Background: The aim of this study was to define the reproducibility and accuracy of magnification chromoendoscopy for the diagnosis of lesions associated with gastric cancer (intestinal metaplasia and dysplasia).
Methods: A total of 136 patients with previously diagnosed lesions and 5 gastrectomy specimens were studied. Endoscopic examination was performed with a magnification endoscope after methylene blue (1%) spraying. According to differences in color and mucosal pattern, groups and subgroups of endoscopic images were defined, and biopsies taken (n = 462). Five endoscopists were asked to classify individually 2 endoscopic images per subgroup on 2 separate occasions.
Results: Three groups of endoscopic images were defined: nonmetaplastic, nondysplastic mucosa (I); metaplastic mucosa (II); and dysplastic mucosa (III). Ten subgroups were defined according to pit pattern: round small (IA), round and tubular small (IB), coarse round (IC), and course round pits with a straight pit (ID); blue irregular marks (IIA), blue round and tubular pits (IIB), blue villi (IIC), and blue small pits (IID); and loss of clear pattern, with depression (IIIA) or with slight elevation (IIIB). The kappa statistic for intraobserver agreement on the classification of endoscopic images in groups was 0.86; for interobserver agreement, it was 0.74. For classification into subgroups, kappa values ranged from 0.48 to 0.78. For 85% of the areas classified endoscopically as Group I (n = 146), no mucosal lesions or gastritis was described at histologic examination; for 83% of those in Group II (n = 198), intestinal metaplasia was found. Subgroups IIA and IIB were more often associated with complete intestinal metaplasia (62%), and IIC and IID with incomplete metaplasia (67%); in Group III (n = 118), dysplasia was diagnosed histopathologically in 33%. For the diagnosis of dysplasia, specificity was 81% (95% CI [77%, 85%]) and negative predictive value 99% (95% CI [99%, 100%]).
Conclusions: Gastric endoscopic patterns with chromoendoscopy and magnification seem reproducible and valid for the diagnosis of lesions associated with gastric cancer. This procedure may improve the follow-up of individuals at high-risk of gastric cancer, at least for the exclusion of severe lesions.