Interobserver variability in the classification of proliferative breast lesions by fine-needle aspiration: results of the Papanicolaou Society of Cytopathology Study.

This study evaluates the applicability of the published cytologic criteria in the categorization of proliferative breast lesions by assessing the diagnostic accuracy and interobserver reproducibility of a panel of experts. Twelve breast fine-needle aspiration (FNA) specimens of biopsy-proven nonproliferative breast lesion (NPL) (1 case), proliferative lesions without atypia (PL) (7 cases), proliferative lesion with atypia (PLA) (1 case), and low-nuclear grade ductal carcinoma in situ (DCIS) (3 cases) were selected. Six FNAs were Papanicolaou (PAP) and 6 were Diff-Quik-stained (DQ). Six expert cytopathologists classified the smears using a summary of published criteria as a guideline. All 6 participants rendered the same cytologic diagnosis in 2/12 (16%) cases. The agreement among the 6 raters was low (Kappa = 0.35). Cytohistologic correlation was achieved in 26/72 (36%) FNA diagnoses. The correlation of the PAP-stained cases was better than the DQ: 17/36 (47%) PAP and 9/36 (25%) DQ correlated. Improving the correlation was achieved by amalgamation of NPL and PL into "low risk" and PLA and DCIS into "high risk" categories: 47/72 (65%) FNA diagnoses then correlated with histology [29/36 (81%) PAP and 18/36 (50%) DQ]. We conclude that the cytologic criteria of proliferative breast lesions need to be further defined and assessed. Consideration should be given to minimizing the number of diagnostic categories and adopting a terminology that has a direct effect on patient management.