Michel Pompeu Barros Oliveira SáI,II,III,IV; Luiz de Albuquerque Pereira de Oliveira NetoI,II,IV; Gabriella Caroline Sales do NascimentoI,II,IV; Erik Everton da Silva VieiraI,II,IV; Gabriel Lopes MartinsI,II,IV; Karine Coelho RodriguesI,II,IV; Giulia Cioffi NascimentoI,II; Alexandre Motta de MenezesI,II; Ricardo Felipe de Albuquerque LinsI,II; Frederico Pires Vasconcelos SilvaI,II; Ricardo Carvalho LimaI,II,III,IV
ABSTRACTObjective: We aimed to determine whether patent foramen ovale closure reduces the risk of stroke, also assessing some safety outcomes.
AHA = American Heart Association
ASA = American Stroke Association
CI = Confidence interval
LILACS = Literatura Latino-Americana em Ciências da Saúde
MeSH = Medical Subject Headings
PFO = Patent foramen ovale
PICOS = Population, Intervention, Comparison, Outcome and Study design
PRISMA = Preferred Reporting Items for Systematic Reviews and Meta-Analyses
RCTs = Randomized controlled trials
RR = Risk ratio
SciELO = Scientific Electronic Library Online
SE = Standard error
The clinical benefit of closing a patent foramen ovale (PFO) after a cryptogenic stroke has been an open question for several decades. Current American Heart Association (AHA)/American Stroke Association (ASA) guidelines do not support the use of PFO closure among patients with PFO and cryptogenic stroke. Nevertheless, new randomized controlled trials (RCTs) were published recently, so that controversy still exists over the preferred management strategy for patients with cryptogenic stroke and PFO. Therefore, it is necessary to review the current state of published medical data in this regard.
We aimed to determine whether PFO closure reduces the risk of stroke, also assessing some safety outcomes. This analysis was planned in accordance with current guidelines for performing comprehensive systematic reviews and meta-analysis with meta-regression, including the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines for RCTs. We prespecified our analytical plan and registered the study protocol with PROSPERO, the international prospective register of systematic reviews (CRD42018084583).
With the PICOS (Population, Intervention, Comparison, Outcome and Study design) strategy, studies were considered if: 1) the population comprised patients with recent stroke or transient ischemic attack who had a PFO; 2) there was an intervention group of device closure; 3) there was a control group receiving medical therapy; 4) outcomes studied included any of the following: stroke, death, major bleeding, atrial fibrillation; 5) studies were RCTs.
The following databases were used (until December 2017): MEDLINE; EMBASE; CENTRAL/CCTR (Cochrane Controlled Trials Register); ClinicalTrials.gov; Scientific Electronic Library Online (SciELO); LILACS (Literatura Latino-Americana em Ciências da Saúde); Google Scholar; and reference lists of relevant articles.
We conducted the search with Medical Subject Headings (MeSH) terms ('Foramen Ovale, Patent' OR 'Patent Oval Foramen ' OR 'Oval Foramen, Patent' OR 'Patent Foramen Ovale') AND ('Stroke' OR 'Cerebrovascular Accident' OR 'Cerebrovascular Accidents' OR 'CVA' OR 'CVAs' OR 'Cerebrovascular Apoplexy' OR 'Apoplexy, Cerebrovascular' OR 'Vascular Accident, Brain' OR 'Brain Vascular Accident ' OR 'Brain Vascular Accidents' OR 'Vascular Accidents, Brain' OR 'Cerebrovascular Stroke' OR 'Cerebrovascular Strokes' OR 'Stroke, Cerebrovascular' OR 'Strokes, Cerebrovascular' OR 'Apoplexy' OR 'Cerebral Stroke' OR 'Cerebral Strokes' OR 'Stroke, Cerebral' OR 'Strokes, Cerebral' OR 'Stroke, Acute' OR 'Acute Stroke' OR 'Acute Strokes' OR 'Strokes, Acute' OR 'Cerebrovascular Accident, Acute' OR 'Acute Cerebrovascular Accident' OR 'Acute Cerebrovascular Accidents' OR 'Cerebrovascular Accidents, Acute').
The following steps were taken: 1) identification of titles of records through databases searching; 2) removal of duplicates; 3) screening and selection of abstracts; 4) assessment for eligibility through full-text articles; and 5) final inclusion in study. One reviewer followed steps 1 to 3. Two independent reviewers followed step 4 and selected studies. Inclusion or exclusion of studies was decided unanimously. When there was disagreement, a third reviewer made the final decision.
The crude endpoints were stroke, death (any cause), major bleeding and atrial fibrillation.
Data Collection Process
Two independent reviewers extracted the data. When there was disagreement about data, a third reviewer checked the data and made the final decision. From each study, we extracted patient characteristics, study design, and outcomes.
Risk of Bias in Individual Studies
Included studies were assessed for the following characteristics: sequence generation (randomization); allocation concealment (selection bias); blinding of participants and personnel (performance bias); blinding of outcome assessors (detection bias); incomplete outcome data addressed (attrition bias) and selective outcome reporting (reporting bias). Considering these characteristics, the papers were classified into A (low risk of bias), B (moderate risk of bias) or C (high risk of bias). Two independent reviewers assessed risk of bias. Agreement between the two reviewers was assessed with kappa statistics for full-text screening and rating of relevance and risk of bias. When there was disagreement about risk of bias, a third reviewer checked the data and made the final decision.
The principal summary measures were risk ratio (RR) with 95% CI and P values (considered statistically significant when P<0.05) for stroke, death, major bleeding and atrial fibrillation. The metaanalysis was completed with the software Comprehensive Meta-Analysis (version 2, Biostat, Inc., Englewood, NJ, USA).
Synthesis of Results
Forest plots were generated for graphical presentations of clinical outcomes, and we performed the I2 test and χ2 test for the assessment of heterogeneity across the studies. Inter-study heterogeneity was explored using the χ2 statistic, but the I2-value was calculated to quantify the degree of heterogeneity across the studies that could not be attributable to chance alone. When I2 was more than 50%, significant statistical heterogeneity was considered to be present. Each study was summarized by the difference in means or RR, depending on the outcome analyzed. The RR and differences in means were combined across studies using a weighted DerSimonian-Laird random effects model.
Risk of Bias Across Studies
To assess publication bias, a funnel plot was generated for each outcome, statistically assessed by Begg and Mazumdar's test and Egger's test.
Sensitivity analyses included the investigation of the influence of a single study on the overall effect - by sequentially removing one study - in order to test the robustness of the main results, so that we could verify whether any study had an excessive influence on the overall results. Moreover, we also analyzed the pool data regarding the outcome "stroke" according to the presence (or absence) of atrial septal aneurysm (hypermobile septum, defined as a septum primum excursion greater than 10 mm).
Meta-regression analyses were performed to determine whether the effects of the PFO closure were modulated by prespecified factors. Meta-regression graphs describe the effect of aspirin on the outcome (plotted on the y-axis) as a function of a given factor (plotted as a mean or proportion of that factor on the x-axis). Meta-regression coefficients show the estimated increase in log RR per unit increase in the covariate. Since log RR > 0 corresponds to RR > 1 and log RR < 0 corresponds to RR < 1, a negative coefficient would indicate that as a given factor increases, the RR decreases, and vice versa.
The predetermined modulating factors to be examined were: age (mean - years), male gender (%), hypertension (%), smoking (%), large shunt before the interventions, atrial septal aneurysm and effective closure (freedom from large shunt after the interventions).
A total of 3,740 citations were identified, of which 9 studies were potentially relevant and retrieved as full-text. Five[7-11] publications fulfilled our eligibility criteria. Interobserver reliability of study relevance was excellent (Kappa = 0.81). Agreement for decisions related to study validity was very good (Kappa = 0.83). The search strategy can be seen in Figure 1.
A total of 3,440 patients (device closure: 1,829 patients; medical therapy: 1,611 patients) were included from studies published from 2012 to 2017. All the trials were multicentric. Most studies consisted of patients whose mean or median age was approximately the fourth decade of life. The medical therapy in the studies was not homogeneous, since different regimens were applied (aspirin, clopidogrel, dipyridamole, combined regimens, etc.). The same goes for the devices used, being the CLOSE trial noteworthy for applying various devices (see Table 1). The overall internal validity was considered "low risk of bias" (Table 2).
|CLOSE (N=473)||REDUCE (N=664)||PC (N=414)||RESPECT (N=980)||CLOSURE (N=909)|
|% of data in metanalysis||13.7||19.3||12.0||28.4||26.4|
|Age ± SD, years||43.3±10.3||45.1±9.45||44.5±10.2||45.4±9.8||45.5±10.2|
|Medical history variables|
|Current smoking (%)||28.9||13.3||23.9||13.3||15.2|
|Coronary artery disease (%)||NR||NR||1.9||2.9||2.1|
|Prior stroke/TIA (%)||3.6||85||37.4||18.6||12.5|
|Atrial septal aneurysm (%)||32.7||NR||23.7||35.6||35.6|
|Large shunt (%)||92.8||39.3||21.7||76.1||61.1|
|Randomized to device closure (%)||50.3||66.4||49.3||50.9||49.2|
|Treated with antiplatelets only (%)||49.6||33.6||80.0||88.0||84.7|
|Amplatzer PFO Occluder or Cribriform; Starflex; CardioSeal; Intrasept PFO; PFOStar; Helex; Premere; PFO occluder OCCLUTECH; PFO occluder GORE (GSO)||EITHER the Helex Septal Occluder device OR the Cardioform Septal Occluder||Amplatzer PFO Occluder (St. Jude Medical)||Amplatzer PFO Occluder (disc occluder)||STARFlex septal closure system (umbrella occluder)|
|Study||Randomization||Selection bias||Performance bias||Detection bias||Attrition bias||Reporting bias|
|CLOSURE I 2012||A||A||A||A||A||A|
Synthesis of Results
The RR for stroke in the "device closure" group compared with the "medical therapy" group in each study is reported in Figure 2. There was evidence of moderate heterogeneity of treatment effect among the studies for stroke. The overall RR (95% CI) of stroke showed a statistically significant difference between the groups, favouring the "device closure" group (random effect model: RR 0.400; 95% CI 0.183-0.873, P=0.021).
The RR for death in the "device closure" group compared with the "medical therapy" group in each study is reported in Figure 3A. There was no evidence of heterogeneity of treatment effect among the studies for death. The overall RR (95% CI) of death showed no statistically significant difference between the groups (random effect model: RR 0.760; 95% CI 0.308-1.877, P=0.552).
The RR for major bleeding in the "device closure" group compared with the "medical therapy" group in each study is reported in Figure 3B. There was evidence of mild heterogeneity of treatment effect among the studies for major bleeding. The overall RR (95% CI) of major bleeding showed no statistically significant difference between the groups (random effect model: RR 0.945; 95% CI 0.468-0.873, P=1.911).
The RR for atrial fibrillation in the "device closure" group compared with the "medical therapy" group in each study is reported in Figure 3C. There was evidence of mild heterogeneity of treatment effect among the studies for atrial fibrillation. The overall RR (95% CI) of atrial fibrillation showed a statistically significant difference between the groups (random effect model: RR 4.000; 95% CI 2.262-7.092, P<0.001).
Risk of Bias Across Studies
Funnel plot analysis (Figure 4) disclosed no asymmetry around the axis for the outcomes stroke, major bleeding and atrial fibrillation, which means that we have low risk of publication bias related to these outcomes. However, we detected a possibility of publication bias for the outcome death.
Sensitivity analyses performed by removing each single study from the meta-analysis to determine the influence of individual data sets to the pooled RR, showed that none of the studies had a particular impact on the results (Figure 5).
Searching for evidence of a particular impact of the presence of an atrial septal aneurysm on the results, we detected no difference between the groups (Figure 6). Unfortunately, the REDUCE trial was left out of this last analysis because the presence of an atrial septal aneurysm was determined at the time of the PFO closure procedure and, therefore, it was not recorded before trial entry or among the patients in the antiplatelet-only group.
Meta-regression coefficients were statistically significant for, age, hypertension, atrial septal aneurysm and effective closure regarding the outcome "stroke". For the variables age, hypertension and atrial septal aneurysm, we observed that the older the patients, the larger the proportion of patients with hypertension and the larger the proportion of patients with atrial septal aneurysm, the higher the risk of stroke (Figures 7A, 7B, 7C). Conversely, the larger the proportion of effective closure, the lower the risk of stroke (Figure 7D).
Summary of Evidence
To our knowledge, this is the largest meta-analysis of studies performed to date that provides incremental value by demonstrating that patients seem to benefit from device closures in comparison to medical therapy in the reduction of the rate of stroke. On the other hand, there was an increase in the rates of atrial fibrillation. We did not identify the group of patients with an atrial septal aneurysm as a particular group that benefits from the device closure in the sensitivity analysis, although we identified this variable as a risk marker for stroke in the meta-regression. We also observed that the benefit of the device closure in the reduction of the rates of stroke hinges on the rate of effective closure.
About 25% of the population has a PFO, but the condition in itself does not increase the risk of ischemic stroke. PFO is more prevalent, however, among patients who had a cryptogenic ischemic stroke than in the general population. Therefore, we must be careful when selecting patients who would receive some benefit of PFO closure (in term of the risk of stroke).
Kent et al. carried out a patient-level analysis of the CLOSURE I, PC and RESPECT trials (before the publication of the CLOSE and REDUCE trials), demonstrating that the device closure was superior to medical therapy, which turned out to be confirmed in our meta-analysis (in terms of the outcome stroke). The improved efficacy in the CLOSE and REDUCE trials might be owing to more strict patient selection. The REDUCE trial had a very strict criteria to exclude patients with other sources of stroke and the CLOSE trial only included those with an atrial septal aneurysm or large shunt.
The Role of Atrial Septal Aneurysm
Our results do not suggest that this purportedly high-risk anatomical feature is, by itself, very useful at discriminating patients likely to benefit from closure from those unlikely to benefit. Nevertheless, von Klemperer et al., in a survey of current practice in the United Kingdom, identified that around 80% of the 120 respondents (including cardiologists, stroke physicians and neurologists) agreed that an aneurysmal septum was more likely to implicate the PFO in stroke. Only the CLOSE and RESPECT trials showed isolatedly this difference (as we can see in the Figure 6A), but the pooled analysis did not confirm this finding. On the other hand, we might well recognize that there is a correlation between the presence of an atrial septal aneurysm and the risk of stroke (as we can see in the meta-regression - Figure 7C), which led us to the conclusion that this factor is rather a risk marker than a risk factor. Nevertheless, there is no evidence that we should see it as a primary discriminator between those who should have a PFO closed by means of a device.
The Role of the Effective Closure
The lack of efficacy observed in the CLOSURE I trial has been put down to ineffective PFO closure in the device arm, with 14% demonstrating significant residual right-to-left shunting, whereas, in the other trials, we observed the following rates: 7% (CLOSE), 5.5% (REDUCE), 6.5% (RESPECT) and 6.5% (PC). Our meta-regression showed that the more successful the closure, the lower the risk of stroke in the device group (see Figure 7D). Therefore, we must bare in mind that "procedural success", which was defined in the studies as successful implantation with no complications, does not mean "success of PFO closure", which was defined in the studies as minimal or no shunt after the procedure.
The Problem of Atrial Fibrillation After the Procedure
The rate of new-onset atrial fibrillation was significantly higher in the PFO closure group than in antiplatelet group in our meta-analysis, with most cases detected within 1 month after the procedure - a finding that suggests that the procedure itself induces atrial fibrillation. Indeed, in the closure group, most of the observed cases of atrial fibrillation were periprocedural. The risk of stroke from atrial fibrillation induced by PFO closure has not been determined in the CLOSE trial. In the REDUCE trial, atrial fibrillation was more commonly reported in the PFO closure group, but it was usually transient and the clinical relevance of atrial fibrillation related to closure and overall risk of stroke requires further investigation. In the CLOSURE I trial, a quarter of the strokes in the closure group were ascribed to atrial fibrillation, and in two of these cases, the patients had deviceassociated thrombus on transesophageal echocardiography.
Future Data to Come Out
At this moment, Song et al. are carrying out the DEFENSE-PFO trial (Device Closure Versus Medical Therapy for Cryptogenic Stroke Patients with High-Risk Patent Foramen Ovale - ClinicalTrials.gov Identifier: NCT01550588), which will shed some additional light on this issue by assessing whether percutaneous device closure of PFO is superior to conventional antithrombotic treatment in preventing stroke recurrence in the cryptogenic stroke patients with high-risk of PFO, which was defined as highrisk of recurrence (PFO size ≥ 2 mm or atrial septal aneurysm or hypermobility by transesophageal echocardiography. This study started in 2012 and will be finished in 2020.
Risk of Bias and Limitations of the Present Study
There are inherent limitations with meta-analyses, including the use of cumulative data from summary estimates. Patient data were gathered from published data, not from individual patient follow-up. Access to individual patient data would have enabled us to conduct further subgroup analysis and propensity analysis to account for differences between the treatment groups. This meta-analysis included only data from randomized studies, which do not reflect the "real world" but, on the other hand, are less limited by publication bias, treatment bias, confounders, and a certain tendency to overestimate treatment effects observed in the observational studies, since patient selection alters outcome and thus make non-randomized studies less robust.
Moreover, besides statistical heterogeneity in some analyses, there is also the issue of the clinical heterogeneity that might have played some role in the pooled results. For instance, in the CLOSE trial, 11 different devices were appplied for PFO closure. In the antiplatelet-only group and the PFO closure group, 410 (86.7%) patients received aspirin, 51 (10.8%) received clopidogrel, 6 (1.3%) received aspirin with extended-release dipyridamole, and 6 (1.3%) received aspirin with clopidogrel. As we can see, not all of patients were 100% equally treated.
This meta-analysis found that stroke rates are lower with percutaneously implanted device closure than with medical therapy alone, being these rates modulated by the rates of effective closure.
What is known?
The results of the firstly published three RCTs (CLOSURE I, PC and RESPECT) revealed that PFO closure had a statistically significant effect on the composite of stroke, transient ischemic attack, and death in adjusted but not unadjusted analyses, as published in a previous pooled analysis of individual participant data.
What is New?
After the publication of the two new RCTs (CLOSE and REDUCE), the pooled results of our meta-analysis with the five RCTs confirmed that PFO closure reduced the rates of stroke, but also reinforced the problem of atrial fibrillation after the procedure, whose impact remains unknow. This meta-analysis revealed that the more effective the closure, the lower the risk of stroke.
What is Next?
The publication of the DEFENSE-PFO trial will add important data to those already available. Longer-term followup of completed trials will enhance our understanding of the effectiveness of PFO closure, but studies of various antithrombotic treatment regimens, including those in patients undergoing PFO closure, are necessary to address important knowledge gaps. We still need to know whether all of the devices are beneficial.
1. Kernan WN, Ovbiagele B, Black HR, Bravata DM, Chimowitz MI,Ezekowitz MD, et al. Guidelines for the prevention of stroke in patients withstroke and transient ischemic attack: a guideline for healthcare professionalsfrom the American Heart Association/American Stroke Association. Stroke.2014;45(7):2160-236.
2. Moher D, Liberati A, Tetzlaff J, Altman DG, for the PRISMA Group.Preferred reporting items for systematic reviews and meta-analyses: the PRISMAstatement. Ann Intern Med. 2009;151(4):264-9.
3. Higgins JP, Thompson SG, Deeks JJ, Altman DG. Measuringinconsistency in meta-analyses. BMJ. 2003;327(7414):557-60.
4. DerSimonian R, Kacker R. Random-effects model for meta-analysis ofclinical trials: an update. Contemp Clin Trials.2007;28(2):105-14.
5. Begg CB, Mazumdar M. Operating characteristics of a rank correlationtest for publication bias. Biometrics. 1994;50(4):1088-101.
6. Egger M, Davey Smith G, Schneider M, Minder C. Bias in meta-analysisdetected by a simple, graphical test. BMJ.1997;315(7109):629-34.
7. Mas JL, Derumeaux G, Guillon B, Massardier E, Hosseini H, MechtouffL, et al.; CLOSE investigators. Patent foramen ovale closure or anticoagulationvs. antiplatelets after stroke. N Engl J Med.2017;377(11):1011-21. [MedLine]
8. Søndergaard L, Kasner SE, Rhodes JF, Andersen G, Iversen HK,Nielsen- Kudsk JE, et al.; Gore REDUCE Clinical Study Investigators. Patentforamen ovale closure or antiplatelet therapy for cryptogenic stroke. N Engl JMed. 2017;377(11):1033-42. [MedLine]
9. Carroll JD, Saver JL, Thaler DE, Smalling RW, Berry S, MacDonald LA,et al.; RESPECT Investigators. Closure of patent foramen ovale versus medicaltherapy after cryptogenic stroke. N Engl J Med.2013;368(12):1092-100.
10. Meier B, Kalesan B, Mattle HP, Khattab AA, Hildick-Smith D, Dudek D,et al.; PC Trial Investigators. Percutaneous closure of patent foramen ovale incryptogenic embolism. N Engl J Med. 2013;368(12):1083-91.
11. Furlan AJ, Reisman M, Massaro J, Mauri L, Adams H, Albers GW, etal.; CLOSURE I Investigators. Closure or medical therapy for cryptogenic strokewith patent foramen ovale. N Engl J Med. 2012;366(11):991-9.
12. Di Tullio MR, Sacco RL, Sciacca RR, Jin Z, Homma S. Patent foramenovale and the risk of ischemic stroke in a multiethnic population. J Am CollCardiol. 2007;49(7):797-802. [MedLine]
13. Homma S, Sacco RL, Di Tullio MR, Sciacca RR, Mohr JP; PFO inCryptogenic Stroke Study (PICSS) Investigators. Effect of medical treatment, instroke patients with patent foramen ovale: Patent Foramen Ovale in CryptogenicStroke Study. Circulation. 2002;105(22):2625-31.
14. Kent DM, Dahabreh IJ, Ruthazer R, Furlan AJ, Reisman M, Carroll JD,et al. Device closure of patent foramen ovale after stroke: pooled analysis ofcompleted randomized trials. J Am Coll Cardiol.2016;67(8):907-17.
15. Von Klemperer K, Kempny A, Pavitt CW, Janssen JC, Uebing A, Nicol E.Device closure for patent foramen ovale following cryptogenic stroke: a surveyof current practice in the UK. Open Heart 2017;4(2):e000636.
16. Song JK, Busan YS, Lee JH, Jeong SS, Shin ES. Device closure versusmedical therapy for cryptogenic stroke patients with high-risk patent foramenovale (DEFENSE-PFO). Available from:
No financial support.
No conflict of interest
Authors' roles & responsibilities
MPBOS Conception and design, analysis and interpretation of data, drafting of the manuscript, revising it critically for important intellectual content; final approval of the version to be published
LAPON Collection of data, drafting of the manuscript, revising it critically for important intellectual content; final approval of the version to be published
GCSN Collection of data, drafting of the manuscript, revising it critically for important intellectual content; final approval of the version to be published
EESV Collection of data, drafting of the manuscript, revising it critically for important intellectual content; final approval of the version to be published
GLM Collection of data, drafting of the manuscript, revising it critically for important intellectual content; final approval of the version to be published
KCR Collection of data, drafting of the manuscript, revising it critically for important intellectual content; final approval of the version to be published
GCN Collection of data, drafting of the manuscript, revising it critically for important intellectual content; final approval of the version to be published
AMM Revising it critically for important intellectual content; final approval of the version to be published
RFAL Revising it critically for important intellectual content; final approval of the version to be published
FPVS Revising it critically for important intellectual contente; final approval of the version to be published
RCL Revising it critically for important intellectual content; final approval of the version to be published