamann J et al (2011) Patient participation in
16. Hamann J et al (2006) Partizipative medical and social decisions in Alzheimer’s dis-
Entscheidungsfindung. Implikationen des ease. J Am Geriatr Soc 59:2045–2052
Modells “Shared Decision Making” für 29. Hamann J et al (2009) Psychiatrists’ use of
Psychiatrie und Neurologie. Nervenarzt 9: shared decision making in the treatment of
1071–1078 schizophrenia: patient characteristics and deci-
17. Reichhart T et al (2008) Patientenbeteiligung sion topics. Psychiatr Serv 60:1107–1112
in der Psychiatrie–eine kritische 30. Vollmann J et al (2004) Mental compe-
Bestandsaufnahme. Psychiatr Prax 35:111–121 tence and neuropsychologic impairments in
18. NICE (2009) Depression in adults: rec- demented patients. Nervenarzt 75:29–35
ognition and management: NICE Clinical 31. Karlawish JHT et al (2000) Caregivers’ pref-
Guideline [CG90]. https://www.nice.org. erences for the treatment of patients with
uk/guidance/cg90/resources/depression- Alzheimer’s disease. Neurology 55:1008–1014
in-a dults-r ecognition-and-management- 32. Karlawish JHT et al (2002) Relationship
pdf-975742636741. Accessed 27 April 2017 between Alzheimer’s disease severity and patient
19. NICE (2014) Psychosis and schizophrenia in participation in decision about their medical
adults: prevention and management: NICE care. J Geriatr Psychiatry Neurol 15:68–72
Clinical Gudeline [CG178]. https://www.nice. 33. Peterson RC et al (2001) Current concepts
org.uk/guidance/cg178/resources/psychosis- in mild cognitive impairment. Arch Neurol
and-schizophrenia-in-adults-prevention-and- 58:1985–1992
management-pdf-35109758952133. Accessed 34. Raschetti R et al (2007) Cholinesterase inhibi-
27 April 2017 tors in mild cognitive impairment: a systematic
20. Hamann J, Heres S (2014) Adapting shared review of randomised trials. PLoS Med 4:e338
decision making for individuals with severe 35. DeKosky ST (2008) Ginkgo Biloba for pre-
mental illness. Psychiatr Serv 65:1483–1486 vention of dementia: a randomized controlled
21. Hirschman KB et al (2004) How does an trial. JAMA 300:2253–2262
Alzheimer’s disease patients’ role in medical 36. Rosen C et al (2013) Fluid biomarkers in
decision making change over time? J Geriatr Alzheimer’s disease–current concepts. Mol
Psychiatry Neurol 17:55–60 Neurodegener 8:20
37. Kivipelto M, Soloman A (2008) Alzheimer’s 51. Alzheimer Europe (2016) Clinical trials watch.
disease–the ways of prevention. J Nutr Health http://www.alzheimer-europe.org/Research/
Aging 12:89S–94S Clinical-Trials-Watch. Accessed 3 May 2017
38. Alonso A et al (2009) Cardiovascular risk fac- 52. Institut für Qualität und Wirtschaftlichkeit
tors and dementia mortality: 40 years of fol- im Gesundheitswesen (IQWiG) (2009)
low-up in the seven countries study. J Neurol Nichtmedikamentöse Behandlung der
Sci 280:79–83 Alzheimer Demenz. Abschlussbericht A05-
39. Feart C et al (2009) Adherence to a 19D(Version 1.0 Stand: 13.01.2009), Köln
Mediterranean diet, cognitive decline, and risk 53. Pinquart M, Forstmeier S (2012) Effects of
of dementia. JAMA 302:638–648 reminiscence interventions on psychosocial
40. Scarmeas N et al (2009) Physical activity, outcomes: a meta-analysis. Aging Ment Health
diet, and risk of Alzheimer disease. JAMA 16:541–558
302:627–637 54. Korczak D, Habermann C, Braz S (2013) The
41. Sindi S, Mangialasche F, Kivipelto M (2015) effectiveness of occupational therapy for per-
Advances in the prevention of Alzheimer’s dis- sons with moderate and severe dementia. GMS
ease. F1000Prime Rep 7:50. 10.12703/P7-50 Health Technol Assess 9:1–7
42. Marckmann G, in der Schmitten J (2014) 55. McLaren AN, Lamantia MA, Callahan CM
Krebsfrüherkennung aus Sicht der Public- (2013) Systematic review of non-pharmacologic
Health-Ethik. Bundesgesundheitsbl interventions to delay functional decline in
57:327. https://doi.org/10.1007/ community-dwelling patients with dementia.
s00103-013-1913-0 Aging Ment Health 17:655–666
43. Williamson J, Goldman J, Marder KS 56. Forbes D et al (2013) Exercise programs for
(2009) Genetic aspects of Alzheimer disease. people with dementia. Cochrane Database Syst
Neurologist 15:80–86 Rev 12:CD006489
44. Jessen F, Dodel R (2014) Prädiktion 57. Schölzel-Dorenbos CJ et al (2009) Quality of
der Alzheimer Demenz. Nervenarzt 85: life and burden of spouses of Alzheimer dis-
1233–1237 ease patients. Alzheimer Dis Assoc Disord
45. Quian W et al (2016) Misdiagnosis of 23:171–177
Alzheimer’s disease: inconsistencies between 58. Chien LY et al (2011) Caregiver support groups
clinical diagnosis. Alzheimers Dement 12:293 in patients with dementia: a meta-analysis. Int
46. Deutsche Alzheimer Gesellschaft (2013) J Geriatr Psychiatry 26:1089–1098
Empfehlungen zum Umgang mit 59. Wolf-Ostermann K et al (2017) Users
Frühdiagnostik bei Demenz. https://www. of regional dementia care networks in
deutsche-alzheimer.de/fileadmin/alz/pdf/ Germany: first results of the evaluation study
empfehlungen/empfehlungen_fruehdiagnos- DemNet-D. Z Gerontol Geriatr 50:21–27
tik_demenz_dalzg.pdf. Accessed 2 May 2017 60. Sozialgesetzbuch (SGB) Neuntes Buch
47. Robert PH et al (2005) Grouping for behav- (IX) - Rehabilitation und Teilhabe behinder-
ioral and psychological symptoms in dementia: ter Menschen - (Artikel 1 des Gesetzes
clinical and biological aspects. Consensus paper v. 19.6.2001, BGBl. I S. 1046) § 145
of the European Alzheimer disease consortium. Unentgeltliche Beförderung, Anspruch auf
Eur Psychiatry 20:490–496 Erstattung der Fahrgeldausfälle
48. Winblad B et al (2006) 3-year study of done- 61. Sozialgesetzbuch (SGB) Neuntes Buch
pezil therapy in Alzheimer’s disease: effects of (IX) - Rehabilitation und Teilhabe behinder-
early and continuous therapy. Dement Geriatr ter Menschen - (Artikel 1 des Gesetzes v.
Cogn Disord 21:353–363 19.6.2001, BGBl. I S. 1046) § 85 Erfordernis
49. NICE (2011) Donepezil, galantamine, riv- der Zustimmung
astigmine and memantine for the treat- 62. Emanuel LL, von Gunten CF, Ferris FD
ment of Alzheimer’s disease. Technology (2000) Advance care planning. Arch Fam Med
appraisal guidance [TA217]. https://www. 9:1181–1187
nice.org.uk/guidance/ta217/resources/ 63. Coors M, Jox R, in der Schmitten J (Hrsg)
donepezil-galantamine-rivastigmine-and- (2015) Advance care planning: Von der
memantine-for-the-treatment-of-alzheimers- Patientenverfügung zur gesundheitlichen
disease-pdf-82600254699973. Accessed 2 Vorausplanung. Kohlhammer, Stuttgart
May 2017 64. Wolter DK (2014) Beginnende Demenz und
50. McShane R, Areosa Sastre A, Minakaran N Fahreignung. Teil 2: Das Assessment und seine
(2006) Memantine for dementia. Cochrane praktischen Konsequenzen. Z Gerontol Geriat
Database Syst Rev 19:CD003154 47:345–355
65. Fastenmeier W, Gstalter H (2014) Fahreignung dementia patients: a focus group study on gen-
älterer Kraftfahrer im internationalen Vergleich. eral practitioners’ views and experiences. PLoS
Literaturrecherche, Analyse und Bewertung. One 9:e84905
Forschungsbericht Nr. 25. Unfallforschung der 69. Callahan CM et al (2009) Integrating care for
Versicherer, Berlin older adults with cognitive impairment. Curr
66. Von Kutzleben M et al (2012) Community- Alzheimer Res 6:368–374
dwelling persons with dementia: what do they 70. Morant N, Kaminskiy E, Ramon S (2016)
need? What do they demand? What do they Shared decision making for psychiatric medi-
do? A systematic review on the subjective expe- cation management: beyond the micro-social.
riences of persons with dementia. Aging Ment Health Expect 19:1002–1014
Health 16:378–390 71. Towle A, Godolphin W (1999) Framework for
67. Hirschman KB, Kapo JM, Karlawish JH (2008) teaching and learning informed shared decision
Identifying the factors that facilitate or hinder making. BMJ 319:766–771
advance planning by persons with dementia. 72. O'Connor AM et al (2009) Decision aids
Alzheimer Dis Assoc Disord 22:293–298 for people facing health treatment or screen-
68. DeVleminck A et al (2014) Barriers to advance ing decisions. Cochrane Database Syst Rev
care planning in cancer, heart failure and 8:CD001431
Index
A D
Aging��15, 16, 18–21, 23–26, 31–35, 45–47, Default mode network (DMN)�� 124, 129, 131, 132,
64, 77, 100, 119–121, 127, 129–133, 135, 138, 141, 180, 134–136, 139, 140
214, 226, 237, 243 Dementia��3, 15, 45, 55, 75, 93, 95, 107, 112,
Alpha linolenic acid (ALA)�� 18, 19 120, 149, 150, 152, 175, 179–181, 189, 193, 213, 214,
Alzheimer’s disease (AD)�� 3–5, 15, 71, 74–76, 236, 249, 250, 254, 259, 260
93, 107, 119, 151, 154, 157–164, 169, 179, 189–196, Dementia subtypes�� 95, 115
199–210, 213, 214, 235–245, 249, 259 Diagnosis�� 5, 16, 18, 55, 58, 72, 76, 79, 80, 87, 88,
Amyloid-beta��72–75, 82, 86, 87, 93, 149, 93, 100, 101, 115, 141, 150–152, 155, 179, 181, 185,
155, 180, 189, 190, 250–252, 254 235–239, 241, 242, 245, 249, 259
Amyloid precursor protein (APP)��73, 74, 76, Diet�� 15, 17, 240
88, 179–186, 238–239 Diffusion tensor imaging (DTI)�� 27, 114, 115
Apolipoprotein E (APOE)�� 16, 19–21, 24, 26, Disease classification��71–88
27, 120, 126, 127, 133–137, 139, 140, 152, 155, 236, Disease staging�� 190, 243
239–241, 252, 254 DNA��22
Drug development��264
B
E
Biomarkers��18, 26, 27, 31–33, 35, 46, 93, 95,
99–104, 114–115, 120–122, 128, 132, 134, 137, 138, 141, Early diagnosis��4, 151, 250, 259
154–155, 157, 169, 180–183, 185, 191, 200, 201, 207, Economic burden�� 4, 5, 9
208, 214, 216, 226–228, 236, 237, 242–245, 249, 251 Eicosapentaenoic acid (EPA)�� 18, 19,
Blood plasma�� 181, 183, 192 27, 31, 34
Blood serum�� 20, 185 Environment�� 16, 67, 99, 100, 265, 266
Blood vessel��199, 202–204, 208 Enzyme-linked immunosorbent assay (ELISA)�� 96, 97,
Brain��3, 45, 63–65, 67, 68, 71, 75, 77–80, 83, 99, 102, 169, 183–185
85, 86, 93, 100, 107–115, 119–141, 149–155, 160, 171, Epidemiology��15–36
180, 190, 199, 200, 214, 218, 236, 241, 250, 262 Ethics��171, 237, 249–256
Brain reserve�� 17, 254, 255
F
C
F-fluorodeoxyglucose�� 27, 29
18
Care costs��6, 7, 9, 259 Functional connectivity�� 66–68, 109, 123, 131,

Cerebral amyloid angiopathy (CAA)�� 32, 72, 75, 76, 132, 135, 136, 139, 157, 161, 162
87, 88, 121, 133, 138, 193 Functional magnetic resonance imaging (FMRI)�� 66,
Cerebrospinal fluid (CSF)�� 27, 93–104, 115, 154, 108, 109, 114, 122–124, 131–136, 138–140
169, 180, 237, 244, 245, 250, 252 Fundus�� 201–208, 210
Cholinergic basal forebrain��213–228
Clinical trial�� 31, 45, 48, 95, 101, 114, 180, G
236, 238–240, 242–244, 253, 254, 256, 263, 264 Genetics�� 17, 236, 260
Cognitive activity��17, 54–56, 63, 100, 253, 264
Cognitive assessment��18, 49, 52–57, 125, 253, 256 H
Cognitive reserve�� 17, 55, 63–69, 113, 125, 130, 141
Histology�� 77, 80, 81, 193, 215,
Cost-of-illness (COI)�� 4–9, 11
217–219, 224, 227
11
C-Pittsburgh compound B��131, 151, 153–155
Hyper-phosphorylated tau�� 16, 72, 74, 75, 83–85, 94
https://doi.org/10.1007/978-1-4939-7674-4, © Springer Science+Business Media, LLC 2018
271
Biomarkers for Preclinical Alzheimer’s Disease
272 Index

L Postmortem issue��191, 217, 241, 243
Predictive modeling�� 26, 110
Lifestyle�� 16, 17, 25, 33, 54, 58, 68, 141, Predictor�� 107, 115, 128, 180, 216
240, 252, 253, 256 Prevention�� 9, 16, 17, 22, 23, 26, 34, 35,
Luminex xMAP��98 114, 236, 240, 253, 262
Prognosis��100, 101, 254, 255
M
Protective factors�� 127, 255
Magnetic resonance imaging (MRI)��20, 27, 31–33, Proteomics��169–171
107–110, 112, 113, 115, 119, 121–125, 130, 132–140,
153, 154, 162, 180, 214–228, 237, 244, 245, 250, 252 Q
Mild cognitive impairment (MCI)��24, 45, 94, Quantitative electroencephalography��157
120, 136, 151, 179, 194, 201, 216, 238, 249, 262
Morphometry��107, 110, 114, 214, 221–223, 225 R
Multivariate models�� 26, 110
Retinal imaging��200
N Risk classification��114
Risk factors�� 9, 15, 57, 120, 126, 130, 133,
Network analysis��163 134, 138, 141, 155, 171, 239, 240, 245, 255, 262
Neuritic plaques (NP)��72, 75, 77, 81, 83, 85, 86
Neurodegeneration��16, 19, 27, 33, 34, 71–77, S
101, 119, 121, 122, 125, 126, 129, 130, 136–138, 179,
Shared decision-making��259–267
190–192, 208, 209, 214, 215, 226, 236, 237, 244, 245,
Single nucleotide polymorphism��16
249–251, 254–256
Social networks��266
Neuroinflammation�� 19, 191
Societal burden��3–11
Neuropathology�� 19, 20, 33, 35, 72–76,
Standardization��97, 102, 104,
121, 131, 193, 195
182, 228
Neuropsychology�� 46, 51–57, 100, 109,
Subjective cognitive impairment�� 46, 47
110, 115, 141, 255, 260
Substantia innominata�� 214, 215, 218
Non-invasive biomarker�� 125, 140
Support vector machine (SVM)��110
Nucleus basalis Meynert (NBM)��213, 216, 218,
Systems biology��171
220, 226
Nutrition�� 17–36, 262 T
P Tau�� 16, 72, 83–85, 94, 113, 120, 180,
189, 236, 243, 250
Pattern recognition�� 108, 111–115
TREM2�� 171, 172
Physical activity�� 17, 54, 240, 253, 264
Polyunsuturated omega-3 fats��18–20 V
Positron-emission-tomography (PET)�� 27, 33, 63, 115,
120–122, 125–133, 138–141, 149–154, 180, 190, 191, Validation��32, 53, 97, 104, 108, 111,
195, 216, 227, 228, 236, 237, 241–245, 250, 253, 254 113, 158, 169, 185, 238

Clinical Biomarker For Alzheimer

Uploaded by

Copyright:

Available Formats

Clinical Biomarker For Alzheimer

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Clinical Biomarker For Alzheimer

Uploaded by

Copyright:

Available Formats

What is the book about?

What is the book about?

What information is contained in the index?

What information is contained in the index?

Neuromethods 137

Robert Perneczky Editor

For further volumes:

ISSN 0893-2336 ISSN 1940-6045 (electronic)

Library of Congress Control Number: 2017964263

© Springer Science+Business Media, LLC 2018

Printed on acid-free paper

This Humana Press imprint is published by Springer Nature

Saskatoon, Canada Wolfgang Walz

London, UK Robert Perneczky

Part I Alzheimer’s Disease as a Societal Challenge

1 Worldwide Economic Costs and Societal Burden of Dementia . . . . . . . . . . . . . . . . 3

Part II Diagnostic Concepts

3 The Dimensional Structure of Subjective Cognitive Decline . . . . . . . . . . . . . . . . . . 45

Part III Established Markers of Pathophysiology and Risk

6 Cerebrospinal Fluid Biomarkers of Preclinical Alzheimer’s Disease . . . . . . . . . . . . . 93

Part IV Biomarker Discovery

11 A Selected Reaction Monitoring Protocol for the Measurement

12 Soluble Amyloid Precursor Proteins in Blood: Methods and Challenges . . . . . . . . . 179

Part V Patient Benefit and Ethical Considerations

Marina Ávila-Villanueva • Alzheimer Disease Research Unit, CIEN Foundation, Carlos

Johannes Hamann • Department of Psychiatry and Psychotherapy, Technische Universität

Chengxuan Qiu • Aging Research Center, Department of Neurobiology, Care Sciences

Alzheimer’s Disease as a Societal Challenge

Worldwide Economic Costs and Societal Burden

Dementia is defined as a clinical syndrome characterized by pro-

2 Dementia: A Priority of Global Public Health

Population aging is a worldwide universal phenomenon. Thus,

3 Estimation of Dementia Costs: Conceptual and Methodological Issues

The economic burden of a disease is often defined by cost-of-­illness

defining Alzheimer’s disease [6]. The severity of dementia and the

Fig. 1 Main cost components in the cost-of-illness studies of dementia

costs are incurred by outpatient visits or inpatient hospitalizations,

However, it is often difficult to tease out the cost of a certain illness

intrinsic value of life to estimate the average costs that an individual

4 Economic and Societal Burden of Dementia: A Brief Summary of Literature

We searched PubMed, Google Scholar, Web of Science, and China

First, methodological issues need to be reported in detail in future

Costs by categories (US$)

Perspectives of Direct Direct Indirect cost Total

Finally, the potential impact of secular trends in dementia

Chengxuan Qiu received research grants from the Swedish

Epidemiology and Risk Factors of Alzheimer’s Disease:

Dementia is the most frequent disease associated with brain aging.

accounting for approximately two thirds of cases. No curative

related alterations of the neurovasculature which constitutes the

on cross-sectional studies which assess exposure and outcome con-

2 Nutrition, Diet, and Alzheimer’s Disease Clinical Outcome

particular for the hippocampus, a key structure for memory which

[22, 23, 25, 26], others reporting association in APOEε4 noncar-

2.1.2 Antioxidant As n-3 PUFA, antioxidant nutrients (vitamins E and C, carot-

or cognitive decline. Most studies on the overall consumption of

Thus overall, epidemiological evidence for an association between

mandatory), an inverse association was found between higher

Mini-Mental State Examination score over 5 years [92]. The Dietary

Saskatoon, Canada Wolfgang Walz

London, UK Robert Perneczky

The economic burden of a disease is often defined by cost-of-illness

1.2 Neuro- Neuropathological assessment of the post-mortem brain is per-

2.5 Immuno- 1. Xylene.