Figure 1 summarises the selection of studies. Of 5746 potentially eligible articles identified through the electronic search, 5714 were excluded after an evaluation of the title and abstract. A further seven were excluded: two used the same cohorts as other studies but had shorter follow-up,12 13
one did not present confidence intervals for the relative risks,14
one was cross sectional and did not present overall estimates for the follow-up data,15
one was cross sectional and the temporal association between migraine and cardiovascular events was not clear,16
one used only descriptive statistics,17
and one was a case-control study from a database that also contained a cohort analysis.18
Overall, 25 studies were suitable for inclusion in the analysis (table 1).
Fig 1Process of study selection
Table 1 Characteristics of included studies
Thirteen of the 25 studies were case-control studies,w4 w5 w10 w12-w16 w20 w22-w25 10 cohort studies,w1-w3 w7-w9 w11 w17-w19 and two cross sectional studies.w6 w21 Both cross sectional studies presented results for cardiovascular events after the diagnosis of migraine.
Eighteen studies presented results for any migrainew1-w3 w7-w13 w15 w16 w18 w19 w22-w25 and 16 for migraine with and without aura separately.w3-w6 w11-w14 w16-w21 w24 w25
The cardiovascular events investigated were heterogeneous. Three studies looked at the association of migraine with major cardiovascular disease,w2 w3 w17 16 with ischaemic stroke,w1-w6 w12-w18 w23-w25 three with transient ischaemic attacks,w8 w11 w12 three with combined ischaemic stroke and transient ischaemic attacks,w10 w20 w22 five with haemorrhagic stroke,w1 w10 w13 w18 w23 five with any stroke,w1 w8 w11 w13 w18 six with myocardial infarction,w1-w3 w7 w8 w17 five with angina,w2 w3 w8 w9 w21 three with coronary revascularisation procedures,w2 w3 w9 one with death due to coronary heart disease,w19 and five with death due to cardiovascular disease.w1-w3 w8 w9
Ten studies presented results for women only,w3 w4 w9 w10 w13 w14 w17 w18 w23 w25 one for men only,w2 seven for overall mixed cohorts of men and women without stratification by sex,w1 w5 w6 w8 w16 w20 w22 three for overall cohorts plus stratification by sex,w11 w12 w24 and four only for women and men separately.w7 w15 w19 w21
The age in the studies ranged from 15 to 80 years. Eleven presented results for participants aged less than 45w5 w7 w10-w14 w20 w23-w25 and eight for those aged 45 or more.w3 w6 w7 w11 w17-w19 w21 Five studies presented results stratified by smoking,w4 w13 w17 w24 w25 four stratified results for women according to oral contraceptive use,w4 w10 w13 w25 and one only included women using oral contraceptives.w23
Table 2 summarises the association between migraine and stroke in the included studies. Table 3 summarises the association between migraine and myocardial infarction, angina, and death due to cardiovascular disease. Table 4 shows the results from the pooled analyses, including measures of heterogeneity and publication bias.
Table 2 Studies included for analyses investigating association between migraine and stroke*
Table 3 Cohort studies included for analyses investigating association between migraine and myocardial infarction, angina, and death due to cardiovascular disease*
Table 4 Association between migraine and cardiovascular events, heterogeneity, and publication bias
Association between migraine and stroke
Nine studies investigated the association between any migraine and ischaemic stroke, which fulfilled the inclusion criteria.w1-w3 w12 w13 w15 w16 w24 w25 The pooled relative risks were 1.73 (95% confidence interval 1.31 to 2.29) for all studies (fig 2), 1.96 (1.39 to 2.76) for the six case-control studies,w12 w13 w15 w16 w24 w25 and 1.47 (0.95 to 2.27) for the three cohort studies.w1-w3 Heterogeneity was moderate across all studies (I2=65%); less for case-control studies (I2=42%) than cohort studies (I2=79%). Meta-regression showed that study type (case-control v cohort) was not a significant source of heterogeneity (P=0.3), and accounted for only 6.8% of the variance across all studies. Further analysis suggested an increased risk of ischaemic stroke among women (pooled relative risk 2.08, 95% confidence interval 1.13 to 3.84) but not among men (1.37, 0.89 to 2.11). Meta-regression did not, however, indicate that sex accounts for significant heterogeneity across the studies (P=0.31). The risk for people with migraine aged less than 45 (2.65, 1.41 to 4.97) was higher than for the overall group, which was more pronounced among women (3.65, 2.21 to 6.04). The risk of ischaemic stroke seemed to be further increased among smokers (9.03, 4.22 to 19.34) and women currently using oral contraceptives (7.02, 1.51 to 32.68). Formal investigation using Begg’s test indicated no publication bias (P=0.095), whereas Egger’s test suggested some publication bias (P=0.045). The funnel plot is presented in fig 3.
Fig 2Association between any migraine and ischaemic stroke (all studies)
Fig 3Funnel plot for studies investigating association between any migraine and ischaemic stroke
Eight studies investigated the association between migraine and ischaemic stroke stratified by migraine aura status.w3 w4 w6 w12 w13 w16 w24 w25 Pooled analyses suggested a significantly increased risk of ischaemic stroke among people who had migraine with aura (2.16, 1.53 to 3.03) but not those who had migraine without aura (1.23, 0.90 to 1.69; fig 4). This agrees with results from meta-regression, which indicate that migraine aura status is a significant source of heterogeneity across studies (P=0.02) and that 42% of the variance between studies is explained by this variable.
Fig 4Association between migraine with and without aura and ischaemic stroke
Three studies each investigated the association between any migraine and transient ischaemic attacksw8 w11 w12 and haemorrhagic stroke.w1 w13 w18 The risk of transient ischaemic attacks seemed to be increased more than twofold (2.34, 1.90 to 2.88), but there was no association with haemorrhagic stroke (1.18, 0.87 to 1.60).
Association between migraine and myocardial ischaemia
Eight studies (four cohorts from one paperw7) investigated the association between any migraine and myocardial infarction.w1-w3 w7 w8 Overall analyses (fig 5) and analyses stratified by sex did not suggest an increased risk. Although heterogeneity was moderate across the studies (I2=59%), it did not seem to be accounted for by sex (meta-regression P=1.00). Publication bias was not indicated (Begg’s test P=0.62; Egger’s test P=0.77). Only one study presented results stratified by migraine aura status.w3 Migraine with aura (relative risk 2.08, 95% confidence interval 1.30 to 3.31) but not migraine without aura seemed to be associated with a twofold increased risk of myocardial infarction.
Fig 5Association between any migraine and myocardial infarction
Three studies investigated the association between migraine and angina.w2 w3 w8 Among participants with any migraine the risk of angina seemed to be slightly but significantly increased (pooled relative risk 1.29, 95% confidence interval 1.12 to 1.47). Results from single studies suggest that the risk is higher in women than in men. The overall heterogeneity was low (I2=44%) and publication bias was not indicated (Begg’s test P=0.12; Egger’s test P=0.36). Only one study presented analyses stratified by migraine aura status,w3 which suggested a significantly increased risk in people who had migraine with aura (relative risk 1.71, 95% confidence interval 1.16 to 2.53) but not migraine without aura.
Association between migraine and death due to cardiovascular disease
Five studies investigated the association between any migraine and death due to cardiovascular disease.w1-w3 w8 w9 The analysis did not suggest an overall association (pooled relative risk 1.03, 95% confidence interval 0.79 to 1.34; fig 6). The two studies investigating the association among women found an increased risk (1.60, 1.06 to 2.42), which was not the case in another study among men. Heterogeneity was moderate across all studies (I2=54%) and publication bias was not indicated (Begg’s test P=1.0; Egger’s test P=0.9). The one study that investigated aura specific associations found an increased risk only among people who had migraine with aura (relative risk 2.33, 95% confidence interval 1.21 to 4.51) not migraine without aura.
Fig 6Association between any migraine and death due to cardiovascular disease
The Galbraith plots for some analyses identified individual studies as important sources of heterogeneity. Sensitivity analyses were carried out after excluding studies that did not fall within two standard deviations of the z score. For most of the analyses evaluating links between migraine and stroke, the associations did not change, albeit the effect estimates were lower. For example, after excluding three studies,w1 w2 w25 the pooled relative risk for the association between ischaemic stroke and any migraine was 1.54 (95% confidence interval 1.18 to 2.00), for migraine with aura (one study excludedw25) it was 1.80 (1.41 to 2.30), and for migraine without aura (one study excludedw25) it was 1.06 (0.83 to 1.37). However, after excluding one studyw25 the overall association between migraine and ischaemic stroke among women did not reach statistical significance (pooled relative risk 1.64, 0.94 to 2.86).
Although no significant overall association was shown between any migraine and myocardial infarction in the sensitivity analysis (1.12, 0.96 to 1.30; two studies excludedw2 w7), the analyses stratified by sex did: the pooled relative risk among women was 1.41 (95% confidence interval 1.10 to 1.81) and among men was 1.38 (1.14 to 1.69, one cohort from one study excluded in each casew7). In a further sensitivity analysis the results between any migraine and death due to cardiovascular disease were virtually unchanged (0.93, 0.81 to 1.10; one study excludedw3).