Estimating Time-Varying PCB Exposures Using Person-SpecificPredictions to Supplement Measured Values: A Comparison ofObserved and Predicted Values in Two Cohorts of NorwegianWomen

Abstract

Background: Studies on health effects of polychlorinated biphenyls (PCBs) call for an understanding of past and present human exposure. Time-resolved mechanistic models may supplement information on concentrations in individuals obtained from measurements and/or statistical approaches, if they can be shown to reproduce empirical data. Objectives: Here we evaluate the capability of one such mechanistic model to reproduce measured PCB concentrations in individual Norwegian women. We also assess individual life course concentrations. Methods: Concentrations of four PCB congeners in pregnant women (n= 310, sampled in 2007-2009) and postmenopausal women (n= 244, 2005) were compared to person-specific predictions obtained with CoZMoMAN, an emission-based environmental fate and human food-Chain bioaccumulation model. Person-specific predictions were also made with statistical regression models including dietary and lifestyle variables and concentrations. Results: CoZMoMAN accurately reproduced medians and ranges of measured concentrations in the two study groups. Further, rank correlations between measurements and predictions from both CoZMoMAN and regression analyses were strong (Spearman`s r >0.67). Precision in quartile assignments from predictions was strong overall as evaluated by Weighted Cohen`s Kappa >0.6. Simulations indicated large inter-individual differences in concentrations experienced in the past. Conclusions: The mechanistic model reproduced all measurements of PCB concentrations within a factor of ten and subject ranking and quartile assignments were overall largely consistent, although weak within each study group. Contamination histories for individuals predicted by CoZMoMAN revealed variation between study subjects, particularly in the timing of peak concentrations. Mechanistic models provide individual PCB exposure metrics that could serve as valuable supplements to measurements.