Early and Middle Phase Clinical Trial Designs for Groups

This dissertation presents phase I and phase II trial designs for groups. In particular, groups completely or partially ordered by dose sensitivity are considered. Groups are completely or partially ordered when the groups can be completely or partially ordered by the probability of an adverse event for any given dose, respectively. A pair of phase I clinical trials are presented in this dissertation: the Quasi-CRM Shift method and the Group Averaged Bayesian Optimal Interval Design (GAB). The Quasi-CRM Shift method is the first design for partially ordered groups considering ordinal toxicity, allowing clinicians to control for the frequency and severity of adverse events during dose selection. GAB is the first model-assisted design for partially ordered groups, a class of designs marked by their simplicity. Simulation studies show that GAB performs as well as more complex model-based designs, demonstrating GAB provides clinicians with a simple design that performs well. Large sample properties show allocation under GAB tends to correct doses. Both the Quasi-CRM Shift and GAB demonstrate that utilizing the group ordering leads to increased accuracy in dose allocation during the trial and dose selection at the end of the trial. In addition to the phase I designs, a pair of phase II designs are presented in this dissertation. These designs consider a trial with multiple doses and two ordered groups. At the end of the trial, doses are determined to be acceptable or unacceptable. The first design is a single-stage design and maximizes power subject to a type I error constraint. The second design is a two-stage design with cutoffs determining if a dose continues onto the second stage for a group. This design minimizes the number of unacceptable doses that continue onto the second stage while meeting power and type I error requirements. Through the development of these designs, clinicians are provided with multi-dose designs in the group framework, extending dose exploration and optimization into phase II.