PURPOSE: This study was designed to determine if hepatic arterial therapy with floxuridine (F), mitomycin, and carmustine (BCNU) (FMB) is superior to hepatic arterial therapy with F alone in previously treated patients with hepatic metastases from colorectal cancer. PATIENTS AND METHODS: Ninety-five patients were randomized to intrahepatic FMB versus intrahepatic F. All patients had tumor progression after systemic chemotherapy (either therapeutic or adjuvant). RESULTS: There was no significant difference in response rate (47% FMB v 33% F; P = .17). Median survival was similar in the two groups, 19.1 months for the FMB group compared with 14.0 months for the F group (P = .23). The overall median survival was 16.8 months. In patients who received prior adjuvant therapy, there was no difference between the two groups, but response rate was high in both (50% FMB v 62% F). The response rate for all patients who had received only prior adjuvant therapy versus all those who had received prior therapy for metastatic disease was 57% and 35%, respectively (P = .066). In the subset of patients whose disease had progressed with prior systemic chemotherapy, the response rate to FMB was greater than that to F (47% v 23%; P = .035). CONCLUSION: The overall partial response rate of 39% and the overall survival of 16.8 months from initiation of intrahepatitis therapy show that hepatic arterial therapy is a reasonable treatment option for patients whose tumor does not respond to systemic therapy or whose disease progresses after adjuvant therapy for colorectal cancer.