Fourteen variables were tested for their ability to predict visceral or bony metastases in 177 patients with clinical Stage I melanoma of intermediate thickness (1.51 - 3.39 mm). A Cox multivariate analysis yielded a combination of four variables that best predicted bony or visceral metastases for these patients: 1) mitoses greater than 6/min 2 (p = 0.0007), 2) location other than the forearm of leg) p = 0.009, 3) ulceration width greater than 3 mm (p = 0.04), 4) microscopic satellites (p = 0.05). The overall prognostic model chi square was 32.40 with 4 degrees of freedom (p less than 10 (-5). Combinations of the above variables were used to separate these patients into at least two risk groups. The high risk patients had at least a 35% or greater chance of developing visceral metastases within five years, while the low risk group had greater than an 85% chance of being disease free at five years. Criteria for the high risk group were as follows: 1) mitoses greater than 6/mm 2 in at least one area of the tumor, irrespective of primary tumor location, or 2) a melanoma located at some site other than the forearm or leg and histologic evidence in the primary tumor of either ulceration greater than 3 mm wide or microscopic satellites. The low risk group was defined as follows: 1) mitoses less than or equal to 6/mm 2 and a location on the leg or forearm, or 2) mitoses less than or equal to 6/mm 2 and the absence in histologic sections of the primary tumor of both microscopic satellites and ulceration greater then 3 mm wide. The number of patients in this series who did not undergo elective regional node dissection (N = 47) was probably too small to detect any benefit from this procedure. Based on survival rates from this and other studies, it is estimated that approximately 1500 patients with clinical Stage I melanoma of intermediate thickness in each arm of a randomized clinical trial would be needed to detect an increase in survival rates from elective regional node dissection.