A0766
Title: Bayes optimal decision-making in adaptive enrichment clinical trials
Authors: Thomas Burnett - University of Bath (United Kingdom) [presenting]
Abstract: Adaptive Enrichment designs allow for the selection of pre-defined patient sub-groups at a pre-planned interim analysis. Given the possible pathways through the trial are pre-defined, hypothesis tests may be defined to ensure strong control of the FamilyWise Error Rate for every possible interim decision. This structure allows complete freedom for how these interim decisions are made. A Bayesian decision framework is used to find optimal decisions in the interim analysis. Evaluating the overall performance of these Bayes optimal Adaptive Enrichment trials is computationally intensive. An algorithm for computing the form of the optimal decision rule shall be discussed. This allows a comparison of the adaptive designs with fixed sampling alternatives, showing the possible benefits of the adaptive methods. Further to optimising the interim analysis, the same optimisation principle may be applied to the whole trial. By optimising across the two-stage adaptive design, fully optimal trials and gain further efficiency in the adaptive design are found. The fixed sampling designs and the previous Enrichment design are special cases in this optimisation, so as a direct result, the most efficient trial design is chosen.
