Elsevier

Toxicology Letters

Volume 220, Issue 1, 20 June 2013, Pages 26-34
Toxicology Letters

Prediction of dose-hepatotoxic response in humans based on toxicokinetic/toxicodynamic modeling with or without in vivo data: A case study with acetaminophen

https://doi.org/10.1016/j.toxlet.2013.03.032Get rights and content

Highlights

  • We built a dynamic model to account for effects on liver cells throughout time.

  • A relevant kinetic model for acetaminophen was calibrated based on non-animal data.

  • The predicted acetaminophen dose–response was relevant to human data.

  • The methods to predict metabolism and elimination should be improved.

Abstract

In the present legislations, the use of methods alternative to animal testing is explicitly encouraged, to use animal testing only ‘as a last resort’ or to ban it. The use of alternative methods to replace kinetics or repeated dose in vivo tests is a challenging issue. We propose here a strategy based on in vitro tests and QSAR (Quantitative Structure Activity Relationship) models to calibrate a dose–response model predicting hepatotoxicity. The dose response consists in calibrating and coupling a PBPK (physiologically-based pharmacokinetic) model with a toxicodynamic model for cell viability. We applied our strategy to acetaminophen and compared three different ways to calibrate the PBPK model: only with in vitro and in silico methods, using rat data or using all available data including data on humans. Some estimates of kinetic parameters differed substantially among the three calibration processes, but, at the end, the three models were quite comparable in terms of liver toxicity predictions and close to the usual range of human overdose. For the model based on alternative methods, the good adequation with the two other models resulted from an overestimated renal elimination rate which compensated for the underestimation of the metabolism rate. Our study points out that toxicokinetics/toxicodynamics approaches, based on alternative methods and modelling only, can predict in vivo liver toxicity with accuracy comparable to in vivo methods.

Introduction

The promotion of alternative approaches to replace, reduce or refine animal use is present in all relevant EU legislations. For chemicals in general, the REACH legislation requires industry to evaluate the toxicity of new and existing chemicals, even those that are in use but have never been subject to regulatory testing, which translates into many thousands of chemicals to be tested. To fulfil this objective, the regulation explicitly encourages the use of methods alternative to animal testing, which would only be used ‘as a last resort’. As far as cosmetics are concerned, the 7th Amendment to the Cosmetic Directive (76/768/EEC) was adopted in 2003. According to this directive, animal testing in the EU for cosmetic products had to stop immediately and a complete ban of animal testing for ingredients was due to enter into force by 11 March 2009. The deadline for most complex endpoints such as repeated dose toxicity, reproductive toxicity and toxicokinetics was set for the 11th of March 2013 but the directive can also accommodate an extension if alternative and validated methods are not available within this time framework. In parallel to the set up of a new legislation, considerable efforts, in particular from EU, were made to develop suitable and appropriate alternative methods. For instance, between 2003 and 2010, European Commission funded research programmes related to such development through FP6 and FP7 programmes for around 150 million euros (Adler et al., 2011).

Most of the toxicological information for a chemical obtained through in vitro testing provides information on the dose response relationship at target cell or, more rarely, at target organ level. Future risk assessment is expected to focus on avoiding significant perturbations in toxicity pathways through in vitro assays on human cells or cell lines (Krewski et al., 2010). To integrate the in vitro results into quantitative in vivo risk assessment, it is thus necessary to relate the toxicokinetics (TK) of the chemical in the body and the toxicodynamics at each target level (Andersen and Krewski, 2010, Adler et al., 2011, Louisse et al., 2012, Yoon et al., 2012). This can be achieved through Physiologically Based Toxicokinetics or Pharmacokinetics models – PBTK or PBPK (Clewell et al., 2008). A PBTK model consists of a series of mathematical equations which, based on the specific physiology of an organism and on the physicochemical properties of a substance, are able to describe the absorption, distribution, metabolism and elimination (ADME) of the compound within this organism. The solution of these equations provides the time-course of the parent compound and its metabolites in the organs and allows for a sound mechanistic description of the kinetic processes including accumulation in tissues.

There are only a few examples for which PBPK models were used to extrapolate in vitro response to in vivo hazard assessment. Rotroff et al. (2010) combined a simple PBPK model with in vitro bioactivity data to rank chemicals according to actual human oral exposure and oral exposure leading to significant activity. As for tentative to go a step further in the prediction, Péry et al. (2011) showed the possibility to relate in vitro and in vivo gene responses in macrophages following exposure to benzo(a)pyrene through PBPK modeling. Punt et al. (2011) proposed a few examples where PBPK models have been used to extrapolate in vitro toxicity results to the in vivo situation. They show that predictions of in vivo effects on the basis of integrated in vitro and PBPK modeling approaches are generally within one order of magnitude of the observed in vivo data. Among the examples, Gubbels-van Hal et al. (2005) attempted to predict acute toxicity and repeated dose toxicity based on PBPK models and cytotoxic data. In general, the predictions over-estimated the toxicity, which could be both due to overestimation of the effects in the in vitro systems and to uncertainty on the toxicokinetics parameters. There are thus improvements to be made both in the experimental field and in the accurate calibration of toxicokinetics models. Punt et al. (2011) asked for more examples that provide proof of principle of deriving in vivo dose–response curves based on in vitro assays and PBPK modeling techniques. Yoon et al. (2012) reviewed different quantitative in vitro to in vivo extrapolations (QIVIVE). They identified key elements that have to be predicted for a successful extrapolation: (i) intestinal absorption and pre-hepatic clearance, (ii) hepatic and extra-hepatic metabolic clearance, (iii) renal clearance, (iv) volume of distribution (for acute in vivo exposures). They also indicated that only rare examples of QIVIVE do not rely, at least partly, on in vivo data.

In this paper, we propose a modeling framework to predict in vivo liver toxicity. The proposed framework is composed by a dynamic model to analyze in vitro concentration-liver cell viability data throughout time, and by a PBPK model to relate dose of exposure and concentration in the liver. As case study, this modeling framework was applied to humans exposed to acetaminophen, which is a pharmaceutical known for its hepatotoxicity (Jaeschke et al., 2011). The human PBPK model was calibrated in three different ways: (i) with in vitro and in silico methods only, (ii) by extrapolating to humans a PBPK model calibrated for rats, as usually performed in human toxicokinetics assessment based on animal testing, (iii) or by using all relevant available data, including actual human data to compare the quality of the kinetics and effects predictions based either on animal or non-animal testing.

Section snippets

Toxicokinetic data in rats for acetaminophen

Male Sprague-Dawley rats were purchased from Elevage Janvier Laboratory (Le Genet Saint Isle, France). They were maintained under standardized conditions (relative humidity 55 ± 15%, 12-h day night cycle, room temperature 22 ± 2 °C). There were five different groups of 16 rats, with different times of exposure per group. For each group, 14 rats were orally exposed to a single dose of acetaminophen in an aqueous suspension containing methycellulose (0.5% w/v) and Tween 80 (0.1% w/v). We selected the

PBPK modeling

The fits of the rat PBTK model to our data are presented in Fig. 2 for plasma. As for excretion, the model predicts that more than 96% of the initial amount of acetaminophen was eliminated after 24 h (10% not absorbed, 41% eliminated through the bile, 45% eliminated through urine as sulfate conjugate (58%), glucuronide (24%), acetaminophen (9%) and other metabolites (9%)). Actual urine data showed a recovery of 45% in urine (sulfate conjugate (56%), glucuronide (26%), acetaminophen (12%) and

Relevance of our PBPK models relative to other kinetics studies for rats and humans

For rats, we estimated a maximum velocity for glucuronide formation of 5.3 μmol/min/kg. This is twice the value found by Watari et al. (1983) for lower exposure doses (2.76 μmol/min/kg). Maximum velocity for sulfate formation at the start of the experiment was estimated at 3.5 μmol/min/kg, which is lower but close to the value from Watari et al. (1983), 4.92 μmol/min/kg. The pool of activated sulfate was estimated at a mean value of 417 μmol. As hypothesized, the rate of formation of new activated

Conflict of interest

There are no conflicts of interest.

Acknowledgements

The research leading to the results presented here has received funding from the European Community's Seventh Framework Program (FP7/2007-2013) and from Cosmetics Europe through the COSMOS project under grant agreement no. 266835. The experimental rat data have been obtained in the project ACTIVISME funded by the French ministry of ecology. The authors are very grateful to Alicia Paini who helped in the finalization of the work, so as to two anonymous reviewers who greatly helped to improve the

References (48)

  • A.R.R. Péry et al.

    Predicting in vivo gene expression in macrophages after exposure to benzo(a)oyrene based on in vitro assays and toxicokinetic/toxicodynamic models

    Toxicology Letters

    (2011)
  • A. Punt et al.

    Evaluation of research activities and research needs to increase the impact and applicability of alternative testing strategies in risk assessment practice

    Regulatory Toxicology and Pharmacology

    (2011)
  • T. Rodgers et al.

    Physiologically based pharmacokinetic modeling 2: predicting the tissue distribution of acids, very weak bases, neutrals and zwitterions

    Journal of Pharmacological Sciences

    (2006)
  • R.M. Rodrigues et al.

    Assessment of an automated in vitro basal cytotoxicity test system based on metabolically-competent cells

    Toxicology In Vitro

    (2013)
  • ACD/PhysChem Suite, version 12.00. Advanced Chemistry Development, Inc., Toronto, ON, Canada. www.acdlabs.com,...
  • ACD/ADME Suite, version 5.0. Advanced Chemistry Development, Inc., Toronto, ON, Canada. www.acdlabs.com,...
  • S. Adler et al.

    Alternative (non-animal) methods for cosmetics testing: current status and future prospects-2010

    Archives of Toxicology

    (2011)
  • M.E. Andersen et al.

    The vision of toxicity testing in the 21st century: moving from discussion to action

    Toxicological Sciences

    (2010)
  • D.N. Bailey et al.

    The binding of selected therapeutic drugs to human serum [alpha]-1 acid glycoprotein and to human serum albumin in vitro

    Therapeutic Drug Monitoring

    (2004)
  • J.R. Bales et al.

    Urinary excretion of acetaminophen andits metabolites as studied by proton NMR spectroscopy

    Clinical Chemistry

    (1984)
  • Z.E. Barter et al.

    Scaling factors for the extrapolation of in vivo metabolic drug clearance from in vitro data: reaching a consensus on values of human microsomal protein and hepatocellularity per gram of liver

    Current Drug Metabolism

    (2007)
  • F. Bois et al.

    MCSim: a simulation program

    Journal of Statistical Software

    (1997)
  • R.C. Brown et al.

    The effect of wheat bran and pectin on paracetamol absorption in the rat

    British Journal of Nutrition

    (1979)
  • T.A. Clayton et al.

    Pharmaco-metabonomic phenotyping and personalized drug treatment

    Nature

    (2006)
  • Cited by (0)

    View full text