Bayesian Adaptive Randomization: A Mixture of Response-Based and Covariate-Balanced Approach

• Activity Number: 615
• Type: Contributed
• Date/Time: Wednesday, August 12, 2015 : 2:00 PM to 3:50 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Shanhong Guan*
• Keywords: Adaptive randomization ; Covariate ; Response-adaptive

Abstract

Traditional randomization scheme uses fixed allocation probabilities in advance. Adaptive randomization approaches can alter the allocation of subjects to treatment groups based on accruing trial information. Among these methods, there are two commonly used types: covariate and response adaptive randomization. Covariate adaptive (CA) randomization provides a mechanism to actively control the imbalance of covariates/prognostic factors by allowing the allocation probabilities to change as a function of the current distribution of covariates. Response-adaptive randomization (RA) approach uses accumulating information from the available subjects' responses to skew the treatment assignment probabilities and assign more patients to the more efficacious and/or safer treatment arms. In this research, we investigate a randomization procedure that naturally combines two methods: the procedure first actively balances the covariates across treatment arms by optimizing a prognostic score based on each covariate's imbalance weight (proportional to covariate's strength of association with response), and then further integrate this into the response-adaptive randomization (RA) scheme.