Population pharmacokinetics of dalbavancin: external validation, model averaging, and implications for precision dosing in prolonged therapy

Dalbavancin is increasingly used off-label for prolonged therapy in gram-positive infections, but optimal repeat-dosing strategies remain undefined, and published population pharmacokinetic (popPK) models lack external validation in real-world cohorts. A popPK model was developed using routine therapeutic drug monitoring (TDM) data from 183 adults (406 concentrations) with gram-positive infections and externally validated in an independent cohort ( n = 30, 92 concentrations). Predictive performance was benchmarked against five published models and a model averaging (MA) approach under three scenarios: a priori prediction and Bayesian forecasting with one or two TDM measurements. Monte Carlo simulations assessed the probability of target attainment ( f AUC ₂₄ /MIC ≥111.1) for repeat-dose regimens (1,500 mg D1/D8 and D1/D15) at MICs 0.125 and 0.250 mg/L. A two-compartment model with eGFR (clearance) and body weight (clearance and volumes) best described the data. In a priori external validation, it achieved the lowest rRMSE (49.8%) among published models. MA with two TDM measurements yielded optimal performance (rRMSE 26.9%), closely followed by our model (31.4%). Simulations showed that empirical regimens maintained target attainment of ≥90% for 3–7 weeks, heavily dependent on PD targets and patient characteristics. A third 1,500 mg dose, timed to target decline, restored prolonged coverage across all strata. This externally validated popPK model, complemented by MA, provides robust pharmacokinetic foundations for precision dosing in gram-positive infections requiring prolonged therapy. Based on these findings, we propose an investigational sequential precision dosing workflow integrating machine learning with Bayesian forecasting; prospective validation is required before clinical implementation.