Statistical Inference Following Covariate-Adaptive Randomization: Recent Advances

• Activity Number: 546
• Type: Topic Contributed
• Date/Time: Wednesday, August 6, 2014 : 2:00 PM to 3:50 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Wei Ma*+ and Feifang Hu
• Companies: University of Virginia and George Washington University
• Keywords: Covariate-adaptive randomization ; Minimization ; Conservative tests ; Testing hypotheses ; Stratification

Abstract

Covariate-adaptive randomization (CAR) has been increasingly implemented in clinical trials to balance important covariates. However, the properties of statistical inference following CAR are not fully understood. In the literature, most studies are based on simulations. In this paper, we summarize some recent advances on theoretical properties of hypothesis testing under CAR, proposed by Ma et al. (2014). We will first give a general review of basic concepts of CAR and motivations to study inference properties following CAR. We next summary the main results in the paper, including describing the framework and assumptions and giving the theoretical properties. In the linear model framework, asymptotical distributions of test statistics are given for testing treatment effects and significance of covariates under null and alternative hypotheses. In particular, it is shown that under a large class of CAR designs, (i) the hypothesis testing to compare treatment effects is usually conservative in terms of small Type I error; (ii) the hypothesis testing to compare treatment effects is usually more powerful than complete randomization; and (iii) the hypothesis testing for significance of covariates is still valid. We close with a discussion of related work and possible future directions.