In a prospective study, endoscopic ultrasonography was compared with transabdominal ultrasonography, computed tomography, and angiography in 60 consecutive patients with pancreatic (n = 46) and ampullary (n = 14) cancer considered to be candidates for surgery. The diagnostic value of these imaging procedures in determining local resectability was assessed. The diagnosis of ampullopancreatic malignancy was made by operation (n = 40) or puncture/biopsy (n = 20). In the 40 patients who underwent surgery, endoscopic ultrasonography was significantly superior to abdominal ultrasonography and computed tomography in determining tumor size and extent and lymph node metastases of pancreatic and ampullary cancer. Furthermore, involvement of the portal venous system as judged by histopathology or surgical exploration was correctly assessed by endoscopic ultrasonography in 95%, whereas angiography (85%), computed tomography (75%) and abdominal ultrasonography (55%) were less sensitive. Of 11 cases of portal venous infiltration found at surgery, endoscopic ultrasonography correctly predicted 10, abdominal ultrasonography only 1, computed tomography 4, and angiography 5 (P less than 0.05 for all three comparisons). Twenty patients did not undergo surgery for different reasons: of those, 9 patients were excluded from operation because of portal venous involvement as shown by angiography. Endoscopic ultrasonography detected portal venous invasion in all these cases. In contrast to the venous system, arterial encasement was less reliably detected by endoscopic ultrasonography. In conclusion, endoscopic ultrasonography is the most effective single imaging procedure for local tumor staging in pancreatic and ampullary cancer. Thus, endoscopic ultrasonography will improve the assessment of tumor resectability and further decrease the need for explorative laparotomy.