Cox's multivariate regression model for survival data was applied to 2,012 patients with primary cutaneous melanoma in order to evaluate the relative prognostic value of numerous clinical and histological variables and to establish their prognostically most efficient combination. The material was divided into 4 groups according to the size of resection margin of the primary lesion (less than 2.0 cm, 2.0 cm, 2.1-4.9 cm, and greater than or equal to 5.0 cm). Data were analysed separately in these 4 groups and equivalent results were obtained. The risk factors were clinical stage, site of tumour, tumour thickness, level of invasion, mitotic activity, ulceration, lymphocytic reaction, predominant type of invasive tumour cell and partial regression. When accounting for these factors, histological type, nuclear pleomorphism, nucleolar size, vascular invasion, pigmentation, verrucous growth pattern, and dermal elastosis were without prognostic influence. The effect of sex and age of patient was uncertain and both variables, therefore, were retained in the model. By using Cox's method it is possible to make a qualified estimate of the survival for the individual patient.