The Impact of Model-Misspecification on Model Based Personalised Dosing

Area of Science:

• Pharmacometrics
• Clinical Pharmacology
• Antifungal Drug Development

Background:

• Model Based Personalised Dosing (MBPD) utilizes population pharmacokinetic (PK) or pharmacodynamic models for optimal drug dosing.
• The selection of a specific PK model can significantly influence the clinical utility of MBPD.
• Voriconazole, a widely used antifungal, has multiple published population PK models, presenting a challenge for MBPD implementation.

Purpose of the Study:

• To evaluate the impact of model misspecification on the clinical utility of MBPD.
• To assess how structural model simplifications affect voriconazole dosing recommendations.
• To determine the influence of covariate removal (body weight, CYP2C9 genotype) on MBPD outcomes.

Main Methods:

• Development of five structurally mis-specified population PK models for voriconazole.
• Simulation of plasma concentrations for 100 virtual subjects under each model.
• Determination of dose adjustments using empirical Bayes estimates to achieve target exposure.
• Prediction of plasma concentrations and clinical outcomes based on recommended doses.

Main Results:

• Models lacking non-linear clearance showed poor performance, recommending 155-310 mg higher doses with one plasma concentration.
• Removal of body weight as a covariate had no clinically significant impact on outcomes.
• Exclusion of CYP2C9 genotype, a known influential covariate, also did not significantly affect clinical outcomes.

Conclusions:

• Structural model misspecification, particularly the omission of non-linear clearance, substantially degrades the clinical utility of MBPD for voriconazole.
• Exclusion of key patient covariates like body weight and CYP2C9 genotype from PK models has a negligible impact on the clinical effectiveness of MBPD.
• Accurate structural model components are crucial for reliable MBPD, outweighing the influence of certain covariates.