Dose-response assessment is used as a criterion in establishing causality from associations observed in epidemiologic studies. It is also used to improve the understanding of biologic mechanisms and as a basis for setting exposure standards in occupational and environmental settings. For this project, we will examine two sources of bias that may alter dose-response relationships. The first bias comes from competing risks as they contribute to the healthy worker survivor effect (HWSE). In another funded project, we are examining four methods for control of the HWSE. None of these methods directly assesses the contribution of competing risks. In this project, we will fit Poisson regression models and separately assess the contribution of exposure-related morbidity and mortality in influencing the shape of the dose-response curve. The second potential source of bias stems from the choice of referent populations in the calculation of relative risks at different dose levels. We will examine concordance between the most commonly used measures of association used in occupational cohort studies, namely -standardized mortality ratios and ratios of directly standardized rates, with respect to dose-response relationships. We will also examine concordance within each method across differing referent populations. Our findings will be validated by examining an outcome not plausibly related to exposure. These objectives will be examined in one or more of three occupational cohorts which have demonstrated (i) a strong dose-response association between arsenic and lung cancer mortality, (ii) a weak association between external penetrating radiation and lung cancer mortality, and (iii) no association between formaldehyde exposed workers and cancer mortality. This investigation will result in an improved understanding of factors influencing dose-response relations. In particular, we will describe possible distortion of dose-response curves due to competing risks that are related to dose and to the choice of referents for cohort mortality studies. Results from this project may then bc used to increase the sensitivity of future occupational studies that utilize quantitative exposure measures, and hence to improve the ability to set appropriate regulatory standards.
