Elmore et al. describe the variability of mammographic interpretation between radiologists and the variability in specificity by age 35. If one examines the variability that they describe, one sees from the receiver operating curves presented that it is along the curve rather than above or below it. This suggests that improvement in interpretive skills to better define features and criteria for biopsy will have a significant beneficial impact on this variability.
Inasmuch as breast cancer is among women’s greatest health concerns, test results suggesting that a woman may have breast cancer are likely to cause anxiety, and the anxiety may be greater and more lasting the further along the diagnostic chain her evaluation proceeds. However, in terms of the consequences and resulting anxiety, from a measurement standpoint there is no justification for treating all false-positive results (i.e., FP1–3 or PPV1–3), as equivalent. Elmore reviewed mammographic recall and biopsy data over a 10-year period and observed that 23.8% of women had at least one false-positive mammogram and 31.7% had either a false-positive mammogram or CBE.
Although relatively high over time (2,227 women had 9,762 examinations over the study period), the rate of false-positive findings by examination was much lower. In other words, the likelihood of a false-positive result is considerably lower per screening event than over the duration of screening events. This is to be expected, and Elmore’s study highlights not only that a higher rate of false-positives can be expected with sporadic, opportunistic screening but also the need to identify best practices in order to reduce the rate of avoidable false-positives. Because women are worried about breast cancer, providers must handle any need for extra procedures with great sensitivity. Although avoiding mammography because of the risk of false-positive results has been proposed, a more valuable strategy would assure that women who undergo screening know what to expect. New programs are needed to improve the overall quality of screening and follow-up services, including greater attention to the psychosocial needs of women who have abnormal findings.
An encouraging report and actually the first glimpse of mammography as performed on a national level was presented by May et al. As noted in the discussion of specificity, the report summarizes findings for 230,143 women who underwent mammography during the first 4 years of the National Breast and Cervical Cancer Early Detection Program. When biopsy was recommended from mammography, the PPV was 26%. The suggested goal from the AHCPR panel is 25% to 40%. The rate of prevalent cancers found was 5.1 in 1,000, and the rate of incident cancers was 2.0 per 1,000, a little lower than the AHCPR suggested goals for these parameters of 6 to 10 per 1,000 and 2 to 4 per 1,000, respectively, but overall the CDC screening population was skewed in early years toward a younger population of women. The recall rate was 5%, and the suggested goal is no greater than 10%.
Finally, minimal cancers (invasive tumors less than 1.0 cm and DCIS) comprised 45% of the cancers, with a suggested goal of more than 30%. Thus, what we see from this report is a first view of screening results in the United States that is not limited to academic centers but is probably representative of a broad cross section of our screening efforts. The fact that overall performance by facilities participating in the CDC program falls within the target ranges that were established by the AHCPR panel is encouraging.
Robert A. Smith and Carl J. D’OrsiR. A. Smith: Cancer Screening, Department of Cancer Control, American Cancer Society, Atlanta, Georgia
C. J. D’Orsi: Diagnostic Radiology, University of Massachusetts Memorial Medical Center, Worchester, MassachusettsReferences