Background For early gastric cancer, submucosal invasion may be unrecognized until histopathologic examination of the specimen obtained by EMR. Gastrectomy wi th lymphadenectomy is the standard treatment for such submucosal cancers. Howeve r,approximately 80%of submucosal cancers do not have lymph node metastasis. Unn ecessary surgery could be avoided if a subgroup of patients with submucosal canc er with negligible risk of lymph node metastasis can be defined. This study was conducted to define such a subgroup. Methods Data from 104 patients surgically t reated for differentiated submucosal cancers were retrospectively collected. A m ultivariate analysis of clinicopathologic factors was performed to identify pred ictive factors for lymph node metastasis. Results Three independent risk factors , namely, female gender (p=0.0174), deep invasion(≥500 μm) into the submucosal layer (p=0.001), and presence of lymphatic involvement (p < 0.0001) were associ ated with lymph node metastasis. Lymph node metastasis was not observed in any p atient who had limited submucosal invasion and absence of lymphatic involvement. The rate of lymph node metastasis was calculated to be 80%in patients who had both deep submucosal invasion and lymphatic involvement. Conclusions If endoscop ic resection specimens exhibit no deep penetration(< 500 μm) into the submucosa l layer and lymphatic involvement is absent, EMR may be sufficient treatment for submucosal well-differentiated early gastric cancers. A long-term followup st udy of patients with such lesions treated by EMR alone is required. Background For early gastric cancer, submucosal invasion may be unrecognized until histopathologic examination of the specimen obtained by EMR. Gastrectomy wi th lymphadenectomy is the standard treatment for such submucosal cancers. Howeve r, approximately 80% of submucosal cancers do not have lymph node metastasis. Unn ecessary surgery could be avoided if a subgroup of patients with submucosal canc er with negligible risk of lymph node metastasis can be defined. This study was conducted to define such a subgroup. Methods Data from 104 patients surgically t reated for differentiated submucosal cancers were retrospectively collected. A m ultivariate analysis of clinicopathologic factors was performed to identify pred ictive factors for lymph node metastasis. Results Three independent risk factors, namely, female gender (p = 0.0174), deep invasion (≥500 μm) into the submucosal layer = 0.001), and presence of lymphatic involvement (p <0.0001) were associ ated with lymph node metastasis. The rate of lymph node metastasis was calculated to be 80% in patients who had both deep submucosal invasion and lymphatic involvement. Conclusions If endoscop ic resection specimens exhibit no deep penetration (<500 μm) into the submucosa l layer and lymphatic involvement is absent, EMR may be sufficient for submucosal well-differentiated early gastric cancers. A long-term follow up st udy of patients with such lesions treated by EMR alone is required.