Refinement of the Population Pharmacokinetic Model for the Monoclonal Antibody Matuzumab

A developed population pharmacokinetic model of the humanized monoclonal antibody (mAb) matuzumab was evaluated by external evaluation. Based on the estimates of the final model, simulations of different dosing regimens and the covariate effect were performed.

The development dataset included 90 patients, and the evaluation dataset included 81 patients; the two sets of patients were from three different studies. In all studies, the patients had different types of advanced carcinoma — mainly colon, rectal and pancreatic cancer. They received matuzumab as multiple 1-hour intravenous infusions in a wide range of dosing regimens (development dataset: from 400 mg every 3 weeks to 2000 mg in the first week followed by 1600 mg weekly; evaluation dataset: from 100 mg weekly to 800 mg weekly). In addition to 1256 serum mAb concentrations for model development, there were 1124 concentrations available for model evaluation. Serum concentration-time data were simultaneously fitted using NONMEM™ software. The developed two-compartment model — with the parameters central volume of distribution (V₁) and peripheral volume of distribution (V₂), intercompartmental clearance and linear clearance (CLL), an additional nonlinear elimination pathway (Michaelis-Menten constant: the concentration with the half-maximal elimination rate and V_max: the maximum elimination rate) and covariate relations — was evaluated by an external dataset. Different simulation scenarios were performed to demonstrate the impact of the incorporated covariate effect and the influence of different dosing regimens and dosing strategies on the concentration-time profiles.

The developed model included the covariate fat-free mass (FFM) on V₁ and on CLL. The evaluation did not support the covariate FFM on V₁ and, after deletion of this covariate, the model parameters of the refined model were estimated. The model showed good precision for all parameters: the relative standard errors (RSEs) were <42% for the development dataset and ≤51% for the evaluation dataset (excluding the higher RSEs for the correlation between V₂ and V_max and the interindividual variability on V₂ for the evaluation dataset). The model showed good robustness for the ability to estimate highly precise parameters for the combined dataset of 171 patients (RSE <29%). Simulations revealed that variability in concentration-time profiles for minimum and maximum steady-state concentrations was reduced to a marginal extent by a proposed dose adaptation.

The population pharmacokinetic model for matuzumab was improved by evaluation with an external dataset. The new model obtained precise parameter estimates and demonstrated robustness. After correlation with efficacy data simulation results in particular could serve as a tool to guide dose selection for this ‘targeted’ cancer therapy.