Background: The prospective, randomized, open, blinded endpoint evaluation (PROBE) design has been proposed as a valid alternative to the double-blind (DB) design for trials comparing new oral anticoagulants (NOAs) with INR-adjusted vitamin K antagonists in patients with non-valvular atrial fibrillation (NVAF).
Objectives: To determine whether the observed treatment effects of NOAs in patients with NVAF differ between PROBE/open-label trials and DB trials.
Methods: All phase II or III trials were eligible. The main efficacy and safety outcomes were stroke/systemic embolism (SSE) and major bleeding, respectively. Other outcomes included ischemic SSE, hemorrhagic stroke, intracranial and extracranial bleeding, myocardial infarction, and all-cause and cardiovascular mortality. Interaction (Cochran's chi-squared test) between PROBE and DB designs was tested.
Results: Thirteen studies (61 620 patients) were included. For SSE, a greater treatment effect of NOAs vs. INR-adjusted warfarin was observed in PROBE trials (RR 0.76, CI 0.65-0.89) compared with DB trials (RR 0.88, CI 0.78-0.98), but the interaction test was non-significant (P = 0.16). A significant 67% enhancement of treatment effect was found with PROBE/open-label trials compared with DB trials (interaction test, P = 0.05) for hemorrhagic stroke. No other interaction was significant. A non-significant interaction (P = 0.07) between oral direct thrombin inhibitors (RR 0.33; 0.22-0.51) and factor Xa inhibitors (RR 0.54; 0.40-0.72) was seen. No heterogeneity was found for any outcome.
Conclusions: Our meta-analysis showed no significant interaction of study design for the main efficacy and safety outcomes. However, the non-significantly exaggerated reduction in SSE suggests interdependence of treatment effect and PROBE design, especially for hemorrhagic stroke.
Keywords: anticoagulants; atrial fibrillation; bias; meta-analysis; stroke.
© 2013 International Society on Thrombosis and Haemostasis.