Defense Date

2020

Document Type

Dissertation

Degree Name

Doctor of Philosophy

Department

Biostatistics

First Advisor

Roy T. Sabo, PhD

Second Advisor

Le Kang, PhD

Third Advisor

Nitai Mukhopadhyay, PhD

Fourth Advisor

Edward Boone, PhD

Fifth Advisor

Amir Toor, MD

Abstract

Response-adaptive (RA) allocation designs can be implemented in clinical trials to skew the allocation of incoming subjects toward the better performing treatment group based on the previously accrued subjects' responses. These designs alleviate potential ethical concerns of equally allocating subjects in a trial when one treatment arm is inferior. The RA design can be generalized to include covariate information in the covariate-adjusted response-adaptive (CARA) design, which aims to maximize treatment successes conditional on a set of patient characteristics. While RA and CARA designs can improve the treatment of patients, they have unstable estimators and increased variability in early stages of clinical trials when sample sizes are small. In order to avoid small sample irregularities, a lead-in period can be considered where adaptation is delayed until sufficient subjects are accrued, though this inclusion will reduce the realized benefits from adaptation. Alternatively, Bayesian methods can be implemented to overcome the small-sample irregularities, in such a way as to constrain the weights from changing too quickly and too early in the trial. Incorporating prior information, such as results from previous studies or known clinical opinion, increases the effective sample size by adding ``prior'' knowledge to the observed data. Increasing the effective sample size reduces small sample variability and thus allows for more stable estimators in the early stages of the trial. As patients are enrolled in the trial and response data is accrued, the allocation weight can become increasingly informed by the data and less by the prior information. This transition of the locus of information can be accelerated through the use of decreasingly informative prior (DIP) elicitation, which is a probability distribution that is a function of the sample size that accounts for decreasing amounts of the posterior likelihood as subjects accrue. These priors can be applied to RA and CARA allocation in a manner such that allocation weights are constrained early in the trial, but where adaptation increases as more subjects are accrued. Simulated clinical trials comparing the behavior of these approaches with traditional frequentist and Bayesian RA and CARA as well as balanced designs show that the natural lead-in approaches support ethical adaptation with less bias and greater power.

Rights

© The Author

Is Part Of

VCU University Archives

Is Part Of

VCU Theses and Dissertations

Date of Submission

3-5-2020

Available for download on Tuesday, March 04, 2025

Share

COinS