Incidence of prostate cancer among auto workers exposed to metalworking fluids

To examine whether exposure to metalworking fluids (MWF) is associated with prostate cancer, a nested case-control study of 872 incident cases was conducted in a cohort of autoworkers with quantitative exposure information. The objectives of this study were to examine the shape of exposure-response curves using semi-parametric modeling and two contrasting sets of exposure windows defined by latency, and differences in susceptibility of prostate cells due to hormonal changes.;Cases in the GM/UAW cohort were identified through Michigan cancer registry from 1985 through 2000. Controls were selected using incidence-density sampling with 5:1 ratio. Using cumulative exposure (mg/m3-years) as the dose metric, we evaluated latency time windows of exposure: ≥25 and <25 years prior to risk age. Using a hormonal biological model, we defined three consecutive windows of exposure: (1) puberty to early adulthood; (2) adulthood; and (3) middle age or andropause. We used penalized splines to model risk as a smooth function of exposure, and controlled for race, and calendar year of diagnosis in a Cox model.;Risk of prostate cancer increased with cumulative exposure to soluble and straight MWF ≥25 years prior to risk age, but not with exposure in the later window (<25 years). The relationship with soluble MWF was piecewise linear, with a small increase in risk at lower exposures followed by a steeper rise. By contrast, the relationship with straight MWF was linear with a relative risk of 1.12 per 10 mg/m3-years. Based on biological exposure windows, association was strongest in the earliest window. Risk of prostate cancer increased linearly with exposure to straight MWF from puberty to early adulthood, with a relative risk of 2.38 per 10 mg/m3-years. Subjects exposed to MWF at young age had also increased risk of prostate cancer due to exposure incurred later in life.;Results provide evidence that the latency period for prostate cancer is at least 25 years, and point to heightened risk with early exposure. Results also suggest that exposure characterization based on the biological approach better captures risk of prostate cancer. Semi-parametric modeling was useful as a guide to choosing the appropriate parametric model.