Bayesian Methods for Response-Adaptive Randomization and Drug Repurposing

Abstract

Phase II clinical trials are an expensive and risky component of the drug development process used to study the safety and efficacy of new treatments. Recent studies suggest that the average cost of a phase II trial ranges from 7.0 to 19.6 million dollars, depending on therapeutic area, and that about one-third of all phase II trials fail. Failures in phase II generally occur when a new toxic side effect arises or the observed treatment effect is smaller than anticipated.To minimize harm to trial participants, drug developers require statistical methods that can be used to reduce patient exposure to ineffective or harmful treatments. This thesis focuses on innovations in early phase development aimed at reducing exposure to ineffective treatments. Chapters 2 and 3 concern response-adaptive randomization (RAR), which alters the allocation ratio based on accruing data in favor of the empirically superior treatment. In contrast to fixed 1:1 allocation, RAR gives participants a greater chance of receiving the treatment during the trial, which tends to reduce the number of participants assigned to the inferior arm. Yet, existing RAR approaches are commonly criticized for reducing power relative to 1:1 allocation, inflating type I error rate when a time-trend is present, and engendering nontrivial probabilities of allocating more participants to the inferior arm. To temper these problematic behaviors, Chapter 2 proposes a new probability model and randomization strategy for implementing Bayesian RAR in a binary outcomes setting. Simulation studies show that the proposed methods engender smaller average sample sizes with similar power, better control over type I error rate, and a negligible chance of a sample size imbalance in the wrong direction compared to the traditional design. Chapter 3 proposes a new metric for comparing group sequential designs that measures the expected number of failures in the fixed group of individuals who are directly impacted by the design choice. In contrast to within-trial metrics, this approach considers designs with equal type I and II error rates and assesses their impact with respect to relevant, equal-sized populations. Simulation studies show that various implementations of group sequential Bayesian RAR offer modest improvements with respect to the proposed metric relative to conventional group sequential monitoring alone. Chapter 4 concerns drug repurposing, which is the process of discovering new therapeutic uses for existing treatments. Drug repurposing involves studying treatments with well-established safety profiles, which can dramatically shorten the drug development timeline and reduce the occurrence of toxic side effects in patients. However, existing approaches for drug repurposing involve complex, computationally-intensive analytical methods that are not widely used in practice. This chapter proposes a novel Bayesian network meta-analysis (NMA) framework that can predict the efficacy of an approved treatment in a new indication and thereby identify candidate treatments for repurposing. We obtain predictions using two main steps: first, we use standard NMA modeling to estimate average relative effects from a network comprised of treatments studied in both indications in addition to one treatment studied in only one indication. Then, we model the correlation between relative effects using various strategies that differ in how they model treatments across indications and within the same drug class. Simulation studies find that the model minimizing root mean squared error of the posterior median for the candidate treatment depends on the amount of available data, the level of correlation between indications, and whether treatment effects differ, on average, by drug class.