Randomized trial evaluates Bayesian sample size estimation in continuous outcomes, suggesting fewer participants needed with adaptive design.
Key Points
This research aims to evaluate a Bayesian method for estimating sample sizes in multi-arm trials using response adaptive randomization for continuous outcomes.
Utilized simulations for a 4-arm trial to compare sample size estimations across different approaches.
Investigated three scenarios: no interim analysis, non-adaptive randomization, and RAR with interim analyses.
Conducted two interim analyses at 25% and 50% participant enrollment.
RAR with interim analyses could reduce participant enrollment by optimizing allocation toward effective treatments.
For larger treatment effects, RAR-based estimates increased due to imbalanced participant allocation.
Application to completed trials confirmed that early interim analyses minimize necessary sample sizes.