A multifactorial analysis was used to identify the dominant prognostic variables affecting survival from a computerized data base of 339 melanoma patients treated at this institution during the past 17 years. Five of the 13 parameters examined simultaneously were found to independently influence five year survival rates: 1) pathological stage (I vs II, p = 0.0014), 2) lesion ulceration (present vs absent, p = 0.006), 3) surgical treatment (wide excision vs wide excision plus lymphadenectomy, p = 0.024), 4) melanoma thickness (p = 0.032), and 5) location (upper extremity vs lower extremity vs trunk vs head and neck, p = 0.038). Additional factors considered that had either indirect or no influence on survival rates were clinical stage of disease, age, sex, level of invasion, pigmentation, lymphocyte infiltration, growth pattern, and regression. Most of these latter variables derived their prognostic value from correlation with melanoma thickness, except sex which correlated with location (extremity lesions were more frequent on females, trunk lesions on males). This statistical analysis enabled us to derive a mathematical equation for predicting an individual patient's probability of five year survival. Three categories of risk were delineated by measuring tumor thickness (Breslow microstaging) in Stage I patients: 1) thin melanomas (<0.76 mm) were associated with localized disease and a 100% cure rate: 2) intermediate thickness melanomas (0.76-4.00 mm) had an increasing risk (up to 80%) of harboring regional and/or distant metastases and 3) thick melanomas (>/=4.00 mm) had a 80% risk of occult distant metastases at the time of initial presentation. The level of invasion (Clark's microstaging) correlated with survival, but was less predictive than measuring tumor thickness. Within each of Clark's Level II, III and IV groups, there were gradations of thickness with statistically different survival rates. Both microstaging methods (Breslow and Clark) were less predictive factors in patients with lymph node or distant metastases. Clinical trials evaluating alternative surgical treatments or adjunctive therapy modalities for melanoma patients should incorporate these parameters into their assessment, especially in Stage I (localized) disease where tumor thickness and the anatomical site of the primary melanoma are dominant prognostic factors.