Penalized GEE for Complex Carry-Over in Repeated-Measures Crossover Designs

Area of Science:

• Biostatistics
• Clinical Research Methodology
• Statistical Modeling

Background:

• Crossover designs are widely used in clinical and behavioral research.
• Existing statistical models often make oversimplified assumptions about carry-over effects, ignoring their complexity.
• Complex carry-over effects, varying by treatment order and interaction, have lacked formal statistical estimation methods.

Purpose of the Study:

• To introduce a novel statistical methodology for estimating complex first-order carry-over effects in repeated-measures crossover designs.
• To establish identifiability conditions for complex carry-over effects.
• To provide theoretical guarantees for the proposed estimation approach.

Main Methods:

• Development of a penalized semiparametric Generalized Estimating Equations (GEE) approach.
• Derivation of identifiability conditions for complex carry-over effects.
• Extension of the sandwich variance formula to provide asymptotic normality guarantees.
• Shrinking negligible carry-over effects towards zero for practical identification.

Main Results:

• The proposed penalized GEE method accurately estimates complex carry-over effects.
• Theoretical guarantees demonstrate asymptotic normality for functional components.
• The method effectively identifies and quantifies complex, order-dependent carry-over effects.
• Simulation studies and real-data applications show improved estimation accuracy over simpler models.

Conclusions:

• This work presents the first rigorous and generalizable approach for modeling complex carry-over effects in repeated-measures crossover designs.
• The penalized semiparametric GEE method offers improved accuracy and practical identification of complex carry-over effects.
• This methodology enhances the statistical analysis of crossover trials where treatment order influences outcomes.