Endoscopic Trimodality Imaging (ETMI) for Detection and Classification of Neoplasia in Patients with Longstanding Ulcerative Colitis

Abstract

Background and Aims: Patients with longstanding ulcerative colitis (UC) are recommended to undergo regular colonoscopic surveillance with targeted and random biopsies for the detection of neoplasia. However, the sensitivity of this method is low. Recently, Endoscopic TriModality Imaging (ETMI) has been developed, incorporating high resolution White Light Endoscopy (WLE), AutoFluorescence Imaging (AFI) and Narrow Band Imaging (NBI). We performed a pilot study, in which we assessed the feasibility and accuracy of ETMI and compared AFI with WLE for the detection of neoplasia in patients with longstanding UC. Methods: Consecutive UC patients, scheduled for surveillance colonoscopy, were invited to participate in this study. During withdrawal, each colonic segment was inspected twice; once with WLE and once with AFI, in randomized order. All detected lesions were subsequently inspected by NBI for assessing Kudo classification. Hereafter, targeted as well as random biopsies were taken. The histology of the biopsies was used as the gold standard. Results: Informed consent was obtained from 58 patients. Three patients had to be excluded because of insufficient colon preparation and 5 because of active inflammation despite a low disease activity index. Among 25 patients randomized to start with AFI, 8 neoplastic lesions were found with AFI (in 4 patients). Second inspection with WLE revealed 1 extra neoplastic lesion (1 additional patient). Among 25 patients examined with WLE first, 4 neoplastic lesions were found with WLE (2 patients). Subsequent AFI yielded 3 additional neoplastic lesions (2 extra patients). A mean number of 40 (SD 7) random biopsies was taken per patient, revealing neoplasia in 4 biopsies (one additional patient with neoplasia in whom WLE and subsequent AFI were both normal). The sensitivity of AFI and WLE for the detection of patients with neoplasia was 75% and 50% respectively (p = 0.580). The positive predictive values were 21% and 12% respectively (p = 0.470). The Kudo classification based on NBI for predicting histological outcome of detected lesions had a sensitivity of 75%, specificity of 80%, positive predictive value of 43% and negative predictive value (NPV) of 94%. A detected lesion which was green with AFI never revealed neoplasia. Conclusion: The use of ETMI in patients with UC is feasible, provided that the colon is properly prepared and the mucosa is not actively inflamed. In this pilot study, the sensitivity of AFI seems to be better than WLE for the detection of neoplasia, but the small sample size of our study prevents drawing any firm conclusion. Once a lesion has been detected, ETMI has a high NPV for histology.