Interobserver Agreement for Pancreatic Endoscopic Ultrasonography Determined by Same Day Back-to-Back Examinations

Abstract

We aimed to determine interobserver agreement between endosonographers for pancreatic morphology by performing same day back-to-back endoscopic ultrasonography (EUS) examinations on patients without known hepatopancreaticobiliary disease. Morphologic EUS examination is often used to make the diagnosis of chronic pancreatitis, although perceived lack of interobserver agreement limits it accuracy. A prospective clinical trial was performed in patients without known hepatopancreaticobiliary disease referred for EUS. All subjects underwent back-to-back same day EUS examinations at our institution performed by 2 blinded endosonographers. The main outcome measurement was the Cohen κ scores between endosonographers calculated for each parenchymal and ductal EUS feature. Twenty-four patients were male and indications for EUS included staging of esophageal malignancy (10), submucosal mass (9), lymphadenopathy (7), gastrointestinal stromal tumor (4), and other (14). Both endosonographers agreed that 32% (14 of 44) had hyperechoic strands, 30% (13 of 44) had hyperechoic duct walls, 16% (7 of 44) had hyperechoic foci, 14% (6 of 44) had a dilated main pancreatic duct, 9% (4 of 44) had parenchymal lobularity, and 5% (2 of 44) had parenchymal cysts. Kappa scores showed "good" or better correlation only for the presence of hyperechoic strands and parenchymal cysts; the presence of parenchymal lobularity, a dilated main pancreatic duct, and hyperechoic foci had "moderate" correlation. The correlation for the presence of hyperechoic foci was "fair." Same day back-to-back EUS examinations on patients without known hepatopancreaticobiliary disease showed a wide variation in interobserver agreement for standard pancreatic morphologic findings. These results suggest the need for improvements in the current EUS classification system of chronic pancreatitis.