Purpose: to analyze interobserver variability in the histopathological diagnosis of premalignant breast lesions before and after training with diagnostic standardized criteria.
Methods: Slides containing histological sections representative of three kinds of breast lesions (atypical ductal hyperplasia, ductal carcinoma in situ and ductal carcinoma in situ with microinvasion), revised by an international specialist in breast pathology whose diagnoses were considered as golden standard, have been used. The same slides have been evaluated at two different times by five pathologists from the community according to a specific protocol for classifying the lesions. In the first evaluation, the cases were analyzed and classified according to the specific criteria adopted in each service. At the second time, the pathologists were given a tutorial containing diagnostic criteria and representative images, and the lesions were classified again, employing the standardized criteria. Interobserver analysis using percent agreement and weighted Kappa index has been performed.
Results: There has been a large diagnostic variation among the pathologists in the initial analysis without the use of standardized diagnostic criteria concerning the diagnostic, nuclear grade and histological grade (weighted Kappa indexes related to diagnosis varied from 0.15 to 0.40). In the second evaluation using standardized criteria, there has been a significant improvement in the diagnostic concordance among the five pathologists concerning the diagnosis, nuclear grade and histological grade (weighted Kappa indexes related to diagnosis have varied from 0.42 to 0.80).
Conclusions: interobserver concordance related to diagnosis and classification of breast premalignant lesions may be improved with specific training and the use of standardized histopathological criteria.