A graphical approach to the identification and estimation of causal parameters in mortality studies with sustained exposure periods.
J Chronic Dis 1987 Jan; 40(Suppl 2):139S-161S
A graphical approach to the identification and estimation of causal parameters in mortality studies with sustained exposure periods was developed. The objective was to account for the effect of independent risk factors that determine later exposure to the study agent (for example, early retirement caused by a chronic disabling illness reducing the long term exposure of the subject). The observed study data were represented on structured tree graphs and causal parameters were identified. Estimable causal parameters were determined from analysis of the graphs using a set of algorithms. Several paradigmatic examples were presented to demonstrate that when there exists a risk factor whose current level both determines subsequent exposure to the study agent and is determined by previous exposure, the observed association of mortality with cumulative exposure can underestimate the true exposure effect. The approach was applied to the effect of arsenic (7440382) exposure on total mortality in a cohort of copper smelter workers. A small adverse effect of arsenic exposure on survival was found, in contrast to standard analysis which was uncontrolled for healthy worker survival effect. The quantitative estimate of causal effect required the fitting of statistical models which may have been biased. However, a nonparametric test of the null hypothesis that there was no exposure effect also showed that arsenic had a statistically significant adverse effect.
NIOSH-Publication; NIOSH-Grant; Humans; Mortality-surveys; Risk-analysis; Arsenic-compounds; Epidemiology; Occupational-exposure; Risk-factors
Environmental Sci & Physiology Harvard Sch of Public Health 665 Huntington Ave Boston, Mass 02115
Journal of Chronic Diseases
Harvard University, Boston, Massachusetts