Biomarker-based Dose-finding Designs for Single- or Multiple-Agent Phase I Trials

Xue, Yuan, Statistics - Graduate School of Arts and Sciences, University of Virginia
Conaway, Mark, PBHS Public Health Sciences Admin, University of Virginia
Hu, Feifang, Department of Statistics, University of Virginia

The primary goal of a Phase I clinical trial is to estimate the maximum tolerated dose (MTD). The MTD is defined as the highest dose which can be administered with a "tolerable" level of toxicity. The "tolerable" level of toxicity is based on the probability that a patient in the trial experiences a dose-limiting toxicity (DLT). This dissertation addresses two practical considerations: heterogeneous toxicity response and a partial ordering for the probability of toxicity for available treatments.

First, the majority of methods for the Phase I trials are designed for a homogeneous toxicity response, resulting in a unique MTD for the broad patient population. However, patients may naturally differ in the way they react to a treatment. Clinical useful biomarkers which affect the probability of a DLT have been developed as more toxicity biomarker studies have been done. We propose a new design which chooses a distinct MTD for individual patient by use of toxicity biomarker information, thus contributing to a proper and better treatment for individual patient.

Second, for multiple-agent trials we may be able to identify the order of the probability of toxicity for only a subset of the available treatments, which is a "partial order", in contrast to single-agent trial problems whose order of the probability of toxicity for all the treatments is fully known, which is a "simple order". We propose a biomarker-based design for which the ordering is not fully known.

The operating characteristics of the proposed designs for simple order and partial order are investigated through extensive simulation studies. A discuss of the theoretical properties is provided. We also employ implementing model selection techniques, specifically BIC model selection method to improve the performance of the biomarker-based designs. We close with some conclusions drawn from the proposed biomarker-based designs, as well as some topics for further research.

PHD (Doctor of Philosophy)
Phase I trials, maximum tolerated dose (MTD), biomarker, continual reassessment method (CRM), partial ordering
All rights reserved (no additional license for public reuse)
Issued Date: