Background: The accuracy of computed tomography (CT) in predicting resectability of pancreatic malignancy has been questioned recently and alternative methods have been recommended.
Methods: To determine the accuracy of CT for predicting resectability and its influence on survival, a standard protocol for performing CT and reporting the results was developed and then compared retrospectively with the ability of one surgeon to perform a resection during 1989-1994. Postoperative survival was determined.
Results: Of 88 consecutive patients 35 (40 per cent) had CT-resectable disease and 53 (60 per cent) had CT-irresectable disease. Twenty-one patients were excluded because of advanced disease or poor performance status. Of the remaining 67 patients, 47 (70 per cent) had pancreatic ductal adenocarcinoma and 20 (30 per cent) had ampullary adenocarcinoma, of whom 32 had a resection, 32 had a palliative bypass and three had only a staging laparoscopy. The sensitivity and specificity for computed tomographic prediction of resectability were 72 and 80 per cent respectively. The positive predictive value was 77 per cent and the negative predictive value 76 per cent. There were seven false-positive and nine false-negative findings. Survival was more dependent on whether or not resection was performed than on computed tomographic predictability of resection.
Conclusion: CT was reasonably accurate in predicting resectability but cannot be relied on entirely, requiring an improvement in staging methods for pancreatic malignancy.