Rebuttal: “Closure of patent foramen ovale and prevention of recurrent thromboembolic events”

We have recently published the results of a metaanalysis of transcatheter closure of a patent foramen ovale (PFO) versus medical therapy for prevention of recurrent thromboembolic events in patients with cryptogenic cerebrovascular events [1]. Percutaneous closure of a PFO in these patients resulted in a 30– 38% reduction of recurrent cerebrovascular events when compared with medical therapy alone. This difference was statistically significant for the “as-treated” population (OR: 0.62; 95% CI, 0.41–0.94, P1⁄4 0.02) and showed a trend for the “intention-to-treat” population (OR: 0.70; 95% CI, 0.47–1.05, P1⁄4 0.08). As Zhang et al. pointed out in their letter to the editor, meta-analyses have limitations and should be interpreted carefully [2]. We agree with the authors of this letter, in that a transient ischemic attack (TIA) may be a softer end-point compared with a cerebrovascular accident (CVA), and that in clinical practice this diagnosis may sometimes be given to nonspecific neurological symptoms. However, a TIA was strictly defined, in the CLOSURE-1 and PC trials for example, as a clear motor deficit or speech abnormality localizable to a particular vascular territory [3,4], and therefore it has significant implications in the prognosis and management of a patient. In our manuscript, we had also acknowledged other limitations, including the different devices used and the heterogeneity in the inclusion criteria, study end-points, and allowed medical therapy amongst the three randomized controlled trials (RCT) included in our meta-analysis. At least two other meta-analyses have reached a similar conclusion with slightly different numerical results due to the different statistical methods used [5,6]. They additionally demonstrated that percutaneous PFO closure is better than medical therapy to prevent recurrent cerebrovascular events in patients with a substantial shunt, and in those who received the Amplatzer PFO occluder. None of the three available RCTs may had been sufficiently powered to demonstrate a significant difference between the two treatment groups, and therefore, we believe that the pooled data from these three RCTs is currently, the best available evidence.