The nitrosourea, fotemustine, was given intravenously in 1 h constant-rate infusion to 66 patients in a multicentric study to assess both fotemustine pharmacokinetic behaviour and the pharmacokinetic-pharmacodynamic relationships. Depending on the tumour type treated, two administration and sampling protocols were used: 100 mg/m2/week as a conventional dose (six samples, 44 patients) and 300-500 mg/m2/day as a high dose (10 samples, 22 patients). The 91 time-concentration curves were best described by either a one-(55) or a two-compartment (36) model, and their mean clearance values did not differ significantly (85.3 +/- 6.5 and 101.3 +/- 9.5 l/h, respectively, P = 0.1727). Fotemustine pharmacokinetics were not influenced by repeated treatment (time-independence) nor by dose level (dose-independence). The pharmacodynamic effect observed on white blood cell count was expressed by a logit regression model involving the area under the curve mainly and the total administered dose. White blood cell toxicity could be predicted as a function of the dose for a given patient with a known fotemustine clearance value.