Hybrid classical-Bayesian approach to sample size determination for two-arm superiority clinical trials

Area of Science:

• Biostatistics
• Clinical Trial Design
• Statistical Inference

Background:

• Traditional sample size determination (SSD) relies on fixed values or estimates for unknown parameters in power analysis.
• Existing methods may not fully account for uncertainty in these parameters.
• A hybrid approach offers a way to integrate prior information within a frequentist framework.

Purpose of the Study:

• To propose a novel hybrid classical-Bayesian procedure for SSD in two-arm superiority trials.
• To formally incorporate uncertainty regarding unknown parameters influencing statistical power.
• To differentiate the use of prior distributions for design expectations versus preliminary estimate modeling.

Main Methods:

• Development of a hybrid classical-Bayesian procedure for SSD.
• Application to two-arm superiority trials with binary data.
• Derivation of hybrid criteria using difference in proportions, log relative risk, and log odds ratio.

Main Results:

• The proposed hybrid procedure allows for formal incorporation of uncertainty in unknown parameters.
• Different prior distributions are utilized to reflect design expectations and preliminary estimate uncertainty.
• The method is illustrated with numerical examples for binary outcomes.

Conclusions:

• The hybrid classical-Bayesian approach offers a more robust method for SSD in superiority trials.
• It provides a flexible framework for handling parameter uncertainty.
• The derived criteria are practical for implementation in clinical trial design.