All ETDs from UAB

Advisory Committee Chair

Inmaculada B Aban

Advisory Committee Members

Gary R Cutter

Charity Morgan

John O'Quigley

John Rinker

Document Type

Dissertation

Date of Award

2015

Degree Name by School

Doctor of Philosophy (PhD) School of Public Health

Abstract

This dissertation considers the presence of patient heterogeneity in dose finding studies using the likelihood-based time to event continual reassessment method (TITE-CRM). First, assuming group membership is known and the groups share some information regarding their maximum tolerated dose (MTD), we extend the two-group two-parameter CRM to the TITE-CRM. We examine the performance of the two-group two-parameter TITE-CRM for various scenarios where the true MTD for each group is different. We also consider the situation where the groups are truly homogeneous but a heterogeneous model is employed. Our simulations show that the method has some loss in accuracy compared to the larger one-sample TITE-CRM, but has the ability to share information and recommend a MTD for both groups in a single trial. The implementation of the two-group two-parameter TITE-CRM requires statistical programs, at minimum, to solve the estimating equations for the maximum likelihood estimate. In order to aid other researchers to utilize designs that accommodate patient heterogeneity, we developed a program to design and conduct a trial using this method. We provide documentation needed to use the program, an example of how the simulation program can be used for sample size guidance and the output resulting from the trial implementation. Finally, we anticipate that in practice assuming heterogeneity from the onset may be weak. Instead, one may wish to start with assuming the groups are homogeneous unless the data collected suggest the presence of heterogeneity. In this situation, the researcher is able to cautiously explore the possibility of heterogeneity. We describe and compare approaches based on models that accommodate patient heterogeneity including the two-group two-parameter TITE-CRM and the shift model TITE-CRM. Approaches are recommended for different scenarios. This dissertation addresses practical issues associated with the application of the TITE-CRM to patient heterogeneity. We show that the TITE-CRM accommodates two-group heterogeneity and have developed a program to implement the method. We also consider the idea of exploring the presence of heterogeneity using the TITE-CRM models in addition to evaluating the homogeneous model.

Included in

Public Health Commons

Share

COinS