In most research settings, the absence of information regarding the presence of symptoms and signs of a cancer in BOTH screened and unscreened persons will limit the ability of cohort studies to provide a valid estimate of the efficacy of screening to prevent mortality from that cancer. Bias from this source generally will be small or absent in cohort studies that seek to measure the impact of those forms of cancer screening that have the potential to decrease cancer incidence, but the analysis of such studies must incorporate assumptions regarding the natural history of the cancer in question.