Abstract:
|
Innovative Bayesian designs have been increasingly adopted in phase 2 clinical trials that aim to gauge the early signals of efficacy and establish an optimal dose through dose ranging studies. We propose a Bayesian multi-stage design to combine both studies and aid in phase 2 trials with reduced trial duration and patient use. The design seeks to optimize patient allocation by controlling both Frequentist and Bayesian error rates, while increasing statistical powers to detect drug efficacy with borrowing of historical data. Flexible prior elicitation is enabled to address known issues in Bayesian borrowing, such as selection bias, prior-data conflict and between-study variability. Further sample size reduction is feasible by borrowing historical or synthetic controls using artificial intelligence and machine learning approaches. The merits of the proposed Bayesian design are demonstrated through operating characteristics from comprehensive simulation studies inspired by phase 2 trials for Inflammatory bowel disease. The examples involve flexible distributional assumptions on the efficacy endpoints, and can be easily adapted to other therapeutic areas.
|