Background: This study was designed to establish a prognostic score for gastric cancer that takes into account factors related to the tumor, the patient, and the treatment.Methods: Two hundred thirty patients with gastric adenocarcinoma admitted to the Department of Abdominal Surgery at Hospital do Caˆncer A. C. Camargo (Sa˜o Paulo) and treated by gastrectomy from January 1992 until December 1996 were included in this retrospective cohort. The prognostic score was created according to the variables identified in the multivariate analysis and by using the regression coefficients generated by the Cox regression.Results: The 5-year overall survival rate was 44.5%. The final multivariate model identified six variables with a significant and independent effect on survival: sex, weight loss, lymphocyte count, tumor-node-metastasis staging, lymphadenectomy, and lymph node ratio. Patients were divided into four groups according to their scores, as follows: group 1, 0 to 3.0; group 2, 3.5 to 5.5; group 3, 6.0 to 8.5; and group 4, 9.0 to 14.0. The 5-year survival rates were 91.5%, 49.3%, 20.3%, and .0% for the score groups 1, 2, 3, and 4, respectively (P < .001). The score was superior in the assessment of prognosis when compared with tumor-node-metastasis staging alone.Conclusions: It is possible to create a prognostic score that simultaneously includes factors related to the tumor, patient, and treatment, thus generating a more effective system in predicting the prognosis than the morphology-based staging systems.