Report from AIUM: Ultrasound BI-RADS needs fine-tuning

A significant number of breast cancers missed by mammography in the last decade may have been misclassified as benign lesions by ultrasound in the first place, according to a study presented at the 2005 American Institute of Ultrasound in Medicine meeting.

Recent studies have validated the American College of Radiology's Breast Imaging Reporting and Data System for ultrasound as a reliable tool for characterization and classification of tumors. The BI-RADS classification began in 2003, however, and it might not yet reflect accurately all the subtleties and features of ultrasound-scanned lesions.

According to current standards, a 2% false-negative yield from lesions originally classified as benign BI-RADS category 3 seems reasonable. But this margin of error should not be acceptable to either patients or their physicians, said principal investigator Dr. Dona J. Hills, director of ultrasound at the Queens-Long Island Medical Group.

Hills and colleagues retrospectively reviewed a database with 246 breast cancer cases diagnosed by ultrasound-guided biopsy dating back to 1991. Forty-one cases had been classified as benign or probably benign on ultrasound in relation to the terminology available at the time. These cases were reassigned the BI-RAD categories 2 or 3 according to the new nomenclature.

The investigators found that 23 of the 41 patients also showed benign or negative findings on mammography. They discovered that the ultrasound characteristics were unlike those usually associated with breast cancer.

The false negatives that ultrasound wrongly classified as benign or probably benign had no shadowing and little or no fat compression, were mostly round or oval in shape, and measured 1 cm or less.

It is not reassuring that mammograms consistently failed to detect breast cancer in patients with probably benign ultrasound examinations. But since the lesion characteristics shown by mammography and ultrasound were similar, physicians need to become more vigilant and suspicious to avoid missing cancers with atypical features, Hills said.

"If you exclude suspicious findings from the 'probably benign' category and put them into a BI-RADS 4, you will probably not be underestimating as many cancers as we may have underestimated before. You will become much more precise," she said.