Dan Gillen
Department of Biostatistics
University of Washington
Monday, February 17, 2003
3:30 PM
Moos 2-620
Minneapolis Campus
Abstract:
Group sequential methods have been widely described and implemented in the clinical trial setting when parametric and semi-parametric models are deemed appropriate. In these situations, the evaluation of group sequential stopping rules used for the planning and design of clinical trials remains relatively straightforward. However, when non-parametric methods are used, as is often the case for non-proportional hazards survival data, the evaluation of stopping rules is not a trivial task. Specifically, non-parametric test statistics do not necessarily correspond to a parameter of clinical interest, thus making it difficult to characterize alternatives at which operating characteristics are to be computed. Further, in the presence of non-proportional hazards, relative treatment effects are changing over time, making it difficult to compare stopping rules because interim testing inherently changes the support of observed survival times. Simulation-based methods for the estimation of frequently considered operating characteristics when pilot data suggest the presence of non-proportional hazards will be demonstrated for the case when testing is to be based upon a non-parametric weighted logrank statistic. In addition, shortcomings of the current methods for evaluating designs in the presence of non-proportional hazards, particularly when testing is based upon the use of weighted logrank statistics, will be discussed and modifications to the more commonly used logrank statistics will be suggested.