Robust optimal design for the estimation of hyperparameters in population pharmacokinetics

Area of Science:

• Pharmacokinetics and Pharmacometric Modeling
• Statistical Modeling and Experimental Design

Background:

• Accurate estimation of population pharmacokinetic (PK) parameters is essential for drug development and clinical application.
• Traditional study designs may not be optimal for capturing PK variability and ensuring precise parameter estimation.

Purpose of the Study:

• To propose and evaluate the expectation of the determinant of the inverse of the population Fisher information matrix (EID) as a criterion for optimizing population PK study designs.
• To determine optimal sampling times within defined constraints to enhance the estimation of population PK parameters.

Main Methods:

• The EID criterion was minimized to find optimal population designs, considering PK models, measurement error models, and prior distributions for parameters.
• A group-based approach was used, where subjects within a group share the same sampling schedule.
• Optimization of sampling times was performed using a stochastic gradient algorithm for a one-compartment open model.

Main Results:

• The EID criterion was successfully applied to optimize sampling schedules for a one-compartment PK model.
• The study investigated the influence of various factors, including the number of sampling schemes, group sizes, and uncertainties in parameter distributions.
• The method demonstrated flexibility in handling different error models (homoscedastic and heteroscedastic) and dose assumptions.

Conclusions:

• The EID criterion provides an effective method for evaluating and optimizing population PK study designs.
• Optimal sampling time determination using this criterion can lead to more precise estimation of population PK parameters.
• The approach is robust to variations in study design parameters and model assumptions.