Determination of safe levels of persistent organic pollutants in toxicology and epidemiology

We reviewed published manuscripts from toxicology and epidemiology reporting harmful health effects and doses of persistent organic pollutants (POPs), published between 2000 and 2021. We found 42 in vitro, 32 in vivo, and 74 epidemiological studies and abstracted the dose associated with harm in a common Molar unit. We hypothesized that the dose associated with harm would vary between animal and human studies. To test this hypothesis, for each of several POPs, we assessed the significance of variation in the dose associated with a harmful effect [categorized as non-thyroid endocrine (NTE), developmental neurotoxicity (DNT), and Thyroid] with study type (in vitro, in vivo, and Epidemiology) using a linear model after adjustment for basis (lipid weight, wet weight). We created a Calculated Safety Factor (CSF) defined as the toxicology dose divided by epidemiology dose needed to exhibit significant harm. Significant differences were found between study types ranging from <1 to 5.0 orders of magnitude in the dose associated with harm. Our CSFs in lipid weight varied from 12.4 (95% confidence interval (CI) 3.3, 47) for NTE effects in Epidemiology relative to in vivo studies to 6,244 (95% CI 2510, 15530) for DNT effects in Epidemiology relative to in vitro in wet weight representing 12.4 to 6.2 thousand-fold more sensitivity in people relative to animals, and mechanistic models, respectively. In lipid weight, all CSF 95% CI lower bounds across effect categories were less than 6.5. CIs for CSFs ranged from less than one to four orders of magnitude for in vivo, and two to five orders of magnitude for in vitro vs. Epidemiology. A global CSF for all Epidemiology vs. all Toxicology was 104.6 (95% CI 72 to 152), significant at p<0.001.