Scher AI, Gudmundsson LS, Sigurdsson S, Ghambaryan A, Aspelund T, Eiriksdottir G, van Buchem MA, Gudnason V, Launer LJ. Migraine Headache in Middle Age and Late-Life Brain Infarcts. JAMA. 2009;301(24):2563-2570. doi:10.1001/jama.2009.932
Author Affiliations: Uniformed Services University, Bethesda, Maryland (Drs Scher and Ghambaryan); University of Iceland, Reykjavik, Iceland (Mr Gudmundsson); The Icelandic Heart Association, Kopavogur, Iceland (Mr Sigurdsson, Drs Aspelund and Gudnason, and Ms Eiriksdottir); Department of Radiology, Leiden University Medical Center, Leiden, the Netherlands (Dr van Buchem); and National Institute on Aging, Bethesda, Maryland (Dr Launer).
Context Migraine is considered to be an episodic condition with no long-term consequences. However, recent studies suggest that migraine attacks may be associated with pathologic changes in the brain, particularly in the cerebellum.
Objective To determine whether individuals not reporting headache compared with individuals reporting migraine symptoms, particularly aura, in midlife are at increased risk of late-life infarct-like lesions found on magnetic resonance imaging (MRI) without consideration of clinical symptoms.
Design, Setting, and Participants A population-based study of men and women in Reykjavik, Iceland (cohort born 1907-1935; n = 4689; 57% women) were followed up since 1967, examined, and interviewed about migraine symptoms in midlife (mean age, 51 years; range, 33-65 years). Between 2002 and 2006, more than 26 years later, brain MRIs were performed. Participants reporting headaches once or more per month were asked about migraine symptoms including nausea, unilateral location, photophobia, visual disturbance, and numbness. These individuals with headache were classified as having migraine without aura, migraine with aura, or nonmigraine headache. A comprehensive cardiovascular risk assessment was performed at both examinations.
Main Outcome Measure Presence of infarct-like lesions (total) and specifically located in the cortical, subcortical, and cerebellar regions.
Results Infarct-like lesions were present in 39.3% of men and 24.6% of women. After adjusting for age, sex, and follow-up time, compared with those not reporting headaches once or more per month (n = 3243), those with midlife migraine with aura (n = 361) had an increased risk of late-life infarct-like lesions (adjusted odds ratio [OR], 1.4; 95% confidence interval [CI], 1.1-1.8) that specifically reflected an association with cerebellar lesions in women (prevalence of infarcts 23.0% for women with migraine with aura vs 14.5% for women not reporting headaches; adjusted OR, 1.9; 95% CI, 1.4-2.6 vs a 19.3% prevalence of infarcts for men with migraine with aura vs 21.3% for men not reporting headaches; adjusted OR, 1.0; 95% CI, 0.6-1.8; P<.04 for interaction by sex). Migraine without aura and nonmigraine headache were not associated with an increased risk.
Conclusions Migraine with aura in midlife was associated with late-life prevalence of cerebellar infarct-like lesions on MRI. This association was statistically significant only for women. This is consistent with the hypothesis that migraine with aura in midlife is associated with late-life vascular disease in the cerebellum and in women.
Migraine, a common neurovascular disorder that affects approximately 11% of adults and 5% of children worldwide, is more common in women than men and is most prevalent in the third and fourth decades of life.1 Although a severe migraine attack is among the most disabling of neurological disorders,2 many individuals with migraine do not consult physicians.3
Approximately one-third of individuals with migraine experience neurological aura symptoms before headache onset (migraine with aura), usually consisting of transient visual, and also sensory, aphasic, or motor disturbances.4 Recent evidence suggests that migraine with aura is associated with an increased risk of clinically evident stroke or coronary artery disease.5- 9
Migraine has also been linked to silent infarct-like lesions (identified on magnetic resonance imaging [MRI] regardless of clinical manifestations) in a community-based cohort evaluated as a part of the CAMERA study,10 which showed that individuals with migraine had a 7-fold increased risk for infarcts in the cerebellum compared with controls, an association that was strongest in those with aura and frequent attacks (at least monthly).
Although the precise etiology linking migraine with aura and vascular disease is uncertain,5,6,11 the degree to which migraine is a marker or risk factor for brain changes that may have functional consequences in old age is a question of public health importance. We had the opportunity to study the relationship of midlife migraine symptoms and late-life infarct-like lesions (hereafter called infarcts) evident on MRI. The study is based on a large population-based cohort of men and women who have been followed up first as part of the Reykjavik Study and later as part of the Age Gene/Environment Susceptibility-Reykjavik Study (AGES-RS).12,13 We examined risk in men and women for infarcts in specific regions of the brain, and secondarily considered whether the risk varied by age at headache assessment and other established risk factors for vascular disease.
Detailed descriptions of the Reykjavik Study12,14 and AGES-Reykjavik Study13,15 have been published previously. In brief, the Reykjavik Study is a population-based cohort study established in 1967 by the Icelandic Heart Association to prospectively study cardiovascular disease in Iceland.12 The cohort included a random sample of men and women born between 1907 and 1935 and living in Reykjavik at baseline. In 2002, the Reykjavik Study continued as the AGES-Reykjavik Study to examine risk factors, genetic susceptibility, and gene-environment interactions in relation to disease and disability in old age.13 Of the 11 549 (58% women) surviving members of the Reykjavik cohort (representing 64% of the original examined cohort), 8030 (68.6% of men and 70.1% of women) were randomly selected and invited to participate in the AGES-Reykjavik Study. Of these individuals, 71.8% participated (74.0% of men and 70.2% of women), deriving a final sample of 5764 (58% women). Participants had a slightly better cardiovascular risk profile (lower midlife cholesterol, systolic blood pressure, and fewer smokers). Recruitment details and comparisons of the AGES-Reykjavik Study with the original cohort have been described.13
We refer to the assessments of relevance to this study from the Reykjavik Study as midlife assessments, and to those from the AGES-Reykjavik Study as late-life assessments. Midlife assessments included questions about headache, measurement of cardiovascular risk factors, and demographic characteristics. Late-life assessments included MRI of the brain, measurement of cardiovascular risk factors, and history of cardiovascular disease. The average year of the midlife assessment was 1978 with 90% occurring between 1972 and 1986. Late-life assessments (including MRI) were conducted from 2002 through 2006.
Headache. Participants were asked about current headache symptoms as part of the Reykjavik Study.16 Those reporting headache once or more per month were asked whether the headaches were accompanied by any of the following 5 features of migraine: nausea or vomiting, unilateral location, photophobia, visual disturbance during or preceding headache, and unilateral numbness preceding headache.
Demographic and Cardiovascular Factors. Cardiovascular risk assessment was performed at the midlife examination concurrently with the migraine assessment. The following variables were considered putative confounders or mediators: educational level (primary, secondary, college, university), self-reported current use of medication for hypertension, smoking history (never, former, current smoker), and history of diabetes, as well as measured body mass index, systolic blood pressure, total cholesterol, and fasting blood glucose.
Brain MRI Protocol. All eligible participants were offered a high-resolution brain MRI acquired on a study-dedicated 1.5-T Signa Twinspeed system (General Electric Medical Systems, Waukesha, Wisconsin). The image protocol consisted of the following pulse sequences: T1-weighted 1.5-mm slice thickness 3-dimensional spoiled gradient-echo sequence (echo time [TE], 8 milliseconds; repetition time [TR], 21 milliseconds; flip angle (FA), 30°; field of view [FOV], 240 mm; matrix, 256 × 256) and in addition, with 3-mm thick interleaved slices, a proton density (PD)/T2-weighted fast spin-echo (SE) sequence (TE1, 22 milliseconds; TE2, 90 milliseconds; TR, 3220 milliseconds; echo train length, 8; FA, 90°; FOV, 220 mm; matrix, 256 × 256), and a fluid-attenuated inversion recovery (FLAIR) sequence (TE, 100 milliseconds; TR, 8000 milliseconds; inversion time, 2000 milliseconds; FA, 90°; FOV, 220 mm; matrix, 256 × 256). All images were acquired to give full brain coverage and slices were angled parallel to the anterior-posterior commissure line to give reproducible image views in the oblique axial plane.
Image Analysis. Infarcts were evaluated based on the T2-weighted fast SE/PD images and FLAIR images.
Infarcts were defined based on radiologic characteristics as described. A parenchymal defect (infarct) was defined as a defect of the brain parenchyma with a signal intensity that is isointense to that of cerebrospinal fluid on all pulse sequences (ie, FLAIR, T2-weighted, PD-weighted). Cortical infarct-like lesions were defined as parenchymal defects involving or limited to the cortical ribbon and surrounded by an area of high signal intensity on FLAIR images. Subcortical infarcts were defined as parenchymal defects not extending into the cortex that are surrounded by an area of high signal intensity on FLAIR images. Defects in the subcortical area without a rim or area of high signal intensity on FLAIR, and without evidence of hemosiderin on the T2*-weighted GRE-EPI scan were labeled as large Virchow-Robin Spaces (VRS); these were excluded from the definition of subcortical infarcts. Defects in the subcortical area with evidence of hemosiderin on the T2*-weighted GRE-EPI scan were labeled as resorbed hematomas and were also excluded from the definition of subcortical infarcts. Lesions 4 mm or larger were recorded except for those in the cerebellum, for which there was no size criterion. Infarcts that spanned 2 areas were assigned to the location with the largest measured (mm) diameter of the defect regardless of orientation. This protocol was comparable with the protocol used in the CAMERA study.10
Image analyses were performed in a 2-step procedure by readers blinded to participant health status, including midlife headache history. An experienced neuroradiologist examined the scan for clinical abnormalities that needed immediate attention. At the same time, the neuroradiologist recorded the slice location of observed cortical and cerebellar infarcts directly into a shared database. Trained raters with access to the shared database identified subcortical infarcts and characterized all of the infarcts in more radiologic detail. Quality control procedures included 6 monthly assessments of intraobserver variability, and 3 monthly assessments for interobserver differences. The intraobserver weighted κ statistic was 0.92 for cerebral infarcts; the interobserver weighted κ statistic was 0.66 for cerebral infarcts. In addition, a 5% random sample was reread by a trained neuroradiologist at Leiden University Medical Center, Leiden, the Netherlands, and differences discussed.
Late-Life Cardiovascular Risk Factors and Disease. Late-life measurements included carotid artery distensibility by ultrasonography and coronary calcification (Agatston units) measured by computed tomography; both these measures were categorized into sex-specific quartiles. Diabetes was defined based on self-reported history of diabetes, use of medication, or fasting glucose levels greater than 126 mg/dL; systolic blood pressure was taken twice and averaged for the final measure; and standard questions were administered to assess smoking history (never, former, current) and history of physician diagnosis of stroke and transient ischemic attack (TIA). History of coronary artery disease (CAD) was defined as a self-reported physician diagnosis of myocardial infarction or angina, or a history of coronary angioplasty or coronary artery bypass graft with supporting evidence from electrocardiography or nitrate use.
Analytic Sample. Of the 5764 AGES-Reykjavik participants, 5003 underwent MRI. Reasons for nonparticipation included contraindications (n = 280), refusal (n = 283), or being examined via home visit rather than clinic (n = 198). An additional 237 participants were not included in the analysis because of either not completing all image sequences needed for infarct assessment or insufficient scan quality for infarct assessment. We excluded an additional 77 individuals who were older than 65 years at the time of the midlife examination. The final analytic sample thus consists of 4689 surviving Reykjavik Study participants who had complete headache and MRI data. Those excluded were older (79 vs 76 years of age), had a higher midlife systolic blood pressure, and a higher prevalence of CAD, stroke, or TIA in late life (P < .001 for all). Sex or midlife migraine status did not differ between individuals who were included and excluded from these analyses.
Statistical Analyses. Based on the midlife headache questions, we classified participants into 4 mutually exclusive headache categories: No headache once or more per month (reference category), nonmigraine headache, migraine without aura, and migraine with aura. The migraine without aura category included individuals with headache with at least 2 of the 3 nonaura symptoms (nausea, unilateral location, photophobia). The migraine with aura category included those reporting visual aura, sensory aura, or both. Individuals with headache but no nonaura symptoms or 1 nonaura symptom (nausea, unilateral location, or photophobia) were defined as having nonmigraine headache. Aura symptoms took precedence over other symptoms.
The classification scheme represents an approximation of International Headache Society diagnostic criteria for migraine with or without aura, which were formalized after the midlife data were collected.17 International Headache Society features for migraine without aura that are missing from these criteria include pulsatility, exacerbation with activity, and phonophobia. International Headache Society criteria for migraine with aura missing from these criteria include duration of aura (aura symptoms must last between 5 and 60 minutes) and speed of onset (aura symptoms must develop gradually over more than 5 minutes). Due to the screening question for headache, our case definition does not include individuals who experience aura exclusively without headache.
A priori analyses were conducted for the total sample and stratified by sex. We used logistic regression to estimate the odds (95% confidence interval [CI]) of prevalent late-life infarcts in those with midlife migraine symptoms relative to individuals without midlife migraine symptoms. Separate models were calculated for cerebellar, cortical, subcortical, and total infarcts for the total sample and by sex. In model 1, we adjusted for age at the midlife examination, sex (for analyses on the total sample), and duration of follow-up. In model 2 we additionally adjusted for possible confounding by midlife cardiovascular factors. We tested for sex differences in the relationship between midlife migraine and late-life infarcts by including an interaction term in model 1 and model 2 (eg, migraine × sex).
In secondary analyses, we adjusted for late-life cardiovascular risk factors and stratified by a history of CAD or TIA/stroke, to examine whether the associations of migraine to infarcts were changed by these factors. We tested for interaction by the age at which migraine symptoms were assessed (age <50 years, age ≥50 years), CAD, and TIA/stroke history by including interaction terms as previously shown. All analyses were performed with Stata version 10.1 (StataCorp LP, College Station, Texas).
The AGES-Reykjavik Study was approved by the Icelandic National Bioethics Committee (VSN-00-063), which acts as the institutional review board for the Icelandic Heart Association and by the institutional review board for the US National Institute on Aging, National Institutes of Health. Written informed consent was obtained from all participants.
Participants were 2693 women and 1996 men with an average age of 50.9 years (range, 33-65) at the midlife interview and 76.2 years (range, 66-96) at the late-life interview (Table 1). Overall, 12.2% of the participants (5.7% of men; 17.0% of women) were classified as having migraine, including 4.5% migraine without aura (1.5% of men; 6.6% of women) and 7.7% migraine with aura (4.2% of men; 10.3% of women). Among participants with aura, the proportion with visual aura, sensory aura, and both visual and sensory aura, respectively, was 77.1%, 14.5%, and 8.4% for men and 66.2%, 17.3%, and 16.5% for women. Within the migraine with aura group, 89% reported having at least 1 other migraine symptom.
Individuals with migraine were slightly younger at the midlife examination compared with others (Table 1). Other differences were that women with migraine with aura were more likely to report a history of CAD or TIA/stroke than those without (P < .005), although most other measures of cardiovascular risk were not obviously different.
Infarcts were present on MRI in 39.3% of men and 24.6% of women. The most common lesion location was the cerebellum (21.0% in men and 14.7% in women; Table 2).
In unadjusted comparisons, infarcts overall were more prevalent in women with migraine with aura compared with women without headache (31% vs 25%; P = .04; Table 2) but there was no difference in prevalence for men (41% vs 39%). Infarcts in the cerebellum, but not in other locations, were more prevalent in women with migraine with aura compared with women without headache (23% vs 15%; P < .001); there was no difference in prevalence for men (19% vs 21%).
After adjusting for age, sex, and follow-up time in a pooled model for men and women, participants with midlife migraine with aura were at increased risk for total infarcts (adjusted odds ratio [OR], 1.4; 95% CI, 1.1-1.8; Table 3). This mainly reflects the risk associated with lesions located in the cerebellum (adjusted OR, 1.6; 95% CI, 1.3-2.2; Table 3). There was no increased risk for cortical or subcortical lesions (Table 3) for participants with midlife migraine with aura, migraine without aura, or nonmigraine headache. Results were similar without (model 1) or after (model 2) adjustment for midlife measures of cardiovascular risk.
The relationship between migraine with aura and cerebellar infarcts was only significant in women (men, adjusted OR, 1.0; 95% CI, 0.6-1.8 vs women, adjusted OR, 1.9; 95% CI, 1.4-2.6; P = .04 for interaction by sex; Table 3), but was not statistically different by the age at which headache symptoms were assessed (age <50 years, adjusted OR, 2.0; 95% CI, 1.4-3.0 vs age ≥50 years, adjusted OR, 1.4; 95% CI, 0.9-2.0; P = .18 for interaction by age, Table 4).
For cortical infarcts in the migraine without aura group, there was an interaction by sex, suggesting a higher risk in men compared with women (P = .04), although the individual sex-stratified ORs were not significant (Table 4). Results were generally similar when stratified by age (Table 4), although there was also a marginally increased risk for cortical infarcts in participants aged ≥50 years with migraine with aura (adjusted OR, 1.6; 95% CI, 1.0-2.5; P = .07).
Results were similar after adjusting for late-life measures of cardiovascular risk and history of CAD or TIA/stroke. The relationship between migraine with aura and cerebellar infarcts was not changed by adjustment for late-life measures of cardiovascular risk and history of CAD or TIA/stroke in the total sample (adjusted OR, 1.5; 95% CI, 1.2-2.0) or when analyzed separately for men (adjusted OR, 1.0; 95% CI, 0.5-1.7) and women (adjusted OR, 1.8; 95% CI, 1.3-2.5). The association did not differ by CAD history (interaction, P<.13) with no CAD history having an adjusted OR of 1.8 (95% CI, 1.2-2.5) and with CAD history having an adjusted OR of 1.2 (95% CI, 0.8-1.9). The relationship did not differ by history of TIA or stroke (no history, adjusted OR, 1.7; 95% CI, 1.2-2.3; vs with history, adjusted OR, 1.6; 95% CI, 0.8-3.5; P = .57 for interaction by TIA/stroke history).
The separate analyses of visual and sensory aura symptoms suggested that the association of cerebellar infarcts with migraine with aura in women was stronger in those (8.6% of all women) with visual aura (adjusted OR, 2.2; 95% CI, 1.5-3.1) compared with those (1.7% of all women) with only sensory aura symptoms (adjusted OR, 1.3; 95% CI, 0.6-2.8).
In a large cohort of Icelandic adults, we found that women who reported migraine with aura in middle age were at increased risk of late-life infarcts relative to those without migraine symptoms. The risk was primarily for cerebellar lesions; there was no increased risk for cortical or subcortical lesions in these women or for those with migraine without aura or nonmigraine headache.
This risk was independent of cardiovascular risk factors measured in midlife or late life. Risk was not statistically different between individuals who were aged 50 years or younger vs those who were older when headache was ascertained or between those with a history of diagnosed CAD or TIA/stroke vs those without.
Our study has substantial strengths. The large well-characterized cohort was established in 1967 when, at the time of headache assessment, participants were aged 33 to 65 years. At those ages, many participants were still experiencing migraines, therefore recall bias is likely reduced. Participants were also at low risk for TIA or stroke, making the identification of migraine visual aura symptoms more robust. Measurement of late-life infarcts on MRI was performed by raters blinded to midlife headache status. Because participants were followed up as part of a cardiovascular disease study, we were also able to rigorously adjust for plausible confounding cardiovascular risk factors. Other strengths include the size of our cohort and broad age range, which gave us statistical power to consider sex, age, and cardiovascular disease in our analyses.
Some limitations of this study should be taken into account when interpreting the findings. Because migraine symptom questions were not asked of those reporting headache less than once per month, we are likely capturing only those with severe migraine occurring with a higher frequency. Participants with aura only and no headache would be classified as having no migraine. Further, our assessment of migraine was based on pre-IHS diagnostic criteria, although the questions addressed 5 symptoms included in the IHS guidelines. We note that our estimated prevalence of migraine overall (eg, with or without aura) is highly consistent with prior studies.1 Our prevalence of aura (as a proportion of the total migraine population) is higher than has been reported in other population studies and may include frequently occurring nonspecific visual symptoms such as blurring. However, the likely effect of this misclassification would be to attenuate the relationship between migraine with aura and infarcts, unless, compared with aura, nonspecific symptoms are differentially more strongly related to the risk for infarcts, a hypothesis we believe is unlikely.
Given the age of our study population, it is worth considering the extent to which overall or cardiovascular-related mortality may have affected our results. In particular, those with migraine with aura have been reported to be at increased risk of cardiovascular death compared with others.7 If individuals with midlife migraine with aura were more likely to die of cardiovascular disease before the late-life examination, and if these individuals were also more likely to have infarcts in the cerebellum or overall compared with others, then our results would have been attenuated. However, if these cerebellar or overall lesions were somehow protective (eg, individuals with migraine with aura and these lesions had lower all-cause mortality compared with those with migraine with aura without these lesions), then our results would have been exaggerated. The second scenario seems unlikely.
Our results are consistent with the cross-sectional CAMERA study,10 the only other study that measured infarcts on MRI, which also found the migraine-associated infarcts to be preferentially located in the cerebellum. This prospective longitudinal study had a long follow-up and an older cohort with a much higher background risk for brain lesions. Our results suggest that the association of infarcts with migraine with aura is detectable in older individuals who typically have cardiovascular risk factors that lead to similar-appearing lesions.18 Further, the study is based on a larger sample of men and women, therefore, sex differences could be investigated. We found that the relationship between migraine with aura and cerebellar infarcts may be specific to women. However, we cannot rule out a possible increased risk for men with migraine with aura due to the relatively small number of men with migraine with aura in our sample.
Why migraine, particularly with aura, is associated with clinical and silent (presumed) ischemic stroke is uncertain. Proposed mechanisms include atherosclerotic and nonatherosclerotic causes,5,6,11 including traditional cardiovascular risk factors,11,19 endothelial dysfunction,11,20- 22 shared genetic risk factors for migraine and stroke,11,23- 25 vasoconstrictor medications taken to treat headache,11,22 cardiac abnormalities including patent foramen ovale,11,26 and diagnostic artifact,11,27 among other factors. These mechanisms do not obviously explain why infarcts associated with migraine with aura would be preferentially located in the cerebellum and in women. There are clinical reports suggesting that the cerebellum is vulnerable in individuals with migraine28- 32 and in familial hemiplegic migraine—a rare Mendelian variant of migraine with aura.33 In population studies, no particular location pattern was evident for clinically evident ischemic stroke among women with aura,9,34 although as mentioned earlier, silent infarcts (as per the CAMERA study) were preferentially located in the cerebellum.10 We also note that secondary analyses suggested an association of migraine with aura to cortical infarcts in some subgroups was stronger (eg, men with migraine with or without aura or men and women who were older than aged 50 years at the time of headache assessment).
In summary, this study suggests that a remote history of migraine with aura is associated with brain lesions commonly found in older populations. Results persisted after controlling for cardiovascular risk factors and history of cardiovascular disease, thus suggesting that the mechanism linking the migraine aura with these lesions is independent of the usual risk factors for ischemic vascular disease and may be specifically related to migraine with aura. Additional longitudinal studies with repeated MRIs are needed to better establish the temporality and dose-response relationship between migraine with aura and brain infarcts. Finally, the clinical implications of the infarct-like lesions identified have not been established and will require investigation.
Corresponding Author: Lenore J. Launer, PhD, LEDB/NIA/NIH, 7201 Wisconsin Ave, Ste 3C-309, Bethesda, MD 20892 (email@example.com).
Author Contributions: Dr Scher had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Scher, Eiriksdottir, van Buchem, Gudnason, Launer.
Acquisition of data: Sigurdsson, Eiriksdottir, Gudnason.
Analysis and interpretation of data: Scher, Gudmundsson, Sigurdsson, Ghambaryan, Aspelund, van Buchem, Gudnason.
Drafting of the manuscript: Scher.
Critical revision of the manuscript for important intellectual content: Gudmundsson, Sigurdsson, Ghambaryan, Aspelund, van Buchem, Eiriksdottir, Gudnason, Launer.
Statistical analysis: Scher, Gudmundsson, Ghambaryan, Aspelund, Launer.
Obtained funding: Eiriksdottir, Gudnason.
Administrative, technical, or material support: Sigurdsson, Eiriksdottir, van Buchem, Gudnason.
Study supervision: Sigurdsson, van Buchem, Gudnason, Launer.
Financial Disclosures: Dr Scher has served on advisory boards for Endo Pharmaceuticals and OrthoMcNeil Neurologics. The other authors report no financial disclosures.
Funding/Support: This study was funded by National Institutes of Health contract N01-AG-12100, the National Institute on Aging Intramural Research Program, Hjartavernd (the Icelandic Heart Association), and the Althingi (the Icelandic Parliament). Components of the study were also supported by the National Eye Institute, the National Institute on Deafness and Other Communication Disorders, and the National Heart, Lung, and Blood Institute. Funding in support of this analysis was provided by the Migraine Research Foundation.
Role of the Sponsor: None of the funding bodies had any role in the study design or conduct; data collection, management, analysis, or interpretation; or preparation, review, or approval of the manuscript.