Interobserver Agreement of a Simplified Classification System of Mucosal Morphology Using Narrow Band Imaging in Barrett's Esophagus; An International Multi-Center Study

Abstract

Introduction: Three different classifications of mucosal morphology in Barrett's esophagus (BE) based on narrow band imaging (NBI) have been published. In order to achieve a uniform evaluation of mucosal morphology in clinical practice we propose a more simplified classification based on these publications. This simplified classification consists of the evaluation of the regularity of the mucosal and vascular patterns .The aim of this study was to assess the interobserver agreement and correlation with histology of this simplified classification on mucosal morphology in BE combining the experience of 2 international expert centers in this field. Methods: In both centers magnified NBI images with corresponding histology were obtained from BE using a RGB-based NBI-endoscopy system (GIF-Q240Z) and stored in a prospective database. From this database 200 NBI-images (99 nondysplastic BE, 44 low-grade intra-epithelial neoplasia (LGIN) and 57 high-grade intra-epithelial neoplasia (HGIN)) were selected. Images were evaluated by 4 NBI-expert endoscopists and 4 endoscopists with no specific expertise in this field; all assessors were blinded to the pathology results. The evaluation was preceded by an automated instruction explaining the proposed simplified classification. The endoscopists assessed 4 items for each image: quality of the image on a 10-cm visual analogue scale (VAS), suspicion for dysplasia, and regularity of the mucosal and vascular patterns. The interobserver agreement was assessed using kappa statistics. Results: The mean VAS score for image quality was 4 (SD 1). Overall interobserver agreement on the suspicion for dysplasia was moderate (k 0.44 (0.40-0.47)). Overall interobserver agreement on the regularity of the mucosal and vascular pattern was 0.42 (0.38-0.46) and 0.42 (0.39-0.46) respectively. There were no significant differences in agreement between expert and non-expert endoscopists. The sensitivity of the endoscopists for diagnosing HGIN was 71% and the specificity 68%. This did not significantly differ between expert and non-expert endoscopists (p = 0.43). In 66% of the evaluations from areas with HGIN, both patterns (mucosal and vascular) were scored irregular and in 71% of the evaluations, at least one of the patterns was scored irregular. Conclusion: Interobserver agreement of the simplified classification was moderate. There was no significant difference in interobserver agreement between expert and non-expert endoscopists suggesting a short learning curve for the proposed classification. The disappointing accuracy for predicting dysplasia questions, however, if detailed inspection of BE with NBI can replace histological sampling.