Causal Inference in Occupational Epidemiology: Asbestos, Lung Cancer Mortality, and the Healthy Worker Survivor Effect Public Deposited

Downloadable Content

Download PDF
Last Modified
  • March 22, 2019
Creator
  • Naimi, Ashley Isaac
    • Affiliation: Gillings School of Global Public Health, Department of Epidemiology
Abstract
  • The healthy worker survivor effect is well recognized as a potential source of bias in occupational epidemiology. Three component associations are necessary for this bias to occur: (i) prior exposure and employment status; (ii) employment status and subsequent exposure; and (iii) employment status and mortality. Together, these associations result in time-varying confounding affected by prior exposure. Previous estimates of the effect of occupational asbestos on lung cancer mortality have been obtained using methods that do not account for such confounding. Recent advances in causal inference provide key tools to examine the severity of the healthy worker survivor effect in a given cohort, and estimate an exposure-outcome relation accounting for this bias. The former relies on the use of causal diagrams developed by Pearl (2000), allowing researchers to assess the magnitude of the component pathways in an assumed causal structure. The latter relies on the work of Robins (1989), who introduced g-estimation of a structural nested failure time model to estimate causal effects using observational data subject to biases such as the healthy worker survivor effect. The research for this dissertation was conducted using data from 3,072 asbestos textile factory workers hired between January 1940 and December 1965 and followed through December 2001. First, we illustrate how the component associations can be assessed using standard regression methods. For a 100 fiber-year/mL increase in cumulative asbestos, the covariate-adjusted hazard of leaving work decreased by 52% (95% confidence interval: 46, 58). The association between employment status and subsequent asbestos exposure was strong due to nonpositivity: 88.3% of person-years at work (95% confidence interval: 87.0, 89.5) were classified as exposed to any asbestos; no person-years were classified as exposed to asbestos after leaving work. Finally, leaving active employment was associated with a 48% (95% confidence interval: 9, 71) decrease in the covariate-adjusted hazard of lung cancer mortality. Second, we estimated the effect of cumulative asbestos exposure on lung cancer mortality using two modeling strategies. For a 100 fiber-year/mL increase in cumulative asbestos, a standard Weibull model adjusting for gender, race, birth year, baseline exposure, and age at study entry yielded a survival time ratio of 0.88 (95% confidence interval: 0.83, 0.93). Further adjustment for work status yielded no practical change. The corresponding survival time ratio obtained using g-estimation of a structural nested model was 0.57 (95% confidence interval: 0.33, 0.96). Accounting for the healthy worker survivor effect resulted in a 35% stronger effect estimate. However, this estimate was considerably less precise. When suspected, methods that account for the healthy worker survivor effect should be used.
Date of publication
Keyword
Resource type
Rights statement
  • In Copyright
Advisor
  • Cole, Stephen
Degree
  • Doctor of Philosophy
Degree granting institution
  • University of North Carolina at Chapel Hill
Graduation year
  • 2012
Language
Publisher
Parents:

This work has no parents.

Items