Abstract

A commonly used estimate of a screening test's sensitivity, based on the prevalence at screening and the incidence of interval cancers, is shown to be logically unsound and quantitatively poor. An alternative simple estimate is proposed, based only on incidence rates.