Date of Award

1-1-2016

Embargo Period

1-1-2019

Document Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Department

Public Health Sciences

College

College of Graduate Studies

First Advisor

Elizabeth Garrett-Mayer

Second Advisor

Jeffrey D. Blume

Third Advisor

Valerie Durkalski

Fourth Advisor

Elizabeth G. Hill

Fifth Advisor

John Wrangle

Abstract

Among anti-cancer agents selected for phase III trials, only about 5% will ever reach the oncology market. The low success rate of phase III oncology trials calls for more efficient phase II designs to better screen experimental agents. Several different approaches to improve phase II oncology trial designs have been proposed in the recent years, including the choice of alternative endpoints, and the push for novel and adaptive designs. Therefore, the objective of my dissertation research is to develop innovative phase II designs. Specific aim 1 introduces an empirical likelihood based group-sequential design for single-arm phase II trials with time-to-event endpoints. The primary objective of this undertaking is to provide an alternative to the frequentist and Bayesian designs without making unrealistic assumptions about the distribution of time-to-event data yet to be collected. The goal of specific aim 2 is to construct likelihood-based group-sequential designs for randomized studies with time-to-event endpoints. The operating characteristics of this proposed design are derived based on the asymptotic joint distribution of the log partial likelihood ratios, and tested by simulation studies. Lastly, tumor burden is assessed at regular intervals in most cancer clinical trials. However, the dynamics of tumor growth trajectories are often ignored in evaluation of treatment effects and a binary indicator of tumor shrinkage is commonly used as the primary efficacy endpoint. Hence, the goal of specific aim 3 is to construct a Bayesian mixture model for surrogate markers of tumor growth. This model assumes natural growth and drug induced decay ii are two latent processes underlying tumor growth among responders of treatment, whereas growth process is the sole driver of tumor growth among non-responders. The proposed model allows a more efficient and informative comparison of treatment effects based on the proportion of responders, time to nadir of tumor growth and the rate of tumor shrinkage in response to treatment.

Rights

All rights reserved. Copyright is held by the author.

Download

Share

COinS