Causes of inconsistency in diagnosing and classifying intraductal proliferations of the breast. European Commission Working Group on Breast Screening Pathology

Abstract

It is now widely recognised that classifying ductal carcinoma in situ (DCIS) of the breast and diagnosing atypical ductal hyperplasia are associated with significant interobserver variation. Two possible reasons for this inconsistency are differences in the interpretation of specified histological features and field selection where morphology is heterogeneous. In order to investigate the relative contribution of these two factors to inconsistent interpretation of intraductal proliferations, histological sections of 32 lesions were sent to 23 European pathologists followed 3 years later by images of small parts of these sections. Kappa statistics for diagnosing hyperplasia of usual type, atypical ductal hyperplasia and ductal carcinoma in situ were 0.54, 0.35 and 0.78 for sections and 0.47, 0.29 and 0.78 for images, respectively, showing that most of the inconsistency is due to differences in morphological interpretation. Improvements can thus be expected only if diagnostic criteria or methodology are changed. In contrast, kappa for classifying DCIS by growth pattern was very low at 0.23 for sections and better at 0.47 for images, reflecting the widely recognised variation in the growth pattern of DCIS. Higher kappa statistics were obtained when any mention of an individual growth pattern was included in that category, thus allowing multiple categories per case; but kappa was still higher for images than sections. Classifying DCIS by nuclear grade gave kappa values of 0.36 for sections and 0.49 for images, indicating that intralesional heterogeneity has hitherto been underestimated as a cause of inconsistency in classifying DCIS by this method. More rigorous assessment of the proportions of the different nuclear grades present could lead to an improvement in consistency.