| Summary: | 碩士 === 國立臺灣大學 === 流行病學研究所 === 97 === Background: Current endoscopic diagnostic criteria for gastro-esophageal reflux disease (GERD) depends on identification of mucosal breaks. However, the diagnosis badly correlates with symptoms and acid reflux. Therefore a refined diagnostic criteria is desirable. Based on personal experience, Haruma et al. proposed that human gastro-esophageal junctional (GEJ) morphology under endoscopy typically varied into 5 categories, and that certain category might be associated with GERD. Prior to prove this association, we plan to define a novel classification of GEJ morphology based on Haruma’s experience. And through category-manipulation and reliability test, we aim to propose a best classification suitable for further studies.
Method: We defined an original 5-category GEJ morphological classification. Two hundreds and fifty standardized endoscopic pictures on GEJ from subjects underwent health check were collected, and subjected to ratings by 7 endoscopists. Categorical prevalence, intra-rater agreement, and inter-rater agreement were calculated. By post-hoc category combination procedures, we repeated the calculations for all 25 possible modified classifications, and compared the result to the original one and each other. The comparison formed the basis for selection.
Result: Intra-rater agreements were kappa between 0.5 and 0.79. Inter-rater agreement of the original classification was presented by average raw agreement of 63.4%, mean kappa of 0.42, and overall weighted kappa of 0.42. Added by 25 modified classifications, a ranking of agreements was made for the total 26. Twelve of them were discarded because they were, by definition, not clinically applicable. Finally 2 classifications were selected, 2-category and the other 4-category. The mean kappa/overall weighted kappa was 0.57/0.55, 0.52/0.51, respectively.
Conclusion: Based on inter-rater agreement and clinical applicability, our study has defined two reliable endoscopic classifications on GEJ. Both classifications are candidates for further association study with GERD.
|
|---|
