Keywords: Subgroup Analysis, Survival Analysis, Benefit-Risk Analysis, Personalized Medicine.
Background: In a clinical trial that evaluates treatment effects on multiple time-to-event endpoints, subjects may experience clinical benefits, harms, or both during the follow-up period. Strategies for analyzing the time to multiple endpoints include the standard analysis based on the time to a composite endpoint or the first event, individual component analysis on the time to each event, and multivariate analysis based on categorized outcomes. These strategies are suitable and powerful in many, but not all, studies with multiple time-to-event endpoints. Methods: This study proposes a personalized benefit-harm analytical tree (PBHAT) approach that explicitly outlines and predicts the benefit-risk profile of a treatment at the patient level based on multiple time-to-event endpoints. The PBHAT approach sequentially classifies the subjects in a clinically meaningful order from the least desirable to the most desirable, e.g. from death to survival with adequate clinical response and without adverse events. Conclusions: The simulations show that the PBHAT approach may help outline and predict the benefit-risk profile of a treatment under many circumstances. There are also challenges waiting to be solved to make the approach suitable and powerful under more circumstances.