BACKGROUND: Ten- to 15-year survival results were analyzed from a prospective multi-institutional randomized surgical trial that involved 740 stages I and II melanoma patients with intermediate thickness melanomas (1.0 to 4.0 mm) and compared elective (immediate) lymph node dissection (ELND) with clinical observation of the lymph nodes as well as prognostic factors that independently predict outcomes. METHODS: Eligible patients were stratified according to tumor thickness, anatomical site, and ulceration, and then prerandomized to either ELND or nodal observation. By using Cox stepwise multivariate regression analysis, the independent predictors of outcome were tumor thickness (P < .001), the presence of tumor ulceration (P < .001), trunk site (P = .003), and patient age more than 60 years (P = .01). RESULTS: Overall 10-year survival was not significantly different for patients who received ELND or nodal observation (77% vs. 73%; P = .12). Among the prospectively stratified subgroups of patients, 10-year survival rates favored those patients with ELND, with a 30% reduction in mortality rate for the 543 patients with nonulcerated melanomas (84% vs. 77%; P = .03), a 30% reduction in mortality rate for the 446 patients with tumor thickness of 1.0 to 2.0 mm (86% vs. 80%; P = .03), and a 27% reduction in mortality rate for 385 patients with limb melanomas (84% vs. 78%; P = .05). Of these subgroups, the presence or absence of ulceration should be the key factor for making treatment recommendations with regard to ELND for patients with intermediate thickness melanomas. CONCLUSIONS: These long-term survival rates from patients treated at 77 institutions demonstrate that ulceration and tumor thickness are dominant predictive factors that should be used in the staging of stages I and II melanomas, and confer a survival advantage for these subgroups of prospectively defined melanoma patients.