Computer-aided detection (CAD) technology fails to spot breast tumors and increases a woman's risk of being called back for needless post-mammography testing, according to a new study published online today in the Journal of the National Cancer Institute.
Computer-aided detection (CAD) technology fails to spot breast tumors and increases a woman's risk of being called back for needless post-mammography testing, according to a new study published online today in the Journal of the National Cancer Institute.
A team led by University of California, Davis assistant professor Joshua Fenton, MD, analyzed 1.6 million mammograms in seven states, providing what the researchers called “the most definitive findings to date on whether the popular mammography tool is effective in helping find breast cancer.” The Breast Cancer Surveillance Consortium provided the mammograms and associated patient data.
The study examined screening mammograms performed on more than 680,000 women at 90 mammography facilities from 1998 to 2006. The false-positive rate typically increased from 8.1 percent before CAD to 8.6 percent after CAD was installed at the facilities in the study. In addition, the detection rate of breast cancer and the stage and size of breast cancer tumors were similar regardless of CAD.
“In real-world practice, CAD increases the chances of being unnecessarily called back for further testing because of false-positive results without clear benefits to women,” Fenton said. “Breast cancers were detected at a similar stage and size regardless of whether or not radiologists used CAD.”
The authors said the results of real-world studies of CAD may differ from results from clinical trials, possibly because radiologists in practice don't adhere as strictly the CAD-use guidelines as do radiologists in protocol-driven studies.
Computer-aided detection software, approved by the FDA in 1998, analyzes the mammogram image and marks suspicious areas for radiologists to review. Its use skyrocketed after Medicare began covering it in 2001. CAD is now applied to the large majority of screening mammograms in the United States, with annual direct Medicare costs exceeding $30 million, according to a 2010 study in the Journal of the American College of Radiology.
According to 2009 Medicare data, insurers including Medicare typically paid about $12 per screening mammogram for CAD in addition to the costs of the mammogram (about $81 for film mammography and $130 for digital mammography), representing a 9 percent to 15 percent additional cost for CAD use.
The current study builds on Fenton's initial assessment of the technology published in the New England Journal of Medicine in 2007. That report, which examined mammography screening results in 43 facilities, including seven that utilized CAD, found that CAD was associated with reduced accuracy of interpretation of screening mammograms but no difference in the detection rate of invasive breast cancer.
Critics of the research findings in the New England Journal of Medicine said the study was based on use of an older kind of CAD technology, and so did not accurately reflect its usefulness. The new study evaluated newer technology for a longer period of time, Fenton said.
In an accompanying editorial, Donald Berry, PhD, of the M.D. Anderson Cancer Center in Houston said the benefit of CAD is likely to be difficult to detect, and that improving the sensitivity of CAD might find less aggressive tumors, early detection of which wouldn't impact breast cancer mortality. He argued that researchers and companies should work to improve the software, but added, “this should happen in an experimental setting and not while exposing millions of women to a technology that may be more harmful than it is beneficial. In the meantime, economic incentives may stoke its continued proliferation.”
Read the full article here, and the accompanying editorial here. (PDF)
Study Assesses Lung CT-Based AI Models for Predicting Interstitial Lung Abnormality
September 6th 2024A machine-learning-based model demonstrated an 87 percent area under the curve and a 90 percent specificity rate for predicting interstitial lung abnormality on CT scans, according to new research.