Towards model-informed precision dosing of intravenous linezolid: a multicentre external evaluation of pharmacokinetic models in critically ill adults.

Abstract

Objectives: Linezolid is a frequently utilised antibiotic to treat serious infections caused by resistant pathogens in critically ill patients. Using the currently recommended one-dose-fits-all strategy leads to insufficient target attainment. Model-informed precision dosing (MIPD), an approach combining mathematical models and therapeutic drug monitoring (TDM), can improve target attainment. The underlying population pharmacokinetic (popPK) model must be selected carefully. The aim of this study was to determine which models are applicable for MIPD in critically ill patients treated with intravenous linezolid.

Methods: Data for intravenous linezolid administration was obtained from three sites and 166 patients (498 TDM samples). The predictive performance of 30 published popPK models was analysed in three scenarios: considering patients' covariates only (a priori, AP), and including one or two TDM samples (Bayesian forecasting, B1/B2). Metrics used for comparison were median relative prediction error (MDPE) [%], median absolute relative prediction error (MDAPE) [%], and the theoretical target attainment (TTA) [%] for trough concentrations of 2-8 mg/L.

Results: The evaluated models were highly heterogeneous in model structure and predictive performance. MDPE, MDAPE and TTA were improved by inclusion of TDM samples. MDPE ranged from -135.9% to 110.9% (AP), -21.1% to 69.7% (B1) and -15.6% to 54.6% (B2). Overall, the lowest MDAPE was demonstrated by the models of Boak, Fang, and Wu. Several models that resulted in similarly good results are also potentially useful for MIPD.

Conclusion: Predictive performance varied substantially underlining the importance of model evaluation prior to MIPD implementation for linezolid. Using the aforementioned models promises target attainment rates up to 80% in critically ill patients.