Send to

Choose Destination
See comment in PubMed Commons below
J Am Coll Radiol. 2016 Nov;13(11S):e72-e79. doi: 10.1016/j.jacr.2016.09.035.

Radiologist Agreement for Mammographic Recall by Case Difficulty and Finding Type.

Author information

  • 1Department of Community and Family Medicine, Dartmouth Medical School, Hanover, New Hampshire. Electronic address:
  • 2Group Health Research Institute, Seattle, Washington.
  • 3Group Health Research Institute, Seattle, Washington; Department of Biostatistics, University of Washington, Seattle, Washington.
  • 4Departments of Family Medicine and Public Health and Preventive Medicine, Oregon Health & Science University, Portland, Oregon.
  • 5University of Vermont, Burlington, Vermont.
  • 6Departments of Medicine and Epidemiology/Biostatistics and Radiology, University of California, San Francisco, San Francisco, California.
  • 7Department of Radiology, University of New Mexico, Albuquerque, New Mexico.
  • 8American Cancer Society, Atlanta, Georgia.
  • 9Department of Biostatistics, Harvard School of Public Health, Boston, Massachusetts.
  • 10Department of Radiology, University of North Carolina at Chapel Hill, Chapel Hill, North Carolina.



The aim of this study was to assess agreement of mammographic interpretations by community radiologists with consensus interpretations of an expert radiology panel to inform approaches that improve mammographic performance.


From 6 mammographic registries, 119 community-based radiologists were recruited to assess 1 of 4 randomly assigned test sets of 109 screening mammograms with comparison studies for no recall or recall, giving the most significant finding type (mass, calcifications, asymmetric density, or architectural distortion) and location. The mean proportion of agreement with an expert radiology panel was calculated by cancer status, finding type, and difficulty level of identifying the finding at the patient, breast, and lesion level. Concordance in finding type between study radiologists and the expert panel was also examined. For each finding type, the proportion of unnecessary recalls, defined as study radiologist recalls that were not expert panel recalls, was determined.


Recall agreement was 100% for masses and for examinations with obvious findings in both cancer and noncancer cases. Among cancer cases, recall agreement was lower for lesions that were subtle (50%) or asymmetric (60%). Subtle noncancer findings and benign calcifications showed 33% agreement for recall. Agreement for finding responsible for recall was low, especially for architectural distortions (43%) and asymmetric densities (40%). Most unnecessary recalls (51%) were asymmetric densities.


Agreement in mammographic interpretation was low for asymmetric densities and architectural distortions. Training focused on these interpretations could improve the accuracy of mammography and reduce unnecessary recalls.


Mammography; agreement; breast cancer; screening

[PubMed - in process]
PubMed Commons home

PubMed Commons

How to join PubMed Commons

    Supplemental Content

    Full text links

    Icon for Elsevier Science
    Loading ...
    Support Center