Breast Cancer Surveillance Consortium: a national mammography screening and outcomes database.

M ammography is the primary method of detecting early stage breast cancer and has been shown in randomized clinical trials to reduce breast cancer mortality, especially among women 50 years old and older [1-5]. Authorities in cancer screening have bong recognized that the level of efficacy of screening demonstrated in randomized clinical trials may not pertain to community practice for several reasons [6]. These reasons include possible differences in the population groups receiving screening, lower accuracy of screening mammography in the community, and lower compliance with diagnostic follow-up and treatment in community practice, which may result in more adverse outcomes. Screening effectiveness in community practice today could exceed that estimated in trials because the technical and interpretative quality of mammography has improved since the trials were performed. Furthermore, clinical trial efficacy has been estimated on the basis of assignment to receive screening; to the extent that women assigned to screening were not screened or that women in the control groups were screened, efficacy in trials may have been underestimated. To optimally evaluate the performance of mammography in a community setting, the screening prevalence and patterns and the associated sensitivity, specificity, and predictive value of mammography in community screening programs should be determined by linkage with cancer outcomes [7, 8]. A program of monitoring should also provide data on specific populations, such as rural and minority subgroups, that are traditionally underserved by screening programs and that may have different breast cancer mortality rates [9]. Before the Mammography Quality Standards Act (MQSA) of 1992, most mammography facilities in the United States did not maintain records that could provide reliable and comprehensive data to evaluate the performance of screening mammography [10]. The concept of a medical audit of outcomes data had been proposed [ 1 1 ] but has not been routinely practiced in the community. The interim regulations of the MQSA mandated maintaining mammography data and performing a medical outcomes audit [12]. In practical terms, the medical audit requirement of the MQSA was limited to an analysis of patients with tests interpreted as “suspicious abnormality” or “highly suggestive for malignancy,” which permits evaluation of the positive predictive value of such interpretations. However, the MQSA does not require linkage to populationbased cancer registry data or another source of pathology data, without which it is impossible to accurately assess the outcomes of patients with mammograms interpreted as having normal findings. To understand the full effect of