BOSTON – A good mammogram reader may do just as well at spotting cancers without expensive new computer systems often used for a second opinion, a new study suggests. Computerized mammography, now used for about a third of the nation's mammograms, too often finds harmless spots that lead to false scares, researchers found. That conflicts with earlier studies showing benefit from the systems.
"It looks like computer-aided detection might not be working like people thought it would," said lead researcher Dr. Joshua Fenton, a family doctor at the University of California-Davis, in Sacramento.
The findings, which appeared Thursday in the New England Journal of Medicine, touch on a rapidly spreading technology first marketed in 1998.
Known as computer-aided detection or CAD, it consists of a computer coupled with software that identifies suspicious spots on mammograms and visibly marks them.
Here's how it works. When mammograms are taken, radiologists first read the X-rays and make their own judgments. But they can then double-check with the computer system to see if they have missed anything that's worth examining further.
There usually isn't. Still, some studies have shown that CAD can turn up 10 percent to 20 percent more cancers. Patients often have no idea if this new technology is being used.
The researchers in this five-year study — backed by the federal government and the American Cancer Society — analyzed mammograms from medical centers in Washington state, Colorado and New Hampshire. Seven of 43 centers used CAD. The mammograms came from 222,135 women and included 2,351 with a cancer diagnosis within a year of their tests.
The researchers found that with computerized mammography, a third more women were called back for suspicious findings and 20 percent more got biopsies than with ordinary mammograms. That might be a good thing, if enough cancers turned up to justify the minor surgeries and anxiety surrounding them.
Yet the computerized method showed no clear capability to turn up more cancer cases than unaided readings: Four cancers were found for every 1,000 mammograms, whatever screening method was used. That means that CAD would give 156 more unneeded callbacks and 14 more biopsies for every additional cancer it finds. And though these extra cancers tend to be early ones that are easier to treat, many would never be threatening anyway.
Dr. Phil Evans, at the University of Texas Southwestern Medical Center, said that "most radiologists that use computer-aided mammograms understand there are many false positives."
Frustratingly, the study ultimately wasn't big enough to reach fully reliable comparisons between the rates of cancers found by the two methods. That means that bigger studies are needed to clarify whether computerized mammography finds enough additional cancers to make it worth all those false alarms and added cost. While the technology adds just $20 or so to a single mammogram, a CAD unit might cost $50,000 to $75,000.
Even so, Dr. Jay Baker, a Duke University radiologist who has studied the technology, said: "I don't think it's a huge stop sign to using CAD."
"CAD won't go away; it will have a place," agreed Dr. Ferris Hall, a mammogram specialist at Beth Israel Deaconess Medical Center, in Boston, who wrote an accompanying editorial. But he added, "This is a setback for it."
Whether computerized or not, periodic mammograms are recommended for healthy women every year or two once they reach age 40. Experts advise women to check the credentials of radiologists at the clinic they plan to use and look for places that do a high volume.