The effect of using different reference dates for control exposure measurement on relative risk estimates in a case-control study.

ABSTRACT

In case-control studies in which case and control enrollment periods are not identical, exposure status for time-dependent variables is often measured relative to a reference date. Using data from a case-control study of the relation between cervical cancer and oral contraceptive (OC) use in which control enrollment began 6 months after the end of case enrollment, we evaluated the effect on odds ratios from using five different reference dates to determine the controls' exposure status. The choice of reference date had little effect on the odds ratios in this study. Reference dates for time-dependent exposure variables should be considered carefully in studies when case and control enrollment periods are not identical.