Author response: “Timing of oral anticoagulants initiation for atrial fibrillation after acute ischemic stroke: a systematic review and meta-analysis”

We appreciate Dr. Shu’s thoughtful comments on our article “Timing of oral anticoagulants initiation for atrial fibrillation after acute ischemic stroke: a systematic review and meta-analysis.” Our study investigated the effects of early versus later oral anticoagulant (OAC) initiation in atrial fibrillation (AF) patients after acute ischemic stroke (AIS) based on data up to January 2024. ¹ Specifically, our meta-analysis included two randomized controlled trials (RCTs) and seven observational studies, showing that early OAC initiation had better efficacy and a similar safety profile compared to later initiation.

We acknowledge the concern regarding the inclusion of both RCTs and observational studies. However, combining these study types can provide a broader perspective: RCTs offer high internal validity, while observational studies reflect real-world practice. Importantly, we performed subgroup analyses stratified by study design and found no significant differences between RCTs and observational studies. Moreover, when restricting the analysis to RCTs alone, the results remained consistent with the overall findings. While the pooled estimates from RCTs and observational studies differed slightly in magnitude, they followed a similar direction, confirming rather than contradicting each other. Thus, in response to the question of “which result should we trust?,” our findings indicate that both sources contribute valuable insights supporting the same conclusion.

Regarding the type of OAC used, we acknowledge the importance of distinguishing between vitamin K antagonists (VKAs) and non-vitamin K antagonist oral anticoagulants (NOACs). A previous study by our group demonstrated a significant reduction in ischemic stroke recurrence in patients treated with NOACs compared to VKAs, regardless of the timing of treatment initiation. ² However, due to data limitations, we were unable to conduct a comprehensive subgroup analysis based on OAC type in the present study.

The choice of random-effects modeling in our study was primarily driven by the inclusion of observational studies, which inherently introduce variability due to differences in study design, population characteristics, and treatment protocols. The random-effects model accounts for this heterogeneity and produces more conservative estimates. However, even with this approach, our findings remained statistically significant for the composite outcome of ischemic and hemorrhagic events, as well as for mortality and ischemic stroke recurrence, all favoring early OAC initiation. While it is true that the random-effects model yields wider confidence intervals, it provides a more appropriate and realistic estimate given the expected variability across studies.

Since the publication of our study, two additional RCTs – the START and OPTIMAS trials – have been released,^3,4 increasing the total number of RCTs addressing this question to four. Taking this opportunity, and in response to Dr. Shu’s comments, we conducted an updated study-level meta-analysis incorporating these four RCTs, employing a fixed-effect model appropriate for their controlled design. Results showed no significant difference between early and later NOAC initiation for the composite outcome (RR: 0.82; 95%CI: 0.65–1.03; p-value = 0.08; I² = 3%), ischemic stroke recurrence (RR: 0.82; 95%CI: 0.61–1.11; p-value = 0.20; I² = 26%), intracranial hemorrhage (RR: 0.93; 95%CI: 0.44–1.96; p-value = 0.84; I² = 0%), major bleeding (RR: 0.79; 95%CI: 0.48–1.29; p-value = 0.34; I² = 23%), and all-cause mortality at follow-up (RR: 0.97; 95%CI: 0.81–1.18; p-value = 0.79; I² = 0%).

A key limitation of these RCTs is the variability in defining early versus late NOAC initiation, making it difficult to determine a uniform initiation time applicable in clinical practice. To address this, an individual patient data (IPD) meta-analysis has been conducted by the CATALYST collaboration, harmonizing definitions across a large dataset. ⁵ In this analysis, timing and outcome definitions were harmonized across the majority of the included participants. Early NOAC initiation was defined as treatment within the first four days post-stroke (n = 2312), while later initiation was defined as five or more days post-stroke (n = 2375), and the follow-up duration was set at 30 days. Based on the presented results, early NOAC initiation was associated with lower odds of all-stroke and ischemic stroke recurrence compared to later-NOAC initiation, while symptomatic intracranial hemorrhage was lower in the early-treated group.

Importantly, IPD meta-analysis offers significant advantages over aggregate data meta-analysis. It enables standardization of participant inclusion and outcome definitions, application of consistent statistical methods, more comprehensive adjustment for baseline characteristics, improved handling of missing data, and more detailed bias assessment. Additionally, it allows investigation of treatment effect modification by individual patient characteristics. These strengths provide a more refined understanding of treatment timing and effects.

The CATALYST collaboration is expected to provide further individualized insights, addressing additional clinical factors such as baseline stroke severity, receipt of acute reperfusion therapies, hemorrhagic transformation, and prior anticoagulation use. These findings will help refine treatment guidelines and improve clinical decision-making.