Date of Award

2020

Embargo Period

8-1-2024

Document Type

Dissertation - MUSC Only

Degree Name

Doctor of Philosophy (PhD)

Department

Public Health Sciences

College

College of Graduate Studies

First Advisor

Viswanathan Ramakrishnan

Second Advisor

Wenle Zhao

Third Advisor

Paul J. Nietert

Fourth Advisor

Jody D. Ciolino

Fifth Advisor

Michael D. Hill

Abstract

When there is a large number of baseline covariates whose imbalance needs to be controlled in sequential randomized controlled trials, minimization is most commonly used for randomizing treatment assignments. The lack of allocation randomness associated with the minimization method has been the source of controversy. The minimal sufficient balance (MSB) method is an alternative to minimization. It prevents serious imbalance from a large number of covariates while maintaining high levels of allocation randomness. However, a formal comparison between covariate-adaptive methods of randomization has not yet been studied. Using a re-randomization of the rt-PA clinical trial dataset with 1:1 equal allocation, minimization and MSB methods are compared with respect to allocation randomness, effectiveness at balancing covariates across treatment arms, and preservation of the nominal type I error rate. Using a simulated dataset, power and bias in the estimation of treatment effect are studied for completely randomized design, stratified permuted blocks, minimization, and MSB. A novel randomization method, known as allocation ratio preserving Minimal Sufficient Balance (ARP MSB) is presented as an alternative to allocation ratio preserving biased coin minimization (ARP BCM). Using a re-randomization of the rt-PA clinical trial dataset, ARP BCM and ARP MSB are compared with respect to the allocation randomness, effectiveness at balancing covariates across treatment arms, and preservation of the nominal type I error rate in unequal allocation clinical trials. MSB and ARP MSB methods proved to have equal or superior effectiveness at controlling imbalance on a combination of continuous and categorical variables, as well as a far greater proportion of completely random treatment assignments compared to the minimization and ARP BCM methods. MSB, ARP MSB, minimization, and ARP BCM all proved to have similar properties with respect to type I error rate preservation, power, and bias in measuring treatment effects. MSB and ARP MSB, while not presented as optimal methods for controlling covariate imbalances in sequential clinical trials, provide an alternative to the minimization and ARP BCM methods. The arguments in this dissertation should be considered by those who wish to use minimization or ARP BCM for subject allocation in clinical trials.

Rights

All rights reserved. Copyright is held by the author.

Download

Share

COinS