Public Release: 

Accuracy of diagnostic mammograms varies by radiologist

Group Health team finds wide variability in nationwide study

Group Health Research Institute

SEATTLE--For women with breast symptoms such as lumps, the ability of diagnostic mammograms to detect breast cancer accurately depends strongly on which radiologist reads them, according to a Group Health study published online on December 11 in the Journal of the National Cancer Institute.

"When a woman gets a mammogram, she wants to know that if she has breast cancer, the mammogram will be likely to detect it," said study leader Diana Miglioretti, PhD, an associate investigator at Group Health Center for Health Studies. "This is especially important when the woman has a breast concern such as a lump."

Ideally, this ability to accurately detect cancer (known as "sensitivity") would be consistently high, with few false-positives--biopsies performed despite the absence of cancer. And it wouldn't depend on which radiologist was reading the mammograms. "But that's not what we found," she added.

The research team examined how well 123 radiologists interpreted nearly 36,000 diagnostic mammograms done to evaluate breast problems, such as lumps, from 1996 through 2003 at 72 U.S. facilities, including six from Group Health, that contribute data to the Breast Cancer Surveillance Consortium.

For different radiologists, sensitivity ranged from 27 percent to 100 percent; and false-positives, from 0 to 16 percent. These differences were only partially explained by the characteristics of the patients and the experience of the radiologists.

The radiologists who read diagnostic mammograms most accurately (with highest sensitivity, without too many false-positives) tended to be those who were based at academic medical centers or spent at least 20 percent of their time on breast imaging. By contrast, unlike in Europe, most U.S. women get mammograms interpreted by general radiologists who interpret mammograms as only a small percentage of their practice.

"We need to reduce the wide variability among radiologists in how they interpret diagnostic--and screening--mammograms," said Miglioretti.

"A good way to do that may be to identify the radiologists who are least accurate at reading mammograms--and to improve their performance with extra training." The national Breast Cancer Surveillance Consortium is working on ways to accomplish these goals, including developing an interactive training program.

"Women should get regular screening mammograms," said Miglioretti. "Mammography isn't perfect, but it's the best way we have to detect breast cancer early, when it has the greatest chance of being cured." She also urged women with breast concerns, such as lumps, to try to get evaluated at a center that has at least one breast imaging specialist: a radiologist who spends a large percentage of the time reading mammograms and performing breast biopsies.

Most mammograms are done to screen women with no symptoms for breast cancer. Previous research has shown that radiologists vary widely in how they read such screening mammograms. This new study is the largest to examine what predicts variability in diagnostic mammograms. Miglioretti's coauthors included colleagues at Group Health; Harborview Medical Center; University of Washington; University of California, San Francisco; and Oregon Health & Science University.


Grants from the Agency for Healthcare Research and Quality, the National Cancer Institute, and the Breast Cancer Surveillance Consortium funded the study.

About Group Health Center for Health Studies

Founded in 1947, Group Health is a Seattle-based, consumer-governed, nonprofit health care system that coordinates care and coverage. Group Health Center for Health Studies conducts research related to prevention, diagnosis, and treatment of major health problems. It is funded primarily through government and private research grants.

Please visit the virtual newsroom on our Web site, under "Newsroom."

Disclaimer: AAAS and EurekAlert! are not responsible for the accuracy of news releases posted to EurekAlert! by contributing institutions or for the use of any information through the EurekAlert system.