Group sequential t-test for clinical trials with small sample sizes across stages.

Area of Science:

• Biostatistics
• Clinical Trial Design
• Statistical Methods

Background:

• Group sequential t-tests are crucial for interim analyses in clinical trials.
• Current methods rely on normal approximations for critical values, which can be inaccurate with small sample sizes.
• Inaccurate critical values can affect the reliability of treatment effect assessments.

Purpose of the Study:

• To develop a more accurate method for determining critical values in group sequential t-tests.
• To address the limitations of normal approximation in clinical trials with varying sample sizes.
• To provide practical tools and values for improved clinical trial analysis.

Main Methods:

• Employed a Monte Carlo method to directly derive critical values, bypassing normal approximation.
• Calculated critical values for specific sample sizes and numbers of interim analyses.
• Developed SAS code for broader application and conducted simulations to validate accuracy.

Main Results:

• The proposed Monte Carlo critical values yield type I error probabilities close to the nominal significance level.
• Existing critical values based on normal approximation showed inaccuracies with small sample sizes across stages.
• Simulations confirmed the superior accuracy of the Monte Carlo method in diverse clinical trial scenarios.

Conclusions:

• The Monte Carlo method offers a more accurate and reliable approach for critical values in group sequential t-tests.
• This method enhances the precision of treatment effect evaluation in clinical trials, particularly when sample sizes are small.
• The findings support the adoption of this method for more robust clinical trial statistical analysis.