Multi-Stage Clinical Trial Design with Multidimensional Mixed Endpoints

• Activity Number: 686
• Type: Contributed
• Date/Time: Thursday, August 8, 2013 : 10:30 AM to 12:20 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Peng Huang*+ and Ming Tan
• Companies: Johns Hopkins University - Oncology Biostatistics and Georgetown University Medical Center
• Keywords: Rank-Based Comparison ; Adaptive Design ; Brownian motion ; group sequential test

Abstract

Multiple primary endpoints are increasingly used in clinical trials to test if a new agent is more efficacious than a control. The commonly used null hypothesis of identical distribution may not be suitable when nuisance effect exists. For example, a new agent for neurodegenerated disease does not possess the desired neuroprotective effect but it is likely to show different symptomatic effects as compared to control. We address this issue in the framework of the Behrens-Fisher problem and propose a rank based sequential test. We show that our stochastic process converges to a Gaussian process. In particular, an upper bound to the asymptotic approximation is provided for finite sample case. This renders existing multi-stage sequential procedures for univariate endpoint applicable to multidimensional endpoints of different data types. Our simulation shows that such design can preserve both type I error and statistical power well even for studies of moderate sample size. The methodology is illustrated using data from two randomized clinical trials.