ABSTRACT
Although Mammography Quality Standards Act requires tracking true positives, tracking false negatives is not required. We describe a peer review process implemented at Lahey Clinic to identify false-negative mammograms. We defined a false-negative mammogram as one which was read as negative within 12 months of a cancer diagnosis, and in which two of three radiologists correctly identified the site of cancer. Reviewing radiologists were blinded to each other and to computer-aided design (CAD), but were aware that somewhere in the mammogram was cancer. 25/64, 983, or 0.038% were classified as misses. The false-negative rate of any one radiologist averaged <0.1% without outliers. Of the false negatives, 60% were in heterogeneously dense tissue, 72% were asymmetries or masses rather than calcifications, and 24% were correctly identified by CAD in two views. We use these data for quality assurance, educational purposes, and to help identify patterns of undetected cancers to aid in earlier and improved detection of breast cancers.
