Effect of ASA on the risk of cerebrovascular ischemic events in patients with PFO

Abstract Background Whether atrial septal aneurysm (ASA) increases the risk of cerebrovascular ischemic events in patients with patent foramen ovale (PFO) remains controversial. Objective We constructed a detailed meta‐analysis to assess the effect of ASA on risk of cerebrovascular ischemic events in patients with PFO. Methods Randomized controlled trials (RCTs) and observational studies (cohort studies and case‐control studies) that compared PFO‐ASA against PFO alone were included. Pooled odds ratios (OR) estimates and 95% CI were calculated using the fixed‐effect and random‐effect models. Results Four RCTs and twelve observational studies (five cohort studies and seven case‐control studies) contributed to the meta‐analysis. The pooled results of case‐control studies showed that ASA increased the risk of cerebrovascular ischemic events in patients with PFO (fixed‐effect model: OR = 3.69; 95% CI: 2.67–5.09; p < 0.01, random‐effect model: OR = 3.63; 95% CI: 2.51–5.24; p < 0.01). However, poole results from RCTs (fixed‐effect model: OR = 1.24; 95% CI: 0.78–1.95; p = 0.36, random‐effect model: OR = 1.27; 95% CI: 0.78–2.08; p = 0.34) and cohort studies (fixed‐effect model: OR = 1.35; 95% CI: 0.81–2.23; p = 0.25, random‐effect model: OR = 1.40; 95% CI: 0.84–2.33; p = 0.20) found no evidence. Overall analysis showed that ASA increased the risk of cerebrovascular ischemic events (fixed‐effect model: OR = 2.30; 95% CI: 1.84–2.87; p < 0.01, random‐effect model: OR = 2.11; 95% CI: 1.48–3.01; p < 0.01). The sensitivity analysis confirmed the stability of all results. Conclusions Although case‐control studies support ASA to increase the risk of cerebrovascular ischemic events in patients with PFO, RCTs and cohort studies challenged the credibility. Further prospective studies are needed to confirm the effect of ASA on patients with PFO.


Introduction
Patent foramen ovale (PFO) has been identified as one of the etiology of unexplained embolism events. 1,2 However, the incidence of PFO in the general population is 25%, indicating that some PFOs usually have no clinical effect. 3 Atrial septal aneurysm (ASA) is a pouch-like structure formed by the expansion of atrial septum. Some studies have shown that PFO patients with ASA or large right-to-left shunt have an increased risk of cerebrovascular ischemic events. 4,5 However, the evidence in some studies is insufficient. 6,7 In addition, there may be differences between different types of studies. Therefore, the effect of ASA on the risk of cerebrovascular ischemic events in patients with PFO is unclear. We conducted a comprehensive systematic review and meta-analysis, including randomized controlled trials studies (RCTs) and observational studies (cohort studies and case-control studies), to fully explore the effect of ASA and provide evidence-based basis for the prevention and treatment of cerebrovascular ischemic events.

Methods
This meta-analysis was performed in accordance with the preferred reporting items for systematic reviews and meta-analyses (PRISMA) statement 8 (File S1).

Study selection and inclusion
The studies enrolled in this meta-analysis included both RCTs and observational studies (cohort studies and casecontrol studies) that evaluated effect of ASA on the risk of cerebrovascular ischemic events (stroke or TIA) in patients with PFO. Comparisons between patients with PFO-ASA and those with PFO alone should be included in these studies. If all patients in one arm of a study had concomitant large shunt, the study was excluded. For RCTs and cohort studies，studies consistent with medical therapy (antiplatelet therapy and/or anticoagulant therapy) as the only preventive measure for recurrent cerebrovascular ischemic events were included in this meta-analysis. If studies were from multiple publications of the same population, only data from the latest publications were included. Reviews, case reports, cross-sectional studies, repeated papers, and conference abstracts were excluded in this meta-analysis. Two reviewers (L.X. and C.Z.) abstracted the data independently according to selection criteria. Any disagreement was resolved by discussion or referral to a third author (X.P.).

Quality assessment
The quality of studies was assessed according to Cochrane Handbook for RCTs 9 and Newcastle-Ottawa Scale for observational studies. 10 Discrepancies were resolved through negotiation.

Statistical analysis
We conducted meta-analyses of RCTs and observational studies (cohort studies and case-control studies) according to different study designs. For observational studies, we performed separate analyses for case-control studies and cohort studies. This was done to examine the consistency of results from different study designs. In addition, we pre-specified subgroup analyses for the three study designs based on neurological characteristics of exposed population and age range of patients. Odds ratios (OR) and 95% confidence intervals (CI) were calculated for each study and pooled values. If zero endpoint events occurred in 1 arm of a study, continuity correction of 1/2 was used. If zero endpoint events occurred in both arms of a study, we did not include them in the meta-analysis. According to heterogeneity detected, we used a fixedeffect model (Mantel-Haenszel method) or a randomeffect model (DerSimonian-Laird method) to calculate pooled OR estimates. In the absence of heterogeneity, the results of fixed effects and random effects models are similar. In the presence of heterogeneity, both models may be biased. Therefore, we conducted interactive tests of the model against the pooled results of RCTs and observational studies. Study heterogeneity was assessed with Cochran Q test and I 2 test. The I 2 value is between 0% and 100%, and larger values show increasing heterogeneity. If I 2 was >50%, heterogeneity was considered significant; If I 2 was <25%, heterogeneity was considered significant. To explore the stability of meta-analysis results, sensitivity analysis was performed by excluding each study one by one. Funnel plots were used to test the possibility of publication bias. For all tests, p < 0.05 was considered statistically significant. Review Manager 5.3 software (The Cochrane Collaboration, Copenhagen, Denmark) was used for the statistical analyses.

Description of included studies
Our initial search identified 628 records. After step-bystep selection and search, four RCTs 7,11-13 and 12 observational studies were finally included in the meta-analysis. The 1observational studies included five cohort studies 6,14-17 and seven case-control studies. 5,[18][19][20][21][22][23] The study selection process is shown in File S2. The main descriptions and patient characteristics of the included studies are shown in Table 1. The quality assessment each study is summarized in File S3.

Meta-analysis of RCTs
Four RCTs were eligible. A total of 425 patients with PFO-ASA and 920 patients with PFO alone were included in the analysis. The overall incidence of cerebrovascular ischemic events was 7.29% in the patients with PFO-ASA and 6.20% in the patients with PFO alone. The results of RCTs showed no evidence that patients with PFO-ASA had an increased risk of cerebrovascular ischemic events compared with patients with PFO alone (fixed-effect model: OR = 1.24; 95% CI: 0.78-1.95; p = 0.36) (Fig. 1). Heterogeneity among studies was not significant (I 2 = 9%; p = 0.35). The results of random-effect model after interaction test were also not statistically significant (OR = 1.27; 95% CI: 0.78-2.08; p = 0.34) ( Table 2). Sensitivity analysis showed that the pooled OR ranged from 1.05 (95% CI: 0.59-1.87) to 1.53 (95% CI: 0.88-2.67) after exclusion of the included studies one by one and the overall conclusions remained unchanged.

Meta-analysis of observational studies
Twelve observational studies were eligible, including five cohort studies and seven case-control studies. For the cohort study, a total of 204 patients with PFO-ASA and 583 patients with PFO alone were included in the analysis. For the case-control study, a total of 532 patients with PFO-ASA and 471 patients with PFO alone were included in the analysis. Meta-analysis of all observational studies showed that patients with PFO-ASA had an increased risk of cerebrovascular ischemic events compared with patients with PFO alone (29.35% vs. 12.24%; fixed-effect model: OR = 2.76; 95% CI: 2.12-3.59; p < 0.01, random-effect model: OR = 2.52; 95% CI: 1.74-3.65; p < 0.01). Heterogeneity among studies was moderate (I 2 = 43%; p = 0.05). Sensitivity analysis showed that the pooled OR ranged from 2.36 (95% CI: 1.77-3.14) to 2.91 (95% CI: 2.22-3.83) and the overall conclusions remained unchanged.
We performed separate analyses of cohort and casecontrol studies. In the analyses of cohort studies, we found that although ASA increased the risk of cerebrovascular ischemic events in patients with PFO, the difference was not statistically significant (12.25% vs. 10

Overall meta-analysis
We performed a combined meta-analysis of RCTs and observational studies. Combined analysis showed that ASA increased the risk of cerebrovascular ischemic events in patients with PFO (fixed-effect model: OR = 2.30; 95% CI: 1.84-2.87; p < 0.01, random-effect model: OR = 2.11; 95% CI: 1.48-3.01; p < 0.01). Heterogeneity among studies was significant (I 2 = 54%; p < 0.01). Sensitivity analysis showed that the pooled OR ranged from 1.92 (95% CI: 1.42-2.59) to 2.30 (95% CI: 1.65-3.19) after exclusion of the included studies one by one and the overall conclusions remained unchanged. The main results of metaanalysis of RCTs and observational studies are presented in Table 2.

Major subgroup analyses
We pre-specified subgroup analyses for the three study designs based on exposure population and age range of patients. In cohort studies, ASA increased the risk of cerebrovascular ischemic events in the subgroup   Table 3.

Publication bias
Funnel plots of cohort studies, case-control studies, and RCTs are shown in Figure 2. For cohort studies, the lower right corner of the funnel, which should include studies with small sample sizes, was missing. For case-control studies and RCTs, funnel plots showed no evidence of publication bias.

Discussion
Whether ASA can increase the risk of cerebrovascular ischemic events in patients with PFO remains controversial. [4][5][6][7] Although some studies 15,24 have shown that patients with PFO-ASA are the most suitable population for closure therapy, there is insufficient evidence that patients with PFO-ASA have a higher risk of recurrence than patients with PFO alone. In this meta-analysis, RCTs, cohort studies and case-control studies were analyzed based on different study designs to help us draw  more comprehensive conclusions. Furthermore, we conducted a model interaction test on each of the three pooled results to increase the reliability of results.
Our meta-analysis showed that the conclusions of RCTs and observational studies were different. The pooled results of observational studies showed that ASA increased the risk of cerebrovascular ischemic events in patients with PFO. However, the pooled results of RCTs showed no evidence that ASA increased the risk of cerebrovascular ischemic events in patients with PFO. Furthermore, the results of separate analysis of cohort and case-control studies were also inconsistent. For cohort studies, although patients with PFO-ASA had an increased risk of recurrence than patients with PFO alone, the difference was not statistically significant. However， the pooled results of case-control studies showed that ASA significantly increased the risk of cerebrovascular ischemic events in patients with PFO. In our opinion, on the one hand, selection bias and event definition heterogeneity of observational studies will have a certain deviation from the research conclusion. On the other hand, most case-control studies did not use multivariable adjustment methods to adjust for confounding factors, resulting in statistical bias. Third, in RCTs and cohort studies, the low recurrence rate of cerebrovascular ischemic events led to some studies unable to find the difference after follow-up. Prior to this study, Rigatelli et al. 25 conducted a meta-analysis of case-control studies and showed that ASA increased the risk of cerebrovascular ischemic events in patients with PFO (OR: 3.38, 95% CI: 2.72-5.51), which was consistent with the conclusion of our separate analysis of case-control studies. In addition, the current prospective studies on the effect of ASA are not comprehensive, and the recurrence risk of PFO-ASA is more shown in the subgroup analysis of some studies. To the best of our knowledge，this is the first time to provide a comprehensive analysis of the effect of ASA. In addition, in order to avoid the possible bias of the model in the presence of heterogeneity, we conducted an interactive test of the model, and the conclusion remained unchanged.
Subgroup analysis was performed for each of the three designs according to the factors that might influence the results. Age as influential factor plays an important role in the risk of cerebrovascular ischemic events in patients with PFO. Some studies 3,26 have shown that the risk of cerebrovascular ischemic events increases with age in patients with PFO. In addition, the neurological characteristics of the exposed population may influence the occurrence of cerebrovascular ischemic events. Therefore, we performed subgroup analysis on the age range included and neurological characteristics of exposed population. The results of subgroup analysis showed that there was significant heterogeneity between different age ranges in cohort and case-control studies. In subgroup analysis of cohort studies, ASA increased the risk of cerebrovascular ischemic events in patients with PFO ≤60 years old. In the subgroup analysis of case-control studies, ASA had a significant effect on patients with PFO ≤60 years old. This is similar to the results of Overell et al. 27 Overell et al. performed a meta-analysis and showed that PFO and ASA were significantly associated with ischemic stroke in patients younger than 55 years. 27 Heterogeneity within comparisons is eliminated by grouping into age bands. As we know, other causes and risk factors of cerebrovascular ischemic events are more likely to play a role in the elderly. Some studies suggest that large shunt may be a high risk factor for patients with PFO. 4,5 In the study of CLOSE 2017, 4 all patients with PFO alone were combined with large shunt, may increase the recurrence rate of cerebrovascular ischemic events, resulting in underestimation of the effect of ASA. Therefore, we deleted this study to avoid the potential heterogeneity of RCTs.
Our meta-analysis had several limitations. First, observational studies included in this meta-analysis were not adjusted for confounders. Therefore, the effect of ASA on cerebrovascular ischemic events in patients with PFO may be affected by other risk factors. Second, results of metaanalyses and pre-specified subgroup data can only be considered exploratory. Third, limitations of non-randomized studies include selection bias, heterogeneity of event definition, and differences in duration and intensity of follow-up to events. Fourth, heterogeneity in the definition of ASA was observed in several studies. In most studies, ASA was defined as an atrial septal excursion of ≥10 mm, while some studies used 11 or 15 mm as the cutoff point. The heterogeneity of these definitions cannot be ignored and may affect the results.

Conclusions
Although case-control studies showed that ASA increased the risk of cerebrovascular ischemic events in patients with PFO, RCTs and cohort studies found no evidence. Further prospective studies are needed to confirm the effect of ASA on patients with PFO. 25. Rigatelli G, Zuin M, Bilato C. Atrial septal aneurysm contribution to the risk of cryptogenic stroke in patients with patent foramen ovale: a brief updated systematic review and meta-analysis. Trends Cardiovasc Med 2022: S1050-1738 (22)

Supporting Information
Additional supporting information may be found online in the Supporting Information section at the end of the article.
File S1. PRISMA 2009 checklist. File S2. Flow diagram of study selection. Table S1. Risk of bias of included randomized trials.