Predictive modelling of the maternal transfer of organic pollutants to next generations in reptile species

Maternal transfer of pollutants between mothers and their offspring presents one of the first exposure routes for newborn organisms before they themselves come into contact with the external environment, and this during sensitive early life stages where many fundamental development and organization processes take place. To support exposure assessment of early life stages, we aimed to develop a predictive model regarding the maternal transfer of organic pollutants that is applicable across reptile species. We developed a Bayesian orthogonal regression model to characterize the relationship between pollutant concentrations in mother and offspring tissues as dependent variables which are both measured quantities. The model explicitly accounts for uncertainty in measured variables by applying a Gaussian error distribution on both variables and constraining their range in relation to the population mean. Parameter estimation using Bayesian inference allowed the full parameter uncertainty to be propagated into the probabilistic model predictions. To support model development, we conducted a systematic search regarding measurements of organic pollutants in coupled samples from mother and offspring tissues. Subsequent data curation, including homogenization, standardization and integration, resulted in a large, compiled database containing heterogeneous records from diverse species, diverse tissues, and diverse chemicals (including legacy POP such as PCB and OCP as well as newer compounds such as plasticizers). Alternative model formulations and extensions were developed to identify the most parsimonious model that can capture the influence of the diversity in species and chemical structures within the database, as well as account for the large proportion of censored data (records where measured concentrations were below the detection limits of the analytical instruments). Model goodness of fit was evaluated against the WAIC criterion and predictive performance assessed via cross-validation. The resulting model allows scientists and decision-makers to quantitatively link internal body burdens of mothers to subsequent generations in reptile species, a group which is underrepresented in ecotoxicology and chemical risk assessment.