Inter-Observer Agreement Amongst Non-Expert Endoscopists for Mucosal Morphology in Barrett's Esophagus

Abstract

Introduction: Mucosal and vascular patterns (e.g. mucosal morphology) can be used to differentiate non-dysplastic Barrett's esophagus (BE) from high-grade dysplasia (HGD) or early cancer (EC). We have recently investigated the mucosal morphology of HGD/EC and non-dysplastic BE using narrow band imaging (NBI) and proposed a 3-step classification for the evaluation of BE: determination of the regularity of the mucosal pattern, the regularity of the vascular pattern and the presence of abnormal blood vessels. The aim of this study was to evaluate the inter-observer agreement of this 3-step mucosal morphology classification model amongst non-expert endoscopists and to evaluate the appreciated difference in image quality between NBI and white light endoscopy (WL). Methods: 7 endoscopists from non-university hospitals with no specific experience in BE or advanced imaging techniques independently evaluated magnified still images of 50 areas of BE. They scored the overall, mucosal and vascular imaging quality on a 15 cm VAS. In addition, the 3 step classification system was scored. The endoscopists first assessed the WL images, followed by the corresponding NBI images (in random order), and finally evaluated the WL images and NBI images together. Results: Assessed independently, the overall imaging quality of WL received a higher rating than NBI (mean VAS score 11.9 vs. 11.0; p < 0.05). When WL and NBI were assessed side-by-side, however, NBI images were scored as having a better overall imaging quality in 69% of the assessments, a better mucosal imaging quality (75%) and vascular imaging quality (83%). The inter-observer agreement of the 7 endoscopists for the items of the 3-step classification system was fair to moderate (kappa 0.37 to 0.53). The mean kappa scores of WL, NBI or WL-NBI for scoring the 3-step classification system and the sensitivities for detecting HGD/EC using the 3-step classification model are shown in Table 1. Conclusion: NBI was appreciated as a better imaging modality for magnified still images of BE when directly compared with WL. Inter-observer agreement of non-expert endoscopists for assessing the 3-step classification model was reasonable. In the setting of this study, however, the addition of NBI to WL did not improve interobserver agreement or sensitivity of detecting HGD/EC. Tabled 1Table 1 HRE NBI HRE-NBI Kappa for regular vs irregular mucosal pattern 0.53 0.51 0.53 Kappa for regular vs irregular vascular pattern 0.50 0.44 0.44 Kappa for presence abnormal blood vessels 0.41 0.37 0.41 Sensitivity for HGD/EC 0.84 0.77 0.86 Open table in a new tab