Augmenting digital screening mammography with computer-aided detection failed to improve diagnostic accuracy in every performance measure and every subgroup of women studied in a series of 625,625 exams performed across the United States during a 7-year period, according to a report published online Sept. 28 in JAMA Internal Medicine.
In fact, the sensitivity of mammography was actually decreased by computer-aided detection (CAD) in a subgroup of radiologists who practiced at some sites that used CAD and others that did not.
“CAD is a technology that does not seem to warrant added compensation beyond coverage of the mammographic examination,” wrote Dr. Constance D. Lehman of the department of radiology, Massachusetts General Hospital and the Avon Comprehensive Breast Evaluation Center, both in Boston. “The results of our comprehensive study lend no support for continued reimbursement for CAD as a method to increase mammography performance or improve patient outcomes.”
Measuring the real-world impact of CAD on mammographic accuracy has been difficult and has yielded inconsistent and contradictory findings; most studies to date have been relatively small, have focused on older women only, or haven’t taken into account the early part of radiologists’ learning curves. To circumvent these problems, the investigators pooled data from five mammographic registries to include more than 625,000 full-field digital mammograms, included a demographically diverse population of women aged 40-89 years, and excluded the first year of CAD use for every radiologist in the study.
Dr. Constance Lehman
They assessed outcomes after routine screening mammography with CAD (495,818) or without CAD (129,807), which were interpreted by 271 radiologists. Breast cancer was diagnosed in 3,159 women within 1 year of these screening mammograms.
The overall sensitivity of mammography was 85.3% with CAD and 87.3% without it; sensitivity for invasive cancer was 82.1% with CAD and 85.0% without it, all of which are nonsignificant differences. Also similar were mammography’s specificity, at 91.6% with CAD and 91.4% without CAD.
The overall cancer detection rate was exactly the same regardless of the use of CAD, at 4.1 cancers per 1,000 women screened. And the invasive cancer detection rate was nearly the same, at 2.9 cancers with CAD versus 3.0 cancers without CAD (JAMA Intern Med. 2015 Sep 28. doi:10.1001/jamainternmed.2015.5231).
Also, diagnostic accuracy was the same with or without CAD regardless of patient age, patient ethnicity, breast density, patient menopausal status, family history, or interval since the last mammogram. In the subgroup of 107 radiologists who sometimes used CAD and sometimes did not, the use of CAD actually decreased the sensitivity of mammography (83.3% with CAD and 89.6% without CAD).
“Given that the evidence of the current application of CAD in community practice does not show an improvement in diagnostic accuracy, we question the policy of continuing to charge for a technology that provides no established benefits to women,” Dr. Lehman and her associates wrote.
But the researchers noted that CAD may offer advantages beyond diagnostic accuracy, such as improved work flow or shorter times spent assessing faint calcifications. CAD also may be useful in guiding treatment decisions, perhaps reducing unnecessary biopsies of lesions that have specific benign features or ensuring biopsy of lesions that have specific malignant features, they added.
The study was supported by the National Cancer Institute, the Breast Cancer Surveillance Consortium, several state public health departments, and cancer registries throughout the United States. Dr. Lehman reported receiving grant support from General Electric Healthcare and serving as a member of the company’s Comparative Effectiveness Research Advisory Board.