Predictability of Vancomycin Concentrations from Nine Approaches for Estimating Pharmacokinetic Parameters

Abstract

Objectives: Doses of vancomycin are frequently estimated using various predictor formulas aiming for trough concentrations between 5 and 15 mg/L and peak concentrations between 25 and 40 mg/L.1 There is however, some controversy about the relationship between vancomycin concentrations and therapeutic response. This project compares the ability of several methods to estimate serum concentrations of vancomycin. Methods: This project was a retrospective look at 243 patient records, the patients were given vancomycin and later had at least one concentration measured. Data collected while the patient was being treated was used in the nine predictor models to determine which model would best predict actual concentrations. The methods compared were Moellering, Matzke, Lake-Peterson, Rodvold, Abbott, Birt, Burton, Ambrose, and Bauer. Results: There were 188 patients included in the analysis, 97 males and 91 women. The method with the least bias (+ 1.0) was the Rodvold method using the actual body weight. None of the models were very precise, with most around 10 (high of 12.83, and a low of 9.35). The r- values for all the models were also low, none of the models had an r- value greater than 0.5. The Lake-Peterson method predicted within 20% and 50% the most often; the Ambrose method the least often within 50%, and both Ambrose and Bauer the least often within 20%. The Lake-Peterson method predicted concentrations within plus or minus 2.5 and 5.0 mg/L of measured concentrations most frequently. The Ambrose method predicted concentrations within plus or minus 2.5 mg/L of measured the least often; Burton and Rodvold the least often within 5.0 mg/L of measured. Implications: With the best model only accurate (defined as ± 20%) less than 25 percent of the time, there is too much error to make a good decision on dose and interval without the feedback of measured serum concentrations. The models may be a good starting point as which dose and interval to choose, but they are not a substitute for measuring steady state concentrations.