Evaluation of Wireless Capsule Endoscopy Interobserver Agreement among Four Different Levels of Expertise: 969

Purpose: To prospectively evaluate multiobserver diagnostic findings and render agreement for lesion and normal anatomical landmarks compared with gastroenterologist capsule expert (E) findings as the gold standard. Methods: 21 subjects underwent wireless capsule endoscopy (CE). CE video was initially read by E, who has read approximately 600 prior CE cases. 2 board-certified gastroenterologist capsule trainees [1 with no experience reading CE (T#1) and 1 with minimal experience reading CE (T#2)] independently read the same 21 CE videos. A nurse (N) with no prior endoscopy experience, but with experience processing 600 prior CE cases and who received certification from an ASGE-sponsored CE training program also read the same 21 CE videos. Thumbnail findings of E, T#1, T#2, and N were reviewed and were categorized as clinically significant, clinically insignificant, or normal anatomical landmarks. Thumbnail findings of E were considered to be the gold standard and were compared against the findings of T#1, T#2, and E. Kappa and other statistical analysis were performed. Results: 18 significant lesions were identified by E: AVM (9), bleeding (4), erosion (2), radiation colitis (1), scalloped mucosa (1), and tumor (1). The number correctly identified by T#1 was 10 (56%), by T#2 was 16 (89%), and by N was 14 (78%). 10 insignificant lesions were identified by E: xanthoma (4) and red spot (6). The number correctly identifed by T#1 was 5 (50%), T#2 was 8 (80%), and N was 9 (90%). 55 normal anatomical landmarks were identified: 1st gastric image (20), 1st duodenal image (19), and 1st cecal image (16). The number correctly identified by T#1 was 53 (96%), T#2 was 54 (98%), and N was 49 (89%). For significant lesions, 18 overcalls occurred: T#1 (10), T#2 (4), and N (4). There were 34 overcalls for insignificant lesions: T#1 (9), T#2 (10), and N (15). Kappa was 0.48 (moderate agreement) for significant lesions, 0.43 (moderate agreement) for insignificant lesions, and 0.88 (almost perfect agreement) for normal anatomical landmarks. Overall, kappa was 0.79 (substantial agreement). Conclusions: Prior endoscopy training alone by gastroenterologists does not assure adequate recognition of pathology on CE. A nurse with no experience reading CE performed remarkaby well when compared to three gastroenterologists. The role of a nurse in reading CE warrants further consideration.