Comparison of four methods of predicting serum gentamicin concentrations in adult patients with impaired renal function.

Four methods of predicting serum gentamicin concentrations (SGCs) in adult patients with stable impaired renal function were compared. Serum samples obtained from 17 patients receiving intravenous gentamicin therapy were assayed for gentamicin using a radioimmunoassay. Kinetic variables derived using four different methods were used to predict peak and trough concentrations. The Sawchuk-Zaske method uses individualized apparent volume of distribution (Vapp) and half-life (t1/2) derived from at least three SGCs following a given dose. The fitted method assumes Vapp equals 0.28 liters/kg lean body weight (LBW) and a t1/2 derived by a fitting technique using only one SGC. Both the Tozer and Hull methods assume Vapp equals 0.28 liters/kg LBW, and t1/2 is estimated from equations that take into account the patient's renal function (creatinine clearance). The predicted and measured SGCs for each method were compared using linear regression analysis. The prediction errors, defined as the predicted minus the measured SGC, were compared for the four methods. The correlation coefficients between the measured and predicted SGCs were significant at p less than 0.05 only for the Sawchuk-Zaske and fitted methods. For the Sawchuk-Zaske method, peak and trough r values were 0.73 and 0.91, respectively. For the fitted method, peak and trough r values were 0.53 and 0.71, respectively. No significant differences in the mean prediction errors were observed, except for the Sawchuk-Zaske method, which had significantly less mean prediction error than the Hull method. The Sawchuk-Zaske method was the most reliable and accurate pharmacokinetic technique. However, because it is more costly and less convenient than the other three methods, both the accuracy and cost should be considered when selecting a method for determining serum gentamicin concentrations for a particular patient.