On Type I and II error adjustment and efficiency of two-stage design in accelerated drug development

In traditional drug development, phase 2 dose-finding studies and phase 3 confirmatory studies are conducted separately. This paradigm may be inefficient because pivotal analyses are based only on data from phase 3 trials as the definitive source of evidence for drug approval, and enrollment momentum may be lost when the program is interrupted for the transition from phase 2 to phase 3.; This thesis describes a method for the design of a clinical trial to combine phase 2 and 3 of clinical development. The method provides a two-stage study design allowing a fixed or data dependent number of treatment arms and sub-populations to be selected for stage 2, based on observations at stage 1 and a futility boundary. Examples are given to illustrate the method and to examine advantages of the two-stage flexible design. The key components of a two-stage flexible design are to combine two separate trials into one trial, and use data collected before and after the adaptation in the final analysis. The motivation behind the design is to save time and be more efficient by incorporating the data from the selection stage into the final analysis. Given a fixed number of patients, the study would be able to have a better chance to be conclusive.; In consideration of limited resources for drug development, we study some key questions in finding a good strategy for detecting the best treatment(s)/population(s) and producing a significant result given a pre-fixed sample size and desire to have strong control of the familywise Type I error rate. For treatment selection given a prespecified fixed sample size, practical recommendations are made on: whether to select one arm or multiple arms, whether to expand the number of experimental arms in a trial, and what is the appropriate timing of interim analysis. For target population selection, we examine the effect of unequal sample size between subgroups, appropriate stopping boundaries, and timing of interim analysis on the study power.