A prospective randomised trial was conducted in critically ill patients to evaluate a computer aided pharmacokinetic method of aminoglycoside dose prediction based on 3 measured plasma concentrations following the loading dose. The ability of this method to achieve therapeutic plasma aminoglycoside concentrations early in the course of treatment was compared with that of a nomogram approach based on creatinine clearance estimated using the formula of Cock-roft and Gault. Ninety-two percent of patients in the computer group achieved peak plasma concentrations within the optimum range of 6–10 mg/l at 48–72 h compared with 21% of control group patients (p=0.0009). The mean peak plasma concentration of 7.45 mg/l at 48–72 h in the computer group was closer to the target concentration of 8 mg/l than was the 5.14 mg/l in the control group (p=0.0004). There was no significant difference between the groups in measured indices of renal function, both groups showing an improvement in mean estimated creatinine clearance from the beginning to the end of the course of treatment. Dosing based on individualised pharmacokinetic data is therefore a more reliable method of achieving therapeutic blood concentrations early in the course of treatment than is nomogram based dosing. Other studies suggest that this should be associated with a reduction in mortality in severe infections.