The Use of Toxicokinetic Information for Setting Concentrations of In Vitro Toxicity Tests and for Interpreting Their Results: A Proposed Workflow

Abstract

Background:

In vitro assays are increasingly used in the toxicological testing of chemicals. These use high concentrations and the only criteria to set an upper concentration are related to solubility and cytotoxicity or by a predefined concentration as a default. However, the top concentrations in vitro should also consider the equivalent dose in vivo. A concentration causing an effect in vitro that exceeds the internal concentration at a maximum external dose in vivo may not be relevant for toxicological assessment since the cellular effect observed in vitro cannot occur in vivo.

Materials and Methods:

To avoid this, we propose the use of measured or modelled toxicokinetic data to extrapolate the top dose in vivo to an upper concentration in vitro. This in vivo to in vitro dose extrapolation (IVIVE) uses pharmacokinetic/toxicokinetic modelling to yield an in vitro concentration (c_maxIVIVE) which reflects the internal dose at the maximum external dose in vivo.

Results:

The c_maxIVIVE can be used to (1) define a relevant upper concentration for the in vitro testing (integrated approach) and (2) interpret the results of an in vitro test, even if the top concentration exceeds the c_maxIVIVE (post hoc approach). We describe a workflow which uses solubility and cytotoxicity data, as well as in vivo and in vitro toxicokinetic data, to derive c_maxvivo or c_maxIVIVE to optimize the setting of the top concentration of a test substance in an in vitro test system. Finally, we demonstrate the benefit of using this approach using the herbicide, Dicamba, as an example.

Conclusion:

A workflow based on solubility and cytotoxicity data, as well as in vivo and in vitro toxicokinetic data, can be used to set the top concentration of a test substance in an in vitro test.

Introduction

Toxicological studies are performed to identify hazardous properties of a test substance and to describe the dose–response relationship. In vitro assays are used, partly not only to reduce or avoid animal testing but also to generate information which cannot directly be obtained from animal testing but is relevant to characterize hazard and risk in humans. There are multiple in vitro assays already used routinely to determine either adverse outcomes, such as skin and eye irritation, or corrosion (Organisation for Economic Co-operation and Development [OECD] Test Guidelines [TG] 439, 435 and 431, 492, etc.^1–4), or key events of adverse outcome pathways (AOPs), such as skin sensitization (OECD TG no. 442C, D and E^5–7). These assays all relate to local effects, whereby in vivo exposure can directly be correlated to in vitro exposure if tissue models with an air–liquid interface are used. Other OECD in vitro methods are related to systemic effects, for example, endocrine effects (OECD TG 455, 458, 493; cf. OECD Guidance document 150⁸), immunotoxicity (OECD TG 444A⁹), and genetic toxicology (OECD TG 473, 476, 487,^10–12 etc.). For these assays, in vivo to in vitro extrapolation (IVIVE) is needed to correlate external doses to in vitro concentrations.

For some substances (e.g., agrochemicals), toxicokinetic information is routinely obtained in in vivo studies (OECD TG 417¹³); limited toxicokinetic information can also be obtained from plasma concentrations of the test substance analyzed in toxicodynamic in vivo studies (e.g., OECD TG 474¹⁴). However, due to ethical and regulatory demands, these studies are intended to be replaced in the future by nonanimal in vitro, ex vivo, and in silico models, the so-called “new approach methods” (NAMs). These approaches are particularly relevant for cosmetic ingredients and chemicals regulated under Registration, Evaluation, Authorisation and Restriction of Chemicals, for which available data are limited or lacking,^15–17 whereas the use of NAMs for pesticides is currently limited to cases where they match the design of the in vivo studies by providing mechanistic data, gap filling, and addressing of endpoints which are currently not or not sufficiently covered in vivo. This would, however, require some “back-translation” into doses relevant for adversity by IVIVE.¹⁸

In in vivo studies, the top dose is limited by general toxicity (e.g., mortality or body weight loss); the maximum tolerated dose (MTD) or is set to a predefined limit dose [e.g., 1000 mg/(kg·d) in OECD TG 414¹⁹] by default—often correlating with limits of hazard classification (Fig. 1A). Obviously, results of an in vivo study are of less relevance to the hazard and risk assessments which are observed at doses above the limit dose. The general procedure of performing and planning an in vitro assay should be based on the OECD Guidance Document on Good In Vitro Method Practices (GIVIMP).⁸ This includes careful consideration of the test concentrations. Currently, in vitro assays are performed by applying concentrations up to a predefined concentration (e.g., “10 mM, 2 mg/mL or 2 μL/mL whichever is the lowest” in OECD TG 473¹⁰) by default or up to the solubility limit or to concentrations causing cytotoxicity (the extent of cytotoxicity defined as limiting varies among in vitro assays; Fig. 1B).

FIG. 1.

Dose setting in (A) in vivo and (B) in vitro assays.

Dose setting for in vivo studies considers technical limitations (e.g., generation of respirable aerosols for inhalation studies), biological limitations (MTD), and relevance (limit dose). Concentration setting for in vitro testing also considers technical and biological limitations (solubility and cytotoxicity). A predefined default upper concentration (e.g., 10 mM) is, however, not applied throughout all methods and is not equivalent to a limit dose in vivo [e.g., 1000 mg/(kg·d)] as it is not directly related to actual external exposures. Instead of using a rather arbitrary concentration limit of, for example, 10 mM (or any other value) the upper concentration in in vitro tests should be related to the highest dose of the corresponding in vivo tests—be it the MTD or the default limit dose. This can be achieved by translating the external in vivo dose to the in vitro assay concentration using in silico modeling (i.e., IVIVE). This will avoid in vitro test results which are of little relevance because they occur at concentrations which would not be achieved as internal doses in vivo.

There are three situations in which in vitro results obtained at high concentrations need to be assessed with caution regarding their relevance: (i)

Responses in vitro which only occur at concentrations inducing general and unspecific cytotoxicity. This has also been recognized by Judson et al.,²⁰ when evaluating responses of 1060 chemicals using high throughput testing. They proposed to differentiate responses into (1) specific molecular interactions against one or more targets and (2) activity which is associated with cell stress or cytotoxicity, which may result from a number of different causes and should not be considered as a specific activity response. Cell stress and cytotoxicity are not always or easily detectable with standard cytotoxicity assays, but can nevertheless impact the physiological responses of the cell and thereby the specificity of the in vitro outcome of the assay. Therefore, Judson et al.²⁰ followed a combination of different assay results and computational approaches to derive cytotoxicity limits per substance.

(ii)

Responses in vitro which only occur at concentrations which exceed the internal concentration in corresponding in vivo studies. Here, key events of an AOP would be identified in vitro, but the adverse outcome of this AOP would not be observed in in vivo studies.

(iii)

Responses in vitro which only occur at concentrations which exceed the internal dose resulting from actual human exposure (but not necessarily from in vivo animal studies). This can be considered for priority setting or risk assessment purposes.^21–23 It is, however, not limiting hazard assessments, which are largely independent of actual human exposure.

To address point (ii) raised above, we propose IVIVE using pharmacokinetic/toxicokinetic (PBPK/PBTK) modeling to design in vitro studies and asses their results. IVIVE yields an in vitro concentration (c_maxIVIVE) which reflects the internal dose obtained by the maximum external dose in vivo. This c_maxIVIVE can be used in two ways: (a) to define a relevant upper concentration for the in vitro testing (integrated approach; Fig. 2C) and (b) to interpret the results of an in vitro test, even if the top concentration exceeded the c_maxIVIVE (post hoc approach; Fig. 2B). We describe a workflow which uses toxicokinetic data to derive c_maxIVIVE and interpret the results of in vitro studies in relation to this. Finally, we demonstrate the benefit of using this approach using the herbicide, Dicamba, as an example.

FIG. 2.

Three approaches to set the upper concentrations in in vitro assays: (A) current, (B) post hoc, and (C) integrated approaches. IVIVE, In vivo to in vitro extrapolation.

Proposed Workflow to Set Upper Concentrations for In Vitro Testing

The current approach to set the upper concentration of in vitro testing

Some test methods include predefined maximum concentrations (c_maxDEFAULT; Fig. 2), for example, “10 mM, 2 mg/mL or 2 μL/mL whichever is the lowest” in OECD TG 473.¹⁰ Most in vitro test systems apply the test substance in cell culture media, the exceptions being those using air-liquid interfaces, for example, OECD TG 431.³ Usually, the test substance are completely dissolved as this avoids interference and artifacts often observed in assays using undissolved material, allows coherent concentration settings, and is similar to the form of delivery of test substances to the target cells in vivo (with exceptions, e.g., solid particles^24,25). Hence, the solubility of the test substance in culture media is limiting the upper concentration tested in vitro, c_{maxSOLUBILITY}. Recently, the solubility of test substances used in in vitro methods for the detection of thyroid hormone system disrupting chemicals was determined by measuring the Tyndall effect as an indicator of insoluble suspension or precipitation.²⁶

Sometimes, in vitro testing is performed at concentrations above the aqueous solubility—as a homogenous suspension, for example, if soluble fractions of the test item are of interest. Most noncytotoxicity-based in vitro methods, such as mutagenicity assays, rely on the presence of viable cells and, hence, limit the upper test concentration to avoid high cytotoxicity and ensure a defined fraction of viable cells (e.g., “the highest concentration should aim to achieve 55 ± 5% cytotoxicity” in OECD TG 473¹⁰ or 75% cellular viability in OECD TG 442D⁶). This ensures that enough cells are available for the noncytotoxic cellular response and that cytotoxicity is not interfering with this response. Hence, cytotoxicity limits the upper concentration, c_{maxCYTOTOXICITY}, in in vitro testing. Recommendations for setting test concentrations according to levels of cytotoxicity have been described previously.²⁷

Obviously, the lowest of the three, c_maxDEFAULT, c_{maxSOLUBILITY}, and c_{maxCYTOTOXICITY} limits the upper concentration used in the in vitro assay. While the solubility and cytotoxicity of a test substance are relevant to effects in vivo, their current use is merely to limit upper in vitro concentrations by addressing technical or formal restrictions. Thus, the current approach (Fig. 2A) is not considering the correlation of the in vitro concentration to in vivo doses when setting upper concentrations.

Introducing toxicokinetic data into concentration setting for in vitro testing

The relevance of a concentration tested in an in vitro assay to the in vivo situation in humans or in experimental animals, can be accounted for by additional considerations: The toxicodynamic relevance is determined by defining the AOP which links in vitro effects to in vivo adverse outcomes.²⁸ The toxicokinetic relevance can be established by reverse dosimetry, extrapolating the in vitro concentrations to in vivo doses.^29,30 By applying this approach, a prediction of in vivo toxicity based on in vitro data can be realized.³¹ This is the concept used in next-generation risk assessment (NGRA).^21,32,33 Extrapolations of in vitro to in vivo doses are often only considered after in vitro effects were observed, hence, this is a post hoc approach to take toxicokinetic data into account (Fig. 2B).

Instead of this post hoc approach, toxicokinetic considerations can be directly integrated into the in vitro test (Fig. 2C) by using in vivo doses to set in vitro concentrations. This uses the IVIVE concept to set in vitro concentrations based on in vivo doses and figuring in toxicokinetic information when linking in vitro to in vivo toxicity.

Figure 3 describes the IVIVE to determine upper concentrations for in vitro testing, starting from the upper dose used in in vivo testing (either a predefined limit dose or the MTD). Exposing an organism to an external dose of a substance will result in differing concentrations of this substance in various compartments of the body. The effective concentration is the concentration in the compartment of the toxicological target. Often, the locations of the targets are widely distributed or not entirely identified. Hence, plasma concentration (c_max) is often used in lieu of the actual target in vivo. This, however, is only valid if the effect is concentration-dependent (Fig. 3A), otherwise other dosimetry (e.g., Area Under the Curve [AUC]) would be appropriate, for example, for reactions between test substances and biological targets that are irreversible or when a toxicological target becomes saturated so that equilibrium conditions are not achieved.³⁴

FIG. 3.

IVIVE: Determining the upper concentration for in vitro testing from the highest in vivo dose. The options to calculate the top concentration are denoted by (A–F). AUC, Area Under the Curve; MTD, maximum tolerated dose; PBPK, pharmacokinetic.

If the in vivo study includes toxicokinetic data, that is, plasma concentrations were measured, this can be used to estimate C_maxVIVO (Fig. 3B). If the plasma concentrations are measured after exposure to a dose different from the highest dose in vivo, an extrapolation to the limit dose or MTD may be necessary considering, for example, dose proportionality of plasma concentrations.

If no measured C_maxVIVO is available, it can be calculated by PBTK modeling (Fig. 3D). There are multiple open source and commercial PBTK models of varying complexities available.^35,36 For all input parameters, some assumptions of uncertainties should be considered for improving the final internal dose. Several publications and guidelines are available which describe the building of a PBTK model, and, most importantly, the validation and qualification of the model.^37,38 A simple kinetic model is the “one-compartment model,” whereby the complete body of an animal/human is defined as one compartment by ignoring different tissues/organs. The numbers of compartments can be increased based on the available data sets. In this article, we have used a 1- and 9-compartment-model for the calculations (based on models described previously³⁹ and³³). This was used to demonstrate the approach, but other PBTK models can, of course, also be applied.

The input parameters for PBTK modeling (Fig. 3C), for example, molecular weight, plasma protein binding, logK_ow, Caco-2 P_app, and hepatic clearance in vitro (CL_{intrinisic in vitro}), can be generated using computational tools (Quantitative Structure Activity Relationship [QSAR]) and/or in vitro methods.^40,41 Since hepatic clearance is one of the main drivers of the outcome of the PBTK modeling, availability of in vitro input data is recommended.^42,43

The maximum plasma concentration at the MTD or limit dose (C_maxVIVO) is usually measured or calculated as total concentration (the fraction of the test substance bound to plasma proteins and the unbound fraction, f_ub). However, only the f_ub is available to interact directly with target structures of the target cell (e.g., receptors). If plasma concentrations are compared to in vitro test concentrations, f_ub should be considered for both (Fig. 3E). In plasma, which contains higher protein concentrations than cell culture medium, f_ub is lower than in in vitro test systems as these are using no or—most often—between 5% and 20% serum in the cell culture medium.

Like any model, the results of this approach do not perfectly predict actual concentrations in plasma. Several studies compared plasma concentrations predicted with kinetic models to measured concentrations.^33,44,44 investigated 349 substances and found the concentrations of 102 to be overestimated, 140 to be in the same order of magnitude, and 6 were underestimated (and 99 substances which could not be assessed either because the protein binding assay failed and f_ub is an important predictor or the substances did not reach steady state). Only two substances were underestimated by a factor more than 10. They proposed a framework for toxicokinetic triage that can group substances into categories associated with varying levels of confidence in kinetic predictions. Obviously, IVIVE and setting upper concentrations in vitro based on IVIVE has some uncertainties which should be discussed and considered (proposed workflow; Fig. 3F). Uncertainty considerations for elements of this workflow are available.^27,44–48

Considering toxicokinetic data when interpreting results of in vitro tests

In vitro tests provide results usually as concentration–response relations. Concentrations causing a certain response are used to interpret the results. A concentration causing a pre-defined effect (e.g., AC₅₀ in ToxCast or x-fold statistically significant or benchmark concentration)²⁷ is used to draw a conclusion. This can be done by a predefined data interpretation procedure (DIP).^49,50,^* As an example the DIP for the LuSens assay (although a study addressing local effects and hence not immediately relevant for this workflow)—an in vitro assay that addresses mechanisms described under the second Key Event of the AOP for skin sensitization, namely keratinocyte activation—is “the luciferase induction of ≥1.5-fold and is statistically significant compared to the solvent control in at least 2 consecutive noncytotoxic tested concentrations (i.e., cellular viability is ≥70%)” (OECD TG 442D⁶).

Besides a prescribed DIP, the effect-, no effect-, or benchmark-concentration derived from an in vitro assay is used as such to define a point of departure (POD_vitro) for further interpretations and assessments (e.g., comparing to exposure levels in NGRA⁵¹). Moreover, in regulatory testing schemes, this may be a trigger for further testing,⁵² including in vivo tests. If, however, POD_vitro is greater than C_maxIVIVE (Fig. 4), this is not warranted. The corresponding plasma concentration in the in vivo study, C_maxVIVO, would not be reached by the highest dose tested in vivo. Hence, the test substance cannot cause the cellular response in vivo which would lead to the corresponding adverse effect. Then the in vivo study is not capable of following-up on the in vitro effect as it will show no effect.

FIG. 4.

Post hoc evaluation of the relevance of effects observed in vitro for subsequent in vivo testing.

If the top dose in vivo was limited by the MTD, a cellular effect observed in vitro at a concentration exceeding c_maxIVIVE could cause an effect in vivo at a dose exceeding the MTD. Despite the ongoing discussion on defining the MTD,^53,54 there is a consensus that testing above a (however defined) MTD is not sensible. If the top dose in vivo was limited by a default described in a test guideline (“limit dose”), this limit dose often correlates with the criteria for hazard classification.^† In this case, a cellular effect in vitro at a concentration exceeding c_maxIVIVE, would only occur in vivo at a dose that is not triggering a hazard classification.

A Preliminary Case Study Applying This Approach

We used the example of Dicamba and its testing in repro-toxicological studies in rats in vivo and in an in vitro endocrine assay (steroidogenesis assay following OCED TG 456⁵⁵). Here, we illustrate how the workflow can be conducted using (1) an integrated approach, and (2) a post hoc approach by applying PBTK-modeling and additionally experimental biokinetic data.

For Dicamba, the highest doses that were tested in repro-toxicological studies in rats were 390 and 362 mg/kg (two generation study following OECD TG 416⁵⁶ and developmental toxicity study following OECD TG 414¹⁹). The limited high doses tested are justified by clinical signs of toxicity, increased liver weights, and reduced feed consumption correlated to body weight loss in the two-generation study, as well as by mortality of individual pregnant animals in the developmental toxicity study.⁵⁷ The results of the steroidogenesis assay following OCED TG 456 showed an estrogenic effect at the highest concentration tested (i.e., 1000 μM) only.

For the integrated approach, the highest testable dose in vivo was set to 390 mg/(kg·d), and this dose was converted to maximum free plasma concentrations of Dicamba. The post hoc approach takes the highest test concentration of 1000 μM in vitro into account that is in accordance with the recommendation for the highest concentration to be tested given in OECD TG 456 and converts this concentration by reverse dosimetry to correlated dose levels.

Physicochemical parameters for Dicamba were obtained from PubChem.⁵⁸ Kinetic modeling for Dicamba was conducted using a 1- and 9-compartment model for which the principles have been described by Chang et al.³⁹ and Fabian et al.³³ The substance-specific input parameters were molecular weight = 221.0 g/mol, fraction unbound in plasma (F_ub) = 10.6% (calculated using an in-house QSAR), log k_ow = −0.55 and an assumed worst-case scenario with a CL_int = 0 μL/min*million hepatocytes. Experimental data on Papp values from Caco-2 or other systems were not used in this case study but were calculated from internally established QSARs and applied for the 9-compartment model.

The 1-compartment model-specific parameters were dose = 390 mg/kg, renal clearance included a glomerular filtration rate of 0.088 L/h, a typical value for the rat in accordance with Katayama et al.,⁵⁹ Sadick et al.,⁶⁰ and Wetmore et al.⁶¹ The 9-compartment model-specific parameters were the same as for the 1-compartment except for the timeframe for the single dose, which was 80 hours and the timeframe and dosing interval for the repeated dose simulation, which were 500 and 24 hours, respectively. The Papp value was calculated by QSAR and resulted in 25.16 × 10^–6 cm/s. The free steady-state concentration (c_ss,u) was calculated for the 1-compartment model and the free maximum concentrations (c_max,u) were calculated for the 9-compartment model. Since the protein concentration in cell culture medium is significantly lower than in plasma, protein binding was neglected for Dicamba in the in vitro steroidogenesis assay.

Integrated approach

The upper concentration for in vitro testing can be set as described in Figure 3. According to the workflow, the first step is to use a kinetic model for the rat to predict total c_ss and total c_max for Dicamba at the dose of 390 mg/kg. In the modeled plasma concentration, F_up was incorporated to derive an upper concentration for in vitro testing. For the purposes of this example, an additional safety factor was not applied. The resulting free concentrations are shown in Table 1.

Table 1.

In Vivo to In Vitro Extrapolation Based In Vitro Concentrations Derived from the Highest Dose of 390 mg/kg Tested in Repro-Toxicological Studies

	Conversion of 390 mg/kg to unbound plasma concentrations
1-compartment model	C_ss,u = 208.5 μM
9-compartment model single dose	C_max,u = 312.9 μM
9-compartment model repeated dose	C_max,u = 375.6 μM
Experimental data^a	C_max,u = 155.9 μM

Effect concentration in the estrogenic in vitro assay was 1000 μM.

Leibold et al. (1998),⁶² c_max in male rats after 14 days pretreatment with Dicamba in feed (4500 ppm) and 1 single oral dose of ¹⁴C-Dicamba by gavage at a dose level of 400 mg/kg, corrected by F_up (BASF study code 02B0266/976009).

The solubility of Dicamba in water is 8310 mg/L at 25°C (equivalent to 37 mM), which is much higher than these concentrations (which are in the range of 155.9–375.6 μM). Therefore, in this example, the upper concentration for the in vitro test could be set at the upper concentration (conservatively using the highest value from the 9-compartment PBPK model) to that is, 375.6 μM. Additionally, experimental kinetic data were available that yielded a plasma concentration in rats after a dose of 400 mg/kg b.w. as shown in Table 1. The measured plasma concentration was corrected by F_up to yield the free concentration of Dicamba in plasma. The experimental value confirmed the results of the kinetic models and showed that these slightly overestimated the plasma concentration but were able to predict the experimental value within a factor of three.

Post hoc approach

For the post hoc approach, the results of the steroidogenesis assay were assessed for their relevance to an in vivo external dose. In the in vitro assay, an estrogenic effect was only observed at 1000 μM. In a first step, the 9-compartment model was used to evaluate whether Dicamba would accumulate by comparing c_max,u after a single and repeated oral doses of 390 mg/kg b.w. in the rat. The results indicated no accumulation in kinetic modeling, since the calculated half-life was 8.9 hours and, consequently, plasma concentrations were comparable after single and repeated doses (313.0 vs. 375.6 μM).

When the in vitro estrogenic effect at 1000 μM was converted to external doses, using the 1- or 9-compartmental models, the external doses of 1871 and 1038 mg/kg were above the limit dose of 1000 mg/kg. and were more than 2.5-fold higher than the highest testable dose in vivo (390 mg/kg; Table 2). When the available experimental kinetic data were applied to reverse dosimetry, the measured plasma concentration, corrected by F_up yields an equivalent dose of 2566 mg/kg b.w. This calculated dose confirms the overprediction of the kinetic models and underlines the irrelevance of the obtained estrogenic effect at 1000 μM within the objective of the presented case study.

Table 2.

In Vivo to In Vitro Extrapolation Based Dose Levels of a Repeated Dose Toxicity Study Derived from an In Vitro Estrogenicity Assay with an Effect Concentration of 1000 μM

	Reverse dosimetry: Conversion of in vitro concentration (1000 μM) to in vivo dose in mg/kg
1-compartment model	1871
9-compartment model Single dose	1246
9-compartment model repeated dose	1038
Experimental data^a	2566

The actual highest dose tested in the developmental toxicity study is 390 mg/kg body weight.

Leibold et al., 1998,⁶² c_max in male rats after 14 days pretreatment with Dicamba in feed (4500 ppm) and 1 single oral dose of ¹⁴C-Dicamba by gavage at a dose level of 400 mg/kg corrected by F_up (BASF study code 02B0266/976009).

Discussion

We propose to consider the internal concentration of substances tested in in vivo studies when setting their concentrations for in vitro studies. Therefore, we have developed a workflow to derive the corresponding concentration, c_maxIVIVE, and two approaches to either integrate the upper concentration setting before the in vitro assay is conducted (integrated approach) or to put the results of an in vitro assay into context when the upper concentration was not limited for the conduct of the study (post hoc approach).

With the integrated approach, the IVIVE-derived upper concentration, c_maxIVIVE, is applied as the highest concentration in the in vitro test if this concentration is lower than the upper concentrations derived from cytotoxicity or solubility estimations. This approach is similar to that developed for nanoparticles, which are often tested in much higher concentrations in in vitro assays than are present in vivo.²³ Ma-Hock et al.²³ proposed a 6-step reverse dosimetry process to convert an in vivo dose to an in vitro dose, considering organ burden for the extrapolation as well as uncertainties and specificities of the in vitro test system.

Currently upper doses for regulatory in vivo testing are set by using prior information—often the results of a so-called “dose range” finding study or a limit dose. The analogue approach for setting in vitro concentrations would be the “integrated approach”—using IVIVE to set the upper in vitro concentration. This approach avoids in vitro results (especially positive in vitro results) at concentrations which are higher than the internal concentrations in vivo, for example, at dose levels above the MTD.

There are, however, differences between in vivo and in vitro testing which needs to be considered: While in vivo testing is limited to use as little animals as possible, in vitro testing can afford testing multiple concentrations and repeating studies: While in vivo testing is usually restricted to three dose groups and setting doses in the relevant range is crucial, testing an additional, higher concentration in vitro would not take away a test concentration in the lower, supposedly more relevant concentration range. Moreover, the “integrated approach” in vitro refers to an MTD (if no general limit dose is used) observed in a specific in vivo study with the very dose setting, test substance application, and animal strain used in this study.

Different studies could result in different MTDs (i.e., when using a different study design or a different animal strain, e.g., following an updated test guideline). In such cases, the highest concentration selected in the in vitro assay based on the “integrated approach” may not correlate to the new in vivo results and may be too low. This would require repeating the in vitro study with higher concentrations. Apart from the potential change of the correlating in vivo dose, there are more uncertainties which are concerning the in vitro model and the IVIVE. The general assumption is that an effect concentration in vitro is relevant for the in vivo situation. Certain in vitro models may, however, be more or less sensitive than the in vivo target by comparison of Css or Cmax in vivo to the total or free concentrations in vitro.

Finally, the estimations of in vitro concentrations correlating to in vivo doses using PBTK models without all relevant transport- and metabolism parameters may err by more than an order of magnitude.³³ All or parts of this may be unknown by the time of the in vitro testing. If upper concentrations were limited according to the “integrated approach,” repeated in vitro testing with higher concentrations may be necessary once more information becomes available.

In the example shown with Dicamba, when the in vitro estrogenic effect at 1000 μM was converted to an external dose, the latter was more than twofold higher than the maximal tolerated dose in vivo. This example shows that the in vitro effects were observed at concentrations corresponding to external doses markedly exceeding what could be tolerated by the rats. Since the estrogenic effect in vitro was only observed at the highest concentration, it indicates that the key event of an AOP observed in vitro could not happen in vivo and hence cannot proceed to the adverse outcome; no corresponding adverse effect will be observed in vivo despite the positive in vitro assay. Often, in vitro tests are used as screening or tier one studies and a positive in vitro test will trigger in vivo testing. It should be obvious that this trigger is not relevant, and an in vivo test will not be appropriate, if the effect observed in vitro occurs at concentrations which cannot be achieved in vivo due to low internal doses. Rather, the bioavailability of the test substance at the in vivo target is too low for this effect and a given top concentration in vivo.

This could be avoided by using a “post hoc-approach” which uses the highest feasible concentrations. This would cover concentrations which are correlating to all in vivo doses and are testable in the in vitro assay. The post hoc approach uses IVIVE to retrospectively determine the relevance of the assay result.

There are several considerations when comparing data from in vivo models with those of in vitro models. Dose and concentration are different dose metrics,^34,63 but often, the plasma concentration (or its integral over time, the area under the time-concentration curve, AUC) are used to describe the internal exposure of organisms and are discussed in general as an internal dose. Potential bioaccumulation of the test substance should be considered; the PBTK modeling should simulate single oral dosing versus repeated oral dosing and assess the respective plasma concentrations over time. If a bioaccumulation potential is determined after multiple dosing, the 1-compartment model should not be used in the proposed workflow because it does not include a fat compartment, in which lipophilic compounds are accumulated. Instead, PBTK modeling with multiple dosing is recommended to reach a steady-state concentration after the corresponding repeated dosing regimen. For bioaccumulative compounds, this steady concentration may be set—as a worst-case scenario—to C_maxIVIVE.

In vivo studies are performed according to different endpoint measurements and therefore vary with respect to the species, exposure route and frequency, duration, and nominal dose levels. All these factors impact the kinetic behavior of the substance and must be considered in the calculations of the internal dose. If information on internal concentrations of a substance is available from in vivo studies, it should be kept in mind that biokinetic information from these are usually based on plasma levels, which do not necessarily correlate with concentrations in the target cells. Hence, cellular concentrations may not directly correlate with external exposures—even if the biokinetic extrapolation is accurate.

It is also important to consider that in vitro models are often limited in their metabolic capacity and therefore only provide information on the parent compound, while being agnostic to metabolites which could have other toxic properties or may be more potent than the parent. This limitation may be addressed by tiered NGRA-approaches, in which the formation of toxic metabolites is quantified and corresponding bioactivity is included in the IVIVE.⁶⁴ Detoxification is, however, considered if hepatic clearance is included in the IVIVE.

In the current workflow, the toxicokinetics of the test substance are incorporated to link in vitro to in vivo and to optimize the design and the concentration setting of in vitro experiments. While substance toxicokinetics are the key for the principles proposed here, time dependency of these processes is not considered and a time-independent, simplified approach using the steady-state concentration from the 1-compartment model or the maximum plasma concentration from the 9-compartment model is recommended.

An important consideration when extrapolating in vitro concentrations to in vivo external doses is that the actual effective concentration of a test substance in a cell in an assay may deviate from the nominal concentration in a time-dependent manner. Effective concentrations can be affected by several processes, such as plastic binding and evaporation (artifacts due to the in vitro test system) and binding to cell culture media proteins and the cell membrane (qualitatively correlating with respect to in vivo processes, but not necessarily correlating quantitatively).^65–67 Hence, the actual concentration of a test substance at the cellular target can differ significantly from the nominal concentration added to the test system and is currently increasingly addressed in so-called “in vitro dosimetry approaches.” Obviously, the actual internal concentration at the target in vivo and thus the upper concentration selected for in vitro testing are not exactly known (therefore, the uncertainties described above should be considered).

Application of in vitro methods for hazard assessment requires knowledge about the in vivo relevance of findings. This has recently been used for genotoxicity data.⁶⁸ The hazard of a substance as an intrinsic property to cause harm is not only characterized by its bioactivity, which can be reliably assessed applying in vitro methods, but also by its ADME parameters, that predetermine the internal exposure at a given dose in vivo. As the latter are not frequently addressed in in vitro studies, we propose a workflow involving PBTK modeling to ensure that in vitro findings are of relevance for hazard identification. By applying an integrated or post hoc approach, our strategy aims to avoid or identify in vitro concentrations not achievable in vivo, thereby guiding in vitro hazard identification. This approach will also reduce the number of unnecessary follow-up animal studies and avoid removing technically promising chemicals from innovation pipelines by enhancing the relevance of in vitro results.

Footnotes

Acknowledgments

We would like to thank Matthew Dent (Unilever), Sue Marty (Dow), Steffi Melching-Kolmus and Christiane Wieman (BASF Agricultural Solutions) for their helpful input to this manuscript.

Authors' Contributions

R.L.: Conceptualization, resources, writing—review and editing. E.F.: Methodology, software, formal analysis, writing—review and editing. B.B.: Methodology, writing—review and editing. P.D., H.M.H., and J.S.: Writing—review and editing. N.J.H.: Writing original draft, review and editing.

Author Disclosure Statement

The authors have no conflicts of interest.

Funding Information

No funding was involved in this study. Nicola Hewitt received financial support from BASF.

References

OECD. Test No. 439: In Vitro Skin Irritation. 2010. Available from: https://www.oecd-ilibrary.org/content/publication/9789264090958-en [Last accessed: June 11, 2021].

OECD. Test No. 435: In Vitro Membrane Barrier Test Method for Skin Corrosion. 2015. Available from: https://www.oecd-ilibrary.org/content/publication/9789264242791-en [Last accessed: June 11, 2021].

OECD. Test No. 431: In Vitro Skin Corrosion: Human Skin Model Test. 2004. Available from: https://www.oecd-ilibrary.org/content/publication/9789264071148-en [Last accessed: June 11, 2021].

OECD. Test No. 492: Reconstructed human Cornea-like Epithelium (RhCE) test method for identifying chemicals not requiring classification and labelling for eye irritation or serious eye damage. 2023. Available from: https://www.oecd-ilibrary.org/content/publication/9789264242548-en [Last accessed: November 1, 2023].

OECD. Test No. 442E: In Vitro Skin Sensitisation. 2023. Available from: https://www.oecd-ilibrary.org/content/publication/9789264264359-en [Last accessed: November 1, 2023].

OECD. Test No. 442D: In Vitro Skin Sensitisation. 2022. Available from: https://www.oecd-ilibrary.org/content/publication/9789264229822-en [Last accessed: November 1, 2023].

OECD. Test No. 442C: In Chemico Skin Sensitisation. 2023. Available from: https://www.oecd-ilibrary.org/content/publication/9789264229709-en [Last accessed: November 1, 2023].

OECD. Revised Guidance Document 150 on Standardised Test Guidelines for Evaluating Chemicals for Endocrine Disruption. 2018. Available from: https://www.oecd-ilibrary.org/content/publication/9789264304741-en [Last accessed: June 11, 2021].

OECD. Test No. 444A: In Vitro Immunotoxicity. 2023. Available from: https://www.oecd-ilibrary.org/content/publication/27b10ba3-en [Last accessed: November 1, 2023].

10.

OECD. Test No. 473: In Vitro Mammalian Chromosomal Aberration Test. 2016. Available from: https://www.oecd-ilibrary.org/content/publication/9789264264649-en [Last accessed: June 11, 2021].

11.

OECD. Test No. 476: In Vitro Mammalian Cell Gene Mutation Test. 1997. Available from: https://www.oecd-ilibrary.org/content/publication/9789264071322-en [Last accessed: June 11, 2021].

12.

OECD. Test No. 487: In Vitro Mammalian Cell Micronucleus Test. 2023. Available from: https://www.oecd-ilibrary.org/content/publication/9789264264861-en [Last accessed: November 1, 2023].

13.

OECD. Test No. 417: Toxicokinetics. 2010. Available from: https://www.oecd-ilibrary.org/content/publication/9789264070882-en [Last accessed: June 11, 2021].

14.

OECD. Test No. 474: Mammalian Erythrocyte Micronucleus Test. 2016. Available from: https://www.oecd-ilibrary.org/content/publication/9789264264762-en [Last accessed: June 11, 2021].

15.

EC. Regulation (EU) No 1107/2009 of the European Parliament and of the Council of 21 October 2009 concerning the placing of plant protection products on the market and repealing Council Directives 79/117/EEC and 91/414/EEC. 2009.

16.

Escher

, Kamp

, Bennekou

, et al. Towards grouping concepts based on new approach methodologies in chemical hazard assessment: the read-across approach of the EU-ToxRisk project. Arch Toxicol, 2019; 93(12):3643–3667; doi: 10.1007/s00204-019-02591-7

17.

Parish

, Aschner

, Casey

, et al. An evaluation framework for new approach methodologies (NAMs) for human health safety assessment. Regul Toxicol Pharmacol, 2020; 112:104592; doi: 10.1016/j.yrtph.2020.104592

18.

Niemann

, Choi

, Kneuer

, et al. Traditional and novel approaches to derive health-based guidance values for pesticides. Curr Opin Food Sci, 2023; 54:101091; doi: 10.1016/j.cofs.2023.101091

19.

OECD. Test No. 414: Prenatal Developmental Toxicity Study. 2018. Available from: https://www.oecd-ilibrary.org/content/publication/9789264070820-en [Last accessed: June 11, 2021].

20.

Judson

, Houck

, Martin

, et al. Editor's highlight: Analysis of the effects of cell stress and cytotoxicity on in vitro assay activity across a diverse chemical and assay space. Toxicol Sci, 2016; 152(2):323–339; doi: 10.1093/toxsci/kfw092

21.

Dent

, Vaillancourt

, Thomas

, et al. Paving the way for application of next generation risk assessment to safety decision-making for cosmetic ingredients. Regul Toxicol Pharmacol, 2021; 125:105026; doi: 10.1016/j.yrtph.2021.105026

22.

Ouedraogo

, Alexander-White

, Bury

, et al. Read-across and new approach methodologies applied in a 10-step framework for cosmetics safety assessment—A case study with parabens. Regul Toxicol Pharmacol, 2022; 132:105161; doi: 10.1016/j.yrtph.2022.105161

23.

Ma-Hock

, Sauer

, Ruggiero

, et al. The use of nanomaterial in vivo organ burden data for in vitro dose setting. Small, 2021; 17(15):e2005725; doi: 10.1002/smll.202005725

24.

Schulze

, Schulze

, Kroll

, et al. Not ready to use—Overcoming pitfalls when dispersing nanoparticles in physiological media. Nanotoxicology, 2008; 2(2):51–61; doi: 10.1080/17435390802018378

25.

DeLoid

, Cohen

, Darrah

, et al. Estimating the effective density of engineered nanomaterials for in vitro dosimetry. Nat Commun, 2014; 5(1):3514; doi: 10.1038/ncomms4514

26.

Cole

, Fortaner Torrent

. Solubility Determination of Chemicals by Nephelometry, editor(s), EUR 31447 EN, Publications Office of the European Union, Luxembourg, JRC132585: 2023.

27.

Krebs

, Nyffeler

, Karreman

, et al. Determination of benchmark concentrations and their statistical uncertainty for cytotoxicity test data and functional in vitro assays. Altex, 2020; 37(1):155–163; doi: 10.14573/altex.1912021

28.

Vinken

. The adverse outcome pathway concept: A pragmatic tool in toxicology. Toxicology, 2013; 312:158–165; doi: 10.1016/j.tox.2013.08.011

29.

DeJongh

, Forsby

, Houston

, et al. An integrated approach to the prediction of systemic toxicity using computer-based biokinetic models and biological in vitro test methods: Overview of a Prevalidation Study Based on the ECITTS11ECITTS = ERGATT/CFN Integrated Toxicity Testing Scheme; ERGATT = European Research Group for Alternatives in Toxicity Testing; CFN = Swedish Board for Laboratory Animals. Project Toxicol In Vitro, 1999; 13(4):549–554; doi: 10.1016/S0887-2333(99)00030-2

30.

Blaauboer

. The necessity of biokinetic information in the interpretation of in vitro toxicity data. Alternat Lab Anim, 2002; 30(2 Suppl.):85–91; doi: 10.1177/026119290203002s13

31.

Louisse

, Beekmann

, Rietjens

IMCM

. Use of physiologically based kinetic modeling-based reverse dosimetry to predict in vivo toxicity from in vitro data. Chem Res Toxicol, 2017; 30(1):114–125; doi: 10.1021/acs.chemrestox.6b00302

32.

, Yuan

, Middleton

, et al. Next generation risk assessment (NGRA): Bridging in vitro points-of-departure to human safety assessment using physiologically-based kinetic (PBK) modelling—A case study of doxorubicin with dose metrics considerations. Toxicol In Vitro, 2021; 74:105171; doi: 10.1016/j.tiv.2021.105171

33.

Fabian

, Gomes

, Birk

, et al. In vitro-to-in vivo extrapolation (IVIVE) by PBTK modeling for animal-free risk assessment approaches of potential endocrine-disrupting compounds. Arch Toxicol, 2019; 93(2):401–416; doi: 10.1007/s00204-018-2372-z

34.

Kisitu

, Hougaard Bennekou

, Leist

. Chemical concentrations in cell culture compartments (C5)—Concentration definitions. Altex, 2019; 36(1):154–160; doi: 10.14573/altex.1901031

35.

Sager

, Yu

, Ragueneau-Majlessi

, et al. Physiologically based pharmacokinetic (PBPK) modeling and simulation approaches: A systematic review of published models, applications, and model verification. Drug Metab Dispos, 2015; 43(11):1823–1837; doi: 10.1124/dmd.115.065920

36.

Upton

, Foster

, Abuhelwa

. An introduction to physiologically-based pharmacokinetic models. Paediatr Anaesth, 2016; 26(11):1036–1046; doi: 10.1111/pan.12995

37.

OECD. Guidance document on the characterisation, validation and reporting of Physiologically Based Kinetic (PBK) models for regulatory purposes, OECD Series on Testing and Assessment, No. 331, Environment, Health and Safety, Environment Directorate, OECD; 2021.

38.

WHO. Project

, No

. Characterization and Application of Physiologically Based Pharmacokinetic Models. IPCS—WHO. Available from: https://apps.who.int/iris/bitstream/handle/10665/44495/9789241500906_eng.pdf. 2010 [Last accessed: June 11, 2021].

39.

Chang

, Kleinstreuer

, Ceger

, et al. Application of reverse dosimetry to compare in vitro and in vivo estrogen receptor activity. Appl In Vitro Toxicol, 2015; 1:33–44; doi: 10.1089/aivt.2014.0005

40.

Bessems

, Loizou

, Krishnan

, et al. PBTK modelling platforms and parameter estimation tools to enable animal-free risk assessment: Recommendations from a joint EPAA—EURL ECVAM ADME workshop. Regul Toxicol Pharmacol, 2014; 68(1):119–139; doi: 10.1016/j.yrtph.2013.11.008

41.

Coecke

, Pelkonen

, Leite

, et al. Toxicokinetics as a key to the integrated toxicity risk assessment based primarily on non-animal approaches. Toxicol In Vitro, 2013; 27(5):1570–1577; doi: 10.1016/j.tiv.2012.06.012

42.

Punt

, Louisse

, Beekmann

, et al. Predictive performance of next generation human physiologically based kinetic (PBK) models based on in vitro and in silico input data. Altex, 2022; 39(2):221–234; doi: 10.14573/altex.2108301

43.

Punt

, Louisse

, Pinckaers

, et al. Predictive performance of next generation physiologically based kinetic (PBK) model predictions in rats based on in vitro and in silico input data. Toxicol Sci, 2022; 186(1):18–28; doi: 10.1093/toxsci/kfab150

44.

Wambaugh

, Wetmore

, Pearce

, et al. Toxicokinetic triage for environmental chemicals. Toxicol Sci, 2015; 147(1):55–67; doi: 10.1093/toxsci/kfv118

45.

Paul Friedman

, Foster

, Pham

, et al. Reproducibility of organ-level effects in repeat dose animal studies. Comput Toxicol, 2023; 28:100287; doi: 10.1016/j.comtox.2023.100287

46.

Pham

, Watford

, Pradeep

, et al. Variability in in vivo studies: Defining the upper limit of performance for predictions of systemic effect levels. Comput Toxicol, 2020; 15:100126; doi: 10.1016/j.comtox.2020.100126

47.

Gabbert

, Mathea

, Kolle

, et al. Accounting for precision uncertainty of toxicity testing: Methods to define borderline ranges and implications for hazard assessment of chemicals. Risk Analysis, 2022; 42(2):224–238; doi: 10.1111/risa.13648

48.

Maertens

, Golden

, Luechtefeld

, et al. Probabilistic risk assessment—The keystone for the future of toxicology. ALTEX Altern Animal Exp, 2022; 39(1):3–29; doi: 10.14573/altex.2201081

49.

Sauer

, Hill

, Curren

, et al. Local tolerance testing under REACH: Accepted non-animal methods are not on equal footing with animal tests. Altern Lab Anim, 2016; 44(3):281–299; doi: 10.1177/026119291604400311

50.

Caloni

, De Angelis

, Hartung

. Replacement of animal testing by integrated approaches to testing and assessment (IATA): A call for in vivitrosi. Arch Toxicol, 2022; 96(7):1935–1950; doi: 10.1007/s00204-022-03299-x

51.

European Chemical

Agency

, Centre

EFSAwttsotJR

, Andersson

, et al. Guidance for the identification of endocrine disruptors in the context of Regulations (EU) No 528/2012 and (EC) No 1107/2009. EFSA J, 2018; 16(6):e05311; doi: 10.2903/j.efsa.2018.5311

52.

Baltazar

, Cable

, Carmichael

, et al. A next-generation risk assessment case study for coumarin in cosmetic products. Toxicol Sci, 2020; 176(1):236–252; doi: 10.1093/toxsci/kfaa048

53.

van Berlo

, Woutersen

, Muller

, et al. 10% Body weight (gain) change as criterion for the maximum tolerated dose: A critical analysis. Regul Toxicol Pharmacol, 2022; 134:105235; doi: 10.1016/j.yrtph.2022.105235

54.

Arts

JHE

, Humfrey

, Sloter

. Letter to the Editors regarding “10% body weight (gain) change as criterion for the maximum tolerated dose: A critical analysis.”. Regul Toxicol Pharmacol, 2023; 143:105440; doi: 10.1016/j.yrtph.2023.105440

55.

OECD. Test No. 456: H295R Steroidogenesis Assay. 2023. Available from: https://www.oecd-ilibrary.org/content/publication/9789264122642-en [Last accessed: November 1, 2023].

56.

OECD. Test No. 416: Two-Generation Reproduction Toxicity. 2001. Available from: https://www.oecd-ilibrary.org/content/publication/9789264070868-en [Last accessed: June 11, 2021].

57.

EFSA. Public consultation on the active substance Dicamba. 2021. Available from: https://www.efsa.europa.eu/en/call/public-consultation-active-substance-dicamba [Last accessed: November 1, 2023].

58.

PubChem Database. Dicamba. Available from: https://pubchem.ncbi.nlm.nih.gov/compound/3030 [Last accessed: February 1, 2022).

59.

Katayama

, Yamaguchi

, Yamashita

, et al. Calculation of glomerular filtration rate in conscious rats by the use of a bolus injection of iodixanol and a single blood sample. J Pharmacol Toxicol Methods, 2010; 61(1):59–64; doi: 10.1016/j.vascn.2009.10.002

60.

Sadick

, Attenberger

, Kraenzlin

, et al. Two non-invasive GFR-estimation methods in rat models of polycystic kidney disease: 3.0 Tesla dynamic contrast-enhanced MRI and optical imaging. Nephrol Dial Transplant, 2011; 26(10):3101–3108; doi: 10.1093/ndt/gfr148

61.

Wetmore

, Wambaugh

, Ferguson

, et al. Relative impact of incorporating pharmacokinetics on predicting in vivo hazard and mode of action from high-throughput in vitro toxicity assays. Toxicol Sci, 2013; 132(2):327–346; doi: 10.1093/toxsci/kft012

62.

Leibold

. ¹⁴C-Dicamba—Study of the plasma kinetics in rats. Unpublished report No. 1998/10553 from BASF AG, Ludwigshafen/Rhein, Germany. Submitted to WHO by BASF Corporation. 1998. WHO report available at https://apps.who.int/pesticide-residues-jmpr-database/Document/86

63.

Kisitu

, Hollert

, Fisher

, et al. Chemical concentrations in cell culture compartments (C5)—Free concentrations. Altex, 2020; 37(4):693–708; doi: 10.14573/altex.2008251

64.

van Tongeren

TCA

, Carmichael

, Rietjens

, et al. Next generation risk assessment of the anti-androgen flutamide including the contribution of its active metabolite hydroxyflutamide. Front Toxicol, 2022; 4:881235; doi: 10.3389/ftox.2022.881235

65.

Armitage

, Wania

, Arnot

. Application of mass balance models and the chemical activity concept to facilitate the use of in vitro toxicity data for risk assessment. Environ Sci Technol, 2014; 48(16):9770–9779; doi: 10.1021/es501955g

66.

Kramer

, Di Consiglio

, Blaauboer

, et al. Biokinetics in repeated-dosing in vitro drug toxicity studies. Toxicol In Vitro, 2015; 30(1 Pt A):217–224; doi: 10.1016/j.tiv.2015.09.005

67.

Kramer

, Krismartina

, Rico-Rico

, et al. Quantifying processes determining the free concentration of phenanthrene in Basal cytotoxicity assays. Chem Res Toxicol, 2012; 25(2):436–445; doi: 10.1021/tx200479k

68.

Beal

, Audebert

, Barton-Maclaren

, et al. Quantitative in vitro to in vivo extrapolation of genotoxicity data provides protective estimates of in vivo dose. Environ Mol Mutagen, 2023; 64(2):105–122; doi: 10.1002/em.22521