Package: BayesianQDM 1.0.1

BayesianQDM: Bayesian Quantitative Decision-Making Framework for Binary, Continuous and Time-to-Event Endpoints

Bayesian quantitative decision-making for proof-of-concept clinical trials with Go, NoGo and Gray decisions. A study is specified by its sample sizes, conjugate prior distributions, external data borrowed through power priors, a concurrent, external, hypothetical or fixed control group, and a decision criterion based on the posterior or posterior predictive probability of the treatment effect. The package computes the decision for observed data, the operating characteristics of a design under scenarios for the true parameters, the decision thresholds that control the probabilities of false Go and false NoGo decisions, and the observed data that lead to each decision. One or two binary endpoints (Beta-binomial and Dirichlet-multinomial models), one or two continuous endpoints (Normal-Inverse-chi-squared and Normal-Inverse-Wishart models) and one time-to-event endpoint evaluated by the milestone survival probability or the hazard ratio (exponential model with Gamma priors) are supported, with larger or smaller effects as favorable. The posterior distributions have closed forms, so Markov chain Monte Carlo (MCMC) sampling is not needed; the probabilities are computed from closed-form distribution functions, by exact summation, numerical integration, a moment-matching approximation or Monte Carlo simulation, with the main computations in C++. See Kang, Yamaguchi, and Han (2026) <doi:10.1080/10543406.2026.2655410> for the methodological framework.

Authors:Gosuke Homma [aut, cre], Yusuke Yamaguchi [aut]