Ignoring breast CAD findings need not breach standard of care

July 22, 2004

Failure to act on computer-aided detection breast markings in areas judged to be benign or normal by experienced radiologists does not constitute a breach in the standard of care, according to a study in the March issue of Radiology. The researchers

Failure to act on computer-aided detection breast markings in areas judged to be benign or normal by experienced radiologists does not constitute a breach in the standard of care, according to a study in the March issue of Radiology.

The researchers recommend that radiologists follow standard interpretation procedures even when CAD marks a nonspecific finding.

"In the current legal climate, the appearance of a CAD mark at an area that eventually is shown to be malignant has caused much concern among radiologists. The standard of care for recalling a lesion for further investigation should not change based on such a CAD mark," said Dr. R. James Brenner, a lawyer and a clinical professor of radiology at the University of California, Los Angeles.

Brenner and colleagues at Stanford University and R2 Technology collected 493 pairs of mammogram studies. The pairs consisted of an initial negative screening mammogram and a follow-up screening mammogram indicating cancer. The average time between the two mammograms was 14.6 months.

Of the 493 pair, 169 cases (172 areas of cancer) had cancer indications so subtle that either none or only one or two of the five blinded radiologists participating in the study recommended a screening recall. They judged 80% of the 172 areas as having either a benign or normal appearance.

The investigators then examined the subtle findings with a CAD system. It marked 42% of the 172 cancer findings, or:
· 29% of the 82 findings not recalled by any of the radiologists,
· 49% of the 51 findings recalled by one radiologist
· 59% of the 39 findings recalled by two of the five radiologists

"If CAD identifies a significant abnormality that might not have been detected by a radiologist, then its value is straightforward. The degree to which this occurs is thus far unresolved," Brenner said.

A CAD mark does not necessarily reach a level of predictability that would necessitate a recall. But if a truly suspicious lesion is marked and not recalled, the interpreting radiologist would be required to provide a rationale for not recalling, he said.

"This would occur with or without a CAD mark. The same is true for lesions that are marked but are unlikely to represent cancer," he said. "The legal test for the law of negligence is not accuracy, but reasonableness."