Personalized treatment selection using data from crossover designs with carry-over effects

Area of Science:

• Biostatistics
• Clinical Trial Design
• Pharmacometrics

Background:

• Crossover designs are efficient for clinical trials but can be complicated by carry-over effects.
• Nonparametric methods for carry-over effects may limit the use of joint patient measurements for treatment comparison.
• Estimating optimal patient-specific treatments requires methods that account for individual covariates and treatment carry-over.

Purpose of the Study:

• To develop a semiparametric method for estimating optimal patient treatment in crossover trials with carry-over effects.
• To address limitations of nonparametric methods in comparing treatments when carry-over is present.
• To propose a robust approach for personalized treatment assignment using covariate information.

Main Methods:

• A semiparametric approach is proposed for optimal treatment estimation.
• The method compares probabilities of treatment dominance based on patient-specific scores derived from covariates.
• Single-index models are utilized to link outcome variables with covariates.

Main Results:

• The proposed method demonstrates highly accurate frequencies of correct treatment assignments.
• Empirical investigations confirm the effectiveness of the approach in identifying optimal treatments.
• The method shows robustness against deviations from the single-index model structure.

Conclusions:

• The semiparametric method provides an effective way to estimate optimal treatments in crossover designs with carry-over effects.
• The approach overcomes limitations of traditional methods by utilizing joint patient measurements.
• Real data analysis validates the practical applicability of the proposed procedure for personalized medicine.