Healthy screened bias in epidemiologic studies of cancer incidence.

Abstract

The term screenee has at least several possible meanings. It was coined by Morrison1 to de? scribe a phenomenon in nonrandomized studies seeking to measure the efficacy of screening for a particular cancer, by comparing the cancer mortality experience of screened persons with that of unscreened persons. In these studies, a bias is likely to result because persons undergoing screening, by definition, have neither a his? tory of the cancer nor symptoms of it. Unless persons in the unscreened, comparison group have their eligibility for screening assessed, there will be some who already have been diagnosed with that cancer, and still others who have not yet been diagnosed but who have symp? toms of the cancer. As a result, the unscreened group would be expected to have a relatively higher mortality from the cancer than would screened persons, at least in the immediate weeks and months after the time of screening, even in the absence of a true mortality reduc? tion resulting from screening. This term also describes the situation in which persons who choose to be screened may for other reasons be at low risk of devel? oping (and therefore dying of) a particular cancer. Finally, healthy screenee bias can be present in studies of the effects of a therapeutic intervention on cancer occurrence, when persons who are being considered for a particular intervention are first evaluated for the pres? ence of one or more forms of cancer (or a condition that gives rise to the cancer). If receiving the intervention requires the absence of a given cancer or its antecedent, then for a certain period of time the reported incidence of that cancer in treated patients will be lower than the incidence in a comparison group of untreated persons who have not been similarly screened, even if there is no biological effect of the intervention on the occurrence of the cancer. This form of healthy screenee bias also can be present in studies of the relation of a particular characteristic or pathologic condition to the subsequent

Ähnliche Arbeiten