Mo1244 IDENTIFICATION OF VISIBLE LESIONS USING NARROW BAND IMAGING: A VIDEO SURVEY STUDY OF ACADEMIC AND COMMUNITY ENDOSCOPISTS

Abstract

The routine use of Narrow Band Imaging (NBI) has been supported by recent guidelines for improved dysplasia detection during endoscopic surveillance for Barrett’s Esophagus (BE). However, the ability for gastroenterologists to identify neoplastic visible lesions (VL) utilizing NBI remains unclear. We sought to assess the rate at which endoscopists correctly identify VL harboring neoplasia using high-definition white light endoscopy (HD-WLE) and NBI. Academic and community endoscopists who routinely perform surveillance endoscopy for BE were recruited for this video survey study. A total of 25 video clips were randomly selected from our multicenter repository with varying degrees of neoplasia. All clips were standardized to include a 30 second pullback with HD-WLE and NBI, retroflexed view, and distances marked every 1 cm. Endoscopists were asked to identify anatomic landmarks, VL using HD-WLE and NBI, and to identify the correct treatment modality for each VL. Demographics for each endoscopist were collected. A total of 5 blinded experts in BE evaluated each clip to establish a gold standard (17 VL by HD-WLE and 20 VL by NBI). Responses were analyzed using a 90% threshold, established a priori, as a benchmark to assess performance based on published quality metrics data or expert opinion. Assuming a 10% increase in VL detection using NBI based on published data, a total of 40 endoscopists and 20 videos were needed to reach 80% power. Multivariate logistic regression was performed per lesion to assess for predictors of VL identification. Of the 50 invited endoscopists, 44 (88%) completed the survey (Table 1). Interrater reliability for each measure ranged from a κ of 0.73-0.82. A total of 100% (44/44) physicians correctly identified the top of the gastric folds and top of intestinal metaplasia within 2 cm. Compared to the gold standard, participants correctly identified 72% (539/748) VL using HD-WLE and 69% (610/880) using NBI, with only 18% (8/44) and 5% (2/44) reaching the 90% threshold for HD-WLE and NBI, respectively. Community endoscopists had a higher rate of VL detection by NBI (OR 2.07, CI: 1.2-3.5, p<0.008), but they had significantly greater volume of surveillance endoscopies per month (p=0.046, Table 1). VL detection using NBI improved significantly with physicians performing >5 surveillance endoscopies/month (AUC: 0.72; 95% CI: 0.56-0.85; p=0.006, Figure 1). As a whole, participants identified the correct treatment modality for 84% of all VL. The current video based survey study suggests identification of VL by NBI significantly increased with volume of surveillance endoscopies and is independent of practice type, though the expected increase in VL detection with NBI was not observed. Given recent ASGE guidelines, further educational efforts focused on use of NBI for VL detection are needed.View Large Image Figure ViewerDownload Hi-res image Download (PPT)