Power Analysis for Testing Treatment Effect with a Covariate and a Covariate-Dependent Stratification Factor

• Activity Number: 471
• Type: Contributed
• Date/Time: Wednesday, August 1, 2012 : 8:30 AM to 10:20 AM
• Sponsor: Biopharmaceutical Section
• Author(s): Chii-Dean Lin*+ and Tao Lin and Chien-Feng Chen and Larry Shen
• Companies: San Diego State University and Allergon and Otsuka Pharmaceutical and Amylin Pharmaceutical
• Keywords: Clinical trials ; RCBD ; ANCOVA ; Power

Abstract

In the analysis of clinical trial data, it is well accepted that including baseline measurements of the outcome variables may improve the accuracy of the estimate of treatment effect. Models using outcome variable or its change from baseline as a response variable and the baseline measurements as a covariate for testing treatment effect are usually employed. To ensure that similarly represented samples with respect to baseline values in both treatment groups are selected for a clinical trial, stratification based on a preselected cut point(s) of the baseline value is used to define strata and randomization is then implemented within each stratum. The ICH and CHMP guidance documents have stated that the stratification factor shall be included in the analysis models. In this paper, we evaluate the performance of 3 types of analysis models, an RCBD, an RCBD using change from baseline values, and an RCBD including the baseline measures as a covariate. Expected values and variances for the estimated mean treatment difference are derived. In addition, monte carlo simulations are performed to compare the power and Type I error rate for testing the treatment effect.