PTH-069 Validating a classification system using ISCAN optical enhancement for detection of early barrett’s oesophagus neoplasia

IntroductionMagnification endoscopy (ME) offers an additional tool for endoscopists undertaking BE surveillance to guide biopsies or direct endoscopic eradication therapy (EET) for early neoplasia. Mucosal patterns and oesophageal mucosal microvasculature can be interrogated. We propose a simple classification based on mucosal (M) and vascular (V) patterns, to differentiate dysplastic BE (DBE) mucosa from non-dysplastic (NDBE) mucosa. We assessed if magnification endoscopy paired with a new imaging platform, iScan OE (Pentax, Hoya, Japan) improves detection of early BE associated neoplasia. We validate our MV classification with this new platform.MethodsPatients undergoing BE surveillance from Jan 2016-Nov 2017, were recruited from 3 international centres. ME was used to image the mucosa at up to 136x magnification. Endoscopists targeted normal and abnormal areas suspicious for neoplasia. Where possible, each area was imaged in HD-WLE and OE and matched histology was acquired from recorded areas. 7 experts blinded to histology classified HD images using our MV classification as NDBE or DBE. Mucosal patterns were graded as M1 (regular/gyric pits); M2 (irregular/featureless pits). Vascular patterns were graded as V1 (regular, normal calibre); V2 (dilated, irregular). M1V1 was considered NDBE; M2V1, M1V2 or M2V2 as DBE. Classifications were compared with histology. Accuracy, sensitivity, specificity were calculated. Krippendorf’s alpha was used to assess interobserver agreement (IA).Results54 patients were recruited: 63 HD-WLE and 90 OE magnified images were obtained. The proportion of dysplastic areas imaged between WLE and OE was similar (29/63 vs 49/90 p=NS). Using the MV classification, experts classified tissue as DBE or NDBE with 66.3%(±0.05) accuracy using HD-WLE vs 80.1% (±0.02) with OE [65–76% p