Analysis of morphologic patterns of fine-needle aspiration of the breast to reduce false-negative results in breast cytology

BACKGROUND: The identification of specific morphologic diagnostic criteria is of paramount importance to optimize the accuracy of fine-needle aspiration cytology (FNAC) and to reduce the rate of false-negative results. In the current study, the authors reviewed a consecutive series of false-negative findings observed in the study center to define the presence and degree of cytologic abnormalities. False-negative cases were randomly mixed with true-negative cases and were reviewed by a panel of expert readers in a blinded fashion. The main objective of the current study was to identify a morphologic pattern that may permit the reduction of false-negative findings while maintaining the specificity of FNAC. METHODS: A blind review of a set of 41 consecutive false-negative and 49 true-negative breast aspiration samples was performed by a panel of 10 expert cytologists who were asked to give a final report and to classify the samples according to classic morphologic parameters. RESULTS: The majority final report sensitivity was 54% (range, 19-61%) and specificity was 73% (range, 65-92%). The average concordance with the majority report, adjusted for chance agreement (kappa statistic), was moderate at 0.54 (range, 0.40-0.65). Enlarged nuclear size, a hyperchromatic nucleus, the absence of naked nuclei, and the absence of apocrine metaplasia were reported more frequently in carcinoma cases, although not to a significant extent. The only variable found to be associated significantly (P = 0.041) with a diagnosis of carcinoma was the presence of microcalcifications, which nevertheless were found to occur in only a minority of carcinoma cases (7 of 41 cases) or controls (2 of 49 controls). Multivariate analysis demonstrated that the presence of microcalcifications (odds ration [OR] of 3.0; 95% confidence interval [95% CI], 1.2-7.4), the absence of naked nuclei (OR of 2.4; 95% CI, 1.3-4.4), and enlargement of the nucleus (OR of 1.9; 95% CI, 1.1-3.4) were all independently associated with false-negative findings. Diagnostic accuracy using a morphology-based score did not appear to improve the results substantially compared with the final report (sensitivity of 0.46 vs. 0.54 [P = 0.508] and a specificity of 0.80 vs. 0.73 [P = 0.218]). CONCLUSIONS: The results of the current study confirm that breast FNAC false-negative results are at least partially the result of underreporting of abnormalities that may be noted at review. Detailed analysis of a single morphologic characteristic was found to be of limited diagnostic value, suggesting that operators do perceive abnormalities but cannot translate these findings into distinct morphologic categories.