Real-Time Characterizations of Diminutive Colorectal Polyps Using Narrow Band Imaging

Overview

Colorectal cancer is the second leading cause of cancer related death in the United States. Colonoscopy is the most commonly performed screening procedure and diminutive polyps (\<5mm) are the most commonly found polyps during colonoscopy. Although these polyps have a very low risk of harboring malignancy, they are routinely removed to determine surveillance intervals. Narrow Band Imaging is equipped on widely available colonoscopes and in expert hands can allow accurate real-time optical histologic diagnosis of colorectal polyps. If this practice can be applied widely, there is significant potential for cost savings. This has led to a 'characterize, resect and discard' strategy where polyps determined to be hyperplastic (benign with no neoplastic potential) can be left in place and those determined to be adenomatous (have neoplastic potential) can be resected and discarded. It is unclear if endoscopists without prior expertise or training in Narrow Band Imaging can achieve adequate diagnostic accuracy to put 'characterize, resect and discard' into wide practice.

Gastroenterologists without prior training in NBI from two affiliated academic hospitals will participate in an ex-vivo training session in which they will view a short audiovisual tool describing previously validated NBI criteria to determine polyp histology, followed by reviewing 80 videos of diminutive polyps under NBI and will record predicted polyp histology and degree of confidence. After each video, targeted feedback regarding actual polyp histology and NBI criteria supporting the diagnosis will be provided. Participants will then employ NBI in real-time colonoscopy (in-vivo) and record predicted polyp histology, degree of confidence and predicted surveillance intervals based on NBI interpretations. Each study polyp will be sent for histology separately. Performance will be assessed by comparing predicted histology with actual histology. Structured performance feedback will be given to promote practice-based learning, establish a real-time learning curve and determine the number of observations required to achieve competency in-vivo. The primary aim of this study was to determine whether endoscopists with no prior experience or training in Narrow Band Imaging can achieve the thresholds set forth by the American Society of Gastrointestinal Endoscopy: For diminutive colorectal polyp diagnoses made with 'high-confidence,' a (1) greater than or equal to 90% negative predictive value in the rectosigmoid colon and a (2) greater than or equal to 90% agreement in surveillance intervals predicted by narrow band imaging and those based on the current gold standard of histology. The secondary outcomes for the in-vivo phase included (1) evaluating overall group performance (accuracy, sensitivity, specificity, predictive values) of optical diagnoses using NBI based on degree of confidence and location within the colon, (2) evaluating individual performance on the ASGE benchmarks, (3) determining predictors of performance, and (4) determining real-time learning effect in the setting of ongoing, structured performance feedback. The secondary outcomes for the ex-vivo (training) phase included (1) evaluating overall performance by degree of confidence, (2) determining predictors of performance during training and (3) evaluating a learning effect in the setting of ongoing feedback during training. Sample size was calculated to show an NPV of 90% or higher assuming that the true NPV is 95% for rectosigmoid polyps characterized with "high-confidence," based on 26 participating endoscopists and within-endoscopist correlation of 0.05. This will require 336 total rectosigmoid non-adenoma polyps characterized with "high-confidence," and assuming approximately 22% rectosigmoid polyps, 70% with high confidence and 80% hyperplastic, the study will require approximately 2,727 polyps and 1,364 colonoscopies in total, assuming approximately two polyps per colonoscopy.