Twenty-two patients with advanced solid tumors were treated with a quinazoline folate antagonist, trimetrexate, to determine the toxicity spectrum, the maximal tolerated dose, and the pharmacokinetics of the drug. Negligible toxicity was seen with single doses of 10-70 mg/m2 given as a 1-h infusion. Single doses of 120 mg/m2 infused over 1 h caused moderate to grade 4 toxicity in five of nine patients treated. Two patients who had no toxicity at this level were escalated to a dose of 213 mg/m2 with mild to moderate toxicity. The primary dose-limiting toxicity was myelosuppression. Moderate transaminase elevations, rash, anorexia, nausea and vomiting, and mucositis were occasionally seen. Although there was variation in dose tolerance to this drug, with selected patients able to tolerate higher doses, we consider 120 mg/m2 every 2 weeks to be the maximal tolerated dose, and the recommended Phase II starting dose. Trimetrexate plasma concentration-time curves were best described as biphasic (N = 9) or triphasic (N = 5) in form. The half-life of the terminal elimination-phase was 16.4 h. The mean residence time was 17.8 h. The volume of distribution of the plasma compartment and the volume of distribution at steady-state were 0.17 and 0.62 liter/kg, respectively. Plasma clearance was 53 ml/min. Plasma concentrations as determined by dihydrofolate reductase enzyme inhibition assay and high-performance liquid chromatography were initially identical, but diverged at later times. Divergences were seen also in urinary recovery as determined by the two methods. Both results suggest the appearance of metabolite(s) of trimetrexate which can inhibit dihydrofolate reductase. Measurable objective solid tumor responses were not seen in this Phase I study, although three patients with colon cancer had stable disease lasting 18, 26, and 26 weeks, respectively.