Bayesian Prediction of Satisfaction Indices via Sequential Analysis and Experimental Design in Hybrid and Full Bayesian Frameworks

via Sequential Analysis and Experimental Design in Hybrid and Full Bayesian Frameworks

Abstract

This paper proposes a methodology for phase II clinical trials, grounded in both hybrid and fully Bayesian designs. The approach leverages predictive probabilities to monitor trial progress, particularly focusing on the probability of rejecting the null hypothesis and satisfaction index inversely related to the $p-value$. The methodology is applied to a negative binomial model, which is adapted to binary outcomes. This model allows for the estimation of the number of trials required to achieve a pre-specified success rate, thereby accounting for variability in observed outcomes. The proposed framework combines operational flexibility with statistical rigor, supporting interim decision-making based on futility or efficacy criteria. Ilustrative simulation demonstrate the clinical relevance of the approach, particularly its potential to optimize resource allocation, reduce patient risk, and expedite the evaluation of promising therapeutic candidates.