NPR logo

Study Casts New Doubt on Mammogram Software

  • Download
  • <iframe src="" width="100%" height="290" frameborder="0" scrolling="no" title="NPR embedded audio player">
  • Transcript
Study Casts New Doubt on Mammogram Software

Health Care

Study Casts New Doubt on Mammogram Software

Study Casts New Doubt on Mammogram Software

  • Download
  • <iframe src="" width="100%" height="290" frameborder="0" scrolling="no" title="NPR embedded audio player">
  • Transcript

New research calls into question the computer technology that helps doctors interpret mammograms. The technology is supposed to spot breast cancers that a doctor might miss. But a study in this week's New England Journal of Medicine concludes that computer-aided detection programs don't work as advertised — and may do more harm than good.

The computer programs are supposed to be like having a second opinion on each mammogram. In a darkened room at Beth Israel Deaconess Medical Center in Boston, Dr. Ferris Hall demonstrates how computer-aided detection works.

"We're looking at this mammogram now. We're looking at all four views up at once. And I'm going to press this little button here — which gives us the computer-aided detection — and lo and behold, we can see that there's a circle on three of the images," Hall says.

That means the computer has spotted something the doctor should pay close attention to.

The problem is, most of the time it's not a problem.

Dr. Joanne Elmore of the University of Washington says that for every 1,000 marks doctors evaluate on a mammogram, only one may mark a cancer.

So she and her colleagues wanted to know the bottom line on computer-aided detection. They collected data from 43 mammography centers in three states to see if those that had adopted the computer technology did better.

Article continues after sponsorship

"We had hoped to find that more cancers would be detected," Elmore said. "Obviously, we do not like missing breast cancers. And unfortunately, we did not find that the programs improved the cancer detection rate."

Elmore says radiologists may be overwhelmed by the high numbers of marks that computer programs are putting on mammograms.

One thing is clear: Centers that use computer-aided mammography bring many more women back for more studies, with a 31 percent increase in the patient recall rate.

And even though the technology doesn't detect any more cancers, the number of breast biopsies went up 21 percent.

Elmore says women should ask whether their hospital or mammography center uses computer-aided detection, as about 40 percent of U.S. centers do. If so, they would know they have a higher chance of being called back for unnecessary tests.

"Some women do not like the experience of getting called back for additional testing," Elmore said. "They find it very stressful, time-consuming and expensive."

But other women, who may have just had a best friend diagnosed with breast cancer, "want to make certain nothing's missed... they don't mind putting up with additional testing if there's a very, very small chance of an improvement. I think each woman should make this decision on her own," Elmore said.

The new study indicates that computer-aided mammography is doing more harm than good, by exposing women to unnecessary tests and anxiety. But Elmore doesn't expect that it will cause centers to abandon the technology.

"So many facilities in the United States have already purchased these programs that we're at the stage where it's hard to make an abrupt u-turn," she said.

Boston radiologist Ferris Hall agrees.

"This is a substantial hit for computer-aided detection, but in no way is it dead," Hall says.

Hall says that's partly because years ago, when computer-aided mammography was new, Medicare and private insurers started paying extra for it. That's one big reason why the technology spread so fast.

"Many of us got this with the idea that we would help patients and pick up more things," Hall said. "But that wasn't proven. What we did know was we were going to be paid a little more. And that we would be keeping up with our neighbors who could advertise... computer-aided detection."

So the main message of the new study is: Doctors who use it need to be more skeptical about it.