Design and analysis of crossover trials for absorbing binary endpoints.

Area of Science:

• Biostatistics
• Clinical Trial Design
• Epidemiology

Background:

• Crossover designs are efficient for patient heterogeneity but not typically used for absorbing binary endpoints.
• Conventional wisdom suggests crossover designs are inappropriate for endpoints such as death or HIV infection.

Purpose of the Study:

• To explore the use of crossover designs for absorbing binary endpoints.
• To introduce and evaluate a novel two-period crossover-parallel design for such endpoints.
• To assess the efficiency and validity of these designs compared to standard parallel group designs.

Main Methods:

• Exploration of crossover designs for absorbing binary endpoints.
• Introduction of a new two-period design with rerandomization of survivors.
• Evaluation of a mixture model and a modified Mantel-Haenszel test for statistical inference.
• Simulation studies to compare design efficiencies.

Main Results:

• Crossover designs can be more efficient than parallel group designs for absorbing binary endpoints when patient risk heterogeneity exists.
• The novel two-period design combines crossover efficiency with parallel design comparability.
• The mixture model and Mantel-Haenszel test provide valid inference methods.

Conclusions:

• Crossover designs are viable and potentially more efficient for absorbing binary endpoints in the presence of patient heterogeneity.
• The proposed two-period design offers a practical solution balancing efficiency and comparability.
• Further exploration and practical implementation are warranted.