Migraine and atrial fibrillation: a systematic review and meta-analysis

Abstract

Background:

Patients with migraines, particularly those with auras, may present with stroke. Atrial fibrillation is a known risk factor for stroke. With common pathophysiological factors between migraines and atrial fibrillation, we aimed to clarify the association between migraine and atrial fibrillation in this systematic review and meta-analysis.

Methods:

A literature search was conducted in EMBASE, PubMed, Scopus and Cochrane electronic bibliographic databases from inception to 5 September 2022 with the following inclusion criteria: (a) cohort or cross-sectional studies; (b) studies that included only patients aged ≥18 years; and (c) studies that examined the association between atrial fibrillation and migraines. Exclusion criteria were case–control studies and the studies that included patients with previous diagnosis of atrial fibrillation or nonmigrainous headache. The Newcastle–Ottawa Scale was used to assess the quality of studies.

Results:

Six studies were included, demonstrating a pooled prevalence of atrial fibrillation of 1.61% (95% confidence interval [CI] 0.51, 3.29) in migraine with aura and 1.32% (95% CI 0.17, 3.41) in migraine without aura. The overall prevalence of atrial fibrillation in migraine was 1.39% (95% CI 0.24, 3.46).

Conclusion:

In this systematic review and meta-analysis, the overall prevalence of atrial fibrillation in patients with migraine was low. Further studies are needed to clarify this relationship.

INTRODUCTION

Migraine is a primary headache disorder characterised by a recurrent unilateral throbbing headache and is often accompanied by symptoms like nausea, vomiting, photosensitivity and hyperacusis. While the cause of migraine, with or without aura, is still currently unknown, there are two main theories. The first postulates that migraine is a result of primary neuronal dysfunction, while the other one suggests that cerebral vasodilation may be a causative factor.[1] Regardless of the underlying mechanism, it has been observed in epidemiological studies that migraines, especially those with aura, are commonly seen in patients with strokes and could represent an independent risk factor.[2,3] The observation that atrial fibrillation (AF) leads to an increased risk of blood clot formation and is the cause of one in four strokes raises the question of whether there is an association between migraine and AF.

Asymptomatic AF can remain undetected for years before complications appear. Therefore, if migraine is identified to be a risk factor of AF, it could have important clinical implications in terms of primary prevention for stroke. Patients with migraine could also benefit from targeted early screening to pick up covert AF. As such, we aimed to perform a systematic review and meta-analysis to examine the co-prevalence of migraine and AF.

METHODS

Study design

The protocol for this review was registered and published in the Prospective Register of Systematic Reviews (PROSPERO; CRD42021260317). In the protocol, it was stated that only cohort studies were to be included. However, as there was a lack of studies investigating the incidence of AF events, we decided to also investigate the prevalence of AF in patients with migraine. A literature search of EMBASE, PubMed, Scopus and Cochrane electronic bibliographic databases was conducted from their inception to 5 September 2022 to identify the relevant studies. Search terms included ‘migraine disorders or migraine with aura or migraine without aura’ or similar terms to ‘headache’ and ‘atrial fibrillation’ or word variations for the same condition. Relevant articles referenced in the studies were also hand-searched and identified. Language and further time restrictions were not applied to the search.

Inclusion criteria were as follows: (a) cohort or cross-sectional studies, (b) studies that included only patients aged ≥18 years and (c) studies that examined the association between atrial fibrillation and migraines. Exclusion criteria were case–control studies that included patients with previous diagnosis of AF or suffering from nonmigrainous headache. Cohort studies selected may include either three groups of patients (those with migraine without aura, with migraine with aura, and without migraine) or two groups of patients (those with migraine and without migraine).

Titles and abstracts were screened independently by two researchers (CL, CHS). Any discrepancies were resolved through discussion. Full texts of the studies were accessed, and the data extracted and tabulated onto a standardised data extraction form. Information on study design, patient medical history and study outcome, including the prevalence of migraine with aura, migraine without aura and no migraine, were extracted. The Newcastle–Ottawa Scale for case–control and cross-sectional studies was used to assess the quality of the studies included in the analysis.

Statistical analyses

Data extracted from the included studies were used to meta-analyse the pooled mean prevalence with 95% confidence intervals (CIs) for AF in patients with migraine. Analysis was performed with the one-step generalised linear mixed-effects model method using the metaprop_one routine in Stata version 16.0 (StataCorp LP, College Station, TX, USA). This has been shown to yield smaller errors, greater coverage probabilities and less biased estimates than traditional two-stage methods.[4,5] When the one-stage model failed to converge, we conducted an inverse variance-weighted random-effects meta-analysis with the Freeman Tukey double-arcsine transformation to pool proportions. Random-effects models were used in all analyses to account for anticipated heterogeneity in the observational estimates.[6] Between-study heterogeneity was measured using the I² statistic,[7] where an I² of <30%, 30%–60% and >60% indicates low, moderate and substantial heterogeneity between studies, respectively. Two-sided P values of <0.05 indicated statistical significance.

Excluded studies

Three of the studies excluded in the eligibility stage had recorded outcomes that were unsuitable for analysis. Of these, one reported on incidence of AF instead of prevalence, one recorded the incidence of AF and atrial flutter combined, and one reported on the prediction of AF and not the actual prevalence.

The study by De Giuli et al.[8] was also excluded, as the inclusion criterion in that study was patients with pre-existing stroke or transient ischaemic attack. This would make the pool of patients more heterogeneous and an outlier, as our interest was the co-prevalence in the general population. Several studies previously suggested that the incidence of newly detected AF is high post stroke.[9] Another paper also suggested that the presence of newly detected AF is precipitated by permanent brain damage–resultant sympathetic and parasympathetic imbalances or insular ischaemic lesions that affect the modulation of cardiac function, rather than AF leading up to the stroke.[10] These factors make the possibility of developing AF higher, and therefore, this paper was excluded from our analysis. We could not obtain full text for two studies despite contacting the authors, and thus, they were excluded from our study.

RESULTS

There were a total of six studies with 2,219,834 patients that fulfilled the final inclusion and exclusion criteria, including four cohort studies and two cross-sectional studies [Figure 1]. Characteristics and data from the included studies were collated [Table 1]. The quality of all the studies was generally good based on the Newcastle–Ottawa Scale, and all six studies were included in the meta-analysis [Table 2].

Figure 1.

PRISMA flow diagram of the process of identifying, screening, and checking for eligibility and inclusion of studies.

Table 1.

Characteristics and data from included studies.

Study	Study design	Age (yr)	n
			Migraine + aura	Migraine + aura + AF	Migraine w/o aura	Migraine w/o aura + AF
Vijiaratnam et al.,[14] Australia, 2016	Cohort (retrospective) (n=505)	Migraine+aura: mean 39	257	2	248	1
		Migraine w/o aura: mean 41	M: 40		M: 31
			F: 217		F: 217
Lantz et al.,[15] Sweden, 2017	Cohort (prospective) (n=53,404)	45.3 (19.3–65.5)a	3,553	107	5,082	158
			M: 742		M: 1,393
			F: 2,811		F: 3,689
Kuybu et al.,[16] USA, 2020	Cross-sectional (retrospective) (n=834,875)	33 (18–44)a	40,909	475	793,966	6,655
			M: 7,590		M: 118,370
			F: 33,319		F: 675,596
Lin et al.,[17] Taiwan, 2014	Cohort (prospective) (n=316,602)	Mean 40.3	Migraine: 158,301
			Migraine + AF: 421
Ghanshani et al.,[18] USA, 2020	Cohort (retrospective) (n=189,684)	Non-triptan users: 45.4±15.8	Migraine: 189,684
		Triptan users: 40.8±13.2	Migraine + AF: 1,438
Patel et al.,[19] USA, 2019	Cross-sectional (retrospective) (n=56,499,788)	Median 48	Migraine: 983,065
			Migraine + AF: 47,285

^aData presented as mean (range). AF: atrial fibrillation, F: female, M: male

Table 2.

Newcastle–Ottawa scale for cohort, case–control and cross-sectional studies included in the study.

Cohort study	Representativeness of the exposed cohort	Nonexposed cohort drawn from the same community as exposed cohort	Ascertainment of exposure (secure record, structured interview)	Demonstrates that outcome was not initially present	Assessment of outcome (record linkage)	Follow-up at least a year	Adequacy of follow-up (complete, or describes characteristics of missing subjects)	Comparability of cohorts on the basis of design or analysis
Lantz, 2017[15]	*	*	*	*		*	*	**
Ghanshani, 2020[18]		*	*	*	*	*	**	*
Vijiaratnam,[14] 2016	*		*	*	*	*	**	*
Case-control study	Representativeness of cases	Is the case definition adequate?	Selection of controls	Definition of controls	Ascertainment of exposure	Same method of ascertainment for cases and controls	Nonresponse rate	Comparability of cases and controls on the basis of design or analysis
Lin, 2014[17]	*	*	*	*	*	*	*	*
Cross-sectional study	Representativeness of sample	Sample size	Nonrespondents	Ascertainment of exposure	Comparability	Assessment of outcome	Statistical test
Kuybu,2020[16]	*	*	*	*	**	**	*
Patel, 2019[19]	*	*	*	*	**	**	*

*Each asterisk denotes the fulfilment of a criterion in the particular category.

Out of the six studies, three cohort studies (two retrospective, one prospective) with a total sample size of 888,784 patients provided prevalence data of patients with migraine with and without aura having AF [Figures 2 and 3]. Three studies provided prevalence data for patients with migraine with aura and patients with migraine without aura, while the remaining three studies provided prevalence data for only patients with migraine in general. The pooled prevalence of AF was 1.61% 1.61% (95% CI 0.51, 3.29) in migraine with aura compared to 1.32% (95% CI 0.17, 3.41) in migraine without aura. From the analysis of the effect of total migraine on AF, three more studies were included and the overall prevalence was 1.39% (95% CI 0.24, 3.46) [Figure 4]. Overall, there was high heterogeneity (I² = 100%).

Figure 2.

Association between migraine with aura and atrial fibrillation.

Figure 3.

Association between migraine without aura and atrial fibrillation.

Figure 4.

Overall association between migraine and atrial fibrillation.

DISCUSSION

In our study, on comparing the prevalence of AF in migraine with and without aura, the prevalence of migraine with aura is found to be higher than migraine without aura. The overall prevalence of AF in the combined cohort, as well as, in the general, non-stroke population were low.

According to Rhee et al.,[11] a cohort study in which 4 million patients were included, mild migraine increased the risk of AF minimally for both men and women. However, for men, the incidence rate (IR) and hazard ratios (HRs) increased for migraine without aura (IR 3.72/1,000 vs. 2.20/1,000 person-years in the control group, HR_adj 1.15 [1.08–1.22]) and with increasing severity of migraine (IR 4.51/1000 person-years, HR_adj 1.21 [1.12–1.31]), but not for migraine with aura. On the other hand, the incidence and the risk of AF greatly increased in those with migraine with aura (IR = 2.93/1,000 vs. 1.80/1,000 person-years in the control group, HR_adj 1.41 [1.15–1.72]), and with increased severity of migraine. This shows that migraine severity may be associated with higher prevalence of AF, and the results may differ between males and females. However, as the severity of migraine was not determined in the patient pools for all the included studies, there were insufficient studies to allow for analysis of the effect size of different migraine severities on AF. We were unable to look into sex-specific risk of migraine with and without aura on AF, as only three of the six studies presented their results for men and women separately.

A systematic review by Scutelnic et al.[12] also attempted to explore the relationship between AF and migraine with and without aura. However, it included studies of patients in specific populations — stroke or catheter ablation for AF. Of the 21 studies included in the review, seven were case reports describing specific cases related to migraine and AF. This was excluded in our analysis, as case reports constitute low level of evidence. Also, we excluded patients diagnosed with AF before migraine with or without aura, as we wanted to examine specifically the development of AF after a diagnosis of migraine.

The majority of people suffering from migraines were younger, reflected by the mean and median age of the patients (in their 40s), while approximately 70% of individuals with AF were between 65 and 85 years old. Prevalence increased from 2.3% in people older than 40 years to around 5.9% in those older than 65 years.[13] Age may possibly explain why the prevalence of AF in those suffering from migraines was low.

On the contrary, the study of Lantz looked into migraine in the Swedish Twin Registry and interestingly reported that AF was less common in patients with no headache compared to those with migraine. It was similarly noted that the no-headache group was older than the group with migraine. While this study did point out that those with migraine were more prone to have hypertension and peripheral artery disease, and more obese than nonmigraineurs, these past medical histories were extracted from a self-reported questionnaire compared to the diagnosis of AF, which was extracted via international classification of diseases codes. As such, there may have been an element of reporting bias and some differences in baseline characteristics between the two groups that were unreported.

To explore the idea of migraine being associated with higher risk of AF, we have to consider the causes of the two conditions. Atrial fibrillation is a condition of heterogeneous nature with multiple underlying and overlapping aetiologies. However, there is a common consensus that autonomic dysregulation is one of the triggers of AF.[20,21] Migraine, as mentioned earlier, is also associated with neuronal dysfunction. Although the exact cause of migraine is not entirely known, there is strong clinical and experimental evidence to show that cortical spreading depression has a strong association with migraine with aura.[22] Cortical spreading depression is a slowly propagated wave of depolarisation preceding the inhibition of electrical brain activity, leading to transient chemical, nervous and vascular changes in the brain, which may be related to pain. This hypothesis is further supported by abnormal electrocardiogram changes seen in patients with migraine attacks.[23] It is also reported that there is a higher frequency of cardiac arrhythmias during migraine versus when the patient is not having migraine.[24] Therefore, the idea of neuronal dysfunction being the reason why migraine, especially migraine with aura, has a higher association with AF seems probable, and this area of interest should be further investigated.

Despite the multiple probable theories behind the relationship between migraine and AF, our study showed that prevalence of AF in patients with migraine in the general population was low. Further studies need to be conducted to find out if the effect of migraine on AF is large enough to warrant modification to risk prediction tools, such as the CHA₂DS₂-VASc Score for Atrial Fibrillation Stroke Risk. Currently, prevention of strokes in patients who suffer from AF does not take into account a history of migraine. If an association can be established, patients with migraine may then need to be considered for stroke prophylaxis, depending on their pre-existing health conditions and comorbidities. Further trials can also be conducted in the future to find out whether migraine patients will benefit from earlier AF detection and if antiplatelets and anticoagulants for primary stroke prevention are effective.

To the best of our knowledge, our systematic review is the first to examine the relationship between migraine and AF in the general population. A limitation we faced was the difference in definitions of migraine across the papers, which could have possibly left some potentially suitable patients out from our study. Patients who only had migraines for a short period of time could have been excluded, depending on the specific migraine diagnosis criteria. We were not able to determine a sex-specific effect of migraine with and without aura on AF because not all studies reported on the sex of the patients who developed AF during the study. However, it was previously observed that females make up a larger portion of the total patients included in the studies compared to males, and this may be attributed to fluctuating oestrogen levels, which may trigger migraine attacks.[25] This is corroborated by studies demonstrating that prevalence of migraine increases in females after menarche, even though boys and girls in the prepuberty stage are equally likely to get migraines.[26] Hence, this could potentially affect the prevalence of migraine in patients who have AF.

We considered conducting a meta-analysis to analyse whether migraine leads to higher risk of developing AF, but were unable to do so, as there was only one cohort study that investigated migraine as a risk factor for AF.[27] However, results from that study suggested that migraine with aura was directly associated with an increased risk of AF compared to no migraine (HR 1.30, 95% CI 1.03, 1.62). Migraine without aura was reported to have a higher risk (HR 1.39, 95% CI 1.05, 1.83). This was from a follow-up period of 20 years, where the incident AF was noted in 232 (15%) of 1,516 individuals with migraine and 1,623 (17%) of 9,405 individuals without migraine.

CONCLUSION

In this systematic review and meta-analysis, the overall prevalence of AF in patients with migraine was low. Further studies are needed to clarify this relationship.

Conflicts of interest

Sia CH and Tan BYQ are members of the SMJ Editorial Board, and were thus not involved in the peer review and publication decisions of this article.

Funding Statement

Sia CH was supported by the National University of Singapore Yong Loo Lin School of Medicine’s Junior Academic Fellowship Scheme.