Estimating the accuracy of screening mammography: a meta-analysis.

OBJECTIVE To estimate the accuracy of mammographic screening. DESIGN A meta-analysis of published literature. DATA SOURCES Published English-language randomized controlled trials, case-control studies, and demonstration projects involving screening mammography were identified using recent review articles. We found additional references using MEDLINE searches combining the MeSH terms "mammography," "screening," and/or study authors and locations. STUDY SELECTION We included all studies that provided information to calculate the true-positive rate (TPR) and the false-positive rate (FPR) for breast cancer screening. DATA EXTRACTION Reported data were reviewed independently by the authors; calculations were compared and discrepancies resolved. We calculated the sensitivity as the number of breast cancers detected during the first round of screening (true positives) divided by the sum of the true positives and the false negatives (defined as cancer discovered within 1 year of screening). False-positives were determined by biopsy. DATA SYNTHESIS TPR and FPR values from each study were plotted in receiver operating characteristic (ROC) space. Tests of homogeneity were performed to assess the validity of using summary ROC curves or a single point to summarize the data. The reported TPR and FPR of mammography ranged from 83% to 95% and 0.9% to 6.5%, respectively. The sensitivity of mammography is higher in women over the age of 50 years. CONCLUSIONS The accuracy of mammography should be recognized and included in discussions about policies for screening for breast cancer. This meta-analysis, by quantifying the expected TPR/FPR, should assist program planners, physicians, and women to understand better the cost and clinical implications of such screening programs.