Jorge Lage1, Pedro Pimentel-Nunes1,2, Pedro C Figueiredo3, Diogo Libanio1, Iolanda Ribeiro4, Manuel Jacome5, Luis Afonso5, Mario Dinis-Ribeiro1,2. 1. a Gastroenterology Department , Portuguese Oncology Institute of Porto , Porto , Portugal ; 2. b CIDES/CINTESIS, Faculty of Medicine, University of Porto , Porto , Portugal ; 3. c Gastroenterology Department , Hospital Garcia De Orta , Almada , Portugal ; 4. d Gastroenterology Department , Centro Hospitalar De Vila Nova De Gaia/Espinho , Vila Nova De Gaia , Portugal ; 5. e Pathology Department , Portuguese Oncology Institute of Porto , Porto , Portugal.
Abstract
OBJECTIVE: Early diagnosis of gastric cancer may be achieved through surveillance of patients with extensive gastric intestinal metaplasia (eGIM). However, diagnosis of eGIM generally implies histology. We aimed at determining the accuracy of high-resolution endoscopy with light-narrow band imaging (NBI) to assess the presence of eGIM on a per-patient basis. MATERIAL AND METHODS: Prospective cohort of 60 patients divided into two groups: derivation cohort (n = 25) to evaluate the reliability and validity, and a real-time validation group (n = 35). In the derivation group, six endoscopists with two levels of expertise were asked to estimate the grade of GIM based in endoscopic images (white light endoscopy, light-NBI and amplification/near focus). In the real-time validation set, experienced endoscopists were asked to similarly record their real-time optical diagnosis. Histology was then considered as the gold standard. RESULTS: In the derivation group diagnosis accuracy was 60% with WLE (non-expert 59% vs. 61% experts), increasing to 73% after NBI magnification (non-expert 63% vs. 83% expert, p < 0.05). Moreover, proportion of agreement with histology was 83%, with a correct diagnosis of eGIM in 87% for experienced observers. In the real-time group experts obtained 89% global diagnostic accuracy correctly identifying 91% of the eGIM. The sensitivity, specificity, LR + and LR- of real-time endoscopic diagnosis of eGIM was 0.92 (CI95%:0.67-0.99), 0.96 (0.79-0.99), 21.1 (3.08-144) and 0.09 (0.013-0.57). CONCLUSION: For the first time the reliability of high-resolution endoscopy with light-NBI for extension of GIM is described. Our results suggest that more than 90% of individuals at risk could be identified without the need for biopsies, simplifying the current recommendations.
OBJECTIVE: Early diagnosis of gastric cancer may be achieved through surveillance of patients with extensive gastric intestinal metaplasia (eGIM). However, diagnosis of eGIM generally implies histology. We aimed at determining the accuracy of high-resolution endoscopy with light-narrow band imaging (NBI) to assess the presence of eGIM on a per-patient basis. MATERIAL AND METHODS: Prospective cohort of 60 patients divided into two groups: derivation cohort (n = 25) to evaluate the reliability and validity, and a real-time validation group (n = 35). In the derivation group, six endoscopists with two levels of expertise were asked to estimate the grade of GIM based in endoscopic images (white light endoscopy, light-NBI and amplification/near focus). In the real-time validation set, experienced endoscopists were asked to similarly record their real-time optical diagnosis. Histology was then considered as the gold standard. RESULTS: In the derivation group diagnosis accuracy was 60% with WLE (non-expert 59% vs. 61% experts), increasing to 73% after NBI magnification (non-expert 63% vs. 83% expert, p < 0.05). Moreover, proportion of agreement with histology was 83%, with a correct diagnosis of eGIM in 87% for experienced observers. In the real-time group experts obtained 89% global diagnostic accuracy correctly identifying 91% of the eGIM. The sensitivity, specificity, LR + and LR- of real-time endoscopic diagnosis of eGIM was 0.92 (CI95%:0.67-0.99), 0.96 (0.79-0.99), 21.1 (3.08-144) and 0.09 (0.013-0.57). CONCLUSION: For the first time the reliability of high-resolution endoscopy with light-NBI for extension of GIM is described. Our results suggest that more than 90% of individuals at risk could be identified without the need for biopsies, simplifying the current recommendations.