A data-analytic approach for estimating lead time and screening benefit based on survival curves in randomized cancer screening trials.

Screening tests are used frequently for control of diseases such as cancer. The increased survival time of screen-detected cases over those that are detected clinically may be due in part to 'lead time', or the length of time by which the disease is diagnosed earlier by screening in the presence or absence of any real extension in survival time. A realistic evaluation of screening needs to assess the true benefit of screening; that is, the length of time by which survival has been extended, beyond merely the time of the advanced diagnosis. The comparison of survival measured from time of entry between cases in a screening arm and in a control arm in randomized studies avoids the lead time bias. If the effects of average lead time and average benefit on survival are additive, these effects can be estimated by recognizing that (a) the difference in survival curves since time of diagnosis confounds benefit and lead time, but (b) the difference in survival curves since time of start of study involves benefit only. The method is evaluated on simulated data for its accuracy and may be used on data from randomized studies.