Extending the multi-arm multi-stage trial design

The multi-arm multi-stage (MAMS) adaptive clinical trial design has been successfully implemented in several randomised phase III trials. Intermediate short-term outcomes identify and stop recruitment to research arms demonstrating insu cient bene t compared to the control arm at interim stages, before the nal analysis on the primary outcome. The design has been shown to reduce the time and resources required to identify an e ective treatment compared to traditional two-arm designs. This PhD extends the applications of the MAMS design to a broader range of research questions, with the aim of increasing uptake of the design. Stopping recruitment early has been introduced to arms demonstrating overwhelming e cacy on the primary outcome, whilst also stopping for lack-of-bene t on the intermediate outcome for the time-to-event setting. The methods could reduce the patients and resources required should any e cacious arm be identi ed early. Guidelines have been developed on how to design a trial of this nature, and it is shown how to modify the design to control the familywise error rate and power at a pre-speci ed level. It may be necessary to restrict the number of arms in each stage of a MAMS design due to budget constraints or limitations on the number of patients available. This thesis explores how pre-speci ed treatment selection could be implemented, where a subset of arms is chosen at each interim analysis, reducing the maximum sample size. Since selection can potentially lead to bias in treatment e ect estimates, this research also addresses estimation concerns in the proposed design by quantifying the extent of potential bias. Programs for designing MAMS trials have been updated in Stata to accommodate the new methods, to encourage easy adoption of the designs. Finally, practical recommendations have been developed for implementing the proposed ideas, and demonstrates the applications of each of the methods using real trials