Breast cancer screening: the effect of self selection for screening on comparisons of randomised controlled trials

In randomised controlled trials of breast cancer screening women are randomly allocated to an intervention group that is offered screening, or a control group that receives the usual medical care. Not all women in the intervention group accept screening; the women who do so may differ from other women in their underlying risk of breast cancer. This self selection for screening can result in either women at high risk or at low risk being overrepresented in the screened group. When comparisons between trials are made it is important to take self selection for screening into account, as a trial with self selection of women at high risk has the potential to have greater efficacy than a trial with self selection of women at low risk. A method of adjusting for self selection when comparing randomised controlled trials of breast cancer screening is described.