Abstract
Screening aims to detect breast cancer at an earlier stage than would occur if symptoms developed. The characteristics of breast cancer that are detectable at screening depend on both the physical properties of the screening test and specific anatomical features of breast cancer. As a result, breast cancer detected by screening is a select subset of all breast cancer existing in the population. Therefore, biomedical, clinical and epidemiological research into breast cancer using populations with access to screening can result in major bias. The biases, with examples, are explained. (C) 2013 Elsevier Ltd. All rights reserved.