Aim: To evaluate the accuracy of diagnosing gastric antral lesions in routine clinical practice using magnifying endoscopy with narrow-band imaging (M-NBI) as a real-time diagnosing technique.
Methods: Consecutive patients undergoing upper endoscopy were selected for the study. In each patient, the mucosa of the gastric antrum was observed by M-NBI, and the gastric microstructure was categorized into five types (A-E). Based on these patterns, histological types were predicted in a real-time manner. The accuracy of these predictions was evaluated based on histological findings. Inter-observer agreement was also assessed.
Results: A total of 207 sites in 90 patients were examined by M-NBI. Compared with type A gastric microstructure, types B and C gastric microstructure showed a significantly higher degree of inflammation (P < 0.001). The sensitivity, specificity and accuracy of types B + C microstructure as a predictor of gastric inflammation were 85.4, 81.7 and 83.1 %, respectively. Similarly, the sensitivity, specificity and accuracy of type D microstructure as a predictor of gastric intestinal metaplasia were 71.8, 95.2 and 90.8 %, respectively, and those of type E microstructure as a predictor of early gastric cancer were 80.0, 98.9 and 97.6 %, respectively. The sensitivity and specificity of type B alone, type C alone and types B + C combined for the detection of Helicobacter pylori infection were 52.2 and 87.0 %, 22.8 and 92.2 %, 75.0 and 79.1 %, respectively. The kappa value for the inter-observer agreement was 0.715 (95 % confidence interval 0.655-0.895).
Conclusions: In conclusion, M-NBI can significantly improve the accuracy of the prediction of histopathology of gastric antral lesions in vivo, implying the possibility of using M-NBI as an effective diagnosis technique.