Background. A univariate and multivariate statistical analysis of a single surgeon's experience with resectable malignant melanoma during 26 years (November 1970 to August 1996) was conducted. Methods. Six hundred twenty consecutive patients were registered. Univariate analysts of disense-fr ee survival (DFS) and melanoma survival (MS) was calculated by the Kaplan-Meier method and correlated to American Joint Committee on Cancer stage, thickness, ulceration, site, lymph node involvement, age, sex, type, and excision margins. Linear trends, log-rank test, and pairwise comparisons were used to discriminate differences in survival curves. A Cox proportional hazards model was used for multivariate analysis and determination of relative risk. Results. Univariate analysis of stage, thickness (in millimeters), ulceration, lymph node involvement, age, type, and margins of excision were predictive of DFS (5 years, 85.7%; 10 years, 82.5%) and MS (5 years, 92.2%; 10 years, 87.8%) (P < .01). Multivariate analysis revealed correlations with thickness, ulceration, and age in predicting DFS (relative risk = 2.75, 2.21, and 1.47, respectively) and MS (relative risk = 2.66, 2.47, and 1.48, respectively). The 5-year MS rate was 73.3% and 93.3% for patients with positive and negative lymph nodes, respectively. Of 133 patients who underwent lymph node dissection, 28 (21.1%) had nodal metastases. Patients with primary melanomas thicker than 4 mm had 50% metastatic involvement of their lymph nodes. Conclusions. Our findings reveal that thickness, ulceration, and age are the most important predicting factors in DFS and MS. The data support including ulceration and age in modifying America joint Committee on Cancer staging for melanoma.