UW News

December 31, 1996

Interpreting mammograms: Radiologists less accurate in detecting breast cancer when they know patient’s medical history

Knowing a patient’s medical history has a small but significant impact on radiologists’ interpretations of mammograms and recommendations for followup,a new study shows.

Dr. Joann G. Elmore, assistant professor of medicine at the University of Washington, and colleagues report their findings in the Jan. 1 issue of the Journal of the American Medical Association.

“Our study found that radiologists are much less accurate in reading mammograms when they know the patient’s clinical history,” said Elmore. “They are adversely biased by the history; it causes them both to miss cancers and to recommend additional work-ups in patients who turn out not to have cancer.”

On two occasions, separated by a five-month “wash-out” period, the researchers gave 10 radiologists mammograms from the same 100 women. The 10 radiologists had a wide range of experience reading mammograms, from 18 months to 20 years.

The mammograms were randomly divided into two groups of 50. For one group, the women’s detailed medical histories were supplied for the first reading and omitted (except for age) for the second reading. The sequence was reversed for the other group. The radiologists were not told they were seeing the same mammograms twice, and were not told what the researchers were endeavoring to measure.

Histories were classified as “alerting” or “non-alerting.” An alerting history contained a patient-reported abnormality (such as a breast lump or nipple discharge) or mention of a family history of breast cancer. A history without any of these features was designated as non-alerting.

The researchers found that radiologists’ recommendations for additional diagnostic testing were adversely influenced by the medical history. For example, they increased the number of work-ups they recommended in patients without cancer if an alerting history was present. Likewise, they were falsely reassured by a non-alerting history in patients who had cancer.

Five cases were shown a third time, with a deliberately false history. The false histories led an average of four of the 10 radiologists to change a previous diagnosis, and an average of one of the 10 to change a previous biopsy recommendation.

As a result of their study, the researchers recommend a simple change in radiologists’ reading of mammograms: that they examine the film and form a preliminary conclusion before reading the clinical history.

“We suspect that many radiologists read the clinical history first, examine the film, and then reach a conclusion,” they state. “… The history should always be made available to the radiologist, particularly for a patient with a breast lump or symptoms, which would help in the selection of additional mammographic views or films.

“If radiologists, however, delayed inspecting the clinical history until after forming an initial impression of the mammogram, (it) might reduce the additional testing provoked by an alerting history in women without cancer. Furthermore, radiologists could avoid the missed opportunities to diagnose breast cancer that can occur, as noted in our study, when a non-alerting clinical history creates false assurance.”

Co-investigators were Carolyn K. Wells and Dr. Alvan R. Feinstein of Yale University, and Dr. Debra H. Howard of New York Hospital-Cornell Medical Center. Their work was supported by the Robert Wood Johnson Foundation and the American Cancer Society.