Lobular neoplasia and invasive lobular breast cancer: Inter-observer agreement for histological grading and subclassification.

Lobular neoplasia (LN), invasive lobular breast cancer (ILBC) and related pleomorphic variants represent a distinct group of neoplastic mammary gland lesions. This study assessed the inter-observer agreement of histological grading in a series of ILBC and LN. 54 cases (36x ILBC, 18x LN) were evaluated by 17 observers. 3978 classification calls on various histological features, including nuclear grade, proliferative activity (Ki67 immunohistochemistry, categorical scoring), histological grade and pleomorphism were obtained. Pairwise Cohen's kappa values were calculated and compared between various features and different observer subsets with variable histomorphological experience. In ILBC, pairwise inter-observer agreement for histological grade ranged from poor to almost perfect concordance and was higher in advanced and experienced histopathologists compared with beginners (P < 0.001). Agreement for proliferation (Ki67) ranged from slight to almost perfect concordance and was also higher in advanced and experienced histopathologists (P < 0.001). Considering different features, agreement for proliferation (Ki67) was superior to agreement for histological grade and nuclear grade, even among advanced and experienced histopathologists (P < 0.001). In LN, agreement for B-classification ranged from poor to almost perfect concordance and was higher in advanced and experienced histopathologists (P < 0.001). Considering different features, agreement for proliferation (Ki67 in LN) was superior to subclassification agreement based on conventional features, such as acinar distention and nuclear grade (P < 0.001). In summary, pairwise inter-observer concordance of histological grading of ILBC and LN is dependent on histomorphological experience. Assessment of proliferation by Ki67 immunohistochemistry is associated with favorable inter-observer agreement and can improve histological grading of ILBC as well as LN.