New Continuous Learning PK Model Outperforms Published PopPK Models for Gentamicin

  • Published September 5, 2025

Background

Gentamicin is an aminoglycoside antibiotic used to treat gram-negative infections. Its clinical use is limited by a narrow therapeutic window and the risk of ototoxicity and nephrotoxicity. Model-informed precision dosing (MIPD) can improve safety and efficacy, but requires a sufficiently predictive population pharmacokinetic (popPK) model. InsightRX Nova currently supports two published models for adult gentamicin patients (Xuan 2004 and Smit 2019). We evaluated their performance on real-world clinical data and developed a new model to improve prediction performance. 

genta_adults_onepager_figure

Figure 1. Predictive performance of published and new continuous learning model for adult patients dosed with gentamicin, evaluated by (left) accuracy, (centre) mean percent error (MPE), and (right) normalized root mean square error (nRMSE). Closed circles (vertical bars) represent the median (5th to 95th percentile) of 1000 bootstraps. Statistical significance was determined by presence of overlapping confidence intervals: ▲ represents the model with best performance; ▽ represents models not statistically different from the best model. The green horizontal line indicates ideal model performance.

Results

Data Collection

After removal of anomalous and outlier data, the dataset contained 3,409 gentamicin concentrations across 1,653 treatment courses from 16 hospital systems.

Model Fit

The new continuous learning model, designated the Tong model, incorporates fat-free mass, sex, and estimated glomerular filtration rate (eGFR) calculated using the 2021 CKD-EPI equation.

Predictive Performance Comparison

The Tong model demonstrated the best overall predictive performance across accuracy, mean percentage error (MPE), and normalized root mean square error (nRMSE). For a posteriori MPE, it performed comparably to the model by Xuan et al. (Figure 1). Improvements were most pronounced for peak concentration predictions, though enhanced predictive performance was observed across all sample timings.

Recommendation for Practice

InsightRX recommends the newly-fit Tong model for gentamicin dosing in adult patients.

Want to Explore More?

Read on for the full analysis below, and learn how InsightRX can support precision dosing at your organization.