New Feedback on High-Tech Cancer Screen
FDA Panel Listens to Debate on Computer-Aided Detection for Breast Cancer
March 4, 2008 -- Ten years after the first high-tech computerized device to
help with breast
cancer screening hit the U.S. market, researchers and regulators are still
trying to figure out just how useful they are.
An FDA advisory panel heard testimony today about the effectiveness of
computer-aided detection, known as CAD. Thousands of clinics and radiologists'
offices now use the devices; the idea is to help improve the chances that
suspicious areas on mammograms won't go unnoticed among the thousands of women
who get screened for breast cancer each day.
Hundreds of studies have been done on CAD since the FDA approved it in 1998.
But many are small or contain biases. Some studies suggest that CAD may improve
a radiologist's chances of catching a hard-to-see cancer. But others have not
shown better accuracy in computer-aided reading of mammograms, including a
study published last year.
CAD is vulnerable to false-positives: red flags that turn out not to be
cancer. That can lead to unneeded office visits, invasive biopsies, and lots of
anxiety for worried patients. But the companies that make
CAD products say their machines improve detection with only minimal
false-positives. And some radiologists back them up.
Part of the problem is that the vast majority of women who get routine
breast cancer screening don't have cancer. That leaves radiologists to sort
through mountains of "noise" in an effort to catch cancers early.
"The task of the mammographer is essentially to find a needle in a
haystack," Akira Hasegawa, an executive with Fujifilm Medical Systems USA,
told an FDA advisory panel Tuesday.
CAD doesn't replace a human radiologist. Instead, it uses a computer
algorithm to "mark" potentially worrisome areas and draw doctors'
attention to them.
Debate Over False-Positives
Doctors sometimes don't find cancers, said Robin Birdwell, an associate
professor of radiology at Brigham and Women's Hospital at Harvard Medical
"Having a tool to improve sensitivity is a good idea," said
Birdwell, who represents the Society for Breast Imaging, a group that receives
And this "good idea" is why researchers lament that few independent
studies are large enough or designed properly to answer the lingering questions
about the technology.
Researchers said they want the government to accelerate an effort to build a
massive database of mammography results to help researchers run unbiased
tests of how computer-aided screening stacks up against humans alone in the
"CAD works. It works well in certain specific areas," said Carl J.
d'Orsi, MD, a professor of radiology at Emory University and a member of the
FDA's advisory panel.
"I think we have to define more where it is good and where it is not
good," he said.
Researchers already know that false-positives can cause a lot of anxiety in
women who fear they may have cancer when they don't. What they don't know is
whether most women are willing to risk anxiety and more biopsies in exchange
for better cancer detection, said David K. Spindell, MD, vice president of
medical affairs at Abbott Pharmaceuticals and a panel member.
"That question really has to be flushed out," he said.