In vivo Detection of Colonic Mucosa Microinflammation by Confocal Laser Endomicroscopy Probe (p-CLE) in Patients With Irritable Bowel Syndrome: 2017 Category Award (Functional Bowel Disease)

Abstract

Introduction:The Miami criteria were developed and validated to standardize definitions of dysplasia and cancer during probe-based confocal laser endomicroscopy (pCLE) of Barrett's esophagus. These criteria do not distinguish high-grade dysplasia from low-grade dysplasia (LGD). The aim of our study was to evaluate the accuracy of Miami criteria for LGD in Barrett's esophagus. Methods: We reviewed medical records of all patients undergoing pCLE for Barrett's esophagus from December 2014 to December 2016. An expert pathologist performed blinded review of histology specimens. Patients with LGD on blinded or un-blinded pathology review were included. Patients with nondysplastic Barrett's (NDB) on blinded and un-blinded pathology review were randomly included in a 1:1 ratio as a control group. Two gastroenterologists with expertise in pCLE blindly interpreted all pCLE sequences based solely on Miami criteria and independently assigned a diagnosis as well as degree of confidence. Results: 20 patients with LGD and 20 patients with NDB were included. Patients with LGD had a significantly longer median Barrett's segment than those with NDB (5 cm vs. 2 cm, P=0.003). Inter-observer agreement was poor between blinded and un-blinded pathologists (kappa=0.2), but substantial between the two blinded pCLE reviewers (kappa=0.78). Among the 4 patients with LGD on both blinded and un-blinded pathology review, Miami criteria for LGD were felt to be satisfied in 3 patients by one pCLE reviewer, and 2 patients by the second pCLE reviewer. In the 20 patients with LGD identified by at least one pathologist, accuracy of pCLE Miami criteria were as follows: sensitivity 35-50%, specificity 70-75%, positive predictive value 58-63%, and negative predictive value 54 to 58% (Tables 1 and 2). Among the 20 patients with NDB on blinded and un-blinded pathology review, both pCLE reviewers identified dysplasia in 5 patients. Conclusion:In patients with LGD, Miami pCLE criteria demonstrate substantial inter-observer agreement but correlate poorly with biopsy. As documented in previous studies, there is significant disagreement between pathologists in determining a diagnosis of LGD, which limits the utility of histology as a gold standard for comparison.