Pharmacokinetic studies of etoposide administered at 100-200 mg/m2 to 33 children are described. Twenty-seven studies were performed in children aged < 10 years. Repeat studies were performed in 11 patients. Median pharmacokinetic parameters were as follows: plasma clearance, 26 ml/min/m2; volume of distribution, 4.9 liters/m2; area under the etoposide plasma concentration-time curve (AUC), 3.9 mg/ml x min per 100 mg/m2. Interindividual variability in pharmacokinetic parameters was large (coefficient of variation (CV) = 30, 28, and 27%, respectively) in comparison with intraindividual variability (CV = 12, 14, and 12%, respectively). Variability in AUC was much greater in those patients treated with 150-200 mg/m2 etoposide than with 100 mg/m2 (CV, 35 versus 13%) and was related to variability in renal function and prior exposure to cisplatin. Data from the first 20 studies were used to develop pharmacokinetic monitoring equations which were validated in a further 13 patients. The most accurate equation relies upon the elimination constant of 51Cr-EDTA and a single blood specimen taken at the end of the etoposide infusion. [formula: see text] where K = 51Cr-EDTA elimination rate constant. This equation showed no significant bias, and the predictive error was small with respect to AUC calculated according to a two-compartment model. Predictive error did not increase with increasing AUC, whereas a marked increase in predictive error was seen for dosing according to body surface area. Dosing according to body surface area alone led to marked over- or underexposure to etoposide in 8 patients. Pharmacokinetic monitoring using the equation described would have identified these patients and permitted dose modification. This approach provides an accurate means of monitoring etoposide AUC for administration times of 1-4 h without the need for detailed pharmacokinetic sampling. It will allow a significant reduction in the variability of exposure seen with surface area-based dosing.