Background: Previous malignant melanoma studies regarding prognostic factors have often selected their patients from hospitals. Unfortunately, most of these studies have had small numbers of patients, consisted of short-term data sets, omitted important factors, did not optimize histopathologic classification, had too short or inadequate follow-up, and did not test their predictive models.
Purpose: Our study goals were to identify independent clinical and histopathologic determinants of survival in malignant melanoma, to analyze changes in prognostic value over follow-up time, and, finally to construct a prognostic index.
Methods: A random sample from the Swedish Cancer Registry of the records of 498 (246 men and 252 women) patients defined by gender, five 5-year time periods of diagnosis from 1960 through 1984, and five anatomic sites formed the cohort on whom data were analyzed by univariate analyses. Multivariate analyses were based on data on 476 patients with complete information about all variables. All patients in the cohort had complete follow-up through December 31, 1989. Clinical information was abstracted and recorded as: date of diagnosis, stage at diagnosis, sex, age, anatomic site of primary tumor, date of death, and cause of death. Histopathologic slides were re-examined and classified with regard to histogenetic type, level of invasion, tumor thickness, ulceration, vascular invasion, regression, lymphocytic reaction, pre-existing nevus, and cell type.
Results: All variables, except pre-existing nevus and cell type, were significant predictors of survival. In the multivariate analyses including all variables, women still had a significant, 33% lower relative hazard than men. The prognosis was poor in the youngest age group. Patients with external ear, scalp-neck, and trunk-located melanoma had increasing relative hazard when all variables were included. Regional metastases and tumor thickness remained independent prognostic factors. No significant association between histogenetic type or level of invasion persisted. Patients whose tumors showed ulceration or vascular invasion had lower relative hazard when all variables were included. Level of invasion, tumor thickness, ulceration, and vascular invasion were significantly associated with the prognosis during both short- and long-term follow-up. The patients were subgrouped according to percentage fractions of their score on the prognostic index. Survival curves for these groups of patients were well separated, thus identifying patients with a low or high risk of death from malignant melanoma.
Conclusion: The present population-based study identifies independent clinical and histopathologic predictors of survival in cutaneous malignant melanoma and emphasizes the role of histopathologic characteristics such as tumor thickness, ulceration, and vascular invasion besides anatomic site.