BACKGROUND : Early gastric cancer that meets the expanded criteria for endoscopic resection (ER) is expected to be associated with a negligible risk for lymph node metastasis (LNM); however, recent studies have reported LNM in submucosal gastric cancer patients who met the existing criteria. In this study, we develop the revised criteria for ER of submucosal gastric cancer with the aim of minimizing LNM. METHODS : We analyzed the clinicopathological data of 2461 patients diagnosed with differentiated, submucosal gastric cancer who underwent surgery at three tertiary hospitals between March 2001 and December 2012, and re-analyzed the pathological slides of all patients. The depth of submucosal invasion was measured histopathologically in two different ways (the classic and alternative methods) to obtain accurate data. RESULTS : Of the enrolled subjects, 306 (17.0%) had LNM. The width of submucosal invasion correlated well with the LNM. We defined the depth and width of submucosal infiltration associated with the lowest incidence of LNM. None of the 254 subjects developed LNM when the following criteria were met: tumor diameter ≤ 3 cm, submucosal invasion depth < 1000 μm (as measured using the alternative method), submucosal invasion width < 4 mm, no lymphovascular invasion, and no perineural invasion; however, LNM was observed in 2.7% of subjects (6/218) who met the existing criteria. CONCLUSIONS : We revised the criteria for ER by adopting the alternative method to measure the depth of submucosal invasion and adding the width of such invasion. Our criteria better predicted LNM than the current criteria used to select ER to treat submucosal gastric cancer.