One Model to Dose Them All: A Population Pharmacokinetic Model for Multiple Aminoglycosides Across the Human Lifespan

Aminoglycosides require therapeutic drug monitoring due to substantial pharmacokinetic variability across individuals. While population pharmacokinetic (popPK) models have been developed for individual drugs across age groups, and cross-drug maturation models have been explored, it is unclear whether pooling data across aminoglycosides improves predictive performance. We developed and evaluated a unified multi-aminoglycoside popPK model spanning neonates, children, and adults. De-identified data from 5659 amikacin, gentamicin, and tobramycin patients at 154 US institutions were split into training and testing datasets. Predictive performance in the test dataset was evaluated using mean percent error, normalized root mean square error (nRMSE), and clinical accuracy (trough concentrations correctly classified as below or above 1 mg/L and non-troughs within 20% of observation), and compared with selected published popPK models. To assess generalizability, the final model was refit using single-drug datasets and leave-one-subgroup-out datasets. The unified model achieved the lowest nRMSE across age groups, with absolute reductions of 0.6%-24.8%. Clinical accuracy was highest for the unified model, with a posteriori accuracy in pediatric patients of 56.4% versus 48% in the best published model. Leave-one-out refits yielded similar estimates but reduced precision, while single-drug models demonstrated substantially reduced precision, and one such model was not identifiable. This unified model improved predictive performance across drugs and age groups and demonstrated robustness to exclusion of individual drug-age group combinations. These findings support use of a unified aminoglycoside model for Bayesian model-informed precision dosing and demonstrate the advantages of pooling drugs within a class and across developmental stages.