The win ratio approach to analyzing composite outcomes: An application to the EVOLVE trial

Background: Unlike conventional time-to-event analysis of composite endpoints in clinical trials, the "win ratio" method allows for flexibility in prioritizing their components. Here, we compare the EVOLVE trial findings using the win ratio with those from time-to-event analysis. Methods: Exposure: Randomization to cinacalcet or placebo. Outcome: The primary composite endpoint combining all-cause mortality and non-fatal myocardial infarction, hospitalization for unstable angina, heart failure, and peripheral vascular events. Analysis: In an unadjusted analysis, we paired each participant from the cinacalcet arm with every participant from the placebo arm within randomization strata. Pairs were classified as "winners" or "losers," according to which participant died first during the shared follow-up time, or experienced the next ranked event first. We ranked non-fatal events in two ways: 1) all ranked evenly; and 2) prioritized by their effect on health-related quality of life. The win ratio equaled the total winners divided by total losers. Further analyses were conducted where the win ratio was stratified by, or adjusted for, age. Results: The unadjusted win ratio for the primary composite endpoint was 1.09 (95% CI 0.97 to 1.21), a statistically non-significant result which supports the primary trial result unadjusted hazard ratio 0.93 (95% CI 0.85 to 1.02). Age-stratified analyses showed a nominally significant benefit of cinacalcet (win ratio 1.14, 95% CI 1.04 to 1.26). Ranking of non-fatal outcomes by their relative effects on quality of life did not materially alter the results. Conclusion: The win ratio method corroborated the findings of EVOLVE based on conventional time-to-event analysis. (C) 2016 Elsevier Inc. All rights reserved.