A recommended analysis for 2 × 2 crossover trials with baseline measurements

Area of Science:

• Biostatistics
• Clinical Trial Design
• Pharmacometrics

Background:

• Two-period, two-treatment (2x2) crossover trials are common in clinical research.
• Continuous outcomes are measured pre- and post-treatment within each period.
• Statistical analysis aims to test the true difference in treatment response means.

Purpose of the Study:

• To investigate how variance-covariance structure and baseline response handling affect statistical power in 2x2 crossover trials.
• To compare various methods for accounting for baseline measurements.
• To recommend an optimal statistical approach for analyzing such trials.

Main Methods:

• Theoretical analysis of statistical power.
• Simulation studies evaluating type I error rates and power properties.
• Comparison of different baseline adjustment methods: ignoring baselines, common change from baseline, covariate adjustment, and joint modeling.
• Analysis of covariance (ANCOVA) with difference scores.

Main Results:

• The structure of the within-subject variance-covariance matrix and the method of handling baseline responses critically influence statistical power.
• Common change-from-baseline analyses are generally not recommended.
• An ANCOVA approach using the difference in treatment responses as the dependent variable and the difference in baseline responses as a covariate demonstrated favorable properties.

Conclusions:

• The choice of statistical method significantly impacts the power of two-period, two-treatment crossover trials.
• Accounting for baseline responses appropriately is crucial for maximizing study power.
• The recommended ANCOVA approach provides a robust and powerful method for analyzing data from these trial designs.