Sa1426 DIAGNOSTIC YIELD AND CLINICAL PREDICTORS OF SIGNIFICANT FINDINGS OF ENDOSCOPIC ULTRASOUND IN EVALUATING ASYMPTOMATIC COMMON BILE DUCT DILATION WITH NORMAL LIVER FUNCTION TESTS

Abstract

While EUS is highly sensitive and specific for the evaluation of common bile duct (CBD) dilation, its yield in asymptomatic patients with normal liver function tests (LFTs) is unclear. To assess the diagnostic yield of EUS, to identify predictors of a pathologic finding, and to create a risk stratification model to aid clinical decision making in patients with incidental CBD dilation found on imaging with normal LFTs. We performed a retrospective review of patients with unexplained CBD dilation on abdominal ultrasound, CT, or MRI/MRCP who underwent EUS at an academic medical center between 2009 and 2019. After excluding patients with right upper quadrant pain, known pancreaticobiliary disease, obvious cause of CBD dilation on imaging, or elevated LFTs, n=288 patients were identified for analysis. Univariate and multivariate analyses were used to identify predictors of findings on EUS. We used backward stepwise selection based on the corrected Akaike information criterion to select the variables included the multivariate analysis. Using the results of our multivariate analysis, we created a risk prediction model to determine the probability of a pathologic finding on EUS. Of the 288 patients, 29 (10.1%) had significant findings: 21 had choledocholithiasis or sludge (7.3%), 3 had ampullary pathology (1%), 2 had head of pancreas masses (0.7%), and 3 had other causes (1%). On univariate analysis, factors associated with a pathologic finding on EUS were age ≥70 (OR 3.31 (1.45, 7.56)), non biliary-type abdominal pain (OR 2.3 (1.05, 5.1)), CBD ≥14 mm on index imaging (4.91 (2.11, 11.45)), and history of ERCP (3.27 (1.32, 8.11)). Using multivariate regression, predictive factors affecting a diagnostic EUS were age ≥70 (OR 3.79 (1.53, 10.12)), non-biliary type abdominal pain (OR 3.28 (1.29, 8.95)), CBD ≥14 mm on index imaging (6.46 (2.53, 17.64)), history of ERCP (OR 3.04 (0.93, 9.18)), and history of cholecystectomy (OR 0.43 (0.16, 1.14). The overall AUC of the model was 0.797. Using the coefficients for each variable, a score was assigned to each to create a clinical prediction model to predict the chance of a pathologic finding on EUS: age ≥70 = 1, non-biliary type abdominal pain = 1, history of ERCP = 1, CBD ≥14 mm on index imaging = 2, and history of a cholecystectomy = -1. A total score of ≤1 predicts a <5% chance of having a significant finding (Figure). Only 10% of patients undergoing EUS for asymptomatic CBD dilation had pathologic findings. Age ≥70, CBD ≥14 on cross sectional imaging, non-biliary type abdominal pain, and a history of prior ERCP predicted a pathologic finding on EUS, while history of a cholecystectomy predicts no finding on EUS. We created a novel clinical prediction model to aid in the utilization of EUS in this context which will require independent, external validation.Figure 1Predicted probability of a significant finding on EUS based on clinical prediction model score.View Large Image Figure ViewerDownload Hi-res image Download (PPT)