A decision theoretic approach to sample size determination in clinical trials

Area of Science:

• Biostatistics
• Clinical Trial Design
• Statistical Inference

Background:

• Determining appropriate sample size is critical for the validity and efficiency of phase III clinical trials.
• Traditional methods often assume known variance, which may not reflect real-world scenarios.
• Phase III clinical trials require robust statistical methodologies to ensure reliable results.

Purpose of the Study:

• To present a Behavioral Bayes approach for sample size determination in phase III clinical trials.
• To extend existing methodologies by incorporating unknown mean and variance in the normal distribution model.
• To develop and describe software that optimizes sample size by minimizing expected net cost.

Main Methods:

• The study employs a Behavioral Bayes framework for statistical decision-making.
• Sample size is determined by minimizing the expected net cost function.
• The methodology assumes data follows a normal distribution with both unknown mean and variance.

Main Results:

• A software tool is described that implements the proposed methodology.
• Numerical examples demonstrate the impact of unknown variance on optimal sample size calculations.
• The generalized model, accounting for unknown variance, significantly affects optimal sample size compared to known variance models.

Conclusions:

• The Behavioral Bayes approach provides a more comprehensive method for sample size determination in phase III clinical trials when variance is unknown.
• The developed software aids researchers in establishing optimal sample sizes, enhancing trial efficiency.
• Accounting for unknown variance is crucial for accurate sample size determination and can lead to substantial differences in required sample sizes.