A Phase I trial of thiotepa (TT) administered as an i.v. bolus was performed in 19 children with refractory malignancies. The starting dose was 25 mg/m2 with escalations to 50, 65, and 75 mg/m2. Seven additional patients were treated with 8-h infusions at 50 or 65 mg/m2. The maximum tolerated bolus dose was 65 mg/m2. Reversible myelosuppression was the dose-limiting toxicity. The plasma and cerebrospinal fluid (CSF) pharmacokinetic parameters of TT and its major active metabolite tepa (TP) were also evaluated. When the bolus or infusion methods of TT administration were compared, there was little difference observed in any pharmacokinetic parameter for either TT or TP. The plasma disappearance of TT was rapid and biphasic with half-lives of 0.14 to 0.32 and 1.34 to 2.0 h. Dose-dependent pharmacokinetics was demonstrated by steadily declining plasma clearance with increasing TT dose. Clearance values declined from 28.6 liters/m2/h at the 25-mg/m2 dose to 11.9 liters/m2/h at the 75-mg/m2 dose. The half-life of TP was longer than that of TT and ranged between 4.3 and 5.6 h. There was evidence of the saturation of TP production. TT and TP both exhibited excellent penetration into the CSF, producing lumbar and ventricular concentrations which were nearly identical to simultaneous plasma concentrations. In one patient with a Rickham reservoir, the CSF:plasma area under the (concentration x time) curve ratios for TT and TP were 1.01 and 0.95, respectively. The above data indicate that TT can be safely administered to pediatric patients at doses higher than conventionally used. The favorable CSF penetration of TT and TP suggests that Phase II studies of TT be considered in patients with central nervous system tumors.