Exact critical values for group sequential designs with small sample sizes

Area of Science:

• Biostatistics
• Clinical Trial Design
• Statistical Methods

Background:

• Group sequential designs enable early stopping of clinical trials.
• Current methods rely on large sample assumptions, potentially inflating type-1 error rates with small sample sizes.
• Existing solutions for small samples use simulations or ad-hoc adjustments.

Purpose of the Study:

• To develop exact joint distributions for test statistics in group sequential designs, applicable to any sample size.
• To derive accurate critical values for common alpha-spending functions.
• To compare the type-1 error rates of the new exact methods against traditional and existing small-sample approaches.

Main Methods:

• Derivation of the exact joint distribution of test statistics for group sequential trials.
• Calculation of exact critical values matching O'Brien-Fleming and Pocock alpha-spending functions.
• Comparative analysis of type-1 error rates using exact, asymptotic, and alternative small-sample methods.

Main Results:

• The study provides a method for calculating exact critical values for group sequential designs, valid for all sample sizes.
• The proposed exact critical values maintain the desired type-1 error rate, unlike asymptotic methods with small samples.
• The new approach offers improved accuracy compared to existing small-sample adjustments.

Conclusions:

• Exact critical values provide a statistically rigorous approach for group sequential clinical trials, particularly with small sample sizes.
• This methodology corrects the type-1 error inflation observed with traditional methods in small samples.
• The findings support the adoption of exact methods for more reliable clinical trial outcomes.