Stage I melanoma encompasses an extraordinary diversity of biologic behavior. In such a setting where numerous parameters appear to influence survival, a multifactorial analysis using Cox's regression model is a valuable statistical model. Using a computerized data base of 394 clinical stage I melanoma patients treated at this institution during the past 20 years, a multifactorial analysis was used to compare the relative prognostic strength of 11 parameters. Two pathological factors (tumor thickness and ulceration) and two clinial factors (initial surgical treatment and anatomic location) were identified as the dominant prognostic variables. Other factors examined simultaneously that did not provide additional predictive influence on survival included the level of invasion, pigmentation, growth pattern, lymphocyte infiltration, pathological state, sex, and age. Melanoma thickness was the most important factor for predicting survival in patients with stage I melanoma (P less than 10(-8). This parameter is easy to measure and provides a quantitative estimate of clinically occult regional and distant metastases. Contrary to other reports using single factor analysis, the type of initial surgical treatment, in fact, did influence survival after other variables were taken into consideration. Thus the multifactorial analysis supports the observation that patients with intermediate thickness melanoma thickness of 1.5 to 3.99 mm had a 78% 8-year survival rate with wide excision of the melanoma and elective node dissection, while none survived more than 8 years if a melanoma of the same thickness was only widely excised. Multifactorial analysis is a useful and important statistical method when comparing treatment alternatives and prognostic factors in patients with melanoma.