Estimating serum polychlorinated biphenyl levels in highly exposed workers: an empirical model.
Taylor-PR; Reilly-AA; Stelma-JM; Lawrence-CE
J Toxicol Environ Health 1991 Dec; 34(4):413-422
A regression model was developed to estimate polychlorinated biphenyl (PCB) serum concentrations according to job exposure categories. Two facilities manufacturing capacitors using PCBs with Aroclor-1254 (11097691), Aroclor-1242 (53469219), and Aroclor-1016 (12674112) as their primary dielectric fluid from 1946 to 1977 were evaluated in 1979 and 1983. One hundred forty seven employees were studied; complete histories were acquired along with serum Aroclor- 1254 measurements. Calculations at this stage assumed first order kinetics and an Aroclor-1254 half life of 3.32 years was estimated. All jobs were categorized into two exposure groups and four specific duration related subgroups. Classical multiple regression techniques based on least squares were employed with serum high homolog PCB as the dependent variable and exposure category months as the independent variables. The regression models were applied in a study of the birth weight and gestational ages of infants born to female employees. The efficiency of regression based exposure estimates were compared to commonly used epidemiological parameters through investigation of dichotomized, ordinal, and continuous exposure surrogates. Of the alternative exposure categorizations, the ever versus never direct measure was found to be a poor indicator of serum PCB levels. In fact, nearly all measures tried were poor predictors; the best after the fact was total months employed in direct exposure jobs. The authors conclude that, compared to the deductive approaches, the empirical model offers a substantially improved way of estimating and categorizing exposure in epidemiological studies.
NIOSH-Publication; NIOSH-Contract; Contract-210-81-5102; Statistical-analysis; Mathematical-models; Blood-serum; Epidemiology; Polychlorinated-biphenyls; Electronics-industry; Occupational-exposure; Humans
11097-69-1; 53469-21-9; 12674-11-2
Journal of Toxicology and Environmental Health