Background and study aims: This study aimed to compare the diagnostic accuracy of endoscopic ultrasonography (EUS) with that of conventional endoscopy for staging depth of invasion (T staging) in early gastric cancer. Patients and methods: A total of 955 patients with suspected early gastric cancer were prospectively registered. EUS staging was carried out prospectively by a single endoscopist using either miniprobe or radial EUS depending on the endoscopic appearance of the tumor. Conventional endoscopy staging was performed retrospectively by consensus between two endoscopists who were blinded to the EUS staging. Conventional endoscopy staging was conducted on the basis of endoscopic features such as surface nodularity and fold convergence. Patients underwent either surgical (n=586) or endoscopic resection (n=369) with curative intent. The staging accuracy of each test was compared with the pathological staging of the resected specimen. Results: The presence of a T1m tumor was histologically confirmed in 644 cases (67.4%) and that of a T1sm tumor in 311 cases (32.6%). The overall accuracy of EUS staging was 67.4% (644/955) and that of conventional endoscopy staging was 73.7% (704/955) (P<0.001). The accuracy of miniprobe EUS was significantly higher than that of radial EUS (79.5% vs. 59.6%, P<0.001), but did not differ significantly from that of conventional endoscopy (79.0%). Conclusions: EUS does not substantially impact on pretreatment T staging of patients with early gastric cancer compared with conventional endoscopy. Therefore, EUS may not be necessary routinely, and conventional endoscopy may be sufficient for determining the optimal therapeutic strategy, especially in relation to endoscopic resection for early gastric cancer.