The reliability of EUS for the diagnosis of chronic pancreatitis: interobserver agreement among experienced endosonographers

Background: Endoscopic ultrasound (EUS) is a minimally invasive, low risk m ethod of diagnosis for chronic pancreatitis (CP). The degree to which endos onographers agree on the features and diagnosis of CP is unknown. For EUS t o be considered an accurate test for CP, there must be good interobserver a greement. Methods: Forty-five pancreatic EUS examinations were videotaped by 3 experi enced endosonographers. Examinations from 33 patients with suspected CP bas ed on typical symptoms, as well as 12 control patients without suspected CP , were included. Eleven experienced endosonographers ("experts") who were b linded to clinical information independently evaluated all videotaped exami nations for the presence of CP and the following 9 validated features of CP : echogenic foci, strands, lobularity, cysts, stones, duct dilatation, duct irregularity, hyperechoic duct margins, and visible side branches. The exp erts also ranked (most to least) which features they believed to be the mos t indicative of CP. Interobserver agreement was expressed as the kappa (kap pa) statistic. Results: There was moderately good overall agreement for the final diagnosi s of CP (kappa = 0.45), Agreement was good for individual features of duct dilatation (kappa = 0.6) and lobularity (kappa = 0.51) but poor for the oth er 7 features (kappa < 0.4), The expert panel had consensus or near consens us agreement (greater than 90%) on 206 of 450 (46%) individual EUS features including 22 of 45 diagnoses of CP. Agreement on the final diagnosis of CP was moderately good for those trained in third tier fellowships (<kappa> = 0.42 +/- 0.03) and those with more than 1100 lifetime pancreatic EUS exami nations (kappa = 0.46 +/- 0.05). The presence of stones was regarded as the most predictive feature of CP by all endosonographers, followed by visible side branches, cysts, lobularity, irregular main pancreatic duct, hyperech oic foci, hyperechoic strands, main pancreatic duct dilatation, and main du ct hyperechoic margins. The most common diagnostic criterion for the diagno sis of CP was the total number of features (median 4 or greater, range 3 or greater to 5 or greater). Conclusions: EUS is a reliable method for the diagnosis of chronic pancreat itis with good interobserver agreement among experienced endosonographers. Agreement on the EUS diagnosis of chronic pancreatitis is comparable to oth er commonly used endoscopic procedures such as bleeding ulcer stigmata and computed tomography of the brain for stroke localization and better than th e physical diagnosis of heart sounds.