All Times EDT
Time-to-first-event analysis of a composite endpoint is often used for studies involving multiple outcomes. Each component outcome is treated equally, even though they may be of different clinical importance. Win Ratio, net benefit, and Win Odds (WO) have been used as alternative summaries that can handle different types of outcomes and allow for a hierarchical ordering in component outcomes, and thus have drawn much attention in the pharmaceutical industry, academia, and regulatory agencies (Pocock et al., 2012). Most existing work has focused on the nonparametric estimation of these measures under the two-sample setting. In this talk, we propose a proportional win odds regression model to evaluate the treatment effect on multiple outcomes while controlling for other risk factors. The model is easily interpretable as a standard logistic regression model. However, the proposed win odds regression is much more advanced in that multiple numbers and types of outcomes are modeled together and the estimating equation is constructed based on all possible and potentially dependent pairings of a treated subject and a control subject. We also carefully distinguish the ties caused by binary outcomes, which imply treatment equivalency and motivate the use of WO, and the ties due to censored observations. The latter should not affect WO and can be handled by using inverse probability of censoring weighting. The asymptotic properties of the estimated regression coefficients are established through the U-statistic theory coupled with the empirical process theory. We demonstrate the finite sample performance through numerical studies.