BACKGROUND Although over 7,000 people die from malignant melanoma each year, there are limited prognostic data for patients with metastatic disease. A retrospective analysis was undertaken to identify variables that accurately predict outcome and to determine if the survival rate of patients with melanoma treated for distant metastases (American Joint Committee on Cancer [AJCC] stage IV disease) at the authors' institution changed between 1971 and 1993. METHODS Data for 1,521 patients with AJCC stage IV melanoma treated by the staff of the John Wayne Cancer Institute were reviewed, and a univariate and multivariate survival analysis against ten clinical and pathological variables was performed using the Cox proportional hazard regression model. RESULTS The median survival time of the 1,521 patients was 7.5 months; the estimated five-year survival rate was 6 percent. Three independent variables predicted survival: initial site of metastases (p < 0.0001); disease-free interval before distant metastases (p = 0.0001); and stage of disease preceding distant metastases (p = 0.0001). Patients could be divided into three distinct prognostic groups based on the initial site of metastases: cutaneous, nodal, or gastrointestinal metastases (median survival of 12.5 months; estimated five-year survival rate 14 percent); pulmonary metastases (median survival of 8.3 months; estimated five-year survival rate 4 percent); and metastases to the liver, brain, or bone (median survival of 4.4 months; estimated five-year survival rate 3 percent). There was no significant change in the survival rate of patients with AJCC stage IV melanoma during the 22-year review period. CONCLUSIONS Despite new treatment options, the survival rate of patients with metastatic melanoma has not changed significantly over the last 22 years; their prognosis remains dismal. The three prognostic variables identified in this study should be considered in the design of future clinical trials.