Very few patients with liver metastases from colorectal cancer can be cured. We have investigated whether a treatment to slow the growth of liver metastases, hepatic-artery infusion of floxuridine, improves palliation in this setting. In a randomised study of 100 patients, we compared quality of life and survival in patients who received hepatic-artery infusion of floxuridine and in those who received conventional symptom palliation. 95% of control patient survival time was spent with normal quality-of-life scores, which suggests that the aim of treatment should be to prolong normal-quality survival rather than merely to sustain quality of life. There was a significant prolongation (p = 0.03) in overall survival in floxuridine-treated patients compared with controls (median 405 vs 226 days). There were similar significant prolongations in normal-quality (ie, normal symptom scores) survival for physical symptoms (p = 0.04), anxiety (p = 0.04), and depression (p = 0.04). This survival benefit was associated with significant reductions in metastasis size on computed tomography (p = 0.001) and in serum carcinoembryonic antigen concentration (p = 0.006) in floxuridine-treated patients. There was no evidence of treatment-related hepatotoxicity as assessed by serum aspartate aminotransferase and bilirubin measurements. This is the first demonstration that survival can be prolonged with normal quality of life in patients with colorectal liver metastases. We conclude that hepatic-artery floxuridine infusion can be recommended for suitable patients.