Accuracy and inter-observer agreement of the nice and kudo classifications of superficial colonic lesions: a comparative study.

Abstract

For superficial colonic lesions, the NICE and Kudo classifications are used in the in vivo prediction of histology and as guide to therapy. The NICE system derives information from unmagnified NBI endoscopic images. The Kudo one necessitates a magnification, but, as this tool is not commonly available, it is applied also to characterize unmagnified pictures to compare their diagnostic performances. We conducted a prospective comparison of the NICE versus the Kudo classification for the differential diagnosis of colonic polyps taking histology as the gold standard. The inter-observer agreement for both classifications among 11 colonoscopists was also evaluated. Short unmagnified NBI videoclips of 64 colonic polyps were sent twice to the participants. In the first round, they classified the lesions according to the NICE classification; 4 months later, the same videos were assessed with the Kudo system. The diagnosis provided by the participants was grouped in non-neoplastic, non-invasive neoplasia, invasive neoplasia. Overall, the diagnostic accuracy was 82% (95%CI: 79-85) with the NICE system and 81% (95%CI: 78-84) with the Kudo one (ρ = 0.78). The accuracy of the NICE classification for non-neoplastic lesions was greater compared with the Kudo's (ρ = 0.03). Sensitivity sub-analyses revealed a higher ability of the NICE in distinguishing between neoplastic vs. non-neoplastic lesions (ρ = 0.01). The overall inter-rater agreement did not differ when the classifications were compared. The NICE and the Kudo classifications might be considered comparable. Our data could allow the use of the NBI Kudo classification even in those centers where magnification is not available.