Mammogram Study Evaluates Computer-Aided Detection
Researchers are reporting that a computer system created to help radiologists interpret mammograms may not be helping after all. The system, known as computer-aided detection, or CAD, uses software to mark suspicious spots on mammograms that could be overlooked by radiologists.
In the first large-scale study of how well the technology works in the clinic, the researchers found that it did not improve the detection of breast cancer. Rather, the use of CAD led to significantly more false-positive mammograms than when radiologists relied on their own experience.
With the technology, 20 percent more women had biopsies than when the system was not used, even though most of these women did not have breast cancer. Dr. Joshua J. Fenton of the University of California, Davis, and his colleagues reported their findings in the April 5 New England Journal of Medicine (NEJM).
"Our goal was to see how computer-aided detection is working in clinical practice, and this study tells us that it's probably not working the way we expected," says co-author Dr. Stephen Taplin of NCI's Division of Cancer Control and Population Sciences (DCCPS).
The findings will surprise and disappoint most mammographers, says Dr. Ferris Hall of Beth Israel Deaconess Medical Center in an accompanying editorial. He points out that the use of CAD not only failed to increase the cancer-detection rate, but also was harmful because it increased the number of false-positive mammograms, which resulted in more testing and biopsies.
CAD systems, which cost between $50,000 and $175,000, are increasingly common in relatively large mammography centers. The Food and Drug Administration approved the technology in 1998 based on limited data, and Medicare began to pay for CAD soon after. Within 3 years of FDA approval, 10 percent of U.S. mammography facilities had adopted CAD, and others have followed.
It is unfortunate, Dr. Fenton notes, that the technology has come into widespread use before researchers could be certain of its clinical benefits. His team and Dr. Hall have called for larger studies to determine whether the routine use of CAD does more good than harm.
The researchers analyzed 429,000 mammograms and 2,351 cases of cancer detected at 43 facilities of the Breast Cancer Surveillance Consortium between 1998 and 2002. During this period, seven of the facilities implemented CAD, which allowed the researchers to compare results before and after. The other facilities were the control group.
The study did not evaluate whether CAD might save lives. But the results reinforce the need to resolve a major controversy in breast cancer: the clinical implications of detecting ductal carcinoma in situ (DCIS), a precancerous condition that may lead to invasive cancer, but also may never cause harm. It is not possible to identify which ones will lead to invasive cancer at this time.
The use of the technology increased the detection of these lesions, but it did not identify more invasive cancers. This raises the question of whether detecting more DCIS will save lives. The editorial suggests that it may not substantially reduce deaths, because detecting DCIS accounts for only about 10 percent of the reduction in deaths associated with screening.
"The critical question is whether detecting more DCIS helps save lives, and we do not have an answer right now," says Dr. Taplin.
The researchers estimate that with CAD, 157 women would be recalled (and 15 would undergo a biopsy) in order to detect 1 additional case of cancer, which might be DCIS. This scenario could increase the national costs of mammography screening by approximately $550 million annually.
In addition to the economic costs, a false-positive mammogram can have long-term effects on a woman's health. Women who receive false-positive mammograms tend to become more anxious and worried about breast cancer, and the concerns can linger for many years, researchers reported in the April 3 Annals of Internal Medicine.
"Receiving a false-positive mammogram can have an enduring effect on a woman's behavior and her well-being," says Dr. Noel Brewer of the University of North Carolina, Chapel Hill, who led the study.
For many women in the NEJM study, the result of a false-positive mammogram was not just some anxiety, but also a biopsy. "These women are not just getting a false-positive and becoming worried - they're also having a procedure," says Dr. Taplin. "We need to understand the implications of this."
One concern that he and others are investigating is whether breast biopsies may distort breast tissue and affect the interpretation of future mammograms.
By Edward R. Winstead