A retrospective study was made of the surgical treatment of distant metastases from melanoma in 143 patients. The determinants of survival benefit were: (1) successful complete resection of all clinical disease; (2) single versus multiple metastatic site; and (3) anatomic location of disease. Complete resection was achieved with single- and multiple-site lesions in 67% and 31% of the patients, respectively, yielding corresponding 5-year survival rates of 33% and 9%, respectively. Incomplete resection yielded little survival benefit. Soft tissue and extraregional nodal lesions were completely resectable in 70%, whereas pulmonary, extrahepatic abdominal visceral, and osseous lesions were in 40% of the patients, resulting in similar 5-year survival rates of 17% to 22%. The operative mortality rate was 1.4% (2/143). Unsuccessful resection frequently led to very long hospitalization periods for complications and disease, with some patients spending more than 50% of their remaining survival time in the hospital. These observations suggest that, with proper selection of patients, guided by these determinants, surgical resection can offer a reasonable expectation of prolonged survival.