U.K. study strengthens case for screening mammography with CAD

October 3, 2008

Findings from a large study by British researchers suggest mammography screening interpretations by a single radiologist using computer-aided detection are just as good as double reads and could yield better cancer detection rates than single reads without CAD. Despite higher recall rates, single reads with CAD could offer staffing costs trade-offs.

Findings from a large study by British researchers suggest mammography screening interpretations by a single radiologist using computer-aided detection are just as good as double reads and could yield better cancer detection rates than single reads without CAD. Despite higher recall rates, single reads with CAD could offer staffing costs trade-offs. The swing of opinion on mammography CAD's utility in one year has not been quite as wild as that on the U.S. Congress' Wall Street bailout plan in a week. But it has not settled down either. Evidence giving CAD a nod for its accuracy and its potentially positive impact on clinical practice, however, continues to pile up.Double reading of screening mammography -- which has proven to increase cancer detection rates up to 14% -- is gold standard procedure in about a fifth of all European states. Not so in the U.S., where having a single dedicated mammographer is nearly a luxury in some areas dealing with a short supply of radiologists. The clinical literature shows that CAD systems have the potential to increase cancer detection rates and that single readers using CAD might be able to match the performance of two readers. CAD could bolster single reading results while posing as an alternative to double+reads, according to results from the Computer-Aided Detection Evaluation Trial II (CADET II) led by Dr. Fiona J. Gilbert, a professor of radiology at the University of Aberdeen in Scotland.

Gilbert and colleagues prospectively enrolled 31,057 women who underwent screening mammography at three centers in England. The researchers randomly assigned batches of film from selected subjects into three groups:

  • A: double reading (1000 subjects)
  • B: single reading with CAD (1000 subjects)
  • C: double reading plus single reading with CAD (28,000 subjects)

The investigators measured cancer detection and recall rates as the primary outcomes. They found that a single expert using CAD can detect as many cancers as can two experts. Findings appeared online Oct. 1 and will be published in the Oct. 16 issue of The New England Journal of Medicine.

The cancer detection rates for double reading and single reading with CAD were, respectively, 87.7% and 87.2%. The overall recall rates were 3.4% for double reading and 3.9% for single reading with CAD (

p

Diagnostic Imaging.According to Dr. Laurie Fajardo, radiology chair at the University of Iowa Carver College of Medicine, the two biggest issues in U.S. healthcare are personnel cost and staffing shortages. Study results, acceptable by U.K. standards, show CAD has the potential to address both by reducing the cost of having a mammogram."What they do not address is quality standards for machinery, image quality, and personnel expertise, all factors which receive higher attention in the U.S.," Fajardo said.The study validates previously published data. Although the recall rates in the trial were higher for CAD, the trade-off remains the associated cost from a double read, according to Dr. David D. Dershaw, director of breast imaging at the Memorial Sloan-Kettering Cancer Center in New York City. "It's important for women to understand that a bad reading along with computer-aided detection is not as good as a good reading with no second readers," Dershaw said. "However, it's also important for those of us who are experts to understand that no matter how good we are, a second set of eyes will make us better."For more information from the Diagnostic Imaging archives:

Large screening mammography study restores faith in CAD

Radiologists pan study critical of computer-aided detection

Why the NEJM study on CAD is wrong

Study questions accuracy of breast CAD