In 1983, the Northern California Oncology Group (NCOG) instituted a randomized trial of intravenous (IV) versus intraarterial (IA) floxuridine (FUDR) administered via an implantable pump for patients with colorectal cancer metastatic to the liver. The study objectives were to compare the hepatic response rate, time to hepatic progression, and toxicity for the two treatment arms. The study design, which allowed patients failing IV FUDR to crossover to the IA arm, prevents a meaningful comparative analysis of survival. Patients with liver-only metastases (N = 143) were randomized, 76 to the IV arm and 67 to the IA arm, and 115 patients (65 IV, 50 IA) were fully evaluable. Of the 65 patients in the IV arm, 28 crossed over to IA treatment after failing IV FUDR. The dose-limiting toxicity of IV FUDR was diarrhea, whereas biliary toxicity limited both the dose and duration of IA FUDR therapy. Of the first 25 patients treated with IA FUDR at a dose of .3 mg/kg/day, 10 developed radiographically evident biliary strictures, and three developed permanent jaundice. With reduction of the initial IA FUDR dose to .2 mg/kg/day, and adoption of a policy of early dosage reduction, treatment interruption, or termination of therapy for persistent elevations in alkaline phosphatase, only two further cases of serious biliary toxicity occurred. However, 26 of the 50 IA FUDR patients ultimately had therapy terminated because of drug toxicity rather than disease progression. When compared with systemic infusion, infusion into the hepatic artery greatly enhanced the antitumor activity of FUDR against colorectal liver metastases. Although biliary toxicity is the most serious limitation of this form of therapy, biliary stricture and jaundice usually can be averted through careful monitoring of liver enzymes and early dosage reduction.