Lung cancer due to chloromethyl ethers: bias in cohort definition.

The mortality risks of respiratory cancer and of all other causes of death combined were studied in two types of cohort: (1) a cross-sectional cohort composed of all employees working in a chemical plant building in January 1963 when the occupational cause (chloromethyl ethers) of a suspected lung cancer epidemic was unknown, and (2) an inception cohort composed of all workers ever exposed in the building from the beginning of exposure, registered after the cause was recognized. Selection bias led to an overestimate of the risk in the cross-sectional cohort. The bias was due to overrepresentation of workers with moderate and high cumulative exposure to chloromethyl ethers in the cross-sectional cohort. The results indicate that a more valid assessment of an epidemic is obtained from a study of an inception cohort than from a study of a cross-sectional cohort.