PURPOSE To retrospectively assess the accuracy of combined multiphasic computed tomography (CT) and direct cholangiography for evaluation of the resectability of hilar cholangiocarcinoma, on the basis of revised criteria for unresectability, by using surgery as the reference standard. MATERIALS AND METHODS Institutional review board approval was obtained, and informed consent was waived. From 1998 to 2003, 55 patients (37 men, 18 women; mean age +/- standard deviation, 59 years +/- 12) with surgically proved hilar cholangiocarcinomas who underwent preoperative CT (single-detector row CT, n = 26; multi-detector row CT, n = 29) and cholangiography were included for study. The authors' revised criteria for unresectable tumor were contralateral hepatic artery invasion; main or contralateral portal vein invasion longer than 2 cm; biliary extension to the contralateral secondary confluence, farther than 2 cm from hepatic hilum; enlarged lymph nodes at the celiac, portacaval, and paraaortic area; and other ancillary findings. Tumor resectability based on these parameters was determined at imaging by two radiologists in consensus. Mann-Whitney U test and weighted kappa coefficient of agreement were used for accuracy determination. RESULTS For depiction of portal vein invasion (in 26 patients), CT yielded an accuracy of 85.5%. Arterial invasion was found at surgery in 19 patients, with CT providing an accuracy of 92.7%. For prediction of node involvement (15 patients, 27%), CT yielded an accuracy of 83.6%. The extent of ductal involvement could be accurately predicted in 46 patients (84%) (weighted kappa = 0.767). In 30 of 42 patients with disease classified as resectable according to revised criteria, disease was found to be resectable at surgery (71.4% positive predictive value). In 11 of 13 patients with disease classified as unresectable according to revised criteria, unresectable disease was confirmed (84.6% negative predictive value). Overall accuracy of resectability was 74.5%. CONCLUSION Combined interpretation of CT and direct cholangiographic images by using our revised criteria resulted in overall accuracy of 74.5% for prediction of resectability for hilar cholangiocarcinoma.