Epidemiologic investigations of associations between protracted low level occupational exposures and cancer mortality routinely encounter the following problems: 1) potential latency effects between exposure and disease; 2) potential bias resulting from exposure measurement error; and, 3) potential bias resulting from health-related selection out of employment (i.e., the healthy worker survivor effect). The identified problems are of direct relevance to worker protection, as each is a source of bias that may lead to spurious conclusions about the adverse effects of occupational hazards. The goal of this project is to improve the analytical tools available to address these problems. We will develop a conceptual description of each problem, develop a simple analytical tool (or tools) to reduce or eliminate the potential bias, evaluate the proposed analytical method via simulation analyses, and then illustrate the application of the proposed method using empirical data. We will begin by exploring the use of flexible latency models for occupational cancer studies. Via simulation analyses we will evaluate the use of these flexible models for reducing bias due to mis-specification of exposure lag assumptions; and, in empirical analyses of rubber hydrochloride and asbestos textile worker cohort data we will illustrate the application of these methods. Next, we will develop an approach to control for bias that can arise when grouped data are used to assign exposure scores, as in a job-exposure matrix. The use of assigned exposure values is often assumed to result in a Berkson error model that does not produce biased risk estimates. Using simulated data, we will evaluate the conditions under which the Berkson model applies, and develop approaches to exploit this error model to reduce bias. We will illustrate these approaches with empirical data from a cohort study of electrical utility workers. Finally, we will identify the conditions under which non-standard regression methods (e.g., G-estimation) are necessary in cohort studies to control for the healthy worker survivor effect. We will use simulation methods to explore these conditions, and develop simple analytical tools to guide investigators on when to use G-estimation. The proposal addresses the NORA priority area on cancer research methods. The results of the proposed research will further improve the analytic methods used in occupational cancer studies.
Showing the most recent 10 out of 22 publications
