Article Text
Abstract
Objective In a cohort study of respiratory cancer incidence and metalworking fluids (MWF) exposure, we assessed the potential for bias due to the Healthy Worker Survivor Effect (HWSE). HWSE exists if underlying health status is a time-varying confounder affected by prior exposure. Using employment status as a proxy for health status, HWSE depends on the presence of three underlying associations: (1) prior exposure and employment status, (2) employment status and future exposure, and (3) employment status and risk of the outcome. Since occupational exposure is necessarily zero after employment termination, (2) is satisfied. We assessed the presence of (1) and (3) in a cohort of Michigan autoworkers.
Methods The study population included 27,841 male hourly workers from three automobile manufacturing plants with potential exposure to MWF. Workers were hired between January 1, 1938 and December 31, 1982 and employed for a minimum of 3 years. Cancer incidence follow-up began in 1985. Cox proportional hazards models for employment termination and for the outcomes, lung and laryngeal cancer incidence, were used to assess associations (1) and (3), respectively. All models were adjusted for age (as the time metric), race, plant, calendar year, year of hire, and duration of employment.
Results Cumulative exposures to straight, soluble, and synthetic MWF were each associated with employment termination, adjusting for the other two exposures. Employment termination status was associated with higher incidence of lung cancer: HR = 1.46 (95% CI: 1.03, 2.08). Results for laryngeal cancer incidence were less conclusive: HR = 1.05 (95% CI: 0.48, 2.32).
Conclusions For studying lung cancer, documentation of employment status as a time-varying confounder affected by prior exposure indicates that we must use alternative methods to avoid bias by HWSE. By contrast, leaving work was not clearly associated with risk of laryngeal cancer. In the absence of evidence of HWSE, standard methods may be unbiased.