Upstream adverse effects in risk assessment: a model of polychlorinated biphenyls, thyroid hormone disruption and neurological outcomes in humans.

BACKGROUND: Increasing data on early biological changes from chemical exposures requires new interpretation tools to support decision-making.
OBJECTIVES: To test the possibility of applying a quantitative approach using human data linking chemical exposures and upstream biological perturbations to overt downstream outcomes.
METHODS: Using polychlorinated biphenyl (PCB) exposures and maternal thyroid hormone (TH) perturbations as a case study, we model three relationships: (1) prenatal PCB exposures and TH changes, using free T(4) (FT(4)); (2) prenatal TH and childhood neurodevelopmental outcomes; and (3) prenatal PCB exposures and childhood neurodevelopmental outcomes (IQ). We surveyed the epidemiological literature; extracted relevant quantitative data; and developed models for each relationship, applying meta-analysis where appropriate.
RESULTS: For relationship 1, a meta-analysis of 3 studies gives a coefficient of -0.27 pg/mL FT(4) per ln(sum of PCBs) (95% confidence interval [CI] -0.82 to 0.27). For relationship 2, regression coefficients from three studies of maternal FT(4) levels and cognitive scores ranged between 0.99 IQ points/(pg/mL FT(4)) (95% CI -0.31 to 2.2) and 7.6 points/(pg/mL FT(4)) (95% CI 1.2 to 16.3). For relationship 3, a meta-analysis of five studies produces a coefficient of -1.98 IQ points (95% CI -4.46 to 0.50) per unit increase in ln(sum of PCBs). Combining relationships 1 and 2 yields an estimate of -2.0 to -0.27 points of IQ per unit increase in ln(sum of PCBs).
CONCLUSIONS: Combining analysis of chemical exposures and early biological perturbations (PCBs and FT(4)) with analysis of early biological perturbations and downstream overt effects (FT(4) and IQ) yields estimates within the range of studies of exposures and overt effects (PCBs and IQ). This is an example approach using upstream biological perturbations for effect prediction.