Although curative resection is the treatment of choice for gastric cancer, controversy exists about the adequate extent of lymph node dissection when resection is performed. We retrospectively assessed 85 patients who underwent a limited lymphadenectomy (D1) and 71 who had an extended lymph node dissection (D2) in a single institution between 1990 and 1998 (median follow-up, 37.3 months). Prognostic factors were assessed by Cox proportional hazard models adjusted for potential confounders. We found no significant difference in the length of hospital stay (median, 12.1 and 13.1 days), overall morbidity (48.2% and 53.5%), or operative mortality (2.3% and 0%) between D1 and D2, respectively. Five-year survival in the D2 group was longer (50.6%) than in the D1 group (41.4%) for tumor stages (tumor-node-metastasis) >I. In multivariate analysis, tumor-node-metastasis stage (hazard ratio for stages >I vs. 0-I, 11.6), the ratio between invaded and removed lymph nodes, the presence of distant metastases, Laurén classification, and the extent of lymphadenectomy (hazard ratio for D1 vs. D2, 2.3; 95% confidence interval, 1.25-4.30) were the only significant prognostic factors. Our experience shows that extended (D2) lymph node dissection improves survival in patients with resected gastric cancer.