Background and study aim: The establishment of precise and valid diagnostic criteria is important for any disease. We determined the interobserver reliability in the endoscopic diagnosis and grading of Barrett's esophagus.
Patients and methods: Video clips of endoscopy in 21 patients with/without Barrett's esophagus were used for training (n = 3) and for diagnosis/grading (n = 18) of Barrett's esophagus by endoscopists from seven hospitals in Asia. Barrett's esophagus was graded using the Prague C & M Criteria whereby the circumferential extent of the Barrett's segment (C value), maximum extent of Barrett's segment (M value), location of the gastroesophageal junction, and location of the diaphragmatic hiatus were scored. The intraclass correlation coefficients (ICC) were calculated as a measure of interobserver reliability.
Results: A total of 34 endoscopists participated. ICC values for the scores of the C value, M value, location of the gastroesophageal junction, and location of the diaphragmatic hiatus were: 0.92 (95 % confidence interval [CI] 0.88 - 0.97), 0.94 (95 %CI 0.90 - 0.98), 0.86 (95 %CI 0.78 - 0.94), and 0.81 (95 %CI 0.71 - 0.92), respectively, indicating excellent interobserver agreement. The differences in region/country, endoscopists' experience, case volume of participating centers, or primary practice type had no significant effect on the reliability. The ICC values for recognition of Barrett's esophagus of > or = 1 cm were 0.90 (95 %CI 0.80 - 1.00) and 0.92 (95 %CI 0.87 - 0.98) for the C and M values, respectively, whereas the corresponding ICC values for Barrett's segment of < 1 cm were 0.18 (95 %CI 0.03 - 0.32) and 0.21 (95 %CI 0.00 - 0.51), respectively.
Conclusions: Despite the uncommon occurrence of Barrett's esophagus in Asia, our endoscopists exhibited excellent agreement in the endoscopic diagnosis and grading of Barrett's esophagus using the Prague C & M Criteria. However, in view of the low interobserver reliability in recognizing Barrett's segments of < 1 cm, future studies in Asia should take this into account when selecting the study population.
Copyright Georg Thieme Verlag KG Stuttgart . New York.