The Effects of Early Treatment, Lead Time and Length Bias on the Mortality Experienced by Cases Detected by Screening

Publisher Website
Google Scholar
Abstract
Morrison, A (Department of Epidemiology, Harvard School of Public Health, 677 Huntington Avenue, Boston, MA 02115, USA). The effects of early treatment, lead time and length bias on the mortality experienced by cases detected by screening.International Journal of Epidemiology 1982, 11: 261–267. This paper describes a means of analysing the effects that the benefit of early treatment, lead time and length-biased sampling (and other forms of prognostic selection bias) have on the mortality rate of cases detected in screening programmes. Both benefit and lead time reduce the mortality rate of screen-detected cases. A beneficial effect of early treatment will lead to a decrease in the number of deaths, the numerator of the rate. The amount of person-time among screen-detected cases, the denominator, is increased by early diagnosis as a result of screening (lead time) as well as by prolongation of life due to early treatment. The numbers of cases experiencing benefit and lead time can be estimated by comparisons of the numbers of deaths and numbers of diagnosed cases between the entire screened population, from which the series of screen-detected cases is drawn, and an other-wise comparable unscreened population. Benefit and lead time, as reflected in these numbers, can be removed from the mortality rate of cases detected by screening. The effects of benefit or lead time on prognosis then can be assessed by comparing the observed mortality rate of screen-detected cases to the rates with benefit or lead time removed. Prognostic selection bias (a tendency of screen-detected cases to be relatively benign or relatively malignant) can be evaluated by comparing the case-mortality rate, with both benefit and lead time removed, to the mortality rate of cases in an unscreened population. The relationships described are illustrated with data from a breast cancer screening programme.