Ryan Woods MD, MPH, Presenter: Nothing to Disclose

Mary Lindstrom PhD, Abstract Co-Author: Nothing to Disclose

Elizabeth S. Burnside MD, MPH, Abstract Co-Author: Stockholder, Cellectar, LLC

Previous research has retrospectively assessed the positive predictive value (PPV) of Breast Imaging Reporting and Data System (BI-RADS) descriptors among masses recommended for biopsy. In contrast, we evaluate the PPV of mass descriptors—shape, margin, stability, and density—prospectively in a non-selected, consecutive series of mammograms.

We analyzed consecutive diagnostic mammograms performed between October 2005 and September 2007, and extracted 2230 prospectively-described non-calcified masses among 1587 patients on 1889 mammograms. Each mass was given a final assessment using BI-RADS descriptors, as required by our structured reporting software, by one of nine radiologists, three of which are fellowship trained. We defined true positives as BI-RADS 0, 4, or 5 masses that were malignant at image-guided or surgical biopsy or appeared in a hospital-based cancer registry within 1 year of the mammogram. Masses were matched to the registry based on patient identifier and mass laterality. The remaining masses were followed for at least 1 year (median = 23 months) to establish benignity. We calculated the PPV (the number of malignant masses divided by the total number of masses) for every level of each mass descriptor. We compared the PPVs within each descriptor using the Chi-Square test.

The overall rate of malignancy was 6.1% (137/2230 masses). The PPVs for mass shape were: round = 4.7%; oval = 3.9%; lobular = 8%; irregular = 26.5% (p-value <0.0001). The PPVs for mass margin were: circumscribed = 2.4%; obscured = 5.8%; indistinct = 16.0%; microlobulated = 28.1%; spiculated = 62% (p-value <0.0001). The PPVs for stability were: size decreasing = 0%; no change = 0.3%; size increasing = 8% (p-value = 0.0015). The PPVs for mass density were: fat-containing = 0%; low density = 0%; iso-dense = 4.2%; high density 19.6% (p-value <0.0001).

Unlike previous retrospective analyses of masses recommended for biopsy, we show that prospectively-collected BI-RADS mass descriptors can stratify the risk of malignancy in a consecutive series of masses. These results may guide future prospective research and identify unique combinations of descriptors important to the decision to biopsy.

Using several BI-RADS mass descriptors to stratify malignancy risk may help guide appropriate management and can potentially improve PPV without missing cancers.

Woods, R, Lindstrom, M, Burnside, E, The Positive Predictive Value of Mass Descriptors among Consecutive Diagnostic Mammograms.  Radiological Society of North America 2009 Scientific Assembly and Annual Meeting, November 29 - December 4, 2009 ,Chicago IL. http://archive.rsna.org/2009/8011175.html