Journal of Cardiac Failure Vol. 17 No.

7 2011

Improved Algorithm to Detect Fluid Accumulation via

Intrathoracic Impedance Monitoring in Heart Failure Patients
With Implantable Devices
SHANTANU SARKAR, PhD,1 DOUGLAS A. HETTRICK, PhD,1 JODI KOEHLER, MS,1 TYSON ROGERS, MS,1
YANINA GRINBERG, MS,1 CHEUK-MAN YU, MD, FRCP,2 WILLIAM T. ABRAHAM, MD,3
ROY SMALL, MD,4 AND W. H. WILSON TANG, MD5

Minneapolis, Minnesota; Hong Kong, China; Columbus and Cleveland, Ohio; Lancaster, Pennsylvania

ABSTRACT
Background: Intrathoracic impedance fluid monitoring has been shown to predict worsening congestive
heart failure (CHF) in patients with implantable devices. We developed and externally validated a modified
algorithm to identify worsening heart failure (HF) by using intrathoracic impedance.
Methods and Results: The modified algorithm was developed by using published data from 81 CHF sub-
jects averaging 259 days of follow-up. Device-measured daily impedance was input to both the existing
and the modified intrathoracic impedance fluid monitoring algorithms to determine a reference impedance
and a fluid index (FI). Separate validation sets included 326 cardiac resynchronization therapy device
(CRT-D) patients with an average 333 days of follow-up (group 1) and 104 CRT-D/implantable cardiovert-
er/defibrillator (ICD) patients followed for an average of 520 days (group 2). Clinicians and patients in
group 2 were blinded to impedance and FI data. HF events included adjudicated HF hospitalizations or
emergency room visits. Sensitivity was defined as the percentage of HF events preceded by FI exceeding
the predefined threshold (60 U-d) within the last 2 weeks. Unexplained detections were FI threshold cross-
ing events not followed by a HF event within 2 weeks. The modified algorithm significantly decreased
unexplained detections by 30% (P 5 .01; GEE) in the development set, 30% (P ! .001) in the group
1 validation set, and 43% (P ! .001) in group 2. Sensitivity did not change significantly in any group.
Simulated monthly review of FI threshold crossings identified subjects at significantly greater risk of wors-
ening HF within the next 30 days.
Conclusions: A modified intrathoracic impedance based fluid detection algorithm lowered the number of
unexplained FI threshold crossings and identified patients at significantly increased immediate risk of
worsening HF. (J Cardiac Fail 2011;17:569e576)
Key Words: Heart failure monitoring, implantable cardiac defibrillator, cardiac resynchronization therapy,
intrathoracic impedance.

Besides providing potentially life-saving therapies for

From the 1Medtronic, Minneapolis, Minnesota; 2Department of Medi-
cine and Therapeutics, Prince of Wales Hospital, Chinese University of tachyarrhythmias and pacing therapies for advanced heart
Hong Kong, China; 3Ohio State University, Columbus, Ohio; 4The Heart failure (HF), some modern implanted devices, including
Group, Lancaster, Pennsylvania and 5The Cleveland Clinic, Cleveland, implantable cardiac defibrillators (ICDs) and cardiac re-
Ohio.
Manuscript received September 21, 2010; revised manuscript received synchronization therapy devices (CRT-Ds), also provide
February 16, 2011; revised manuscript accepted March 2, 2011. important patient diagnostic monitoring capabilities. The
Reprint requests: Shantanu Sarkar, PhD, Medtronic, Inc, 8200 Coral Sea chronically monitored diagnostic parameters provided by
St NE, Mounds View, MN 55112. Tel: 763-526-0287; Fax: 763-526-5726.
E-mail: shantanu.sarkar@medtronic.com implantable devices may include heart rate (both day and
All decisions regarding this manuscript were made by a guest editor. night time), heart rate variability, patient daily activity,
See page 575 for disclosure information. and atrial and ventricular tachyarrhythmia frequency and
1071-9164/$ - see front matter
Ó 2011 Elsevier Inc. All rights reserved. duration.1 The diagnostic parameters are communicated
doi:10.1016/j.cardfail.2011.03.002 to the managing clinician through various clinical reports

569
570 Journal of Cardiac Failure Vol. 17 No. 7 July 2011

generated by the device.1 These detailed diagnostic reports a postimplantation reduction in impedance, presumably due to
contain serial data recorded over several months and may surgically related device pocket inflammation, followed by a sus-
be accessed directly through device programming modules tained period of gradually increasing and plateauing daily imped-
or remotely by using automatic or manual telemetric patient ance. Based on the trends in the daily impedance, the device also
calculates a daily ‘‘reference impedance’’ that represents the ex-
monitoring systems.1
pected trend in the daily impedance for the subject specific for
Recently, the monitoring of trends of intrathoracic im-
that particular day (Fig. 1). Sustained differences between the ref-
pedance measurement has been demonstrated to be a clini- erence impedance and the measured daily impedance are high-
cally useful tool for managing HF patients with implantable lighted on a separate trend known as the FI (Fig. 1). The
devices.2e7 To help interpret this data, an algorithm applied magnitude of the FI on a given day depends on both the magni-
to daily impedance measurements generates a ‘‘fluid in- tude of the difference between the reference and daily impedance
dex’’ (FI) that can surpass a preprogrammed threshold to and the duration of the reduced impedance event. The FI is auto-
potentially indicate worsening HF or acute decompensa- matically reset to a value of 0 when the daily impedance is sus-
tion. Two prospective double-blinded clinical trials have in- tained above the reference impedance. Besides the FI,
vestigated the relative sensitivity of FI threshold crossings a threshold, which is also displayed, can be manually pro-
to HF hospitalizations at the nominal FI threshold value grammed to identify values of the FI of particular concern for
an individual patient. FI values in excess of this threshold can re-
of 60 U-d. The Medtronic Impedance Diagnostics in Heart
motely generate a written ‘‘CareAlert’’ on the printed diagnostic
Failure Patients Trial (MIDHEFT) reported a sensitivity of
report and an audible patient alert (for devices implanted outside
77% with an unexplained event rate of 1.5 per patient per of the United States) to alert clinicians to potentially important
year.2 More recently, the Fluid Accumulation Status Trial fluid accumulation events.
(FAST) reported a similar patient adjusted sensitivity
(76%) and unexplained event rate of 1.9 per patient per
Development Data Set
year.8 Furthermore, 3 recent trials have shown that all FI
threshold crossings, not just those directly associated with The modified algorithm was developed by using published
HF hospitalization, identify patients at significantly higher data from 2 double-blinded clinical trials including 81 HF sub-
prospective risk of HF hospitalization.3,5,7 jects averaging 259 6 109 days of follow-up and 18 adjudicated
The observation that the existing algorithm may result in HF events.2,8 Both clinicians and patients in these trials were
unexplained threshold crossings remains one of the biggest blinded to the daily impedance, reference impedance, and FI
challenges limiting broad acceptance of this technology, in- data. From this development set, key features of the existing al-
gorithm that contributed to the occurrence of unexplained thresh-
cluding the adoption of alert implementation regarding
old crossings were identified. Modifications to the algorithm
threshold crossing. In an observational study, Vollmann
were made with the goal of decreasing the overall number of
et al4 reported that 60% of patients clinically presenting FI threshold crossings while simultaneously preserving the clin-
shortly after an audible patient alert demonstrated observ- ical utility of the FI.
able signs and symptoms of worsening HF. Likewise, The modified algorithm pertains only to the proprietary algo-
60% of all clinical events associated with HF were pre- rithms for calculation of the reference impedance and the FI.
ceded by FI threshold crossings. The modified algorithm pertained to 3 fundamental algorithmic
The present study was conducted with the goal to de- changes in the computation of the FI, identified from exploratory
velop and test a modified algorithm to determine the FI review of device data not included in the present validation or de-
based on continuous intrathoracic daily impedance mea- velopment sets. First, the behavior of the reference impedance
surements. The goal was to decrease the overall number was modified to transiently adjust more readily to expected in-
creases in daily impedance occurring within the first several
of FI threshold crossings while simultaneously preserving
months after device implantation. Second, a variability algorithm
the clinical utility of the FI. The modified algorithm was
was developed to decrease the probability of an FI threshold
developed by using data from several clinical data sources crossing in response to naturally occurring variability in the daily
and validated on 2 additional and separate prospectively impedance measurement. By this variability algorithm, higher
collected clinical trial data sets. day-to-day variability in the daily impedance measurement results
in relatively lower increments in the FI by using a fraction of the
Methods differences in the daily impedance and reference impedance. The
fractional increment is also dependent on the duration of the FI
Measurement of Daily Impedance and Derivation
of the FI event. Third, the behavior of the FI for sustained threshold cross-
ing events was adjusted algorithmically to allow the FI to de-
The intrathoracic impedance-based fluid monitoring diagnostic crease, without resetting, when the daily impedance consistently
feature (Optivol; Medtronic, Minneapolis, Minnesota) has been increases toward the reference impedance. Note that no changes
previously described in detail.2 Briefly, the feature operates by au- were made to the measurement of the daily impedance, or to
tomatically performing multiple intermittent impedance measure- the nominal value of the threshold. These changes are further
ments within a defined time period between the right ventricular summarized in Table 1. To facilitate the development process,
lead and the metallic case of the implantable device. The imped- raw device-measured daily impedance was input to both the exist-
ance measurements are combined to determine the average daily ing and the modified intrathoracic impedance fluid monitoring al-
impedance (Fig. 1). The daily impedance typically displays a pre- gorithms to determine the reference impedance and an FI using
dictable pattern in the months after implantation, characterized by a custom computer simulator.
 HF Monitoring Using Intrathoracic Impedance
Sarkar et al 571

Fig. 1. Example comparison of original and modified fluid index calculation for an individual subject. The lower graph shows the measured
daily impedance and the calculated reference (or ‘‘expected’’) impedance over a 14-month period. The daily impedance measurement is
identical for both the original and the modified algorithms. However, the calculation of both the reference impedance and the fluid index
in response to observed changes in the daily impedance has been modified. The resultant fluid indexes for both algorithms, along with the
nominal fluid index threshold (60 U-d), are shown in the upper graph. Note that the number of threshold crossings has decreased from 4 to 2
over the 14-month period analyzed for this subject.

The new algorithm for calculating the reference impedance and However, the impedance and FI data was blinded to both clini-
deriving the FI from the accumulated difference between the refer- cians and subjects in group 2.8
ence impedance and the daily impedance was then developed by
using HF events within this development set to define the modified Statistical Analyses
algorithm. Calculation of the reference impedance was altered to
Demographic data between the development and validation co-
more closely track transient changes in impedance, especially
horts were compared via t test for age and chi-square analysis for
those occurring in the first several months after implantation.
categoric variables. Because the modified algorithm did not alter
Also, the calculations determining the FI were modified to better
how the device measures daily impedance, the resultant FI could
respond to changing impedance associated with HF events.
be compared directly between the modified and original algo-
rithms for each subject and each HF event. HF events were defined
Validation Data Sets specifically for the development and validation trial cohorts as ad-
judicated hospitalizations or emergency room visits primarily at-
The FI for both the original and the modified algorithms was tributed to HF.2,3,8 HF events were considered to be distinct if
calculated from the measured daily impedance in 2 additional separated by $7 days. Sensitivity was defined as the percentage
and separate validation cohorts from previously reported trials.3,8 of HF events preceded by an FI value exceeding threshold within
Validation cohort subjects were required to have a minimum of 60 2 weeks. Unexplained detections (ie, false positives) were FI
days of stored device follow-up data. The first validation set threshold crossing events not followed by a HF event within
(group 1) included 326 CRT-D patients with a mean 333 6 96 2 weeks. Differences in sensitivity and unexplained detection rates
days of follow-up and 44 total adjudicated HF events.3 The second between the original and modified algorithm were compared by
validation set cohort (group 2) included 109 CRT-D/ICD subjects. using the generalized estimating equation.
Three group 2 subjects were excluded because of missing follow- To determine the relative ability of the modified algorithm to
up data, and the remaining subjects were followed for a mean prospectively identify patients at near-term risk of worsening HF
520 6 264 days with a total of 52 adjudicated HF events. Two ad- and to evaluate the potential clinical utility of regular monthly re-
ditional group 2 subjects were excluded from hazard ratio analysis mote monitoring of the FI, evaluations for the relative risk were
for insufficient follow-up duration. In group 1, clinicians had ac- simulated at the nominal threshold of 60 U-d for both algorithms.
cess to the impedance data and the original derived FI and were These simulations were performed for consecutive 30-day periods,
able to use device diagnostic data to make clinical decisions.3 including a retrospective 30-day diagnostic evaluation period (to
572 Journal of Cardiac Failure Vol. 17 No. 7 July 2011

Table 1. Summary of Changes Between Original and Modified Algorithms

Algorithmic Change Description Illustration Rationale

Transiently increase sensitivity of

Improve sensitivity after
calculated reference impedance to
1. Reference impedance calculation reference impedance
changes in the daily impedance
initialization
after initialization

Transiently attenuate accumulation

of the fluid index in patients
2. Fluid index calculation Reduce false detections
with higher day-to-day variability
in daily impedance

Accumulation of the fluid index is Allows the fluid index to decrease

3. Fluid index temporal accumulation limited to a finite time period without resetting, to reflect
limit during sustained threshold improving fluid status in response
crossing events to increasing daily impedance

assess the subject’s HF risk based on the diagnostic trends) and the FI are unchanged with the modified algorithm, but
a 30-day prospective risk evaluation period (to observe the first with the modified algorithm, threshold crossings decreased
HF event occurrence) as previously described.7 Univariate and from 4 to 2.
multivariate Cox regression analyses were then performed to de- The modified algorithm significantly decreased unex-
termine a hazard ratio (ie, the 30-day prospective relative risk of
plained detections by 30% (P 5 .01) in the development
an HF event based on the occurrence of an FI threshold crossing
set, 30% (P ! .001) in the group 1 validation set, and
within the preceding 30 days).
43% (P ! .001) in group 2 (Fig. 2). Although the point es-
timate of sensitivity increased by 27% in the development
Results set and by 54% in the group 1 validation set, and decreased
in the group 2 validation set by 10%, these changes were
Demographic data for the development and group 1 and not statistically significant. Unexplained crossings were sig-
group 2 validation cohorts are presented in Table 2. Most nificantly (P ! .001) lower at every programmable thresh-
patients were New York Heart Association (NYHA) func- old in the blinded validation set (Fig. 3). In contrast, no
tional class III and receiving beta-blocker, angiotensin- significant difference in sensitivity was observed at any
converting enzyme inhibitor (ACEi), and loop diuretic threshold.
therapy. Figure 1 demonstrates the differences in the de- The Cox univariate regression analysis concluded that
rived reference impedance and FI resulting from the identi- the FI threshold crossings in the preceding 30 days identi-
cal actual measured daily impedance for 1 subject in the fied patients at significantly greater risk of an HF event
group 1 validation set. Note that the daily impedance and within the following 30 days for both the original and the
 HF Monitoring Using Intrathoracic Impedance
Sarkar et al 573

Table 2. Patient Demographics for Development and [original: HR 4.57 (1.97e10.63), P ! .001; modified:
Validation Cohorts HR 3.91 (1.64e9.32), P 5 .002]. The multivariate analysis
Validation adjusted for age, gender, NYHA I/II vs III/IV, coronary ar-
tery disease, ACEi/angiotensin receptor blocker, beta-
Development Nonblinded Blinded blocker, diuretic, and antiarrhythmic drug usage.
n 5 80* n 5 326 n 5 109
Discussion
Male 58 (73%) 244 (75%) 83 (76%)
Age, y 72 6 11 70 6 11 70 6 11
Cardiovascular disease history The key finding of the present analysis was that the mod-
Coronary artery disease 47 (59%) 200 (61%) 79 (73%)y ified algorithm enhanced diagnostic accuracy by reducing
Chronic obstructive 4 (9%) 50 (15%) 10 (9%) unexplained threshold crossings of an intrathoracic
pulmonary disease
Pharmacologic therapy impedance-derived HF FI. Key features were identified in
b-Blocker 57 (71%) 290 (89%)y 94 (86%)y the existing algorithm that contributed to the occurrence
ACEi/ARB 67 (84%) 259 (79%) 89 (82%) of unexplained threshold crossings. The present results
Antiarrhythmic agents 11 (14%) 94 (29%)y 28 (26%)y
(class III) demonstrated statistically improved algorithm performance
Calcium channel blockers 8 (10%) 23 (7%) 7 (6%) in 2 independent data sets. Such enhancements are vital to
Diuretics 74 (93%) 267 (82%)y 89 (82%)y the adoption and implementation of an alert system with the
y y
NYHA functional class
I 13 (16%) 2 (1%) 9 (8%) hope to improve management and risk stratification of HF
II 27 (34%) 49 (18%) 67 (62%) patients.
III 38 (48%) 219 (78%) 33 (30%) The original algorithm to determine the FI based on
IV 2 (2%) 9 (3%) 0 (0.0%)
changes in the measured daily intrathoracic impedance
ACEi, angiotensin-converting enzyme inhibitor; ARB, angiotensin re- was entirely developed and initially validated by using
ceptor blocker; NYHA, New York Heart Association. data from the previously published MIDHEFT.2 That trial
*Demographic data available for 80 of 81 subjects in the development
cohort. included 35 enrolled subjects and reported a sensitivity of
y
P ! .05 versus development. 77% and an unexplained detection rate of 1.5 events per pa-
tient per year. The recently reported results of the FAST8
modified algorithm (Fig. 4). Similar results were observed confirmed this sensitivity (76%) in a population of 156 sub-
for the multivariate regression analysis for the nonblinded jects, but reported 1.9 unexplained detections per patient-
validation set [original: hazard ratio (HR) 2.83 (95% confi- year. Although other trials have reported sensitivity values,
dence interval 1.21e6.64), P 5 .02; modified: HR 2.94 the aforementioned trials stand out in that the data were col-
(1.18e7.36), P 5 .02] and the blinded validation set lected in a blinded fashion over a long period of time after
implantation. For example, Vollmann et al4 reported 60%
sensitivity and 60% positive predictive value by using a dif-
ferent definition of sensitivity in a nonblinded study design
evaluating the clinical utility of an audible patient alert. De-
spite these relatively high reported sensitivities, the obser-
vation that many threshold crossings do not, in fact,
precede actual HF events provided motivation to leverage
newly available clinical data to try to improve the overall
algorithm performance. The modified algorithm was devel-
oped on previously reported datasets including over 57
patient-years of intrathoracic impedance monitoring. The
algorithm was then prospectively validated on 2 additional
independent data sets, including 1 double-blinded data set,
accounting for a combined 151 patient-years of intratho-
racic impedance monitoring and event tracking.
The present results demonstrated that a novel modified
algorithm to derive an HF FI parameter from sequential de-
vice measurements of intrathoracic impedance was able to
reduce the rate of unexplained events. Furthermore, these
reductions were observed separately in both a nonblinded
and blinded validation data set. These performance en-
hancements were achieved without making any change to
Fig. 2. Observed decreases in the average rate of unexplained de- the measurement of the raw daily impedance and without
tections in the development set as well as both the nonblinded meaningfully altering sensitivity to HF hospitalizations.
(group 1) and the blinded (group 2) validation sets. All decreases Likewise, FI threshold crossings for the modified algorithm
were statistically significant. identified patients at greater prospective risk for worsening
574 Journal of Cardiac Failure Vol. 17 No. 7 July 2011

and symptoms of worsening HF (not requiring hospital ad-

mission),4 dietary or pharmacologic nonadherence,9 acute
onset of atrial tachyarrhythmias10 or ventricular tachyar-
rhythmias,11 acute anemia, pneumonia, pericardial or pleu-
ral effusion, and hospital procedures including surgical
revision of the implanted system.12 Thus, the term ‘‘false
positive’’ may be somewhat of a misnomer, and such events
are referred to as unexplained detections in the present
analysis. However, despite the potential clinical utility of
specific events identified by unexplained FI threshold cross-
ings, most of these events are not emergent and therefore do
not require immediate patient or clinical notification.
Therefore, the modified algorithm seems to have the ad-
vantage of lowering the overall alert burden while simulta-
neously maintaining sensitivity to the most severe clinical
events, such as HF hospitalization. In addition, the results
of the present monthly risk analysis (Fig. 4) show that FI
threshold crossings for both the original and the modified
algorithms identify subjects with significantly increased
30-day prospective risk of an HF event.

Alert Versus No Alert

Two distinct clinical models for the application of

device-derived HF diagnostics have evolved since the fea-
ture became clinically available in 2004. In regions where
the audible alert feature is available, it may be applied to
try to predict an immediately impending HF hospitaliza-
tion. In contrast, in regions where no audible alert is avail-
able, intrathoracic impedance and other device-derived
diagnostic variables may be reviewed at regular routine in-
tervals (eg, monthly), either in the office or remotely via te-
lemetry, to identify patients at increased risk of worsening
HF.13,14 Identified subjects may then be counseled and fur-
Fig. 3. Differences in (A) unexplained detections and (B) sensitiv-
ther interviewed or examined to determine the precise cause
ity between the original and modified algorithm for all fluid index
thresholds in the blinded validation set (group 2). The rate of un-
of the observed diagnostic changes, as well as the appropri-
explained detections was lower for the modified algorithm com- ate clinical action. Within this ‘‘routine review’’ frame-
pared with the original algorithm at every threshold value. work, some diagnostic events may be identified only long
after an actual FI threshold crossing has occurred. Rather
than identifying imminent hospitalization, device diagnos-
HF. These improvements have the potential to improve re- tics are used within this framework to stratify risk and to
mote and in-clinic monitoring of HF patients with implant- identify potential issues with therapy nonadherence or other
able devices. causes. The present and previous3,5,7 trial results showing
risk stratification support this remote monitoring frame-
False Positives work as a viable alterative to the identification of imminent
hospitalization (Fig. 4). Therefore, improved performance
The goal of the modified intrathoracic impedance FI al-
of the FI may have significant clinical benefit to devices
gorithm was to reduce the number of FI threshold crossings
both with and without audible alert.
not correlated with immediate HF hospitalization. Previ-
ously, such events have been referred to as false positives.2 Limitations
However, several recent trials examining the clinical utility
of intrathoracic impedance monitoring have determined These results should be interpreted within the context of
that all threshold crossing events identify patients with sig- inherent limitations. The group 1 trial data was collected in
nificantly greater prospective risk of HF hospitalization.3,5,7 a retrospective fashion, and the impedance data was avail-
More frequent crossings, and longer-duration crossings, able to the treating clinicians during the evaluation period.3
compound the risk. In addition, threshold crossings are fre- In contrast, the group 2 data were collected prospectively in
quently associated with other clinically relevant comorbid a double-blinded trial design.8 Despite these design differ-
events. Some of these other events include less severe signs ences, both validation trials were designed specifically to
 HF Monitoring Using Intrathoracic Impedance
Sarkar et al 575

original algorithm, simulated monthly review of FI thresh-

old crossings identified subjects at significantly greater risk
of worsening HF within the next 30 days. Thus, the new al-
gorithm may improve the clinical utility of both in office
and remote monitoring of HF via implantable devices.

Disclosures

Drs Sarkar and Hettrick, Ms Koehler, Mr Rogers, and

Ms Grinberg are full-time employees of Medtronic. Profes-
sor Yu, and Drs Small, Abraham, and Tang have received
honoraria and consulting fees from Medtronic. The clinical
trials contributing data for the present investigation were
sponsored by Medtronic.

References

1. Santini M, Ricci RP, Lunati M, Landolina M, Perego GB,

Marzegalli M, et al. Remote monitoring of patients with biventricular
defibrillators through the CareLink system improves clinical manage-
ment of arrhythmias and heart failure episodes. J Interv Card Electro-
physiol 2009;24:53e61.
2. Yu CM, Wang L, Chau E, Chan RH, Kong SL, Tang MO, et al. Intra-
thoracic impedance monitoring in patients with heart failure: correla-
tion with fluid status and feasibility of early warning preceding
hospitalization. Circulation 2005;112:841e8.
3. Small R, Wickemeyer W, Germany R, Hoppe B, Andriulli J, Brady P,
et al. Changes in intrathoracic impedance are associated with subse-
quent risk of hospitalizations for acute decompensated heart failure:
clinical utility of implanted device monitoring without a patient alert.
Fig. 4. Subjects with threshold crossing in preceding 30 days had J Card Fail 2009;15:475e81.
significantly increased risk of HF event within following 30 days 4. Vollmann D, N€agele H, Schauerte P, Wiegand U, Butter C, Zanotto G,
for both algorithms in both (A) the development and the (B) non- et al., European InSync Sentry Observational Study Investigators.
blinded and (C) blinded validation data sets. The hazard ratios Clinical utility of intrathoracic impedance monitoring to alert patients
were statistically significant for both the original and the modified with an implanted device of deteriorating chronic heart failure. Eur
Heart J 2007;28:1835e40.
algorithms.
5. Perego GB, Landolina M, Vergara G, Lunati M, Zanotto G,
Pappone A, et al, Optivol-CRT Clinical Service Observational Group.
quantify the performance of the intrathoracic impedance- Implantable CRT device diagnostics identify patients with increased
risk for heart failure hospitalization. J Interv Card Electrophysiol
derived FI. Because of the distinct nature of the validation 2008;23:235e42.
clinical trial designs, we did not combine the validation 6. Catanzariti D, Lunati M, Landolina M, Zanotto G, Lonardi G,
data sets. Rather, each data set was treated independently. Iacopino S, et al., Italian Clinical Service Optivol-CRT Group. Mon-
Furthermore, both trials showed similar decreases in the itoring intrathoracic impedance with an implantable defibrillator re-
rate of unexplained detections (Fig. 2). Although no signif- duces hospitalizations in patients with heart failure. Pacing Clin
Electrophysiol 2009;32:363e70.
icant difference in sensitivity was observed between algo- 7. Whellan DJ, Ousdigian KT, Al-Khatib SM, Pu W, Sarkar S, Porter CB,
rithms, the present analysis was underpowered to et al., PARTNERS Study Investigators. Combined heart failure device
specifically test for noninferiority, because the calculation diagnostics identify patients at higher risk of subsequent heart failure
of sensitivity may only be performed on the minority of pa- hospitalizations: results from PARTNERS HF (Program to Access and
tients with HF hospitalizations. Review Trending Information and Evaluate Correlation to Symptoms
in Patients With Heart Failure) study. J Am Coll Cardiol 2010;55:
1803e10.
8. Abraham WT, Compton S, Haas G, Foreman B, Canby RC, Fishel R,
Conclusions et al. Superior performance of intrathoracic impedance-derived fluid
index versus daily weight monitoring in heart failure patients: results
Compared with the currently available algorithm, a mod- of the Fluid Accumulation Status Trial (FAST). Congestive Heart Fail
ified intrathoracic impedance-based fluid detection algo- 2011 2011; [In Press].
9. Rathman L, Repoley J, Small R, Dukes Graves D, Trynowsky K,
rithm lowered the number of FI threshold crossings Delgado SS, et al. Adherence by interrogation: heart failure patients
uncorrelated with imminent HF hospitalization in both an with lower treatment adherence emonstrate frequent decreases in in-
unblinded and a blinded validation data set. Similar to the trathoracic impedance. [abstract]. J Card Fail 2007;13:S184.
576 Journal of Cardiac Failure Vol. 17 No. 7 July 2011

10. Jhanjee R, Templeton GA, Sattiraju S, Nguyen J, Sakaguchi S, Lu F, impedance-based indices of worsening heart failure. [abstract].
et al. Relationship of paroxysmal atrial tachyarrhythmias to volume J Card Fail 2006;6:S66.
overload: assessment by implanted transpulmonary impedance moni- 13. Mullens M, Oliveira LPJ, Verga T, Wilkoff BL, Tang WHW. Insights
toring. Circ Arrhythm Electrophysiol 2009;2:488e94. from internet-based remote intrathoracic impedance monitoring as
11. Moore HJ, Peters MN, Franz MR, Karasik PE, Singh SN, Fletcher RD. part of a heart failure disease management program. Congest Heart
Intrathoracic Impedance preceding ventricular tachyarrhythmia epi- Fail 2010;16:159e63.
sodes. Pacing Clin Electrophysiol 2010;33:960e6. 14. Broesch L, Heywood JT. Device diagnostic data by remote monitoring
12. Repoley JL, Dukes-Graves D, Kiser C, Worley S, Gohn DC. Surgical provides time efficient method of managing heart failure patients. [ab-
revision of implantable device ‘‘pocket’’ or lead can alter thoracic stract]. J Card Fail 2009;17:S111.