This first metanalysis of melanoma from treatment centers worldwide consisted of 15,798 patients with localized melanoma (stages I and II) and 2,116 stage II melanoma patients with nodal metastases. Comparisons of dominant prognostic variables showed consistent results from center to center, despite the heterogeneity of the patient population. Six of eight centers that performed a multivariate analysis ranked ulceration among the first three most dominant prognostic factors. Men had a higher proportion of ulcerated lesions than did women. There was a positive correlation between ulceration and thickness. Patients with melanoma of the scalp had a worse prognosis than did those with lesions of the face and neck; those with melanomas on the hands had a significantly worse prognosis than did those with lesions on the arms or legs. In this study, women had a statistically significant survival advantage over men. Their melanomas arose in more favorable sites, were thinner, and less ulcerative and had a lower stage of disease at presentation. Stage 111 melanomas were more common in males, thicker, and more ulcerated and had a nodular growth pattern. Patients with clinically occult nodal metastases detected by pathological examination and those with a single metastatic node fared the best. Five of six centers identified the number of metastatic nodes to be the most significant prognostic factor. Distant metastases (stage IV were analysed at only two centers, which found that the number and site of metastases appeared to be the dominant prognostic features of stage IV melanoma. When all factors were analyzed in a Cox regression analysis, the dominant factors for stage IV melanoma patients were (I) the number of metastatic sites, and (2) the remission duration. There were no histologic criteria of the primary melanomas that predicted the patient's clinical course once distant metastases had developed.