Variation in response rates to chemotherapy and survival in patients with hepatic metastases from colorectal carcinoma may be due to patient selection factors. The prognostic importance of 13 factors were analyzed in 112 patients with only hepatic metastases, who were eligible for hepatic artery infusional chemotherapy. When individually analyzed, six factors were found to significantly (less than 0.001) affect survival: the percentage of tumor involvement of the liver, assessed medically or surgically; initial serum albumin and lactic dehydrogenase; initial Karnofsky performance status; and weight loss. Patients with less than or equal to 30% liver involvement had a median survival of 24 months versus 10 months if they had greater than 30% involvement. There was a highly significant agreement between medical and surgical assessment of liver involvement (P = 0.0001). When the variables affecting survival were studied together by multivariable analyses, the most important factor was the medical assessment of liver involvement accomplished by evaluation of radionuclide liver scan and CTT scans. The next two most important factors in the model were the ability of the patient to obtain a tumor response and the presence or absence of weight loss. Only one factor helped predict response to chemotherapy, the type of perfusion seen on a 99Technetium-macroaggregated albumin (MAA) arterial flow scan. Forty-five percent of patients with good perfusion had a partial response while 13% of patients with poor perfusion had a tumor response (P = 0.006). We recommend that future studies, dealing with patients who have hepatic metastases from colorectal carcinoma and are eligible for hepatic arterial infusion, document and stratify for the following factors: the percentage of liver involvement, the presence or absence of weight loss, and the type of perfusion seen on MAA scans.