Coincidental Impact of Transcatheter Patent Foramen Ovale Closure on Migraine with and without Aura – A Comprehensive Meta-Analysis

Abstract

Background

We analyzed the literature to assess the coincidental impact on migraines of transcatheter patent foramen ovale (PFO) closure performed for secondary stroke prevention.

Methods

We searched Medline, EMBASE, and the Cochrane database for studies published up until August 2013. We included English-language studies that provided information on complete resolution or improvement in migraine headaches following PFO closure. Two study authors identified 375 original articles and both independently reviewed 32 relevant manuscripts. Data including study methodology, inclusion criteria, PFO closure and migraine outcomes were extracted manually from all eligible studies. Pooled odds (and probability) of resolution or improvement of migraine headaches were calculated using random-effects models.

Results

Twenty studies were analyzed. Most were uncontrolled studies that included a small number of patients with cryptogenic stroke who had undergone PFO closure and had variable time of followup. The probability of complete resolution of migraine with PFO closure (18 studies, 917 patients) was 0.46 (95% confidence interval 0.39, 0.53) and of any improvement in migraine (17 studies, 881 patients) was 0.78 (0.74, 0.82). There was evidence for publication bias in studies reporting on improvement in migraines (Begg’s p=0.002), but not for studies on complete resolution of migraine (p=0.3). In patients with aura, the probability of complete resolution of migraine post-PFO closure was 0.54 (0.43, 0.65), and in those without aura, complete resolution occurred in 0.39 (0.29, 0.51).

Conclusions

Among patients with unexplained stroke and migraine undergoing transcatheter PFO closure, resolution of headaches occurred in a majority of patients with aura and for a smaller proportion of patients without aura.

Introduction

Migraine is a primary headache disorder. Twelve percent of the adult U.S. general population have migraines, of which 30% have associated aura [1,2]. A foramen ovale is an embryological inter-atrial shunt that remains patent in up to 25% of the general adult population based on autopsy evaluation [3]. Patent foramen ovale (PFO) has been noted in 40% to 60% of migraineurs [4], especially in those with preceding aura [5–7]. In addition, migraineurs have been known to have intracardiac right-to-left shunting more commonly [5]. A PFO provides the potential for shunting portal and systemic venous metabolites like serotonin to the cerebral blood flow, bypassing their removal during pulmonary circulation passage. Serotonin can lead to platelet activation and aggregation in the cerebral circulation and is thought to be the initiating trigger of migraines in susceptible individuals [8]. Similar to PFO, the creation of an ASD with transseptal puncture for catheter ablation of atrial fibrillation has been suggested to be associated with migraine headaches that resolve over time, presumably as the ASD closes [9].

The results of available studies involving PFO and migraine, and the influence of PFO closure on resolution of migraine symptoms are variable. In a population-based cohort study, PFO detected with transthoracic echocardiography and agitated saline was not associated with self-reported migraine [10]. The MIST trial (Migraine Intervention With STARFlex Technology) was a negative randomized sham-controlled study on PFO closure for drug refractory migraines [11]. However, studies in migraineurs with aura have shown a higher prevalence of PFO, and the migraine frequency was reduced with PFO closure as compared to migraineurs without preceding aura [12]. Partial or complete relief of migraine symptoms have been reported with transcatheter PFO closure performed for secondary stroke prevention or prevention of decompression illness [13]. However, most studies have not elucidated the distinction between partial and complete relief from migraines with transcatheter closure of PFO. The difference between migraine with and without aura in patients undergoing PFO device-based closure has not been well defined. The pathogenesis of migraine with aura may differ from that without aura, and this difference may have implications on the effect of PFO closure on migraine outcome.

Given the discrepant results among studies and the limited sample sizes of some of these studies, we performed a meta-analysis to evaluate the effect of PFO closure on resolution or improvement of migraine. Our objective was to conduct a systematic review of relevant existing literature to examine the effect of PFO closure on the migraine symptoms and to identify a subgroup, if any, which could benefit more from PFO closure.

Methods

Search strategy

Following the Meta-analysis of Observational Studies in Epidemiology (MOOSE) guidelines [14], we conducted a comprehensive literature search of studies published in Medline, EMBASE, and the Cochrane Library databases for studies published until August 2013. Key words used were migraine, migraineurs, migraine headaches, patent foramen ovale or PFO, closure of PFO, heart septal defect, atrial septal defect, atrial septal aneurysm, percutaneous or surgical or operative procedure, transient ischemic attack, stroke, and atrial fibrillation.

Studies that met each of the following criteria were considered eligible for the meta-analysis: 1) published in English, and 2) studies reporting a formal assessment of migraine before PFO closure and at followup. We did not exclude studies without a control arm if they reported frequency and severity of migraine before and after the PFO closure. Studies that were not subjected to a formal peer-review process were excluded from the meta-analysis. The title and abstract of all articles meeting the search criteria were reviewed by two of the study authors (S.K., A.N.). Full reports of 32 potentially relevant articles were then independently reviewed by these two authors to establish eligibility based on the defined inclusion criteria.

Data extraction

A standardized, piloted data extraction form was used for recording information. Data extraction was completed by one of the authors (SK), and verified by a second author (AN). Study populations with migraine were characterized by the presence or absence of aura, and the following information was recorded: study design (observational or randomized; controlled or uncontrolled), primary author, publication year, indication for PFO closure, followup duration, number and duration of antiplatelet therapy, number of participants, resolution or improvement (defined as ≥ 50% improvement in migraine headaches). Disagreement between reviewers during the selection process was resolved through consensus.

The quality of the studies was assessed on the basis of elements from the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) checklist for observational studies [15]. We did not assign a threshold for study inclusion. Inconsistencies among studies were found in determination of bias, selection of participants in the study, determination of study size, handling of quantitative variables, and determination of the source of funding.

Each study contributed only one effect size per analysis. If data were duplicated between studies, the most recent study was used. If frequencies were not given, they were estimated from percentages and rounded to the nearest integer. If any cell had a zero count, odds were calculated by adding 0.5 to avoid division by zero.

Statistical analyses

For the two primary analyses, we pooled the log odds of 1) resolution of migraine and 2) improvement (partial/complete resolution) in migraine using random-effects models, with corresponding 95% confidence intervals (CI). Pooled odds were converted to pooled probability of outcome. The I² statistic (percentage of variance in the pooled estimate due to inter-study variance) was used to examine the heterogeneity of effect sizes in the overall aggregations: I²≤25% indicated low heterogeneity and I²≥75% indicated high heterogeneity.

Publication bias was evaluated using a Begg’s funnel plot, and statistical testing for publication bias performed was assessed using Begg’s and Egger’s methods with p<0.05 suggesting evidence for presence of publication bias.[16,17] Statistical analyses were performed using Review Manager (RevMan) Version 5.3 (Copenhagen, The Cochrane Collaboration, 2014) and www.wessa.net. Two-tailed P values less than 0.05 were considered statistically significant.

Results

We identified 375 articles from our defined search strategy. After review of the abstracts, we identified 32 potentially relevant studies and these manuscripts were reviewed in detail. We then excluded 12 studies that did not report relevant data or were duplicate reports. Twenty studies were thus included in the pooled analyses (Table 1, Figure 1). PFO closure was performed in all but one study for secondary prevention of stroke or other embolic events, and in one study for closure of a hemodynamically significant atrial septal defect [18].

Table 1.

Studies included in the meta-analysis

No.	Author	Year	Design	Patient subset	Migraine assessment	Follow-up (months)	Antiplatelet drugs
1	Wilmshurst[26]	2000	Uncontrolled	Decompression illness Paradoxical embolism Large ASD	Questionnaire for frequency, duration, and intensity of migraine episodes	9–30m	Aspirin 6m
2.	Morandi[27]	2003	Uncontrolled	Cryptogenic stroke	Questionnaire for frequency, duration, and intensity of migraine episodes	12	Aspirin 6m
3.	Post[28]	2004	Uncontrolled	Paradoxical embolism Systemic desaturation	Questionnaire for frequency, duration, and intensity of migraine episodes	6	Low dose aspirin (duration not mentioned)
4.	Schwerzmann[29]	2004	Uncontrolled	Paradoxical embolism	Neurologist classified the headache	12	Clopidogrel 1m Aspirin 6m
5.	Azarbal[30]	2005	Uncontrolled	Cryptogenic stroke	MIDAS	12	Clopidogrel 3m Aspirin indefinitely
6.	Mortelmans[18]	2005	Uncontrolled	ASD	Questionnaire for frequency, duration, and intensity of migraine episodes	29	-
7.	Reisman[31]	2005	Uncontrolled	Paradoxical cerebral embolism	Standardized migraine embolism	12	Clopidogrel 3m Aspirin 6m
8.	Anzola[32]	2006	Case-control	Stroke	MSS	12	Aspirin 300mg
9.	Giardini[13]	2006	Uncontrolled	Stroke/TIA	MIDAS	12	Aspirin 12m Ticlopidine 3m Warfarin 6m with thrombophilia
10.	Slavin[33]	2007	Uncontrolled	Cryptogenic stroke TIA Desaturation	MIDAS Frequency and severity assessed	30	-
11.	Dubiel[34]	2008	Uncontrolled	Paradoxical embolism	Questionnaire for frequency, duration, and intensity of migraine episodes	38	Aspirin 6m
12.	Jesurum[35]	2008	Uncontrolled	Paradoxical cerebral embolism	Questionnaire for frequency, duration, and intensity of migraine episodes	24	Clopidogrel 3m Aspirin ≥6m
13.	Luermans[36]	2008	Uncontrolled	Paradoxical embolism	Questionnaire for frequency, duration, and intensity of migraine episodes	6	Clopidogrel 1m Aspirin ≥6m
14.	Chessa[37]	2009	Uncontrolled	TIA Ischemic lesions on CT/MRI Severe migraine	MSS	6	Aspirin + Clopidogrel 1m Aspirin 5m
15.	Kimmelstiel[38]	2009	Case-control	Stroke/paradoxical embolism	MIDAS	3	Clopidogrel 6 m Aspirin indefinitely
16.	Vigna[39]	2009	Case-control	Subclinical brain MRI lesions	Questionnaire for frequency, duration, and intensity of migraine episodes	16	Clopidogrel 3m Aspirin 6m
17.	Wahl[40]	2010	Uncontrolled	Paradoxical embolism	Questionnaire for frequency, duration, and intensity of migraine episodes	60	Clopidogrel 1–6m Aspirin 6m
18.	Trabattoni[41]	2011	Uncontrolled	Embolic CVA	MSS	28	Aspirin 6m
19.	Rigatelli[42]	2012	Uncontrolled	High risk for paradoxical embolism	MIDAS	18	-
20.	Nagpal[43]	2013	Uncontrolled	Cryptogenic stroke (93%) Peripheral embolism Hypoxemia Intractable migraine	Frequency and severity on a 0–10 scale	55.2	Aspirin (duration not mentioned) Clopidogrel (duration not mentioned) Warfarin (8%)

ASD, atrial septal defect; CT, computerized tomography; CVA, cerebrovascular accident; MIDAS, migraine disability assessment test; MRI, magnetic resonance imaging; MSS, migraine severity score; TIA, transient ischemic attack

Figure 1.

Study flow diagram

Effect of PFO closure on migraine

The pooled result for resolution of migraine was derived from 18 studies comprising 917 patients undergoing PFO closure. The pooled odds of resolution of migraine at follow-up were 0.85 (95% CI 0.64, 1.14), translating to a pooled probability of resolution of 0.46 (0.39, 0.53). There was large heterogeneity between study results (I²=76%). Improvement in migraine (partial/complete resolution) was obtained from 17 studies involving 881 patients. Odds of improvement were 3.63 (2.80, 4.71) or probability 0.78 (0.74, 0.82), with moderate heterogeneity (I²=54%) (Table 2, Figure 2).

Table 2.

Pooled estimates (odds and probability) of resolution or any improvement of migraines following PFO closure

Outcome	Studies #	Patients #	Odds [95% CI]	Probability [95% CI]	Heterogeneity I2 (%)	Publication bias (p-value)*
Complete resolution of migraine
Overall population	18	917	0.85 [0.64, 1.14]	0.46 [0.39, 0.53]	76	0.3
Patients with migraine with aura	13	427	1.17 [0.74, 1.84]	0.54 [0.43, 0.65]	77	0.4
Patients with migraine without aura	12	206	0.64 [0.40, 1.04]	0.39 [0.29, 0.51]	54	0.3
Improvement of migraine (partial/complete resolution)
Overall population	17	881	3.63 [2.80, 4.71]	0.78 [0.74, 0.82]	54	0.002
Patients with migraine with aura	13	499	4.40 [3.22, 6.01]	0.81 [0.76, 0.86]	37	0.06
Patients with migraine without aura	12	211	2.47 [1.37, 4.46]	0.71 [0.58, 0.82]	64	0.06

^*Begg’s test for publication bias

CI, confidence interval

Figure 2.

Metaanalysis: (A) complete and (B) partial/complete resolution of migraine headaches following PFO closure.

Publication bias

Both Begg’s and Egger’s tests suggested no apparent publication bias for the outcome of resolution of migraine (p=0.26 and 0.21 respectively). However, there was asymmetry in results for the improvement in migraine outcome with smaller studies suggesting a larger benefit (Begg’s p=0.002; Egger’s p=0.004) (Table 2, Figure 3).

Figure 3.

Begg’s publication bias plots: (A) complete resolution, (B) partial/complete resolution of migraine with PFO closure

Migraine with aura

Following PFO closure among the subgroup of migraineurs with aura, the pooled probability of resolution of migraine was 0.54 (0.43, 0.65), and the pooled probability for any improvement was 0.81 (0.76, 0.86) (Table 2, Figure 4A).

Figure 4.

Migraine with aura: (A) complete and (B) partial/complete resolution of migraine with aura with PFO closure Migraine without aura: (C) complete and (D) partial/complete resolution of migraine without aura with PFO closure

Migraine without aura

For patients without aura, the pooled probability of resolution of migraine with PFO closure was 0.39 (0.29, 0.51) and pooled probability of improvement was 0.71 (0.58, 0.82) (Table 2, Figure 4B).

Discussion

We found in studies primarily evaluating transcatheter closure of PFO for secondary prevention of stroke, that 46% (95% CI 39%, 53%) of patients with migraines had a resolution of their symptoms. There was a large heterogeneity in the results reported from different studies. The numerical proportion of patients with resolution of migraines was more for those with aura 54% (43%, 65%) compared to those without aura 39% (29%, 51%). The assessment of any improvement in migraine headaches (78% [74%, 82%] patients) had moderate heterogeneity, and suggested evidence for publication bias.

Pathophysiological attributes linking right-to-left shunting across PFO and migraine include microemboli and chemicals like serotonin which bypass clearance/metabolism that is thought to occur within the pulmonary circulation [8,19]. Migraineurs with aura have a higher prevalence of PFO than migraineurs without aura and non-migraineurs [20]. In our meta-analysis, we found this subgroup had higher rates of migraine improvement and resolution after PFO closure.

There have been two randomized controlled trials (RCTs) for PFO closure as treatment for migraines with aura refractory to multiple drugs – MIST[11] and PRIMA [21]. Both studies were negative for the primary efficacy endpoint. We did not include the results of these RCTs in our analysis because PRIMA has not yet been published and MIST did not report the results on the proportions of patients with varying degrees of benefit. MIST showed that at 6 months, PFO closure was not statistically different from a sham procedure in eliminating headache, headache frequency, or migraine disability scores; though there was a suggestion of reduction in headache days. PRIMA has not yet been published, but results show a non-statistically significant reduction in migraine days in the PFO closure group compared to medical therapy. However, a significant reduction in migraine with aura days was seen at 1 year, with a mean reduction of migraines with aura days of −2.4 (PFO closure group) versus −0.6 (medical therapy group) days/month from a baseline of 4.1 and 4.0 days/month, respectively. Limitations of these RCTs include small sample size, inclusion of patients with chronic headache, operator learning curve in PFO closure, procedural complications, and short follow-up periods.

The patients included in these RCTs were being treated primarily for refractory migraines, and thus are inherently different from those in our meta-analysis which almost only included patients with stroke and associated migraine. It is possible that there is a relationship between migraine and stroke and benefit of PFO closure as seen commonly in this subgroup, especially if they present with aura. A meta-analysis of 14 studies demonstrated 2-fold increased association between migraines and ischemic stroke, especially in patients with aura (relative risk 2.27) [22]. It is also known that migraineurs with stroke have higher prevalence of PFO (84%) as compared to migraineurs without a neurological event (38%) [23]. It is, therefore, conceivable that there is a relationship between cerebrovascular events and migraines in presence of PFO and it is this subset that may benefit from PFO closure. Irrespectively, our results are not necessarily contradictory to the negative RCTs and might reflect the absence of a control group. This is exemplified by the PFO closure group in PRIMA where 40% became free of migraine with aura. The placebo effect in migraineurs is known to influence their improvement in symptoms and that response may be exaggerated with PFO closure [24,25]. Recollection bias may have also influenced the results of uncontrolled studies.

There was a large heterogeneity in reported results from the included studies. This variability could be due to the differences in patient population, differences in techniques, methodology and quality of migraine assessment, bias in recall and reporting, differences in followup and other factors. Most studies in our analysis had a sample size of less than 100. The type and duration of anticoagulation therapy following PFO closure was disparate with many studies relying on aspirin monotherapy but others using dual antiplatelet therapy with clopidogrel or ticlopidine, and occasional use of oral anticoagulation. PFOs were detected during evaluation for cryptogenic stroke and influence of PFO closure on frequency and severity of migraine was detected using different scoring systems, thus making inter-study comparisons difficult. We also noted some evidence of publication bias in assessment of any benefit, and it is possible that studies reporting negligible benefit did not get published and were unavailable to be included in our pooled estimates.

In conclusion, the collective evidence in patients with migraine undergoing PFO closure for other indications (mostly secondary stroke prevention) suggests that a majority of those with aura have resolution of their headache symptoms. Small RCTs of PFO closure for drug-refractory migraine with aura have failed to show statistical benefit from PFO closure. Whether this discordance is because of placebo effect in uncontrolled studies, recollection or publication bias, or benefit restricted to migraines in setting of thromboembolic neurological events needs to be teased out with well-designed prospective studies.

Abbreviations

CI

confidence interval

PFO

patent foramen ovale

RCT

randomized controlled trial