BackgroundWin ratio (WR) is a newer analytic approach for trials with composite end points that accounts for the relative importance of individual components. Our objective was to compare the results of the Cardiovascular Outcomes for People Using Anticoagulation Strategies (COMPASS) trial analyzed using WR with those obtained using conventional statistical approaches. MethodsWe used an unmatched WR analysis for first and total (first plus recurrent) events to examine effects of rivaroxaban with aspirin and rivaroxaban alone vs aspirin alone on primary efficacy (cardiovascular death, stroke, myocardial infarction), safety (modified International Society on Thrombosis and Haemostasis major bleeding), and net clinical benefit (primary efficacy plus fatal or critical organ bleeding) end points. We compared the WR results with those obtained using the Cox proportional hazards regression model for first events and Anderson-Gill method for total events. We calculated the win difference to estimate absolute treatment effects. ResultsThe WR approach produced results consistent with those obtained using conventional statistical methods for the primary composite end point (first event: WR, 1.32 [95% confidence interval (CI), 1.14-1.52]; 1/Cox hazard ratio, 1.32 [95% CI, 1.16-1.52]; total [first plus recurrent] events: WR, 1.32 [95% CI, 1.14-1.52]; 1/Anderson-Gill hazard ratio, 1.32 [95% CI, 1.16-1.54]) as well as for main safety and net clinical benefit end points. The absolute benefits of the combination of rivaroxaban and aspirin compared with aspirin alone calculated using the win difference were greatest in those with multiple high-risk features. ConclusionsReanalysis of the COMPASS trial results using WR produced results that were consistent with those obtained using conventional statistical approaches. Clinical Trial RegistrationNCT01776424
Read full abstract