OBJECTIVES: The purpose of this study was to determine whether the choice for performance of endoscopic retrograde pancreatography (ERP) could be tailored to findings on computed tomography (CT) in patients with suspected autoimmune pancreatitis (AIP). METHODS: Eighty-four AIP patients and 73 pathology-proven pancreatic cancer patients from a prospectively maintained database were retrospectively included. Computed tomography and ERP images were reviewed in consensus by 2 blinded radiologists. The diagnostic performance of CT alone and combined use of CT and ERP (CT-ERP) were compared. RESULTS: The area under the receiver operating characteristic curve of CT-ERP was significantly greater than that of CT alone (0.97 vs 0.87, P < 0.001). When patients with AIP were divided into 2 subgroups according to CT features (typical vs atypical), 24 (69%) of 35 AIP patients with atypical CT findings were correctly diagnosed with AIP at CT-ERP and received benefits from additional ERP. Endoscopic retrograde pancreatography had little added benefit in patients with typical CT findings for AIP (n = 49), because no alternative diagnoses were established after ERP. CONCLUSIONS: In patients with suspected AIP, the decision to perform ERP could be tailored to findings on CT.