PREDICTING 10-YEAR SURVIVAL OF PATIENTS WITH PRIMARY CUTANEOUS MELANOMA - CORROBORATION OF A PROGNOSTIC MODEL

BACKGROUND, Recently, the pigmented Lesion Group at the University of Pennsylvania described a 4-variable model for predicting 10-year survi val for patients with primary cutaneous melanoma. The variables are tu mor thickness, anatomic site of the lesion, age, and gender. The objec tive of the current study was to test the validity of this model, empl oying the large data base of the New York University Melanoma Cooperat ive Group. METHODS. The predicted probabilities of 10-year survival fo r 780 patients with primary cutaneous melanoma were determined by mult ivariate logistic regression, using the 4 variables. RESULTS, The over all 10-year survival rate of the current study group was 78.4%. Of the four variables, tumor thickness, anatomic site of the lesion, and age were found to be independent predictors of survival. Although surviva l was better for women, gender was not a statistically significant fac tor in predicting 10-year survival when entered into the multivariate logistic regression model. In the current study, the probability of 10 -year survival of patients with melanomas < 0.76 mm ranged from 93-99% , depending on the age and primary site. Age and site had more impact on the prognosis of intermediate and thick melanomas than on thin mela nomas. Thus, for melanomas 0.76-1.69 mm, 1.70-3.60 mm, and thicker tha n 3.60 mm, the probabilities of survival ranged from 70-94%, 39-82%, a nd 23-68%, respectively. CONCLUSIONS, The wider ranges in survival rat es for thicker melanomas, depending on the other variables, emphasize the importance of including variables in addition to tumor thickness i n a prognostic model. Using a large data base from a medical center, t he current study supports the prognostic multivariate model of the Pig mented Lesions Group of the University of Pennsylvania; however, the a uthors of the current study did not find gender to be statistically si gnificant in this multivariate model. (C) 1997 American Cancer Society .