Patent foramen ovale closure for patients with cryptogenic stroke: A systematic review and comprehensive meta‐analysis of 5 randomized controlled trials and 14 observational studies

Summary

Background

Previous review from randomized controlled trials (RCT) showed that patients with cryptogenic stroke may benefit from patent foramen ovale (PFO) closure. However, the findings from the systematic review were not clear when observational studies were also included.

Methods

We searched MEDLINE, Embase, and Cochrane databases. The primary endpoints were recurrent stroke or transient ischemic attack (TIA). The secondary outcomes were all‐cause death, atrial fibrillation (AF), and hemorrhagic events.

Results

Five randomized trials and fourteen observational studies (6301 participants) were eligible. PFO closure was superior to medical therapy for stroke prevention risk ratios ([RR], 0.38; 95% CI, 0.24‐0.60), but showed increased risk of AF (RR, 4.96; 95% CI, 2.31‐10.7). There was no significant difference in TIA recurrence, death, and hemorrhagic events. Subgroup analyses showed that patients with factors such as substantial residual shunt, the presence of atrial septal aneurysm (ASA), male, and age <45 years had a lower risk of recurrent stroke when PFOs were closed.

Conclusions

In patients with cryptogenic stroke, PFO closure does appeared to be superior to medical therapy in stroke prevention, with an increased incidence of AF. Male, age <45 years, substantial residual shunt, and the history of ASA are the factors that will predict the benefit when PFO is closed.

1. INTRODUCTION

Stroke can be devastating. About 30%‐40% stroke patients without a clear cause were likely related to cryptogenic stroke.1 The prevalence of a patent foramen ovale (PFO) in patients younger than 60 years of age with cryptogenic stroke is about 50%, almost twice as much as those in the general population.2 Paradoxical embolism has presumed to be the cause in these patients.3

The eleven observational studies in a meta‐analysis suggested PFO closure was a better intervention than medical treatment.4 Until now, 5 randomized controlled trials (RCTs) have evaluated mechanical closure in patients with PFO but without obvious other causes of their index stroke. Recently, AHA/ASA (American Heart Association/American Stroke Association) guideline indicated that all 5 trials had potential bias because of unblinded investigators were deciding on which events should be referred for blinded endpoint adjudication.5 Three of the 5 trials had more patients lost to follow‐up. Other 2 trials had different and highly restrictive eligibility criteria, used different closure devices, and had different guidelines for antithrombotic therapy. In contrary, the results from recent RCTs showed a lower risk of recurrent stroke when PFO was closed comparing to medical therapy.6, 7, 8, 9, 10, 11 This finding has been supported by a meta‐analysis of RCTs recently.12

In order to further examine the side effects of PFO closure, especially the associated incidence of atrial fibrillation (AF),12, 13, 14 we aim to test the relative effectiveness of PFO closure compared with medical therapy in patients enrolled in RCTs and observational studies.

2. METHOD

2.1. Search strategy

This systematic review was performed according to the Preferred Reporting Items for Systematic Reviews and Meta‐Analyses guidelines.15 MEDLINE, Embase, and Cochrane databases were reviewed using the search terms “patent foramen ovale,” “PFO,” “stroke,” “percutaneous closure,” “cryptogenic stroke,” “atrial septal aneurysm,” “heart septal defects,” “interatrial shunt,” “PFO closure,” “patent foramen ovale closure,” “transcatheter closure,” “percutaneous closure,” “atrial percutaneous closure,” “antiplatelet therapy,” “anticoagulant therapy,” “medical therapy,” “TIA,” “transient ischemic attack,” “recurrent stroke,” “mortality,” “recurrent thromboembolism,” “recurrence of embolic events,” “brain infarctions,” and “transcatheter closure,” with no language restrictions up to October 2017. The relevant research references were also reviewed for the purpose of finding more trials.

Candidates included in this analysis met the following criteria: (i) patients with a history of cryptogenic stroke or transient ischemic attack (TIA) and the presence of PFO; (ii) comparison of PFO closure and medical therapy in preventing recurrence of embolic events; (iii) clinical and outcome data available for both PFO closure and medical therapy groups; (iv) follow‐up ≥1 year; and (v) RCTs or observational studies. Studies were excluded if the data had duplication; if the later studies were showing updated long‐term outcomes; if the studies were ongoing or unpublished; and if the articles were published as conference proceedings or abstracts.

2.2. Data extraction

Data extraction was performed independently by 2 researchers (Xi Chen and Shi‐Dong Chen) according to the selection criteria. Any uncertainties or disagreements were resolved by other authors. The primary endpoints of the meta‐analysis were incidence of recurrent ischemic stroke (fatal or non‐fatal) or TIA. The secondary endpoints included death (cardiovascular and noncardiovascular death during the follow‐up), new‐onset AF, hemorrhagic events (minor or major intracerebral and non‐intra‐cerebral), and any adverse results (device or catheterization‐related complications, hospitalizations related to PFO or its treatment, conduction abnormality or new arrhythmias, acute myocardial infarction, etc.).

2.3. Study quality assessment

Two researchers independently assessed the study quality using the Review Manager (version 5.3) risk‐of‐bias assessment tool according to the 4 bias domains: selection, performance/detection, attrition, and reporting bias for RCTs and the Newcastle‐Ottawa Scale (NOS) for assessing the quality of nonrandomized studies.16, 17

2.4. Statistical analysis

Dichotomous data were presented as risk ratios (RR) and corresponding 95% confidence intervals (CI) using both fixed‐effect and random‐effect models.18 Heterogeneity between trials was assessed by measuring inconsistency using the I ² statistic and chi‐squared test (P < 0.10 for statistical significance).16 Interestingly, comparing the characteristics of patients in REDUCE, CLOSE, and RESPECT, which released their updated results in September 2017 to that in PC and CLOSURE I, it seemed that small proportion of patients with small shunt or without being on ASA showed significant difference in efficacy between closure and medicine groups. Besides, results were contradictory from the subgroup analysis among RESPECT, REDUCE, CLOSURE, and PC. Therefore, we selected age (<45 years old (yrs) or ≥45 years), gender (male or female), the presence of a shunt of substantial, moderate or trace size, and the presence or absence of ASA to test whether these 4 factors would enhance the benefits from the PFO closure in our subgroup analyses.

Sensitivity analyses were tested by excluding individual studies, and the systematic bias was examined by funnel plots. More quantitative syntheses of efficacy endpoints of observational studies were conducted by excluding trials with patients’ number fewer than 100/200, trials with quality scores lower than 7, and trials with fewer than 3 years of follow‐up. Testing for systematic bias was conducted using funnel plots. All analyses were performed with the R, V3.4.3 (package “meta”). All P values were 2‐sided with an alpha of 0.05.

3. RESULTS

The study selection process was illustrated in Figure 1. We identified a total of 728 articles, from which 5 were RCTs6, 7, 8, 10, 11 with 3627 participants and 14 were observational studies 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32 with 2674 patients. They were ultimately selected for the present meta‐analysis. Overall, publication bias in this meta‐analysis was demonstrated in Figure S1, and the quality of the included articles was modest (Table S1 and Figure 2). The baseline characteristics of the included RCTs and observational studies are shown in Tables S2 and S3. The primary endpoints were recurrent stroke or TIA, of which data were analyzed on the basis of intention‐to‐treat (ITT) (if possible). There was statistically significant difference in the incidence of recurrent stroke compared with that of medical therapy group (RR = 0.38, 95% CI = 0.24‐0.60; I ²  = 31%, P = 0.1); RR = 0.45 (95% CI = 0.23‐0.90; I ² = 51%, P = 0.09) among 5 RCTs, and RR = 0.32 (95% CI = 0.17‐0.60; I ²  = 21%, P = 0.22) among 14 observational studies. However, it did not reach a statistical significance in the rate of TIA recurrence (RR = 0.71, 95% CI = 0.42‐1.21; I ² = 58%, P < 0.01) and with significant heterogeneity. For 4 RCTs, RR = 0.80 (95% CI = 0.55‐1.18; I ² = 0%, P = 0.88); while for observational studies, RR = 0.60 (95% CI = 0.22‐1.64; I ²  = 69%, P < 0.01) (Figure 3). Sensitivity analysis of TIA recurrence is shown in Figure S2, which suggested that studies from Schuchlenz (2005) and Faggiano (2012) might be the source of significant heterogeneity. Once these 2 studies were excluded, there was no evidence of heterogeneity in the remaining studies (I ²  = 30%, P = 0.19).

Figure 1.

Flow diagram of the study selection process. RCT, randomized controlled trial; non‐RCT, nonrandomized controlled trial

Figure 2.

Risk of bias summary demonstrated by risk of bias graph. Authors’ judgments about each risk of bias item presented as percentages of all included randomized controlled trials; authors’ judgments about each risk of bias item of all included randomized controlled trials

Figure 3.

Forest plot of risk ratios (RR) for recurrent stroke (top) and transient ischemic attack (bottom). Closure = percutaneous patient foramen oval closure; BMT = best medical therapy; RCT = randomized controlled trials; Observational = observational studies; CI = confidence interval

New‐onset AF occurred more frequently among patients undergoing PFO closure than that in the medical therapy (RR = 4.96, 95% CI = 2.31‐10.7; I ² = 28%, P = 0.23). Among the RCTs, the RR was 5.16 (95% CI = 2.20‐12.1; I ² = 41%, P = 0.15), while in the observational studies, the RR was 2.90 (95% CI = 0.12‐70.5).

The pooled RR of all‐cause mortality was 0.73 (95% CI = 0.41‐1.30; I ² = 0%, P = 0.90). For the RCTs, RR was 0.73 (95% CI = 0.35‐1.55; I ² = 0%, P = 0.63), and for the non‐RCTs, RR was 0.73 (95% CI = 0.30‐1.77; I ² = 0%, P = 0.81). Also, the results in both RCTs and observational studies were consistent in the rate of hemorrhagic events (RR = 0.73, 95% CI = 0.39‐1.36; I ² = 38%, P = 0.12). Among observational studies, RR was 0.50 (95% CI = 0.15‐1.73; I ² = 30%, P = 0.23), while among the RCTs, RR was 0.86 (95% CI = 0.40‐1.87; I ² = 49%, P = 0.1, Figure 4). In addition, any adverse results (RR = 1.05, 95% CI = 0.95‐1.17; I ² = 0%, P = 0.41) showed no significant difference between PFO closure arm and medical treatment arm (Figure S3). In subgroup analyses of the included RCTs, the above outcomes were analyzed by ASA, age, gender, and residual shunt size to explore the impact of patient characteristics. In a random‐effect model, Table 1 indicates that patients with substantial residual shunt (RR = 0.26, 95% CI = 0.10‐0.67; I ² = 0%, P = 0.55), the presence of ASA (RR = 0.19, 95% CI = 0.06‐0.59; I ² = 0%, P = 0.56), age <45 years (RR = 0.40, 95% CI = 0.18‐0.86; I ² = 0%, P = 0.54), and male (RR = 0.31, 95% CI = 0.10‐0.96; I ² = 51%, P = 0.13) had a lower risk of a recurrent stroke when their PFO was closed.

Figure 4.

Forest plot of risk ratios (RRs) for atrial fibrillation (top), all‐cause mortality (median), and any hemorrhagic event (bottom). Closure = percutaneous patient foramen oval closure; BMT = best medical therapy; RCT = randomized controlled trials; Observational = observational studies; CI = confidence interval

Table 1.

The results of subgroup analyses of randomized controlled trials

RCT	Outcome	Subgroup	Studies, n	Random effects	Fixed effects	I 2, %	Heterogeneity, P
Age	Stroke	<45 years’ old	3	0.40 (0.18, 0.86)	0.40 (0.18, 0.86)	0	0.54
		≥45 years’ old	3	0.37 (0.13, 1.06)	0.49 (0.27, 0.91)	43	0.17
Gender	Stroke	Male	3	0.31 (0.10, 0.96)	0.42 (0.22, 0.80)	51	0.13
		Female	3	0.51 (0.25, 1.05)	0.51 (0.25, 1.05)	0	0.73
Shunt size	Stroke	None, trace or moderate	1	1.07 (0.50, 2.30)	1.07 (0.50, 2.30)	1	NA
		Substantial	2	0.26 (0.10, 0.67)	0.26 (0.10, 0.67)	0	0.55
ASA	Stroke	Present	2	0.19 (0.06, 0.59)	0.19 (0.06, 0.59)	0	0.56
		Absent	1	0.97 (0.48, 1.95)	0.97 (0.48, 1.95)	1	NA
	Composite	Present	2	1.05 (0.46, 2.42)	1.05 (0.46, 2.42)	0	0.32
		Absent	2	0.64 (0.27, 1.54)	0.72 (0.41, 1.29)	41	0.19

RCT, randomized controlled trials; ASA, atrial septal aneurysm; Composite = outcome (death/vascular events); Significant results are in bold.

In stratified analyses of RCTs and observational studies (Table 2), the occurrence of recurrent stroke was lower in those received Helex or Cardioform Septal Occluder device (RR = 0.25, 95% CI = 0.10‐0.66) from one RCT trial and Amplatzer PFO Occluder device (RR = 0.47, 95% CI = 0.23‐0.97) from 2 RCTs and 2 observational studies. Those received Helex or Cardioform Septal Occluder (RR = 14.7, 95% CI = 2.10‐107) or STARFlex Septal Closure (RR = 8.73, 95% CI = 2.64‐28.9) had an increased risk of new‐onset AF. In addition, there was no difference in other endpoints.

Table 2.

The results of stratified analyses of randomized controlled trials (RCTs) and observational studies compared with different device

Outcome	Device	Studies, na	Studies, na	Random effects	Fixed effects	I 2, %	Heterogeneity, P
Stroke	Helex or Cardioform Septal Occluder	1	0	0.25 (0.10, 0.66)	0.25 (0.10, 0.66)	1	NA
	Amplatzer PFO Occluder	2	2	0.47 (0.23, 0.97)	0.53 (0.31, 0.90)	9	0.35
	STARFlex Septal Closure	1	0	0.95 (0.44, 2.07)	0.95 (0.44, 2.07)	1	NA
TIA	Amplatzer PFO Occluder	2	2	0.70 (0.30, 1.64)	0.71 (0.42, 1.19)	34	0.21
	STARFlex Septal Closure	1	0	0.79 (0.39, 1.61)	0.79 (0.39, 1.61)	1	NA
All‐cause death	Helex or Cardioform Septal Occluder	1	0	2.53 (0.12, 52.50)	2.53 (0.12, 52.50)	1	NA
	Amplatzer PFO Occluder	2	1	0.71 (0.30, 1.68)	0.71 (0.30, 1.68)	0	0.4
	STARFlex Septal Closure	1	0	0.51 (0.09, 2.79)	0.51 (0.09, 2.79)	1	NA
Hemorrhagic events	Helex or Cardioform Septal Occluder	1	0	0.67 (0.24, 1.92)	0.67 (0.24, 1.92)	1	NA
	Amplatzer PFO Occluder	2	1	0.64 (0.14, 2.98)	0.69 (0.31, 1.50)	50	0.13
	STARFlex Septal Closure	1	0	2.47 (0.78, 7.82)	2.47 (0.78, 7.82)	1	NA
AF	Helex or Cardioform Septal Occluder	1	0	14.66 (2.01, 106.95)	14.66 (2.01, 106.95)	1	NA
	Amplatzer PFO Occluder	2	1	1.94 (0.72, 5.20)	1.94 (0.72, 5.20)	0	0.46
	STARFlex Septal Closure	1	0	8.73 (2.64, 28.87)	8.73 (2.64, 28.87)	1	NA
Any serious adverse event	Helex or Cardioform Septal Occluder	1	0	0.83 (0.63, 1.09)	0.83 (0.63, 1.09)	1	NA
	Amplatzer PFO Occluder	2	0	1.13 (0.97, 1.31)	1.13 (0.97, 1.31)	0	0.76
	STARFlex Septal Closure	1	0	1.02 (0.76, 1.37)	1.02 (0.76, 1.37)	1	NA

AF, atrial fibrillation; TIA, transient ischemic attack; PFO, patent foramen ovale.

Significant results are in bold.

Number of RCT.

^aNumber of observational studies.

Results of the quantitative syntheses from the data of observational studies comparing the efficacy between PFO closure and medication are listed in Table S4. The findings showed that PFO closure did have an effect on stroke prevention in observational studies, however, not on TIA or mortality.

4. DISCUSSION

Our meta‐analysis including both RCTs and observational studies showed a possible benefit from PFO closure, compared to medical treatment, for the prevention of recurrent ischemic stroke. However, there was no effect on TIA recurrence, all‐cause mortality, other serious adverse events, or hemorrhagic events. On the other hand, PFO closure might increase the risk of new‐onset of AF but mostly transient. Subgroup analyses indicated that patients who had substantial residual shunt or the presence of ASA might benefit more from closure with less recurrent strokes. Males or patients at an age <45 years also had the benefit of stroke prevention if PFO was closed.

Our findings were consistent with the findings of one recent meta‐analysis of 5 RCTs.12 We added all high‐quality observational studies in our meta‐analysis, which showed more general efficacy from PFO closure. In our analysis, we also added subgroup of gender, age, ASA, and shunt size for which all had clinical significance. We therefore recommend that patients’ characteristics need to be taken into consideration when choosing a treatment method. For example, if patients with substantial size of residual shunt or on ASA have a stroke, they may have a better outcome if their PFOs are closed. Male or younger (under 45 years) patients were also more likely to benefit from PFO closure.

The potential mechanism of having benefits when PFO are closed in patients with ASA might be from the benefit of preventing thrombi formation in the septal aneurysm or as a result of atrial arrhythmias. In addition, venous source of paradoxical embolism could be avoided.33 Similarly, a stronger association between the presence of PFO with ASA on board and cryptogenic stroke in younger patients (<55 years) was confirmed by the large case‐control studies, which probably explained that the ideal patient population showing superiority of the PFO closure are the younger population (under 45 years) with a lower atherosclerotic burden.34 Furthermore, we can hypothesize that larger shunt size could increase the risk of paradoxical flow and hence cause stroke. However, gender factor that impacting the outcome remained unexplained and requires further studies.

We also provided an updated meta‐analysis of 14 observational studies to enhance our findings, which was different from an analysis in 2014 that only included 11 observational studies.4, 35 Some studies that were not included in the previous articles such as the Korean Population study from Moon et al19 were included in our analysis. The design of van de Wyngaert et al27 study was different from the other observational studies for it was a before‐after type of study that might account for the heterogeneity at the endpoint.

There was no significant heterogeneity among the included studies except for the outcome of TIA in the observational studies. Notably, we found that studies by Schuchlenz et al22, 32 and Faggiano et al might be the sources of statistical heterogeneity revealed by the sensitivity analysis on recurrent TIA. Therefore, we removed these 2 studies and avoided potential heterogeneity in the endpoint analysis of stroke recurrence from other 10 remaining observational studies. The study from Schuchlenz et al was the earliest research conducted, but the quality of methodological (score = 5) was the lowest among all studies with a much shorter duration. The accurate incidence of recurrent stroke and TIA was unavailable from Faggiano et al.

Data in network meta‐analysis showed that Amplatzer PFO occluder was superior to medical treatment, while no significant difference was found in STARFlex and Helex PFO occluder.36 When updated RCTs were included, we found patients who had their PFO closed with Helex or Cardioform Septal Occluder or Amplatzer PFO may have a potential benefit in the prevention of recurrent stroke. For adverse safety events, Helex or Cardioform Septal Occluder and STARFlex Septal Closure might increase the risk of new‐onset AF. Therefore, Amplatzer PFO might be better in terms of safety compared with Helex or Cardioform Septal Occluder. However, the device‐specific differences were not clarified by the traditional meta‐analysis; hence, the need for an updated network meta‐analysis was needed to generate better hierarchies of treatment efficacy and safety from different device.36 In addition, a pooled data analysis of RCTs is still needed to explain more details of the role of gender, age, ASA, or shunt size.

The 2016 American Academy of Neurology guidelines recommended that antiplatelet medications be routinely offered to patients with cryptogenic strokes and PFO rather than anticoagulation.37 It also recommended that percutaneous PFO closure with Amplatzer PFO Occluder could also be applied in rare circumstances, such as for those with recurrent strokes despite adequate medical therapy and with no other mechanism identified.37 However, from the new findings based on our meta‐analysis of the data of 5 updated RCTs and 14 observational studies, an updated recommendation on closing PFO for stroke prevention is on the horizon.

Strengths of our study included the use of Cochrane risk‐of‐bias assessment tool and Newcastle‐Ottawa Scale for assessing the quality of studies, the conduction of this meta‐analysis according to the Preferred Reporting Items for Systematic Reviews and Meta‐Analyses guideline and subgroup analyses of gender, age, shunt size and ASA usage.

There were several limitations in our meta‐analysis. Firstly, there were differences in the study design, endpoint definition, and lengths of follow‐up among selected studies. We have tried to list out all the causes contributed to heterogeneity. Secondly, medical treatment group was not always standardized. Most were at physician’s discretion. Thirdly, data from 2 trials remained unavailable after we have tried to acquire the data from the authors.22, 26 Lastly, various kinds of devices were applied in studies and there were publication biases.

5. CONCLUSION

Based on this updated meta‐analysis of RCTs and observational studies, percutaneous PFO closure for patients with cryptogenic stroke/TIA and PFO was superior to medical therapy in secondary stroke prevention, but with an increased incidence of transient AF. The benefit of PFO closure might be greater in the patients who are male, below 45 years, having substantial residual shunt or ASA. Further studies are still needed to differentiate the effectiveness between different closing devices.

CONFLICT OF INTEREST

The authors declare no conflict of interest.

Supporting information

 

Chen X, Chen S‐D, Dong Y, Dong Q. Patent foramen ovale closure for patients with cryptogenic stroke: A systematic review and comprehensive meta‐analysis of 5 randomized controlled trials and 14 observational studies. CNS Neurosci Ther. 2018;24:853–862. 10.1111/cns.12980