Abstract
Introduction
Observations that migraine increases risk of cardiovascular disease and ischemic brain changes may suggest sustained vascular differences between migraineurs and controls. In a population-based setting, we compared cerebral blood flow between migraineurs in the attack-free period and controls.
Methods
Between 2006 and 2008, 2642 participants, aged 45–65, from the Rotterdam Study completed a migraine questionnaire and had complete usable MRI data. Participants were classified into controls (N = 2033), probable migraine (N = 153), or migraine (N = 456). Using 2D phase contrast MRI, we performed a cross-sectional analysis of the effect of migraine on total cerebral blood flow (tCBF), parenchymal cerebral blood flow (pCBF), and blood flow in each intracranial arterial using linear regression. Additionally, we performed stratified analysis of subtypes of migraine.
Results
Compared with controls, migraineurs had higher pCBF (1.07 ml/min/100 ml, 95% CI 0.08; 2.05). In particular, migraineurs had significantly higher blood flow in the basilar artery (4.70 ml/min, 95% CI 0.77; 8.62).
Discussion
Migraineurs in the attack-free period have higher pCBF, particularly basilar artery flow, compared to controls, supporting the notion of sustained vascular differences between these groups outside of migraine attacks.
Introduction
In recent years, evidence has mounted that migraineurs are at increased risk for a number of cardiovascular and ischemic cerebrovascular outcomes, particularly in the posterior circulating territory (1–9). Rather than a direct causal relationship with migraine, the combination of vascular outcomes associated with migraine is suggestive of systemic vascular differences between migraineurs and controls (10,11).
One way to explore vascular differences is through measuring blood flow to the brain. Over several decades, many studies have described blood flow and perfusion changes in migraineurs during attacks (12–17). Yet, in view of the above-described vascular outcomes, blood flow differences in attack-free periods are of particular interest. Only a few studies, using various methods, have studied cerebral blood flow (CBF) and perfusion differences between interictal migraineurs and controls, but the results thus far have been inconclusive (10,13,18–21). Most of these studies have been conducted in clinical settings with limited sample size. Moreover, studies on sustained blood flow differences in migraineurs after a long period without an attack are lacking. Thus, the question remains whether changes in CBF and perfusion in migraineurs are sustained in between migraine attacks and over long periods of time after attacks cease.
In this study, we explored differences in total cerebral blood flow (tCBF) and parenchymal cerebral blood flow (pCBF) between migraineurs in a headache-free period and controls, aged 45–65 from the population-based Rotterdam Study. To assess potentially different blood flow patterns in subtypes of migraineurs, we additionally distinguished between migraineurs who were <1 year without an attack and >1 year without an attack as well as migraineurs with and without aura. Finally, we explored blood flow in the individual large intracranial cerebral arteries in migraineurs compared to controls, giving particular notice to the posterior circulating territory.
Methods
Setting and study population
For this study, we performed a cross-sectional analysis on participants of the Rotterdam Study, a prospective population-based cohort study among community-dwelling middle-aged and elderly in Ommoord, the Netherlands (22). The original cohort of the Rotterdam Study began in 1990 and an expansion cohort began in 2000, which both included participants 55 and older. In 2006, the study expanded again, which is known as Rotterdam Study III (RS-III) and included 3932 individuals, this time who were 45 years of age or older. The migraine questionnaire was introduced to participant interviews in RS-III. Between February 2006 and December 2008, participants from RS-III underwent their baseline visits, which consisted of a home visit by trained interviewers and physical examination in the research facility (22). Since 2005, a prospective brain magnetic resonance imaging (MRI) study has been included as part of the protocol for the larger Rotterdam Study (23), and all eligible participants from RS-III were invited to undergo an MRI scan.
This current study included individuals aged 45 to 64 years from RS-III. In this age range, 3582 participants were invited, and 3577 people underwent baseline interview, but 21 participants were excluded for missing data on migraine characteristics. This left 3556 individuals with complete migraine data. Of these, 2725 underwent an MRI scan (597 ineligible; 254 non-responders). We excluded 47 participants with prevalent cortical infarcts because these infarcts affect MR image processing, making the image-derived data less reliable. Additionally, we excluded 30 participants because of bad scan quality (e.g. excessive movement or large imaging artifacts) and three participants because of incorrect positioning in scans that prevented CBF measurements. We excluded an additional three people for biological conditions leading to abnormal flow values (one arteriovenous malformation leading to extremely high right carotid flow; two arterial occlusions leading to no flow in an artery). This left a total of 2642 participants with complete information on migraine characteristics and usable MRI scans for our analysis.
Assessment of migraine
The migraine questionnaire was a component of the home, face-to-face structured interview. For logistic reasons, 634 people received the migraine questionnaire by a phone interview instead of a face-to-face interview. The migraine questionnaire was based on the revised International Headache Society (IHS) criteria of 2004 (International Classification of Headache Disorders, second edition (ICHD-II)) (24) and was a modified version of the questionnaire used in the Genetic Epidemiology of Migraine (GEM) study of Leiden (see online Supplementary Table S1) (25).
All individuals who met the criteria for lifetime history of migraine (with or without aura) were classified as migraine. In our questionnaire, there are two modifications from the ICHD-II criteria (24). First, participants were first asked if they had ever experienced a headache with severe pain that affected their daily activities. If the answer was negative or if the participant clearly indicated that severe headache was due to other causes, such as a tumor, sinusitis, stroke, trauma or meningitis, no further questions on headaches were asked. Thus, with this survey, we miss participants who might have migraine without aura with moderately painful headaches as well as participants who might have aura without migraine headache.
Second, our classification for migraine with aura was modified from the ICHD-II criteria because of limited questions related to aura in our survey (see online Supplementary Table S2). We classified migraine with aura as individuals who met all criteria for migraine without aura and also reported headache attacks accompanied by visual or sensory aura symptoms that lasted more than 5 minutes and less than 60 minutes. This modification did not affect the distinction between migraineurs and controls, but may have affected the distinction between migraine with and without aura.
For sub-analyses, migraineurs (with and without aura) were dichotomized into <1 year since last attack or >1 year since last attack. Probable migraine was defined as having headache attacks that fulfilled all but one of the criteria A–D for migraine without aura.
Brain MRI protocol
We performed a multi-sequence MRI protocol on a single 1.5 T scanner (23). Participants with migraine were attack free at the time of the scan. For assessment of flow, two-dimension (2D) phase-contrast imaging was performed (26). A sagittal 2D phase-contrast MRI angiographic scout image was performed (repetition time (TR) = 24 ms, echo time (TE) = 9 ms, field of view (FOV) = 32 × 32 cm2, matrix = 256 × 160, flip angle = 10 degrees, number of signals averaged (NEX) = 1, bandwidth (BW) = 8.06 kHz, velocity encoding = 60 cm/sec, slice thickness = 60 mm, acquisition time = 12 seconds). From this image, a transverse imaging plane perpendicular to both the precavernous portion of the internal carotid arteries and to the middle part of the basilar artery was chosen for a 2D gradient-echo phase-contrast sequence (TR = 20 ms, TE = 4 ms, FOV = 19 × 19 cm2, matrix = 256 × 160, flip angle = 80, NEX = 8, BW = 22.73 kHz, velocity encoding = 120 cm/second, slice thickness = 5 mm, acquisition time = 51 seconds). No cardiac gating was performed (26). Additionally, T1-weighted (T1w), proton density-weighted (PDw), and fluid-attenuated inversion recovery (FLAIR) sequences were included in the protocol (23). Infarcts were rated visually on these sequences and categorized as cortical and lacunar. Brain volume was assessed using a fully automated process previously described elsewhere (23). Briefly, the T1w, PDw and FLAIR sequences of each participant were first co-registered and then underwent non-uniformity correction and variance scaling. A k-nearest neighbor classifier was used to classify scans into brain tissue and cerebrospinal fluid (23). Using elastix software (27), a manually segmented brain mask (excluding the cerebellum, eyes and skull) was non-rigidly registered to participants’ T1w image. Total brain volume (TBV) was calculated by summing all voxels of the gray matter, white matter, and white matter lesions across the brain, resulting in volumes in ml.
CBF and pCBF
Using interactive data custom software (Cinetool version 4, General Electric Healthcare, Milwaukee, WI, USA), CBF was calculated from the phase-contrast images. Circular to elliptical regions of interest (ROIs) were drawn around both carotid arteries and the basilar artery on the phase-contrast images (Figure 1). The entire lumen of the vessel was incorporated in the ROIs. Velocity in the vessel (cm/second) was determined as the mean signal intensity in each ROI. To calculate arterial flow, the average velocity was multiplied by the cross-sectional area of the vessel. Flow rates from each artery were summed and multiplied by 60 sec/min to calculate tCBF. Two experienced technicians performed all ROI drawings and flow measurements (26). For 533 scans, double rating was performed and yielded inter-rater correlations of >0.94 for all vessels (26). To calculate pCBF (ml/min per 100 ml), each individual’s tCBF was divided by his or her TBV (ml) and the result was multiplied by 100 (26).
Two-dimensional (2D) phase-contrast magnetic resonance imaging (MRI) used to measure flow. (a) Sagittal 2D phase-contrast scout image used to localize the phase-contrast imaging plane, which is denoted by the white line perpendicular to the left and right carotid and basilar arteries. (b) 2D phase-contrast MRI (orientation indicated by the white line in (a)), which shows the right carotid (1), left carotid (2), and basilar artery (3) in black. White outlines around the vessels are drawn manually. For improved visualization of the vessel boundaries, the contrast between the arteries and the background in panel (b) is inverted compared to panel (a).
Covariates
Smoking status and alcohol use were self-reported during participant interviews. Heavy alcohol consumption in this study was considered drinking alcoholic beverages four or more days per week. Smoking was categorized as past, current, or never smokers. Serum cholesterol, fasting glucose levels, systolic and diastolic blood pressure as well as medication status were all obtained during participants’ center visit and were included as continuous measures. Diabetes mellitus was classified as people with fasting glucose level ≥7.0 mmol/l or the use of antidiabetic medication. Hypertension was classified as individuals with systolic blood pressure ≥140 or diastolic blood pressure ≥90 or the use of antihypertension medicine.
Statistical analysis
We assessed the effect of migraine on tCBF and pCBF using multiple linear regressions comparing any migraine to controls. In addition to analyses of any lifetime migraine, we compared controls to two sub-type stratifications of migraine: first, time since last migraine attack as <1 year or >1 year, and second, migraine with and without aura. Individuals with probable migraine were compared to controls separately. Subsequently, we assessed the effect of any migraine and migraine subtypes on blood flow in the basilar artery and the right and left carotid arteries compared to controls. For basilar flow, we also assessed potential interaction between sex and migraine through an additional model with an interaction term as well as a sex-stratified analysis. For each analysis, we used two models: Model 1 was adjusted for age and sex; Model 2 was additionally adjusted for total brain volume (except in pCBF analyses), presence of lacunar infarcts, systolic blood pressure, serum cholesterol, hypertension, diabetes, smoking status, and heavy alcohol drinking. All analyses were based on complete cases. Statistical significance was considered two-sided p values <0.05. Neither the reported p values nor the significance threshold were adjusted for multiple testing. All analysis were performed using the statistical software package SPSS (Chicago, IL, USA), version 21.0 for Windows.
Results
Characteristics of the study population.
Values are presented either as mean ± SD or number (%). Data missing for: blood pressure (N = 9), cholesterol (N = 30), diabetes (N = 27).
Total cerebral blood flow and total brain perfusion across migraine categories using linear regression models.
Values are differences (95% confidence intervals (CI)) between each migraine category and controls. Model 1 was adjusted for age and sex. Model 2 was additionally adjusted for presence of lacunar infarcts, systolic blood pressure (mmHg), total serum cholesterol (mmol/l), hypertension, diabetes, current smoking, and past smoking, heavy alcohol (consuming alcohol >3 times/week). Models for total cerebral blood flow were additionally adjusted for total brain volume (ml). R2 values denote the variance explained by the fitted model. aParenchymal cerebral blood flow is calculated as: total cerebral blood flow (tCBF) (ml/min)/total brain volume (ml)* 100. bThe estimated marginal mean (EMM) of tCBF of controls Model 1 was 562.2 ml/min, standard error 2.2. The EMM of tCBF of controls in Model 2 was 561.5 ml/min, standard error 2.0. The EMM of parenchymal cerebral blood flow (pCBF) of controls Model 1 was 58.6 ml/min per 100 ml, standard error 0.2. The EMM of pCBF of controls in Model 2 was 58.5 ml/min per 100 ml, standard error 0.2.
Blood flow in cerebral arteries across migraine categories using linear regression models.
Values are differences (95% confidence intervals (CI)) between each migraine category and controls. Model 1 was adjusted for age and sex. Model 2 was additionally adjusted for total brain volume (ml), presence of lacunar infarcts, systolic blood pressure (mmHg), total serum cholesterol (mmol/l), hypertension, diabetes, current smoking, and past smoking, heavy alcohol (consuming alcohol >3 times/week). R2 values denote the variance explained by the fitted model. aThe estimated marginal mean (EMM) of basilar artery flow of controls Model 1 was 109.9 ml/min, standard error 0.8. The EMM of basilar artery flow of controls Model 2 was 109.8 ml/min, standard error 0.8. The EMM of right carotid artery flow of controls in Model 1 was 228.1 ml/min, standard error 1.2. The EMM of right carotid artery flow of controls in Model 2 was 227.7 ml/min, standard error 1.1. The EMM of left carotid artery flow of controls Model 1 was 224.4 ml/min, standard error 1.2. The EMM of left carotid artery flow of controls Model 2 was 224.1 ml/min, standard error 1.1.
When we stratified the study population by sex (Supplementary Table 2), we found the difference in basilar flow between migraineurs and controls both in males and females was consistent with estimates seen in the overall analysis., In both sexes, the largest differences in basilar flow were observed in migraineurs with <1 year without an attack. We found no significant statistical interaction of migraine and sex in basilar flow between any migraine compared to controls (p = 0.71) (data not shown).
In all analyses, probable migraine did not differ significantly from controls but had a tendency toward decreased flow compared to controls.
Discussion
In a population of community-dwelling middle-aged and elderly individuals, migraineurs in the attack-free period had higher tCBF and pCBF compared to controls, with the latter difference reaching statistical significance. The higher CBF seemed to be driven by higher basilar artery flow in migraineurs, particularly migraineurs with <1 year since a last attack, compared to controls.
A major strength of our study is the large sample size of community-dwellers, which reduces the likelihood of selection bias and facilitates the assessment of average migraineurs, and not exclusively the most severe clinical cases. The study had several limitations. The complete migraine questionnaire was given only to participants who reported history of severe headache. Thus, participants who may experience migraine without aura with moderately painful headaches or aura without migraine headache would have been misclassified as controls. However, if there was substantial misclassification of migraineurs as controls in our population, we feel this would have underestimated the true effect of migraine on CBF. Additionally, because of limited information on aura symptoms, our classification for migraine with aura was modified from the ICHD-II criteria. Since all people classified in our study as migraine with aura also fulfilled the criteria for migraine without aura, this modification did not affect comparisons to controls but may have led to misclassification in the distinction of migraine with aura versus without aura. We also had limited additional clinical information, like prevalence of other headache types in our study population or detailed use of antimigraine medication. Another methodological consideration is our assessment of CBF with (non-contrast-enhanced) 2D phase-contrast MRI, which differs from other studies on migraine and blood flow and did not allow us to assess blood flow on a regional level. However, our non-invasive scan protocol allows for larger numbers of participants than more invasive, but more accurate, measures of blood flow like single photon emission computed tomographic (SPECT) and PET scans or contrast-enhanced computed tomography (CT) and MR imaging.
We found higher tCBF and pCBF in migraineurs compared with controls, with only pCBF reaching statistical significance. One explanation for this juxtaposition is that the variability in the tCBF measure was greater than in the pCBF measure, which may have precluded statistically significant differences in the former group despite notable effect sizes. Only a few studies have compared tCBF and regional CBF differences between interictal migraineurs and controls, using methods different from this study (10,19,21,28). The earlier SPECT studies reported regions of hypoperfusion (18–20) and no observable difference (13) in interictal migraineurs, and at least one study found cerebral hyperperfusion in interictal migraine without aura but cerebral hypoperfusion in interictal migraine with aura (21). More recently, a study of female interictal migraineurs using dynamic susceptibility contrast magnetic resonance imaging (DSC-MRI) found frontal lobe hyperperfusion in all migraineurs as well as unique regions of hyperperfusion and hypoperfusion in migraine with and without aura (10). Our method for measuring blood flow cannot take into consideration potential regional differences, so we cannot say whether our results represent global hyperperfusion or the effects from specific regions. Moreover, our statistical models explain a rather small proportion of the variance observed in CBF in our study population. However, our findings do partially correspond with previous studies and suggest that attack-free migraineurs, even migraineurs with >1 year without an attack, have subtly higher pCBF compared to controls.
When exploring CBF per artery, we found, most notably, that migraineurs had higher basilar flow compared to controls. When stratified by sex, we found similar increased basilar flow both in male and female migraineurs compared to controls. Given the apparent vulnerability of the posterior circulating territory in migraineurs, others have hypothesized interictal differences in blood flow in this region. Posterior hypoperfusion in migraineurs during an attack (14,15,20) and posterior regional CBF asymmetries in the interictal period (18,28) have been reported, though findings have not always been consistent (10,21). Some researchers have speculated that periods of hyperperfusion in the interictal period could be the brain’s compensation method for periods of hypoperfusion, e.g. caused by vasoconstriction during a migraine attack (10,29), or in reaction to cortical spreading neuronal depression during headache (30). Alternatively, hyperperfusion in migraineurs during the attack-free period could also be a reflection of systemic autonomic nervous system dysregulation in migraineurs, which has also been postulated and supported by other studies. (31–33). Our findings of increased basilar flow in attack-free migraineurs do correspond with regions known to experience decreased flow and perfusion during attacks (14,15,20) and known to be vulnerable to ischemic outcomes (7–9). Though the mechanism and pathways leading to such outcomes remains unclear, our findings bolster the notion that in migraineurs, the posterior circulating territory is particularly affected.
When exploring subcategories of migraine, we found that migraineurs with and without aura had similar differences in total and arterial flow compared to controls. This may be due to suboptimal differentiation between the two subtypes in our population, though other studies have also found no differences between interictal migraineurs with and without aura compared to controls (18,20,34). Only a few studies have explored the long-term impact of migraine as a risk factor for cerebrovascular (7) and cardiovascular (1,2) outcomes and are suggestive of a long-term impact of migraine. The largest differences in basilar artery flow were observed in migraineurs with <1 year without an attack and basilar flow in migraineurs with >1 year without an attack appears close to normal, suggesting that perhaps these vascular changes are relevant only in people still having migraine attacks. However, we find evidence against this idea because the largest differences in tCBF and pCBF compared to controls were observed in migraineurs with >1 year without an attack, indeed suggesting overall increased flow in this group that is not attributed to the basilar artery alone. These findings support sustained vascular differences between controls and both migraineurs with <1 year without an attack as well as migraineurs with >1 year without an attack, but additional research will be needed to draw more definitive conclusions. Overall, research on migraineurs with >1 year without an attack is largely lacking, and more work should be conducted to investigate the effect of history of migraine after migraine attacks cease.
In conclusion, we find in a large study of the general population in the Netherlands higher CBF, particularly basilar arterial flow, in attack-free migraineurs compared to controls. Though cross-sectional, our study supports the idea that vascular differences between migraineurs and controls exist outside the attack period, even long after the last attack, and indicates the posterior circulation territory is an ROI in migraineurs.
Clinical implications
Community-dwelling middle-aged and elderly individuals with (history of) migraine had higher parenchymal cerebral blood flow compared to controls. Higher blood flow in migraineurs seemed to be driven by higher basilar artery flow. Our study supports the idea that vascular differences between migraineurs and controls exist outside the attack period, even long after the last attack, and indicates the posterior circulation territory is a region of interest in migraineurs.
Footnotes
Funding
The Rotterdam Study is supported by the Erasmus MC University Medical Center and Erasmus University Rotterdam; The Netherlands Organisation for Scientific Research (NWO); The Netherlands Organisation for Health Research and Development (ZonMw); the Research Institute for Diseases in the Elderly (RIDE); The Netherlands Genomics Initiative (NGI); the Ministry of Education, Culture and Science; the Ministry of Health, Welfare and Sports; the European Commission (DG XII); and the Municipality of Rotterdam. None of the funding organizations had any role in the design and conduct of the study, in the collection, the analysis, or the interpretation of data, the writing of the manuscript, or in the decision to submit for publication.
Conflict of interest
None declared.
Ethics and institutional review board approval
The Rotterdam Study has been approved by the medical ethics committee according to the Population Study Act Rotterdam Study, executed by the Ministry of Health, Welfare and Sports of the Netherlands. A written informed consent was obtained from all participants.
Acknowledgment
The authors wish to acknowledge the efforts of Joyce van den Ende for her help with conducting migraine questionnaire interviews and coding of the migraine variables from the interview data.
References
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
