Aims: To assess the impact of cytokeratin (CK) 5/6 and E-cadherin immunohistochemistry on diagnostic agreement of non-invasive proliferative breast lesions.
Methods and results: Twenty pathologists classified 105 cases of non-invasive proliferative breast lesions into 10 diagnostic categories. One haematoxylin and eosin (H&E) slide of each case was analysed on a first round and one H&E slide with corresponding CK5/6 and E-cadherin immunohistochemistry was analysed on a second round. Interobserver reproducibility for category-specific and management-specific lesions was measured on each round. CK5/6 and E-cadherin had little impact on diagnostic agreement, which remained moderate between the first and second rounds (overall kappa coefficients of 0.47 and 0.53, respectively, P = NS). Levels of agreement slightly improved for lesions with specific CK5/6 and E-cadherin immunoprofiles (usual ductal hyperplasia, atypical ductal hyperplasia, atypical lobular hyperplasia, lobular carcinoma in situ, non-high-grade ductal carcinoma in situ), but the differences observed were not statistically significant. However, diagnostic agreement improved when lesions were grouped according to their management category (overall kappa coefficients of 0.58 and 0.66 in the first and second rounds, respectively).
Conclusions: CK5/6 and E-cadherin immunohistochemistry has little impact on interobserver reproducibility for non-invasive breast lesions. Diagnostic agreement can, however, be improved by grouping lesions in management categories.