Assessment of clinically relevant bleeding as a surrogate outcome for major bleeding: validation by meta‐analysis of randomized controlled trials

Essentials Surrogacy of clinically relevant bleeding (CRB) for major bleeding has never been validated. Our meta-analysis evaluated CRB surrogacy in trials of new versus traditional anticoagulants. Surrogacy was not validated in orthopedic surgery, venous thromboembolism or atrial fibrillation The difficulty in demonstrating the surrogacy may reflect a lack of homogeneity in its definition SUMMARY: Background Clinically relevant bleeding (CRB), comprising major bleeding and clinically relevant non-major bleeding, has been used as a surrogate for major bleeding in most anticoagulant trials. The validity of this surrogate to estimate trade-off between thrombotic and bleeding events in clinical trials was never assessed. Methods We systematically reviewed randomized phase III trials comparing new anticoagulants with the standard of care for venous thromboembolism prevention following major orthopedic surgery, venous thromboembolism (VTE) treatment, or stroke and systemic embolism prevention in atrial fibrillation (AF), and reporting both major bleeding and CRB rates. The validity of CRB as a surrogate for major bleeding was assessed according to the strength of the association between the relative risks of major bleeding and CRB, measured by the use of R²_trial and its 95% confidence interval (CI). Results In the postoperative prophylactic setting (13 studies), major bleeding and CRB rates were 1.12% and 3.56%, respectively, and R²_trial was 0.69 (95% CI 0.34-0.93). For acute VTE studies (n = 12), major bleeding and CRB rates were 1.87% and 9.07%; the corresponding R²_trial values were 0.28 (95% CI 0.01-0.80) and 0.68 (95% CI 0.09-1.00) when only double-blind studies were considered (n = 7). For AF studies (n = 7; 22 strata), major bleeding and CRB rates were 4.82% and 15.3%, and R²_trial was 0.59 (95% CI 0.15-0.82). Conclusion Despite an apparent correlation between CRB and major bleeding in major orthopedic surgery, AF, and double-blind acute VTE studies, the wide CIs suggest that CRB might not be an acceptable surrogate outcome in any of these settings.