Thirty patients in whom an implantable pump was used for hepatic arterial perfusion of chemotherapeutic agents to treat hepatic malignancy were evaluated. Three patients with primary hepatocellular carcinoma demonstrated less than 50% reduction in tumor size. One patient with metastatic gastric carcinoma and one patient with metastatic islet cell carcinoma also showed a decrease of less than 50% in the size of the tumor mass. Among the 25 patients with metastatic carcinoma from the colon or rectum, 23 had elevated carcinoembryonic antigen (CEA) levels before surgery and 75% of these revealed a reduction of at least 50%. Only three of 20 patients followed with sequential imaging studies showed a 50% decrease in the size of the tumor mass. No increase in the duration of survival could be defined when the treated patients were compared with 13 patients who had not received chemotherapy. The pump functioned well in all patients, but 77% showed signs of toxicity. This experience, coupled with a review of the literature, suggests that the procedure should be regarded as experimental and should not be applied liberally until a definite benefit can be demonstrated.