A Bayesian framework for estimating parameters of a generic toxicokinetic model for the bioaccumulation of organic chemicals by benthic invertebrates: Proof of concept with PCB153 and two freshwater species

Toxicokinetic (TK) models are relevant and widely used to predict chemical concentrations in biological organisms. The importance of dietary uptake for aquatic invertebrates has been increasingly assessed in recent years. However, the model parameters are estimated on limited specific laboratory data sets that are bounded by several uncertainties. The aim of this study was to implement a Bayesian framework for simultaneously estimating the parameters of a generic TK model for benthic invertebrate species from all data collected. We illustrate our approach on the bioaccumulation of PCB153 by two species with different life traits and therefore exposure routes: Chironomus riparius larvae exposed to spiked sediment for 7 days and Gammarus fossarum exposed to spiked sediment and/or leaves for 7 days and then transferred to a clean media for 7 more days. The TK models assuming first-order kinetics were fitted to the data using Bayesian inference. The median model predictions and their 95% credibility intervals showed that the model fit the data well. From a methodological point of view, this paper illustrates that simultaneously estimating all model parameters from all available data by Bayesian inference, while considering the correlation between parameters and different types of data, is a real added value for TK modeling. Moreover, we demonstrated the ability of a generic TK model considering uptake and elimination routes as modules to add according to the availability of the data measured. From an ecotoxicological point of view, we show differences in PCB153 bioaccumulation between chironomids and gammarids, explained by the different life traits of these two organisms. [Display omitted] •A generic TK model in a Bayesian framework was proposed.•Each uptake and elimination route was considered as a module in the model.•Model parameter estimates are relevant regarding species difference.