Impact of the specificity of the exposure metric on exposure-response relationships

Exposure misclassification may occur when nonspecific exposure indicators are used. Developing estimates of more specific measures may be difficult due to sampling limitations or a paucity of historical measurements and, thus, often requires substantial effort. We examine the impact on exposure-response relationships of moving from 2 measures of exposure mixtures (dust, chlorophenols) to more specific exposure indicators (wood dust, pentachlorophenol, tetrachlorophenol) in a retrospective cohort.

The study population consisted of 26,847 male sawmill workers (> or =1 year employment between 1950 and 1995) with linkage to national cancer registries. A subcohort (n = 11,273 employed more than 1 day between 1985 and 1995) was linked to hospital discharge records. We evaluated the shape (log-linear vs log-log models), goodness of fit, precision, and expected versus observed attenuation of the exposure-response relationships.

The correlation between the cumulative exposure indices was moderately high (dust/wood dust, r = 0.68; total chlorophenol/pentachlorophenol, r = 0.88; total chlorophenol/tetrachlorophenol, r = 0.78). An increase in chronic obstructive pulmonary disease hospitalizations was found with wood dust but not with total dust. Stronger associations for non-Hodgkin lymphoma and kidney cancer incidence were observed with pentachlorophenol than with total chlorophenol; no association was observed with tetrachlorophenol. We observed greater attenuation than expected using total dust, but less than expected using total chlorophenol.

The relationships between health outcomes were substantially attenuated when nonspecific exposure indicators were used. This study demonstrates the importance of developing exposure metrics as specific to the disease-causing agent as possible, particularly when the composition of mixed exposures varies by work areas.