Clinical Decisionmaking

Radiologists vary in how accurately they interpret diagnostic mammograms

Women who need diagnostic mammograms to diagnose or rule out breast cancer may be best served when their radiologist's primary affiliation is with an academic medical center, a new study finds. A radiologist's years of experience and focus on breast imaging were also beneficial in diagnosis. Researchers explored radiologists' variability in interpreting diagnostic mammograms, which are taken when a woman has a clinical sign or symptom such as a breast lump.

The researchers linked survey responses of 123 radiologists from 72 facilities to their performance in reading 35,895 diagnostic mammograms from January 1, 1996, to December 31, 2003, by using 3 mammography registries in Washington, New Hampshire, and Colorado.

The median sensitivity for detecting breast cancer was 79 percent for 54 radiologists who read at least 10 mammograms associated with a cancer diagnosis. The median rate for false positives (recommendation for a biopsy when there was no cancer) for the 118 radiologists who read at least 10 mammograms not associated with a cancer diagnosis was 4.3 percent. The seven radiologists affiliated with academic medical centers had the highest sensitivity (88 percent) in the study with slightly higher false positive rates (7.8 percent). The authors suggest that working in a teaching hospital and following up with patients may allow this group to better identify lesions likely to be cancerous.

Thresholds for what radiologists considered abnormal varied. For example, radiologists with less than 10 years of reading diagnostic mammograms were more apt to recommend a biopsy than radiologists with a decade or more of experience. The annual volume of mammograms that a radiologist reviewed did not affect sensitivity or false positive rates.

The authors suggest that future studies should determine whether mammography readings may be improved through double reading of mammograms, continuing medical education, and direct feedback to radiologists. This study was funded in part by the Agency for Healthcare Research and Quality (HS10591).

See "Radiologist characteristics associated with interpretive performance of diagnostic mammography," by Diana L. Miglioretti, Ph.D., Rebecca Smith-Bindman, M.D., Linn Abraham, M.S., and others in the December 19, 2007, Journal of the National Cancer Institute 99(24), pp. 1854-1863.

Return to Contents
Proceed to Next Article