Background: Although the overall incidence of non-cardia gastric cancer has declined, the incidence of localized disease (confined to the stomach) has remained unchanged (1 per 100,000). The overall survival with gastric cancer remains poor and unchanged over the past 2 decades. This brings into question the quantity and quality of gastrectomy and lymphadenectomy, the main treatment for non-metastatic gastric cancer. To ensure accurate staging and hence proper treatment, lymph node (LN) sampling of at least 15 nodes is recommended (American Joint Committee on Cancer, 1997). We examined the determinants of gastric resection and adequate LN sampling for gastric cancer. Methods: Data from Surveillance, Epidemiology, and End Results (SEER) registries was used to identify patients with non-cardia gastric cancer diagnosed during 1983-2002. Logistic multivariable regression was used to examine determinants of gastric resection and adequacy of lymphadenectomy. Cox proportional hazard (PH) models were used to examine trends in mortality risk. All models adjusted for age, race, gender, geographic region, and cancer stage (localized, regional, metastatic); Cox PH models also adjusted for treatment (gastrectomy, gastrectomy and radiation). Results: Resection for non-cardia gastric cancer has steadily declined between 1983 and 2002 from 66% to 60% of all cases. In multivariable models, gastrectomies were less likely to be performed in 2002 (-48% compared to 1983), patients +70 (-39% compared to patients younger than 40), White race (-54% compared to Asian), and localized disease (-78% compared to regional disease). In localized disease, only 75% underwent resection. Wide geographic variability was found also (lowest in New Mexico [-45%] as compared to highest in Hawaii ). Adequate LN sampling (15+ LN) was recorded in only 25% overall and 19% of localized disease. Improvement in LN collection since 1997 has been modest, with only a 7% increase. Again, the greatest disparity was related to geographic region; for example, adequate sampling occurred less in Utah (-83%) compared to Hawaii . Cox PH models showed a 66% (95% CI, 65%-68%) and 71% (95% CI, 68%-73%) increased chance of survival with gastrectomy and with both gastrectomy and radiotherapy compared to those who received no treatment. Conclusion: There is apparent underutilization of gastrectomy for gastric cancer, even for early stage disease. Furthermore, in up to 65% cases where gastrectomies are performed, LN collection is inadequate. Geographic variations in performing gastric resection and adequate LN sampling were almost as significant as stage of the cancer.