Magnifying narrowband imaging is more accurate than conventional white-light imaging in diagnosis of gastric mucosal cancer.


BACKGROUND & AIMS It is difficult to accurately diagnose patients with depressed gastric mucosal cancer based on conventional white-light imaging (C-WLI) endoscopy. We compared the real-time diagnostic yield of C-WLI for small, depressed gastric mucosal cancers with that of magnifying narrow-band imaging (M-NBI). METHODS We performed a multicenter, prospective, randomized, controlled trial of patients with undiagnosed depressed lesions ≤10 mm in diameter identified by esophagogastroduodenoscopy. Patients were randomly assigned to groups that were analyzed by C-WLI (n = 176) or M-NBI (n = 177) immediately after detection; the C-WLI group received M-NBI after C-WLI. We compared the diagnostic accuracy, sensitivity, and specificity between C-WLI and M-NBI and assessed the diagnostic yield of M-NBI conducted in conjunction with C-WLI. RESULTS Overall, 40 gastric cancers (20 in each group) were identified. The median diagnostic values for M-NBI and C-WLI were as follows: accuracy, 90.4% and 64.8%; sensitivity, 60.0% and 40.0%; and specificity, 94.3% and 67.9%, respectively. The accuracy and specificity of M-NBI were greater than those of C-WLI (P < .001); the difference in sensitivity was not significant (P = .34). The combination of M-NBI with C-WLI significantly enhanced performance compared with C-WLI alone; accuracy increased from (median) 64.8% to 96.6% (P < .001), sensitivity increased from 40.0% to 95.0% (P < .001), and specificity increased from 67.9% to 96.8% (P < .001). CONCLUSIONS M-NBI, in conjunction with C-WLI, identifies small, depressed gastric mucosal cancers with 96.6% accuracy, 95.0% sensitivity, and 96.8% specificity. These values are better than for C-WLI or M-NBI alone.

DOI: 10.1053/j.gastro.2011.08.007

4 Figures and Tables

Showing 1-10 of 52 extracted citations
Citations per Year

1,184 Citations

Semantic Scholar estimates that this publication has received between 528 and 2,252 citations based on the available data.

See our FAQ for additional information.