The research was led by a UCSF team and is published in the March 2 issue of the Journal of the National Cancer Institute.
Screening mammograms are the type most women get. These images are designed to discover any cancer that might be hidden in women with no breast symptoms. Women who have a history of breast cancer or other symptoms -- or who experience symptoms at the time of their exam -- receive diagnostic mammograms. After two to five asymptomic years, their physicians will usually revert to screening mammography.
Study findings showed that the more experienced physicians not only properly identified cancer, they also were 50 percent less likely to falsely characterize a mammogram as abnormal. In addition, the study found dramatic variability in physician accuracy. For example, physicians on average detected 77 percent of cancers, but for individual physicians the detection rate ranged from 29-97 percent. The average false positive rate was 10 percent, while the rate for individual physicians ranged from 1-29 percent.
According to the UCSF researchers, it has already been well documented in scientific literature that patient characteristics -- such as breast tissue density -- can be negatively associated with mammogram accuracy and those characteristics largely cannot be changed. They chose, therefore, to focus their study on what physician characteristics could be associated with accuracy. They studied more than 1.2 million mammograms paired with data from statewide cancer registries in four states.
The study found that physicians who read a greater number of mammograms identified cancers without mistaking normal tissue for a malignancy with greater accuracy than physicians whose volume was not as high. In general, the false-positive rate declined with increasing physician age, with increasing time since receipt of medical degree, and with increasing annual volume. Additionally, those physicians who focused their practice on screening, as opposed to diagnostic mammography, had a lower false-positive rate.
Study data came from four sites that participate in the National Cancer Institute- funded Breast Cancer Surveillance Consortium. Those sites -- San Francisco, New Mexico, Vermont and Colorado -- prospectively collect mammographic interpretations. Researchers linked the interpretation data to cancer registries and then to physician information from the American Medical Association. They studied 209 physicians who interpreted 1.2 million mammograms from January 1, 1995 through December 31, 2000 including 7143 that were associated with breast cancer.
The minimum number of mammograms a US physician can read and still be considered qualified under FDA regulations is 960 within a two-year period, or approximately 10 mammograms per week, according to the Mammography Quality Standards Act of 1992.
"Raising the annual volume requirements in the Act might improve the overall quality of screening mammography in the US," said principal investigator Rebecca Smith Bindman, MD, UCSF associate professor of radiology, epidemiology/biostatistics, and obstetrics, gynecology and reproductive sciences. She also is director of the UCSF Radiology Outcomes Research Lab and part of the breast oncology program of the UCSF Comprehensive Cancer Center.
The authors hypothesize that physicians who read more mammograms overall and who have been reading them longer develop a higher threshold for identifying cancer and overall do better. Those who read a higher proportion of diagnostic mammograms, they say, may tend to develop a lower threshold, expecting a higher rate of cancer.
"Compared with those who read the minimum number allowed by the Act and who have less focus on screening, physicians who focus on screening and who interpret 2,500-4,000 mammograms annually have 50 percent fewer false-positive diagnoses while detecting approximately one less cancer per 2,500 mammograms," the authors wrote.
Smith Bindman added, "The variation we found was dramatic. We recommend explicit discussion of what the goals of mammography should be, and sooner rather than later."
If the goal is to maximize the identification of cancer while achieving a reasonable false-positive rate, raising the minimum required under the Act would make sense, according to Smith-Bindman. Although fewer doctors would subsequently be available for reading mammograms, any resulting increased barriers to healthcare delivery could be managed, she said. "Change would need to occur slowly to prevent a shortage of physicians. But the payoff would be fewer false alarms for women."
The study was funded in part by the National Cancer Institute and the Department of Defense.
Additional authors are Philip Chu, UCSF; Diana L. Miglioretti, University of Washington; Chris Quale, UCSF; Robert D. Rosenberg, University of New Mexico; Gary Cutter, University of Nevada; Berta Geller, University of Vermont; and Peter Bacchetti, Edward A. Sickles and Karla Kerlikowske, UCSF.