Background Formulas for forecasting gentamicin concentrations have been proposed, such as the Sawchuk-Zaske method. These formulas use weight and creatinine clearance to determine the volume of distribution and clearance.

Purpose The aim of this study was to compare the power of this formula to a Bayesian method of forecasting pharmacokinetic parameters.

Materials and methods The files of 20 patients treated with gentamicin were used. The Sawchuk-Zaske formula was used to determine the volume of distribution and clearance in these patients. These parameters were introduced into a unicompartmental pharmacokinetic model to forecast the concentrations in each patient. Software for Bayesian forecasting of individual pharmacokinetic parameters (USC*Pack) was used to predict serum concentrations. Forecasts from the two methods were compared to concentrations actually measured in each patient.

Results There was a men/women ratio of 14/6, and patients included were a mean 84±5.9 years old, weighed 68.3 ±18.4 kg with a creatininemia of 95.2 ±3.2 µmol/l. A total of 102 serum concentrations were estimated or a mean 5.1±2.9 concentrations per patient. The Sawchuk-Zaske formula predicted concentrations with a bias of -1.31 mg/l, and a precision of 9.37 mg²/l². Prediction after individual estimation of pharmacokinetic parameters included a bias of -0.27 mg/l (p<4.10^-6) and a precision of 2.28 mg²/l² (p<4.10^-6). Considering only the first measured concentration (early treatment), the Sawchuk-Zaske formula presents a systematic error of -1.24 mg/l versus -0.08 mg/l for the Bayesian estimation method (p=0.001).The integration of additional data made during follow-up treatment can increase the precision of the method.

Conclusions The formula tested seems to forecast gentamicin concentrations less accurately than a Bayesian method of individual estimation of pharmacokinetic parameters, which was significantly more accurate. The formulas do not incorporate information obtained from patients after the first results. The limitations of the Sawchuk-Zaske formula may be due to an underestimation of non-renal gentamicin elimination. Bayesian approaches seem best suited to support older patients.