Sixty-five patients with hepatic metastases from colorectal primaries were studied in a prospective randomized fashion. The five patients with solitary metastases all had resection of metastases and 50% were randomized to pump therapy. Of the 16 patients with multiple resectable metastases, 7 had pump only and 9 had resection plus pump. Although the difference was not significant, there was a trend of improved survival for the patients with resection plus pump. For the patients with unresectable disease, those patients with positive portal nodes had poor survival matching those patients with extra hepatic metastases. Patients with unresectable disease treated with pump had a 73% therapeutic response rate and a median survival of 22 months. Significant complications included chemical hepatitis and biliary stenosis. The long-term efficacy of continuous hepatic artery infusion versus the hazards of treatment and the financial cost will need further investigation.