DEVELOPMENT AND INITIAL EVALUATION OF THE VALID-CLASSIFICATION FOR PREDICTION OF COLORECTAL POLYP HISTOLOGY - A MULTICENTER INTERNATIONAL TRIAL

Aims The recently introduced VIST chromoendoscopy technique allows for detailed analysis of surface and vascular pattern morphology. The study aimed to create a new and unique classification for differentiating colorectal polyps using the VIST technology. We additionally aimed to assess the new classification’s interobserver concordance, and evaluate its accuracy in predicting colorectal polyp histology.

Methods A digital library containing 26 videos/still images from 26 histologically-confirmed polyps with VIST was first evaluated by seven endoscopists from five countries in order to identify possible descriptors. Secondly, the descriptors were categorized using a modified Delphi methodology to propose the VALID-classification. Later, the seven endoscopists independently reviewed a new dataset comprising of 45 videos/still images, providing for each a diagnosis based on the VALID-classification, in which the interobserver agreement of individual descriptor was assessed, and its accuracy in predicting colorectal polyp histology evaluated.

Results Using a modified Delphi process, the endoscopists agreed on summarizing 8 descriptors into three main domains. Those include polyp surface pattern morphology (regular, yes/no; irregular, yes/no; ulcerated, yes/no), and pit pattern morphology (regular, yes/no; tubular appearance, yes/no; villous appearance, yes/no; tubulovillous appearance, yes/no, no pits visible yes/no). Interobserver reliability calculated in Gwet’s AC1 for the polyp surface and pit pattern morphology were 0.78 (0.57,0.82, 95%CI, the same as below) and 0.56 (0.46,0.67) respectively. The accuracy of the VALID-classification in predicting diagnosis (Non-neoplasia, Adenoma and Cancer) was 0.90 (0.84, 0.94), and 0.70 (0.63, 0.77) in predicting histology (Hyperplastic Polyp, Tubular Adenoma, Tubulovillous Adenoma, Villous Adenoma and Cancer).

Conclusions The new VALID-classification was developed allowing prediction of colorectal polyp histology using the newly introduced VIST chromoendoscopy. A good concordance was shown among the observers and a rather high accuracy of the VALID-classification has also been demonstrated in the initial video/still-image evaluation. Further in vivo trials are required to validate this new classification in a clinical setting.