One hundred and eightysix cancer patients with different types, site and stage of disease, consecutively admitted to the Istituto Nazionale Tumori, Milan, were examined. The purpose of the study was to assess the prevalence of subjective anorexia and to determine whether some nutrition-related indices are different within homogeneous categories of cancer patients, anorectic and non-anorectic. Anorexia was measured on the basis of an appetite score, evaluated with a visual analogue rating scale. Nutritional intake was evaluated by a computerised programme of the habitual food consumption. Nutritional assessment included anthropometric evaluations (triceps skinfold, arm circumference, arm muscle circumference) and biochemical indices (serum albumin, and total iron binding capacity). A nutritional index was then calculated: GNS = P + I + M × 100, where P: actual weight/premorbid weight; I: actual weight/ideal weight; M: actual arm muscle mass circumference/ideal arm muscle mass circumference. The evaluation of the data showed that: (1) The prevalence of subjective anorexia ranged from 1 3 to 2 5 of the patients depending on the type and site of tumour, and stage of disease. (2) Most of the nutritional-related variables were significantly worse in anorectic compared with non-anorectic patients. (3) Within groups of patients with different tumour location, but homogeneous for the presence or absence of subjective anorexia, there is no difference with respect to nutritional indices; except for patients with cancer of the oesophagus and stomach who were always more depleted. (4) Anorexia did not seem to influence the nutritional status in patients with cancer at an early stage (except in cancer of oesophagus and stomach) but it was associated to malnutrition as the disease progressed to more advanced stages.