Interaction in crossover studies: a modified analysis with more sensitivity

Area of Science:

• Clinical Trials Methodology
• Biostatistics
• Medical Research Design

Background:

• Crossover trials offer intuitive patient-as-own-control designs, eliminating between-subject variability.
• However, these designs are susceptible to treatment-by-period interaction bias, including carryover effects.
• The standard Hills-Armitage analysis for interaction lacks sensitivity in common clinical scenarios.

Purpose of the Study:

• To present an alternative statistical method for analyzing crossover trials.
• To improve the sensitivity of detecting treatment-by-period interactions, particularly carryover effects.
• To compare the performance of the alternative method against the standard Hills-Armitage analysis.

Main Methods:

• Developed an alternative analysis focusing on individual treatment group performance rather than group means.
• Evaluated the sensitivity of the new method in detecting treatment-by-period interactions.
• Compared the sensitivity of the alternative method with the standard Hills-Armitage analysis.

Main Results:

• The alternative method demonstrated higher sensitivity than the standard approach when interaction occurred in only one treatment group.
• The standard Hills-Armitage analysis remained more sensitive in cases of two-group interaction.
• The study confirmed the enhanced sensitivity of the alternative method in specific interaction scenarios.

Conclusions:

• The proposed alternative method offers improved sensitivity for detecting treatment-by-period interactions in crossover trials.
• This alternative approach complements the standard analysis, providing a more comprehensive evaluation of treatment effects.
• Clinicians can benefit from this enhanced method for more accurate interpretation of crossover trial data.