CHICAGO (MCT) — Doctors reading mammograms miss an average of 2 in every 10 cases of breast cancer, even for women with lumps and other symptoms, researchers reported Tuesday.

In a stark reminder of the limitation of the common diagnostic test, the researchers found wide variation in radiologists' ability to detect cancer in breast X-rays, with some missing as many as 7 out of 10 cases. In other words, the test's ability to detect cancer is strongly dependent on who is reading it.

"Women think mammography is perfect, so if they get a negative (normal) mammogram, they think they're safe for at least the next year," said Diana Miglioretti, lead author of the study, which appears in this week's Journal of the National Cancer Institute. "The reality is, they shouldn't be falsely reassured by a negative mammogram."

That message was seconded by Dr. Leonard Berlin, chief of radiology at Rush North Shore Medical Center in north suburban Skokie, Ill., who was not involved in the study.

"You can have cancer and still have a normal mammogram," Berlin said. "If you have any sign or symptom, you need to pursue it. That's the bottom line."

Researchers have known for years that two mammographers looking at the same images will often come up with different results. For example, a 1994 study showed that when 10 radiologists were asked to read the same 150 mammograms, there was substantial disagreement in 25 percent of the cases over what the mammogram showed.

But most previous studies were done in screening mammograms, routine tests performed on women who are assumed to be healthy. Little was known about the accuracy of diagnostic mammograms, which are done to evaluate symptoms or rule out cancer in women with a suspicious screening mammogram.

Miglioretti and her colleagues examined the performance of 123 radiologists who interpreted nearly 36,000 diagnostic mammograms from 1996 to 2003 at 72 U.S. facilities. All the mammograms had been ordered for women who found a lump themselves or whose doctors discovered something of concern.

The researchers found that sensitivity — the ability to detect cancer when it is present — ranged from 27 percent to 100 percent, with a median of 79 percent. The false-positive rate — women who got a tentative diagnosis of cancer when they did not have it — ranged from zero to 16 percent, with a median of 4.3 percent. (A definitive diagnosis of cancer depends on a biopsy.)

The radiologists who were most accurate — that is, had the highest sensitivity without too many false alarms — tended to be those based at academic medical centers, followed by those who spent at least 20 percent of their time on breast imaging.

But even for radiologists at teaching hospitals, median sensitivity was only 88 percent, which means that on average 12 out of every 100 cases of breast cancer in symptomatic patients were not detected.

And in the U.S. — unlike most European countries — the vast majority of mammograms are read by general radiologists in the community, not by specialists at academic centers. In the current study, one-quarter of the radiologists read fewer than 1,000 mammograms a year.

"A lot of American women can't see a breast imaging specialist, especially in rural areas," said Miglioretti.

Miglioretti, a biostatistician at Group Health Center for Health Studies in Seattle, said her results were worrisome, especially since the women whose mammograms were examined in this study all had some reason to believe there was a problem.

"Because the rate of breast cancer is tenfold higher among diagnostic mammograms than among screening mammograms, and the majority of women with breast cancer have a physical sign or symptom at the time of diagnosis, this variability in interpretive performance is concerning," the researchers wrote. "We need to identify ways to improve accuracy and reduce variability among all radiologists who interpret mammograms."

Miglioretti said one option might be to provide additional training to general radiologists. But she cautioned that even the best mammographer is going to miss some cancers.

"It's not necessarily the mammographer's fault," she said. "It's the limitation of the technology."