The aim of this study was to evaluate the independent influence of clinical and pathological variables on survival of patients with gastric carcinoma using the Cox regression proportional hazard model. Of 156 patients operated on for gastric adenocarcinoma, 46 (29.5%) underwent palliative operation, 24 (15.5%) had a palliative resection, and 86 (55%) had a curative resection. The overall 5-year survival rate was 25 +/- 4%. After curative resection, the 5-year survival rate was 44 +/- 6%. Univariate analysis applied to these patients showed that poor survival was related (p less than 0.01) to: age (over 80 years), absence of epigastric pain, vomiting and dysphagia, total gastrectomy, tumor size (more than 4 cm), lymph node involvement (LNI), invasion through the muscularis propria, absence of intestinal metaplasia near the tumor, and linitis plastica. In multivariate analysis, lymph node involvement was found to be the only independent prognostic factor. The 5-year survival rate was 75.5 +/- 8% without LNI, 28 +/- 10% with proximal LNI and 7 +/- 6% with distal LNI. Our results suggest that classification into 3 LNI groups is the best staging system for curative resection in gastric carcinoma.