Abstract
A primary concern regarding current breast cancer detection and diagnosis is the large number of benign
biopsies being performed. Annually in the United States, radiologists recall over five million women for
diagnostic workup and perform over one million breast biopsies, with fewer than one in four diagnosed as
cancer. The new BI-RADS 4 subset assessment categories of 4A/B/C enables new tailored approaches for
“risk” based decisions in different subsets of cases. In this classification scheme, lesions rated as 4A have an
expected probability of malignancy more than 2% and up to 10% and those rated as 4B have a probability of
malignancy more than 10% and up to 50%. A recent publication evaluating the positive predictive value (PPV3)
for 125,447 diagnostic exams collected in the American College of Radiology (ACR) National Mammography
Database (NMD) demonstrated that 88% of biopsies were performed on BI-RADS 4A and 4B lesions. Thus,
opportunity exists to identify novel ways to better classify these low probability lesions more accurately in order
to improve PPV3 and reduce biopsy of actually benign findings for hundreds of thousands of women annually.
We believe that an operationally simple, cost effective, contrast enhanced mammogram (CEM), performed
during the patient's diagnostic evaluation, would be the best approach to improve accuracy of radiologists'
decisions for need to biopsy lesions classified with mammography, tomosynthesis (DBT) or US as 4A or 4B.
CEM uses iodine contrast with dual low and high KeV mammogram images to create a contrast enhancement
map of the breast that directly overlays the mammogram, thus providing anatomic and kinetic information,
similar to MRI. We found in a preliminary clinical trial that radiologists had higher true-positive rates and lower
false-positive rates for biopsy recommendation with CEM than when using DBT and US. To validate those
initial findings, we propose to prospectively and sequentially perform CEM on 1855 consenting women with BI-
RADS 4A or 4B lesions detected on mammography, DBT or US. Prospectively radiologists will provide BI-
RADS ratings for every lesion using DBT alone, then with US and finally with CEM. With pathology known and
based on the study design to minimize case by radiologist potential biases, we plan to estimate the NPV level
of CEM-based recommendations (overall and within the cases with conventionally confirmed biopsy
recommendation) and demonstrate that it is sufficiently high, while leading to substantial reduction in biopsy
recommendations for actually benign lesions. Our primary expectation is that the number of recommendations
to biopsy benign lesions will decrease significantly (~20%), while maintaining high NPV (>95%) among the
initial recommendations. Using a subset of prospectively collected verified cases we will conduct a multi-
reader (MRMC) study to assess heterogeneity of CEM effects across radiologists, obtain generalizable results,
and elucidate supplemental specific CEM performance by lesion characteristics.