Individuals who are hired and remain at work longer are generally healthier than those who are unemployed or leave work, causing downward bias in studies of the health effects of occupational exposures. We present results from a study of ischemic heart disease mortality in a cohort of autoworkers exposed to metalworking fluid (MWF) using two approaches designed to reduce the bias. MWF is a mixed particulate containing metals and polyaromatic hydrocarbons and although previous studies of this cohort have found associations with selected cancers, this is the first investigation of heart disease. The cohort includes all workers hired between 1938 and 1981 who worked more than 3 years in one of three automobile manufacturing plants in Michigan (N = 39,927). The cohort was followed-up for vital status from 1941 to1995, and cause of death was ascertained from state health records and the National Death Index. Date of birth, race, gender and work history, including time off work, were obtained from company records. Annual exposure to oil-based MWF was treated as binary and health status was defined as the time off work in every year of follow up. To adjust for time off work as a time-varying confounder, we used two causal modeling approaches: Marginal Structural Models with Inverse Probability of TreatmentWeights (IPTW) and Structural NestedModels using g-estimation. We compared these results based on 2,725 heart disease deaths, with those from a standard Cox model. Finally, we applied the same three approaches to examine MWF exposure in relation to all causes of death combined and all cancers combined in order to examine the hypothesis that bias due to healthy worker effect is stronger for chronic diseases, such as heart disease, than for cancer.