Several cohort studies of herbicide manufacturing workers have been conducted over the last years. Most of these studies used simple proxies of exposure in the analysis such as a crude grouping in exposure categories based on job titles, presence in certain production areas over a period of time or during an accident, and duration of exposure. Current serum 2,3,7,8-tetrachlorodibenzo- p-dioxin (TCDD) levels available for a subset of workers can be used to back-extrapolate TCDD levels at the end of exposure using first order kinetic models, and relate TCDD levels to job history using regression models. The regression model obtained can be used to estimate TCDD levels for all cohort members. In this paper, the effect of changes in model assumptions on estimated TCDD levels is explored. TCDD levels are back-extrapolated assuming different values for TCDD half-life. A range of regression models with different sets of exposure determinants is used to relate back-extrapolated TCDD levels to determinants of exposure. These models were used to predict TCDD levels in the epidemiological analysis of data from a Dutch cohort study. The results show that the predicted serum TCDD level is strongly dependent on the assumed half-life. However, the ranking of all individuals on the exposure axis (from low to high) is not affected by changes in the half-life. Predicted serum TCDD levels seem not sensitive to changes in assumption regarding TCDD half-life. Predicted TCDD levels were positively associated with increased (cause specific) mortality.
Read full abstract