Su1997 - Generation and Validation of New Confocal Laser Endomicroscopy Criteria for the Diagnosis of Low-Grade Dysplasia in Barrett's Esophagus

Abstract

Introduction The diagnosis and interobserver agreement amongst pathologists for low-grade dysplasia in Barrett’s oesophagus (BO) is sub-optimal. Probe-based confocal laser endomicroscopy (pCLE) allows real-time histologic assessment of BO. pCLE criteria for high-grade dysplasia (HGD) are established, however criteria for low-grade dysplasia (LGD) are lacking. The aim of the study was to develop and validate novel diagnostic criteria for LGD in BO. Method In Phase I, one pathologist and one endoscopist expert in pCLE unblinded assessed 30 good quality pCLE videos (10 non-dysplastic BO, 10 LGD and 10 HGD) to identify criteria for LGD. These criteria were assessed blindly by these two investigators in an independent set of 25 videos (15 non-dysplastic BO and 10 LGD). Criteria with mean accuracy >80% and interobserver agreement κ >0.4 were taken forward. In Phase II, 6 endoscopists evaluated the criteria in an independent set of 37 videos (15 non-dysplastic BO and 22 LGD). The raters assessed each criterion separately and made an overall diagnosis. Sensitivity, specificity and interobserver agreement were calculated for each criterion and the overall diagnosis. A receiver operating characteristic (ROC) curve was constructed to evaluate the best cut-off to diagnose dysplasia. Results Of the initial set of 8 criteria, 6 achieved the agreement and accuracy thresholds in Phase I. These were: (1) dark non-round glands, (2) irregular gland shape, (3) lack of goblet cells, (4) variable degree of darkness with sharp cut-off, (5) variable cell size and (6) loss of nuclear polarity. In Phase II the interobserver agreement among the 6 endoscopists for the criteria ranged from fair (K=0.242; criterion 4) to substantial (K=0.637; criterion 2), with a substantial interobserver agreement for the overall diagnosis (κ=0.6). The best cut-off for LGD diagnosis was 3 positive criteria out of 6, which corresponded to a sensitivity and specificity of 81.6% and 67.6%, respectively and an area under the ROC curve of 0.860. The overall diagnosis had sensitivity and specificity of 77.2% and 72.2%, respectively an area under the ROC curve of 0.895. Conclusion We have developed and validated pCLE criteria for LGD in BO. The performance of these criteria compares favourably with the interobserver agreement among pathologists in a conventional histologic diagnosis. Disclosure of Interest None Declared