报告题目： Statistical Considerations When Applying Restricted Mean Survival Time in Immune-Oncology Trials
报告摘要：A challenge in analyzing survival endpoint in immune-oncology (IO) RCTs is that the proportional hazards assumption is often violated. Restricted mean survival time (RMST) has been proposed as an alternative to compare the survival benefit of immune-oncology treatment. In this paper, we discuss three topics that are relevant to the use of RMST in such trials. First, we discuss the selection of restricted time window for such analysis because the width of the window impacts the conclusion of trials. While the window should be determined based on clinical importance, it is also often limited due to the feasibility of study duration and the censoring types. Here we provide the conditions for which a selection of upper limit of window for which the estimation of RMST by weighted Kaplan-Meier estimate is consistent and statistical test is still validated. Second, we compared the statistical test for the alternative hypothesis that the IO treatment group has a stochastically greater survival time than the control arm within the restricted time window. This is a slightly more restricted alternative than the increasing in RMST but less restrictive to an alternative by the Cox model. Third, we discuss the use of a semi-parametric regression model for adjustment of covariates in RMST comparisons in trials. We performed simulation studies to compare their performance under the null and alternative hypotheses with finite sample size.