Randomized trial evaluates decision-making strategies that influence efficiency in multi-arm multi-stage trials, suggesting optimization improves power without increasing error rates.
Key Points
The aim is to evaluate how different decision-making strategies impact the efficiency of Bayesian multi-arm multi-stage trials.
Applied the Nelder–Mead optimization algorithm for setting decision thresholds.
Conducted a simulation study comparing conventional methods to three alternatives.
Derived posterior probabilities using a Normal–Gamma conjugate model at each interim analysis.
All treatment comparison methods exhibited similar power across various scenarios.
Optimal decision thresholds differed significantly among the evaluated methods.
Adjusted thresholds showed potential for improved efficiency without increasing the type I error rate.