The endoscopic assessment of esophagitis: a progress report on observer agreement.

BACKGROUND & AIMS The study and management of reflux esophagitis require an endoscopic classification system founded on esophageal lesions that can be reproducibly identified. The aim of this study was to investigate interobserver agreement for the identification of endoscopic lesions typical of reflux esophagitis. METHODS Paired comparisons of observers' descriptions were obtained. Seventeen endoscopists assessed 100 still images, and 42 endoscopists, including 13 endoscopists in training, assessed 23 endoscopic video recordings. In a third, ancillary study, using a simpler evaluation sheet, 219 gastroenterologists recorded their assessments of 20 still images. RESULTS The agreement between endoscopists was similar for still images and video recordings. Agreement between experienced endoscopists was acceptable to good for recognition of minimal changes (erythema, friability, mucosal edema; kappa = 0.46 to kappa = 0.8), mucosal breaks (discretely, demarcated areas of slough or erythema; kappa = 0.84), and complications (ulceration, kappa = 0.92; stricturing, kappa = 0.80; columnar metaplasia, kappa = 0.81), although there was poor agreement when the circumferential extent and number of mucosal breaks were assessed. However, total circumferential extent of the mucosal break had a kappa value of 0.59. Agreement between inexperienced endoscopists was poor for recognition of minimal changes but was good for recognition of complications (kappa, 0.70-0.90). CONCLUSIONS Endoscopists can identify mucosal breaks confined to a mucosal fold and lesions that extend throughout the esophageal circumference. Complications of reflux disease can be reproducibly recorded. Criteria for assessing the number of mucosal breaks and their radial extent must be defined more clearly, as must the features of minimal change esophagitis.