A prospective study was undertaken to evaluate selective visceral angiography (SVA) in the management of patients with pancreatic and periampullary cancer. Over a 30-month period 52 patients with potentially resectable pancreatic or periampullary cancer underwent SVA; 4 patients had obvious angiographic evidence of widely disseminated disease and were not subjected to laparotomy while 2 further patients were eventually considered too frail for resection. The remaining 46 patients (median age 58 years, range 37-73 years, males 26, females 20) had no evidence of disseminated disease on ultrasonography and/or CT scanning and had both SVA and surgery and form the basis of this study. Vascular anomalies were detected in 16/46 (35%) patients. Hepatic metastases were wrongly diagnosed by angiography in 7 out of 9 patients (77%). SVA correctly predicted resectability or irresectability in 28/46 patients (overall predictive value 61%). Of the 27 patients who proved to have irresectable disease at operation, 11 were correctly identified by SVA (sensitivity 41%). Of the 13 patients reported to have irresectable disease, 2 underwent resection (false-positive rate 15%). Of the 33 patients reported to have resectable disease, 16 were irresectable (false-negative rate 48%). Overall there was a poor relationship between resectability and the angiographic features. On the basis of these data, SVA cannot be considered a sufficiently accurate means of assessing resectability. Its use for this purpose in patients with pancreatic and periampullary cancer is not justified.