On group sequential designs comparing two binomial proportions

Area of Science:

• Biostatistics
• Clinical Trial Design
• Statistical Inference

Background:

• Determining appropriate sample size is crucial for the statistical power and efficiency of hypothesis tests.
• Comparing two binomial proportions is common in various scientific fields, including medicine and social sciences.
• Existing methods for sample size determination may not fully account for early stopping opportunities.

Purpose of the Study:

• To develop and evaluate exact group sequential designs for one- and two-stage hypothesis tests comparing two binomial proportions.
• To identify designs that allow for early stopping based on futility or efficacy, or both.
• To assess the sample size savings and properties of these novel designs compared to existing methods.

Main Methods:

• A FORTRAN-compiled search algorithm was employed to identify exact group sequential designs.
• The algorithm considered one- and two-stage hypothesis testing frameworks.
• Designs were evaluated for their ability to stop early for futility, efficacy, or both.

Main Results:

• The proposed exact group sequential designs offer substantial sample size savings in many practical scenarios.
• These designs demonstrate desirable statistical properties, ensuring robust hypothesis testing.
• Comparisons show advantages over other exact one- and two-stage designs and asymptotic methods.

Conclusions:

• Exact group sequential designs provide an efficient approach for sample size determination in comparative binomial studies.
• The ability to stop early for futility or efficacy enhances the practicality and resource-effectiveness of clinical trials.
• These findings support the adoption of group sequential methods for improved statistical trial design.