Testing Hypotheses of Continuous-Covariate-Adaptive Randomized Clinical Trials

• Activity Number: 42
• Type: Contributed
• Date/Time: Sunday, August 9, 2015 : 2:00 PM to 3:50 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Xiaoming Li* and Jianhui Zhou and Feifang Hu
• Companies: University of Virginia and University of Virginia and The George Washington University
• Keywords: Covariate-adaptive design ; Continuous covariate ; Linear model ; Conservative tests ; Type I error ; Power

Abstract

Covariate-adaptive design is widely used in clinical trials. Restricted-randomization procedures based on discrete covariate and its asymptotic properties are well understood and documented in the literature. However, studies using continuous covariate-adaptive designs are entirely based on simulations without a rigorous theoretical derivation. Thus conventional hypothesis testing for clinical trial using continuous covariate-adaptive designs is not well understood. Herein, our paper addresses understanding the testing hypotheses under continuous covariate-adaptive designs. We investigate the theoretical properties of the testing hypotheses by proving the asymptotic distributions of the test statistic under null and alternative hypotheses for continuous covariate-adaptive designs satisfying certain conditions. We studied a class of continuous covariate-adaptive randomization: the p-value based method, the Su's percentile method, the Empirical cumulative-density method, the Kullback-Leibler divergence method, and the Kernel-density method etc. The type I error and power of the hypothesis testing under those adaptive designs are compared to those under complete randomization method.