OBJECTIVES: The win ratio approach aims to improve analyses of composite endpoints in clinical trials by accounting for the relative importance of components of the composite endpoint and enhancing interpretability of the treatment effect measure. We propose a new application as an alternative to hierarchical testing of endpoints to improve statistical power in small trials (e.g., in rare diseases). This is illustrated with a reanalysis of the COMET trial (NCT02782741) comparing avalglucosidase alfa (AVA) and alglucosidase alfa (ALGLU) in patients with LOPD, in which statistical noninferiority was achieved, but superiority was not met.

METHODS: Analyses focused on forced vital capacity (FVC) %-predicted and six-minute walk test (6MWT) - the first and second endpoints in the trial’s hierarchical testing sequence. All possible patient pairs from both study arms were compared sequentially on changes at week 49 in FVC %-predicted and 6MWT using thresholds representing clinically meaningful improvement or decline (FVC ±4%, 6MWT ±39 m). Each comparison assessed whether the outcome with AVA was better than (win), worse than (loss), or equivalent to (tie) the outcome with ALGLU. If tied on FVC, 6MWT was compared. The treatment effect is the win/loss ratio, with ties excluded. Analyses were repeated with the sequence of endpoints reversed.

RESULTS: In the 2499 possible pairs (51 receiving AVA × 49 receiving ALGLU), 1237 (49%) ended in a win, 523 (21%) were a loss for AVA, and 739 (30%) were tied when FVC was compared before 6MWT. The win ratio was 2.37 (95% confidence interval [CI], 1.30-4.29, p=0.005). When the order was reversed, the win ratio was 2.02 (95% CI, 1.13-3.62, p=0.018).

CONCLUSIONS: The win ratio approach provides a comprehensive assessment of treatment benefits while borrowing strength across the endpoints. Meaningful improvement in respiratory or motor outcomes was over twice as likely among patients receiving AVA versus ALGLU.