Computer-Aided Detection Doesn't Help Mammography

It leads to too many false-positive results in breast screenings, study finds

WEDNESDAY, April 4, 2007 (HealthDay News) -- Computer-aided detection in breast cancer screening does not offer much benefit and may actually reduce the accuracy of mammograms, a new report contends.

In the most comprehensive analysis of the technology to date, U.S. researchers found that it increased the rate of false-positive results from mammos, which led to more return visits and unnecessary biopsies.

"Our study certainly points to the need for additional studies to clarify whether there any benefits of computer-aided detection and, if so, what types of cancers are detected," said study lead author Dr. Joshua Fenton, assistant professor of family and community medicine at the University of California, Davis. "We need studies that explore whether any benefits of computer-aided detection are outweighed by its harms and costs."

Dr. Jay Brooks, chairman of hematology/oncology at the Ochsner Health System in Baton Rouge, La. added: "It looks like technology doesn't win out. CAD (computer-aided detection) is a very sensitive test, but it's not very specific, and we don't clearly have ways to prevent doing unnecessary biopsies or unnecessary testing on women."

Since its approval in 1998, use of computer-aided detection software has soared, with Medicare and many insurance companies now reimbursing for its use. Within three years of U.S. Food and Drug Administration approval, 10 percent of mammography facilities had computer-aided detection and more have undoubtedly followed suit.

Yet, there's no clear evidence supporting the accuracy of the technology, which helps radiologists interpret conventional mammograms by analyzing the image and highlighting suspicious areas.

For the new study, the most comprehensive analysis of computer-aided detection in breast screening to date, researchers looked at 222,135 women who had 429,345 mammograms at 43 different facilities in the United States between 1998 and 2002. The sample included 2,351 women who received a diagnosis of breast cancer within one year of screening.

During the study period, seven facilities -- 16 percent of the study sites -- implemented computer-aided detection.

With computer-aided detection, 32 percent more women were recalled for further testing, and 20 percent more women underwent a breast biopsy.

But for every real cancer identified by computer-aided detection, radiologists had to contend with 2,000 additional false-positives.

Computer-aided detection also didn't seem to improve the early detection of breast cancer. If anything, the software promoted the detection of the least dangerous forms of breast cancer, such as ductal carcinoma in situ. According to an accompanying editorial in the journal, only an estimated 10 percent of the decrease in deaths associated with screening mammography is linked to the diagnosis of ductal carcinoma in situ.

Using computer-aided detection at every mammography facility in the United States would result in an additional $550 million in costs, the study authors estimated.

The results are published in the April 5 issue of the New England Journal of Medicine.

"Mammography is still a great screening test. The problem is, it's not a perfect test," Brooks said. "We're trying to make it better, but this is an example where we've done something that doesn't really make it any better."

Nevertheless, he added, "we have still made major inroads, because breast cancer is diagnosed much earlier, and survival rates have improved tremendously. It's just that we can't always distinguish between what's nothing and what's not."

Dr. Debra Monticciolo, professor of radiology with the Texas A&M Health Science Center, said, "The results were not unexpected, although I didn't expect accuracy to be reduced as much as it was.

"I'm hesitant to change practice based on one study, but this has a lot of weight. We really need to pay attention to it. This is an unbiased, broad-based report that maybe we shouldn't count on CAD. This will raise a lot of questions as to whether it's useful," Monticciolo added.

More information

Visit the National Cancer Institute for more on detecting breast cancer.

SOURCES: Joshua J. Fenton, M.D., assistant professor of family and community medicine, University of California, Davis, Sacramento; Jay Brooks, M.D., chairman of hematology/oncology, Ochsner Health System, Baton Rouge, La.; Debra Monticciolo, M.D., professor of radiology, Texas A&M Health Science Center, College Station, and vice chair of research and chief of breast imaging, Scott & White Hospital, Temple, Texas; April 5, 2007, New England Journal of Medicine
Consumer News