This paper describes the construction, validation and use of a simple prognostic score suitable for predicting survival of patients undergoing a curative gastric resection. Using death from all causes as outcome, the prognostic significance of age, sex, tumour site, stage of disease (nodal status and wall invasion), surgical treatment and histological type was investigated in a set of 213 patients recruited in a multi-centre clinical trial. A Weibull multiple regression model was adopted to evaluate the joint effect of these variables on survival. From a full model, containing all the variables, a final parsimonious model was obtained by means of a backward selection procedure. The prognostic score is based on the final model, including four variables which are easily detected in every institution: age, wall invasion, site of tumour, and nodal status. Three groups of patients with different probabilities of surviving 5 years from surgery were identified: group I (survival probability > or = 70%), group II (30%-69%) and group III (< 30%). The prognostic score, obtained from the multicentre trial patients, was tested on a set of 135 consecutive patients in an independent institution, confirming its reliability in predicting survival. The score system presented can supply a simple tool for classifying patients radically operated for gastric cancer into three well discriminated groups from the prognostic point of view.