A Modified Hybrid Wald's Approximation Method for Efficient Covariate Selection in Population Pharmacokinetic...

Area of Science:

• Pharmacokinetics and Pharmacometrics
• Statistical Modeling
• Drug Development

Background:

• Population pharmacokinetic (PPK) analysis is crucial for understanding drug behavior in diverse patient populations.
• Identifying covariate relationships enhances PPK model interpretability and predictive power.
• Existing methods for covariate model development, such as stepwise procedures and Wald's approximation method (WAM), have limitations.

Purpose of the Study:

• To introduce and evaluate an innovative covariate model development method, H-WAM-BE (hybrid first-order conditional estimation/Monte-Carlo parametric expectation maximization-based Wald's approximation method with backward elimination).
• To compare the performance of H-WAM-BE against established methods like LRT-based SCM and H-WAM-F using simulation and real-world datasets.

Main Methods:

• The H-WAM-BE method integrates first-order conditional estimation (FOCE) for parameter estimation, MCPEM for covariance matrix estimation, and backward elimination (BE) for covariate selection.
• Performance was assessed using simulated datasets across various scenarios (sample size, sampling schemes) and a rituximab dataset.
• Evaluation metrics included the identification of true and false positive covariates, reference model identification, computation time, and predictive performance.

Main Results:

• The best-performing H-WAM-BE approaches (M2 and M4) demonstrated results comparable to the likelihood ratio test (LRT)-based stepwise covariate method (SCM).
• H-WAM-BE consistently required shorter or comparable computation times compared to LRT-based SCM and H-WAM-F.
• The method's efficiency was maintained across different model structures, sample sizes, and sampling designs.

Conclusions:

• H-WAM-BE offers an effective and computationally efficient alternative for covariate model development in population pharmacokinetic analyses.
• The method provides a robust approach for identifying significant covariate-parameter relationships, aiding in the refinement of PPK models.
• H-WAM-BE represents a valuable advancement in statistical methodologies for drug development and clinical pharmacology research.