CAD is used by an increasing number of radiologists when interpreting mammograms, and as such, is of interest primarily to the radiology community. As much as 70 per cent of American women undergo annual mammography, and a recent article directly affects them.

by Leonard Berlin, MD*

Computer-aided detection (CAD) is used by an increasing number of radiologists when interpreting mammograms, and as such, is of interest primarily to the radiology community. But as much as 70 per cent of American women undergo annual mammography, and a recent New England Journal of Medicine article directly affects them as well.

In 'Influence of computer-aided detection on performance of screening mammography', a number of nationally-known and respected radiology researchers reported their results of a study of nearly a quarter-of-a-million mammograms, some interpreted with the aid of CAD, some without the aid of CAD. Their findings are astonishing, if not shocking: they concluded that CAD is associated with reduced accuracy of mammograms readings, along with increased rate of biopsy, without any increase in detection of invasive breast cancer!

CAD has been engrained on the radiologic community for nearly a decade. Early reports proclaimed that it increased accuracy -- or more specifically detection of breast cancer -- by at least 20 per cent. Almost all radiologists believed it. Congress was lobbied (primarily by General Electric Corp., described in a Forbes Magazine article on March 19, 2001) and approved additional reimbursement for CAD, thus accelerating sales of these units. Radiologist-researchers wrote numbers of articles praising the value of CAD. The most recent evaluation, by Morton at Mayo Clinic, was somewhat less optimistic and probably more realistic (Radiology 2006; 239:375). He reported that CAD was not quite as good as some previous reports, but still increased detection of cancer by 7.6 per cent.

Whether CAD was really helpful in clinical practice, however, was never proven, and in fact was questioned from time to time, but gingerly, by various skeptics. During some of the lectures on Missed Breast Cancer, Mammography, and Malpractice that I've presented in the past few years, I would ask the audience how many had CAD, and about 50-65 per cent said they did. Then I asked how many thought CAD was helpful, and about half indicated it was.

But nobody suggested that CAD could be harmful. Until now.

Also of note: CAD increased the finding of Ductal carcinoma-in-situ (DCIS) by 30 per cent. Unlike the MRI article that appeared in last week's NEJM that implied that most DCIS goes on to invasive cancer, this article is somewhat more forthcoming: "The natural history of DCIS is certainly more indolent than that of invasive cancer... The effect of CAD on mortality from breast cancer may be limited if it chiefly promotes the identification of DCIS rather than invasive cancer."

This is a whopper of a report, and will be sure to shake up the radiology community, and in turn the women of America who have been 'educated' by periodic and ongoing news reports, as well as paid advertisements by certain radiology facilities and CAD manufacturers, that CAD is a valuable -- if not a sine-qua-non -- technology to diagnose early breast cancer by mammography. R2, the original CAD manufacturer, has been acquired by Hologic. However, there are presently several other CAD manufacturers out there.

What is the impact of this article? Certainly it is not the end of CAD, and we must keep in mind that the article based its conclusions on older versions of CAD. Improvements have been made in the interval, and will continue to be made. But the real question to be answered is, "Does CAD have a future?"

My own personal opinion is that its future is questionable. We have been striving for more than 60 years to decrease errors and improve accuracy in radiological diagnosis. Similar efforts have been expended to improve mammographic diagnosis. Thus far we have not been successful in achieving any significant reduction in error-rates, however. The goal of radiology-researchers has been (and still remains) that a substantial reduction if not total elimination of radiological errors and 'misses' can be achieved by utilizing computers. That goal has thus far been extraordinarily elusive; the NEJM article suggests that that goal may in fact be unreachable.

No one is going to abandon CAD. Once the equipment has been purchased, and reimbursement is being received for its use, it would be foolhardy to stop using it. However, I imagine that it will slow down sales of new CAD equipment -- at least for the time being.

I believe Dr. Joshua Fenton's (of the University of California, Davis, leader of the study) estimate of 25-30 per cent of the facilities using CAD is accurate. I do not believe the overall quality of mammographjy in the US has been changed, or will be changed, by CAD.

CAD has been shown to be more useful for the less-experienced mammography radiologists, which probably accounts for the increase in false-positives. I suppose that can be considered a defense of CAD. The more-experienced mammography radiologists rely less on CAD. Because the mimimum required number of annual mammogram readings for American radiologists to be certified in mammography is far lower than it is for for radiologists in the UK, I would expect that in general UK radiologists are more secure in their mammographic interpretations, and probably are not particularly attracted to CAD.

Medicare pays an extra $20 for mammograms that are read by computer. Can this be an obstacle to criticizing CAD? That's a point well taken. If a radiologist gets paid for using CAD, and wants to continue getting paid by using CAD, how can he/she criticize it? That would be the same as criticizing oneself, which no one wants to do.

As for financial incentives for doing more biosies, I would hope that it does not exist. Radiologists do get paid for performing needle biopsies, true. But I would find it hard to believe that a radiologist would 'overcall' a reading in order to generate a biopsy. However, I do believe American radiologists will overcall a reading for medical-legal purposes, i.e., because of defensive medicine. American radiologists do not get sued for ordering an unnecessary breast biopsy; they get sued for not ordering one, if a cancer later is diagnosed.



* Dr. Leonard Berlin is chairman of the Department of Radiology at Rush North Shore Medical Center, Skokie, IL., USA