Limited sampling models are able to estimate the area under the concentration-time curve (AUC) from plasma concentrations measured at only a few time points. The purpose of this study was to establish a model estimating etoposide AUC independently of specific chemotherapy protocols, underlying malignancies, concomitant diseases and age. Pharmacokinetic parameters were measured in 30 patients treated with polychemotherapy including etoposide (80-150 mg/m2). Etoposide analysis was performed by thin layer chromatography and consecutive quantitative sample detection by 252Cf-plasma desorption mass spectrometry. Data from the first 15 patients formed the training set. Based on the training data, five different models were generated, with the multiple regression coefficient r ranging from 0.91 to 0.96. The following model was selected as "most accurate": AUC = 343 (min)C4h(micrograms/ml) + 650(min)C8h(micrograms/ml) + 1252 (min micrograms/mol), where C4h is the plasma concentration of etoposide at 4 h after the end of infusion and C8h at 8 h. This model was validated on the test set, comprising the data of the remaining 15 patients. The mean predictive error (MPE) was -0.2% and the root mean square predictive error (RMSE) was 4.7%. When used for a large number of patients, this practicable and simple model is an instrument for use in prospective studies, to measure a correlation between drug dosage and efficacy or toxicity of the drug.