A Bayesian feedback method of aminoglycoside dosing

We assessed the accuracy of a Bayesian method in providing dosing regimens to achieve desired serum aminoglycoside concentrations. This method calculates individual kinetics based on serum drug concentration data. Performance was analyzed by determining accuracy, bias, correlations of observed to desired serum drug concentrations, and the ability to achieve a target serum drug concentration. We also compared results from the Bayesian method with those resulting from the use of the predictive algorithm portion of the computer program and with routine physician dosing. The Bayesian method resulted in a high correlation coefficient (r = 0.913) between observed and predicted serum concentrations. Analysis of peak aminoglycoside concentrations indicated that the Bayesian method was more accurate and less biased than the predictive algorithm portion of the program or routine physician dosing. A similar trend occurred for trough concentrations. Finally, there were no statistically significant differences between the predicted and observed peak (6.4 ± 1.5 and 5.9 ± µg/ml) and trough (1.2 ± 0.9 and 1.4 ± 0.8 µg/ml) serum aminoglycoside concentrations with the Bayesian dosing method. There were significant differences for peak concentrations with the predictive algorithm portion of the program and for peak and trough concentrations with physician dosing. These data demonstrate the accuracy of the Bayesian dosing method in attaining desired peak and trough serum aminoglycoside concentrations.