Statistical Methods in Clinical Trial Design
Embargo until
Date
2018-04-04
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Johns Hopkins University
Abstract
Numerous human medical problems or diseases have been aided by the development of effective treatments such as drugs and medical devices. Clinical trials are an integral part of the development process, determining the safety and efficacy of the new proposed treatment, as required by the Food and Drug Administration of the United States. A reliable, efficient and cost-effective way of conducting the clinical trials is important for advancing useful treatments/devices to market and screening out the useless ones, thus benefiting public health in a timely manner. I developed several statistical methods and applications toward this purpose, ranging from early, small scale Phase I studies to late, large scale Phase III studies in clinical trials.
In Phase I studies, I establish a general framework for a multi-stage adaptive design where I jointly model a continuous efficacy outcome and continuous toxicity endpoints from multiple treatment cycles, unlike the traditional method that only considers a binary toxicity endpoint (joint work with Mayo Clinic). Extensive simulations confirmed that the design had a high probability of making the correct dose selection and good overdose control. To our best knowledge, this proposed Phase I dual-endpoint dose-finding design is the first to incorporate multiple cycles of toxicities and a continuous efficacy outcome.
I also propose and evaluate a two-stage, adaptive clinical trial design for Phase II studies. Its goal is to determine whether future phase 3 (confirmatory) trials should be conducted, and if so, which population should be enrolled. I compute an approximate Bayes optimal design considering a combination of future health benefits and costs.
Turning to Phase III studies, I analyze the performance of adaptive enrichment designs with delayed outcome, leveraging information in baseline variables and short-term outcomes to improve precision by using semiparametric, locally efficient estimators at each interim analysis. I also propose a prediction method for analyzing heterogeneity in treatment response, as a secondary analysis, through the identification of treatment covariate interactions honoring different hierarchical conditions.
Description
Keywords
Clinical Trial Design, Heterogeneity, Statistical Methods