Piritrexim, an orally administered, lipid-soluble antifolate, was evaluated in a multi-institutional phase I trial in children. The starting dose was 10 mg/m2/dose administered every 8 h daily for 5 days for 3 consecutive weeks, with dose escalations in increments of 5 mg/m2/dose. Eighteen patients (16 with metastatic sarcoma, 1 with acute lymphoblastic leukemia, and 1 with a brainstem glioma), 3.5-20 years of age, with malignancy refractory to therapy, were entered into the study. The dose-limiting toxicities (DLTs), which were myelosuppression and mucositis, occurred in 4 of 4 patients treated at the 25-mg/m2/dose level but in none of the patients treated at the 15- and 20-mg/m2/dose levels. The recommended dose for phase II trials is 20 mg/m2/dose. Pharmacokinetic monitoring was performed in 15 of the 18 children. The area under the concentration-time curve (AUC) was linearly related to the dose administered. Piritrexim was rapidly absorbed, with the median time to peak level occurring 1.5 h after an oral dose. The terminal half-life of piritrexim ranged from 1.5 to 4.5 h. A limited sampling strategy developed earlier, capable of predicting the AUC based on the plasma concentrations at 3 and 6 h after an oral dose, was prospectively tested in this trial and proved to be highly predictive of the AUC (r = 0.98, P = 0.0001). Pharmacodynamic-pharmacokinetic correlations were obtained after combining data from this and the prior phase I pediatric trial. Trough plasma piritrexim concentration strongly correlated with DLT (P = 0.0016). A trough plasma piritrexim concentration greater than 0.5 microM appeared to be predictive of toxicity. Eleven of 15 patients with trough concentrations exceeding this threshold experienced DLTs. Therapeutic drug monitoring may thus play an important role in adjusting the dose and schedule of piritrexim in future trials.