Fine-needle aspiration biopsy (FNAB) has been used with variable success as a diagnostic test for benign and malignant breast lesions. The goal of this study was to examine the effects of training physicians in the fine-needle aspiration sampling-technique on the diagnostic accuracy of FNAB of palpable breast masses. The settings for this study were private physicians' offices and university clinics of primary care physicians, surgeons, and cytopathologists.
We reviewed 1043 consecutive FNAB specimens of the breast obtained during 1 year (1992): 729 FNABs were performed by formally trained physicians (at least 150 FNABs performed previously under supervision during fellowship training or the equivalent) who had done at least 100 FNABs during the year; 314 FNABs were performed by physicians without formal training who had done a median of only 2 FNABs during the year (range, 1-43 FNABs). All FNAB specimens were reviewed microscopically and evaluated for cellularity and type of material present, for diagnostic accuracy, and for the rate of surgical intervention. A minimum of 2 years of follow-up was obtained by matching all cases to the population-based Northern California Cancer Registry. FNAB specimens were correlated with histologic specimens when they were available.
Using FNAB, the formally trained physicians missed 2% of cancers, whereas the physicians without formal training missed 25%. Among the patients with benign lesions seen by the formally trained physicians, 8% went on to surgery, whereas 30% of those seen by physicians without formal training did so. Specimens obtained by the formally trained physicians were significantly more cellular and were significantly less likely to be nondiagnostic.
FNAB, when performed by physicians who are well trained in the technique, is a highly accurate, cost-effective diagnostic method that carries minimal morbidity and could replace a large number of surgical biopsies. When performed by physicians without adequate training, FNAB is often misleading and potentially harmful.