boston.com Your Life your connection to The Boston Globe
White Coat Notes: News from the Boston-area medical community
Comments
Send your comments and tips to whitecoat@globe.com
Categories


Blogger
Elizabeth Cooney is a health reporter for the Worcester Telegram & Gazette.
Contributors
Boston Globe Health and Science staff:
Scott Allen
Alice Dembner
Carey Goldberg
Liz Kowalczyk
Stephen Smith
Colin Nickerson
Beth Daley
Karen Weintraub, Deputy Health and Science Editor, and Gideon Gil, Health and Science Editor.
 Short White Coat blogger Ishani Ganguli
 Short White Coat blogger Jennifer Srygley
Week of: November 11
Week of: November 4
Week of: October 28
Week of: October 21
Week of: October 14
Week of: October 7

« Pregnancy history overlooked in stem cell studies, Tufts researcher says | Main | Today's Globe: Boston climate change, human role in global warming, El Salvador reunions »

Wednesday, April 4, 2007

Computers don't increase mammogram accuracy and may hurt, study says

Computer-aided detection systems not only failed to detect more breast cancer in women who had screening mammograms, researchers report in tomorrow's New England Journal of Medicine, but it also may have harmed them by generating a higher number of false-positive readings, resulting in significantly more call-backs for repeat mammograms and biopsies.

Dr. Joshua J. Fenton of the University of California, Davis, and his co-authors studied CAD in 429,345 mammograms, the largest trial to date since it was approved by the Food and Drug Administration in 1998. They found that about 157 women would be called back for another mammogram and 15 women would undergo biopsies in order to detect one additional case of cancer, possibly a ductal carcinoma in situ, which is noninvasive and highly treatable.

Fenton's results "constitute a substantial hit to this technology," Dr. Ferris M. Hall of Beth Israel Deaconess Medical Center writes in an editorial in the same issue. Early studies had said CAD helped detect 10 to 15 percent more cases of breast cancers.

CAD programs analyze digitized mammograms and mark suspicious areas for review by radiologists.

CAD picks up a disproportionate number of clustered microcalcifications, which are the sign of ductal carcinoma in situ. DCIS makes up 25 percent of breast cancers diagnosed in the United States, Hall writes, but DCIS may never progress to cancer in a woman's lifetime.

"CAD does find a few more cancers in DCIS, so it's got some usefulness, but in all likelihood it's not going to save many lives," he said in an interview.

A study of CAD and mortality will take more time, Hall said, as it did for screening mammography.

"Mortality is the gold standard," he said. "Finding extra cancers does not prove that you have affected anything. You may have found a cancer that you would find next year anyway."

MRI screenings do a better job, Hall said, but cost and the need for interpretive expertise are delaying its adoption, as they did for mammography before it was accepted, he said.

Posted by Elizabeth Cooney at 05:00 PM
Sponsored Links