Date of Award

1-1-2017

Embargo Period

12-1-2020

Document Type

Dissertation - MUSC Only

Degree Name

Doctor of Philosophy (PhD)

Department

Public Health Sciences

College

College of Graduate Studies

First Advisor

Valerie Durkalski-Mauldin

Second Advisor

Wenle Zhao

Third Advisor

Caitlyn Ellerbe

Fourth Advisor

Ying Yuan

Fifth Advisor

Andrew Lawson

Sixth Advisor

William Meurer

Abstract

Bayesian response adaptive randomization (BRAR) has been utilized in early phase trials with the motivation of improving trial efficiency and/or subject ethics. The performance of BRAR in large phase III trials remains unclear. Different response adaptive allocation algorithms have been used in practice, with little information available on their rationale and performance. It is believed that the timing and frequency of treatment allocation updates, and the interim analysis stopping boundaries can affect the trial operating characteristics. However, the impact of these BRAR implementation parameters have not been discussed in the literature. Additionally, clinical trial design and analysis often assume study population homogeneity. The impact of the possible time-trend in patient baseline profile and response to treatment during the study period on the performance of the response adaptive randomization and the trial operating characteristics are rarely studied. This dissertation research accomplishes three objectives: 1) to establish a quantitative evaluation framework for BRAR in both two-arm and three-arm trial scenarios with a binary endpoint; 2) to explore the mechanism of time-trend impact on the performance of BRAR in a two-arm trial, and to develop a method for handling time-trend in the randomization; and 3) to redesign a previously completed phase III randomized controlled trial using BRAR in order to examine how BRAR performs in the presence of time-trend.

Rights

All rights reserved. Copyright is held by the author.

Download

Share

COinS