Predictive performance of Sawchuk-Zaske and Bayesian dosing methods for tobramycin.

We evaluated the reliability of pharmacokinetic parameter estimations determined by the Sawchuk-Zaske method and by a Bayesian method for predicting steady-state tobramycin concentrations on day 4 and day 10 of therapy in 30 patients treated for gram-negative infections. We also assessed the effect of using only trough tobramycin concentrations on the predictive performance of the Bayesian method. The mean estimation of tobramycin volume of distribution was significantly different for the Bayesian methods compared with the Sawchuk-Zaske methods when the same number of L/kg). Comparing the Bayesian and Sawchuk-Zaske methods when the same number of serum concentrations was available, the Bayesian method overpredicted peak concentrations on day 4 of therapy. When only trough concentration data were used, the Bayesian method significantly overpredicted peak concentrations compared with the Sawchuk-Zaske method on days 4 and 10 of therapy. Each method predicted steady-state trough concentrations without significant differences in bias or accuracy. The Bayesian method may be useful in providing aminoglycoside dosing recommendations.