OBJECTIVE: This prospective study was undertaken to evaluate the accuracy of a noninvasive "all-in-one" staging method in predicting surgical resectability in patients with pancreatic or periampullary tumors. SUMMARY BACKGROUND DATA: Despite progress in imaging techniques, accurate staging and correct prediction of resectability remains one of the chief problems in the management of pancreatic tumors. Staging algorithms designed to separate operable from inoperable patients to save the latter an unnecessary laparotomy are becoming increasingly complex, expensive, time-consuming, invasive, and not without risks for the patient. METHODS: Between August 1996 and February 1997, 58 consecutive patients referred for operation of a pancreatic or periampullary tumor were examined clinically and by 5 staging methods: 1) percutaneous ultrasonography (US); 2) ultrafast magnetic resonance imaging (UMRI); 3) dual-phase helical computed tomography (CT); 4) selective visceral angiography; and 5) endoscopic cholangiopancreatography (ERCP). The assessment of resectability by each procedure was verified by surgical exploration and histologic examination. RESULTS: The study comprised 40 male and 18 female patients with a median age of 63 years. Thirty-five lesions were located in the pancreatic head (60%), 11 in the body (19%), and 1 in the tail of the gland (2%); there were 9 tumors of the ampulla (16%) and 2 of the distal common duct (3%). All five staging methods were completed in 36 patients. For reasons ranging from metallic implants to contrast medium allergy or because investigations already had been performed elsewhere, US was completed in 57 (98%), UMRI in 54 (93%), CT in 49 (84%), angiography in 48 (83%), and ERCP in 49 (84%) of these 58 patients. Signs of unresectability found were vascular involvement in 22 (38%), extrapancreatic tumor spread in 16 (26%), liver metastases in 10 (17%), lymph node involvement in 6 (10%), and peritoneal nodules in only 2 patients (3%). These findings were collated with those of surgical exploration in 47 patients (81 %) and percutaneous biopsy in 5 (9%); such invasive verification was deemed unnecessary and therefore unethical in 6 clearly inoperable patients (10%). In assessing the four main signs of unresectability (extrapancreatic tumor spread, liver metastases, lymph node involvement, and vascular invasion), the overall accuracy of UMRI was 95.7%, 93.5%, 80.4%, as compared to 85.1%, 87.2%, 76.6% for US and 74.4%, 87.2%, 69.2% for CT. In assessing vascular invasion, the sensitivity, specificity, and overall accuracy of angiography were 42.9%, 100%, and 68.8%, respectively. There were 3 complications (12.5%) after 24 resections, 5 in 17 palliative procedures, and none after 6 explorations only. The hospital stay was 14 days after resection, 13 after palliative bypass, and 6 after exploration alone. There was no operative or hospital mortality in these 58 cases. CONCLUSIONS: Although it is by no means 100% accurate, UMRI is equal or even superior to all other staging methods. It probably will replace most of these, because it provides an "all-in-one" investigation avoiding endoscopy, vascular cannulation, allergic reactions, and x-radiation. But because even UMRI is not perfect, the final verdict on resectability of a tumor still will depend on surgical exploration in some cases.