Mammography's Predictive Value May Be Improved

May 1, 1997

COLUMBUS, Ohio--Standardized descriptions of mammographic findings and standardized reporting of final assessments continue to play an important role in improving the predictive value of mammography, Lawrence W. Bassett, MD, said at the Ohio State University James Cancer Hospital and Research Institute's Third Oncology Update.

COLUMBUS, Ohio--Standardized descriptions of mammographic findings andstandardized reporting of final assessments continue to play an importantrole in improving the predictive value of mammography, Lawrence W. Bassett,MD, said at the Ohio State University James Cancer Hospital and ResearchInstitute's Third Oncology Update.

Using standardized terms, such as the nomenclature defined by the AmericanCollege of Radiology Breast Image Reporting and Data System (BI-RADS),radiologists can review and compare various studies with more confidence,allowing more effective communication with other radiologists, said Dr.Bassett, who is the Iris Cantor Professor of Breast Imaging at UCLA.

Databases created by collecting standardized descriptions of abnormalities,and the course of the treatment and disease that followed, could be usedto develop algorithms to predict prognosis based on a description of theinitial lesion, Dr. Bassett said.

Accurate computer-aided diagnosis using standardized descriptions mayalso become possible. At Duke University a computer using artificial intelligencereviewed mammographic findings described with standardized terms (Bakeret al: Radiology 198:131-135, 1996). The computer's ability to make a diagnosisand recommendations for biopsy were impressive, Dr. Bassett said, and itsspecificity was better than that of a human.

In the meantime, radiologists can improve their medical practice bymeasuring the predictive value of their imaging techniques and comparingit to emerging national goals, Dr. Bassett said. The positive predictivevalue (PPV) of mam-mography is defined as the number of cancers found atbiopsy divided by the number of cases recommended for biopsy after an abnormalmammogram.

The ideal PPV will vary according to the priorities of those involved.In some parts of Europe, for instance, where cost effectiveness of health-caredelivery is a high priority, the PPV is 50% to 75%.

A random survey conducted by the FDA of 50 US facilities showed a PPVof 21% (Brown et al: AJR 165:1373-1377, 1995). The problem with a low PPVis that a large number of biopsies for benign conditions can discouragereferring physicians from ordering screening mammograms, Dr. Bassett said.

A less aggressive approach to biopsy producing a PPV of 25% to 40% hasbeen described as a desirable goal (Linver et al: AJR 165:19-25, 1995).With this approach, Dr. Bassett said, more than 50% of the tumors discoveredat biopsy should be stage 0 or I with a node positivity rate of less than25%. More than 30% of the cancers should be minimal.

Physicians can use several strategies to increase the PPV of their practice.Among the most valuable is to initiate a complete workup for abnormalitiesfound on screening mammograms. This may include using magnification mammog-raphyor taking a number of special views, he said. Ultrasonography will identify20% to 30% of the abnormalities as definitely benign, he noted.

Another method is the medical audit, which involves collecting and analyzingdata regarding the mammography report and any subsequent outcome data suchas biopsy or clinical follow-up. An audit is a useful way for mammographersto see whether their practice is operating under current guidelines. Medicalaudits are part of the Mammography Quality Standards Act of 1992.