Stovner et al.

The Journal of Headache and Pain (2022) 23:34

The Journal of Headache
https://doi.org/10.1186/s10194-022-01402-2 and Pain

REVIEW ARTICLE Open Access

The global prevalence of headache:

an update, with analysis of the influences
of methodological factors on prevalence
estimates
Lars Jacob Stovner1,2* , Knut Hagen1,3, Mattias Linde1,4 and Timothy J. Steiner1,5

Abstract
Background: According to the Global Burden of Disease (GBD) study, headache disorders are among the most
prevalent and disabling conditions worldwide. GBD builds on epidemiological studies (published and unpublished)
which are notable for wide variations in both their methodologies and their prevalence estimates.
Our first aim was to update the documentation of headache epidemiological studies, summarizing global preva-
lence estimates for all headache, migraine, tension-type headache (TTH) and headache on ≥15 days/month (H15+),
comparing these with GBD estimates and exploring time trends and geographical variations. Our second aim was to
analyse how methodological factors influenced prevalence estimates.
Methods: In a narrative review, all prevalence studies published until 2020, excluding those of clinic populations,
were identified through a literature search. Prevalence data were extracted, along with those related to methodol-
ogy, world region and publication year. Bivariate analyses (correlations or comparisons of means) and multiple linear
regression (MLR) analyses were performed.
Results: From 357 publications, the vast majority from high-income countries, the estimated global prevalence of
active headache disorder was 52.0% (95%CI 48.9–55.4), of migraine 14.0% (12.9–15.2), of TTH 26.0% (22.7–29.5) and of
H15+ 4.6% (3.9–5.5). These estimates were comparable with those of migraine and TTH in GBD2019, the most recent
iteration, but higher for headache overall. Each day, 15.8% of the world’s population had headache. MLR analyses
explained less than 30% of the variation. Methodological factors contributing to variation, were publication year, sam-
ple size, inclusion of probable diagnoses, sub-population sampling (e.g., of health-care personnel), sampling method
(random or not), screening question (neutral, or qualified in severity or presumed cause) and scope of enquiry
(headache disorders only or multiple other conditions). With these taken into account, migraine prevalence estimates
increased over the years, while estimates for all headache types varied between world regions.
Conclusion: The review confirms GBD in finding that headache disorders remain highly prevalent worldwide, and it
identifies methodological factors explaining some of the large variation between study findings. These variations ren-
der uncertain both the increase in migraine prevalence estimates over time, and the geographical differences. More
and better studies are needed in low- and middle-income countries.

*Correspondence: lars.stovner@ntnu.no
1
Department of Neuromedicine and Movement Science, NTNU
Norwegian University of Science and Technology, Trondheim, Norway
Full list of author information is available at the end of the article

© The Author(s) 2022. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which
permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the
original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or
other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line
to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory
regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this
licence, visit http://​creat​iveco​mmons.​org/​licen​ses/​by/4.​0/. The Creative Commons Public Domain Dedication waiver (http://​creat​iveco​
mmons.​org/​publi​cdoma​in/​zero/1.​0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data.
Stovner et al. The Journal of Headache and Pain (2022) 23:34 Page 2 of 17

Keywords: Headache, Methodology, Migraine, Prevalence, Narrative review, Tension-type headache, Global
campaign against headache

Introduction GBD and the Global Campaign against Headache [1]

Documenting the burden of headache disorders has from GBD2010 onwards. Since the review, many new
become an important task, brought to attention by the studies have provided more insight into the importance
Global Campaign against Headache [1]. Through the and influences of methodological issues, and this addi-
Global Burden of Disease (GBD) study, headache disor- tional insight informed the methodological guidelines
ders are revealed as one of the major public-health con- [4]. Here we update that review, and the documentation
cerns globally and in all countries and world regions [2]. of headache epidemiological studies, summarizing global
In the 2019 iteration (GBD2019), migraine alone was prevalence estimates for headache, migraine, tension-
second among the causes of disability, and first among type headache (TTH) and headache on ≥15 days/month
women under 50 years of age [3]. GBD estimates are now (H15+), comparing these with GBD estimates and
updated annually to monitor changes in disease burden exploring time trends and geographical variations. We
around the globe, and thereby forecast future needs in also use the data and material from all the studies to ana-
health services. lyse empirically how methodological factors might influ-
For the various disorders it considers, GBD uses ence prevalence estimates.
multiple data sources (epidemiological studies, health
registers, official statistics, hospital data etc) to make Methods
best-informed estimates of prevalence and burden. For Literature search and data extraction
headache disorders, however, epidemiological studies are A new literature search through PubMed, using the
the only source, but these are now available from many search terms “headache epidemiology”, “migraine epi-
countries, including several yet unpublished from the demiology”, “headache prevalence” or “migraine preva-
Global Campaign. While GBD handles data for all dis- lence”, identified studies published till the end of 2020.
eases in a relatively uniform way, the quality of its esti- The search had been performed several times by LJS,
mates for each disease depends on a good understanding KH and ML since our 2007 publication [5], and was con-
of the particular methodological issues that may influ- cluded by LJS during the first month of 2021.
ence them. Headache epidemiology is a relatively young This proved to be a time-consuming task. The search
and immature discipline [4]. Studies follow general epi- terms “headache” and “prevalence” returned more than
demiological principles, but the field has its peculiar 18,000 articles in PubMed, and “migraine” and “epidemi-
methodological problems [4]. It is believed that the large ology” more than 6000, the great majority irrelevant for
variations in reported prevalences from country to coun- our purpose, but we found no way to restrict the search:
try, and sometimes within countries, are to a large extent additional terms such as “population-based” excluded
caused by methodological differences between studies, many relevant papers. To include all parts of the world
but this has not been explored empirically. and all settings that were possible, we accepted data from
Guidelines on performing and reporting headache epi- all non-clinical samples as well as those that were strictly
demiological studies were published in 2014 [4]. They population-based, which entailed browsing all the pages
included criteria for judging the quality of studies from in PubMed returned by these search terms. While most
their reported methodology, and some adjustments to irrelevant papers could be dismissed from the title alone
prevalence estimates were based upon these in the most and some from the abstract, with only a few requiring
detailed analysis of headache data, from GBD 2016 [2]. perusal of their full texts. This introduced a subjective
Nevertheless, these criteria were mostly theoretical, element in the search that was not in conformity with
derived from general principles and expert opinion (see accepted methodology for systematic review. But by not
[2] Appendix). A better understanding of which meth- restricting the literature search to strictly population-
odological factors influence findings, and how, may be based samples we created a database similar to (although
gained by reviewing metadata from all studies regarding larger than) the one used in GBD studies of headache.
how they collected and processed prevalence data, and This made our investigation of methodological factors
analyzing findings across these studies accordingly. more relevant to future evaluations of GBD estimates.
In 2007 we reviewed all published studies of the preva- Many additional studies were detected from the ref-
lence and burden of headache [5]. This review, highly erence lists in relevant reviews (e.g. [6–8]). All articles
cited, constituted the basis for collaboration between were screened for population of interest, and excluded
Stovner et al. The Journal of Headache and Pain (2022) 23:34 Page 3 of 17

if reporting only samples derived from clinical sources probable TTH (dTTH and pTTH). Otherwise, when
(hospitals, medical practices, etc) or those described as these were not separately provided, we used prevalences
patients or identified through patient groups. as reported.
We extracted prevalence data for all headache,
migraine, TTH and H15+, overall and for each gender.
We also extracted data on the various aspects of meth- Screening question
odology and design described below. When studies As far as we could, we distinguished between studies
failed to provide value for a variable of interest, we stip- with a neutral question (“have you had a headache?”)
ulated a value, assigned a separate value, or registered from those using questions qualified in some manner
it as ‘unknown’ (specific examples are given below: Age in terms of severity, frequency or presumed but unveri-
of population, Timeframe of headache, Application of fied cause (“do you suffer from headache?”, “have you
ICHD criteria). had severe/recurrent headaches?”, “have you had a head-
ache not caused by hangover, common cold, flu or head
Age of population sample trauma?”, etc) [12].
Far from all studies reported either mean or median age
of the population of interest or of the sample, but almost
all indicated the age range (youngest to oldest). Therefore, Other study properties and quality measures
we could not extract data for each age group but, instead, In addition to geographical origin and publication year,
used the mid-range value as a uniform, albeit imprecise, we extracted data related to the quality criteria [4]: those
indicator of population age. In the relatively few studies describing the population of interest (the general popu-
not even specifying this, for example only stating “adults”, lation or a specified sub-population), sampling method
we stipulated the range to be 20–85 years. (randomness and representativeness), size of sample, par-
ticipating proportion, methods of data collection (access
Timeframe of headache to and engagement with participants) and validation of
The International Classification of Headache Disorders diagnostic questions. These, and whether ICHD criteria
(ICHD-3) specifies that “For most purposes, patients or reasonable modifications of them, and whether a suit-
receive a diagnosis according to the headache pheno- able timeframe had been indicated (see above for each),
type currently present or that has presented within the were each scored 1–4, with a penalty, punitively set at − 4
last year” [9] (our emphasis). Nevertheless, to capture [4], applied whenever a study provided no information
all studies that might be relevant for our purposes, we on any criterion except timeframe (no information on
included those reporting 1-year, 6-month or 3-month this was almost invariably understood to mean “current
timeframes, “current headache” or no specific time- headache”). A summed score for all eight criteria, theo-
frame, subsuming their findings into prevalence of “active retically ranging from − 28 to 32, was calculated for each
headache disorder”. We separately considered lifetime study.
prevalence and very short timeframes (headache now, or For MLR analyses we dichotomized the quality meas-
headache yesterday). ures [4] that were not interval or ordinal variables:

Application of ICHD criteria with regard to definite • population of interest (unselected [general] popula-
and probable diagnoses tion of a country, community or tribe, or pupils of
In studies using ICHD criteria or modifications of them obligatory schools, versus selected subpopulations
(almost all those published after 1988), we expected [e.g., university students, factory/workplace employ-
prevalence estimates of migraine and TTH to be highly ees, minorities, etc], or unstipulated [additionally,
dependent on how these criteria were applied when diag- we registered whether selected subpopulations were
noses were uncertain because one of the criteria was not health-care personnel such as medical students, hos-
fulfilled. ICHD-I used the terms “migrainous disorder pital employees, neurologists, etc]);
not fulfilling above criteria” (coded 1.7) and “headache of • sample representativeness of the population of inter-
tension-type not fulfilling above criteria” (coded 2.3) [10], est (random sampling versus non-random sampling
whereas later editions referred to “probable migraine” or failed attempt to secure randomness);
(coded 1.5) and “probable TTH” (coded 2.4) [9, 11]. We • access to and engagement with participants (face-
extracted prevalence data for each of these diagnostic to-face or telephone interview versus unsupervised
categories when they were provided, and, when sum- questionnaire completion or unstipulated);
marizing data, summed the estimates for definite and • validation of diagnostic questions (effort at validating
probable migraine (dMig and pMig) and for definite and versus none or unstipulated);
Stovner et al. The Journal of Headache and Pain (2022) 23:34 Page 4 of 17

• application of ICHD criteria and distinction between each individual variable of significance, β-standardised
definite and probable diagnoses versus not or unstip- (the direction and strength of its correlation with the
ulated. dependent variable [prevalence]) and the square of the
semipartial correlation ­(SPCsqr: the proportion of the var-
We rated participating proportion in 10 categories (0: iability explained by that variable alone). While all varia-
unknown; 1–8 for the deciles 0–9 to 70–79, and 9 for bles used in the bivariate analyses (listed in Table 6) were
≥80%). Studies in obligatory schools performed in class entered into the MLR models for all headache disorders,
during school hours were accorded the value 9 even when this was an exploratory exercise; therefore, a stepwise
the exact proportion was not given, or uncertain because method excluded variables when F > 0.06 and included
of unrecorded absences, since such studies strongly those when F ≤ 0.05. Although MLR is quite robust with
encourage participation. Studies using convenience sam- regard to non-normal distributions, the variable “num-
pling (encountering people on the streets, or mailing or ber of participants” was very skewed, most samples being
emailing questionnaires), which were generally unable in the hundreds or low thousands but some very large
to specify the denominator, and studies not stating the (tens or hundreds of thousands). Therefore, this variable
number addressed to achieve the desired number of par- was logarithmically transformed (ln). For the seven GBD
ticipants, were rated − 4. superregions (High Income countries [HI], Central and
Additionally, although not a quality criterion, we distin- Eastern Europe and Central Asia [CEEAO], Latin Amer-
guished between studies according to scope (restricted, ica and Caribbean [LAC], North Africa and Middle East
with headache and/or one or more specified headache [NAME], South Asia [SA], SouthEast and East Asia and
types the only diagnoses of interest, versus broader Oceania [SEEAO], and Sub-Saharan Africa [SSA]), we
scope, headache being only one among many [often a created dichotomized dummy variables using HI super-
range of neurological disorders or pain conditions]). region, which had more studies than all others combined,
as the reference.
Data handling and statistics
Since many variables were not normally distributed,
Many studies reported imbalance in the proportions of we undertook a sensitivity analysis using non-parametric
male and female participants, and not all adjusted the methods (correlations with Spearman’s rho and compari-
overall prevalence estimates accordingly. Since our inter- sons of group differences with Mann-Whithney U-test
est was in estimates for the general population, with a for two groups or Kruskall Wallis Test for several groups).
gender distribution of 50/50, we calculated overall prev- No good non-parametric method exists for MLR-analy-
alence in all sampled populations as the mean of the sis. All dependent variables were checked for normality
estimates for males and females, even when, in sub-pop- (Shapiro-Wilk [SW] test), both before and after attempt-
ulations, distributions might be different. For this reason, ing to normalize them by ln-transformation and removal
we present some estimates that diverge somewhat from of outliers (outside 1.5 *interquartile range). In the MLR
those published. analysis, multicollinearity was evaluated by Variance
To summarize prevalence estimates from multiple Inflation Factor (VIF, < 4.0), and residuals in all models
studies (Tables 1, 2, 3, 4 and 5), we took the mean with were checked for normal (histograms and P-P plots) and
95% confidence intervals (CIs), also presenting median equal distribution (homoscedasticity with scatter plots)
values for comparisons of the main prevalence estimates. (see Additional file 2).
CIs were calculated by bootstrapping (drawing 1000 All analyses were performed with IBM SPSS (ver-
random samples), a robust method when the underly- sion 27). We considered p < 0.05 (two-sided) to indicate
ing distribution is not normal or unknown. SPSS did not significance. No adjustments for multiple testing was
calculate CIs when there were ≤ 5 observations. In bivari- performed since testing was done only for exploratory
ate analyses (Table 6), correlations were measured with purposes.
Pearson’s r, and means compared with Student’s t-test
for independent samples or one-way ANOVA for several Results
groups. Since testing was for exploratory purposes, we Studies
did not correct for multiple tests. A total of 357 published papers were identified that
We used multiple linear regression (MLR) analyses to were judged relevant to this review (Additional file 1).
explore associations between prevalence (dependent var- The table has more than 357 rows because some papers
iable) and study variables (geography, publication year, were referred to multiple times, reporting data from
and the variables related to method and quality described two or more studies performed in different years, in
above, reporting R ­ 2 (“Goodness-of-fit”, the proportion different countries or places, with different meth-
of the variation explained by the whole model) and, for ods in the same place, or with different timeframes.
Table 1 Prevalence (%) of active headache disorder by type, age (mid-range values of studies) and gender
Gender/age (years) Headache Migraine TTH H15+
N Mean (median) 95% CI N Mean (median) 95% CI N Mean (median) 95% CI N Mean (median) 95% CI
Stovner et al. The Journal of Headache and Pain

Both genders
0–9 10 42.7 30.6–55.4 12 5.0 3.6–6.5 4 5.9 1.3–12.8 0 – –
10–19 60 57.9 52.0–63.9 61 13.4 11.4–15.6 40 25.1 18.8–31.6 14 3.3 2.3–4.0
20–64 107 50.2 46.0–54.8 165 15.2 13.8–16.6 65 27.6 24.0–31.3 43 5.1 4.2–6.2
(2022) 23:34

  
≥ 65 6 40.2 30.0–50.6 6 7.9 4.9–10.3 1 32.2 – 2 4.2
All ages 183 52.0 (53.5) 48.9–55.4 244 14.0 (11.6) 12.9–15.2 110 26.0 (21.4) 22.7–29.5 59 4.6 (3.6) 3.9–5.5
Males
0–9 2 40.4 – 11 4.7 3.3–6.1 4 5.4 – 0 – –
10–19 39 51.4 43.8–58.6 36 10.2 8.4–12.2 16 25.5 16.2–35.5 6 1.4 0.7–2.2
20–64 71 41.8 36.5–47.6 128 8.7 7.6–9.8 43 24.2 18. 7–30.0 33 3.2 2.4–4.3
  
≥ 65 5 27.8 – 5 4.5 – 1 28.1 – 2 2.2 –
All ages 117 44.4 (46.2) 40.3–48.3 180 8.6 (7.4) 7.8–9.5 64 23.4 (17.9) 19.4–28.0 41 2.9 (1.9) 2.2–3.8
Females
0–9 2 41.5 – 11 5.6 3.9–7.3 4 6.4 – 0 – –
10–19 38 62.3 55.0–69.0 37 15.3 12.7–18.1 17 27.1 17.7–37.2 6 3.1 1.4–4.7
20–64 74 56.5 51.4–61.4 131 18.6 17.0–20.5 44 28.8 23.2–35.4 33 6.6 5.2–8.0
  
≥ 65 5 50.4 – 5 10.6 – 1 36.3 – 2 5.8 –
All ages 119 57.8 (60.0) 53.9–61.7 184 17.0 (15.5) 15.6–18.5 66 27.1 (21.0) 22.7–32.3 41 6.0 (5.2) 4.9–7.4
N number of studies, TTH tension-type headache, H15+ headache on ≥15 days/month, CI confidence interval (not given when N ≤ 5), SD standard deviation
Page 5 of 17
Table 2 Prevalence (%) of active headache disorder by type and geographical area (estimated in this study and, for migraine and TTH, in GBD2019)
Stovner et al. The Journal of Headache and Pain

Headache Migraine TTH H15+

1 2 1 1
GBD superregion % N Mean (95%CI) GBD2019 N Mean (95% CI) GBD2019 N Mean (95% CI) GBD2019 N1 Mean (95% CI)

Global 100 183 52.0 (48.9–55.4) 35 (32.3–37.7) 244 14.0 (12.9–15.2) 15.2 (13.2–17.4) 110 26.0 (22.7–29.5) 26.8 (23.5–30.1) 59 4.6 (3.8–5.4)
(2022) 23:34

CEECA 5.3 14 53.6 (41.8–65.2) 41.8 (38.6–44.9) 18 16.6 (12.6–21.3) 15.6 (13.6–17.9) 15 31.1 (22.8–39.2) 35.2 (31.3–39.0) 8 5.2 (3.3–7.2)
HI 14.7 88 51.9 (47.5–56.6) 43.4 (40.2–46.5) 117 13.7 (12.3–15.5) 17.2 (15.1–19.9) 41 30.4 (24.4–36.5) 36.4 (32.4–40.4) 24 3.4 (2.9–4.1)
LAC 7.4 24 55.0 (44.9–65.4) 36.7 (33.8–39.7) 33 17.0 (12.9–22.0) 16.3 (14.1–19.1) 11 26.8 (16.1–39.8) 27.9 (24.4–31.4) 6 6.4 (5.0–7.5)
SEEAO 27.6 10 38.7 (20.4–56.5) 37.3 (34.3–40.4) 16 9.0 (6.0–13.1) 17.5 (15.1–20.2) 8 11.1 (3.5–19.5) 27.7 (24.2–31.2) 5 2.8
NAME 7.7 20 60.5 (50.9–68.5) 35.4 (32.3–38.6) 30 11.8 (9.4–14.4) 16.7 (14.4–19.4) 19 20.5 (14.4–27.8) 25.7 (22.1–29.4) 7 4.9 (3.0–6.9)
SA 23.4 7 53.2 (29.0–74.1) 35.6 (32.7–38.6) 8 22.2 (16.1–27.8) 15.5 (13.4–17.9) 5 33.1 (20.3–42.9) 27.4 (24.1–31.0) 4 6.9
SSA 14.0 19 43.8 (32.4–55.6) 27.4 (24.6–30.2) 21 12.4 (9.1–16.8) 11.3 (9.5–13.3) 10 16.6 (12.4–21.0) 21.0 (17.9–24.4) 4 6.7
1% of the world’s population living in the superregion
2 The global total of studies is higher than the sum of studies per superregion because some studies give data from more than one superregion, i.e. it cannot be assigned to a single superregion
H15+ headache on ≥15 days/month, %: percent of world population, N: number of studies, TTH: tension-type headache, CI: confidence interval (not given when N ≤ 5)
CEECA Central Europe, Eastern Europe and Central Asia, HI High-income countries, LAC Latin America and Caribbean, SEEAO Southeast Asia, East Asia and Oceania, NAME North Africa and Middle East, SA South Asia, SSA
Sub-Saharan Africa
Page 6 of 17
Stovner et al. The Journal of Headache and Pain (2022) 23:34 Page 7 of 17

Table 3 Prevalence (%) of active headache disorder by type, gender and timeframe
Timeframe Headache Migraine TTH H15+
N Mean 95% CI N Mean 95% CI N Mean 95% CI/ N Mean 95% CI

Lifetime
Both genders 42 66.6 58.8–73.9 49 17.5 14.6–21.0 18 40.6 30.0–51.4 9 4.5 1.8–7.7
Males 28 65.0 53.8–74.4 39 11.6 9.1–14.8 9 35.7 19.6–52.6 6 1.9 0.9–3.1
Females 28 73.3 63.3–83.2 41 21.0 17.3–25.4 9 43.8 26.5–61.8 6 4.5 2.2–7.5
1-year
Both genders 110 53.5 49.1–57.7 150 15.3 13.7–16.9 77 26.4 22.9–30.3 50 4.8 4.0–5.7
Males 73 46.4 41.1–51.7 112 9.1 7.9–10.3 47 24.2 19.2–29.6 39 2.9 2.2–3.8
Females 72 60.5 55.1–65.8 113 18.1 16.3–29.9 48 28.5 22.7–34.3 39 6.2 4.9–7.6
6-month, both genders 15 60.9 50.6–71.3 12 11.1 6.7–16.9 9 32.2 15.1–50.8 2 6.0 4.8–7.2
3-month, both genders 12 52.2 36.5–66.6 10 13.4 9.1–17.9 5 20.2 2.1–38.6 3 2.7 2.2–3.3
“Current”, both genders 3 37.2 – 10 12.9 7.7–18.3 0 – – 0 – –
Unknown, both genders 43 45.9 38.5–53.1 62 11.9 10.0–14.0 19 22.9 17.2–29.2 4 3.4 1.5–5.4
1-day
Both genders 12 15.8 10.0–22.0 5 7.0 – 4 8.7 – 3 2.5 –
Males 8 8.0 4.0–12.8 2 2.4 – 1 5.1 – 1 1.8 –
Females 8 14.9 8.9–21.7 2 5.5 – 1 10.1 – 1 4.4 –
All timeframes, both genders 237 52.7 49.7–55.8 298 14.5 13.4–15.5 132 27.4 24.2–30.9 71 4.5 3.9–5.2
N number of studies, TTH tension-type headache, H15+ headache on ≥15 days/month, CI confidence interval (not given when N ≤ 5)

Some rows have more than one reference because data Descriptive overview
from the same study but on different headache types Many studies were deficient in their reporting, hindering
(e.g. migraine and TTH) were reported in multiple our analyses.
publications. For age, many gave data for different groups (e.g., 5-, 10- or
The number of publications from each GBD superre- 20-year categories), some but not all did so for each gender,
gion varied considerably: 183 from HI, 25 from CEEAO, some did so for only one of the headache types and some did
44 from LAC, 42 from NAME, 11 from SA, 24 from so for each type. All studies on very specific age groups (e.g.,
SEEAO and 27 from SSA. It should be noted that all children, adolescents, elderly) gave age ranges, these being
GBD superregions are continuous geographical areas the source of all mid-range values below 10, from 10 to 19 or
except HI, which includes countries in Australasia, Asia above 65 years. Most of the many studies with mid-range val-
Pacific, Western Europe, North America and Southern ues between 20 and 64 years were conducted in adults (young
Latin America. and middle-aged), but some included all ages (5–99 years).

Table 4 Prevalence (%) of active headache disorder by type, gender and diagnostic certainty (definite or probable)
Studies specifying both Studies not making Definite+probable
or not specifying the (where given) or
Definite Probable All (definite + probable) distinction unspecified
N Mean 95% CI N Mean 95% CI N Mean 95% CI N Mean 95% CI N Mean 95% CI

Migraine
Both genders 59 10.5 9.0–12.0 59 8.7 7.1–10.3 59 19.2 16.7–21.7 185 12.4 11.2–13.8 244 14.0 12.8–15.2
Males 27 6.2 5.1–8.3 27 6.3 4.7–8.0 27 12.9 10.5–15.3 153 7.9 7.0–8.8 180 8.6 7.8–9.6
Females 28 14.5 11.9–17.2 28 9.7 7.5–12.0 28 24.2 21.2–27.5 156 15.7 14.2–17.4 184 17.0 15.5–18.4
TTH
Both genders 33 19.4 15.3–24.1 33 10.0 7.9–12.6 33 29.4 24.4–34.3 77 24.5 20.7–29.0 110 26.0 22.9–29.6
Males 12 18.1 11.4–25.4 12 8.6 4.6–13.3 12 26.8 20.6–33.5 52 22.6 17.7–28.4 64 23.4 19.1–28.0
Females 12 22.4 13.2–32.9 12 9.4 5.3–14.6 12 31.8 23.6–40.4 54 26.1 21.0–31.7 66 27.1 22.7–32.2
N number of studies, TTH tension-type headache, CI confidence interval
Stovner et al. The Journal of Headache and Pain (2022) 23:34 Page 8 of 17

Table 5 Prevalences (%) of probable migraine and probable TTH by age and gender, and as proportions (%) of all (probable +
definite) migraine and TTH
Gender/age (years) Probable migraine Probable TTH Total migraine Total TTH pMig as % of pTTH as
(pMig+dMig) (pTTH+dTTH) total migraine % of total
TTH
N Mean 95% CI N Mean 95% CI N Mean 95% CI N Mean 95% CI

Both genders
0–9 0 – – 0 – – – – – 0 – –
10–19 13 10.5 7.0–14.4 8 14.0 9.5–19.8 13 20.3 14.3–26.4 8 29.2 19.3–45.1 52 48
20–64 44 8.4 6.6–10.2 24 9.0 6.5–11.9 44 19.4 16.5–22.5 24 29.4 24.6–34.4 43 31
≥65 2 3.1 – 2 6.6 4.3–8.9 2 6.8 – 2 20.6 – 46 32
All ages 59 8.7 7.1–10.4 34 10.0 7.7–12.6 59 19.2 16.7–21.7 34 28.9 24.4–34.2 45 35
Males
0–9 0 – – 0 – – 0 – – 0 – –
10–19 5 9.1 5.4–12.9 3 9.8 – 5 18.0 12.6–24.0 3 18.9 – 51 52
20–64 21 5.7 4.0–7.7 8 8.8 2.9–15.6 21 12.1 9.6–14.6 8 29.6 19.7–38.0 47 30
≥65 1 3.7 – 1 4.3 – 1 5.1 – 1 28.1 – 73 15
All ages 27 6.3 4.7–8.0 12 8.6 4.6–13.3 27 12.9 10.5–15.3 12 26.8 20.6–33.5 49 32
Females
0–9 0 – – 0 – – 0 – – 0 – –
10–19 5 13.4 7.8–18.5 3 11.3 7.2–16.9 5 26.0 20.0–34.2 3 21.7 21.2–22.4 52 52
20–64 22 8.9 6.6–11.2 8 9.4 3.8–16.6 22 24.2 20.7–27.9 8 35.1 22.8–47.8 37 27
≥65 1 8.0 – 1 4.3 – 1 16.0 – 1 36.3 – 50 12
All ages 28 9.7 7.6–11.7 12 9.4 5.3–14.6 28 24.2 21.2–27.4 12 31.8 23.6–40.4 40 30
N number of studies, TTH tension-type headache, CI confidence interval (not given when N ≤ 5), pMig probable migraine, dMig definite migraine, pTTH probable TTH,
dTTH definite TTH

Some, however, were restricted to the actively working Prevalence of an active headache disorder
population (18–65 or 20–65 years), and some to narrow Globally, an active headache disorder of any type
age ranges (e.g., a single-year cohort of 40 years, or 20–35, was present in 52.0% of the populations studied
or 45–64 years). Hence, our study was not well able to show (males 44.4%, females 57.8%), migraine in 14.0%
age-specific effects: we could relate prevalences only to four (males 8.6%, females 17.0%) and TTH in 26.0%
ranges of mid-range values (< 10, 10–19, 20–64, ≥65 years). (males 23.4%, females 27.1%) (Table 1). The overall
With regard to timeframe, many studies reported value is not always the mean of the values for males
1-year prevalence defined appropriately (a positive and females because some studies were restricted to
response to “have you had headache during the last one gender while others gave only the overall value.
year?”), but a number used shorter timeframes, often H15+ was reported by 4.6% (males 2.9%, females
with the assumption that questions about headache dur- 6.0%). All these data are detailed in Table 1 by gen-
ing the last year were sufficiently answered by consid- der and age. Headache prevalences were reportedly
ering only the last 3–6 months. Others asked whether similar in all age groups in both genders (overlap-
participants had headache “currently”, or specified no ping CIs). On the other hand, studies with mid-range
timeframe (e.g., “do you suffer from headaches?”), but age values below 10 or above 65 years reported lower
clearly implied the present rather than throughout the migraine prevalences in both males and females, and
lifetime. Studies with even shorter timeframes (usually studies with values below 10 years reported lower
24 h, “headache yesterday”, or “now”) always defined the TTH prevalences in both genders. H15+ estimates
timeframe when questioning participants. were lower among females with mid-range age val-
Many studies did not describe the screening question. ues of 10–19 years and among males. Mean values
As a group, they probably included a mixture of screen- were approximately 20% higher than the medians,
ing questions (neutral or qualified). for migraine (mean 14.0% versus median 11.6%),
Many studies were silent on the distinction between TTH (26.0% versus 21.4%) and H15+ (4.6% versus
definite and probable diagnoses, and may or may not 3.6%), but not for all headache (52% versus 53.5%)
have included the latter. (Table 1).
Stovner et al. The Journal of Headache and Pain (2022) 23:34 Page 9 of 17

Table 6 Bivariate analyses of active headache disorders according to methodological aspects of studies
Headache (183 studies) Migraine (244 studies) TTH (110 studies) H15+ (59 studies)

Correlations r p r p r p r p
Year of publication 0.09 0.22 0.30 < 0.0005 −0.03 0.8 0.20 0.14
Number (ln transformed) of participants −0.27 < 0.001 −0.26 < 0.001 −0.17 0.07 −0.33 0.01
Participating proportion (9 categories) 0.11 0.15 0.02 0.7 −0.11 0.2 −0.02 0.9
Age (mid-range value) (years) −0.14 0.05 0.03 0.6 0.13 0.2 0.14 0.3
Summed quality score 0.03 0.7 −0.02 0.8 − 0.02 0.8 0.00 1.0
Means of reported prevalences (%) Mean (SD, N) Mean (SD, N) p Mean (SD, N) p Mean (SD, N) p
Populationa General/schools 51.9 (22.7, 142) 1.0 13.5 (8.4, 177) 0.2 26.5 (17.0, 85) 0.6 4.8 (3.1, 54) 0.2
Selected 52.2 (22.7, 41) 15.5 (11.1, 66) 24.4 (21.1, 25) 3.0 (2.1, 5)
Health-care ­personnelb No 50.2 (25.4, 33) 0.3 13.4 (9.1, 52) 0.03 23.3 (22.9, 19) 0.7 3.0 (2.1, 5) –
Yes 60.0 (22.9, 8) 23.0 (14.7, 15) 27.8 (15.2, 6) –
Sampling ­methodc Random 54.7 (23.2, 44) 0.4 12.8 (8.1, 184) 0.002 25.7 (18.6, 83) 0.8 4.5 (3.2, 51) 0.4
Not random 51.1 (23.1, 139) 18.0 (11.4, 60) 26.8 (16.0, 27) 5.5 (2.1, 8)
Mode of ­engagementd Telephone/face-to face 50.8 (23.6, 95) 0.5 13.1 (8.8, 136) 0.07 26.3 (18.7, 66) 0.8 5.5 (3.5, 32) 0.02
Questionnaire 53.2 (22.6, 88) 15.3 (9.8, 108) 25.5 (17.0, 44) 3.6 (2.0, 27)
Validation of diagnostic ­methode Yes 51.2 (24.3, 93) 0.7 14.3 (10.3, 107) 0.7 25.6 (18.0, 56) 0.8 4.6 (3.2, 41) 0.9
No 52.8 (22.1, 90) 13.8 (8.5, 137) 26.(18.0, 54) 4.7 (2.6, 18)
dMig and ­pMigf Yes 55.4 (20.8, 46) 0.3 19.2 (9.9, 59) < 0.001 27.6 (15.9, 41) 0.5 5.6 (3.9, 22) 0.1
No 50.8 (23.9, 137) 12.4 (8.5, 185) 25.0 (19.1, 69) 4.1 (2.2, 37)
­ TTHg
dTTH and p Yes 61.5 (17.8, 29) 0.02 20.0 (10.3, 32) < 0.001 29.5 (13.7, 33) 0.2 6.2 (4.1, 18) 0.04
No 50.2 (23.7, 154) 13.1 (8.8, 212) 24.5 (19.4, 77) 3.9 (2.2,41)
Number of ­conditionsh Only headaches 52.9 (23.0, 173) 0.02 14.3 (9.3, 235) 0.04 26.0 (18.0, 110) – 4.6 (3.0, 59) –
Including other conditions 35.2 (20.4, 10) 7.8 (5.6, 9) –
Type of screening question Neutral 61.8 (22.0, 58) < 0.0005 17.1 (9.3, 65) 0.006 31.9 (22.5, 42) 0.02 5.0 (2.1, 24) 0.8
Unknown 50.8 (23.7, 71) 12.9 (10.0, 111) 23.0 (14.2, 41) 4.3 (2.4, 16)
Severity qualification 41.8 (19.1, 52) 12.7 (6.6, 66) 20.9 (12.1, 26) 4.4 (4.5, 18)

TTH tension-type headache, H15+ headache on ≥15 days/month, r Pearson’s r, p two-tailed p-value (Pearson’s r, independent-samples Student’s t-test, or one-way
ANOVA for screening question), SD standard deviation, N number of studies
a
General population, community-based sample or obligatory schools versus selected populations (among workplace employees, college students, health-plan
members etc)
b
Among selected populations: medical students, hospital staff or neurologists versus other selected populations
c
Random sampling versus non-random sampling or failure to obtain random sample
d
Telephone or face-to face interview versus self-administered questionnaire or not specified
e
Effort at validating versus none or unstipulated
f
Studies distinguishing between dMig and pMig and reporting both versus those not
g
Studies distinguishing between dTTH and pTTH and reporting both versus those not
h
Studies restricted to headache (or migraine, TTH etc) versus studies investigating multiple other disorders (neurological, other pain conditions etc)

Table 2 shows prevalences by geographical area (GBD migraine or TTH according to the proportions expected
superregions). All headache reportedly varied between to arise from each.
38.7% (SEEAO) and 60.5% (NAME), migraine between Table 2 also shows the percentages of the global pop-
9.0% (SEEAO) and 22.2% (SA), TTH between 11.1% ulation living in each GBD superregion. Adjusting for
(SEEAO) and 33.1% (SA), and H15+ between 2.8% these (not shown in the table), the global prevalence
(SEEAO) and 6.9% (SA). of headache was 45.6%, of migraine 13.6%, of TTH
For comparison, Table 2 also provides the latest avail- 21.1%, and of H15+ 4.7%. This simple adjustment does
able estimates from GBD [13]. For all headache, these not allow CIs to be calculated, and is valid only if esti-
were generally somewhat lower than our findings, glob- mates per superregion are correctly representative for
ally and in several of the GBD superregions (with non- the whole of each superregion. The adjusted values are
overlapping 95% CIs globally and in HI, LAC, NAME more similar to those of GBD for all headache, but not
and SSA). This was despite that GBD estimates for for migraine or TTH.
migraine and TTH were generally higher. GBD does In studies estimating prevalences of an active head-
not give separate data on H15+, although more recent ache disorder and of specific headache types, there were
iterations subsume medication-overuse headache within clear positive correlations between them: for headache
Stovner et al. The Journal of Headache and Pain (2022) 23:34 Page 10 of 17

with migraine (r = 0.46, p < 0.01, 142 studies), with TTH Bivariate analyses
(r = 0.48, p < 0.01, 84 studies) and with H15+ (r = 0.45, Table 6 shows bivariate correlations of active headache
p < 0.01, 42 studies), for migraine with TTH (0.36, disorders with continuous and ordinal variables (pub-
p < 0.0000, 105 studies) and with H15+ (0.45, p < 0.01, 43 lication year, number of participants [ln-transformed],
studies), and for TTH with H15+ (0.37, p = 0.2, 43 studies). participating proportion and summed quality score), as
well as comparing means of reported prevalences accord-
Prevalence related to methodological and other study ing to various categorical variables. Migraine prevalence
variables was positively associated with publication year (r = 0.30).
Timeframe Headache (r = − 0.27), migraine (r = − 0.26) and H15+
Table 3 shows that reported prevalences varied with (r = − 0.33) were negatively associated with number of
timeframe. Excepting 1-day estimates, this was mostly participants, TTH showing a trend in the same direction
with widely overlapping 95% CIs: of all headache from (r = − 0.17). There were no significant associations with
37.2% (“current”) to 66.6% (lifetime), of migraine from summed quality score (p ≥ 0.7). As to the population of
11.1% (6-month) to 17.5% (lifetime), of TTH from 20.2% interest, estimates for migraine from general populations,
(3-month) to 40.6% (lifetime), and of H15+ from 2.7% defined regions or schools were not significantly differ-
(3-month) to 6.0% (6-month). ent from those from selected sub-populations, although,
As to 1-day prevalence, 15.8% of the global popula- among the latter, estimates were higher in health-care
tion reportedly had headache on any particular day (7.0% personnel than others (23.0% versus 13.4%, p = 0.030).
migraine, 8.7% TTH and 2.5% H15+). Many of the stud- With regard to sampling method (random versus not
ies reporting 1-day prevalence (as headache yesterday) or failed attempt to secure randomness), estimates for
did not attempt to diagnose headache type. migraine were significantly lower in studies with random
sampling (12.8% versus 18.0%, p = 0.002), but this was not
so for all headache or the other headache types (p ≥ 0.4).
Definite and probable (or unspecified) diagnoses
Mode of engagement (telephone or face-to-face interview
Table 4 makes the distinction between definite and prob-
by untrained, trained or specialized interviewer versus
able diagnoses of migraine (reported in 59 studies) or of
self-administered questionnaire or unknown) influenced
TTH (33 studies). Mean reported prevalences for migraine
estimates for H15+ (5.5% versus 3.6%, p = 0.02), but not
were 10.5% for dMig and 8.7% for pMig, the sum being
those for the other headache types. There was no evident
19.2%. In studies providing gender-specific estimates,
influence of validation of the diagnostic method (p ≥ 0.7).
these were 6.2 and 6.3% for males, 14.5% and 9.7% for
Distinction or not between pMig and dMig made a dif-
females. For TTH they were 19.4% for dTTH and 10.0%
ference for migraine (19.2% versus 12.4%), and distinc-
for pTTH (sum 29.4%; males 18.1% and 8.6%, females
tion between dTTH and pTTH did so for all headache
22.4% and 9.4%).
(61.5% versus 50.2%), migraine (20.0% versus 13.1%), and
Table 4 also provides mean estimates from the 185
H15+ (6.2% versus 3.9%). Studies restricted to headache
studies not making or not specifying the distinction.
and not including other conditions reported higher esti-
These studies yielded a prevalence of migraine (12.4%)
mates for all headache (52.9% versus 35.2%, p = 0.02) and
that was somewhat higher than that of dMig (10.5%)
for migraine (14.3% versus 7.8%, p = 0.04). Neutral screen-
from the 59 studies that did distinguish, but lower than
ing questions were associated with higher estimates for
their sum of dMig and pMig (19.2%). The combination
all headache, migraine and TTH than were unspecified
of all studies gave the 14.0% global estimate reported
questions (intermediate) and those implying some level of
above. Similarly for TTH, the mean reported prevalence
severity, frequency or presumed causation.
of unspecified TTH (24.5%) was between that of dTTH
(19.4%) and sum of dTTH and pTTH (29.4%), with
26.0% from the combination of all studies. Multiple linear regression (MLR)
Table 5 relates definite and probable diagnoses to gen- None of the dependent variables (headache, migraine,
der and age. In the contributory studies, pMig consti- TTH and HA15+, in both genders combined) were
tuted 45% (8.7/19.2*100) of all migraine (pMig+dMig), normally distributed (p ≤ 0.01, SW). Normalizing the
somewhat more among males (49%) than females (40%), dependent variables (by ln- transformation and removal
and among adolescents 10–19 years of age (52%) than of outliers, see Statistics) was successful for three
adults aged 20–64 years (43%) or ≥ 65 years (46%). For (migraine, TTH and HA15+; p > 0.14), but not for head-
TTH, proportions of pTTH were somewhat lower (over- ache (p < 0.01). No significant collinearity was found in
all 35%, males 32%, females 30%), but markedly higher any of the models (VIF < 1.40). For distribution of resid-
among adolescents (48%) than among the two adult uals, see Additional file 2 (Histograms, P-P plots and
groups (31% and 32%). Scatterplots).
Stovner et al. The Journal of Headache and Pain (2022) 23:34 Page 11 of 17

Table 7 Multiple linear regression with stepwise inclusion of variables (HI superregion serves as reference for other superregions)

Model 1: Original Headache, N = 181 Migraine, N = 242 TTH, N = 109 H15+, N = 58

data set (5 variables) (8 variables) (4 variables) (2 variables)
Whole model F = 10.3, p < 0.001, ­R2 adj: 0.208 F = 13.9, p < 0.001, ­R2 adj: F = 6.9, p < 0.001, ­R2 adj: 0.181 F = 6.6, p = 0.003, ­R2 adj:
0.299 0.165
Variables with β standardised SPCsqr β standardised SPCsqr β standardised SPCsqr β standardised SPCsqr
significance
Year of publication 0.258 0.060
Number (ln trans- −0.207 0.040 − 0.146 0.018
formed) of partici-
pants
dMig and pMig 0.269 0.064
dTTH and pTTH 0.321 0.102
Health-care person- 0.173 0.022
nel
Sampling method −0.169 0.022
Type of screening −0.307 0.090 −0.285 0.080
question
Number of conditions −0.162 0.026
Mode of engagement 0.281 0.078
SSA superregion −0.148 0.021 −0.173 0.028 −0.232 0.052
SEEAO superregion −0.171 0.029 −0.136 0.018 −0.294 0.084
NAME superregion −0.113 0.011 −0.231 0.051
Model 2: ln trans‑ Headache, N = 176 Migraine, N = 238 TTH, N = 103 H15+, N = 57
formed dependent (4 variables) (7 variables) (4 variables) (2 variables)
variable and outli‑
ers (±1.5 interquar‑
tile range) removed
Whole model F = 11.3 p < 0.001, ­R2 adj: 0.191 F = 13.6, p < 0.001, ­R2 adj: 0.271 F = 5.5, p < 0.001, ­R2adj:0.150 F = 6.5, p = 0.003, ­R2 adj: 0.164
sqr sqr sqr
Variables with β standardised SPC β standardised SPC β standardised SPC β standardised SPCsqr
significance
Year of publication 0.259 0.064
Number (ln trans- −0.186 0.032 −0.137 0.016 −0.354 0.125
formed) of partici-
pants
dMig and pMig 0.291 0.078
dTTH and pTTH 0.166 0.027
Health-care person- 0.133 0.013
nel
Sampling method −0.169 0.022
Type of screening −0.277 0.072 −0.233 0.054
question
SEEAO superregion −0.110 0.012 −0.206 0.058 −0.250 0.062
SSA superregion −0.207 0.042 −0.180 0.031 −0.245 0.042
NAME superregion −0.264 0.068
2 2
R adj (Adjusted ­R ): The proportion of variability explained by the whole model (“Goodness of fit”), adjusted for overestimation caused by the number of effects
β-standardised: The direction and strength of its correlation with the dependent prevalence variable
SPCsqr: the proportion of the variability explained by that variable alone

Table 7 shows the MLR results for original (Model 1) included in Model 2. For H15+, the two models are very
and normalized (Model 2) dependent variables. Model different, probably showing that the robustness of MLR
2 should be the more robust. For TTH, the two models models declines with smaller numbers of observations,
are identical with regard to the variables included and particularly when the dependent variable is not normally
directions of the associations, and for migraine, the prin- distributed. For headache, two methodological variables
cipal difference is that one superregion (NAME) is not and one superregion occur in both models.
Stovner et al. The Journal of Headache and Pain (2022) 23:34 Page 12 of 17

Overall, a relatively small proportion of the varia- (summary measures), without conforming fully with
tions in reported headache prevalence was explained by accepted methodology for the former (see Methods: Lit-
these variables, ranging from 29.9% (whole model R ­ 2) in erature search and data extraction). Such reviews, following
migraine to 16.5% in H15+ (Model 1), and from 27.1% published guidelines, are usually made to evaluate effects of
(migraine) to 15.0% (H15+, Model 2). interventions. While many of the same principles remain
For all headache the methodological variables included relevant to other types of data, guidance and reporting
in both models were the type of screening question standards for systematic reviews of prevalences are lacking
­(SPCsqr, explaining 9.0% in Model 1 and 7.2% in Model [14]. Hence, we conceived this as a narrative rather than a
2), and the number of participants in the study (4.0% and systematic review of publications reporting headache prev-
3.2%, negatively correlated). Whether the study inves- alences, updating our previous review of 2007 [5].
tigated only headache or included other disorders was In comparison with our previous review, we found an
included in Model 1(2.6%, lower estimates in the latter), apparent increase in prevalence of migraine but not those
and whether it included both dTTH and pTTH in Model of other headache types. We also found wide geographical
2 (2.7% more when it did). For migraine, the most impor- variations, as did the 2007 review. Estimates for migraine
tant factor was distinction between dMig and pMig (6.4% were very high in populations with very good knowledge
and 7.8% in the two models), followed by publication year and interest in the disorder (e.g., neurologists, and other
(6.0% and 6.4%, positively correlated), sampling method health personnel). While estimates were influenced to
(2.2% in both models lower estimates when random), varying degrees by several methodological factors (nature
selection of health-care personnel (2.2% and 1.3% higher of the screening question, number of conditions investi-
estimates) and number of participants (1.8% and 1.6%, gated, sampling method, number of participants, how
negatively correlated). For TTH, the only methodological patients were engaged and not least – for migraine and
factor was the type of screening question (8.0% and 5.4%). TTH – how ICHD criteria were applied with regard to
For both migraine and TTH, certain world superregions definite and/or probable diagnoses), many of these influ-
(SSA, SEEAO, and NAME for TTH) had lower estimates ences were not significant. Since the review included
than the reference (HI superregion). For H15+, the only studies of all populations of interest other than clinical
methodological variable included in Model 2 was num- populations, and extracted many data on methodological
ber of participants (12.5%, negatively correlated). SEEAO and other aspects of each study, we believe it provides a
superregion had lower prevalence (6.2%, Model 2) than HI. strong basis for examination of these influences, which we
In Model 1 were also included whether the study distin- discuss in the following sections. Meanwhile we note that
guished between dTTH and pTTH (10.2%, higher when our prevalence estimates in this review derive from stud-
they did) and mode of engagement (7.8%, higher with tele- ies varying with regard to these factors, so that it remains
phone/face-to-face than with unsupervised questionnaire). in doubt whether the observed differences over time and
between geographical regions are real.
Sensitivity analyses Estimated medians for the different headache types
Correlations between the different headache types were were approximately 20% lower than the means, whereas,
very similar using a non-parametric method (Spearman’s for headache overall, the median estimate was 3% higher.
rho, 0.36–0.49, p < 0.01 for all). In the bivariate analyses It could be argued that medians, uninfluenced by outliers
(Table 6), two of the 56 p-values became non-significant with very high prevalences, were more correct and con-
with non-parametric methods (headache correlated with servative summary measures. On the other hand, we had
age, and TTH related to dTTH or pTTH) while four no a priori reason to believe that studies with high esti-
became significant (migraine related to mode of engage- mates were less correct or representative for the popula-
ment, TTH to participating proportion, age and screen- tion at large than those with low estimates. In many ways,
ing question). Thirteen analyses remained significant and studies finding high prevalences might have been method-
37 nonsignificant with both methods. ologically better (neutral screening question, considering
In addition, a second MLR analysis (not shown), with both probable and definite diagnoses, face-to face inter-
GBD superregions omitted from the modelling, found views, etc). Therefore, we based estimates on the means.
exactly the same methodological variables to be significant. The sensitivity analysis showed that the bivariate analy-
ses were little influenced by statistical method (paramet-
Discussion ric or non-parametric).
General comments
This review has many of the properties of a systematic Relation to case definition: “active headache disorder”
review (e.g., specification of search terms and databases, Case definition is of over-riding importance, and likely
evaluation of heterogeneity), and even of a meta-analysis to be the single most influential factor in any enquiry
Stovner et al. The Journal of Headache and Pain (2022) 23:34 Page 13 of 17

dependent on prevalence estimation (including burden In summary, with regard to case definition, we believe
estimation) [4, 15]. For headache disorders, case defini- that best estimates from available data are made by
tion revolves around timeframe, diagnostic criteria and including studies with various timeframes, and including
how the latter are applied. Since 1988, use of ICHD cri- definite and probable (or unspecified) diagnoses. These
teria has become mandatory in the absence of acceptable inclusions allow many more studies to be taken into
alternatives [4], but, from a pragmatic perspective, some account, providing data from countries and regions that
flexibility in their application to epidemiological enquiry otherwise would have few or none. This is not to say that
is invariably necessary. Empirically, we found different future studies should not give due regard to these very
approaches towards this flexibility among the reviewed important methodological issues.
studies, and these were not always described.
As to timeframe, we might have restricted this review to
studies reporting 1-year prevalence, but we believe the ICHD Relation to publication year and geography
definition of “active headache disorder” [9] does not exclude Unlike case definition, there were no obvious a priori rea-
studies reporting “current”, 6- or 3-month prevalences (see sons for expecting these factors to be significantly influ-
Methods: Timeframe of headache). These latter timeframes ential. Although geographical variation of course implies
were more restrictive, and therefore given to underestima- genetic, environmental, cultural, economic, lifestyle and
tion, as appears from Tables 1 and 3 (although only of epi- general health variations, among these only dwelling
sodic headache, by exclusion of those with low-frequency altitude has clearly demonstrated impact, and this only
episodes). On the other hand, those reporting lifetime prev- in Nepal [16]. Urban/rural divide and relative wealth or
alences would tend to give higher estimates through inclu- poverty are weakly related to headache prevalence in
sion of “non-active” headache. Those in which timeframe some studies, not always in the same direction. Popula-
was unknown (not specified) might go either way. Overall, tion age may be a factor: age itself is strongly associated
our estimates may be somewhat conservative because we with prevalence of all headache types, and some national
subsumed shorter timeframes into our case definition of populations are relatively young in comparison with the
“active headache disorder”, but this approach is validated by global mean (but, it should be noted, this factor is not
the fact that, for all timeframes subsumed (1-year, 6-month, independent of population poverty or general health).
3-month, “current” and unknown), 95% CIs were widely Compared to our previous summary of prevalence data
overlapping (except for H15+, with only 3 studies). from 2007 [5], this review has found increased preva-
As to ICHD criteria and their application, we might lences of active headache (from 46% to 52%), of migraine
also have restricted our case definitions for migraine and (from 11% to 14%) and of H15+ (from 3% to 4.6%) but
TTH to definite diagnoses (meeting all ICHD criteria). decreased prevalence of TTH (from 42% to 26%). How-
This would have had a large influence: probable diagnoses ever, in both bivariate and MLR analyses, association
accounted for > 40% of total migraine and > 30% of total between prevalence and publication year was significant
TTH. The persuasive arguments for including probable only for migraine. An increase in prevalence was also
diagnoses have been set out before [4]. In clinic, prob- found in some [17, 18] but not all [19, 20] studies per-
able diagnoses are, or should be, confirmed or refuted formed in the same populations at different time points.
during follow-up, and have utility in allowing manage- As to geography, mean reported prevalences of all head-
ment plans to be set meanwhile. This utility is lacking in ache types varied considerably, albeit with wide and
epidemiological studies, with subsequent enquiry rarely mostly overlapping CIs. Compared with HI superregion,
possible, but on the one hand excluding probable cases lower prevalences were estimated for all headache and
would clearly provide a partial account of headache prev- migraine in SSA and SEEAO, and for TTH in NAME.
alence while, on the other, probable cases are probably However, it is uncertain whether or to what extent
what they appear to be [4]. Most studies did not explic- these differences over time and place are real: overall,
itly state how ICHD criteria were applied, and, while the MLR analyses show that the present models explain
some gave estimates for both definite and probable cases, relatively little of the large variations in prevalence esti-
a few reported only the sum of these. We would, there- mates between studies (for migraine less than 30%, and
fore, have excluded the majority of studies had we lim- even less for other headache types, possibly because of
ited the review in this way, while grossly underestimating fewer studies). There are many other aspects of method-
true prevalences, as evidenced for migraine in particular ology (see below), some that are difficult to identify and
in Tables 4 and 6. In the MLR analyses (Table 7), this was impossible to quantify but which are likely or certain to
the factor with the highest ­SPCsqr, alone explaining 6.4% have varied between studies in different times and places,
(Model 1) to 7.8% (Model 2) of the variation in migraine and perhaps been influential. Standardisation accord-
prevalence. ing to accepted guidelines will help in future studies [4],
Stovner et al. The Journal of Headache and Pain (2022) 23:34 Page 14 of 17

but these factors, and the uncertainties they generate, neurologists are in many ways a highly selected and
will persist in the large corpus of historical studies from unrepresentative group.
mostly HI countries.
The relatively few studies from the other GBD super- Relation to age and gender
regions are vulnerable to incidental findings. There- This review mostly confirmed earlier studies on age and
fore, with the large and unexplained variations, it may gender distributions (Tables 1, 4 and 5). Headache in gen-
be risky (less accurate) to adjust the estimate of global eral was most prevalent in the group with mid-range age
prevalence for proportions of the global population liv- values between 10 and 19 years. Overall, and in females,
ing in each region. For example, SA and SEEAO have half migraine was most prevalent between 20 and 64 years,
the world’s population but only 9% (17/180) of headache whereas in males it was somewhat higher from 10 to
prevalence studies. Better estimates of global prevalences 19, although with overlapping 95% CIs. TTH was most
may be obtained by considering all studies as a way of prevalent in the 20–64 years age group for both genders
cluster sampling the world, although adjustment could (except for the single study in those above 64 years), as
and should still be made for the huge oversampling from was H15+. The age group 20–64 years is, of course, a very
HI countries. Arguably, the rich variety of methods used broad category, and this grouping only (and imprecisely)
make the overall estimates more resistant to yet unknown distinguishes adults from children and adolescents on the
sources of variation. one hand and from the elderly on the other. However, the
data did not allow a more refined analysis.
Relation to population of interest All headache types were more common in females,
This review included studies of all populations of inter- most markedly for migraine (17.0% versus 8.6%) and
est other than clinical (patient) populations. In contrast, H15+ (6.0% versus 2.9%), the more disabling types, while
the review from 2007 [5] included only studies with > 500 the difference was small and not significant in TTH. Gen-
participants on whole populations or representative sam- der differences were least in the youngest age groups
ples of these within specified age ranges in communi- (Table 1).
ties, towns or countries, together with studies performed The relationships between age and prevalences of dMig
in schools. This means we took account here of studies and pMig, and of dTTH and pTTH (Table 5), are worthy
on smaller and more selected populations (e.g., among of comment. The proportion of pMig to total migraine
employees of a company or factory, hospital staff, univer- was higher in the age group 10–19 years than in those
sity students, neurologists, ethnic minorities, inhabitants who were older, and also tended to be higher among
of a slum area, a tribe, a monastery, etc). Such popula- males. This suggests that ICHD migraine criteria are
tions were likely to differ from the general population in somewhat less fitting for children and adolescents than
the country in question (i.e., more or less healthy, poorer for adults, and perhaps for males than for females. The
or richer, more or less educated, etc) but, by including proportion of pTTH to total TTH was equal between
these, we gathered data from more diverse settings and genders, but again much higher in young people. In the
many more countries. Again, we believe this approach latter group, a high prevalence has been noted in several
results in a more representative sample of the world’s studies of headaches unclassifiable within ICHD, meeting
population, while, with one exception (see below), nei- the criteria for neither definite nor probable migraine or
ther bivariate nor MLR analyses found significant differ- TTH, and described as “undifferentiated headache” [24].
ences between estimates from selected or more general
populations. Comparisons with GBD2019
The exception was studies on health-care personnel For migraine and TTH, our estimates of global preva-
(medical students, hospital staff, neurologists). These lence and those of GBD2019 had widely overlapping
yielded significantly higher prevalence estimates for CIs (Table 2), but for all headache our estimate (52.0%
migraine in both bivariate and MLR analyses. The studies [48.9–55.4]) was markedly higher than that of GBD2019
on neurologists are an interesting example: three studies (35.0% [32.3–37.7]). Possibly, this is because, in GBD,
from USA [21], France [22] and Norway [23], considering ‘“all headache’” is the sum of migraine and TTH, whereas
either 1-year prevalence or “current” headache (unknown our estimates are based on studies among which the
timeframe), found a mean prevalence of 42.2%. It has majority asked about headache in general, therefore also
been argued that this very high prevalence is a true find- including other headache types. However, it is puzzling
ing because neurologists are expected to diagnose them- that in GBD2019 [13] (and earlier versions [25, 26]) total
selves with very high sensitivity and specificity [23]. This headache prevalence is markedly lower than the sum of
is not to say it reflects the true population prevalence: migraine and TTH prevalences. The explanation does
Stovner et al. The Journal of Headache and Pain (2022) 23:34 Page 15 of 17

not, apparently, lie in corrections for comorbidity of the afford more sensitive methods (personal interview, face
two headache types. to face or by phone) to detect cases. Alternatively, it may
GBD estimates are based on sophisticated mathematical be caused by greater selection bias in small studies. This
modelling [2], which combines methods of meta-analysis would explain the absence of similar effect on estimates
with regression techniques (“meta-regression”). It not only for TTH, since people whose lives are less impacted by
adjusts for gender and age composition of populations, for a disease have less interest in surveys enquiring into it.
comorbidity and for some methodological variables, but That more random sampling – generally more meth-
also takes account of multiple geographical levels (country, odologically rigorous – was negatively associated with
region, superregion, global) and year of study (“borrowing migraine prevalence estimates may similarly be explained
strength over space and time”) [27]. These are smooth- by reduced selection bias.
ing methods, whereby GBD is able to make estimates for It seems obvious that studies including pMig would
all gender- and age-groups, for smaller regions, individual report markedly higher estimates of migraine prevalence,
countries and even parts of countries. GBD estimates are since > 40% of total migraine prevalence in these stud-
more even across regions and time periods than ours: we ies was attributed to pMig. Including pTTH did not sig-
can make reasonable estimates for global prevalences over nificantly affect estimates of TTH, perhaps because cases
a long time span but, for countries, smaller regions and that might have been classified as TTH met criteria for
defined time periods, in the absence of data specific to pMig, counterbalancing the otherwise expected increase.
these, GBD methods achieve what we cannot. This is not That inclusion of pTTH was associated with a higher
a comment on which estimates might be truer in popula- prevalence of all headache (Model 2) suggests this cap-
tions from whom data have been directly derived in one or tured cases, probably of mild headache, that were other-
more methodologically sound studies. wise overlooked.
It also seems obvious that a screening question set-
Influences of methodological factors ting some threshold of severity, and therefore excluding
In the MLR analyses, publication year appeared impor- milder cases, would lead to lower prevalence estimates
tant as a factor explaining variation in migraine preva- of all headache (9.0% and 7.2% of variation) and of TTH,
lence estimates (6.4% of variation in Model 2, higher a less severe headache type (8.0% and 5.4% of variation),
estimates associated with more recent publication), but but not of migraine or H15 + .
it played no role in other headache types. The apparent
increase in migraine prevalence over time may be real, Lifetime and 1‑day prevalences
perhaps related to environmental, physical, behavioural Lifetime prevalences are higher than those of active head-
or psychological changes, but more probably it has to do ache disorders, except for H15+, but information about
with methodological developments over the years, lead- headache earlier in life is not relevant to estimations of
ing to better techniques of access and engagement and population burden deriving from active headache. In
improved diagnostic instruments, both likely to enhance addition, recall error is probably a bigger problem when
case ascertainment. The positive correlations between asking people about their whole lifetimes rather than
prevalence estimates of all diagnoses indicate that case recent time periods.
ascertainment is an important factor also for the other The opposites apply to 1-day prevalence. Enquiry
headache types, and for headache in general. Certainly into headache yesterday may almost eliminate recall
they speak against a hypothesis that, if one diagnosis is error. Although 1-day prevalence is substantially lower
made more often, it is at the expense of another. It may than 1-year prevalence, at least for episodic headache,
also be that better knowledge and greater awareness of it offers a very sound basis for measuring population
a condition lower the thresholds for reporting relevant burden, provided that samples are large enough. How-
symptoms. That knowledge of migraine and its diagnosis ever, questions about headache yesterday provide reli-
can affect estimates is supported by the findings of higher able data only in “on-the-spot” interviews (face-to-face
migraine prevalence among medical personnel, although or by telephone): in internet surveys or those relying
these are, largely, findings specific to the high end of the on mailed questionnaires, participants with headache
spectrum of knowledge. There has been no similar focus on the day they receive it may postpone answering it
on other common headache types. until the first day without, then truthfully but spuri-
The negative association of prevalence estimates of ously report headache yesterday. The relatively few
all headache, migraine and H15+ with number of study studies reporting 1-day prevalence suggest that 15.8% of
participants (Model 2: 3.2%, 1.6% and 12.5% of varia- the world’s population have headache on any day, and
tions respectively) may indicate that smaller studies can almost half of them migraine (7.0%).
Stovner et al. The Journal of Headache and Pain (2022) 23:34 Page 16 of 17

Study quality The fact that dependent variables were not normally dis-
We found no significant association between study tributed might have rendered the MLR analysis less reli-
quality score [4] and prevalence. This does not mean able. However, we believe that, by also presenting a model
that the quality criteria are meaningless: higher qual- with normalized variables, we have generated results that
ity lends credibility to findings. While the criteria may for explorative purposes are robust, although the exact size
need refinement, they are based on sound general epide- of the contribution of each variable remains uncertain.
miological principles, to which all future studies should The fact that < 30% of the variability could be explained
endeavour to adhere. However, higher scores on one cri- by the variables included in this analysis (time and place
terion may tend to increase prevalence estimates (e.g., of studies, age and gender of populations, and methodo-
distinction between definite and probable migraine), and logical variables) indicates that there are other influential
higher scores on another have the opposite effect (e.g., variables that are not yet understood or measured. We can
random sampling), with the net result being very small. only speculate as to what these may be: population differ-
This probably explains why quality score explains so little ences in genes, general health, quality of health services for
of the variability in our models (Table 7). headache, exposure to climatic factors, light or altitude,
nutritional factors, chemicals and air pollution, level of
stress, attitudes towards pain and knowledge about head-
Burden estimates ache in the population. Even differences in connotations in
While prevalence is a determining component of dis- the words used for headache or migraine may play a role.
ease-attributed burden, it provides no estimate of it in
the population. Headache-attributed burden not only
has multiple and diverse components [15] but also is Conclusions
very unevenly distributed in populations with current While this review updates our earlier documentation of
headache, being quite low in the relatively large propor- headache epidemiological studies [5], it also highlights
tions who have infrequent attacks of mild or moderate the dependence of prevalence estimates on a small num-
intensity [28]. Of the two further factors that must be ber of methodological factors (and relative independ-
known to estimate burden – the level of burden associ- ence of others that might be expected to be influential).
ated with the ictal (symptomatic) state and the propor- All future studies should use validated questionnaires,
tion of time spent in this state – both can be assessed as describe the screening question verbatim and explain
population averages. The former may be a constant at how they deal with definite and probable diagnoses. Ide-
population level for each headache type (the view taken ally, future prevalence estimates from all parts of the
in GBD studies, with “disability weights” assessing lost world will be derived from studies performed in a rela-
health), so that a sound estimate of population burden tively standardized way, in accordance with published
for all headache disorders may be based on the mean recommendations.
Time in Ictal State (TIS) (or Time in Symptomatic State Also, future studies should not assess prevalence alone
in the GBD studies) [29]. Estimates of disease-attributed but include data allowing TIS to be estimated, preferably
burden based on mean TIS may be relatively insensitive among the various age and gender subgroups.
to estimated prevalence if higher estimates of the latter
are generated by more searching enquiry, identifying Supplementary Information
The online version contains supplementary material available at https://​doi.​
participants with very low TIS. But all studies are likely org/​10.​1186/​s10194-​022-​01402-2.
to capture those with high burden, so that burden esti-
mates are highly sensitive to participation bias.
Additional file 1. Studies included in the analysis.
It is worth noting that most cases of pMig fail the attack
Additional file 2.
duration criterion (< 4 h for adults, < 2 h for children)
[30–32]. Since duration is a factor in TIS, inclusion or not
of probable diagnoses has rather less impact on burden Acknowledgements
None.
estimates than on prevalence estimates.
Authors’ contributions
Lars Jacob Stovner, Knut Hagen, Mattias Linde and Timothy J. Steiner partici-
Study limitations pated in the conception and design of the study, and in the analysis and inter-
pretation of the data. Lars Jacob Stovner, Knut Hagen and Mattias Linde also
Including not only strictly population-based prevalence participated in the acquisition of data. Lars Jacob Stovner wrote the first draft of
studies but also studies on all other non-clinical sam- the manuscript. The author(s) read and approved the final manuscript.
ples made a rules-based literature search strategy diffi-
Funding
cult. This might have introduced some subjectivity and The study was funded by the Norwegian University of Science and Technol-
unknown study-selection biases. ogy and St. Olavs Hospital, Trondheim, Norway.
Stovner et al. The Journal of Headache and Pain (2022) 23:34 Page 17 of 17

Availability of data and materials 13. IHME (2020) GBD Results Tool. http://​ghdx.​healt​hdata.​org/​gbd-​resul​ts-​
All data are available in the published studies referred to in the supplementary tool. Accessed 20 Mar 2021
material (Additional file 1). 14. Migliavaca CB, Stein C, Colpani V, Munn Z, Falavigna M (2020) Quality
assessment of prevalence studies: a systematic review. J Clin Epidemiol
127:59–68
Declarations 15. Steiner TJ, Gururaj G, Andree C, Katsarava Z, Ayzenberg I, Yu SY, Al Jumah
M, Tekle-Haimanot R, Birbeck GL, Herekar A, Linde M, Mbewe E, Manand-
Ethics approval and consent to participate har K, Risal A, Jensen R, Queiroz LP, Scher AI, Wang SJ, Stovner LJ (2014)
Not relevant. Diagnosis, prevalence estimation and burden measurement in popula-
tion surveys of headache: presenting the HARDSHIP questionnaire. J
Consent for publication Headache Pain 15:3
All authors have approved the submitted version of the paper and agreed to 16. Linde M, Edvinsson L, Manandhar K, Risal A, Steiner TJ (2017) Migraine
be accountable for their contribution. associated with altitude: results from a population-based study in Nepal.
Eur J Neurol 24:1055–1061
Competing interests 17. Linde M, Stovner LJ, Zwart JA, Hagen K (2011) Time trends in the preva-
The authors have no competing interest with regard to the present review. lence of headache disorders. The Nord-Trondelag Health Studies (HUNT 2
and HUNT 3). Cephalalgia 31:585–596
Author details 18. Wang SJ, Fuh JL, Juang KD, Lu SR (2005) Rising prevalence of migraine in
1
Department of Neuromedicine and Movement Science, NTNU Norwegian Uni- Taiwanese adolescents aged 13-15 years. Cephalalgia 25:433–438
versity of Science and Technology, Trondheim, Norway. 2Department of Neurology 19. Lipton RB, Scher AI, Kolodner K, Liberman J, Steiner TJ, Stewart WF (2002)
and Clinical Neurophysiology, Norwegian Advisory Unit on Headache, St Olavs Migraine in the United States: epidemiology and patterns of health care
University Hospital, Trondheim, Norway. 3NTNU Norwegian University of Science use. Neurology 58:885–894
and Technology, Trondheim, Norway. 4Tjörns Headache Clinic, Rönnäng, Sweden. 20. Lyngberg AC, Rasmussen BK, Jorgensen T, Jensen R (2005) Has the preva-
5
Division of Brain Sciences, Imperial College London, London, UK. lence of migraine and tension-type headache changed over a 12-year
period? A Danish population survey. Eur J Epidemiol 20:243–249
Received: 16 December 2021 Accepted: 9 February 2022 21. Evans RW, Lipton RB, Silberstein SD (2003) The prevalence of migraine in
neurologists. Neurology 61:1271–1272
22. Donnet A, Becker H, Allaf B, Lanteri-Minet M (2010) Migraine and migraines
of specialists: perceptions and management. Headache 50:1115–1125
23. Alstadhaug KB, Hernandez A, Naess H, Stovner LJ (2012) Migraine among
References Norwegian neurologists. Headache 52:1369–1376
1. Steiner TJ, Saraceno B, Prilipko L, McGillicuddy L, Shannon K, Tfelt-Hanm- 24. Wober C, Wober-Bingol C, Uluduz D, Aslan TS, Uygunoglu U, Tufekci A,
sen P, South V, Lainez M (2004) Lifting the burden: The global campaign Alp SI, Duman T, Surgun F, Emir GK, Demir CF, Balgetir F, Ozdemir YB, Auer
against headache. Lancet Neurol 3:204–205 T, Siva A, Steiner TJ (2018) Undifferentiated headache: broadening the
2. Stovner LJ, Nichols E, Steiner TJ et al (2018) Global, regional, and national bur- approach to headache in children and adolescents, with supporting evi-
den of migraine and tension-type headache, 1990-2016: a systematic analysis dence from a nationwide school-based cross-sectional survey in Turkey. J
for the Global Burden of Disease Study 2016. Lancet Neurol 17:954–976 Headache Pain 19:18
3. Steiner TJ, Stovner LJ, Jensen R, Uluduz D, and Katsarava (2020) Migraine 25. Vos T, Flaxman AD, Naghavi M, Lozano R et al (2012) Years lived with dis-
remains second among the world’s causes of disability, and first among ability (YLDs) for 1160 sequelae of 289 diseases and injuries 1990-2010: a
young women: findings from GBD2019. J Headache Pain 21: 137 systematic analysis for the Global Burden of Disease Study 2010. Lancet
4. Stovner LJ, Al Jumah M, Birbeck GL, Gururaj G, Jensen R, Katsarava Z, 380:2163–2196
Queiroz LP, Scher AI, Tekle-Haimanot R, Wang SJ, Steiner TJ (2014) The 26. GBD (2015) Disease and Injury Incidence and Prevalence Collabora-
methodology of population surveys of headache prevalence, burden and tors (2016) Global, regional, and national incidence, prevalence, and
cost: Principles and recommendations from the Global Campaign against years lived with disability for 310 diseases and injuries, 1990-2015: a
Headache. J Headache Pain 15:5 systematic analysis for the Global Burden of Disease Study 2015. Lancet
5. Stovner LJ, Hagen K, Jensen R, Katsarava Z, Lipton R, Scher A, Steiner T, 388:1545–1602
Zwart JA (2007) The global burden of headache: a documentation of 27. Global Health Data Methods. https://​globa​lheal​thdata.​org/​stati​stical-​
headache prevalence and disability worldwide. Cephalalgia 27:193–210 estim​ation-​techn​iques/ . Accessed 1 Nov. 2021
6. Woldeamanuel YW, Cowan RP (2017) Migraine affects 1 in 10 people world- 28. Linde M and Dahlöf C (2004) Attitudes and burden of disease among
wide featuring recent rise: A systematic review and meta-analysis of commu- self-considered migraineurs--a nation-wide population-based survey in
nity-based studies involving 6 million participants. J Neurol Sci 372:307–315 Sweden. Cephalalgia 24: 455–465
7. El-Metwally A, Toivola P, AlAhmary K, Bahkali S, AlKhathaami A, Al Ammar 29. Steiner TJ, Linde M, Schnell-Inderst P (2021) A universal outcome measure
SA, Altamimi IM, Alosaimi SM, Jawed M, Almustanyir S (2020) The Epide- for headache treatments, care-delivery systems and economic analysis. J
miology of Migraine Headache in Arab Countries: A Systematic Review. Headache Pain 22:63
ScientificWorldJournal 2020:4790254 30. Song TJ, Cho SJ, Kim WJ, Yang KI, Yun CH, Chu MK (2019) Sex Differences in
8. Abu-Arafeh I, Razak S, Sivaraman B, Graham C (2010) Prevalence of Prevalence, Symptoms, Impact, and Psychiatric Comorbidities in Migraine
headache and migraine in children and adolescents: a systematic review and Probable Migraine: A Population-Based Study. Headache 59:215–223
of population-based studies. Dev Med Child Neurol 52:1088–1097 31. Dent W, Spiss H, Helbok R, Matuja W, Scheunemann S, Schmutzhard E
9. Headache Classification Committee of the International Headache Soci- (2004) Prevalence of migraine in a rural area in South Tanzania: a door-to-
ety (IHS) (2018) The International Classification of Headache Disorders, door survey. Cephalalgia 24:960–966
3rd edition. Cephalalgia 38:1–211 32. Alp R, Alp SI, Palanci Y, Sur H, Boru UT, Ozge A, Yapici Z (2010) Use of the
10. Headache Classification Committee of the International Headache Soci- International Classification of Headache Disorders, Second Edition, crite-
ety (1988) Classification and diagnostic criteria for headache disorders, ria in the diagnosis of primary headache in schoolchildren: epidemiology
cranial neuralgias and facial pain. Cephalalgia 8(Suppl 7):1–96 study from eastern Turkey. Cephalalgia 30:868–877
11. Headache Classification Subcommittee of the International Headache
Society (2004) The International Classification of Headache Disorders. 2nd
Edition. 24: 1–160 Publisher’s Note
12. Hagen K, Zwart JA, Aamodt AH, Nilsen KB, Brathen G, Helde G, Stjern M, Springer Nature remains neutral with regard to jurisdictional claims in pub-
Tronvik EA, Stovner LJ (2008) A face-to-face interview of participants in lished maps and institutional affiliations.
HUNT 3: the impact of the screening question on headache prevalence. J
Headache Pain 9:289–294