Observational Studies That Seek to Emulate a Randomized Trial of Screening to Reduce the Incidence of Cancer: Do They Address the Question to Which We'd Like to Have an Answer?

Abstract

Some forms of cancer screening have the potential to reduce cancer incidence, if the screening modality can identify not only a malignancy but a treatable premalignant condition (such as a colon polyp) as well. Cohort studies of the efficacy of these forms of screening in reducing the incidence of cancer face many challenges, notably the difficulty in distinguishing whether a test performed in a given individual was screening or diagnostic in nature. Downward bias in the estimated efficacy of screening resulting from misclassification of test indication is a particular problem in cohort studies that seek to gauge cancer incidence beginning at the time of screening (and a corresponding point in time among unscreened persons). The downward bias is accentuated in those cohort studies that have sought to mimic the "intention-to-treat" analytical approach used in randomized trials, in which initially unscreened persons are retained in this category even if later they themselves undergo screening.