Diagnostic value of MRI/MRCP direct and indirect signs for identifying biliary etiology in first-episode acute pancreatitis: A two-center external validation study.

BACKGROUND: Identifying biliary etiology during the index admission is important in acute pancreatitis (AP) because it may influence endoscopic evaluation, cholecystectomy timing, and recurrence prevention. Ultrasound and liver biochemical tests remain first-line tools, but they may be inconclusive in some patients. MRI with magnetic resonance cholangiopancreatography (MRCP) can show pancreaticobiliary anatomy when further ductal or biliary assessment is clinically needed. In this setting, the incremental diagnostic value of indirect MRI/MRCP signs, particularly bile signal abnormality, remains insufficiently defined. METHODS: This two-center retrospective diagnostic study included consecutive adults with first-episode AP who underwent non-contrast abdominal MRI with MRCP during hospitalization. Center 1 served as the development cohort and Center 2 served as an independent external validation cohort. Two abdominal radiologists independently assessed predefined direct and indirect MRI/MRCP features while blinded to final etiology. Individual signs, a direct-sign model, an indirect-sign model, a direct plus indirect MRI/MRCP model, and secondary clinical or clinical-MRI models were evaluated. Diagnostic indices were reported with 95 % confidence intervals, and interobserver agreement was summarized with kappa statistics or intraclass correlation coefficients (ICCs), as appropriate. RESULTS: The development cohort included 322 patients, 183 with ABP and 139 with non-biliary AP; the external validation cohort included 144 patients, 83 with ABP and 61 with non-biliary AP. Bile signal abnormality was more frequent in ABP than in non-biliary AP in both cohorts (development, 79.2 % vs 36.7 %; validation, 72.3 % vs 19.7 %; both P < 0.001). In multivariable MRI analysis, gallbladder stones, bile signal abnormality, gallbladder wall edema/thickening, and common bile duct diameter were independently associated with ABP. The direct plus indirect MRI/MRCP model achieved AUCs of 0.929 in development and 0.920 in external validation. In external validation, this model achieved sensitivity of 89.2 %, specificity of 77.0 %, PPV of 84.1 %, NPV of 83.9 %, and accuracy of 84.0 %. The clinical-MRI model had a validation AUC of 0.932, without a clinically meaningful improvement over the MRI-based strategy. CONCLUSION: In first-episode AP patients who underwent clinically indicated MRI/MRCP, combined assessment of direct and indirect MRI/MRCP signs helped identify biliary etiology. Bile signal abnormality provided diagnostic information beyond visible stones. The findings support structured radiology reporting after MRI/MRCP has been obtained, but they should not be interpreted as support for routine MRI/MRCP screening in all AP admissions.