A SIGNATURE ENRICHMENT DESIGN WITH BAYESIAN ADAPTIVE RANDOMIZATION FOR CANCER CLINICAL TRIALS

Clinical trials in the era of precision medicine demand more flexible and efficient trial designs. Adaptive clinical trial designs allow pre-specified modifications of an on-going clinical trial and could shorten the trial duration. We reviewed five common types of adaptive clinical trials based on adaptation methods. In particular, outcome-randomization becomes more popular as it can assign more patients to the promising treatments based on the accumulated trial data. This data-driven allocation allows more patients to benefit from the trial, which is especially important for cancer patients. We compared different Bayesian outcome-adaptive randomization methods and discussed them from both methodological and ethical aspects. When the group of patients who are likely to benefit from the test treatment is known, the clinical trial should focus on that sensitive subpopulation. The use of biomarkers in adaptive clinical trials can guide the assignment of individually optimal treatments to patients. To address this objective, we proposed a cross-validated signature enrichment design combined with Bayesian response-adaptive randomization. We evaluated the performance of this design using four criteria based on the benefits and losses for individuals inside and outside of the clinical trial. The proposed design allows more patients to receive optimal personalized treatments, thus yielding a higher response rate. This design can identify therapies that are globally beneficial as well as treatments that are effective only in a sensitive subset. Simulation studies demonstrate the advantages of the proposed design over alternative designs. The approach is also illustrated by an example based on an actual clinical trial in non-small-cell lung cancer. In contrast to the traditional two-arm trial, an umbrella trial is a master protocol that studies multiple drugs within a single indication, which can result in more accurate and efficient drug development. Based on SEDAR, we proposed a SEDAR-U design that is suitable for umbrella trials. We extended SEDAR to allow multiple arms and incorporate an early termination rule. Simulation studies showed that the SEDAR-U design can allow more enrolled patients to benefit from the promising treatment and achieve favorable operating characteristics