Reproducibility of diagnostic criteria associated with atypical breast cytology

The aim of this project was to assess the interobserver and intraobserver variability when identifying cytological criteria, which were found to have a statistically significant association with C3 within a workplace environment.

Sixty C3 cases with known endpoints of malignant, benign proliferative or benign non-proliferative diagnoses were blindly and independently screened by seven experienced cytologists to identify previously reported statistically significant criteria associated with the C3 category. The criteria included the presence of myoepithelial cells or bare bipolar nuclei, cohesiveness, cystic background, papillary fragments with fibrovascular cores and tubular structures. Kappa statistics were used to measure interobserver variability. Two cytologists repeated the process 6 months later to obtain intraobserver data.

The interobserver agreement was poor for all criteria except tubules which performed badly. The intraobserver variability for the two cytologists showed that one cytologist achieved moderate intraobserver agreement for all the criteria except cohesion which was poor, whilst the second cytologist showed poor agreement for all criteria. The reasons for the variability are multifactorial and include threshold effects where criteria lack good definition or error in identifying the criteria.

Interobserver and intraobserver variability remains a significant challenge for cytologists. Despite attempts to define significant criteria associated with C3, good reproducibility could not be achieved. The C3 category is imprecise and highlights the inadequacy of the current classification reporting categories for breast FNA. The impending review of reporting breast cytology by the International Academy of Cytology is timely and appropriate.