Optimal two-stage Phase II Designs with Time to Event Endpoints
Department of Statistics
University of Wisconsin, Madison
Friday, November 17, 2006, 12:00-1:00
A phase II study of cancer is undertaken to initially assess the activity of a new treatment. Because of the mechanism of actions of the new targeted therapies, it may not be appropriate to use traditional tumor shrinkage to evaluate the activity of these agent. Instead, the primary endpoint is a survival time to event variable. Due to ethical issues, phase II trials are usually designed with an interim analysis so they can be stopped if early results are disappointing (ineffective). However, with survival as the primary endpoint, interim analyses are challenging due to inconvenient suspension of accrual while patients are followed.
There have been some designs using Kaplan-Meier or Nelson-Aalen estimates of the survival probability assuming constant accrual rate (Lin et al, 1996; Case and Morgan, 2003). We propose an extended design by incorporating a flexible study-entry time distribution. We present a detailed optimization algorithm to find the optimal design that minimizes either the expected duration of accrual or the expected total study length
under H0. We also implement the extended design with an R package called OptimPhase2.
Return to seminar list