Recently, a Bayesian-frequentist hybrid approach was proposed for general hypothesis testing problems including noninferiority testing. The method is hybrid in the sense that it uses Bayes factor as measure of evidence and determines the rejection region by minimizing a weighted sum of average type I and type II errors. To address regularity concerns, an analytical result guides the choice of the weight such that the average type I error is strictly controlled under some pre-specified level. Additionally, because prior probabilities are based on a semiparametric structure, the method has flexibility of accommodating situations with or without reliable prior information. This method was initially developed for clinical trials with binary endpoints. Extensions to survival data will also be discussed.