Validation of simplified classification of magnifying endoscopy for diagnosis of Barrett's dysplasia with blue laser imaging

摘要

A new simplified classification of magnifying endoscopy for diagnosis of Barrett's dysplasia has been developed by Barrett's international narrow band imaging group. Further validation is needed to investigate its feasibility. The aim of our study was to evaluate the inter-observer agreement, accuracy, sensitivity, and specificity for the new classification system with a novel blue laser imaging modality. Total of 12 endoscopists (6 more experienced and 6 less experienced) from five medical centers in Taiwan joined this endoscopic training program, which was composed of three sessions: pretest, educational, and post-test. A set of 80 endoscopic images with non-dysplastic or dysplastic Barrett lesions was used for these tests. All raters were asked to predict presence or absence of dysplasia based on the newly developed classification after an educational module. The overall inter-observer agreement (kappa value) before and after the educational module were 0.4419 and 0.5573. The inter-observer agreement before and after training were 0.5471 and 0.5837 in more experienced group, 0.3625 and 0.5499 in less experienced group, respectively. The overall accuracy, sensitivity, and specificity values before and after training did not change significantly, except for increased specificity in more experienced group from 0.7083 to 0.7526. Our validation study for diagnosis of Barrett's dysplasia by using simplified classification with blue laser imaging showed moderate inter-observer agreement and good accuracy. After an educational module, inter-observer agreement in both groups and specificity in more experienced group got improved in a different scale.