Date of Award

12-2019

Degree Name

Master of Science (MS)

Advisor(s)

HONGJIAN ZHU

Second Advisor

HAN CHEN

Abstract

Covariate adaptive randomization is an allocation procedure used in clinical trials that seeks to balance treatment groups. While this method has been shown to reduce bias due to imbalanced treatment groups, the effects of covariate adaptive randomization have not been studied under seamless Phase II/Phase III clinical trials that incorporate short-term endpoint information and treatment selection. Therefore, this analysis sought to determine whether these adaptive randomization methods can be applied to seamless Phase II/Phase III trials while preserving Type I error. In addition, this analysis sought to create R-packages that employ seamless Phase II/Phase III techniques to provide analysis tools for future research. Two covariate adaptive randomization schemes, Pocock and Simon’s procedure and stratified permuted block randomization were applied to simulated datasets to determine treatment groups. Seamless Phase II/Phase III clinical trials with two interim analyses were then simulated with 10,000 repetitions to determine overall Type I error rates. Only the most promising treatment group was selected to continue to the second interim analysis. Both randomization procedures were compared against trials that used simple randomization to allocate treatment groups. Ultimately, Type I error rates under the two adaptive techniques were not preserved. Pocock and Simon’s randomization saw an inflated level of Type I error, while stratified permuted block randomization resulted in lower levels of Type I error compared to simple randomization. In addition, power calculations revealed that both allocation methods resulted in lower levels of power compared to simple randomization.

Share

COinS