1. Home>
2. Specialties>
3. Women's Health>
4. Summary and Comment

Is Computer-Aided Detection Good for Screening Mammography?

CAD seemed to raise mammography recall rate in a U.K. study, but the implications for U.S. practice are unclear.

When two radiologists interpret the same screening mammogram, sensitivity is enhanced. Although double-reading is common in Europe, this procedure is not performed widely in the U.S. In a U.K. study, women who were scheduled for screening mammography were randomized to have their films read by one reader using computer-aided detection (CAD), read by two readers, or read by one reader with CAD and also double-read. Mammogram readers (radiologists, breast cancer clinicians, and trained technologists) recorded results as "recall for further assessment" or "return to routine screening." Discordant double-reading results were arbitrated by a third reader or a pair of readers.

In 28,204 evaluable participants, 227 cancers were detected for an overall screening detection rate of 8.0 per 1000. The cancer detection rate for single-reading with CAD was similar to that for double-reading (7.0 and 7.1 per 1000; P=0.89); likewise, the proportions of cancers detected were similar for both reading modalities (87.2% and 87.7%; P=0.89). Recall rates were 3.9% for single-reading with CAD and 3.4% for double-reading (P<0.001). The distribution of histologic type, tumor size or grade, or lymph node status was similar for tumors that were detected with single-reading with CAD alone or with double-reading alone.

Comment: Breast screening practice in the U.K. (i.e., regular use of double-reading as well as use of nonphysician readers) is substantially different from that in the U.S., which makes the implications of this study’s findings for U.S. practice unclear. Although breast cancer detection rates were comparable for single-reading with CAD and double-reading of screening mammograms, recall rates were higher for mammograms read by a single reader using CAD. This finding echoes those of other reports; nonetheless, the authors of a 2007 U.S. study suggested that CAD was potentially harmful in raising screening mammography recall rates (JW Womens Health May 10 2007). Moreover, routine use of CAD is expensive. Screening modalities that maintain or enhance sensitivity without heightening recall rates would be most advantageous.

— Andrew M. Kaunitz, MD

Published in Journal Watch Women's Health October 1, 2008

Citation(s):

Gilbert FJ et al. Single reader with computer-aided detection vs. two readers of screening mammograms. N Engl J Med 2008 Oct 2; [e-pub ahead of print]. (http://dx.doi.org/10.1056/NEJMoa0803545)

Image Gallery