HIGHLIGHTS
- The study evaluates the Japanese Esophageal Society classification for superficial esophageal squamous cell carcinoma.
- Interobserver agreement was moderate (κ=0.455), similar for experts and novices.
- Type B2 had the highest sensitivity (75%), but overall accuracy was 47.5%.
- Extensive training is needed for accurate magnifying endoscopy image evaluation.
ABSTRACT
Background –Accurate evaluation of the invasion depth of superficial esophageal squamous cell carcinoma (SESCC) is crucial for optimal treatment. While magnifying endoscopy (ME) using the Japanese Esophageal Society (JES) classification is reported as the most accurate method to predict invasion depth, its efficacy has not been tested in the Western world. This study aims to evaluate the interobserver agreement of the JES classification for SESCC and its accuracy in estimating invasion depth in a Brazilian tertiary hospital. Methods – We retrospectively selected ME with Narrow Band Imaging (ME-NBI) images of 30 suspected SESCC cases. The best images of each case were included in online forms, which were evaluated by ten endoscopists (five experts and five novices). The evaluators classified the lesions according to the JES-IPCL classification and estimated the depth of invasion. Interobserver agreement was assessed using kappa values. Histological comparison was possible for 17 lesions. Results – The overall interobserver agreement for the JES-IPCL classification was moderate (κ=0.455, P<0.001). Agreement among experts (κ=0.437) and novices (κ=0.483) was also moderate. Sensitivity, specificity, and accuracy for IPCL types were: B1 (41.3%, 78.9%, 59.9%), B2 (75%, 66.7%, 68.7%), and B3 (46%, 91.7%, 78.6%). Overall accuracy of the JES classification for estimating depth of invasion was 47.5%. Conclusion – The moderate interobserver agreement suggests the JES-IPCL classification may be useful in the Western world, but extensive training is needed. The findings indicate a longer learning curve for accurate ME-NBI image evaluation using the JES classification.
AUTORES
Fauze MALUF-FILHO1,2, Ossamu OKAZAKI1,2, Beanie Conceição Medeiros NUNES1,2, Adriana Vaz SAFATLE-RIBEIRO1,2, Luciano LENZ1,2 and Bruno Costa MARTINS1,2

