Interobserver Reproducibility in Diagnosis of Breast Lesions. A Panel in Puglia (Italy)

Abstract

Background Given the incidence of breast cancer, histologic agreement is necessary to select the proper treatment. Methods & study design Twelve pathologists working in Puglia (Italy) independently evaluated a set of 88 slides of breast lesions. The set included 31 cases which presented difficulties at the moment of first diagnosis (problem cases) and 57 cases without problems (routine cases). Each pathologist was requested to classify lesions according to four categories: A, benign; B, atypical proliferation; C, in situ carcinoma; D, invasive carcinoma. For each case, the definite diagnosis was considered that provided by most of the pathologists (prevalent diagnosis). For the evaluation of degree of agreement, kappa statistics were utilized. Results Among routine cases, agreement was observed in 68.4% of cases (38/57) and in 29% of problem cases (9/31). The pair-wise comparison between all participating pathologists showed a good overall agreement, (kappa mean, 0.66; range, 0.57-0.76). The agreement of each pathologist with the prevalent diagnosis was high for invasive (mean kappa, 0.88) and benign lesions (mean kappa, 0.83) followed by in situ carcinoma (mean kappa, 0.64). The lowest value observed was for atypical lesions (mean kappa, 0.25). Conclusions The results showed a good overall degree of diagnostic concordance among participating pathologists, all working in the Italian region of Puglia. The findings emerging from the study are comparable to those of other studies performed with selected pathologists specifically specialized in breast pathology. Moreover, the study confirmed the diagnostic difficulties for borderline lesions and the necessity of further investigation for sclerosing lesions with discordant diagnoses.