Exploring desirability of outcome ranking and generalised pairwise comparisons as endpoints in serious infections: Post hoc application with the MERINO trial

<h2>Abstract</h2><h3>Objectives</h3> Randomised controlled trials in serious infections typically use a single primary endpoint such as mortality. Ordinal or rank-based composite endpoints that evaluate mortality and other measures are a novel approach in clinical trials. Desirability of outcome ranking (DOOR) has been increasingly used in infectious disease trials and generalised pairwise comparisons (GPC) in cardiovascular trials. <h3>Methods</h3> We undertook a post hoc analysis of the MERINO trial, comparing piperacillin-tazobactam (PTZ) with meropenem in bloodstream infections due to ceftriaxone non-susceptible <i>Escherichia coli</i> and <i>Klebsiella</i> spp. to demonstrate and compare the DOOR and GPC approaches. The primary composite endpoint included mortality, microbiological relapse and secondary infection. A further analysis was performed by adding length of stay as a tiebreaker. We applied DOOR to evaluate the DOOR distribution, probability and conducted partial credit analyses. We applied GPC to estimate the win statistics: win ratio, win odds and net benefit. A secondary analysis with both approaches was conducted with trial participants with PTZ susceptible isolates only as recorded by a central laboratory. <h3>Results</h3> The DOOR probability was 44.3 % (95 % CI: 40.8 %, 47.9 %), indicating superiority of meropenem. Partial credit analyses demonstrated similar conclusions across different patient preferences. The win ratio, win odds and net benefit were 0.40 (95 % CI: 0.23, 0.70; <i>p</i> = 0.001), 0.80 (95 % CI: 0.69, 0.92; <i>p</i> = 0.002) and −11.3 % (95 % CI:18.7 %, −4.0 %; <i>p</i> = 0.003), also indicating superiority of meropenem. With the addition of length of stay, the DOOR probability was 43.7 % (95 % CI: 37.9 %, 49.6 %) and the win ratio, win odds and net benefit were 0.77 (95 % CI: 0.60, 0.99; <i>p</i> = 0.04), 0.78 (95 % CI: 0.62, 0.99; <i>p</i> = 0.04), −12.1 % (95 % CI:23.8 %, −0.3 %; <i>p</i> = 0.04). Superiority of meropenem was not able to be demonstrated in the secondary analysis where only patients with PTZ susceptible isolates were assessed. <h3>Conclusions</h3> Though the functionality and interpretation of DOOR and GPC differ, both approaches can enhance the overall understanding of treatment effect in survivors beyond a single endpoint such as mortality.