Heterogeneity in historical cohort studies: a source of bias in assessing lung cancer risk.

Historical cohorts that are heterogeneous with respect to era in which exposure to a hazard began are subject to chronology bias. Those historical cohorts assembled some time after onset of exposure are also subject to selection bias through attrition of the original populations prior to the time of registration. Such cohorts may be called multiserial cross-sectional. The two forms of bias in such cohorts may account for some of the elevated risk of lung cancer reported in occupational studies of this type. Historical cohorts comprising all workers who began exposure in a relatively limited era can be called inception cohorts. Three occupational investigations are reviewed in which both multiserial cross-sectional cohorts and inception cohorts were studied. In all three investigations the inception cohorts showed lower risks of lung cancer than the multiserial cross-sectional cohorts.