Sample sizes for randomized controlled trials utilizing Bayesian response adaptive randomization for continuous...

Area of Science:

• Biostatistics
• Clinical Trial Design
• Statistical Modeling

Background:

• Sample size estimation is crucial for multi-arm randomized controlled trials (RCTs).
• Response adaptive randomization (RAR) dynamically adjusts treatment allocation based on accumulating data.
• Bayesian approaches offer a flexible framework for complex trial designs.

Purpose of the Study:

• To present a Bayesian approach for sample size estimation in multi-arm RCTs with continuous outcomes using RAR.
• To compare sample size requirements between hypothesis testing and Bayesian methods under different interim analysis scenarios.
• To evaluate the impact of RAR on sample size estimation, particularly with large treatment effects.

Main Methods:

• Developed a Bayesian framework incorporating outcome data to update treatment effect estimates and modify allocation proportions.
• Simulated a 4-arm trial to compare sample sizes from hypothesis testing versus Bayesian approaches with and without interim analyses and RAR.
• Applied the proposed method to two completed RCTs to validate findings.

Main Results:

• Two interim analyses (at 25% and 50% enrollment) can reduce the overall number of participants needed.
• RAR-based sample size estimates can increase, especially when treatment effects are large, due to imbalanced allocation.
• Bayesian approach with interim analyses and RAR showed potential for participant reduction compared to non-adaptive designs.

Conclusions:

• Early and frequent interim analyses can decrease the sample size required for conclusive trials.
• The benefits of within-trial patient allocation to effective treatments via RAR must be weighed against potential increases in sample size requirements.
• The proposed Bayesian approach provides a robust method for sample size determination in adaptive RCTs.