The large interpatient variability in gentamicin half-life and distribution volume in children and the lack of accurate methods of predicting these parameters necessitate individualization of doses based on measured serum concentrations. For practical reasons, a predose nadir and postdose peak (obtained before and after the same dose) have been used to make dose adjustments. To assess the accuracy of this approach, pharmacokinetic parameters estimated from pre- and postdose concentrations measured in 21 children were used to adjust doses and simulate new steady-state peak and nadir concentrations. A total of 53 dose adjustments was evaluated in these 21 patients. The accuracy of this method was compared to results obtained in a control group of 40 children who had steady-state serum concentrations simulated from kinetic parameters calculated from three serial serum concentrations (0.5, 3, and 6 hr after a dose). A one-compartment kinetic model was used for all dosage calculations and for simulation of serum concentrations at the new dosages. All patients in both groups were between 2 and 15 years of age, had stable normal renal function, and were being treated for bacterial infections. The accuracy of both methods was assessed by comparing concentrations predicted for the new dosage to concentrations measured at steady state. The line of best fit for the rectilinear plot of measured (x) versus predicted (y) gentamicin serum concentrations for the control group was described by the equation:y= 0.922x+ 0.23,r= 0.901. The same data for the study group were described by the equation:y= 0.873x+ 0.59,r= 0.813. Simultaneous testing of the slopes and intercepts (jointF-test) indicated that neither of these lines was significantly different from the line of reference. These data show that the use of predose nadir and postdose peak serum concentrations to make dose adjustments results in significantly greater error than using three serial samples (r2= 0.66 vs. 0.81,p< 0.02). Although the error with the former method is comparable to that encountered with nomograms and equations commonly used to estimate gentamicin kinetic parameters in adults, and offers an alternative in children, this method is only recommended as an adjunct with more traditional pharmacokinetic approaches to individualizing therapy.