The toxic effects and tissue uptake of both cisplatin and oxaliplatin--[(1R, 2R)-1,2-cyclohexanediamine-N,N'] [oxalato(2-)-O,O']platinum--were previously shown to vary similarly according to dosing time in mice. A 4-hour infusion of cisplatin resulted in fewer side effects and allowed administration of higher doses at 16 hours than at 4 hours in patients with cancer. We hypothesized that the continuous venous infusion of oxaliplatin for 5 days would be less toxic and would deliver a higher dose to the patient if the drug were infused at a circadian rhythm-modulated rate (peak at 16 hr; schedule B) rather than at a constant rate (schedule A). We tested this hypothesis in a randomized phase I trial. We escalated the dose of oxaliplatin to the patient by 25 mg/m2 per course. Courses were repeated every 3 weeks. An external, multichannel, programmable-in-time pump was used for the infusions. Toxicity was assessable for 94 courses in 23 patients (12 patients with breast carcinoma, nine with hepatocellular carcinoma, and two with cholangiocarcinoma). The incidence of neutropenia of World Health Organization grades II-IV and the incidence of distal paresthesias were 10 or more times higher (P less than .05) with schedule A than with schedule B. In addition, vomiting was 55% higher (P = .15) with schedule A than with schedule B. Furthermore, with schedule B, the mean dose of oxaliplatin (P less than .001) and its maximum tolerated dose (P = .06) could be increased by 15% over those doses with schedule A. An objective response was achieved in two of the 12 patients with previously treated breast cancer. We recommend that the dose of oxaliplatin for phase II trials be 175 mg/m2, delivered according to the circadian rhythm-modulated rate.