AIM To clarify the diagnostic efficacy and limitations of endoscopic ultrasonography (EUS) and the characteristics of early gastric cancers (EGCs) that are indications for EUS-based assessment of cancer invasion depth. METHODS We retrospectively investigated the cases of 153 EGC patients who underwent conventional endoscopy (CE) and EUS (20 MHz) before treatment. RESULTS We found that 13.7% were "inconclusive" cases with low-quality EUS images, including all nine of the cases with protruded (0-I)-type EGCs. There was no significant difference in the diagnostic accuracy between CE and EUS. Two significant independent risk factors for misdiagnosis by EUS were identified-ulcer scarring [UL(+); odds ratio (OR) = 4.49, P = 0.003] and non-indication criteria for endoscopic resection (ER) (OR = 3.02, P = 0.03). In the subgroup analysis, 23.1% of the differentiated-type cancers exhibiting SM massive invasion (SM2) invasion (submucosal invasion >= 500 mu m) by CE were correctly diagnosed by EUS, and 23.1% of the undifferentiated-type EGCs meeting the expanded-indication criteria for ER were correctly diagnosed by EUS. CONCLUSION There is no need to perform EUS for UL(+) EGCs or 0-I-type EGCs, but EUS may enhance the pretreatment staging of differentiated-type EGCs with SM2 invasion without UL or undifferentiated-type EGCs revealed by CE as meeting the expanded-indication criteria for ER.