Not All Visual Data Acquired During Withdrawal Phase (WP) of Colonoscopy Is Interpretable. Introducing the Concept of “Effective” WP Time: 2016 ACG Presidential Poster Award

Abstract

Introduction: A high quality colonoscopy improves adenoma detection rate (ADR) and low ADR is linked to development of interval colon cancer. Both patient and endoscopist-related factors are associated with ADR. A Longer withdrawal phase (WP) time is associated with higher ADR. Six minutes or longer WP time has been recommended. We wanted to investigate how much of WP time is actually spent on inspection of the mucosa as opposed to uninterpretable visual data. We wanted to test a novel image-based approach in assessing meaningful adequate information acquired during WP. We call this adequate visual data as the “effective” WP time. Aim1) Characterize each representative image from withdrawal phase of colonoscopy videos as adequate or inadequate. 2) Calculate proportion of visual data acquired during WP that is effective. Methods: Ten randomly selected WP from screening colonoscopy videos that had adequate cleansing preparation score were selected. Every 200th image was selected to represent the entire visual data acquired during WP. Each image was classified as 1) Adequate(Figure 1): >75% of image was interpretable 2) Inadequate(Figure 2): < 50% interpretable 3) Ambiguous: >50% but < 75%. 2 readers in a blinded manner analyzed each image independently and each ambiguous image was analyzed twice. Both readers had >10 year experience and ADR>40%. Kappa statistics were used for agreement scores.Figure 1Figure 2Results: 1005 individual images from WP of 10 colonoscopies was analyzed. On average for each WP, 32% of acquired visual data was inadequate, 57% adequate and 11% ambiguous for reader one. For the second reader, 27% was inadequate, 61% adequate 12% ambiguous (Table 1). Inter-observer agreement scores between two readers for inadequate images was good (Kappa Score 0.64, p< 0.001). Intra-observer agreement score was excellent (Kappa 0.8 and 0.81) for each reader (p < 0.001).Table 1: Percentage of individual images for each WP video, that had inadequate visual data, for interpretation of mucosa.Conclusion: Not all visual data acquired during WP of colonoscopy is adequate for visual interpretation of mucosa. In our study, up to 32% of the WP images were inadequate, introducing the concept of “effective” withdrawal time. This can be a potential target for improvement for endoscopist with low ADRs. Several approaches, including training programs for better hand-eye coordination can be employed to reduce the inadequate data acquired during WP and maximize the effective time during the WP. Our study generates feasibility of an image analyzing software that can be employed to track this particular indicator.