1. This large retrospective cohort study comparing digital screening mammography with and without computer-aided detection (CAD) did not show differences in diagnostic accuracy.
2. For radiologists who interpreted digital mammograms both with and without CAD, sensitivity for breast cancer detection was lower with CAD use.
Evidence Rating Level: 2 (Good)
Study Rundown: Since its approval by the US Food and Drug Administration (FDA) and its reimbursement by the Center for Medicare and Medicaid Services, computer-aided detection (CAD) for digital mammography has been rapidly adopted in the United States. As of 2008, 74% of screening mammography utilized CAD. Despite this, the benefit of CAD integration to patient care has been controversial. The authors compared diagnostic accuracy of digital mammograms with and without CAD based on ability to detect breast cancer diagnosed in the subsequent year. The study found the inclusion of computer-aided detection for mammograms was associated with no difference in sensitivity, specificity, or overall cancer detection rate. In subset analysis of intra-radiologist performance of those who interpreted mammograms both with and without CAD, there was significantly lower sensitivity associated with the use of CAD.
The results from this study did not support the widespread use of CAD for patent-centered benefits to care. The strengths of this study include its large sample size in terms of mammograms, patients, and radiologists studied. Moreover, there were sufficient numbers of radiologists who both did and did not use CAD to compare intra-radiologist performance. Limitations of the study include the unbalanced patient characteristics, such as race, that may signal the presence of confounders between the comparisons. Additionally, the specific type of CAD software that was used was not assessed, nor was whether the type of software has changed since the study period.
Click to read the study, published today in JAMA Internal Medicine
Relevant Reading: Short-Term Outcomes of Screening Mammography Using Computer-Aided Detection: A Population-Based Study of Medicare Enrollees
In-Depth [retrospective cohort]: This study used screening mammography data from 5 registries in the United States that participate in the Breast Cancer Surveillance Consortium. Digital mammography examinations from January 1, 2003 to December 31, 2009 for 323 973 women aged 40 to 89 with demographic data were used. In total, 495 818 digital mammograms with and 129 807 without CAD were obtained, representing readings by a total of 271 radiologists. There were 107 radiologists who read mammograms both with and without CAD. Positive readings were defined as Breast Imaging, Reporting & Data System (BI-RADS) scores of 0, 3 with recommended immediate follow up, 4, or 5. Negative readings were defined as BI-RADS scores of 1, 2, or 3 with no recommended immediate follow up. Patients were followed to the time of their next mammogram or 12 months, whichever came first for breast cancer diagnosis. Digital mammogram examinations were balanced with respect to patient age, menopausal status, family history, breast density, and time from last mammogram; however they were not balanced with respect to patient race.
Overall, 3159 women had a breast cancer diagnosis after 12 month follow up. Cancer detection rates were 4.1 per 1000 women screened with CAD, vs 4.1 per 1000 women without CAD. Mammography screening with and without CAD demonstrated similar sensitivity (85.3% vs. 87.3%, p = 0.18), and specificity (91.6% vs. 91.4%, p = 0.58). For 107 radiologists who interpreted mammograms both with and without CAD, the techniques demonstrated similar specificity, but lower sensitivity for diagnosis with CAD (OR 0.53; 95%CI 0.29-0.97).
Image: PD
©2015 2 Minute Medicine, Inc. All rights reserved. No works may be reproduced without expressed written consent from 2 Minute Medicine, Inc. Inquire about licensing here. No article should be construed as medical advice and is not intended as such by the authors or by 2 Minute Medicine, Inc.
