Thirteen variables were studied for their relative usefulness in predicting recurrent disease in 107 patients with clinical Stage I melanoma of the upper extremity. After a mean follow-up period of 54 months, the only patents who have had recurrent disease to date are those who primary lesions were located either on the hand or posterior upper arm. The five-year disease-free survival role for 44 patients with melanoma at these sites was 68%. None of 63 patients with melanoma located on the forearm of anterior upper arm have had recurrent disease (i.e., the five-year, disease-free survival rate was 100% (p = 0.00004), compared with the hand or posterior arm group). A Cox proportional hazards (multivariate) analysis demonstrated that two primary tumor histologic variable, thickness in millimeters and ulceration, interacted to produce the best prognostic model for those 44 patients with melanoma of the hand or posterior upper arm. Twenty-one patients with primary lesions at these sites had primary tumors less than 2.25 mm in thickness and no evidence of ulceration histologically. Their five-year, disease-free survival role was 95%. For the remaining 23 patients with primary tumors on the hand or posterior upper arm who had either histologic evidence of ulceration or primary tumors greater than or equal to 2.25 mm, the five-year disease-free survival rate was 37% (p = 0.002, compared with group nonulcerated, thin lesions). The excellent survival rate for patients with melanomas on the forearm or anterior upper arm was not completely explained by pathologic stage, by primary tumor thickness, or by histologic ulceration of the primary tumor.