Sondergaard K, Schou G
Department of Pathology, Finsen Institute, Rigshospitalet, Copenhagen, Denmark.
Am J Dermatopathol. 1985;7 Suppl:1-4. doi: 10.1097/00000372-198501001-00004.
Cox's multivariate regression model for survival data was applied to facts in cases of 2,012 patients with primary cutaneous malignant melanomas in order to evaluate the relative prognostic value of numerous clinicohistologic variables and to establish their prognostically most meaningful combination. The variables that were found to be independent risk factors were clinical stage, site of lesion, thickness of lesion, level of invasion, mitotic figures, ulceration, lymphocytic infiltrate, predominant type of neoplastic cell, and regression. The factors that were found to be without prognostic significance were histological type, nuclear pleomorphism, nucleolar size, vascular invasion, pigmentation, and attendant solar elastosis. The prognostic effects of sex and age of patients were uncertain, and both variables therefore were retained in the model.