Study Questions Accuracy of Mammogram Exams

Apr 06, 08:59 AM

Current Headlines: By Gina Kolata

A highly touted and widely used computerized system for examining mammograms is leading to less accuracy, not more, a new study has found.

The system, known as computer-aided detection, did not find more breast cancer, according to researchers. But it did lead to many more false alarms that resulted in additional testing and biopsies for spots on mammograms that turned out to be harmless.

Such detection systems, approved by the Food and Drug Administration in 1998, are sold by several companies, including Hologic of Bedford, Massachusetts; iCAD, of Nashua, New Hampshire; and Eastman Kodak of Rochester, New York. According to the National Cancer Institute, the systems are now being used in about 30 percent of mammography centers.

The equipment is expensive, costing $50,000 to $175,000, but Medicare pays an extra $20 for each mammogram read with it, making it profitable for large centers to use it. Doctors also worry about lawsuits if they were to not use it and miss a cancer.

But all along, as more and more mammography centers bought the software, the assumption was that the computer would find cancers that radiologists would miss, saving women's lives.

The new findings are likely to surprise radiologists, said Dr. Ferris Hall, a radiology professor at Harvard Medical School. Hall wrote an editorial accompanying the paper, which was being published Thursday in The New England Journal of Medicine.

"I was surprised," he said. "A lot of people will be amazed."

But executives at the company whose equipment was used in the study, Hologic, said they interpreted the results differently.

If there is a suspicious spot on a mammogram, women will want to have a biopsy to rule out invasive cancer, said Robert Cascella, the company's president and chief operating officer. And the study showed that computer-aided detection was finding proportionately more very early precancerous growths. "That's a valuable finding," Cascella said. The company has improved its software since the study was conducted, he added.

The lead author of the study, Dr. Joshua Fenton of the University of California, Davis, emphasized that women should continue to have mammograms. But, Fenton said, they may want to ask if their mammography center uses computer-aided detection. His study, he said, "does raise concerns that technology is causing harm without clear benefit."

The new look at computer-aided detection is the latest sally in the changing era of breast cancer detection. New technology, like computer-aided detection and digital mammography and ultrasound, can be so sensitive that doctors have trouble deciding which findings are worrisome and which are not. The only screening method that has been rigorously evaluated is old-fashioned X-ray film mammograms, but it is likely to be replaced by something, or some combination of things, whose benefits and risks are largely unknown.

"We are getting ourselves out on thinner and thinner ice," said Dr. Suzanne Fletcher, a professor emeritus of ambulatory care and prevention at Harvard Medical School. "With mammography, we have multiple studies showing this is what happens to mortality rates if you get this versus if you don't. With these newer technologies, we don't."

The new study of computer-aided detection was an analysis of 429,345 mammograms obtained from 1998 to 2002 at 43 mammography centers. During that time, seven of the centers switched to computer- aided detection. That enabled the investigators to compare results with and without computer software to help radiologists find suspicious spots.

Computer-aided detection, the researchers wrote, "was associated with significantly higher false positive rates, recall rates, and biopsy rates and with significantly lower overall accuracy."

With computer-aided detection, 31 percent more women were called in for additional tests and 20 percent more had biopsies.

And there was another potential problem: Computer-aided detection did not clearly increase the detection of breast cancer. If anything, it seemed to increase the detection of a precancerous condition, DCIS, for ductal carcinoma in situ.

While all invasive breast cancer is believed to start as DCIS, DCIS is often harmless.

At times, it either never develops into cancer or it grows so slowly that it is not a danger during the woman's lifetime. But, unable to tell which lesions will become deadly, doctors generally treat them all.

(c) 2007 International Herald Tribune. Provided by ProQuest Information and Learning. All rights Reserved.