Simulation Controlled Seamless Phase II/III Clinical Trials

Clinical trials have become a standard part of modern medicine, creating great interest in the field of adaptive clinical trials. The goal of adaptive trials is greater efficiency in the use of patients and time, higher probability of demonstrating treatment effect, and more informative trials. This paper focuses on seamless phase II/III trials, an adaptive trial consisting of two stages: the first stage compares multiple treatment arms to a control and determines the appropriate arm, and the second stage undertakes a more traditional comparison of this arm to a control arm. The data from both stages is used in analysis of the treatment, with the goal of more power than two separate trials where data is not shared. This paper begins with a discussion of an established method of analyzing seamless trials, referred to as the Posch method. Proofs are presented that conclude that the Posch method has analytical control over type one error. This paper also introduces a new method of analyzing seamless trials called the simulation method. This simulation method relies on simulation of clinical trials under the null hypothesis to find critical values that control type one error. This paper includes a model of the simulation method as a first step in a potential proof that the simulation method has analytical control over type one error. Power comparisons between the Posch method, the simulation method, and two separate trials conclude that seamless trials are more powerful than separate trials, and the simulation method is slightly more powerful than the Posch method when analyzing seamless trials. Power comparisons of separate trials and the simulation method with a fixed sample size reveal that implementation of the simulation method would result in larger phase II trials than when separate trials are run. This paper serves as an initial step in establishing the simulation method as a method of analyzing seamless phase II/III clinical trials.