When workers with poor health decrease their exposure but healthy workers do not, it becomes difficult to detect an association even when exposure causes disease. This healthy worker survivor effect is a well known and ubiquitous bias that affects occupational studies of a wide range of health outcomes and exposures. It is particularly problematic in studies of long-term exposures and chronic diseases. There are no satisfactory solutions to the healthy worker survivor effect using conventional methods for analyzing longitudinal occupational data. Fortunately, alternative methods have been developed that do eliminate this type of bias. G-estimation of structural nested accelerated failure time models is one such technique. It unlinks observed exposure from pre-hire prognosis within strata of prior exposure and covariates, thereby ensuring that the healthy worker survivor effect does not bias results. The parameter obtained represents the natural log of the factor by which one year of exposure decreases survival time. Over the past 15 years, several authors have encouraged the adoption of this method, but without success. Very recently, in the first application of g-estimation analysis in an occupational study, longer duration of exposure to oil-based metalworking fluids was clearly related to decreased time to heart disease mortality. This evidence from the UAW-GM autoworkers cohort is the first in the literature for this effect. The importance of this new method is underscored by the failure of conventional analyses to detect this association. Methods for controlling the healthy worker survivor effect and similar biases (known more generally as "causal methods") have up to now focused on binary annual exposure measures. If workers are exposed at various levels, higher exposures may be more likely to cause disease, or may cause disease sooner. Unlike binary exposure measures that yield estimates for the effect of exposure duration, quantitative measures can distinguish etiologically relevant levels of exposure. Quantitative exposure-response characterization is also necessary for risk assessment providing guidance for policy. This project therefore aims to extend causal methods in order to evaluate, without healthy worker survivor bias, the effect of total quantitative exposure on survival time. These new methods will then be applied to the UAW-GM autoworkers cohort to explore the causal effect of quantitative exposure to oil-based metalworking fluids on cardiovascular outcomes. Successful application of g-estimation to total exposures in this case would be a major breakthrough that could pave the way for other occupational studies to handle the healthy worker survivor effect correctly while considering quantitative exposure rather than duration of exposure.
When workers in poorer health are more likely to decrease their exposure by transferring to unexposed jobs or leaving work entirely, it becomes difficult to detect an association even if exposure causes disease. This project aims to improve control of this healthy worker survivor effect by extending causal methods (hitherto applied only to binary exposures) to cumulative exposures, in order to investigate the health impact of interventions to limit occupational exposures. The method will be applied in the UAW-GM autoworkers cohort to study the effects of cumulative exposure to oil-based metalworking fluids on cardiovascular outcomes.
|Picciotto, Sally; Ljungman, Petter L; Eisen, Ellen A (2016) Straight Metalworking Fluids and All-Cause and Cardiovascular Mortality Analyzed by Using G-Estimation of an Accelerated Failure Time Model With Quantitative Exposure: Methods and Interpretations. Am J Epidemiol 183:680-8|
|Picciotto, Sally; Peters, Annette; Eisen, Ellen A (2015) Hypothetical exposure limits for oil-based metalworking fluids and cardiovascular mortality in a cohort of autoworkers: structural accelerated failure time models in a public health framework. Am J Epidemiol 181:563-70|
|Picciotto, Sally; Brown, Daniel M; Chevrier, Jonathan et al. (2013) Healthy worker survivor bias: implications of truncating follow-up at employment termination. Occup Environ Med 70:736-42|