P212 Location and Kudo pit pattern reflect neoplastic histology of lesions detected at surveillance colonoscopy in inflammatory bowel disease
Iacucci M.*1,2,3, Oluseyi A.4, Panaccione R.4, Gui X.5, Urbanski S.3, Minoo P.3, Kaplan G.G.3, Novak K.3, Lowerison M.W.3, Lethebe B.C.3, Leung Y.3, Seow C.H.-T.3, Ghosh S.1
1University of Birmingham, Institute of Translational Medicine, Birmingham, United Kingdom 2University of Birmingham, UK, United Kingdom 3University of Calgary, Calgary, Canada 4University of Calgary, Gastroenterology, Calgary, Canada 5University of Calgary, Pathology, Calgary, Canada
Effective colonoscopic surveillance of IBD benefit from having reliable predictors of neoplasia, since targeted biopsies and endoscopic resection are increasingly used as standard of practice. It is not clear whether Kudo pit patterns may be applicable in characterizing IBD associated lesions.
We aimed to identify the specific clinical and endoscopic features of colonic lesions which predict dysplasia in IBD.
All lesions identified in a randomized study to determine the detection rates of neoplastic lesion (NL) in patients with long standing colitis in IBD (ClinicalTrials.gov NCT02098798) were included. Endoscopic NL were classified by the Paris classification and Kudo pit pattern, and by the Vienna classification histologically. Exploratory univariate analysis was performed, and age, duration of disease, extra-intestinal manifestations family or personal history of polyps/cancer, smoking, size of lesion, Paris classification, Kudo pattern, localisation/extension of disease were considered in the patients with IBD- associated NL. Subsequently a multivariate logistic regression model analyses was created with candidate variables which had p values ≤0.05 based on univariate analysis
A total of 270 patients (55% men; age 20–77y) were assessed by High Definition – white light (n=90), virtual chromoendoscopy (n=90) or dye chromoendoscopy (n=90). Ninety-one (33.7%) colonic dysplastic lesions and 1 adenocarcinoma were found. Sixty-two (68.8%) were polypoid and twenty-nine (31.8%) were non polypoid. Most of these lesions (92.3%) had Kudo pattern III–V (Table). By univariate analysis, age – Odds Ratio (OR) 1.05 (95% CI: 1.02–1.08), localization of the lesions in the right colon – OR 6.15 (95% CI: 3.12–12.12), Kudo pattern IIO, III-IV and V – OR 20.91 (95% CI: 9.34–46.7) and Paris Is/Ip classification OR – 3.29 (95% CI: 1.69–6.38) were associated with NL. Subsequently proportional multivariate logistic regression model for the prediction of colonic neoplasia confirmed that the endoscopic Kudo pit pattern – OR 21.50 (95% CI: 86.5–60.1) and localization of the lesions in the right colon – OR 6.52 (95% CI: 1.98–22.5) were predictors of colonic neoplasia at surveillance colonoscopy in IBD (Table). The overall accuracy of independent variables which predict neoplastic changes was 78% (95% CI 68–88%), sensitivity 82% (95% CI 68–97%), specificity 68% (95% CI 47–89%), PPV 85% (95% CI 76–95%) and NPV64% (95% CI 42–86%) which were significant in the multivariate analysis.
We demonstrated that the endoscopic Kudo pattern and localisation of the lesions in the right colon were predictors of colonic neoplasia in IBD. This may guide management strategy of NL detected at IBD surveillance.