Improved power in crossover designs through linear combinations of baselines

Area of Science:

• Biostatistics
• Clinical Trial Design
• Statistical Methods

Background:

• Including period-specific baselines as covariates in analysis of covariance (ANCOVA) increases precision in crossover trials.
• The efficiency gain depends on the covariance structure of baseline and post-treatment responses.

Purpose of the Study:

• To leverage the underlying covariance structure to find an optimal linear combination of baselines.
• To minimize the theoretical variance of the ANCOVA-based estimated treatment effect.
• To propose an adaptive method for unknown covariance structures in complete crossover designs (2x2, 3x3, 4x4).

Main Methods:

• Developed a method to find an optimal linear combination of baselines by leveraging the covariance structure.
• Proposed an adaptive approach where the covariance structure is selected using an information criterion.
• Evaluated the method's performance in complete designs up to four periods.

Main Results:

• The proposed adaptive method maintains the type I error rate.
• Significant gains in statistical power are achieved compared to previously published methods.
• Illustrative examples from 2x2 (renal function) and 3x3 (heart rate) trials demonstrate the method's applicability.

Conclusions:

• The adaptive method effectively optimizes baseline utilization in crossover trials.
• This approach offers a statistically powerful and reliable tool for clinical trial analysis.
• The method provides a practical solution for enhancing treatment effect estimation when covariance structures are unknown.