Comment on the article “Timing of oral anticoagulants initiation for atrial fibrillation after acute ischemic stroke: A systematic review and meta-analysis”

Dear editor

Anticoagulation therapy with oral anticoagulants (OACs) reduces the risk of ischemic stroke and systemic embolism in atrial fibrillation patients. ¹ However, it remains unclear whether the timing of OAC initiation affects the risk of stroke recurrence and hemorrhage following acute ischemic stroke. The principal finding of the recent article by Palaiodimou, ² namely “Timing of oral anticoagulants initiation for atrial fibrillation after acute ischemic stroke: A systematic review and meta-analysis,” reflects a beneficial role of early OAC for non-valvular atrial fibrillation patients. However, I would like to provide some critical insights into specific statements made in this paper.

First, it is imperative to emphasize that mixing randomized controlled trials (RCTs) and observational studies in meta-analysis is inappropriate due to their inherent differences in study design, which may be one of the reasons of high heterogeneity across paper. This methodological drawback may also contribute to the high heterogeneity observed in the present study. Unfortunately, Palaiodimou et al.’s conclusions are derived from these mixed results. Furthermore, as presented in Figures 2 and 3, the pooled results of RCTs and observational studies are diametrically opposite. In other words, results of observational studies do not support the author’s conclusions. In this setting, which result should we trust?

Second, the authors claim that “The random-effects model (DerSimonian and Laird) was used to calculate the pooled estimates.” However, I believe this is worth reconsidering. In general, the selection of the effect model is undertaken with deliberate care and caution, given its profound impact on the results. In this study, all data was pooled using a random-effects model, resulting in wider confidence intervals and thereby diminishing the clinical confidence in the conclusions. For this issue, Tufanaru provided a decision-making tree to help choice effect model. ³ According to the Tufanaru method, for instance, when analyzed using the fixed-effect model (Mantel-Haenszel method), the result for the observational subgroup in Figure 4(a) yields a risk ratio (RR) of 0.96 (95% CI 0.63–1.46), which is more precise compared to the estimate reported by Palaiodimou et al. (RR = 0.95, 95% CI 0.50–1.81).

Third, the definitions of “Early” and “Late” vary considerably across the included studies. Although the authors conducted subgroup analyses on this issue, further exploration is still needed to determine the optimal timing for initiating oral anticoagulation. Furthermore, the subgroup based on type of OAC also needs to be conducted, as previous study demonstrating a difference between vitamin K antagonists and non-vitamin K antagonist oral anticoagulant. ⁴

In conclusion, the paper undoubtedly provided important evidence regarding OAC in the stroke secondary prevention and I am in concurrence with the core tenets of Palaiodimou et al. I sincerely look forward to the authors’ response to these points, as their insights could help refine our understanding and guide future clinical practice.