Su1516 Which Visual Characteristics Identify Proximal Sessile Serrated Polyps? A Quantitative Analysis of Digital Video-Clips of 160 Polyps

15-20% of incident colorectal cancers and probably many “missed” interval cancers arise from sessile serrated polyps (SSPs). Endoscopically, SSPs are often inconspicuous, and differ in appearance from adenomatous polyps (APs). APs are more prevalent than SSPs and colonoscopists are more familiar with their appearance and neoplastic potential. The endoscopic characteristics of SSPs during non-enhanced colonoscopy have not been analyzed in detail; the current study addresses this gap in knowledge with the hope of increasing SSP detection rate by colonoscopists. We systematically reviewed video clips of 160 unique SSPs located proximal to the splenic flexure. All SSPs were video recorded digitally during high-resolution, low definition, non-enhanced colonoscopy. All patients underwent bowel preparation using a 2-L polyethylene glycol, split-dose regimen. By consensus of the investigators, 9 striking visual characteristics of SSPs were chosen, and verbal descriptors were assigned for each characteristic. Inter-observer agreement was assessed by Kappa statistics. We then applied the descriptors in two ways. (1) To determine the prevalence of each characteristic in the SSP population, 2 high-SSP detectors (Freedman et al, A J Gastroenterol 2010; Vol 105 Supplement 1: S563) independently scored the video clips for the presence or absence of each characteristic. (2) To determine the most frequent “sentinel sign(s)” (the polyp characteristic that first strikes the observer), the colonoscopists reviewed an assortment of SSP, AP and “no polyp” videos and selected one descriptor by consensus as the polyp's sentinel sign. We report an interim analysis of 52 polyp videos. For 8 of 9 polyp descriptors, good or excellent inter-observer agreement was achieved (Table 1); the other descriptor—“surface telangiectasias”—did not achieve this level of agreement (Kappa < 0.65). The prevalence of 8 validated descriptors in the population of SSPs ranged from 12.5% to 52.5% (Table 1). The frequency distribution of sentinel signs differed between the populations of SSPs and APs (Table 2). The colonoscopists accurately identified all “no polyp” videos. The most frequent sentinel signs were “distortion of the mucosal fold” for SSPs and “mound-like elevation” for APs.Table 1Inter-observer agreement and prevalence of descriptors (20 SSPs)DescriptorKappa CoefficientPrevalence⁎Each lesion could score positive for > 1 descriptorRim of Debris or Bubbles (≥25%)0.80045.0%Nodularity1.00015.0%Bile-stained (≥50%) Mucus Coating0.87547.5%Clear Mucus Coating0.90027.5%Red/Pink Color Difference0.77312.5%Interruption of a local blood vessel0.68837.5%Mound-like elevation0.89032.5%Distortion of a Mucosal Fold0.90052.5% Each lesion could score positive for > 1 descriptor Open table in a new tab Table 2Frequency distribution of sentinel signsDescriptorSSP (n=15)AP (n=17)Rim of Debris or Bubbles (≥25%)13.3%0.0%Nodularity0.0%5.9%Bile-stained (≥50%) Mucus Coating13.3%0.0%Clear Mucus Coating13.3%0.0%Red/Pink Color Difference0.0%5.9%Interruption of a local blood vessel0.0%5.9%Mound-like elevation13.3%70.6%Distortion of a Mucosal Fold46.7%11.8% Open table in a new tab Close analysis of video clips allows identification of specific visual characteristics of proximal SSPs during routine colonoscopy. Verbal descriptors for these visual characteristics can be assigned, validated, and applied to populations of SSPs and comparator polyps. This method of analysis could provide better insight into (1) the distinct appearances of SSPs and (2) the characteristics that allow colonoscopists to detect them.