Sixty-four patients were entered into a randomized trial that evaluated intra-arterial (I.A.) versus intravenous (I.V.) 5-fluorodeoxyuridine (FUDR) for colorectal liver metastases. There was a significant improved response rate for I.A. (62%) compared with I.V. (17%) treatment (p less than 0.003). However, the improved response rate for patients in whom I.A. therapy was used did not translate to a significantly improved survival rate. The 2-year actuarial survival rates for the groups for which I.A. and I.V. therapy was used were 22% and 15% respectively, with the survival curves not differing significantly (p = 0.27). These results may have been due to the inclusion of patients with tumor in draining hepatic lymph nodes. The presence of tumor in hepatic lymph nodes was associated with a poorer prognosis. Analysis of a subgroup of patients with negative hepatic lymph nodes suggested an improved actuarial survival rate in patients for whom I.A. versus I.V. therapy was used (p less than 0.03). The toxicity of I.A. FUDR was considerable, and side effects included chemical hepatitis (79%), biliary sclerosis (21%), peptic ulcers (17%), and gastritis/duodenitis (21%). The only major effect of toxicity of I.V. FUDR was severe diarrhea (59%). Regional I.A. FUDR allowed more drug delivery to liver tumors, which resulted in increased tumor responses when compared with use of systemic therapy. However, the small gain in survival seen in a select subgroup of patients with negative hepatic nodes appeared to be offset by the toxicity of I.A. FUDR.