Developing new methods for computing confidence intervals for clinical trials with subpopulation selection

It is widely recognised that an experimental therapy may only benefit a subset of a patient population. An efficient clinical trial design to test whether an experimental therapy is effective and for which part of the population is the two-stage adaptive enrichment design. The trial initially recruits from the full patient population and then partway through the trial, an interim analysis is performed to select the subpopulation where the experimental therapy is promising. Recruitment after the interim analysis is restricted to this subpopulation. Sometimes, the existing method for computing a confidence interval (CI) for the effect size, which is an interval that contains true effect size with a specified probability, gives CIs that are known to be inaccurate. This is not helpful to decision makers so that the aim of the PhD project is to develop methods that give confidence intervals with better properties.