PMCCPMCCPMCC

Search tips
Search criteria 

Advanced

 
Logo of toxsciLink to Publisher's site
 
Toxicol Sci. 2013 November; 136(1): 4–18.
Published online 2013 August 19. doi:  10.1093/toxsci/kft178
PMCID: PMC3829570

Incorporating New Technologies Into Toxicity Testing and Risk Assessment: Moving From 21st Century Vision to a Data-Driven Framework

Abstract

Based on existing data and previous work, a series of studies is proposed as a basis toward a pragmatic early step in transforming toxicity testing. These studies were assembled into a data-driven framework that invokes successive tiers of testing with margin of exposure (MOE) as the primary metric. The first tier of the framework integrates data from high-throughput in vitro assays, in vitro-to-in vivo extrapolation (IVIVE) pharmacokinetic modeling, and exposure modeling. The in vitro assays are used to separate chemicals based on their relative selectivity in interacting with biological targets and identify the concentration at which these interactions occur. The IVIVE modeling converts in vitro concentrations into external dose for calculation of the point of departure (POD) and comparisons to human exposure estimates to yield a MOE. The second tier involves short-term in vivo studies, expanded pharmacokinetic evaluations, and refined human exposure estimates. The results from the second tier studies provide more accurate estimates of the POD and the MOE. The third tier contains the traditional animal studies currently used to assess chemical safety. In each tier, the POD for selective chemicals is based primarily on endpoints associated with a proposed mode of action, whereas the POD for nonselective chemicals is based on potential biological perturbation. Based on the MOE, a significant percentage of chemicals evaluated in the first 2 tiers could be eliminated from further testing. The framework provides a risk-based and animal-sparing approach to evaluate chemical safety, drawing broadly from previous experience but incorporating technological advances to increase efficiency.

Key Words: in vitro and altenatives, biotransformation and toxicokinetics, predictive toxicology, risk assessment, safety evaluation, exposure.

Shortly after the turn of the century, there was increasing recognition and acceptance within government agencies that new approaches were needed to evaluate the safety of the relatively large number of chemicals in commerce and the environment (EPA, 2003; Kavlock et al., 2005; Meek and Armstrong, 2007; NTP, 2004). Following this recognition, the release of the National Research Council’s Report “Toxicity Testing in the 21st Century: A Vision and a Strategy” (NRC, 2007) initiated a broad-based movement in the toxicology community to reassess how toxicity testing and risk assessment are performed. Since the release of the report, multiple efforts in the United States and abroad have added to the momentum with the shared goal of transitioning toxicity testing and risk assessment from an outdated, inefficient, costly, and animal-centric process to one that is more efficient, economical, less animal intensive, and more relevant to human health by utilizing new technologies that provide a better understanding of the underlying biological system. However, the majority of these efforts have focused more on a vision of how things should be done rather than the development of a pragmatic path forward that can be iteratively refined as greater understanding is achieved.

Over the past 5 years, a series of studies has been conducted that together may contribute first steps toward executing the shared vision. This article considers the implications of these studies and develops a new, data-driven tiered toxicity testing framework with potentially broad, international application across multiple regulatory agencies. The framework evolved from tiered approaches developed previously to address regulatory mandates to prioritize and assess large numbers of substances (Meek and Armstrong, 2007; Meek et al., 2011). The primary application of the proposed framework is for chemicals with little or no safety-related data. In the United States, the proposed framework could, for example, be applied to new and legacy manufactured chemicals regulated under the Toxic Substances Control Act (TSCA). Conceptually, it could also be applied to other nonpharmaceutical chemicals that have defined toxicity testing requirements, such as pesticides. In Europe, the proposed framework could be applied to identify substances of concern within Registration, Evaluation, Authorization and Restriction of Chemicals (REACH) that may have adverse effects that are not indicated in the basic information requirements linked to production volume.

TIER 1

The first tier of the proposed framework consists of 5 components—(1) use of high-throughput in vitro assays to separate chemicals into selective and nonselective modes of action; (2) in vitro genotoxicity assays to identify potential genotoxic and nongenotoxic chemicals; (3) in vitro-to-in vivo extrapolation (IVIVE) pharmacokinetic modeling to convert in vitro assay concentrations to applied doses; (4) high-throughput exposure modeling to estimate human exposures to chemicals; and (5) calculation of a margin of exposure (MOE) (Fig. 1). The data collection and analysis associated with each of these components would ideally occur simultaneously for inclusion in a Tier 1 data package associated with each chemical, analogous to the chemical dossiers currently being assembled for REACH. Within the current framework, the high-throughput in vitro screening would be composed of assays that are similar to those employed in the ToxCast project (Houck et al., 2009; Judson et al., 2010; Knudsen et al., 2011; Martin et al., 2010; Reif et al., 2010; Rotroff et al., 2010a). The in vitro ToxCast assays are primarily repurposed biochemical and cell-based assays used in drug discovery and cover 327 genes and 293 Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways (Kavlock et al., 2012). The details surrounding their selection and application have been discussed in the publications cited earlier. Although this selection of in vitro assays may not be ideal for covering the mechanistic landscape of chemically mediated toxicity, the ToxCast data set is the largest currently available for establishing the associations underlying this framework and provides a significant (though recognizably imperfect) starting point for a pragmatic path forward that can be refined as additional experience is acquired.

Fig. 1.
A flowchart outlining Tier 1 in the proposed framework. The green boxes illustrate the Tier 1 data package that includes experimental data and computational modeling results that serve as inputs into the framework. The yellow boxes are separate chemical ...

In Vitro Assays for Bioactivity

In the ToxCast phase I effort, 309 unique chemicals were screened in more than 600 in vitro assays. The 309 chemicals were primarily pesticides and high-production volume chemicals for which significant in vivo animal testing has been performed (Judson et al., 2009; Martin et al., 2009a). Although prediction of specific in vivo hazards with the current set of ToxCast in vitro assays is limited (Thomas et al., 2012a; Wetmore et al., 2012a), the in vitro assays do provide a broad biological profile of the potential proximal biochemical and cellular targets for a chemical. Instead of using the high-throughput in vitro assays to predict hazard, the initial role of high-throughput in vitro toxicity screening in our framework is to separate chemicals into either those that cause toxicity primarily through nonselective interactions with cells and cellular macromolecules or those that act through more selective interactions (eg, receptor-mediated chemicals).

To demonstrate how chemicals could be separated into these categories, the ToxCast phase I data were analyzed for surrogate measures of selectivity. Additional work in this area will need to be performed to identify which selective interactions are important for toxicity. The ToxCast in vitro assays were first grouped by gene to eliminate redundancy. For example, all of the in vitro assays evaluating the binding or transcriptional activation of estrogen receptor alpha (ESR1) were grouped together. The in vitro assays that did not correspond to a specific gene (eg, general cytotoxicity assays) were removed from the analysis. If any of the in vitro assays associated with a gene possessed an AC50 value (concentration at 50% of maximum activity), then the chemical was considered to have activity (ie, either activates or inhibits) toward that gene. Thus, the more genes with AC50 values, the less selective a chemical would be. For the ToxCast phase I chemicals, approximately 80% of chemicals have activity toward 10 or more genes, and the average chemical has activity toward approximately 20 genes (Fig. 2A).

Fig. 2.
Cumulative frequency plots summarizing the selectivity of the ToxCast phase I chemicals across the high-throughput in vitro assays. The in vitro assays were grouped by gene to eliminate redundancy. A, The number of genes with an AC50 value for each chemical. ...

A complementary measure of selectivity expands on this concept and takes into consideration the dose range across which the different in vitro assays were activated. For each chemical, the in vitro assays were again grouped by gene, and the in vitro assays that did not correspond to a specific gene were removed from the analysis (eg, general cytotoxicity assays). The potency of each chemical toward a specific gene was summarized by taking the minimum AC50 value among the associated in vitro assays. For example, if there were 5 in vitro assays evaluating the binding or transcriptional activation of ESR1, the potency value for the most sensitive assay was selected. For each chemical, the summary values for each gene were then sorted, and the lowest summary value across all genes was divided by the 10th percentile. In other words, for each chemical, the potency of the in vitro assay for the most sensitive gene was divided by the relative potency when 10% of the genes were affected. According to this measure of selectivity, approximately 80% of the ToxCast phase I chemicals have a ratio of 3 or less between the most sensitive gene target and the 10th percentile (Fig. 2B). Taken together, these data suggest that the majority of chemicals represented in the ToxCast phase I library likely act via nonselective interactions with cellular macromolecules. These results may not be that surprising given that most chemicals in the ToxCast phase I library were nonpharmaceuticals and were developed either for their functional properties in a range of different products or pesticidal qualities and not optimized to interact specifically with a target protein.

In Vitro Assays for Genotoxicity

In addition to separating chemicals based on selectivity, chemicals are also grouped based on their genotoxic potential. This second screen was included as a practical consideration because most regulatory organizations either label or deal with substances that are likely to be carcinogenic through a mutagenic mode of action (MOA) differently than those considered not to be genotoxic. The in vitro assays for genotoxicity within the proposed framework include the bacterial reverse mutation test and the mammalian cell micronucleus test. These 2 in vitro genotoxicity assays cover the 3 critical genetic endpoints implicated in carcinogenesis and heritable diseases, namely gene mutations, and structural and numerical chromosome alterations. An expert panel convened by the European Food Safety Authority (EFSA) concluded that the inclusion of additional in vitro mammalian cell tests apart from these 2 assays would “significantly reduce specificity with no substantial gain in sensitivity” (EFSA, 2011). This is consistent with previous recommendations of multiple regulatory agencies and the International Conference on Harmonization (ICH) for registration of pharmaceuticals (HC, 2000). Adaptations have been proposed that reduce the number of false positives in the mammalian cell genotoxicity assays (Fowler et al., 2012). If the 2 in vitro genotoxicity assays are negative, the chemical is placed in the nongenotoxic category, whereas positive results in both assays would classify the chemical as potentially genotoxic. If a chemical is positive in one of the in vitro assays, the chemical would be classified as potentially genotoxic and additional in vitro or in vivo assays could be performed to confirm the classification. For example, if a chemical is negative in the bacterial reverse mutation test and positive in the mammalian cell micronucleus test, an in vivo micronucleus analysis could also be performed to confirm the clastogenicity as recommended in the standard ICH guidelines (Müller et al., 1999). Ideally, the in vivo micronuclei studies would be performed in Tier 2 on the same animals as the transcriptomic analysis using red blood cell micronuclei as the endpoint. If a chemical is positive in the bacterial reverse mutation test and negative in the mammalian cell micronucleus test, an in vivo gene mutation study could be performed using 28-day exposures of transgenic rodents carrying mutation reporter genes (OECD, 2011), but this is not cost effective for many chemicals at this point in time.

In Vitro Pharmacokinetic Assays and IVIVE Modeling

The internal dose of a chemical is an important determinant of toxicity. To incorporate pharmacokinetics into the framework, data from in vitro hepatic metabolic clearance and plasma protein binding assays are included in the Tier 1 data package. These 2 pharmacokinetic parameters are critical for estimating steady-state blood concentrations. In previous studies, experimental data from the in vitro pharmacokinetic assays were used to parameterize an IVIVE model to estimate the daily human oral dose, called the oral equivalent dose, necessary to produce steady-state in vivo blood concentrations equivalent to the in vitro AC50 value for each of the high-throughput in vitro assays (Rotroff et al., 2010b; Wetmore et al., 2012b). Oral equivalent dose values calculated using this approach are inherently conservative due to various assumptions in the IVIVE modeling (Rotroff et al., 2010b; Wetmore et al., 2012b). Monte Carlo sampling was also incorporated into the IVIVE model to account for interindividual variability and as a basis to derive an oral equivalent dose that represents the 95th percentile of the population. A similar approach takes place in our framework and when appropriate, other biokinetic factors within the in vitro assays will be taken into account (Blaauboer, 2010). Human equivalent concentrations could also be calculated to allow comparisons to air exposure levels.

Selective-Acting Chemicals and MOA

For the selective chemicals, the high-throughput in vitro assays are used to identify key events in a potential MOA. As previously defined, the MOA is a biologically plausible series of key events leading to an outcome that could be adverse (Sonich-Mullin et al., 2001). Key events are empirically observable steps that are critical to the outcome. Although originally and often simply conceptualized as a linear series of key events, a MOA actually involves interdependent networks of events with feedback loops. A weight of evidence for the hypothesized MOA is established based on modified Bradford Hill considerations including dose response and temporal concordance of key events, consistency and specificity, and biological plausibility (Boobis et al., 2006, 2008).

For application to selective chemicals in Tier 1 of the proposed framework, the key event must be triggered at doses lower than or equal to doses at which the adverse outcome is observed (ie, dose concordance between a key event and the adverse outcome). An example of this approach has been provided for the relationship between in vitro peroxisome proliferator–activated receptor alpha (PPARA) activation and rat liver proliferative lesions and liver tumors (Wetmore et al., 2012a). Among a subset of ToxCast phase I chemicals, 8 were active in the in vitro PPARA assays and cause proliferative lesions in the livers of rats, while 4 also cause rat liver tumors. For each chemical, the in vivo-derived lowest effect level (LEL) and no effect level for rat liver proliferative lesions and tumors were overlaid with the oral equivalent dose values for the 3 high-throughput in vitro assays measuring PPARA-related activation (Fig. 3). In all cases, activation of the PPARA in vitro assays occurred at or below the dose for the in vivo effects, thereby demonstrating dose concordance between this potential key event and the adverse outcome. A similar approach is taken in the proposed framework. From these selectively activated or inhibited assays, probable key event(s) will be determined and the oral equivalent dose value from a relevant in vitro assay will be compared with doses causing the associated in vivo adverse outcome (when available).

Fig. 3.
Comparison of in vivo low effect and no effect doses for rat liver proliferative lesions and the in vitro PPARA-activating oral equivalent doses in the ToxCast phase I assays. The in vivo effects of 8 chemicals were plotted as LEL (blue triangles) and ...

Apart from dose concordance, the biological relevance linking the in vitro assay to the in vivo adverse outcome will also be qualitatively assessed. If the in vitro assay representing the key event demonstrates dose concordance and biological relevance, the oral equivalent dose value will be used as the point of departure (POD) in a dose-response assessment. It is anticipated that there will be a limited number of MOAs that will be identified in Tier 1 for the selective chemicals. The MOAs will have a defined set of criteria based on chemicals that are known to act through those key events. Additional MOAs will be added as our biological knowledge improves and the assay suite is refined.

Nonselective Chemicals and Biological Perturbation

Identifying the MOA will generally be neither efficient nor economical for the nonselective chemicals because the chemicals will interact with and perturb multiple cellular processes. In a previous study, rat-specific oral equivalent doses were calculated for 59 ToxCast phase I chemicals and compared with LEL values for a variety of in vivo apical responses (Wetmore et al., 2012a). Although not significantly correlated (r 2 = .046), the oral equivalent dose for the most sensitive in vitro assay was less than the LEL for the most sensitive rat in vivo endpoint for approximately 95% of the chemicals (Fig. 4A). The LEL values spanned 38 unique in vivo endpoints across multiple tissue and organ systems and across reproductive, chronic, and developmental study types. The median difference between the most sensitive in vitro assay and the most sensitive LEL value was 66-fold (Fig. 4B). These results indicate that the most sensitive in vitro assay was protective regardless of the nature of the adverse outcome and that bioactivity based on the oral equivalent dose for the most sensitive in vitro assay could be used as a reasonable estimate of the POD for the nonselective chemicals in a dose-response assessment.

Fig. 4.
A comparison of ToxCast in vitro assay possessing the lowest oral equivalent dose with the in vivo response possessing the lowest LEL for each chemical. A, Scatter plot of the log-transformed minimum LEL in ToxRefDB (y-axis) versus the minimum oral equivalent ...

Human Exposure Characterization

The final component of the Tier 1 data package relates to exposure. An understanding of human exposure is critical for placing the dose-response data for toxicity into context and ultimately assessing health risks. Human exposure estimates can be obtained from multiple sources and are associated with varying levels of uncertainty, depending principally on the complexity and accuracy of input data and model assumptions. Simple surrogates such as use and emission profiling and physicochemical properties offer potential to reasonably discriminate amongst chemicals with respect to exposure potential (Arnot et al., 2012; Meek and Armstrong, 2007; Meek et al., 2011).

For some of the Tier 1 chemicals, human exposure estimates can be derived from biomonitoring data such as the U.S. National Health and Nutrition Examination Survey (NHANES) (CDC, 2009). Measurements of parent chemical concentrations in the blood or urinary metabolite concentrations can be used to infer exposure concentrations using reverse dosimetry methods (Tan et al., 2007). Using Monte Carlo sampling, probabilistic information about pharmacokinetics and exposure patterns can be included to estimate variability in the exposure distribution across a diverse population (Tan et al., 2007). In Tier 1 of our framework, a conservative 95th percentile upper confidence bound on the exposure estimate is used, where available.

For chemicals without biomonitoring data, it is proposed to use a high-throughput exposure modeling that has been developed by combining existing environmental fate and transport models (ie, far field) with an empirical adjustment for indoor or consumer exposure (ie, near field) (Wambaugh et al., 2013). In addition to a high-level assessment of consumer use, physicochemical properties and an estimate of environmental release are required as inputs to the model. The approach was calibrated using 82 chemicals from NHANES and then used to predict human exposures for 1,936 chemicals. The approach provides uncertainty bounds on the aggregate exposure estimates. The 95th percentile upper confidence bound of the exposure estimate is utilized in our framework.

MOEs

The human exposure estimates are combined with the POD from the dose-response assessment to calculate a MOE. For the components of the framework discussed above, the MOE is based on administered dose. A second alternative is the calculation of MOE based on internal dose using human biomonitoring data. For this alternative, concentrations in blood are directly compared with the concentrations in the in vitro assays so as to provide a direct measure of a MOE (Aylward and Hays, 2011; Aylward et al., 2013). Although not all chemicals are amenable to biomonitoring measurements, the advantage of this approach is that it eliminates inherent uncertainty in the pharmacokinetic and exposure modeling. Interindividual variability could be incorporated by using either the percentiles of the measured distribution in large biomonitoring studies (CDC, 2009) or estimated from smaller studies using pooled samples (Caudill, 2010, 2012).

Regardless of whether the MOE is calculated using administered or internal dose, the MOE is the primary metric for determining whether a substance advances on for consideration in subsequent tiers. For the selective chemicals, the MOE is calculated relative to the oral equivalent dose from the in vitro assay(s) associated with the proposed key event. For the nonselective chemicals, the MOE will be calculated relative to the oral equivalent dose for the most sensitive in vitro assay. For those chemicals with a MOE greater than some defined cutoff, no further testing would be required, whereas chemicals with a MOE below the cutoff would progress on to Tier 2. In a previous study on the ToxCast phase I chemicals (Wetmore et al., 2012a), a Tier 1 MOE cutoff of > 100 would eliminate additional testing on approximately 40% of chemicals, whereas a MOE cutoff of > 1000 would eliminate additional testing on approximately 25% of chemicals (Fig. 5). Although the MOE cutoff will be primarily set through policy decisions and not necessarily based solely on scientific criteria, it should be set to provide adequate protection for susceptible subpopulations.

Fig. 5.
Cumulative frequency plot of the ratio between in vitro bioactivity relative to human exposure in the most highly exposed subpopulation for the ToxCast phase I chemicals. The MOE was calculated by comparing the minimum oral equivalent dose across the ...

Reference Values

For those chemicals that do not advance to the next tier, Tier 1 reference values could be used to provide exposure guidelines. For certain regulatory agencies these could be considered screening-level values, whereas other agencies may wish to apply them in alternative ways. For the nongenotoxic chemicals, the POD and associated lower confidence bound can be divided by a set of uncertainty/safety factors to derive a Tier 1 reference value as a basis to provide health-related guidance. A previous study using in vitro data has provided one possible approach (Judson et al., 2011). For the genotoxic chemicals, the POD and associated lower confidence bound can be divided by uncertainty/safety factors, or, alternatively, a linear, no threshold dose-response extrapolation could be applied. In this case, the POD could be assumed to be equivalent to a defined increase in adverse response (eg, 10%) and used to calculate a Tier 1 slope factor. However, it should be noted that activation of the in vitro assays analyzed here cannot yet be quantitatively linked to a specific incidence in the in vivo adverse response.

Chemicals Not Amenable to In Vitro Screening

Although efforts are underway to broaden the applicability of high-throughput in vitro screening platforms, certain chemicals, such as volatile substances, those not soluble in dimethyl sulfoxide or those that bind avidly to plastic, are currently not readily amenable for high-throughput in vitro screening. In the proposed framework, these chemicals would be identified using an initial structure-activity relationship or molecular properties filter and are automatically directed to Tier 2 for evaluation. No measure of selectivity would be available for these chemicals, and they would be assumed to act through nonselective mechanisms.

TIER 2

The second tier of the proposed framework consists of 5 components—(1) short-term in vivo transcriptomic studies to identify the transcriptional POD values for nonselective chemicals; (2) in vivo MOA studies to identify POD values for selective chemicals; (3) in vitro and in vivo pharmacokinetic studies to improve linkage between internal and applied dose; (4) refined human exposure estimates; and (5) calculation of a MOE (Fig. 6). Similar to the first tier, the data collection and analysis associated with each of these components would ideally occur simultaneously for inclusion in a cohesive Tier 2 data package. The data requirements would be different for the selective and nonselective chemicals.

Fig. 6.
A flowchart outlining Tier 2 in the proposed framework. The green boxes illustrate the Tier 2 data package that includes experimental data and computational modeling results that serve as inputs into the framework. The chemical categories determined by ...

Selective-Acting Chemicals and MOA

Focused in vivo MOA studies would be performed on the selective chemicals to evaluate the putative MOA defined in Tier 1 and would be defined by the type of MOA proposed. For proposed nuclear receptor agonists such as the constitutive androstane receptor, pregnane X receptor, or PPARA, these studies could include knockout and humanized rodent models. For proposed estrogen receptor agonists, these studies may also include standard in vivo studies such as the immature rat uterotrophic assay (OECD 440). Following the in vivo MOA studies and consideration of the weight of evidence, qualitative and quantitative human concordance would be evaluated (Boobis et al., 2008). The POD for relevant MOAs would be determined based on the associated endpoints in the in vivo studies, but scaled, taking into account relevant kinetic and dynamic data.

Nonselective Chemicals and Biological Perturbation

The premise for analysis of nonselective chemicals has been developed based on the results of a series of previous studies examining the relationship between transcriptional and apical responses (Thomas et al., 2007, 2011, 2012b, 2013). Both noncancer and cancer apical responses were analyzed using benchmark dose (BMD) methods to identify PODs. The dose response for changes in gene expression was also analyzed using BMD methods, and the responses were grouped based on signaling pathways. In a comparison of transcriptional BMD values for the most sensitive pathway with BMD values for the noncancer and cancer apical endpoints, there was a high degree of correlation (Thomas et al., 2012b) (Fig. 7). The correlation was also robust across multiple time points (Thomas et al., 2013). In this proposed framework, dose-response studies would be performed on the chemical at any single time point between 5 days and 13 weeks in mice and rats of both sexes. A battery of 8 tissues that include those most frequently impacted in rodent cancer bioassays (liver, lung, mammary gland, stomach, vascular system, kidney, hematopoietic system, and urinary bladder) would be harvested. These 8 tissues cover 92% and 82% of targets for all mouse and rat carcinogens, respectively (Gold et al., 2001). For noncancer effects, a previous analysis of 69 different pesticides demonstrated that adverse responses in a chronic rat bioassay generally occurred at lower doses than either a rat 2-generation reproductive study or a rat developmental toxicity study (Dourson et al., 1992). In addition, potential reproductive and developmental toxicants could be identified based on consideration of weight of evidence of available data including structure activity modeling (Blackburn et al., 2011). For chemicals flagged for these potential responses, additional tissues and alternate study designs (eg, OECD 421) would be incorporated to cover potential reproductive and developmental effects. Gene expression microarray analysis on these tissues would allow the estimation of pathway transcriptional BMD and BMDL values. The signaling pathway with the lowest transcriptional BMD value among the analyzed tissues would be used to derive the POD.

Fig. 7.
Scatter plots of the relationship between the BMD values for (A) cancer-related apical endpoints or (B) noncancer apical endpoints and transcriptional BMD values for the most sensitive signaling pathway following 13 weeks of exposure. The data points ...

In Vivo and In Vitro Pharmacokinetic Studies and Modeling

The pharmacokinetic data from Tier 1 would also be expanded through focused in vitro studies and the collection of additional samples within the existing in vivo studies. Potential metabolites of each chemical would be identified using rodent and human microsomes, S9, and plasma. Recent advances in semiautomated metabolite identification have been made by combining in silico metabolite prediction software with ultrahigh pressure liquid chromatography and mass spectral analysis (Bonn et al., 2010). In essence, computational approaches are used to align the raw mass spectral data with a predicted site of metabolism in order to assign the structure of a metabolite. In our framework, these studies are performed to qualitatively identify probable metabolites in both the rodent and human. Reactive metabolites will also be evaluated using primary hepatocytes for analysis of lipid peroxidation, reactive oxygen species generation, and glutathione depletion.

Apart from the metabolite characterization, blood samples will be collected in either the in vivo mode-of-action studies for the selective chemicals or the in vivo transcriptomic studies for the nonselective chemicals. A previous study has demonstrated that the collection of blood samples at 3 specific time points during a single day of a repeat dose study was sufficient to estimate systemic bioavailability of a chemical (Saghir et al., 2006). The human volume of distribution of each chemical will be estimated using quantitative structure-activity relationships (Peyret and Krishnan, 2011; Peyret et al., 2010; Poulin and Theil, 2002; Rodgers and Rowland, 2006, 2007; Rodgers et al., 2005), and the oral bioavailability will be estimated using the in vitro bidirectional permeability assay with Caco-2 cells (Wetmore et al., 2012b). Together with the metabolic clearance and plasma protein binding data from the Tier 1 studies, the combined data will be used to develop pharmacokinetic models to define the administered dose in humans leading to the active internal doses seen in the Tier 2 studies.

Human Exposure Characterization

Human exposure estimates from Tier 1 could potentially be refined through a combination of experimental measurements of physical-chemical properties and environmental degradation rates, refined release rates, detailed consumer and indoor use profiling, product formulation data, environmental and household exposure factor estimates, and targeted biomonitoring studies. The fate and transport models used for the high-throughput exposure modeling in Tier 1 rely heavily on physical-chemical properties, environmental half-lives, and release rates. For many chemicals, the physical-chemical properties and environmental half-lives are not experimentally available and have to be estimated using quantitative structure-activity relationships (Wambaugh et al., 2013). Higher throughput methods are available for measuring both physical-chemical properties (Kerns, 2001) and the degradation half-lives in some environmental media (Hussain et al., 2007). In our proposed framework, these higher throughput methods could be applied to all Tier 2 chemicals. In addition, more accurate industrial release rates could be obtained from industry. The combination of these data may be used to reduce the uncertainty in the fate and transport models.

To refine the near-field component of the exposure estimates, one option is to obtain more detailed consumer and indoor use profiles and product formulation data. These use profiles and formulation data could be aggregated from existing online databases such as the National Library of Medicine (NLM) Hazardous Substances Data Bank, NLM Household Products Database, Walmart’s online material safety data sheets search engine, U.S. Environmental Protection Agency’s (EPA) Exposure Factor Handbook (EPA, 2011), and the Substances in Products in the Nordic Countries (SPIN) database. A significant amount of the exposure-related data aggregation is currently underway (Egeghy et al., 2012; Mattingly et al., 2012). In the proposed framework, the more detailed use profile and product formulation data could potentially be used to select relevant consumer exposure models that exist for specific exposure scenarios (Kephalopoulos et al., 2007). The resulting near-field exposure estimates could be combined with the far-field exposures to obtain an aggregated exposure estimate.

MOEs

Similar to Tier 1, the human exposure estimates will be combined with the POD from the in vivo dose-response studies to calculate a MOE based on administered dose. Within Tier 2, additional alternatives could be explored to integrate high-throughput biomonitoring studies in order to calculate a MOE based on internal dose. Advances in analytical chemistry and Fourier transform mass spectrometry have enabled what has been termed “high-performance metabolic profiling” (Johnson et al., 2010; Park et al., 2012; Soltow et al., 2011). Using these methods, up to 7,000 chemicals can be measured in 20 µl of biofluids in 20min (Soltow et al., 2011). The high-performance metabolic profiling could be applied to biomonitoring samples from a range of different subpopulations, and mass labeled standards could be used to quantify specific chemicals of interest. The concentrations in the human blood samples would then be compared with estimated blood concentrations at the POD in the in vivo transcriptomic or MOA studies to provide a MOE.

Regardless of whether the MOE is calculated using administered or internal dose, the POD for the selective chemicals will be based on the associated endpoints in the in vivo MOA studies, whereas the POD for the nonselective chemicals will be derived from the signaling pathway with the lowest transcriptional BMD value across all analyzed tissues. For those chemicals with a MOE greater than a defined cutoff, no further testing would be required, whereas chemicals with a MOE below the cutoff would progress on to Tier 3. Using data from ToxRefDB on the in vivo LEL values for the ToxCast phase I chemicals (Knudsen et al., 2009; Martin et al., 2009a,b) and the corresponding human exposure estimates (Rotroff et al., 2010b; Wetmore et al., 2012b), MOE values were calculated using the minimum in vivo LEL value (ie, the most sensitive in vivo endpoint) and the exposure estimate for the most highly exposed subpopulation. Assuming that this ratio is representative of the MOE values that will be obtained from Tier 2 in our proposed framework, a MOE cutoff of > 100 would eliminate additional testing on approximately 97% of chemicals, whereas a MOE cutoff of > 1000 would eliminate testing on approximately 85% of chemicals (Fig. 8).

Fig. 8.
Cumulative frequency plot of the ratio between the in vivo LEL relative to human exposure in the most highly exposed subpopulation for ToxCast phase I chemicals. The MOE was calculated by comparing the minimum LEL value of the guideline animal studies ...

Reference Values

For those chemicals that are set aside from additional testing, Tier 2 reference values could be published to provide exposure guidelines. Similar to those described for Tier 1, the Tier 2 reference values may have different applications among regulatory agencies. For the nongenotoxic chemicals, the POD and associated lower confidence bound can be divided by a set of uncertainty/safety factors to derive a Tier 2 reference value as a basis to provide health-related guidance. For the genotoxic chemicals, the POD and associated lower confidence bound can be divided by uncertainty/safety factors, or, alternatively, a linear, no threshold dose-response extrapolation could be applied. The POD for the most sensitive pathway among the analyzed tissues could be used as the equivalent to the BMDL10 for a tumor response (ie, the lower confidence limit of the BMD associated with a 10% increased risk of tumors). The rationale for this equivalence is that our data showed that transcriptional BMD values for the most sensitive pathway were, on average, within a factor of 2 of the BMD10 values for tumor responses in the corresponding rodent bioassays (Thomas et al., 2013). The pathway transcriptional BMDL value could then be used to estimate a slope factor and the desired risk specific dose.

TIER 3

The toxicity testing proposed for chemicals passing into Tier 3 is not explicitly defined but would be conceptually equivalent to the traditional in vivo studies performed on high-value chemicals with significant potential for human exposure. Alternatively, they would be specified based on understanding of the toxicological profile acquired in lower tier testing. These studies could include rodent cancer bioassays, developmental toxicity studies, and reproductive toxicity studies. Depending on the MOE cutoff values imposed, it is anticipated that the majority of chemicals will be screened out in the preceding tiers and the estimated number of chemicals requiring these studies would be between 3% and 15%. Testing for these remaining compounds could be prioritized by endpoint and compounds based on the results from both Tier 1 and 2 studies.

SUMMARY

The proposed framework integrates data from new technologies into toxicity testing using what is currently the best available science. The proposed framework provides a risk-based, efficient, and animal-sparing approach to evaluate chemical safety. It is consistent with and draws broadly from previous experience in risk assessment but incorporates recent advances in technology to increase efficiency. The framework relies on the initial separation of chemicals into selective and nonselective MOAs. The prevailing thought is that the MOA approach can be efficiently applied to all chemicals. In our view, applying the MOA approach to the nonselective chemicals is neither practical, scientifically justifiable, nor consistent with evolving experience in increasing efficiency in risk assessment. First, imposing an MOA approach on these chemicals would waste valuable resources and unnecessarily delay decision making because each MOA requires agreement on the underlying key events followed by extensive peer review. Second, for chemicals which interact with numerous molecular targets in a cell, identifying a prevailing MOA among many operating in parallel will be a complex and potentially uncertain process. Thus, these efforts should be focused on the selective chemicals where a series of standardized MOAs could be developed that are associated with specific cellular targets.

A shift in thinking will also be required for the nonselective chemicals. The current hazard-based labeling approach that relies on apical responses will need to be transitioned to a “region of safety” approach where the most sensitive adverse apical effect is not known, but disruption of important biologic processes can be measured and the dose response characterized. The subsequent decisions are based on potential biological perturbations (Andersen and Krewski, 2010). This shift in thinking may be more challenging to implement due to the entrenched reliance on apical responses, the potentially conservative nature of utilizing molecular perturbations as PODs, and the difficulty in associating molecular perturbations with actual risk. To increase experience in using these approaches, the proposed framework could be tested in human health risk assessment programs tasked with evaluating chemicals with limited toxicity data, such as the U.S. Environmental Protection Agency’s (EPA) Superfund Health Risk Technical Support Center (STSC; http://www.epa.gov/superfund/health/research.htm). This program evaluates chemicals of interest to Superfund that commonly possess limited toxicity data compared with the generally more data-rich chemicals assessed by the EPA’s Integrated Risk Information System (IRIS). Once a critical mass of chemicals has been evaluated, a reassessment of the framework would identify the strengths and limitations as well as incorporate any subsequent advances in technology or biological knowledge.

Several organizations have proposed using high-throughput in vitro assays for prioritizing chemicals for in vivo testing. In theory, under prioritization, all chemicals would still need to be tested, just in a different order. However, due to resource constraints, chemicals of lower concern may never be tested in practice. In the proposed framework, only a subset of chemicals would likely be elevated for additional testing in successive tiers based on the MOE, resulting in a significant monetary, time, and animal savings. However, it should be noted that depending on the needs of a given regulatory program, the appropriate “capture rate” for testing at each tier could be established, based on transparent delineation of considerations addressing this aspect as one of the components for acceptability of MOEs.

The proposed framework and underlying data sources also have some potential limitations that require improvement during the implementation of the framework. First, the current battery of in vitro assays may not represent one or more critical MOAs, which may falsely identify a chemical as nonselective. In this scenario, the POD for the chemical would be the most sensitive in vitro assay while the chemical may actually selectively activate or inhibit a pathway, process, or macromolecule at a significantly lower dose. In the proposed framework, it is important that the initial screen for selectivity be as comprehensive as possible. To broaden the biological space covered in Tier 1, additional in vitro assays could be included from those reported in PubChem (Li et al., 2010). Alternatively, in vitro transcriptomic measurements could also be performed in concentration-response format across a panel of cell lines that express a diverse range of cellular targets and pathways (Lamb, 2007).

The second limitation that requires improvement is the properties associated with the current ToxCast in vitro assays. The limitations of these properties have been previously discussed (Judson et al., 2011; Thomas et al., 2012a) but include the lack of metabolism, use of single cell types that fail to replicate tissue-level cell-cell interactions, lack of biological context, and the short-term nature of the in vitro assays relative to responses following chronic duration exposure. As these assays are improved, the new assays can be incorporated into the Tier 1 screening process.

A third limitation is related to the exposure modeling across the tiers. In the current state of exposure modeling, the environmental fate and transport models (ie, far field) are generally more mature and have been more thoroughly validated than the indoor and consumer exposure models (ie, near field). Ironically, far-field exposures are usually the least important by orders of magnitude when considering the aggregate exposure to a particular chemical. The lack of well-validated indoor and consumer exposure models can be at least partially attributed to both a lack of data on the use and the relative quantity of specific chemicals in various consumer products and the lack of measurement data for model validation (Egeghy et al., 2012). If exposure information is going to be used to make better decisions on chemical safety in either this framework or more generally, then the science relating to predicting indoor and consumer exposure needs to be improved.

A fourth limitation is adapting these new approaches to risk assessment of chemical mixtures. Current mixtures risk assessment methods focus either on component-based approaches or whole mixture toxicity evaluations (EPA, 2000). The framework described in this manuscript is readily adapted to a components-based approach. For example, for the selective chemicals that target the same protein or pathway, in vivo relative potency factors could be estimated using the results from the associated in vitro assays and IVIVE pharmacokinetic modeling to convert in vitro assay concentrations to applied doses. Mixtures risk estimates could then be derived using a relative potency factor approach similar to the toxic equivalency factors used for dioxins and dioxin-like chemicals (Van den Berg et al., 2006). Alternatively, a hazard index approach could be applied to all chemicals of concern (EPA, 2000). A hazard quotient could be estimated by dividing the estimate of the human exposure by the Tier 1 reference value for each chemical. The cumulative hazard index is then estimated by summing the hazard quotients for all chemicals of concern. If the hazard index is > 1, this would indicate possible exposures of concern. Finally, the new in vitro approaches are also well suited for whole mixtures toxicity evaluation. The advantage of in vitro methods is that large numbers of relevant mixtures could be evaluated for biological activity. The proposed in vitro cumulative risk approach has many of the same limitations as the in vitro risk approach for individual chemicals. Although this proposed in vitro cumulative risk approach is theoretically possible, additional experimental data needs to be collected using this approach prior to application.

In other initiatives, frameworks have or are being developed that should be considered as complementary or congruent to that proposed here. For example, the Health and Environmental Sciences Institute (HESI) RISK21 project has proposed a more general and overarching scheme that easily incorporates the tiered evaluation approaches in this proposed framework. The WHO framework on MOA is currently being updated to reflect increasing experience in hypothesizing potential effects based on information pertaining to putative key events in established modes of action from appropriate in vitro or in silico systems and other evidence.

In summary, the proposed framework provides a risk-based, rapid, cost-effective, and animal-sparing means to evaluate chemicals for safety. The framework represents only an initial step that can be iteratively refined along the way to a fully knowledge-based approach for evaluating chemical safety. Progress toward a better future in chemical safety assessment will require implementation of pragmatic approaches, such as the proposed framework or one similar to it, by national regulatory agencies, cooperation of international organizations such as the Organization for Economic Co-operation and Development (OECD), and stakeholder education and involvement.

ACKNOWLEDGMENTS

The work by R.S.T., M.E.A., H.J.C. III, and B.A.W. on this manuscript was supported by the American Chemistry Council’s Long-Range Research Initiative. The views expressed by authors affiliated with the U.S. Environmental Protection Agency are those of the authors and do not necessarily reflect the views or policies of the U.S. Environmental Protection Agency. This article may be the work product of an employee or group of employees of the National Institute of Environmental Health Sciences (NIEHS), National Institutes of Health (NIH); however, the statements, opinions, or conclusions contained therein do not necessarily represent the statements, opinions, or conclusions of NIEHS, NIH, or the U.S. government. The authors wish to thank Chad Blystone and Alex Merrick for reviewing the manuscript.

REFERENCES

  • Andersen M. E., Krewski D. (2010). The vision of toxicity testing in the 21st century: Moving from discussion to action. Toxicol. Sci. 117, 17–24. [PubMed]
  • Arnot J. A., Brown T. N., Wania F., Breivik K., McLachlan M. S. (2012). Prioritizing chemicals and data requirements for screening-level exposure and risk assessment. Environ. Health Perspect. 120, 1565–1570. [PMC free article] [PubMed]
  • Aylward L. L., Hays S. M. (2011). Consideration of dosimetry in evaluation of ToxCast™ data. J. Appl. Toxicol. 31, 741–751. [PubMed]
  • Aylward L. L., Kirman C. R., Schoeny R., Portier C. J., Hays S. M. (2013). Evaluation of biomonitoring data from the CDC National Exposure Report in a risk assessment context: Perspectives across chemicals. Environ. Health Perspect. 121, 287–294. [PMC free article] [PubMed]
  • Blaauboer B. J. (2010). Biokinetic modeling and in vitro-in vivo exptrapolations. J. Toxicol. Environ. Health. Part B 13, 242–252. [PubMed]
  • Blackburn K., Bjerke D., Daston G., Felter S., Mahony C., Naciff J., Robison S., Wu S. (2011). Case studies to test: A framework for using structural, reactivity, metabolic and physicochemical similarity to evaluate the suitability of analogs for SAR-based toxicological assessments. Regul. Toxicol. Pharmacol. 60, 120–135. [PubMed]
  • Bonn B., Leandersson C., Fontaine F., Zamora I. (2010). Enhanced metabolite identification with MS(E) and a semi-automated software for structural elucidation. Rapid Commun. Mass Spectrom. 24, 3127–3138. [PubMed]
  • Boobis A. R., Cohen S. M., Dellarco V., McGregor D., Meek M. E., Vickers C., Willcocks D., Farland W. (2006). IPCS framework for analyzing the relevance of a cancer mode of action for humans. Crit. Rev. Toxicol. 36, 781–792. [PubMed]
  • Boobis A. R., Doe J. E., Heinrich-Hirsch B., Meek M. E., Munn S., Ruchirawat M., Schlatter J., Seed J., Vickers C. (2008). IPCS framework for analyzing the relevance of a noncancer mode of action for humans. Crit. Rev. Toxicol. 38, 87–96. [PubMed]
  • Caudill S. P. (2010). Characterizing populations of individuals using pooled samples. J. Expo. Sci. Environ. Epidemiol. 20, 29–37. [PubMed]
  • Caudill S. P. (2012). Use of pooled samples from the National Health and Nutrition Examination Survey. Stat. Med. 31, 3269–3277. [PubMed]
  • CDC (2009). Fourth National Report on Human Exposure to Environmental Chemicals (N. C. f. H. Statistics, Ed.) Centers for Disease Control and Prevention, Atlanta, GA.
  • Dourson M. L., Knauf L. A., Swartout J. C. (1992). On reference dose (RfD) and its underlying toxicity data base. Toxicol. Ind. Health 8, 171–189. [PubMed]
  • EFSA (2011). Scientific opinion on genotoxicity testing strategies applicable to food and feed safety assessment. EFSA J. 9, 2379–2447.
  • Egeghy P. P., Judson R., Gangwal S., Mosher S., Smith D., Vail J., Cohen Hubal E. A. (2012). The exposure data landscape for manufactured chemicals. Sci. Total Environ. 414, 159–166. [PubMed]
  • EPA (2000). Supplementary Guidance for Conducting Health Risk Assessment of Chemical Mixtures. U.S. Environmental Protection Agency, Washington, DC: EPA/630/R-00/002.
  • EPA (2003). A Framework for a Computational Toxicology Research Program in ORD. U.S. Environmental Protection Agency, Washington, DC: EPA/600/ R-03/065.
  • EPA (2011). Exposure Factors Handbook. U.S. Environmental Protection Agency, Washington, DC: EPA/600/R-09/052F.
  • Fowler P., Smith R., Smith K., Young J., Jeffrey L., Kirkland D., Pfuhler S., Carmichael P. (2012). Reduction of misleading (“false”) positive results in mammalian cell genotoxicity assays. II. Importance of accurate toxicity measurement. Mutat. Res. 747, 104–117. [PubMed]
  • Gold L. S., Manley N. B., Slone T. H., Ward J. M. (2001). Compendium of chemical carcinogens by target organ: Results of chronic bioassays in rats, mice, hamsters, dogs, and monkeys. Toxicol. Pathol. 29, 639–652. [PubMed]
  • HC (2000). Genotoxicity: A Standard Batter for Genotoxicity Testing of Pharmaceuticals. Therapeutic Products Program, Ottawa, Canada: ICH Topic S2B.
  • Houck K. A., Dix D. J., Judson R. S., Kavlock R. J., Yang J., Berg E. L. (2009). Profiling bioactivity of the ToxCast chemical library using BioMAP primary human cell systems. J. Biomol. Screen. 14, 1054–1066. [PubMed]
  • Hussain S., Arshahd M., Saleem M., Zahir Z. A. (2007). Screening of soil fungi for in vitro degradation of endosulfan. World J. Microbiol. Biotechnol. 23, 939–945.
  • Johnson J. M., Yu T., Strobel F. H., Jones D. P. (2010). A practical approach to detect unique metabolic patterns for personalized medicine. Analyst 135, 2864–2870. [PMC free article] [PubMed]
  • Judson R., Richard A., Dix D. J., Houck K., Martin M., Kavlock R., Dellarco V., Henry T., Holderman T., Sayre P., et al. (2009). The toxicity data landscape for environmental chemicals. Environ. Health Perspect. 117, 685–695. [PMC free article] [PubMed]
  • Judson R. S., Houck K. A., Kavlock R. J., Knudsen T. B., Martin M. T., Mortensen H. M., Reif D. M., Rotroff D. M., Shah I., Richard A. M., et al. (2010). In vitro screening of environmental chemicals for targeted testing prioritization: The ToxCast project. Environ. Health Perspect. 118, 485–492. [PMC free article] [PubMed]
  • Judson R. S., Kavlock R. J., Setzer R. W., Hubal E. A., Martin M. T., Knudsen T. B., Houck K. A., Thomas R. S., Wetmore B. A., Dix D. J. (2011). Estimating toxicity-related biological pathway altering doses for high-throughput chemical risk assessment. Chem. Res. Toxicol. 24, 451–462. [PubMed]
  • Kavlock R., Ankley G. T., Collette T., Francis E., Hammerstrom K., Fowle J., Tilson H., Toth G., Schmieder P., Veith G. D., et al. (2005). Computational toxicology: Framework, partnerships, and program development. September 29-30, 2003, Research Triangle Park, North Carolina. Reprod. Toxicol. 19, 265–280. [PubMed]
  • Kavlock R., Chandler K., Houck K., Hunter S., Judson R., Kleinstreuer N., Knudsen T., Martin M., Padilla S., Reif D., et al. (2012). Update on EPA’s ToxCast program: Providing high throughput decision support tools for chemical risk management. Chem. Res. Toxicol. 25, 1287–1302. [PubMed]
  • Kephalopoulos S., Bruinen de Bruin Y., Arvanitis A., Hakkinen P., Jantunen M. (2007). Issues in consumer exposure modeling: Towards harmonization on a global scale. J. Expo. Sci. Environ. Epidemiol. 17(Suppl. 1), S90–S100. [PubMed]
  • Kerns E. H. (2001). High throughput physicochemical profiling for drug discovery. J. Pharm. Sci. 90, 1838–1858. [PubMed]
  • Knudsen T. B., Houck K. A., Sipes N. S., Singh A. V., Judson R. S., Martin M. T., Weissman A., Kleinstreuer N. C., Mortensen H. M., Reif D. M., et al. (2011). Activity profiles of 309 ToxCast™ chemicals evaluated across 292 biochemical targets. Toxicology 282, 1–15. [PubMed]
  • Knudsen T. B., Martin M. T., Kavlock R. J., Judson R. S., Dix D. J., Singh A. V. (2009). Profiling the activity of environmental chemicals in prenatal developmental toxicity studies using the U.S. EPA’s ToxRefDB. Reprod. Toxicol. 28, 209–219. [PubMed]
  • Lamb J. (2007). The Connectivity Map: A new tool for biomedical research. Nat. Rev. Cancer 7, 54–60. [PubMed]
  • Li Q., Cheng T., Wang Y., Bryant S. H. (2010). PubChem as a public resource for drug discovery. Drug Discov. Today 15, 1052–1057. [PMC free article] [PubMed]
  • Martin M. T., Dix D. J., Judson R. S., Kavlock R. J., Reif D. M., Richard A. M., Rotroff D. M., Romanov S., Medvedev A., Poltoratskaya N., et al. (2010). Impact of environmental chemicals on key transcription regulators and correlation to toxicity end points within EPA’s ToxCast program. Chem. Res. Toxicol. 23, 578–590. [PubMed]
  • Martin M. T., Judson R. S., Reif D. M., Kavlock R. J., Dix D. J. (2009a). Profiling chemicals based on chronic toxicity results from the U.S. EPA ToxRef Database. Environ. Health Perspect. 117, 392–399. [PMC free article] [PubMed]
  • Martin M. T., Mendez E., Corum D. G., Judson R. S., Kavlock R. J., Rotroff D. M., Dix D. J. (2009b). Profiling the reproductive toxicity of chemicals from multigeneration studies in the toxicity reference database. Toxicol. Sci. 110, 181–190. [PubMed]
  • Mattingly C. J., McKone T. E., Callahan M. A., Blake J. A., Hubal E. A. (2012). Providing the missing link: The exposure science ontology ExO. Environ. Sci. Technol. 46, 3046–3053. [PMC free article] [PubMed]
  • Meek M. E., Armstrong V. C. (2007). The management of industrial chemicals in Canada. In Risk Assessment of Chemicals: An introduction (C. J. Van Leeuwen and T. G. Vermeire, Eds.), pp. 591–622Springer, Dordrecht, Netherlands.
  • Meek M. E., Boobis A. R., Crofton K. M., Heinemeyer G., Raaij M. V., Vickers C. (2011). Risk assessment of combined exposure to multiple chemicals: A WHO/IPCS framework. Regul. Toxicol. Pharmacol. 60, S1–S14. [PubMed]
  • Müller L., Kikuchi Y., Probst G., Schechtman L., Shimada H., Sofuni T., Tweats D. (1999). ICH-harmonised guidances on genotoxicity testing of pharmaceuticals: Evolution, reasoning and impact. Mutat. Res. 436, 195–225. [PubMed]
  • NRC (2007). Toxicity Testing in the 21st Century: A Vision and a Strategy. National Research Council of the National Academies, Washington, DC.
  • NTP (2004). A National Toxicology Program for the 21st Century: A Roadmap for the Future. National Toxicology Program, National Institute of Environmental Health Sciences, Research Triangle Park, NC.
  • OECD (2011). Test No. 488: Transgenic Rodent Somatic and Germ Cell Gene Mutation Assays. In OECD Guidelines for the Testing of Chemicals, Section 4. Organisation for Economic Co-operation and Development, Paris, France.
  • Park Y. H., Lee K., Soltow Q. A., Strobel F. H., Brigham K. L., Parker R. E., Wilson M. E., Sutliff R. L., Mansfield K. G., Wachtman L. M., et al. (2012). High-performance metabolic profiling of plasma from seven mammalian species for simultaneous environmental chemical surveillance and bioeffect monitoring. Toxicology 295, 47–55. [PMC free article] [PubMed]
  • Peyret T., Krishnan K. (2011). QSARs for PBPK modelling of environmental contaminants. SAR QSAR Environ. Res. 22, 129–169. [PubMed]
  • Peyret T., Poulin P., Krishnan K. (2010). A unified algorithm for predicting partition coefficients for PBPK modeling of drugs and environmental chemicals. Toxicol. Appl. Pharmacol. 249, 197–207. [PubMed]
  • Poulin P., Theil F. P. (2002). Prediction of pharmacokinetics prior to in vivo studies. 1. Mechanism-based prediction of volume of distribution. J. Pharm. Sci. 91, 129–156. [PubMed]
  • Reif D. M., Martin M. T., Tan S. W., Houck K. A., Judson R. S., Richard A. M., Knudsen T. B., Dix D. J., Kavlock R. J. (2010). Endocrine profiling and prioritization of environmental chemicals using ToxCast data. Environ. Health Perspect. 118, 1714–1720. [PMC free article] [PubMed]
  • Rodgers T., Leahy D., Rowland M. (2005). Physiologically based pharmacokinetic modeling 1: Predicting the tissue distribution of moderate-to-strong bases. J. Pharm. Sci. 94, 1259–1276. [PubMed]
  • Rodgers T., Rowland M. (2006). Physiologically based pharmacokinetic modelling 2: Predicting the tissue distribution of acids, very weak bases, neutrals and zwitterions. J. Pharm. Sci. 95, 1238–1257. [PubMed]
  • Rodgers T., Rowland M. (2007). Mechanistic approaches to volume of distribution predictions: Understanding the processes. Pharm. Res. 24, 918–933. [PubMed]
  • Rotroff D. M., Beam A. L., Dix D. J., Farmer A., Freeman K. M., Houck K. A., Judson R. S., LeCluyse E. L., Martin M. T., Reif D. M., et al. (2010a). Xenobiotic-metabolizing enzyme and transporter gene expression in primary cultures of human hepatocytes modulated by ToxCast chemicals. J. Toxicol. Environ. Health. B. Crit. Rev. 13, 329–346. [PubMed]
  • Rotroff D. M., Wetmore B. A., Dix D. J., Ferguson S. S., Clewell H. J., Houck K. A., Lecluyse E. L., Andersen M. E., Judson R. S., Smith C. M., et al. (2010b). Incorporating human dosimetry and exposure into high-throughput in vitro toxicity screening. Toxicol. Sci. 117, 348–358. [PubMed]
  • Saghir S. A., Mendrala A. L., Bartels M. J., Day S. J., Hansen S. C., Sushynski J. M., Bus J. S. (2006). Strategies to assess systemic exposure of chemicals in subchronic/chronic diet and drinking water studies. Toxicol. Appl. Pharmacol. 211, 245–260. [PubMed]
  • Soltow Q. A., Strobel F. H., Mansfield K. G., Wachtman L. M., Park Y., Jones D. P. (2011). High-performance metabolic profiling with dual chromatography-Fourier-transform mass spectrometry (DC-FTMS) for study of the exposome. Metabolomics 9, 132–143.
  • Sonich-Mullin C., Fielder R., Wiltse J., Baetcke K., Dempsey J., Fenner-Crisp P., Grant D., Hartley M., Knaap A., Kroese D., et al. ; International Programme on Chemical Safety (2001). IPCS conceptual framework for evaluating a mode of action for chemical carcinogenesis. Regul. Toxicol. Pharmacol. 34, 146–152. [PubMed]
  • Tan Y. M., Liao K. H., Clewell H. J., 3rd (2007). Reverse dosimetry: Interpreting trihalomethanes biomonitoring data using physiologically based pharmacokinetic modeling. J. Expo. Sci. Environ. Epidemiol. 17, 591–603. [PubMed]
  • Thomas R. S., Allen B. C., Nong A., Yang L., Bermudez E., Clewell H. J., 3rd, Andersen M. E. (2007). A method to integrate benchmark dose estimates with genomic data to assess the functional effects of chemical exposure. Toxicol. Sci. 98, 240–248. [PubMed]
  • Thomas R. S., Black M. B., Li L., Healy E., Chu T. M., Bao W., Andersen M. E., Wolfinger R. D. (2012a). A comprehensive statistical analysis of predicting in vivo hazard using high-throughput in vitro screening. Toxicol. Sci. 128, 398–417. [PubMed]
  • Thomas R. S., Clewell H. J., 3rd, Allen B. C., Wesselkamper S. C., Wang N. C., Lambert J. C., Hess-Wilson J. K., Zhao Q. J., Andersen M. E. (2011). Application of transcriptional benchmark dose values in quantitative cancer and noncancer risk assessment. Toxicol. Sci. 120, 194–205. [PubMed]
  • Thomas R. S., Clewell H. J., 3rd, Allen B. C., Yang L., Healy E., Andersen M. E. (2012b). Integrating pathway-based transcriptomic data into quantitative chemical risk assessment: A five chemical case study. Mutat. Res. 746, 135–143. [PubMed]
  • Thomas R. S., Wesselkamper S. C., Wang N. C., Zhao Q. J., Petersen D. D., Lambert J. C., Cote I., Yang L., Healy E., Black M. B., et al. (2013). Temporal concordance between apical and transcriptional points of departure for chemical risk assessment. Toxicol. Sci. 134, 180–194. [PubMed]
  • Van den Berg M., Birnbaum L. S., Denison M., De Vito M., Farland W., Feeley M., Fiedler H., Hakansson H., Hanberg A., Haws L., et al. (2006). The 2005 World Health Organization reevaluation of human and Mammalian toxic equivalency factors for dioxins and dioxin-like compounds. Toxicol. Sci. 93, 223–241. [PMC free article] [PubMed]
  • Wambaugh J. F., Setzer R. W., Reif D. M., Gangwal S., Mitchell-Blackwood J., Arnot J. A., Joliet O., Frame A., Rabinowitz J., Knudsen T. B., et al. (2013). High-throughput models for exposure-based chemical prioritization in the ExpoCast project. Environ. Sci. Technol. 47, 8479–8488. [PubMed]
  • Wetmore B. A., Wambaugh J. F., Ferguson S. S., Li L., Clewell H. J., 3rd, Judson R. S., Freeman K., Bao W., Sochaski M. A., Chu T. M., et al. (2012a). Relative impact of incorporating pharmacokinetics on predicting in vivo hazard and mode of action from high-throughput in vitro toxicity assays. Toxicol. Sci. 132, 327–346. [PubMed]
  • Wetmore B. A., Wambaugh J. F., Ferguson S. S., Sochaski M. A., Rotroff D. M., Freeman K., Clewell H. J., 3rd, Dix D. J., Andersen M. E., Houck K. A., et al. (2012b). Integration of dosimetry, exposure, and high-throughput screening data in chemical toxicity assessment. Toxicol. Sci. 125, 157–174. [PubMed]

Articles from Toxicological Sciences are provided here courtesy of Oxford University Press