Evaluation of the updated confocal laser endomicroscopy criteria for Barrett's esophagus among gastrointestinal pathologists

Abstract

Previously developed novel probe-based confocal laser endomicroscopy (pCLE) criteria have been found to have high accuracy and substantial interobserver agreement (IOA) for diagnosing dysplasia in Barrett's esophagus (BE) when used by endoscopists. These updated criteria are: (i) epithelial surface: saw toothed, (ii) cells: enlarged, (iii) cells: pleomorphic, (iv) glands: not equidistant, (v) glands: unequal in size and shape, and (vi) goblet cells: not easily identified. The accuracy and IOA among pathologists in the diagnosis of dysplasia using the novel pCLE criteria is not known. The primary objective of the study was to evaluate the accuracy, overall IOA and learning curve among three gastrointestinal (GI) pathologists in diagnosing dysplasia in BE using the updated pCLE criteria. The secondary aim was to compare the accuracy and IOA between GI pathologists and gastroenterology endoscopists. Ninety pCLE videos and respective histology were retrieved from a previously conducted multicenter, prospective, randomized, controlled trial evaluating the utility of pCLE in BE patients. Videos were obtained from 101 BE patients previously enrolled for surveillance or endoscopic treatment of high-grade dysplasia or early esophageal adenocarcinoma. Three GI pathologists reviewed 90 pCLE video clips for dysplasia versus no dysplasia, confidence in their diagnosis, and image quality. The overall accuracy for the diagnosis of dysplasia (low-grade dysplasia/high-grade dysplasia/esophageal adenocarcinoma) was 77.8% (95% confidence interval [CI]: 72.4-82.3). The accuracy was higher when pathologists had 'high confidence' in their assessment of the videos (93.8% vs. 69.3%, P < 0.001). There was no significant difference in accuracy between the first set of 30 and second set of 60 videos (84% vs. 74%, P = 0.065). IOA among GI pathologists was substantial, k = 0.65 (95% CI: 0.53-0.73). The sensitivity for detecting dysplasia was 85% (95% CI: 78.1-90.7) and the specificity was 70% (95% CI: 61.91-77.92). These results were comparable with the evaluation of the same set of videos by endoscopists. GI pathologists have high accuracy and substantial IOA for diagnosing BE dysplasia with pCLE. Pathologists appear to have similar accuracy and IOA as endoscopists. These results provide further support of endoscopists accurately interpreting the in vivo optical histology provided by pCLE.

Keywords: Barrett's esophagus; accuracy; confocal laser endomicroscopy; dysplasia; interobserver agreement; sensitivity and specificity.