Diagnostic performance of imaging criteria for distinguishing autoimmune cholangiopathy from primary sclerosing cholangitis and bile duct malignancy

To determine the diagnostic performance of imaging criteria for distinguishing Ig-G4-associated autoimmune cholangiopathy (IAC) from primary sclerosing cholangitis (PSC) and bile duct malignancy. A medical records search between January 2008 and October 2013 identified 10 patients (8 M, 2 F, mean age 61 years, range 34-82) with a clinical diagnosis of IAC. Fifteen cases of PSC (6 M, 9 F, mean age 50, range 22-65) and 15 cases of biliary malignancy (7 M, 8 F, mean age 65, range 48-84) were randomly selected for comparative analysis. Three abdominal radiologists independently reviewed MRI with MRCP (n = 32) or CT (n = 8) and ERCP (n = 8) for the following IAC imaging predictors: single-wall bile duct thickness > 2.5 mm, continuous biliary involvement, gallbladder involvement, liver disease, peribiliary mass, or pancreatic and renal abnormalities. Each radiologist provided an imaging-based diagnosis (IAC, PSC, or cancer). Imaging predictor sensitivity, specificity, accuracy, and association with IAC using Fisher's exact test. Inter-reader agreement determined using Fleiss' kappa statistics. For diagnosis of IAC, sensitivities and specificities were high (70-93%). Pancreatic abnormality was strongest predictor for distinguishing IAC from PSC and cancer, with high diagnostic performance (70-80% sensitivity, 87-97% specificity), significant association (p < 0.01), and moderate inter-reader agreement (kappa = 0.59). Continuous biliary involvement was moderately predictive (50-100% sensitivity, 53-83% specificity) and trended toward significant association in distinguishing from PSC (p = 0.01-0.19), but less from cancer (p = 0.06-0.62). It remains difficult to distinguish IAC from PSC or bile duct malignancy based on imaging features alone. The presence of pancreatic abnormalities, including peripancreatic rind, atrophy, abnormal enhancement, or T2 signal intensity, strongly favors a diagnosis of IAC.