Win ratio analysis for the 1-year composite endpoint: a substudy of the LANDMARK trial

Abstract

Abstract Background Conventional time-to-first event analyses for composite endpoints consider only the first event, whereas win-ratio analyses account for event severity and recurrence. Methods We performed a win-ratio analysis of the 1-year composite endpoint in the LANDMARK trial, a multicentre, non-inferiority study that randomized patients with severe native aortic valve stenosis to receive either the Myval transcatheter heart valve (THV) series (n=384) or contemporary THVs (n=384), consisting of the Sapien and Evolut series (n=192 each). The 1-year composite endpoint included the following components, ranked in hierarchical order: all-cause death, disabling stroke, non-disabling stroke, and procedure- or valve-related hospitalization. For death and stroke events, wins or losses were determined according to the time-to-first-event principle. For procedure- or valve-related hospitalizations, the total number of events was compared. Win-ratio analysis was also conducted for an extended composite endpoint that also included quality-of-life outcomes. Results Among 147,456 (384x384) unmatched patient pairs, the win ratio for the primary composite endpoint was 1.02 (95% CI: 0.68–1.51; p=0.94), with 17,870 (12.1%) wins for the Myval group and 17,599 (11.9%) for the contemporary group. For the extended composite endpoint, the win ratio was 1.13 (95% CI: 0.82–1.55; p=0.45), with 27,007 (18.3%) wins for the Myval group and 23,920 (16.2%) for the contemporary group. Comparisons between the Myval and Sapien groups, and between the Myval and Evolut groups, were also not statistically significant. Conclusions Win-ratio analysis did not demonstrate a significant difference between Myval and contemporary THVs, consistent with the primary time-to-first event analysis of the LANDMARK trial.