PURPOSE This study was undertaken to determine the indications for and value of liver resection for metastases from colorectal cancer. METHODS From 1978 through 1991, 66 patients were operated on for liver metastases from colorectal cancer. All patients had had a curative resection of their colorectal cancer. Forty resections of the liver were major anatomic resections. RESULTS Five patients died in the postoperative period. All resections were intended to be curative, but in 16 of the patients the resection became noncurative. None of these patients lived more than two years after liver resection. Fifty patients with a curative resection had a three-year survival rate of 36 percent, postoperative death included. Recurrence in the liver was observed in 30 patients (60 percent) from 3 to 33 (median, 11) months after the liver resection. Four patients had repeated resections performed. Two of them are alive without recurrences 34 and 60 months after the first liver resection, respectively. The difference in survival between curative and noncurative liver resection was highly significant (P = 0.01). CONCLUSIONS Sex, age, Dukes stage of primary colorectal cancer, synchronous or metachronous appearance of metastases, or number of metastases could not predict long-term prognosis. The only factors of predictive value were tumor size less than 4 cm in diameter, a free resection margin, and no extrahepatic tumor. If it is possible to do a curative resection, there should be few contraindications against liver surgery as it is the only treatment that can demonstrate long-term survival for approximately one-third of the patients, and it is the only possibility of a cure.