Cefic-lri Programme | European Chemical Industry Council

LRI-ECO47: Improving IVIVE extrapolation models to predict bioconcentration using in vitro biotransformation rates for bioaccumulation assessment in fish


Biotransformation represents the largest source of uncertainty in in silico bioaccumulation predictions as part of PBT assessment for chemicals in the absence of an in vivo bioaccumulation study (commonly OECD TG 305) in fish. To close this gap, in vitro systems measuring biotransformation rates of chemicals to refine BCF model estimates in fish have been established. The reliability and reproducibility of in vitro substrate depletion assays using rainbow trout hepatocytes or liver subcellular fractions have been demonstrated in a recently completed multi-laboratory ring trial. Two OECD draft test guidelines and a guidance document have recently been approved 1-3.

It has been shown that incorporation of biotransformation rates determined in vitro substantially improves model performance. However, there remains a general trend towards underprediction of apparent in vivo biotransformation rates resulting in overprediction of bioaccumulation potential. Overprediction is particularly observed when hepatic clearance is assumed to be controlled by the unbound (free) chemical concentration in vitro and in vivo.

In vitro biotransformation rates are used to calculate a whole-body metabolism rate (kMET) which is an important parameter in the in vitro-in vivo extrapolation (IVIVE) model to predict BCFs. The model includes a term (fraction unbound, fU) that corrects for potential binding effects on clearance. FU is calculated (fU calc) as the ratio of free chemical fraction in blood plasma and the in vitro S9 (or hepatocyte) system 4. This correction assumes that biotransformation enzymes operate against the free or unbound chemical fraction in vivo (in plasma) and in vitro, with faster metabolic rates in vitro vs. in vivo due to lower protein concentration in the test system compared to in vivo plasma protein concentrations. IVIVE results are often improved for pharmaceutical compounds in mammalian in vitro systems when fU is calculated. However, it has been shown by several authors that fU calc results frequently in an overprediction of the BCF in fish. For some hydrophobic chemicals setting fU equal to 1.0 resulted in much better predictions 5-8. Setting fU to 1.0 assumes that chemical availability to metabolic enzymes in vitro and in vivo is the same, either because binding does not limit kinetic turnover under the dynamic flow-through situation of the liver, or, because the fraction available in both systems is identical. Most mechanistic work in this field had been done on polyaromatic hydrocarbons (PAHs), and generalization to industrially relevant chemicals is lacking.

The difference in predicted BCFs using fU calc and setting fU = 1.0 is in particular important for B assessment for values close to the regulatory cut-off criterion for bioaccumulation (i.e. >2000 L/kg for EU REACH; >1000 L/kg for US EPA). For slowly to moderately biotransformed chemicals which also have a high log Kow, predicted BCFs are frequently <2000 L/kg if fU is set to 1.0 and substantially greater than this B cut-off value if fU calculated is used, raising the critical question which values are relevant for final bioaccumulation assessment.

Two draft OECD test guidelines on determination of fish in vitro hepatic clearance using hepatocytes and liver S9 sub-cellular fractions from rainbow trout and an accompanying guidance document were approved by the OECD WNT in April 2018. Regulatory acceptance will be needed to support a broad application of IVIVE models for bioaccumulation assessment. Thus, relevance of these models has to be established, too. Further understanding of the impact of the fraction unbound and particularly improvement of modelling for the extrapolation from the in vitro metabolic rate to the whole-body metabolism rate (kMET) is therefore important to evaluate in vitro biotransformation measurements and BCF predictions based on IVIVE. This is considered as a key issue for regulatory acceptance of in vitro data. Only with broad regulatory acceptance of the IVIVE predictions, unnecessary higher-tier testing can be avoided to fulfil the 3 Rs approach to significantly reduce the use of vertebrate testing.

The goal of this CEFIC LRI project is to improve the IVIVE approach for fish by addressing the major uncertainties of the predictions. If improvements of the modelling of e.g. protein binding as one of the major uncertainties does not resolve the discrepancy between predicted and in vivo measured BCFs, a fundamentally different, refined IVIVE model is needed. This project will provide complementary results with the following CEFIC LRI projects:  ECO 34, ECO 37, ECO 41, ECO 44 and ARC.3 which are briefly discussed below.

The objective of ECO 34 (“A tiered testing strategy for rapid estimation of bioaccumulation by a combined modelling - in vitro testing approach.”) is to reduce the uncertainty related to the estimation of bioaccumulation of organic chemicals in fish. The project focuses on different in vitro approaches to estimate chemical uptake and biotransformation (liver, gill and intestine) with toxicokinetic and QSAR models. So, there are overlapping activities regarding the modelling part with this project. However, the ultimate goal of this current project is to improve IVIVE by addressing major uncertainties like protein binding and extrapolation from in vitro metabolic rates to predicted whole-body metabolism rates (kMET) which will be complementary to ECO 34 and will result in a further improvement of the models for bioaccumulation assessment.

ECO 37 (“D-BASS: Developing a Bioaccumulation Assessment Strategy for Surfactants”) aims to validate the combined use of partition coefficients, in vitro intrinsic hepatic clearance and IVIVE for ionogenic compounds (mainly surfactants), whereas this current project evaluates neutral hydrophobic chemicals from industrially relevant chemicals. Due to the chemical properties that distinguish ionogenic chemicals from neutral chemicals, specific considerations have to be applied for IVIVE extrapolation in ECO 37.

The project ECO 41 (“Enhanced screening methods to determine bioaccumulation potential of chemicals in air-breathing species.”) aims to develop an approach for assessing bioaccumulation of neutral hydrophobic organic chemicals in air-breathing species. Whereas the methodology (i.e. in vitro biotransformation assays, IVIVE) is similar, ECO 41 targets metabolism of air-breathing vertebrates using rat as model species, which requires IVIVE models which are adopted for air-breathing species with e.g. different routes of uptake and excretion compared to fish.

In contrast to this current project, ECO 44 (“Integrating Bioaccumulation Assessment Tools for Mammals (iBAT-Mam)”) focus like ECO 41 on bioaccumulation assessment in mammals. ECO 44 aims to develop a toxicokinetic modelling framework to assess the bioaccumulation behaviour of chemicals in mammals.

The general objective of ARC.3 (“Development of the bioaccumulation assessment tool (BAT VER.1.0) to aid in the bioaccumulation assessment of organic chemicals”) is to develop a tool for integrating various lines of evidence in a quantitative weight of evidence approach to aid regulatory decision-making for bioaccumulation assessment. Improvement of IVIVE for neutral hydrophobic organic chemicals as outlined in this current project may be possibly integrated in this approach to improve bioaccumulation assessment in fish for regulatory applications.


This CEFIC LRI project intends to refine the current IVIVE models as an important step towards the regulatory acceptance of the IVIVE predictions using in vitro data for bioaccumulation assessment.  This will be done by a thorough investigation of the major uncertainties of IVIVE models, like the fraction unbound and other uncertainties with respect to physiological parameters.

Importantly, the project aims to investigate chemicals of different industrially relevant classes. Cefic LRI project monitors and the research team for this project will discuss the range of chemistries to be covered at the project start, and appropriate substances will be defined based on appropriate property threshold criteria. High quality empirical BCF data should be available for the test chemicals for comparison.

The objectives of the project are:

  1. Review the accuracy and applicability of existing methods to determine the fraction unbound for chemicals in order to choose a method which is suitable for a broad range of industrially relevant classes with different chemical properties, e.g. PDMS depletion method 6, thin-film solvent dosing 9, vial equilibration method 10-12.
  2. Determine the fraction unbound for chemicals of different industrially relevant classes and a broad range of log Kow values (i.e. >~4.5 and <8) including particularly higher log Kow substances (i.e. >~6) in fish plasma vs. liver S9 fractions or hepatocytes to refine binding assumptions. As current method to determine the fraction unbound for chemicals with different properties, the vial equilibration method seems to be the most appropriate for chemicals with different properties 10-12. However, this has to be verified according to objective 1.
  3. Perform experimental studies to evaluate the accessibility of the bound chemical fraction for biotransformation enzymes using industrially relevant chemicals (e.g. in cellular systems 13 and ideally in isolated perfused liver 14,15).
  4. Understand to which extent kinetics of mass transfer limits substrate availability of the bound fraction under the dynamic flow-through conditions of the liver.
  5. Improve current IVIVE models 4 with a focus on refined fU calculations based on the experimentally determined fU values (results from objective 2) to predict the bioaccumulation potential for industrially relevant, hydrophobic chemicals. Alternative IVIVE models may be explored (optional) which use other types of corrections for the impact of protein binding.
  6. Address other uncertainties in current IVIVE models 4 with respect to physiological parameters like the apparent volume of distribution (VD) to further improve the prediction models.
  7. An improved estimate of fU may not resolve the discrepancy between BCF estimated by IVIVE and in vivo values as indicated in liver perfusion assays 15. Thus, in this case, a fundamentally different, refined model, which takes e.g. into consideration the dynamic flow-through conditions of the liver, is needed to derive a whole-body metabolism rate (kMET) from in vitro metabolism rates.
  8. Apply the refined IVIVE model for comparison of predicted and measured BCFs for industrially relevant chemicals. This will involve a comparison of predicted BCFs based on in vitro biotransformation rates using the improved IVIVE model with empirical BCFs (OECD 305 studies, good quality data, same fish species, i.e. rainbow trout; in ideal case even matched species using the same strain).
  9. Validate as alternative approach the refined IVIVE model by a comparison of modelled in vivo biotransformation rates based on in vitro data vs. in vivo biotransformation rates determined in fish. This approach allows a direct evaluation of the IVIVE without the potential influence of factors associated with the uptake and biotransformation independent depuration of the chemical, e.g. gill uptake rate constant, gill elimination rate constant.

Related links

Download here the full version of the RfP LRI-C47.

Timing: 2 years

Cost: €300.000

Cefic-Lri Programme Responsible Care

© Copyright 2017 Cefic | European Chemical Industry Council. All rights reserved | Terms and Conditions of Use