Endoscopic ultrasonography (EUS) was compared to ultrasonography (US) and CT scan (CT) in order to evaluate its role in the diagnosis and the locoregional spread assessment of pancreatic cancer. Sixty-four patients suspected of pancreatic cancer were studied prospectively, and the results of imaging techniques were compared to histology and surgical exploration. There were 49 cases of pancreatic adenocarcinoma, 11 of pancreatitis, 2 of common bile duct carcinoma, 1 lymphoma and 1 hepatocellular carcinoma with peripancreatic metastatic lymph nodes. EUS was significantly more accurate (91%) than CT (66%) and US (64%) for diagnosis of pancreatic cancer. EUS was able to image all 7 cancers less than 25 mm in diameter, US and CT only one. There were 4 false positives with EUS which were all cases of pseudotumorous pancreatitis. For detecting lymph node involvement, EUS was significantly more sensitive (62%) and accurate (74%) than US (8% and 37%) and CT (19% and 42%), respectively. Invaded lymph nodes adjacent to large tumors and micrometastatic involvement were responsible for this lack of sensitivity. EUS was significantly more sensitive (100%) than CT (71%) and US (17%) for detecting venous involvement. The specificity of EUS was lower (67%) because of duodenal bulb stenosis and large tumors. In conclusion, this prospective and comparative study confirms that EUS is an accurate tool for diagnosis and locoregional spread assessment of pancreatic cancer when performed in a reference center. EUS is of particular interest for small tumours. However, EUS does not enable differentiation of pseudotumorous pancreatitis from adenocarcinoma.