All Times EDT
Keywords: Treatment Switching, Oncology, overall Survival, Causal Inference
A randomized control trial is the preferred design for evaluating the efficacy and safety of therapies, but it may not provide basis for valid inference on effect of investigational agent when control patients switch to investigational agent. Switching of control patients to investigational agent only occurs after progressive disease in some oncology trials, and requiring progressive disease to precede switching could usher in left truncation in assessing probability of switching, since patients do not enter risk set for switching until progressive disease. Treatment switching can be addressed using appropriate analytic methods—e.g., the g-methods or two-stage estimation—for evaluating effect of therapies on overall survival, but the analytic methods have different assumptions. However, when treatment effect on overall survival based on intent-to-treat is not significant or imperceptible, the results of the analyses adjusting for switching, if perceptible, may become a moot point. This session hopes to address the following questions: 1) What are approaches to reinforce results of overall survival following adjustment for treatment switching? 2) What are the implications of ignoring left truncation for trials in which progressive disease must precede treatment switching? 3) What are steps to ensure successful implementation and to avoid biased results when adjusting for treatment switching?