Observer agreement in the assessment of narrow-band imaging system surface patterns in Barrett's esophagus: a multicenter study.

Background and study aims: The clinical utility of narrow-band imaging (NBI) for Barrett's esophagus is limited by the multiplicity of classification schemes. We evaluated the interobserver agreement and accuracy of a new consensus-driven simplified binary classification of NBI surface patterns. Patients and methods: NBI images from macroscopically normal areas in patients with Barrett's esophagus were retrieved from endoscopy databases and were randomized for review by seven endoscopists (three experts, four nonexperts). A simplified binary classification of NBI mucosal and vascular patterns was used: (1) regular pattern (nondysplastic Barrett's esophagus); (2) irregular pattern (dysplastic Barrett's esophagus). Agreement in relation to surface patterns and predicted histology (dysplasia vs. no dysplasia) was calculated using κ statistics. Results: A total of 252 NBI images from 75 patients with Barrett's esophagus were assessed: 93 showed intestinal metaplasia, 91 low-grade dysplasia, and 68 high-grade dysplasia/esophageal adenocarcinoma. The median score for image quality was 4 (very good). Interobserver agreement for mucosal and vascular patterns and dysplasia prediction was fair: κ = 0.40 (95%CI: 0.37-0.42), 0.36 (95%CI: 0.33-0.38), and 0.39 (95%CI: 0.36-0.41) respectively, with comparable results for experts and nonexperts. Intraobserver agreement was substantially better among experts than among nonexperts, with mucosal κ = 0.63 vs. 0.49, vascular κ = 0.58 vs. 0.43, and predicted histology κ = 0.68 vs. 0.54 (all P < 0.005). Mean sensitivity and specificity of NBI surface patterns for predicting dysplasia were 47% (95%CI: 44%-55%) and 72% (95%CI: 69%-76%) respectively. Conclusions: Using a consensus-driven simplified classification of NBI surface patterns for Barrett's esophagus, the interobserver agreement was fair with suboptimal sensitivity and specificity. Significant improvements in NBI interpretation are needed prior to the routine use of NBI surface patterns for the assessment of Barrett's esophagus. [ABSTRACT FROM AUTHOR]