First report from International Evaluation of Endoscopic classification JNET: an international multicenter web trial.

AIMS: Narrow-band imaging (NBI) contributes to real-time optical diagnosis and classification of colorectal lesions. The Japan NBI Expert Team (JNET) was introduced in 2011. The aim of this study is to explore the diagnostic accuracy of JNET when applied by European and Japanese endoscopists not familiar with this classification. METHODS: The study was conducted by 36 ESGE and 49 JGES non-JNET endoscopists using still images of 150 lesions. For each lesion, non-magnified white-light, non-magnified NBI, and magnified NBI images were presented. In the magnified NBI, the evaluation area was designated by ROI. The endoscopists scored histological prediction for each lesion. RESULTS: In ESGE members, the sensitivity, specificity, and accuracy were respectively 73.3%, 94.7% and 93.0% for JNET type 1; 53.0%, 64.9% and 62.1% for type 2A; 43.9%, 67.7% and 55.1% for type 2B; and 38.1%, 93.7%, 85.1% for type 3. When type 2B and 3 were considered as one category of cancer, the sensitivity, specificity, and accuracy for differentiating high-grade dysplasia and cancer from the others were 59.9%, 72.5%, and 63.8%, respectively. These trends were the same for JGES endoscopists. CONCLUSIONS: The diagnostic accuracy of the JNET classification was similar between ESGE and JGES and considered to be sufficient for JNET type 1. On the other hand, the accuracy of for Type 2 and 3 is not sufficient, however, JNET 2B lesions should be resected en-bloc due to the risk of cancers and JNET 3 can be treated by surgery due to its high specificity.