One hundred patients were entered on a randomized prospective protocol to evaluate the effectiveness of hepatic resection of single as well as multiple hepatic metastases from colorectal primaries in combination with continuous hepatic artery infusion (CHAI) of fluorodeoxyuridine (FUDR) via the implantable pump (Infusaid, Intermedics Infusaid Inc., Norwood, MA). The eight patients with single metastases were randomized to hepatic resection alone (three patients) or hepatic resection plus CHAI (five patients). The 22 patients with resectable multiple metastases were randomized between receiving CHAI only (12) or CHAI after resection of all metastases (10). Patients who had positive portal lymph nodes (14) were all treated with CHAI. Patients with unresectable metastases (31) were randomized between intravenous 5-fluorouracil or CHAI of FUDR. FUDR was alternately infused every 2 weeks at a dose of 0.1 mg/kg/24 hour escalated to .3 mg/kg/24 hour with heparinized saline as the alternative infusate. The median follow-up of all patients was 20 months. All patients with multiple resectable metastases had at least a partial response (PR) to the CHAI (PR defined as greater than or equal to 50% decrease of the sum of the products of the diameters of the lesions measured on computerized axial tomography scans), and four patients given CHAI only had no metastases in the liver on relaparotomy. Patients with resection and CHAI had a better survival than patients with CHAI only; however, the difference was not significant. Patients with positive portal nodes and CHAI had a lower PR (36%) than patients with unresectable disease treated with CHAI (52%). Patients with positive portal nodes or metastatic disease outside of the liver did significantly worse than patients with unresectable disease treated with CHAI.