August 1, 2011 — The use of computer-aided detection software to help analyze and interpret mammograms does not improve accuracy for all types of breast cancer and is associated with more false-positives among some kinds, according to a study published online in the Journal of the National Cancer Institute, CNN reports. The study lends credence to earlier findings by the same researchers that also found the software was flawed.
According to a 2010 analysis of Medicare data, the software is used to read roughly 75% of screening mammograms in the U.S. (CNN, 7/27). The technology is used to identify suspicious areas on the mammogram for further investigation and to provide a backup for a radiologist's own reading.
For the new study, researchers analyzed 1.6 million mammograms performed at 90 facilities -- 25 of which adopted the software -- between 1998 and 2006 (Hobson, "Health Blog," Wall Street Journal, 7/28). The researchers addressed criticism directed at the earlier study by including a large number of facilities that used the software and excluding mammograms read within three months of a facility adopting the software.
According to the study, the detection rate for noninvasive breast cancers improved at radiology facilities that adopted the software, but the detection rate for invasive cancers did not improve. Further, the percentage of women with abnormal mammograms who were accurately diagnosed dropped from 4.3% to 3.6% (CNN, 7/27). False-positive rates increased by 0.5% during the same period, meaning that if 100,000 women were screened, 500 could be unnecessarily called back for further testing (La, AP/Sacramento Bee, 7/28).
Carol Lee, chair of the American College of Radiology's Breast Imaging Commission, said it is not clear whether radiologists used the software correctly or at all. "It makes me think that we as a medical community need to further evaluate the use of CAD," Lee said (CNN, 7/27).
In an editorial accompanying the study, Donald Berry, chair of the department of biostatistics at M.D. Anderson Cancer Center, noted that "an argument for the use of CAD with film or digital mammograms is that it will get better with time." He added, "Fine. ... But this should happen in an experimental setting and not while exposing millions of women to a technology that may be more harmful than it is beneficial" (Wall Street Journal, 7/28).