Prognostic models in melanoma

Abstract

Predicting which patients with primary melanoma are at risk of developing metastastic disease is important for making rational therapeutic decisions. Tumor thickness alone is the most commonly used predictor of survival, but other clinical and pathologic variables also play an important role. We have developed two multivariate logistic regression models to predict survival in patients who have primary melanoma. The first of these models assigns patients to two groups based on radial or vertical growth phase. The probability of survival for those patients with vertical growth phase tumors was further determined based on a model using six variables (mitotic rate, tumor infiltrating lymphocytes, tumor thickness, anatomic site of the primary tumor, sex, and histologic regression) that have the greatest strength as independent predictors of survival. This model is 89% accurate for predicting survival in patients with vertical growth phase tumors. A second model has been developed that uses readily available clinical parameters to predict survival. Four variables (tumor thickness, anatomic site, age, and sex) entered into the model as powerful independent predictors. Clinical algorithms for assessing patient risk are provided.