There is good prognostic correlation for the two microstaging systems, Breslow depth and Clark level, commonly used to stage melanomas. Many investigators have reported that Breslow depth is the superior microstaging method. Although Clark level has been dropped from most of the proposed American Joint Committee on Cancer (AJCC) melanoma staging system, the AJCC system still includes Clark Level IV as a criterion for upstaging thin melanomas. The authors sought to determine whether this is appropriate, based on melanoma patient data in the Duke Comprehensive Cancer Center database. Of the 8833 patients registered between January 1, 1970 and December 31, 1995, complete data on Breslow depth and Clark level was available for 4560 patients who were without nodal or metastatic disease at presentation. Ten-year survival was measured from the date of excision of the primary tumor until death from melanoma and analyzed using Kaplan-Meier and Cox proportional hazard methodologies. When analyzed separately, both increased Breslow thickness and Clark level correlated with shorter survival times. During subgroup analysis, Breslow thickness remained a significant prognostic indicator of survival at Clark Levels III and IV. Conversely, at narrow levels of Breslow thickness (i.e., 0-0.75 mm, > 0.75 -1.0 mm, > 1.0-1.5 mm) survival times were indistinguishable between Clark Levels III and IV. For the broader Breslow thickness interval of 0-1.0 mm, a barely significant difference between Clark Levels III and IV could be obtained. However, for this thickness range, even greater differences in survival could be obtained by merely comparing Breslow subgroups (i.e., < or = 0.8 mm vs. > 0.8-1.0 mm, < or = 0.9 mm vs. > 0.9-1.0 mm). The authors' data suggested that, after controlling for Breslow depth, Clark level was not a good prognostic indicator for survival. If the AJCC's objective is to design a classification system that will reliably predict the higher risk melanomas, then the system should be based on tumor thickness, which is clearly a better prognostic indicator, and should not be modified because of Clark level.