Prognostic index for clinical Stage I cutaneous malignant melanoma

Thirteen variables were evaluated for their significance in predicting the survival of 148 patients in a retrospective study with clinical Stage I cutaneous malignant melanoma. The variables studied were histological type, tumor thickness, level of invasion, mitotic activity, pigmentation type, existence of ulceration, presence of lymphocytic infiltration, cell type, sex and age of the patient, site of melanoma, and wide local excision preceded or not by contact radiotherapy and associated or not with lymphadenectomy. When these variables were individualized, only seven were significantly related to survival: histological type; tumor thickness; level of invasion; mitotic activity; pigmentation type; existence of ulceration; and sex of the patient. Multivariate analysis based on Breslow's version of the Cox proportional-hazards model was performed on a group of 110 patients. This analysis demonstrated that 5 of the original 13 variables (i.e., mitotic activity, tumor thickness, sex, lymphadenectomy, and site of primary melanoma) could be used to develop a prognostic model. A Gompertz distribution which provided for an appropriate smoothing of the Breslow model estimates was used to derive a simple prognostic index and to predict the survival of individual patients. Fifty-four patients in a prospective study were subsequently evaluated with the Gompertz model in order to test the prognostic accuracy of the model for the five variables.