A retrospective study was performed to define patient selection, safety, and efficacy of hepatic resection for colorectal metastases. The recently proposed preoperative clinical risk score (CRS) for selection of patients for surgery was also assessed. In all, 146 consecutive hepatic resections in 137 patients operated in the period between 1977 and 1999 were studied. Of these patients, 113 were classified into five CRS groups. Perioperative mortality was 1.4% (2 patients; no death in 120 patients operated after 1985) and morbidity was 38%. Five-year actuarial survival (perioperative mortality included) was 29% (median 37 months), and actual 5-year survival was 25% (17/69 patients). Patients operated after 1995 lived longer than those operated before 1995. Multiple regression analyses identified preoperative carcinoembryonic antigen CEA <100 mg/L, nodal status at resection of primary tumor, and R0 vs. R1/R2 resection as prognostic parameters. CRS grouping had prognostic importance. The relative risk (hazard rate) of tumor recurrence in patients with CRS 3-4 was 2.1, compared to that of patients with CRS 0-2. Five-year actuarial survival in the two groups was 12% and 40%, respectively. Fourteen of 15 long-term survivors (>5 years) classified by the CRS system had CRS of 2 or less. Resection for colorectal liver metastases is safe, and long-term survival rates are acceptable. CRS predicts patient outcome, but the clinical role in patient selection will have to be defined in prospective studies.