Adjusting for screening history in epidemiologic studies of cancer: why, when, and how to do it.

In epidemiologic studies of cancer, differences between exposed and nonexposed persons with regard to a history of cancer screening during the time the malignancy (or an antecedent lesion) typically is present prior to diagnosis can be a source of confounding if one of the following conditions is present: (1) the screening modality identifies premalignant changes whose treatment has the potential to prevent the cancer from developing; or (2) the number of cases included in the study would have been smaller but for the presence of screening. These situations occur commonly, arguing that consideration be given to screening history in the design (with attention to distinguishing true screening tests from those administered to persons with signs or symptoms of cancer) and analysis of epidemiologic studies of those cancers for which screening modalities are in use in the study population.