Bayesian randomized non-inferiority trial design

A few random comments. Joint data analysis is in my view better than summarizing previous studies through priors. For one thing you can do covariate adjustment. See this which is not a meta-analysis, in the sense that it does not assume the data sources are exchangeable. Bayes doesn’t need non-inferiority designs but rather just computes for example P(HR < 1.1). The margin that is chosen is critical. It must be the same margin one would claim superiority with.