A prognostic score for gastric cancer

Background: Scoring systems are generally used for predicting prognosis in the intensive care unit, but there is no score being used frequently for predicting prognosis in gastric cancer. The aim of this study was to develop a prognostic score for gastric cancer patients. Methods: Demographic, clinical, laboratory, radiologic, histopathologic data, and operative findings for 128 patients who had curative or palliative resection for gastric cancer were analyzed for their effect on overall and disease-free survival. Ten variables—invasion depth of tumor, node status (American Joint Committed on Cancer, 1992), metastasis, node status (Union Internationale Contre le Cancer, 1997), metastatic lymph node ratio, resectability, tumor location, extent of lymphadenectomy, Borrmann type, Lauren type—that have independent significant effect or borderline significance on both overall and disease-free survival according to multivariate analysis were chosen. Coefficients were calculated for these variables by using Cox regression analysis, and thus the Prognostic Score for Gastric Cancer (PSGC) was designed. All patients were scored using the PSGC and also staged clinically (AJCC 1992) and histopathologically (AJCC 1992 and UICC 1997). Results: Patients were grouped according to their scores: group 1, patients with scores 20 to 50 (probability of 5-year overall survival 50% to 95%); group 2, patients with scores 51 to 80 (probability of 5-year overall survival 10% to 50%); and group 3, patients with scores 81 and higher (probability of 5-year overall survival <10%). Overall survival and disease-free survival decreased significantly with increasing scores. The association of PSGC and staging systems with survival was analyzed by stepwise logistic regression and Cox regression analyses. PSGC was proved to have the most significant association with overall and disease-free survival. Conclusions: Inclusion of more variables in PSGC seems to make it superior than staging. It is easy to adapt PSGC to different patient populations, which may make it accepted as a practical and useful scoring system in clinical practice.