Evaluation of vancomycin Bayesian dose-optimizing software in a neonatal intensive care unit population

Abstract Purpose Infectious Diseases Society of America 2020 guidelines recommend Bayesian-guided area under the curve (AUC) monitoring for vancomycin dosing in pediatric patients. Our institution’s neonatal intensive care unit (NICU) utilizes single-level kinetics to monitor AUC using 2 pharmacokinetic models validated in neonates and infants. Anecdotally, preterm infants corrected to a higher postmenstrual age (PMA) may be better suited to the pediatric dosing model (PDM) versus the neonatal dosing model (NDM). This study aims to identify which model is best suited to the NICU population based on factors including gestational age (GA), postnatal age (PNA), and PMA. Methods A retrospective review including 64 NICU patients who received vancomycin was conducted. Measured vancomycin trough concentrations were compared to Bayesian model predictions using both the NDM and PDM. Model performance was evaluated using median absolute prediction error and stratified by GA, PNA, and PMA. Predicted AUC and pharmacokinetic parameters were also compared between models. Results The PDM had greater predictive accuracy and less variability in trough concentrations across GA and PNA groups when compared to the NDM. Stratified analysis showed that the PDM was most accurate in term infants and those with a PMA of <29 or >52 weeks. The NDM consistently overpredicted trough levels in extremely preterm and early postnatal infants, leading to larger AUC discrepancies. Alternative models may improve predictive performance, particularly in high-risk neonatal subgroups. Conclusion The PDM had superior predictive performance when compared to the NDM. Use of the PDM may improve vancomycin dosing accuracy in NICU patients, especially those with more advanced PMA.