Radiologists' experience and affiliation often influenced results, study found
TUESDAY, Dec. 11 (HealthDay News) -- Radiologists are human, too, and vary widely in their interpretations of diagnostic mammograms, a new study found.
The differences weren't due to differences in patients so much as they were due to differences in the radiologists' experience and affiliation.
The authors of the study, published in the Dec. 11 issue of the Journal of the National Cancer Institute, did find certain factors that predicted more accurate mammography readings.
"This is a little bit of a warning that there is a wide variation," said Diana Miglioretti, study lead author and an associate investigator at Group Health Center for Health Studies in Seattle. "If possible, go to a facility with a breast-imaging specialist, someone who also does biopsies and ultrasounds."
"You definitely want to have radiologists interpreting mammograms who do a substantial number of mammograms in a year -- that's number one," added Dr. Jay Brooks, chairman of hematology/oncology at Ochsner Health System in Baton Rouge, La. "Number two, you want to make sure that they read mammograms in a continuum so that they have previous mammograms to read it against. That's why I encourage patients to stay in one system where X-rays are looked at longitudinally. The third thing is that the radiologist is not reading this by him or herself, that there are other radiologists that can come into the room and look at the films."
Previous studies have also shown variability in how radiologists interpret mammograms, but those studies primarily involved screening mammograms. Diagnostic mammograms are performed after an abnormality of some kind has already been detected either from a screening mammogram or a physical exam.
The accuracy of diagnostic mammograms is extremely important, given that the rate of breast cancer is 10 times higher with those images than in screening mammograms, the study authors said.
For the study, the researchers looked at the records of 123 radiologists who had collectively interpreted 35,895 diagnostic mammograms at 72 U.S. facilities.
Close to 80 percent of breast cancers were diagnosed correctly. But sensitivity -- or the ability to accurately detect cancer -- ranged from 27 percent to 100 percent for different radiologists. False-positives (biopsies performed when there was no cancer) ranged from zero to 16 percent.
But 4.3 percent of women who had no cancer were tentatively told they had the disease, based on the mammogram (i.e. they received a false-positive reading).
The strongest factor linked to better accuracy was if the radiologist was affiliated with an academic medical center. Such radiologists were correct 88 percent of the time, compared to 76 percent for other radiologists. Radiologists at academic institutions were also less likely to report false-positive findings.
But the study results have to be taken with a grain of salt, the authors said, because only seven radiologists affiliated with academic institutions were represented in the study. This is more or less in keeping with the "real world," where academic radiologists interpret only 6.5 percent of mammograms across the nation, the authors said.
Radiologists who spent 20 percent or more of their time on breast imaging were more accurate than those who spent less time on breast imaging -- 80 percent versus 70 percent.
Interestingly, more experienced radiologists were less likely to recommend biopsies, but they missed more cancers than those with less experience, the study authors said.
"Radiologists who had been in practice longer had a lower threshold for recalling women for biopsies," Migliorettii said. "They had fewer false-positives but lower sensitivity, meaning they missed more cancers."
The Radiological Society of North America has more on diagnostic mammography.
SOURCES: Diana Miglioretti, Ph.D., associate investigator, Group Health Center for Health Studies, Seattle; Jay Brooks, M.D., chairman of hematology/oncology, Ochsner Health System, Baton Rouge, La.; Dec. 11, 2007, Journal of the National Cancer Institute
All rights reserved