New Approach Methodologies (NAMs) for Human-Relevant Biokinetics Predictions: Meeting the Paradigm Shift in Toxicology Towards an Animal-Free Chemical Risk Assessment

For almost fifteen years, the availability and regulatory acceptance of new approach methodologies (NAMs) to assess the absorption, distribution, metabolism and excretion (ADME/biokinetics) in chemical risk evaluations are a bottleneck. To enhance the field, a team of 24 experts from science, industry, and regulatory bodies, including new generation toxicologists, met at the Lorentz Centre in Leiden, The Netherlands. A range of possibilities for the use of NAMs for biokinetics in risk evaluations were formulated (for example to define species differences and human variation or to perform quantitative in vitro-in vivo extrapolations). To increase the regulatory use and acceptance of NAMs for biokinetics for these ADME considerations within risk evaluations, the development of test guidelines (protocols) and of overarching guidance documents is considered a critical step. To this end, a need for an expert group on biokinetics within the Organisation of Economic Cooperation and Development (OECD) to supervise this process was formulated. The workshop discussions revealed that method development is still required, particularly to adequately capture transporter mediated processes as well as to obtain cell models that reflect the physiology and kinetic characteristics of relevant organs. Developments in the fields of stem cells, organoids and organ-on-a-chip models provide promising tools to meet these research needs in the future. This is an Open Access article distributed under the terms of the Creative Commons Attribution 4.0 International license (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution and reproduction in any medium, provided the original work is appropriately cited. Disclaimer: The views expressed in this paper are those of the authors and do not necessarily reflect the views of their institutions.


Introduction
There are clear societal and scientific needs for the development and validation of predictive animal-free methods for safety evaluations to prevent adverse effects in humans by exposure to chemicals (EU, 2010;OECD, 2019a;Oredsson et al., 2019). Whereas much work has been devoted to the development of in vitro screening methods to capture biological effects (toxicodynamics) of chemicals, insight in the absorption, distribution, metabolism, and excretion (i.e. ADME/biokinetics) of chemicals also plays a central role in next-generation risk evaluations that move away from animal experimentation towards animal-free methods (Albrecht et al., 2019;Bessems et al., 2014;Coecke et al., 2013;Desprez et al., 2018;Thomas et al., 2019;Wambaugh et al., 2018). These biokinetic data are needed for the adequate design of in vitro toxicity studies with respect to the application of physiologically relevant chemical test concentrations and inclusion of relevant metabolites for testing. In addition, biokinetic data play a crucial role in the process of quantitative in vitro-to-in vivo extrapolation (QIVIVE) of in vitro toxicity results to obtain equivalent oral/skin/inhalation human potency estimates that can be compared with human exposure to define the risk (Blaauboer, 2010;DeJongh et al., 1999;Louisse et al., 2017;Wambaugh et al., 2018;Yoon et al., 2012). New approach methodologies (NAMs) for biokinetics include both in vitro approaches (experiments using preferably human tissue material or cells), computational (in silico) approaches, and combinations thereof. Examples of different in vitro methods for kinetics that can be performed with animal or human tissue material are: Transwell studies using Caco-2, MDCK, or LLC-PK1 monolayer for intestinal absorption and transporter studies, artificial human skins for skin absorption studies, metabolism models with (primary) human or animal (liver) cells or tissue fractions, plasma protein and tissue binding assays, and placenta transport experiments with human BeWo cells (Punt et al., 2017;Strikwold et al., 2017;Wilk-Zasadna et al., 2015). Recent research also focusses on the development of organ-on-a-chip models to improve individual methods on accuracy by better physiological resemblance and to integrate the methods for different organs within one platform (Maschmeyer et al., 2015;McAleer et al., 2019;Prantil-Baun et al., 2018;Santbergen et al., 2019;van der Made et al., 2019).
Different types of computational (in silico) approaches for biokinetics can be distinguished. In silico approaches are first of all used for predictions of the behavior of a chemical in a body based on the structural properties of the chemical. Examples are quantitative structure activity relationships (QSARs) for absorption (Hou et al., 2004) and different prediction models for tissue partitioning (Berezhkovskiy, 2004;Poulin and Theil, 2002;Rodgers and Rowland, 2006). Secondly, computational approaches are very powerful for the integration of the various in vitro and in silico findings with either classical (one-or two-compartment models ) or physiologically based kinetic (PBK)/physiologically based pharmacokinetic (PB(P)K) models (Bessems et al., 2014;Jamei, 2016;Louisse et al., 2017;Prantil-Baun et al., 2018;Punt, 2018). For example, for predicting human biokinetics in vitro methods using human cells or tissue fractions are generally used in combination with in silico calculated partition coefficients to parameterize PB(P)K models (Jamei, 2016(Jamei, , 2020Jones and Rowland-Yeo, 2013;Wambaugh et al., 2018).
Despite the relevance of human-(fully animal-free) and animal-derived in vitro methods and the different in silico methods for biokinetics, the adoption of NAMs for biokinetic considerations in regulatory risk evaluations is lagging behind (Gellatly and Sewell, 2019;Punt, 2018;Tan et al., 2018;Zhuang and Lu, 2016). This challenge and the ambition of the Dutch government to become world leading with respect to animal free research by 2025, formed the starting point for Wageningen Food Safety Research (WFSR) and the National Institute for Public Health and the Environment (RIVM) to organize a workshop on this topic at The Lorentz Centre in Leiden in The Netherlands in October 2017. The general goal of the Lorentz Centre is to grant and facilitate scientific discussions that can lead to ground-breaking changes in any specific field of research. In total, a team of 24 experts from science, industry, and regulatory institutions, including new generation toxicologists came together to define a strategy to move the field of biokinetics forward. A priori defined goals were (1) to make an inventory of ADME considerations within regulatory risk evaluations that can be addressed using NAMs for biokinetics in both current and next generation risk evaluations that move away from animal experimentation, (2) to define what is needed to increase regulatory use and acceptance of available innovative biokinetic approaches, and (3) to define opportunities of new scientific developments. Based on the discussions around these questions and an evaluation of the recent literature, recommendations (see summary Fig. 1) were made on the steps that are needed to achieve a better incorporation of alternative methods for biokinetics in regulatory chemical risk evaluations.

2
Opportunities for NAMs for biokinetics in current and next generation risk evaluations that move away from animal experimentation # As the field of toxicology is moving away from "black box" animal experimentation towards approaches to better understand the internal concentrations of compounds in relation to their mechanism of toxicity in humans (NRC, 2007;Thomas et al., 2017), the potential of NAMs for biokinetics is increasingly recognised by regulatory bodies (ECHA, 2014;EFSA, 2014;EMA, 2018;FDA, 2020). Formal incorporation of these approaches in regulatory toxicology (for e.g. plant protection products, pharmaceuticals, chemicals, or food additives) is, however, still limited (Gellatly and Sewell, 2019;Punt, 2018;Tan et al., 2018;Zhuang and Lu, 2016). To find means to increase the use of NAMs for biokinetics, an inventory was made of ADME considerations within regulatory risk evaluations that can be addressed with these approaches based on different EU regulations and guidance documents in which the use of in vitro and in silico methods for kinetics methods has been mentioned (Table 1 and 2). Such an inventory is key to the future development of Organisation for Economic Co-operation and Development (OECD) guidelines. These can be guidance documents (GD) on how to use alternative methods to inform risk evaluations and test guidelines (TG) on how to perform a specific in vitro biokinetic study that are generally combined with OECD harmonized templates (OHT; standard data formats for reporting) (OECD, 2019a). Fig. 1: Developments that are needed to increase regulatory use and acceptance of NAMs for biokinetics in risk evaluations and support a transition towards next generation risk evaluations that move away from animal experimentation Tab.1: Inventory of ADME considerations within EU regulatory risk evaluations that can be addressed using NAMs for biokinetics ADME consideration NAMs currently included in guidance documents/regulations Are biokinetic data required? Species differences -Are there differences in the profile of compound metabolites between the laboratory animals used in the toxicity study and humans? -Are there differences in metabolic rate constants? (rate by which a chemical is converted into its metabolites) Regulation (EU) No 283/2013 on the data requirements of active substances in plant protection products: "Comparative in vitro metabolism studies shall be performed on animal species to be used in pivotal studies and on human material (microsomes or intact cell systems) in order to determine the relevance of the toxicological animal data and to guide in the interpretation of findings and in further definition of the testing strategy." (EC, 2013).
Yes. However only the qualitative aspects of species differences in metabolism are being evaluated (i.e. differences in types of metabolites), while quantitative differences in rates of metabolic conversion can be just as important.
-Is the default uncertainty factor for species differences in kinetics appropriate in risk evaluations? -What is the difference in plasma/tissue concentrations of a parent chemical/toxic metabolite between the laboratory animals of the toxicity study and humans? EMA/CHMP/458101/2016 guideline on the reporting of physiologically based pharmacokinetic (PB(P)K) modelling and simulation: "Presently, the main purposes of PB(P)K models in regulatory submissions are to qualitatively and quantitatively predict drug-drug interactions (DDIs) and support initial dose selection in paediatric and first-in-human trials." (EMA, 2018).
No, but increasingly included in regulatory dossiers (Jamei, 2016) Is the default uncertainty factor for human variation in kinetics appropriate in risk evaluations?
Possibilities for PB(P)K modelling of human variation are mentioned in (e.g.): SCCS Notes of Guidance for the Testing of Cosmetic Ingredients (SCCS, 2018).
-ECHA Endpoint specific guidance, including a fictional example of using PBK modelling and the development of assessment factors (ECHA, 2017b).

No
Can any drug-drug interactions be expected? EMA/CHMP/458101/2016 guideline on the reporting of physiologically based pharmacokinetic (PB(P)K) modelling and simulation: "Presently, the main purposes of PB(P)K models in regulatory submissions are to qualitatively and quantitatively predict drug-drug interactions (DDIs) and support initial dose selection in paediatric and first in human trials." (EMA, 2018).

Read-across
Is the selected source chemical appropriate for read-across, having similar kinetics, including similar metabolites?
Mentioned in (e.g.) -REACH. Grouping of substances and read-across approachan illustrative example.  Table 1 provides a summary of ADME considerations within regulatory risk evaluations that can be addressed with NAMs for biokinetics. This overview was obtained by exploring EU guidance documents for chemical risk evaluations within different domains (e.g. pharmaceuticals, chemical substances, food additives, cosmetics) (ECHA, 2013(ECHA, , 2014(ECHA, , 2016(ECHA, , 2017aEFSA, 2010EFSA, , 2014EMA, 2018;EU, 2010;SCCS, 2018). Of the different in vitro and in silico approaches that are available for biokinetics, PB(P)K modelling is the most frequently mentioned tool. For example, in the field of pharmaceuticals, PB(P)K modelling plays a significant role to support initial dose selection in pediatric or first-in-human trials as well as in the prediction of drug-drug interactions (EMA, 2018;Jamei, 2016;Sato et al., 2017;Shebley et al., 2018;Taskar et al., 2019). PB(P)K models within this context are generally developed bottom-up, starting from in vitro and/or in silico kinetic data, after which the models are evaluated and/or fine-tuned against in vivo kinetic data. The label of several approved drugs already contains information on potential drug-drug interactions that is obtained applying such PB(P)K modelling approaches (Jamei, 2016;Sato et al., 2017;Shebley et al., 2018;Taskar et al., 2019). Other documents, such as the guidance for testing of cosmetic ingredients by the European Scientific Committee on Consumer Safety (SCCS, 2018) and the guidance on information requirements and chemical safety assessment by the European Chemical Agency (ECHA) (ECHA, 2014), refer to (bottom-up) PB(P)K modelling as well. In those cases, possibilities for the use of PB(P)K models are most frequently seen to predict species difference and/or human variation in plasma concentrations for refinement of the default inter/intra species uncertainty factors for kinetics, or to perform dose or route-to-route extrapolations. However, so far, the actual use of PB(P)K modelling to meet these latter regulatory topics has been limited (Gellatly and Sewell, 2019;Punt et al., 2017;Tan et al., 2018).

Opportunities in current risk evaluations
Apart from PB(P)K modelling, stand-alone in vitro kinetic data (i.e. not integrated in a kinetic model) can also be requested by regulatory authorities. A recent example is EC Regulation No 1107/2009 which requires to include in vitro metabolism studies (with microsomes or intact cell systems) in the regulatory dossiers of plant protection active substances in order to determine the human relevance of the animal species chosen for the toxicological studies in a dossier. A recent EFSA workshop focused on the key elements to be considered for the interpretation of comparative in vitro metabolism studies and the minimum amount of information that should be obtained to satisfy the data requirement (EFSA, 2019).
Stand alone in vitro/in silico kinetic data may also be used for exposure-based waiving of toxicity tests. An example can be found in the EFSA "guidance for submission for food additive evaluations" (EFSA, 2012). In this guidance document it is indicated that a negligible metabolic conversion of an additive by gastrointestinal fluids or the gut microbiota (in vitro) and negligible absorption (in vitro), together with absence of genotoxicity, provides scientific justification for not undertaking higher tier kinetic and toxicological studies. Another example of the use of kinetic data for exposure-based waiving can be found within the European REACH legislation for industrial chemicals for some cases. Within REACH there is no specific requirement to generate biokinetic data (ECHA, 2017b). For chemicals that are manufactured or imported in quantities of 10 tons or more, an assessment needs to be made of the biokinetic behavior of the substance to the extent that can be derived from the relevant available information (Annex IX Regulation (EC) No 1907/2006). These kinetic data, that may include data from NAMs, can be used to guide the design of appropriate toxicity studies or waive studies. For example, no reproductive toxicity tests are required if a substance "is of low toxicological activity (no evidence of toxicity seen in any of the tests available), it can be proven from biokinetic data that no systemic absorption occurs via relevant routes of exposure (e.g. plasma/blood concentrations below detection limit using a sensitive method and absence of the substance and of metabolites of the substance in urine, bile or exhaled air) and there is no or no significant human exposure" (Annex IX Regulation (EC) No 1907. Given that all these three requirements need to be met as waiver for reproductive toxicity tests, the impact of the biokinetic data as such will remain limited in this context.
A final area in which NAMs for biokinetics are emerging is the justification of read-across. The principle of readacross is to predict endpoint information for a data-poor substance, by using data from (an)other related substance(s) (Cronin and Yoon, 2019;Escher et al., 2019). The role of biokinetics in read-across becomes clear from different ECHA read-across guidance documents for REACH as well as the biocidal product regulation (ECHA, 2013(ECHA, , 2017a(ECHA, , 2017b. According to these guidance documents, the similarity in the biokinetics (e.g. type of metabolites and speed of metabolism) between the target and source substance(s) needs to be considered in the justification of the read-across (ECHA, 2013).

Opportunities in next generation risk evaluations that move away from animal experimentation
Compared with current risk evaluation procedures (Table 1), an even more important role of biokinetics is foreseen in nextgeneration risk evaluations that move away from animal experimentation (summarized in Table 2). Questions related to the extrapolation of animal data to humans are expected to become less important, but methodologies that incorporate human variation and allow route-to-route extrapolation, read-across and exposure-based waiving/priority setting without the use of experimental animals will remain and may even become more important. In addition, new applications of biokinetics will emerge. For example, NAMs for biokinetics will play a crucial role in the design of in vitro toxicity experiments. Particularly given that not all in vitro biological effect assays include metabolism, it is important to characterize the metabolites that are formed in the human body and, if possible, to subsequently assess the potential toxic effects of these metabolites (Escher et al., 2019;Wilk-Zasadna et al., 2015). In addition, biokinetic information provides insights in whether specific exposure scenarios such as irregular peak exposure concentrations, may lead to different cellular responses than with continuous exposure of lower dose levels. In some cases there can also be a significant time delay between in vitro medium concentrations and the internal cell concentration (McAleer et al., 2019;Rozman and Doull, 2000). Such temporalconcentration effects can be important to consider in the design of an in vitro toxicodynamic experiment and the subsequent translation of the results to actual human exposure scenarios. Finally, PB(P)K modelling can also be used as a tool to guide the design of in vitro experiments, for example by providing insight in dimensions and operating conditions for organ-on-achip systems (Abaci and Shuler, 2015;Hartung et al., 2017;Sung et al., 2014).
Another new role of biokinetics (and particularly PB(P)K modelling) in next-generation risk evaluations is to put in vitro-obtained biological activity data in the context of human exposure, also called QIVIVE (quantitative in vitro to in vivo 6 extrapolations). This can be done based on a forward dosimetry approach (i.e. comparing predicted or observed internal plasma or tissue levels with the in vitro concentration-response curves) or based on a reverse dosimetry approach (i.e. extrapolating the in vitro effect concentrations to equivalent oral exposures) (Dancik et al., 2013;Rostami-Hodjegan, 2012;Yoon et al., 2012). An example of forward dosimetry is the comparison of an internal exposure with in vitro biological effect concentrations in so-called exposure-to-activity ratios (EARs) (Becker et al., 2014;Dent et al., 2019). This approach provides predictions on whether certain in vitro measured biological effects are likely to be induced at a certain exposure scenario and provides a means to prioritize the key biological targets, focusing on those that are elicited at exposures (concentration, duration, frequency, delay) that are relevant for the expected internal levels. Furthermore, EARs of different compounds can be compared to place the exposure-activity data of a chemical relative to a known reference compound (Becker et al., 2014;Dent et al., 2019). In contrast, reverse dosimetry approaches focus on the extrapolation of in vitro effective or benchmark concentrations (e.g. EC10, BMC10) into equivalent in vivo exposure-response curves (Louisse et al., 2017;Wetmore, 2015;Yoon et al., 2012). Kinetic modelling (including PB(P)K modelling) plays an important role in obtaining the required insights in the relation between external and internal exposure for QIVIVE either based on forward dosimetry or reverse dosimetry. A prerequisite within the context of animal-free testing strategies is that these models are developed based on in vitro and/or in silico input data themselves (Rostami-Hodjegan, 2012). Furthermore, good PB(P)K modelling practice as well as the quantification of uncertainty and variability within PB(P)K models play a crucial role in QIVIVE to gain confidence in the predictions and to estimate the credible interval around the predicted external dose levels that are linked to the in vitro effect concentrations McNally et al., 2018;Paini et al., 2019;Wambaugh et al., 2019). Alternatively to kinetic (computer) modelling, exposure measurements from human biomonitoring studies (blood, urine, or milk) could be used for QIVIVE, though only for chemicals that are already marketed (Becker et al., 2014). Also then, PB(P)K modelling can be used for the interpretation of the human biomonitoring data, for example to translate a urinary concentration back into the corresponding blood concentration (Clewell et al., 2008;McNally et al., 2012McNally et al., , 2019Rostami-Hodjegan, 2012).
A final role for biokinetics in next-generation risk evaluations is to establish freely available concentrations in in vitro assays. Irrespective of whether the internal predicted exposure levels are compared with the in vitro biological activity data based on forward dosimetry or based on reverse dosimetry, it is important to account for potential differences in the free concentration of the chemical in plasma (or the relevant organ or even cell under consideration) vs. the chemical in the in vitro toxicity test. Apart from protein and lipid binding, the free concentration of a chemical in an in vitro system can also be impacted by evaporation and/or binding to plastic or filter (e.g. microtiter plates), reviewed in Kramer et al. 2015. Calculation tools to predict the free available concentration in an in vitro experiment (Armitage et al., 2014;Fischer et al., 2017;Fisher et al., 2019;Groothuis et al., 2015;Kramer, 2010;Zaldivar Comenges et al., 2017) and the fraction unbound in plasma (Lobell and Sivarajah, 2003;Yamazaki and Kanaoka, 2004) will therefore become crucial tools in next-generation risk evaluations.

Tab. 2: Inventory of regulatory topics in next-generation risk evaluations for which insights in biokinetics (using NAMs) are relevant
Topics related to human variation, route-to-route extrapolation, dose-extrapolations, read-across, and exposure-based waiving/priority setting, as described in Table 1 for current risk assessment procedures, remain of importance in next-generation risk evaluations.
Goal Regulatory topics that can be addressed Guiding the design of in vitro toxicity studies -What are physiologically relevant internal exposure conditions of the chemical in humans? -Which metabolites need to be taken into account in the in vitro studies? -Does the in vitro toxicity assay have the adequate metabolic competence? -Which tissues are relevant to include in the toxicity determination? -How does the chemical behave in an in vitro system with respect to protein, lipid, and plastic binding and in terms of evaporation? -Aiding the development of novel in vitro systems like organ-on-chip systems to determine dimensions and operating conditions Quantitative in vitro-to in vivo extrapolations (QIVIVE) -What are the internal levels of a chemical that are reached at a certain exposure scenario and how do these relate to in vitro effect concentrations? (forward dosimetry approach) -Translation of an in vitro biological effect concentration/exposure (e.g. EC10, BMC10) to an equivalent oral/skin/inhalation dose/exposure. -Improving the interpretation of human biomonitoring studies to reconstruct internal exposure levels for exposure assessment, QIVIVE and IVIVE-PBPK

3
Increasing the regulatory acceptance and use Though the possibilities for NAMs for biokinetics are manifold (see above evaluation), formal inclusion is still limited amongst others due to the lack of robust and reproducible methods (Punt et al., 2017;Sewell et al., 2017;Tan et al., 2018;Zhuang and Lu, 2016). To this end, the experts at the workshop recommended to set up a continuous OECD expert group on biokinetics, analogous to the expert groups on e.g. genotoxicity and skin sensitization, to stimulate and guide the generation of test guidelines for methods to generate these biokinetic parameters in a robust and reproducible way and of guidance documents for their application in risk assessment procedures (see summary Figure 1). In the meantime, this working group has been formally established. Within the context of the OECD, a guidance document focuses on how experimental results can be used to inform risk evaluations (i.e. to answer a specific regulatory question) and test guidelines focus on how to perform a specific study (OECD, 2009). In the area of biokinetics, the OECD TG 417 (with relative harmonised template OHT58) describes in vivo studies. TG 417 provides the protocol for the conduct of studies on mass balance, absorption, bioavailability, tissue distribution, metabolism, excretion, and basic kinetic parameters (e.g. Cmax, AUC) (OECD, 2010).

Standardization and development of test guidelines
The experimental conditions of in vitro kinetic studies can have a large effect on the kinetic parameters obtained (e.g. metabolic clearance, absorption rates). This leads to a high variability in the estimates for these parameters for the same chemical among different studies as recently shown for metabolic clearance experiments (Louisse et al., 2020;Ring et al., 2011). One currently needs to evaluate each study in detail to find out whether the results are in agreement with what is expected in vivo and reliable enough to use for e.g. PB(P)K modelling (e.g. is the intrinsic clearance measured at concentrations below the Km, are the Michaelis-Menten kinetics (Km and Vmax) obtained at incubation conditions in which not too much of the substrate is depleted, which solvents were used as vehicle and at which concentrations, are the right cofactors added, and how was the data processed?). To increase quality and transparency in the biokinetic results, to reduce inter-laboratory difference, and to derive kinetic parameters with a high in vivo (human) relevance, the development of standardized methods and test guidelines are needed. Such test guidelines should include information on the principle of the test, experimental set-up, the inclusion of reference compounds, calculation for the in vitro results (correction factors applied or modelling of the in vitro data), and the reporting of the results (Choi et al., 2019;Fisher et al., 2019;. Recently an OECD guidance document on good in vitro method practices (GIVIMP) for the development and implementation of in vitro methods for regulatory use in human safety assessment has been published. GIVIMP addresses key aspects of good in vitro practice including: clear definition of roles and responsibilities, procedures for storing and handling cells and tissues, ways to prepare test items and avoid cross-contamination, defining and describing standard operating procedures and how to properly report results (OECD, 2018a). When developing test guidelines for NAMs for biokinetics it is important to take the GIVIMP guidance with respect to the critical aspects that need to be considered in the design of in vitro kinetic experiments in general. For regulatory use, the test guidelines should furthermore provide specific boundaries (test concentrations, time-point, etc) in which experiments need to be performed to obtain results that are relevant to the human situation.

Development of guidance documents
Providing that the kinetic data are obtained with protocols that lead to robust and reproducible results, an important second step is to develop guidance documents that aid to systematically interpret submitted kinetic data (e.g. in vitro-in vivo scaling, defining uncertainties) and to define acceptance criteria (e.g. biological and technical considerations like maximum incubation times and acceptable concentration ranges). An example of an OECD guidance document in the field of biokinetics is the recently published OECD guidance document on the determination of in vitro intrinsic clearance using primary hepatocytes or S9 from rainbow trout and extrapolation hereof to an in vivo intrinsic clearance (OECD, 2018b). This guidance document provides information on the selection of the in vitro test system (S9 or primary hepatocytes), consideration of properties of the test compound (solubility), the inclusion of positive and negative controls, and how to use the in vitro data to meet the regulatory topic to improve predictions of chemical bioaccumulation in fish (OECD, 2018b). Similar guidance documents will be needed for the regulatory topics for human risk assessment that are included in Table 1 and 2. However, it is recognized that for many of these questions the development of a guidance document may not be straightforward.
The complexity of developing a guidance document can be illustrated by taking the evaluation of species differences as an example. Although the evaluation of species differences can be expected to be phased out in next-generation risk evaluations, the development of guidance documents for this regulatory topic will still be crucial in the transition towards NAMs to gain confidence in outcomes by making use of available in vivo data for evaluation of the results. An approach to define species differences in kinetics based on in vivo data is for example to compare measured plasma concentrations and/or AUCs of a toxic agent (parent compound or metabolite) between experimental animals and humans (corrected for differences in e.g. dosing and protein binding) (Meek et al., 2003). To obtain such a comparison with NAMs is not straightforward as there can be various causes for differences in plasma concentrations. Interspecies difference in metabolism are the most recognized of these (Cao et al., 2006;Musther et al., 2014). Many metabolizing enzymes, such as cytochromes P450 (Martinez et al., 2019;Nishimuta et al., 2013), UGT (Chiu and Huskey, 1998;Deguchi et al., 2011) and SULT (Punt et al., 2007;Wang et al., 2009) are differently expressed/active in rat, dog, monkey and mice compared to humans. By measuring metabolic conversions with primary hepatocytes or tissue fractions, insights in species differences in metabolic rates and profiles can be obtained. However, focusing on species differences in liver metabolism may overlook other potential kinetic processes that influence plasma concentrations. Expression and/or activity of intestinal transporters for example, can also differ between human and animals. While rodents for instance express Mdr1a and Mdr1b transporters, humans only express one MDR1-encoded protein (P-glycoprotein, P-gp) and BCRP expression is high in rodent kidneys but low in humans (Chu et al., 2013). Differences in physiology may need to be accounted for as well when predicting absorption on the basis of NAMs, including differences in the small intestinal transit time (Sutton, 2004), the intestinal pH and radius (Dressman and Yamada, 1991), microbiome composition (Behr, 2019), and differences in the structure of the gastro-intestinal tract, e.g. no circular folds in the dog (Slatter, 2003) that affect the surface area. Differences in distribution in rat, mouse, dog and human have been associated with differences in protein binding and active transport into organs (Grover and Benet, 2009). In case of renal clearance, species differences can range from 1.6-13-fold as a result of differences in passive renal clearance alone (Walton et al., 2004). Ideally, all these kinetic processes are integrated in a PB(P)K model to allow evaluation of species differences on plasma concentrations of the chemical (Musther et al., 2017). Though PB(P)K models can be made as complex as needed, in practice a balance must be found between accuracy (and therefore complexity) and simplicity (ease of use) (Bois, 2010). This also needs to be kept in mind for the development of guidance documents. Rather than directly focusing on covering all sources of species differences in kinetics, it will be important first to enhance the use of individual (human-based) NAMs for kinetics and to make sure that these are sufficiently understood and provide robust results for application in a regulatory context. In case of hepatic clearance studies, the European Union reference laboratory for alternatives to animal testing (EURL ECVAM), recently established for example, a systematic framework to characterize in vitro methods for human hepatic metabolic clearance in terms of their design, applicability and performance (Gouliarmou et al., 2018). In addition, guidelines are needed on the integration of in vitro kinetic data into in silico kinetic models .

Methods that are considered ready for the development of test guidelines
In Table 3, priorities are set for the methods available for different kinetic processes for standardization and the development of test guidelines and research needs. Particularly, the generation of standardized approaches and a test guideline for hepatic metabolic clearance with primary hepatocytes or hepatic microsomes/S9 is considered as a high priority. Gouliarmou et al. (2018) already made a start with establishing a systematic framework to characterize in vitro methods for human hepatic metabolic clearance to provide a basis for harmonization within the OECD (Gouliarmou et al., 2018). A test guideline for hepatic clearance can potentially be developed further into test guidelines for metabolite identification and extrahepatic (i.e. intestine, kidney, lung, skin) clearance.
The development of test guidelines for different in vitro and/or in silico approaches that predict partition coefficients, fraction unbound in plasma, blood-plasma ratio, and in vitro biokinetics (free fraction as a result of protein binding, plastic binding or evaporation) are also considered relevant. Based on the ample experience with the use of these in silico approaches in kinetic model-development and QIVIVE (Armitage et al., 2014;Fischer et al., 2017;Fisher et al., 2019;Kramer, 2010;Zaldivar Comenges et al., 2017) guidance on their use and applicability domain can effectively be made. A final priority can be given at obtaining a technical guideline for Caco-2 (MDCK, LLC-PK1, or PAMPA) absorption studies to measure passive intestinal permeability. Though alternatives have been developed for Caco-2 permeability experiments (cocultures, 3D, medium flow, peristaltic stretching) (Costa and Ahluwalia, 2019), Caco-2 transwell studies remain the most established among different protocols for permeability measurement (Hubatsch et al., 2007) and can serve as a starting point for a technical guidance development. Opportunities for the development of test guidelines for lung absorption experiments and transporter studies (intestine, kidney, lungs) are currently considered to be limited as further scientific development are needed to obtain physiologically relevant cells and scaling methods (see section 4.2). An OECD test guideline already exists for in vitro dermal absorption studies (OECD, 2004) and a test guideline on measuring induction and inhibition processes of metabolic enzymes is currently being developed (Bernasconi et al., 2019;OECD, 2019b).

Tab. 3: Overview of the current availability of NAMs for different kinetic parameters and the research needs Kinetic processes
Research needs Priority for OECD test guideline development Related regulatory topic (see Table 1 and 2 and references therein) a Liver clearance -Standardization of incubation methods with (cryopreserved) primary human hepatocytes or S9/microsomes -Evaluation of the applicability non-primary cells (e.g. HepaRG TM cells) Hepatic metabolic clearance studies with (cryopreserved) primary human hepatocytes or S9/microsomes are considered a high priority for the development of a TG. Well-established protocols are available. Hepatic clearance data are relevant to a large range of regulatory topics as provided (see Table  1 and 2), as a major determinant of blood and tissue concentrations (Cao et al., 2006;Musther et al., 2014). Potential for extension of the TG to HepaRG clearance studies as an alternative to limitedly available primary cells -Species differences -Human variation -Route-to-route extrapolations -QIVIVE

Metabolite identification
-Standardization of methods that are used based on incubations with S9/microsomes/ (cryopreserved) primary hepatocytes -Higher sensitivity and standardization of analytical chemical techniques and improved methods to characterize metabolites -Applicability of nonprimary cell cultures (e.g. HepaRG TM as potential an alternative limitedly primary cells) TGs for comparative metabolite studies with S9/microsomes/ (cryopreserved) primary hepatocytes are considered a high priority, particularly as comparative metabolite studies are a data requirement within the EU regulation on plant protection products (EC, 2013 Although protocols for Caco-2 transwell absorption studies are well established (Hubatsch et al., 2007), the applicability domain of a TG for these studies will be limited to passive absorption as challenges still exist in the adequate scaling of transporter-mediated influx/efflux processes (Clerbaux et al., 2018).
-Route-to-route extrapolations -Exposure based waiving and priority setting -QIVIVE -Guiding in vitro toxicity test design Tissue binding (fraction unbound in plasma, partition coefficients) -Standardization and the development of guidance documents of available in vitro and/or in silico methods Well-established in vitro methods and in silico calculation tools (Peters, 2012)  Various in silico tools are available for animals and humans with different degrees of complexity, but outcomes may vary depending on the applicability domain (Groothuis et al., 2015).
-QIVIVE -Guiding in vitro toxicity test design (Passive) lung absorption -Development of nonprimary cell cultures with physiologically relevant transporter activity and metabolic capacity Availability of in vitro tools is still limited. Calu-3 monolayer assays are currently the most promising for regulatory use (Ong et al., 2013).
-Route-to-route extrapolations -Exposure based waiving and priority setting -QIVIVE -Guiding in vitro toxicity test design Transporter mediated absorption or excretion processes -Development of in vitro methods that contain physiologically relevant transporter activity (required for intestine, kidney, liver, placenta, blood-brain barrier) -Development of adequate scaling methods Low priority due to limitations in current available methods as described section 4.2 (Clerbaux et al., 2018(Clerbaux et al., , 2019 , current capabilities and challenges are highlighted in a recent review (Taskar et al., 2019) -Species differences -Human variation -Route-to-route extrapolations -QIVIVE -Guiding in vitro toxicity test design

Induction and inhibition processes of metabolic enzymes
Currently being developed (Bernasconi et al., 2019;OECD, 2019b). CYP induction/inhibition can be seen as biomarker for metabolic competence of cells and as potential mechanism of action. The mentioned in vitro and in silico methods for different kinetic processes should be considered as providing parts of the information that is needed to answer the mentioned regulatory topics. Integrated approach (combined in vitro and in silico approaches) will be needed to cover the regulatory topics as a whole.

Opportunities of new scientific developments
Whereas much can be achieved with the standardization and the development of guidelines to increase regulatory acceptance and use of NAMs for biokinetics, this is currently not possible for all aspects of biokinetics. Especially transporter-mediated absorption (intestine), distribution (e.g. blood-brain barrier, placenta, organ uptake) or excretion (e.g. active renal excretion, biliary secretion) is still difficult to capture with NAMs (Clerbaux et al., 2018(Clerbaux et al., , 2019Taskar et al., 2019). Clerbaux et al. (2018) performed a survey among seventy-three experts in the field of transporter kinetics to evaluate the applicability of transporter data for ADME considerations in regulatory risk evaluations. Respondents highlighted the complex interplay with metabolic enzymes and other transporters, species differences, lack of specific transporter substrates and inhibitors, loss of cell polarity, lack of negative controls, problems of inter-laboratory variability, and lack of standardized protocols as key challenges of transporter models (Clerbaux et al., 2018(Clerbaux et al., , 2019. For the use of transporter data in PB(P)K models also scaling approaches are needed to account for the quantitative difference of transporters in cell lines versus tissues (Clerbaux et al., 2018;Galetin et al., 2017;Taskar et al., 2019). Apart from the specific challenges with transporter kinetics, there is also a general need for new cell models to overcome many of the hurdles of the classically used tumor-derived cell lines as well primary cell cultures. Widely used tumor-derived cell lines (e.g. Caco-2 intestinal cells, HepaRG or HepG2 liver cells and A549 lung cells) all have shortcomings to some extent. These either relate to the expression of metabolic enzymes and transporters or to the barrier properties (e.g. the ability to grow to confluence, polarized cell layer(s), mucus secretion) (Ehrhardt et al., 2017;Van Breemen and Li, 2005;Wilk-Zasadna et al., 2015). These shortcomings do not only affect kinetic studies (measurements of metabolism or permeability rates), but also in vitro toxicodynamic studies that require for example metabolic activation. In contrast to tumor-derived cell lines, primary cells closely mimic the physiological state of cells in vivo but can only be kept shortly in culture. Hence, developments are needed to overcome these different challenges.
New technologies involving telomerase overexpression in primary cells, induced pluripotent stem cells (iPSCs), embryonic stem cells (ESCs), or adult stem cells (ASCs) as well as new culturing techniques like 3D spheroid or organoid cultures and the organ-on-a-chip technology are expected to provide novel solutions in the field of NAMs for biokinetics. Relevant examples include the following developments of: -a non-cancerous human renal proximal tubular cell line, RPTEC/TERT1 that has recently been successfully utilised for several long-term kinetic studies of cyclosporine A, cisplatin and adefovir, showing promising results due to the expression of many relevant transporters and metabolizing enzymes (Aschauer et al., 2015;Wilmes et al., 2015).
-bioengineered kidney tubule, intestinal tubule and bile ducts obtained by culturing immortalized cell lines or ASCs on collagen-coated hollow fiber (Chen et al., 2018;Faria et al., 2019;Jochems et al., 2019), -alternatives to Caco-2 cells, such as intestinal organoids from iPSCs, differentiated in the presence of small-molecule compounds, to increase the expression of intestinal markers and pharmacokinetic-related genes (Onozato et al., 2018), -HepaRG cells as metabolic competent tumor cell line that provides promising results with respect to intrinsic hepatic clearance measurements (Zanelli et al., 2012) and metabolic activation of chemicals in toxicity studies (Louisse et al., 2019).
-spheroids from primary hepatocytes, or HepaRG cells that express relevant bile acid transporters (Hendriks et al., 2016), -three-dimensional cultures from primary alveolar cells (Huang and Hsu, 2014) -multi-organ platforms (fluidic coupling of different organs-on-a-chip) that allow for example to mimic the effect of liver metabolic/detoxification on other target organs (e.g. McAleer et al., 2019;Wang et al., 2018).
-the combination of cells or organoids derived from iPSCs, ESCs or ASCs in microfluidic systems, showing that particularly the combination of novel cell sources with new culturing techniques are promising to improve the metabolic and transporter competence of cells in vitro as a result of shear stress (e.g. Homan et al., 2019;Starokozhko et al., 2018).
Although these different examples show that in vitro approaches that better reflect the human physiology are emerging fast, the applicability of these approaches to obtain kinetic parameters still needs be explored (see summary Figure  1).

Integration of NAMs for biokinetics in a next-generation risk assessment framework
New developments in animal-free methods for biokinetics should not stand by themselves. Efforts are needed to seamlessly connect different approaches within next-generation risk assessment, particularly to link the biokinetics and the adverse effects (toxicodynamics) of a chemical. With next generation risk assessment frameworks, methods like high-throughput transcriptomic and high-content imaging, followed by more specific assays play a crucial role in the screening of chemicals for biological effects (Thomas et al., 2019). As long as the results of these assays provide quantitative information (e.g. concentration/exposure response curves), kinetic data can be linked to these biological effect assays to obtain equivalent oral/skin/inhalation human potency estimates that can be compared with human exposure estimates to define the risk. For the use of in vitro biological effect data and QIVIVE, in vitro toxicity studies should be able to pick up all relevant interactions with biological targets or pathways. The Adverse Outcome Pathway (AOP) framework currently provides the most effective framework to organize and describe biological effects in terms of molecular initiating events (MIEs) through key events (KEs), and key event relationships (KERs) in relation to an adverse outcome (Leist et al., 2017;Vinken et al., 2017). At present, a direct link between biokinetics (including QIVIVE) and AOPs is not straightforward. Consideration of biokinetics is not incorporated directly into an AOP description, particularly given that AOPs are not chemical-specific. In addition, AOPs are currently mainly qualitative in nature, providing (narrative) descriptions of the KERs, i.e. the relation between MIEs, KEs, and an adverse outcome (Edwards et al., 2015;Vinken, 2018). Though qualitative AOPs are relevant for hazard identification, moving towards the use of AOPs in risk evaluations also requires quantitative dose-response and time-course information. A relevant new development in this respect are so-called quantitative AOP (qAOP) approaches that focus on the simulation of the dynamic processes linking a molecular initiating event with an adverse outcome using different modelling approaches (Conolly et al., 2017;Schultz and Watanabe, 2018;Zgheib et al., 2019). Such qAOPs can be linked to PB(P)K modelling results. The in vitro exposure that is expected to perturb a MIE or KE would then provide a starting point to estimate an external exposure scenario of possible concern. Therefore, the development of qAOPs will be an important next step to make the required link between biokinetics and toxicodynamics within nextgeneration risk assessment (Edwards et al., 2015;Vinken, 2018).

Conclusion
Figure 1 provides a schematic overview of the workshop results. Overall, it can be concluded that opportunities for the implementation of NAMs for biokinetics in current regulatory risk assessment are to • evaluate interspecies differences, • simulate special populations and human interindividual variability, • perform read-across, • perform route-to-route extrapolations, • set priorities (e.g. exposure-based waiving), • perform dose extrapolations, or • guide in vivo toxicity test selection and design. When switching to a fully animal-free regulatory risk assessment for humans in the future, these methods will be necessary to also • guide in vitro toxicity test selection and design, and • enable QIVIVE. Given the importance of NAMs for biokinetics in a successful transition towards next-generation toxicity testing strategies, both regulatory and scientific experience with in vitro/in silico biokinetic approaches must increase. To facilitate the use of human relevant NAMs for biokinetics, guidance documents are needed that provide information on • how NAMs for biokinetics should be designed to address a regulatory topic, including for example how the different parameters should be scaled and integrated in a kinetic model, • the essential kinetic parameters that at least need to address the regulatory topic, and • the applicability domain of the different methods available for each of these parameters, • the use of different existing in silico tools to calculate e.g. partition coefficient, fraction unbound plasma or blood:plasma ratios, and • how to link NAMs for biokinetics and for toxicodynamics for risk assessment purposes. In addition, the validation of NAMs for biokinetics and development of OECD test guidelines (on how to perform a specific biokinetic study) is important. At the workshop, the development of OECD test guidelines was prioritized based on the availability of current methods and research needs. It was recognized that test guidelines are required for: • intrinsic clearance and identification of metabolites (liver and at later stages also other organs), • in vitro approaches for tissue binding (e.g. fraction unbound plasma, blood:plasma ratio), and • (passive) intestinal absorption (later stages also passive and active permeation into other organs). In case of in vitro skin absorption, a TG already exist, which may need to be further developed. Before test guidelines can be developed for each of these kinetic processes, first protocols are needed that lead to robust and reproducible results. Moreover, it is advised to have a continuous OECD expert group for performing this work on guidance documents and test guidelines. In the meantime, the OECD Expert Group on Toxicokinetics (previously Ad hoc OECD Expert Group on biotransformation) has been established.
Furthermore, it was concluded that method development is required to • adequately address transporter mediated absorption (intestine, lung), distribution (e.g. blood-brain barrier, placenta, organ uptake) or excretion (e.g. active renal excretion, biliary secretion), • obtain cell models that reflect the physiology and kinetic characteristics of different organs, e.g. expression of transporters and metabolizing enzymes. Recent progress in the fields of immortalized primary cells, iPSCs, ASCs or ESCs, 3D organoids and body-on-a-chip systems provide promising tools to meet these research needs. To facilitate the transition towards next-generation risk evaluations, NAMs within the field of biokinetics should be linked to conceptual frameworks, such as (quantitative) AOPs.