Predictive performance of a vancomycin-aminoglycoside population model

To evaluate the Wragge-Cooper method of predicting vancomycin serum concentrations utilizing knowledge of aminoglycoside pharmacokinetic parameters in general medicine and intensive care unit populations, and to develop a revised model if necessary.

This study consists of two phases evaluating 50 adults receiving concurrent vancomycin and aminoglycoside therapy. Patients were identified by a retrospective review of medical records. Bayesian analysis of measured serum aminoglycoside and vancomycin concentrations was performed to determine the individualized pharmacokinetic parameters. Phase I of the study tested the predictive performance of a published model incorporating aminoglycoside elimination (Wragge-Cooper) in 25 patients (group 1), and a revised model was developed. Phase II determined the predictive performance of the revised model (revised) and its performance relative to the Wragge-Cooper model and a traditional model incorporating estimated creatinine clearance (traditional) in an additional 25 patients (group 2).

Two tertiary care university teaching hospitals.

The predictive performance of the models was determined by comparing predicted with measured vancomycin serum concentrations. Bias and precision were evaluated by calculating the mean prediction error (ME) and mean absolute error (MAE), respectively. Linear regression was performed to determine relationships between parameters.

The Wragge-Cooper model consistently underpredicts vancomycin serum concentrations in general medicine and intensive care unit populations (ME = -5.18, MAE = 6.63). Relative predictive performance analysis indicates no significant difference in bias or precision between the traditional and Wragge-Cooper models (delta ME 1.17, delta MAE -0.80). Regression analysis of individualized aminoglycoside and vancomycin elimination derived from patients in group 1 reveals the following relationship: vancomycin k10 (1/h) = 0.081 + 1.037ke,amg, r = 0.73. The revised model is significantly less biased and more precise compared with the traditional model (delta ME -4.48; delta MAE 1.22), and is significantly less biased (delta ME 4.29) but no more precise than the Wragge-Cooper model (delta MAE -0.58), using patients from group 2.

The revised model is an accurate method of predicting vancomycin serum concentrations in both general medicine and intensive care unit populations. Use of this model enables individualization of vancomycin dosage in patients receiving concurrent aminoglycoside therapy and minimizes vancomycin serum concentration monitoring.