Time-related factors as potential confounders and effect modifiers in studies based on an occupational cohort

Time-related factors which are potential confounders and effect modifiers in studies based on an occupational cohort are reviewed. The most frequently considered ones include age at first exposure, duration of exposure, interval from exposure to disease recognition, and age at risk. These factors are related to the "healthy worker effect," which appears to be more pronounced among workers with the longest durations of employment and older ages, at date of hire, but weaker with longer length of follow-up and older age at risk. Hence, use of an internal comparison group may not eliminate bias since confounding will occur if the exposed and unexposed groups differ in their distributions across these factors. It is also shown, using the multistage model of carcinogenesis, that these factors may be important effect modifiers. Fortunately, generally straightforward methods of control exist both for stratified analyses and for the commonly used mathematical modeling approaches. Although no firm recommendations can be made, it would appear to be important to control for length of follow-up in the design or analysis of most studies based on an occupational cohort, and controlling for age at first exposure may also be desirable under many circumstances.