Mammogram tool improves some breast cancer detection but also increases false alarms

Published in the April 16 issue of the Annals of Internal Medicine, the study of women enrolled in Medicare is the largest comparison of clinical outcomes of routine screening mammography with and without computer-assisted detection (CAD) -- software developed to enhance the detection of breast cancer during screening mammography.

CAD was most strongly associated in the study with higher incidence of ductal carcinoma in situ (DCIS), a non-invasive breast lesion typically referred to as "stage 0" because it does not affect surrounding tissue or mortality.

"DCIS progresses slowly, if at all," said lead author Joshua Fenton, an associate professor of family and community medicine at UC Davis and a national leader in research to improve the quality of health-care services. "Some of these early noninvasive lesions may never have come to clinical attention in women's lifetimes if CAD were not applied to their mammograms."

CAD also was associated with slightly higher rates of early-stage invasive breast cancer detection.

"There may be benefits if CAD detects early-stage invasive cancer before it progresses," said Fenton. "A longer-term study would be needed to see whether fewer women die of breast cancer on account of the technology."

The study also showed that CAD mammography was associated with increased diagnostic testing, including breast biopsy, among women who did not have breast cancer.

"This means that CAD increases the chances of being unnecessarily called back for further imaging or tests because of a false alarm, which is already a major problem without CAD," said study co-author Joann Elmore, a professor of internal medicine and epidemiology at the University of Washington and specialist in breast cancer screening. "No woman likes receiving a notice saying that there was an abnormality on her mammogram that might or might not be cancer and more testing is necessary. This can understandably be a very stressful experience for women."

CAD algorithms are designed to assist radiologists by marking areas on mammograms that could be associated with breast cancer. Application of CAD has increased rapidly since being approved for Medicare reimbursement in 2001, and it is now used on approximately three out of every four screening mammograms in the U.S. Despite this broad acceptance, some health-care providers have questioned if the benefits of CAD outweigh its annual direct costs to Medicare, which are estimated by Fenton based on CAD prevalence and Medicare reimbursement and enrollment data to exceed $100 million each year.

For the study, Fenton, Elmore and their colleagues analyzed Medicare claims data that were linked with a National Cancer Institute database of cancer diagnoses, treatments and survival. More than 163,000 women nationwide who were 67 to 89 years old when they underwent mammography with or without CAD during 2001 to 2006 were included in the study. The participants received more than 409,000 mammograms -- an average of 2.4 mammograms per person -- during the six-year period.

"The large sample size gave us the opportunity to provide precise estimates of CAD's clinical impact at the national level," said Fenton.

Compared to women screened without CAD, women screened with CAD had a 17 percent increase in diagnoses of noninvasive DCIS breast lesions and a 6 percent increase in diagnoses of early-stage invasive breast cancer. Among women who did not have breast cancer after all, there was a 19 percent increase in additional diagnostic imaging after screening and a 10 percent increase in breast biopsies.

Fenton and Elmore expect their research will inspire debate about the value of Medicare's investment in CAD. They hope their findings also emphasize the need for research to distinguish early-stage breast cancers that are likely to progress from those that are likely to follow a benign course.

"CAD is expensive technology that has been nearly universally adopted in the U.S. due to Medicare's support and the hope that it can help us identify and treat invasive breast cancer early," said Fenton. "Our study suggests that we still don't know whether the benefits outweigh the harms for the average woman on Medicare."

UC Davis Health System co-authors on the study were Guibo Xing, Heejung Bang and Karen Lindfors; co-author Steven Chen is with the City of Hope Medical Center in Duarte, Calif.; and senior author Laura-Mae Baldwin is with the University of Washington in Seattle.

A copy of "Computer-Aided Detection and Short-Term Screening Mammography Outcomes: A Population-Based Study of Medicare Enrollees" can be requested via e-mail to acollom@acponline.org.

The study was funded by the Clinical and Translational Science Center, Center for Healthcare Policy and Research and Department of Family and Community Medicine at UC Davis, and by a National Cancer Institute grant (K05 CA-104699).