Discordance Among Pathologists in the United States and Europe in Diagnosis of Low-Grade Dysplasia for Patients With Barrett's Esophagus

Abstract

Background & Aims There is suboptimal inter-observer agreement, even among expert gastrointestinal pathologists, in the diagnosis of low-grade dysplasia (LGD) in patients with Barrett's esophagus (BE). We analyzed histopathologic criteria required for a diagnosis of LGD using the new subcategories of LGD with inflammatory and dysplastic features. We categorized each diagnosis based on the level of confidence and assessed inter-observer agreement among gastrointestinal pathologists from 5 tertiary centers in the United States and Europe. Methods In the first phase of the study, 3 pathologists held a consensus conference at which they discussed the diagnostic criteria for LGD. In the second phase, 79 slides from patients with BE (23 samples of non-dysplastic BE, 22 samples of LGD, and 34 samples of high-grade dysplasia) were identified, randomly assigned to 7 pathologists (4 from the United States and 3 from Europe), and interpreted in a blinded fashion. κ Values were calculated for inter-observer agreement. We performed multinomial logistic regression analysis to assess the weighting of histologic features with the diagnosis. Results The overall κ value for diagnosis was 0.43 (95% confidence interval [CI], 0.42−0.48). When categorized based on degree of dysplasia, the κ value was 0.22 (95% CI, 0.11−0.29) for non-dysplastic BE, 0.11 (95% CI, 0.004−0.15) for LGD, and 0.43 (95% CI, 0.36−0.46) for high-grade dysplasia. When all pathologists made a diagnosis with high confidence, the inter-observer agreement was substantial among the US pathologists (κ, 0.63; 95% CI, 0.61−0.66) and European pathologists (κ, 0.80; 95% CI, 0.74−0.97). The κ values for all diagnoses made by European pathologists were higher than those made by US pathologists. Conclusions In an analysis of criteria used in histopathologic diagnosis of LGD, we did not observe improvement in level of agreement among experienced pathologists, even after accounting for inflammation. The level of inter-observer agreement increased with level of pathologist confidence. There was also a difference in reading of histopathology samples of BE tissues between US and European pathologists.