OBJECTIVES: Endoscopic biopsies for nonampullary duodenal epithelial neoplasms (NADENs) can induce submucosal fibrosis, making endoscopic resection difficult. However, no biopsy-free method exists to distinguish between NADENs and non-neoplasms. We developed a diagnostic algorithm for duodenal neoplasms based on magnifying endoscopy findings and evaluated the model's diagnostic ability. METHODS: Magnified endoscopic images and duodenal lesion histology were collected consecutively between January 2015 and April 2016. Diagnosticians classified the surface patterns as pit, groove or absent. In cases of nonvisible surface patterns, the vascular pattern was evaluated to determine regularity or irregularity. The correlation between our algorithm (pit-type or absent with irregular vascular pattern) and the lesion histology were evaluated. Four evaluators, who were blinded to the histology, also classified the endoscopic findings and evaluated the diagnostic performance and interobserver agreement. RESULTS: Endoscopic images of 114 lesions were evaluated (70 NADENs and 44 non-neoplasms, 31 in the superior and 83 in the descending and horizontal duodenum). Of the NADEN surface patterns, 88% (62/70) were pit-type, while 79% (35/44) of the non-neoplasm surface patterns were groove-type. Our diagnostic algorithm for differentiating NADENs from non-neoplasms was high (sensitivity 96%, specificity 95%) in the descending and horizontal duodenum. The evaluators' diagnostic performances were also high, and interobserver agreement for the algorithm was good between each diagnostician and evaluator (κ = 0.60-0.76). CONCLUSION: Diagnostic performance of our algorithm sufficiently enabled eliminating endoscopic biopsies for diagnosing the descending and horizontal duodenum.
OBJECTIVES: Endoscopic biopsies for nonampullary duodenal epithelial neoplasms (NADENs) can induce submucosal fibrosis, making endoscopic resection difficult. However, no biopsy-free method exists to distinguish between NADENs and non-neoplasms. We developed a diagnostic algorithm for duodenal neoplasms based on magnifying endoscopy findings and evaluated the model's diagnostic ability. METHODS: Magnified endoscopic images and duodenal lesion histology were collected consecutively between January 2015 and April 2016. Diagnosticians classified the surface patterns as pit, groove or absent. In cases of nonvisible surface patterns, the vascular pattern was evaluated to determine regularity or irregularity. The correlation between our algorithm (pit-type or absent with irregular vascular pattern) and the lesion histology were evaluated. Four evaluators, who were blinded to the histology, also classified the endoscopic findings and evaluated the diagnostic performance and interobserver agreement. RESULTS: Endoscopic images of 114 lesions were evaluated (70 NADENs and 44 non-neoplasms, 31 in the superior and 83 in the descending and horizontal duodenum). Of the NADEN surface patterns, 88% (62/70) were pit-type, while 79% (35/44) of the non-neoplasm surface patterns were groove-type. Our diagnostic algorithm for differentiating NADENs from non-neoplasms was high (sensitivity 96%, specificity 95%) in the descending and horizontal duodenum. The evaluators' diagnostic performances were also high, and interobserver agreement for the algorithm was good between each diagnostician and evaluator (κ = 0.60-0.76). CONCLUSION: Diagnostic performance of our algorithm sufficiently enabled eliminating endoscopic biopsies for diagnosing the descending and horizontal duodenum.
Authors: Maxime Amoyel; Arthur Belle; Marion Dhooge; Einas Abou Ali; Rachel Hallit; Frederic Prat; Anthony Dohan; Benoit Terris; Stanislas Chaussade; Romain Coriat; Maximilien Barret Journal: Endosc Int Open Date: 2022-01-14