CAD software serves as a "second
read" for screening mammography,
marking image patterns for the
radiologist to review. (Image
courtesy of Hologic, Inc.)

Real World Study Questions Effectiveness of CAD  Mammography

Computer-aided detection tools add to the cost of screening mammography for breast cancer but add little in terms of finding undetected harmful tumors, while increasing the number of false positives, a new study shows.

In what may be the largest study of real-world use of CAD mammography, researchers with the Breast Cancer Surveillance Consortium led by Dr. Joshua Fenton analyzed 1.6 million mammograms from 680,000 women in seven states.  Their study was published online today in the Journal of the National Cancer Institute. (Link to published site)

Fenton, who is an assistant professor in the UC Davis Department of Family and Community Medicine, told The Hub in an interview that CAD is designed to be an automated second reader of mammograms. He noted that in the United Kingdom, every mammogram is double-read, meaning two radiologists independently read mammograms for suspicious lesions.

That practice has been shown in studies to increase the number of cancers captured and decrease the number of false positives. In this study, however, the use of computer-aided detection technology for the “second read” produced the opposite result.

“After installation of CAD the number of (false-positives) detected increased by about 0.5 percent,” Fenton said by phone. “Another way to say that is that for every 200 women who get a screening mammogram, CAD would cause one additional woman to be recalled unnecessarily for further testing to exclude breast cancer.”

The study confirms an earlier smaller study Fenton and colleagues had published in the New England Journal of Medicine in 2007. That study was criticized as being too small, including only seven facilities using CAD, and for relying on earlier generations of the CAD technology.

The current study analyzed the results of film mammograms provided at 90 Breast Cancer Surveillance Consortium centers between Jan. 1, 1998 to Dec. 31, 2006. During that time 25 of the centers (27.8 percent) installed CAD technology and software, which was used for an average of 27.5 months during the study period. The film mammograms were scanned and digitally analyzed by the CAD software.

At facilities that never implemented CAD, radiologists increased specificity slightly but statistically significantly from 91.1 percent between 1998 and 2002 to 91.3 percent between 2003 and 2006. At the same time the recall rate declined statistically significantly from 9.3 percent in the first four-year period to 9.1 percent in the second period.

By contrast, at facilities that implemented CAD, specificity decreased statistically significantly from 91.9 percent to 91.4 percent, while the recall rate increased from 8.4 percent to 8.9 percent.

Those real-world results sharply contrast with the clinical studies done prior to FDA approval of CAD use. One study of 12,860 patients in a Community Breast Center published in Radiology 2001, (220:781-786) found that 26.2 percent of cancers missed by a radiologist would be detected with the use of CAD.

In an editorial accompanying the current study, Dr. Donald A. Berry, of the Department of Biostatistics at M.D. Anderson Cancer Center, wrote that the significance of Fenton’s study is that it reflects real-world results where the daily pressures on radiologists to perform quantitatively, likely has an impact on quality, while financial incentives also may play a role in the use of CAD.

“Why is CAD so popular?” Berry asked. “An obvious reason is that it is built into digital mammography equipment, which is increasingly common in the United States. Another is financial: In 2008, Medicare’s global reimbursement for CAD was $16.50. Still another is that CAD marks are comforting to the reader, even though the comfort may be misplaced.”

Since the FDA approved CAD for mammography in 1998, its use has jumped dramatically. Medicare data shows that since it began paying for CAD in 2001 at a reimbursement rate of $12 to $16.50 per screening mammogram Medicare’s cost for screening mammography has increased anywhere from 9 percent to 15 percent.

Outside of Medicare, Fenton said the information about how widespread the use of CAD mammography is not well known.

“We don’t have good national data on prevalence. Medicare data shows that that about 75 percent of mammograms for reimbursement include a charge for CAD,” he said. “Those are obviously all women over 65, so we don’t know what the prevalence is over the general population. I would guess it is probably around 50 percent.”

In any case Fenton says the additional cost to the healthcare system is substantial.
He has seen the startup and capital equipment costs for CAD range from $75,000 to $200,000 for installation. After that there are additional fees for the interpretation that range from $12 to $16 dollars for Medicare recipients, and may range from $25 to $45 for those covered by private health insurance. And finally there is the cost incurred by additional diagnostic testing for women who receive false positive tests.

“Our study raises important questions about how CAD is being used in practice right now,” Fenton said. “This is a study of real world utilization and the effects in real world practice. And our study suggests that the benefits of CAD are not being realized in practice right now. It is important for us to figure out why. Is it a limitation of the technology, or is not being used as it was designed, or is it a combination of both of these factors.” (Link to published site)