WASHINGTON--A highly publicized journal article about the rate of false positives in mammography distorts both the state of the field and the goals and methods of breast cancer screening, said Edward A. Sickles, MD, chief of the Breast Imaging Section, University of California at San Francisco (UCSF).
Presenting the biennial Wendell Scott, MD, Memorial Lecture to a plenary session of the American College of Radiology’s 28th National Conference on Breast Cancer, Dr. Sickles cited the paper’s suggestion that new technologies are needed to reduce false positives as an example of what happens "when investigators write in areas outside their clinical expertise. They write nonsense."
Although technological progress is always welcome, he said, the necessary methods for reducing false positives already exist and "are not rocket science technology." Rather, they are strategies for carefully using existing techniques.
But, lest the field pay too much attention to cutting the false-positive rate, he cautioned that any method that reduces false positives inevitably also results in some reduction in true-positive cases. "It is very important that we understand that," he said. "Do not react by immediately changing your thresholds of sensitivity. Cancer detection, not false-positive rate, is the measure of a screening program. If we wanted no false positives, we wouldn’t screen."
Missing from the paper by Elmore et al (N Engl J Med 338:1089-1096, 1998), he said, is an understanding that in mammography the definition of false positive varies according to the situation.
With screening mammogram, a rating of 4 or 5 on the BI-RADS scale indicates a possibly "positive" finding (ie, mammograms that need more evaluation because they are suggestive or highly suggestive of malignancy). This finding leads to a call-back for more views that generally show no cancer. Thus, the original "false-positive" finding resulted only in the need for more noninvasive studies. At biopsy, on the other hand, a false-positive finding would be a tissue study that did not indicate pathology despite positive mammography findings. In this case, the false-positive meant that the patient underwent an invasive procedure unnecessarily.
Dr. Sickles believes that radiologists can use several existing methods to bring their false-positive level down as far as they safely can. No mammography practice, however, should attempt to lower its false-positive rate, he emphasized, "unless and until you can demonstrate by rate of detection that the techniques used produce acceptable sensitivity." Then and only then should attention go to reducing false positives, he insisted.
An effective way of reducing recalls in screening is to compare a patient’s current mammograms with previous films. Doing so cuts the recall rate in half, he said. Women should be educated to bring their old films with them when using a new mammography facility, he said.
Summation Artifacts
Another cause of recalls is summation artifacts, which are harmless objects photographically superimposed to resemble cancerous lesions. Because they result from the perspective from which a particular view is taken, these supposed lesions disappear when the breast is viewed from another angle. Three percent of all exams produce such single-view artifacts, which, in a moderately busy practice, can occur as often as once a day. Taking two views at each examination can therefore prevent needless call-backs, Dr. Sickles recommended.
Radiologists can also reduce recalls by electing to watch over time, instead of immediately investigating, selected kinds of "doubtful findings," Dr. Sickles said. Between 30% and 70% of cancers found at screening are visible on previous films in "subtle forms below the intervention threshold." If watched over time, however, the great majority of such findings will prove to be benign, he said.
A soon-to-be published study of 550 doubtful findings on mammography, for example, found that over 2½ years of surveillance, 73% of the lesions did not change, 20% went away, 3% became slightly more prominent for technical reasons, and 1% were read as clearly abnormal on a second exam.
Of these seven "abnormal" cases, three were shown to be benign with additional workup, three were benign at biopsy, and one was diagnosed as ductal carcinoma in situ (DCIS).
The strategy of watchful waiting of findings just below the intervention threshold therefore produced results well within the thresholds of avoiding malpractice, he noted.
In diagnosis, Dr. Sickles said, some summation artifacts will require more than two views, but in this study, none of the findings that looked suspicious on only a single view were found to be breast cancer.
Certain Findings Usually Benign
Certain kinds of findings, Dr. Sickles said, are "almost certainly" benign, including circumscribed, noncalcified solid masses. Ultrasound can ascertain for sure whether they are solid. Findings that appear generalized over large areas of both breasts are also likely to be benign. "The more of the same thing you have scattered throughout both breasts, the less likely it is to be cancer," he said.
By using follow-up, rather than biopsy, for patients with doubtful findings, UCSF physicians achieve a biopsy yield of 38% rather than the national average of 24%, Dr. Sickles said. He emphasized that every practice should make vigorous efforts to calculate its own biopsy yield (the ratio of positive biopsies to patients referred for biopsies), to determine whether sensitivity is within an acceptable range.
By choosing follow-up instead of immediate tissue study in cases that are probably benign, practices can reduce their rates of false-positive mammograms in both screening and diagnosis without changing their rate of finding early cancers, he said. This task requires care and precision, but no new technology, he added. "Everyone can do this," Dr. Sickles concluded.
