Computer-Aided Detection Cut Mammogram Accuracy


Computer-aided detection decreased rather than improved the accuracy of mammogram interpretation in a nationwide study involving over 429,000 screening mammograms.

Compared with standard mammography, computer-aided detection (CAD) mammography “was associated with significantly higher false-positive rates, recall rates, and biopsy rates and with significantly lower overall accuracy,” Dr. Joshua J. Fenton and his associates reported. In an editorial comment accompanying the report, Dr. Ferris M. Hall termed the study “the most comprehensive analysis of computer-aided detection in breast screening to date,” and characterized the results as surprising and disappointing. They may not spell the demise of CAD mammography, “but they constitute a substantial hit to this technology,” he said.

According to Dr. Fenton and his associates, CAD has been incorporated into mammography practices rapidly, despite only “tentative” evidence of its clinical benefit, in part because it is Food and Drug Administration approved and its use is reimbursed by Medicare and insurers.

He and his associates studied the results of routine screening mammograms and cancer outcomes for over 222,000 women screened at 43 facilities participating in the Breast Cancer Surveillance Consortium in 1998–2000. During that time, seven of the facilities (16%) implemented CAD. The study design allowed for review of the real-life experience at numerous, diverse facilities with over 150 radiologists across the country, said Dr. Fenton of the University of California, Davis, and his associates.

A total of 2,351 women were diagnosed as having invasive breast cancer or ductal carcinoma in situ within 1 year of their screening mammograms.

The use of CAD proved to be “of uncertain clinical benefit,” the researchers said (N. Engl. J. Med. 2007;356:1399–409).

The technique raised the rates of false-positive results and patient recalls for further assessments, and raised the biopsy rate by nearly 20%. Diagnostic specificity decreased from 90% before implementation of CAD to 87% afterward, and the positive predictive value of screening mammograms declined from 4% to 3%.

CAD slightly increased the diagnostic sensitivity of mammography, but the difference was nonsignificant and was largely accounted for by a slight increase in detection of carcinoma in situ rather than in invasive breast cancer, Dr. Fenton and his associates said.

In his editorial comment, Dr. Hall of Beth Israel Deaconess Medical Center and Harvard Medical School, Boston, concurred with the researchers that larger, controlled trials of CAD are needed, with a particular eye toward determining whether use of the technology has an impact on mortality.

“But such studies will be expensive, controversial, indeterminate, or quickly passé owing to the emergence of new technology. It took 2–3 decades of controversy before it was proved that screening mammography saves lives,” he noted (N. Engl. J. Med. 2007;356:1464–6).

Next Article: