A simulation study to assess the influence of population pharmacokinetic model selection on Bayesian forecasting of vancomycin exposure in preterm neonates.

Abstract

Background: Discrepancies in vancomycin AUC24 estimates across population pharmacokinetic (popPK) models raise concerns about their impact on dosing recommendations and probability of target attainment (PTA) after multiple therapeutic drug monitoring (TDM) occasions. This study assesses model selection's impact on model-based dosing and PTA following simulated TDM occasions and explores predictors of PTA.

Research design and methods: Using a virtual neonatal population and 15 validated vancomycin popPK models, TDM events were simulated with initial dosing, followed by Bayesian-guided adjustments over three TDM occasions. The number of non-therapeutic patients was compared across models, and associations between model characteristics and simulated PTA were explored.

Results: Across models, geometric mean (gMean) AUC24 values varied: 496 mg·h/L (35.7% geometric coefficient of variation, gCV) for the first TDM, 570 mg·h/L (53.1% gCV) for the second, and 484 mg·h/L (62.2% gCV) for the third. Non-therapeutic patients ranged from 0 (Mulubwa et al.) to 530 (Allegaert et al.). Interindividual variability in clearance correlated with non-therapeutic patients (R² = 0.88), while residual variability showed no correlation.

Conclusion: Model selection affects simulated vancomycin AUC24 values and PTA in neonates, influencing dosing decisions. These findings highlight model characteristics correlating with target attainment.