Abstract B82: Performance of digital diagnostic mammography by race

Background: Diagnostic mammography is used to evaluate women who have signs or symptoms of breast cancer or who have an abnormal screening mammogram. Diagnostic mammography typically involves the acquisition of additional images and is usually done to assess a specific area of concern. The performance of diagnostic mammography (how well this tool detects cancers) is known to vary by the radiologist interpreting the image. However, other factors such as race and type of imaging modality may also impact the performance. A prior study found that diagnostic mammography performance differed by race for film examinations but to our knowledge, no studies have examined if performance differs by race for digital examinations. Hence, we sought to ascertain if the diagnostic mammography performance of digital differed among black and white women in the United States. Methods: Using the Carolina Mammography Registry (CMR) data from 2004-2011 we included all digital diagnostic mammograms performed among black and white women ages 18 and older. The CMR is a prospective population-based breast imaging registry in North Carolina that collects patient demographic, risk factor, and clinical information each time a woman receives mammography. This data is linked to pathology and cancer outcome data. We included diagnostic mammograms, defined as those being performed for symptomatic reasons or for the workup of an abnormal screening examination. To determine the performance of diagnostic mammography, each diagnostic mammogram was categorized as true positive, false negative, true negative, and false positive according to the BI-RADS code assigned by the radiologist and cancer outcomes ascertained in the one-year follow-up period. We computed the diagnostic mammography performance measures of sensitivity, specificity, positive predictive value of the biopsy (PPV2), and the cancer detection rate (CDR) using standard definitions. We compared these performance measures and the CDR among black and white women. We adjusted for age, breast density, history of previous breast procedures, family history of breast cancer, time since last mammogram, and tumor characteristics, including size, grade, stage, type, estrogen receptor, progesterone receptor, and HER2 which could impact the ability to detect cancers. We used a multilevel model to adjust for differences between radiologists interpreting the images and to account for correlated observations within women. Results: We included 42,