Evaluation of a Bayesian Dosing Calculator for Vancomycin in Critically Ill Geriatric Patients

BACKGROUND: Aging and critical illnesses are associated with physiological changes that alter vancomycin pharmacokinetics. These changes are associated with increased rates of nephrotoxicity and mortality. This study aimed to assess the performance of freely available online dose calculators for predicting vancomycin concentrations in critically ill elderly patients. METHODS: This study included patients aged >65 years in the intensive care unit. Seven population pharmacokinetic models available through the NextDose and TDMx platforms were assessed. Each model was evaluated a priori and a posteriori. RESULTS: A total of 83 patients were included. For the a priori predictions, none of the models met the predefined thresholds; however, the Goti model showed the best performance, with a bias of -4.4% and an imprecision of 31.4%. For the a posteriori predictions, several models, including those of Thomson and Roberts, met the predefined criteria for both bias and imprecision. The Goti model met the imprecision threshold but did not meet the bias threshold, with a bias of 10.3%. CONCLUSIONS: The Goti et al model (available on the TDMx platform) showed the best performance for a priori predictions, followed by the Roberts model; however, neither model met the predefined thresholds. In the a posteriori analysis, the Thomson and Roberts models satisfied the bias and precision criteria, whereas the Goti model slightly exceeded the bias threshold. The limited availability of models tailored to critically ill geriatric patients, coupled with a moderate overall predictive performance, may warrant pharmacokinetic studies in this population to support more individualized dosing.