Exact Randomized Two-Stage Phase 2 Clinical Trial Designs for Two Binary Co-Primary Endpoints

Area of Science:

• Biostatistics
• Clinical Trial Design

Background:

• Phase 2 trials often require demonstrating efficacy across multiple co-primary endpoints using intersection-union hypothesis testing.
• Existing designs for binary endpoints are limited, especially for two-arm trials with small sample sizes, often relying on inadequate normal approximations.

Purpose of the Study:

• To develop and evaluate exact and simulation-based methods for designing two-arm, two-stage phase 2 trials with two co-primary binary endpoints.
• To identify optimal and minimax designs that minimize expected and maximum sample sizes, respectively, with early futility stopping.

Main Methods:

• Utilized the bivariate binomial distribution to model two co-primary binary endpoints with potentially different correlations across treatment arms.
• Developed exact methods for small sample sizes and evaluated simulation-based and normal approximation methods for larger sample sizes by comparing against the exact method.

Main Results:

• The exact method provides analytically exact operating characteristics for small sample sizes (n ≤ 25 per arm).
• The simulation-based approach yielded designs closely matching the exact method, outperforming the normal approximation, which often deviated.
• The simulation method is computationally efficient and accurate for moderate-to-large sample size trials.

Conclusions:

• The simulation-based method is a recommended, computationally efficient, and accurate alternative for designing phase 2 clinical trials with multiple co-primary binary endpoints.
• This framework enhances the design of phase 2 trials by providing robust methods for complex endpoint scenarios.