All Times EDT
Keywords: Clinical trial, Dose finding, Functional uniform prior, PoC, MCMC
Dose finding is a critical and challenging step in drug development. The dose selection in a Phase II study has a considerable impact on the outcome of the later Phase III confirmatory trial. In practice, severe dose underestimation may result in inadequate clinical benefit and statistical power for trials in later phases. To address this problem, we choose the sigmoid Emax model for dose-response profiling. Through specifying proper prior distributions of model parameters, standard Bayesian model fitting is implemented to extract their posterior distributions. Provided the minimum effective dose (MED), a new rule of target dose estimation is founded upon the posterior conditional probability of overestimation, accompanied by an exploratory "go/no-go" decision rule for PoC. Considering a variety of dose-response profiles, a thorough evaluation of our proposed method is done by extensive simulation studies. We also compare our solution with other commonly accepted dose finding approaches, and demonstrate their strengths and weaknesses.