Statistical identifiability and convergence evaluation for nonlinear pharmacokinetic models with particle swarm...

Area of Science:

• Pharmacokinetics
• Nonlinear Dynamical Systems
• Computational Biology

Background:

• Statistical identifiability is crucial for nonlinear pharmacokinetic (PK) models, particularly those using Michaelis-Menten (MM) kinetics.
• Conventional derivative-based estimation methods often fail to converge for non-identifiable models due to singularity issues.

Purpose of the Study:

• To develop a robust, derivative-free global optimization algorithm for assessing the statistical identifiability of nonlinear PK models.
• To enhance the convergence rate of particle swarm optimization (PSO) by combining it with a local optimization algorithm.
• To establish an efficient method for checking estimation convergence and detecting model identifiability.

Main Methods:

• A hybrid global optimization algorithm combining particle swarm optimization (PSO) with a derivative-free local optimization algorithm was developed.
• The approach focuses on overcoming convergence issues in statistically non-identifiable models.
• The method was validated using PK simulation studies and applied to clinical PK data with a two-compartmental model.

Main Results:

• The proposed derivative-free global optimization approach demonstrated efficient detection of convergence and identifiability for nonlinear PK models.
• The hybrid algorithm improved the convergence rate compared to standard PSO.
• Successful application to clinical PK data confirmed the approach's practical utility.

Conclusions:

• The developed derivative-free global optimization strategy effectively addresses the statistical identifiability challenges in nonlinear PK models.
• This method provides a reliable tool for analyzing complex PK models, including those with Michaelis-Menten kinetics.
• The approach enhances the accuracy and efficiency of PK model parameter estimation and identifiability assessment.