ABSTRACT
BACKGROUND: The Breast Imaging Reporting and Data System (BI-RADS) ultrasound (US) categorization revised in 2013 by the American College of Radiology resulted in unquestionable standardization of reports and confirmed category 3 and 5 as benign and malignant lesions, respectively. In contrast, suspected images (category 4) have subcategorization criteria, although theses have been detailed difficult to apply. The aim of the present study was to determine the role of the US 4A to 4C BI-RADS subcategories in predicting malignancy. PATIENTS AND METHODS: We performed a cross-sectional study of diagnostic tests to estimate the performance of the US BI-RADS categorization to clearly differentiate benign from malignant lesions. A total of 975 US examinations performed at the Hospital Femina, Grupo Hospitalar Conceição teaching hospitals from January 2012 through March 2015 were included in the present study. The US BI-RADS lexicon was used to classify the examination findings. Suspicious lesions underwent core needle biopsy, and the US and histology reports were compared to determine the performance using receiver operating characteristic curves. RESULTS: Overall, the BI-RADS US categorization showed good discriminating accuracy with a receiver operating characteristic curve of 91% (95% confidence interval [CI], 88%-93%). However, BI-RADS subcategory 4b had a positive predictive value of 25% (95% CI, 20%-31%) and subcategory 4A had a positive predictive value of only 6% (95% CI, 3.5%-9.8%). CONCLUSION: Our results have shown that US BI-RADS subcategories 4A and 4B are clearly unfit for use in screening tests, because they cannot rule out the need for biopsy. Therefore, management will not be improved by subcategorizing category 4, because all suspicious lesions will still require definite biopsy.
