A Hybrid Geometric Select-and-Test Design Based on Treatment Failure Time and Toxicity

• Activity Number: 378
• Type: Invited
• Date/Time: Tuesday, August 3, 2010 : 2:00 PM to 3:50 PM
• Sponsor: International Indian Statistical Association
• Author(s): Peter F. Thall*+
• Companies: MD Anderson Cancer Center
• Keywords: Bayesian design ; Brain tumors ; Clinical trial ; Phase II-III design ; Selection ; Two-Stage Design

Abstract

Comparing several experimental treatments to a standard is a common problem in medical research. Many multi-stage designs have been proposed that select promising new treatments and compare them to a standard while controlling overall error rates. We address settings where the goal is to increase mean treatment failure time and control toxicity while accounting for patient covariates. We propose a hybrid two-stage phase II-III design based on two-dimensional treatment effects obtained from Bayesian regression models for failure time and the probability of toxicity. Decisions at each stage of the trial are based on the posterior odds of null and alternative two-dimensional covariate-adjusted treatment effect parameter sets. Design parameters are chosen to minimize expected sample size subject to overall error rate constraints. The method is illustrated by a pediatric brain tumor trial.