In this large-scale population-based 15-year follow-up study, MA had 55% increased risk for stroke, most evident in individuals aged < 55 years and among women. Individuals aged < 55 years with MA had an approximately two-fold higher risk of stroke compared to headache-free participants. No increased risk of stroke was found for MO and TTH.
Migraine and Stroke
Combining MA and MO, we found no increased risk of stroke. Several prospective cohort studies have evaluated the association between migraine and stroke with varying results. Three longitudinal matched cohort studies in South Korea, Denmark and Taiwan reported increased risk of ischemic stroke among migraineurs with HRs ranging from 1.18 to 2.26 [7, 14, 16]. Likewise, a case-control study in the US also showed a significant association between migraine and stroke (HR 1.54, 95% CI 1.16–2.05) [15]. In the sub-group analysis by aura status, studies from Taiwan and the US showed increased risk only for MA (HRs 1.64 to 2.78) [14, 15], whereas the other two studies reported increased risk of stroke for individuals both with (HRs 1.44 to 2.49) and without aura (HRs 1.15 to 1.81) [7, 16]. The differences in the results could partly be explained by methodological differences, because several studies are prone to selection bias, since most of the available reports are from the hospital-based matched cohort study [7], and migraine patients were identified from medical registries or insurance claims [7, 14, 16]. Using such strategy, migraine patients with frequent attacks may be overrepresented in the publications. In addition, the US study used self-reported data for cardiovascular events [15]. Headache diagnosis in the present study were questionnaire-based, without optimal sensitivity, which probably at least partly explains the lack of significant association between total migraine and stroke. Individuals with infrequent mild symptoms may not report specific symptoms for migraine which might lead to underestimation of the association between migraine and stroke.
Migraine with aura and Stroke
Previous studies [7, 14–16], including two meta-analyses [8, 11] have consistently shown greater risk in individuals with MA compared to those without aura. A study in the US reported a significant association between migraine with visual aura and ischemic stroke (HR 1.7, 95% CI 1.2–2.4) but not in individuals without visual aura (HR 1.2, 95% CI 0.8–1.7) [26]. Data from the Women’s Health Studies that included women aged ≥ 45 years observed increased risk of unspecified stroke and ischemic stroke only among women with MA (HRs ranging from 1.53 to 1.91) [13, 27, 28]. Another study observed the strongest association between MA and ischemic stroke among individuals with a low risk of vascular events as measured by Framingham risk score [29]. The result in the present study is in accordance with most of the studies and demonstrates that only subtype MA is associated with an increased risk of stroke. In our supplementary analysis, individuals with MA defined by presence of photophobia or phonophobia but without nausea/vomiting had even greater risk of stroke. The differences in the inclusion criteria, definition of variables, discrepancy in lifestyles, as well as genetic background should be considered when interpreting the study. Consistent with our findings, a recent meta-analysis including all these studies suggested significant association only between MA and ischemic stroke [11]. The HRs for ischemic stroke in people with MA was 1.67 (95% CI 1.26–2.22) and for those without aura was 1.18 (95% CI 0.94–1.49) [11].
Stroke by Sex and Age
Few studies reported migraine and stroke relationship by age and sex. In our subgroup analysis stratified by age and sex, only individuals with MA aged < 55 years and women were at greater risk of developing stroke. The risk of stroke among women with MA was highest among those aged less than 55 years. Consistent to our findings, available evidence indicates a strong association in young people and in women [7, 11, 14, 16, 17]. The Nurse’s Health Study that included US female nurses aged 25–42 years showed 1.6-fold increased risk of stroke in patients with migraine [30]. The Women’s Health Study also agrees that the risk of ischemic stroke is twice as high in young women (< 55 years) with MA compared to non-migraineurs [13]. A study in Denmark, Taiwan and South Korea that included both men and women reported greater than 2-fold increased risk of ischemic stroke among women migraineurs aged less than 50 years [7, 14, 16]. Consistent with our data, result for men and those aged > 55 years in these studies were not significant. One of the reasons could be that the occurrence of MA decreases in older age, and the effect of MA as a risk factor for stroke decreases with age [31]. In Taiwanese matched cohort study, the risk of stroke was more pronounced in women aged ≤ 45 years old with MA (HR 5.79, 95% CI 1.69–19.89) [14]. A recent meta-analysis including 18 cohort studies demonstrated that only women with migraine had a significant association with stroke (HR 1.32, 95% CI, 1.10–1.59) but not men (HR 1.53, 95% CI 0.74–3.17) [11]. Since migraine is more prevalent in women with estimated women to men ratio of 3–4:1 [18], the association is uncertain for men. Although insignificant, we observed men with MA had 54% increased risk of stroke compared to those without headache. Thus, we performed additional analysis for men aged < 55 years, the risk increased to 68% but the results were still not statistically significant (HR 1.68, 95% CI 0.76–3.71) (data not shown). Our study including previous evidence support a difference in migraine-stroke link by age and sex, highest risk among women below 55 years old with MA, suggesting a mechanism dependent to sex and age. Since migraine prevalence is high among childbearing age, thus affects during pregnancy and among oral contraceptive users. We do not have data on oral contraceptive use to present in this study.
Mechanism behind
The exact mechanism underlying the association between migraine and stroke is still not clear and is likely to be multifactorial. Several cardiovascular risk factors, such as smoking, BMI, hypertension, diabetes, and hyperlipidemia have been linked with migraine [32, 33]. A recent Swedish study observed that young men with lower cardiovascular fitness are at higher risk of developing migraine in the future [34]. Our results after adjustment for most of the traditional risk factors demonstrated a significant association between MA and stroke, particularly in young women with MA. Several studies including meta-analysis have shown that the risk was even higher in women with MA who use oral contraceptives and are smokers [8, 35–38]. This indicates an interaction between hormonal changes and MA that increase the risk of stroke in young women. Furthermore, it has been suggested that coagulation abnormalities and endothelial dysfunction play a significant role in the pathogenesis of ischemic stroke [16, 39]. Patients with MA had a higher prevalence of hypercoagulation including endothelial dysfunction and are susceptible to thrombosis [40, 41]. Endothelial dysfunction is recognized for a key role in the development of atherosclerosis, activation of the coagulation pathway, impaired vascular reactivity and enhance inflammatory responses [42]. However, a Dutch study showed that migraine was not associated with atherosclerosis-related ischemic stroke [43]. More studies are required to elucidate this link. Furthermore, cortical spreading depression (CSD), a self-propagating wave of depolarization across cerebral cortex has been associated with MA which plays a significant role in the pathophysiology of ischemic stroke [42]. Additionally, the shared genetic factors may also explain the association between migraine and stroke. A genome-wide analysis has found common genetic factors in migraine and ischemic stroke [44, 45]. Factor V Leiden G1691A and prothrombin G20210A were identified as a risk factor for stroke in young people and these variants were overrepresented in patients diagnosed with MA [46]. Furthermore, several studies have suggested the relationship between stroke and patent foramen ovale (PFO), and migraineurs particularly MA are at higher risk of PFO [47, 48]. However, the benefit of PFO closure with regard to cerebrovascular events is still unclear. A recent meta-analysis that included 3 randomized controlled trials and 4 observational studies reported 3 to 8 fold decrease in migraine frequency after PFO closure compared to those without closure [49]. On the other hand, a recent case-control studies in young adults observed that MA is linked to an increased risk of stroke, irrespective of PFO [50, 51]. Further studies may require identifying PFOs that may cause migraine and not just innocent bystander. Moreover, several studies have suggested increased risk of stroke for those with higher migraine frequency and active migraine status [27, 38, 52], but no data is available on whether migraine prophylaxis drugs such as anti-thrombotic or statin therapy reduces the risk of stroke [11].
Strengths and limitations
Our study has several strengths, including population-based study with follow-up for 15 years, large number of participants with a broad age range, available stroke data on all participants, and use of validated headache questionnaires. Headache status including migraine sub-types were classified based on ICHD3 criteria. Stroke diagnosis was identified from the quality Norwegian National Stroke Registry, reducing the risk of misclassification compared to self-reported data. Excluding stroke events, the first four years reduces the possibility for reverse causation [53]. Our study is the first study evaluating the relationship between stroke and headache types including TTH. Previous studies evaluating the association between migraine and stroke have been limited the comparison group to non-migraineurs. Furthermore, we evaluated stroke risk by aura status, following a separate analysis by age and sex groups. However, several limitations should be considered when interpreting the results. First, generalization of the results to the entire population must be done with caution because only 42% of the invited population answered headache questions. Nevertheless, the risk of stroke was similar in individuals who responded to headache questions compared to those who did not. Second, although validated headache questionnaire was used for headache-diagnosis, misclassification could not be ruled out. It is known that some aura characteristics are difficult to differentiate from symptoms of transient ischemic attack, and the latter is a risk factor for stroke [54]. Thus, potential misclassification of self-reported aura status would likely exaggerate a migraine stroke association. Although with some symptomatic overlap, we believe that this would be a rare situation knowing that the average age of migraine onset and vascular events differ. Furthermore, in our validation study, we have shown excellent agreement between self-reported migraine and ICHD-II based migraine classification [22]. In addition, the prevalence of MA (5.7%) and incidence of stroke (2.9%) is similar to those seen in other large population-based studies in Denmark [55, 56]. Third, we do not have stroke data for the first 4 years of follow-up (2008–2012) because the National Stroke Registry has data only from its inception in 2012. Thus, it is difficult to predict whether the inclusion of stroke diagnoses during the first four years would have changed our results, and in which direction. Fourth, analysis was adjusted for several important confounding factors, but we were unable to evaluate some of the recognized risk factors, such as use of hormonal contraceptives, use of medications including triptans, and NSAIDs (non-steroidal anti-inflammatory drugs) that might play a role in the association [57–59]. However, migraine specific drugs are used by both with and without aura, it seems an unlikely explanation to our results.