S4-1: Overdiagnosis in Breast Cancer Screening: Methodological Considerations of Current Estimates.

Background A harm that has received considerable attention in breast cancer screening is overdiagnosis, i.e., the detection through screening of breast cancer that is non-progressive. Since there is no way to distinguish a truly non-progressive tumor from one that is progressive, overdiagnosis is a statistical phenomenon that is estimated as the difference between observed and expected incidence rates. However, there is no agreement on the true rate of overdiagnosis in breast cancer screening. Estimates vary widely, with the highest estimates reported in the scientific literature having received significant media attention under the premise that a signifiant fraction of newly diagnosed breast cancers are non-progressive and thus overtreated. We examined the range of overdiagnosis estimates in the context of factors that influence observed and expected breast cancer incidence rates over time. Methods Only studies that compared incidence rates of invasive breast cancer, or invasive breast cancer and ductal carcinoma in situ (DCIS), in a group invited or exposed to mammography compared with a group not invited or not exposed to mammography were included in our analysis. Studies were evaluated based on whether or not they adjusted for contemporaneous trends in the underlying incidence of disease, which may differ in different age groups, and age-specific effect on incidence rates due to lead time. Results We identified 17 studies (Table 1) that attempted to estimate overdiagnosis of breast cancer resulting from mammography screening. The main determinant of high estimated rates of overdiagnosis was a failure to adjust for lead time. The highest rate of overdiagnosis (33%) occurred in the one study that did not correct for either lead time or confounders. Studies that only partially corrected for lead time and confounding also had high estimates of overdiagnosis (mean = 30%), whereas studies that corrected for both lead time and confounders had the lowest estimate of overdiagnosis (mean =5%). Conclusion The possibility that some breast cancers detected by mammography are indolent and not life threatening within the patient9s natural life is real. However, it is clear from our analysis that the wide range of estimates of overdiagnosis are due to whether or not studies have adjusted for factors in and outside of screening programs known to influence incidence rates over time. Our findings suggest that the true rate of overdiagnosis associated with breast cancer screening, to the extent that it exits at all, is small. Citation Information: Cancer Res 2011;71(24 Suppl):Abstract nr S4-1.