In this paper, we have focused on the metastatic behavior of EGC and its particularities. The main factors that are currently considered as predictors of the metastatic behavior and that are used in the therapeutic decision (endoscopic resection vs surgical removal) are the tumor size (upper or bellow 2 cm), depth of infiltration, angiolymphatic invasion, the presence or absence of ulceration, and histologic type (undifferentiated vs differentiated carcinomas). However, most of the metastatic cases were published as case reports or case series. This is the reason why a proper estimation of metastatic risk in EGC is not well known. To date, 79 cases presenting bone metastases, three reports of brain metastases, and one EGC that was associated with skin metastasis were published. However, occult metastasis, lymph node micrometastasis, and skip metastasis can also occur and should be identified. Making a synthesis of the literature data that is correlated with our experience, we finally proposed the inclusion of the six Japanese subgrouping system, tumor size, angiolymphatic invasion, and micrometastasis as components of the pTNM staging system, which should be particularly adapted for EGC.