A value of information framework for assessing the trade-offs associated with uncertainty, duration, and cost of chemical toxicity testing.

A number of investigators have explored the use of value of information (VOI) analysis to evaluate alternative information collection procedures in diverse decision-making contexts. This paper presents an analytic framework for determining the value of toxicity information used in risk-based decision making. The framework is specifically designed to explore the trade-offs between cost, timeliness, and uncertainty reduction associated with different toxicity-testing methodologies. The use of the proposed framework is demonstrated by two illustrative applications which, although based on simplified assumptions, show the insights that can be obtained through the use of VOI analysis. Specifically, these results suggest that timeliness of information collection has a significant impact on estimates of the VOI of chemical toxicity tests, even in the presence of smaller reductions in uncertainty. The framework introduces the concept of the expected value of delayed sample information, as an extension to the usual expected value of sample information, to accommodate the reductions in value resulting from delayed decision making. Our analysis also suggests that lower cost and higher throughput testing also may be beneficial in terms of public health benefits by increasing the number of substances that can be evaluated within a given budget. When the relative value is expressed in terms of return-on-investment per testing strategy, the differences can be substantial.

A Framework that Considers the Impacts of Time, Cost, and Uncertainty in the Determination of the Cost Effectiveness of Toxicity-Testing Methodologies.

Regulatory agencies are required to evaluate the impacts of thousands of chemicals. Toxicological tests currently used in such evaluations are time-consuming and resource intensive; however, advances in toxicology and related fields are providing new testing methodologies that reduce the cost and time required for testing. The selection of a preferred methodology is challenging because the new methodologies vary in duration and cost, and the data they generate vary in the level of uncertainty. This article presents a framework for performing cost-effectiveness analyses (CEAs) of toxicity tests that account for cost, duration, and uncertainty. This is achieved by using an output metric-the cost per correct regulatory decision-that reflects the three elements. The framework is demonstrated in two example CEAs, one for a simple decision of risk acceptability and a second, more complex decision, involving the selection of regulatory actions. Each example CEA evaluates five hypothetical toxicity-testing methodologies which differ with respect to cost, time, and uncertainty. The results of the examples indicate that either a fivefold reduction in cost or duration can be a larger driver of the selection of an optimal toxicity-testing methodology than a fivefold reduction in uncertainty. Uncertainty becomes of similar importance to cost and duration when decisionmakers are required to make more complex decisions that require the determination of small differences in risk predictions. The framework presented in this article may provide a useful basis for the identification of cost-effective methods for toxicity testing of large numbers of chemicals.

Data Mining Approaches for Assessing Chemical Coexposures Using Consumer Product Purchase Data.

The use of consumer products presents a potential for chemical exposures to humans. Toxicity testing and exposure models are routinely employed to estimate risks from their use; however, a key challenge is the sparseness of information concerning who uses products and which products are used contemporaneously. Our goal was to demonstrate a method to infer use patterns by way of purchase data. We examined purchase patterns for three types of personal care products (cosmetics, hair care, and skin care) and two household care products (household cleaners and laundry supplies) using data from 60,000 households collected over a one-year period in 2012. The market basket analysis methodology frequent itemset mining (FIM) was used to identify co-occurring sets of product purchases for all households and demographic groups based on income, education, race/ethnicity, and family composition. Our methodology captured robust co-occurrence patterns for personal and household products, globally and for different demographic groups. FIM identified cosmetic co-occurrence patterns captured in prior surveys of cosmetic use, as well as a trend of increased diversity of cosmetic purchases as children mature to teenage years. We propose that consumer product purchase data can be mined to inform person-oriented use patterns for high-throughput chemical screening applications, for aggregate and combined chemical risk evaluations.

Consensus Modeling of Median Chemical Intake for the U.S. Population Based on Predictions of Exposure Pathways.

Prioritizing the potential risk posed to human health by chemicals requires tools that can estimate exposure from limited information. In this study, chemical structure and physicochemical properties were used to predict the probability that a chemical might be associated with any of four exposure pathways leading from sources-consumer (near-field), dietary, far-field industrial, and far-field pesticide-to the general population. The balanced accuracies of these source-based exposure pathway models range from 73 to 81%, with the error rate for identifying positive chemicals ranging from 17 to 36%. We then used exposure pathways to organize predictions from 13 different exposure models as well as other predictors of human intake rates. We created a consensus, meta-model using the Systematic Empirical Evaluation of Models framework in which the predictors of exposure were combined by pathway and weighted according to predictive ability for chemical intake rates inferred from human biomonitoring data for 114 chemicals. The consensus model yields an R2 of ∼0.8. We extrapolate to predict relevant pathway(s), median intake rate, and credible interval for 479 926 chemicals, mostly with minimal exposure information. This approach identifies 1880 chemicals for which the median population intake rates may exceed 0.1 mg/kg bodyweight/day, while there is 95% confidence that the median intake rate is below 1 µg/kg BW/day for 474572 compounds.

Temporal Trends in Exposures to Six Phthalates from Biomonitoring Data: Implications for Cumulative Risk.

Phthalates are used in a wide range of consumer goods, resulting in exposures to specific phthalates that vary over time in accordance with changes in product use and how phthalates are utilized. We investigated trends in estimates of daily intake dose and several cumulative risk metrics, including the Hazard Quotient (HQ), Hazard Index (HI), and Maximum Cumulative Ratio (MCR) for six phthalates from 2005 to 2014 using metabolite biomonitoring data collected from spot urine samples under the National Health and Nutrition Examination Survey (NHANES). Over this period, there was a 2.2-fold decrease in the mean HI (0.34 to 0.15) and a 7.2-fold decrease in the percentage of participants with an HI > 1 (5.7% to 0.8%), indicating an overall decrease in combined exposure to these phthalates. Children (aged 6-11 years) had higher mean HI values than either adolescents (aged 12-19 years) or adults (aged 20+ years) during this period. MCR values were generally low and inversely correlated with HI. This indicated that a single phthalate usually drove the hazards for highly exposed individuals. However, the average value of MCR increased 1.2-fold (1.7-2.1) over this period indicating an increasing need to consider exposures to multiple phthalates in this group.

Use of a probabilistic PBPK/PD model to calculate Data Derived Extrapolation Factors for chlorpyrifos.

A physiologically based pharmacokinetic and pharmacodynamic (PBPK/PD) model combined with Monte Carlo analysis of inter-individual variation was used to assess the effects of the insecticide, chlorpyrifos and its active metabolite, chlorpyrifos oxon in humans. The PBPK/PD model has previously been validated and used to describe physiological changes in typical individuals as they grow from birth to adulthood. This model was updated to include physiological and metabolic changes that occur with pregnancy. The model was then used to assess the impact of inter-individual variability in physiology and biochemistry on predictions of internal dose metrics and quantitatively assess the impact of major sources of parameter uncertainty and biological diversity on the pharmacodynamics of red blood cell acetylcholinesterase inhibition. These metrics were determined in potentially sensitive populations of infants, adult women, pregnant women, and a combined population of adult men and women. The parameters primarily responsible for inter-individual variation in RBC acetylcholinesterase inhibition were related to metabolic clearance of CPF and CPF-oxon. Data Derived Extrapolation Factors that address intra-species physiology and biochemistry to replace uncertainty factors with quantitative differences in metrics were developed in these same populations. The DDEFs were less than 4 for all populations. These data and modeling approach will be useful in ongoing and future human health risk assessments for CPF and could be used for other chemicals with potential human exposure.

Use of the Maximum Cumulative Ratio As an Approach for Prioritizing Aquatic Coexposure to Plant Protection Products: A Case Study of a Large Surface Water Monitoring Database.

This paper uses the maximum cumulative ratio (MCR) as part of a tiered approach to evaluate and prioritize the risk of acute ecological effects from combined exposures to the plant protection products (PPPs) measured in 3 099 surface water samples taken from across the United States. Assessments of the reported mixtures performed on a substance-by-substance approach and using a Tier One cumulative assessment based on the lowest acute ecotoxicity benchmark gave the same findings for 92.3% of the mixtures. These mixtures either did not indicate a potential risk for acute effects or included one or more individual PPPs that had concentrations in excess of their benchmarks. A Tier Two assessment using a trophic level approach was applied to evaluate the remaining 7.7% of the mixtures. This assessment reduced the number of mixtures of concern by eliminating the combination of endpoint from multiple trophic levels, identified invertebrates and nonvascular plants as the most susceptible nontarget organisms, and indicated that a only a very limited number of PPPs drove the potential concerns. The combination of the measures of cumulative risk and the MCR enabled the identification of a small subset of mixtures where a potential risk would be missed in substance-by-substance assessments.

Conceptual Framework To Extend Life Cycle Assessment Using Near-Field Human Exposure Modeling and High-Throughput Tools for Chemicals.

Life Cycle Assessment (LCA) is a decision-making tool that accounts for multiple impacts across the life cycle of a product or service. This paper presents a conceptual framework to integrate human health impact assessment with risk screening approaches to extend LCA to include near-field chemical sources (e.g., those originating from consumer products and building materials) that have traditionally been excluded from LCA. A new generation of rapid human exposure modeling and high-throughput toxicity testing is transforming chemical risk prioritization and provides an opportunity for integration of screening-level risk assessment (RA) with LCA. The combined LCA and RA approach considers environmental impacts of products alongside risks to human health, which is consistent with regulatory frameworks addressing RA within a sustainability mindset. A case study is presented to juxtapose LCA and risk screening approaches for a chemical used in a consumer product. The case study demonstrates how these new risk screening tools can be used to inform toxicity impact estimates in LCA and highlights needs for future research. The framework provides a basis for developing tools and methods to support decision making on the use of chemicals in products.

Risk Assessment of Combined Exposure to Multiple Chemicals at the European Food Safety Authority: Principles, Guidance Documents, Applications and Future Challenges.

Human health and animal health risk assessment of combined exposure to multiple chemicals use the same steps as single-substance risk assessment, namely problem formulation, exposure assessment, hazard assessment and risk characterisation. The main unique feature of combined RA is the assessment of combined exposure, toxicity and risk. Recently, the Scientific Committee of the European Food Safety Authority (EFSA) published two relevant guidance documents. The first one "Harmonised methodologies for the human health, animal health and ecological risk assessment of combined exposure to multiple chemicals" provides principles and explores methodologies for all steps of risk assessment together with a reporting table. This guidance supports also the default assumption that dose addition is applied for combined toxicity of the chemicals unless evidence for response addition or interactions (antagonism or synergism) is available. The second guidance document provides an account of the scientific criteria to group chemicals in assessment groups using hazard-driven criteria and prioritisation methods, i.e., exposure-driven and risk-based approaches. This manuscript describes such principles, provides a brief description of EFSA's guidance documents, examples of applications in the human health and animal health area and concludes with a discussion on future challenges in this field.

A reanalysis of the evidence for increased efficiency in benzene metabolism at airborne exposure levels below 3 p.p.m.

An analysis of monitoring data on workers in Tianjin, China, reported a 9-fold increase in the production of benzene metabolites per unit exposure as air concentrations declined from 88.9 to 0.03 p.p.m. The increase is attributed to an enhanced efficiency of benzene metabolism at lower air concentrations. This finding, however, is not consistent with other studies demonstrating that adsorbed benzene is almost completely metabolized at airborne levels ranging from <1 to 70 p.p.m. In this article (i) the modeling performed in Kim et al. is repeated and the model predictions are reproduced; (ii) the impacts of technical issues in the corrections for background levels of metabolites, accounting for biases in the regression modeling, and the uncertainties introduced by the use of a calibration model to estimate benzene air levels for certain workers are evaluated and (iii) alternative methods of correcting for background levels of metabolites are examined. The new analysis indicates that findings of increased production are probably smaller and are highly uncertain, 4.8 fold [0.1-18] (mean and [95% confidence limits]). Defining background levels as either the levels in all workers with no occupational exposures or in workers with predicted air levels of <0.03 p.p.m. results in estimates of 2.4 fold [<0.1-15] and 3.3 fold [<0.1-19] increases, respectively. Based on this reanalysis, the Tianjin data appear to be too uncertain to support any conclusions of a change in the efficiency of benzene metabolism with variations in exposure.

Statistical methodology to determine kinetically derived maximum tolerated dose in repeat dose toxicity studies.

Several statistical approaches were evaluated to identify an optimum method for determining a point of nonlinearity (PONL) in toxicokinetic data. (1) A second-order least squares regression model was fit iteratively starting with data from all doses. If the second order term was significant (α<0.05), the dataset was reevaluated with successive removal of the highest dose until the second-order term became non-significant. This dose, whose removal made the second order term non-significant, is an estimate of the PONL. (2) A least squares linear model was fit iteratively starting with data from all doses except the highest. The mean response for the omitted dose was compared to the 95% prediction interval. If the omitted dose falls outside the confidence interval it is an estimate of the PONL. (3) Slopes of least squares linear regression lines for sections of contiguous doses were compared. Nonlinearity was suggested when slopes of compared sections differed. A total of 33 dose-response datasets were evaluated. For these toxicokinetic data, the best statistical approach was the least squares regression analysis with a second-order term. Changing the α level for the second-order term and weighting the second-order analysis by the inverse of feed consumption were also considered. This technique has been shown to give reproducible identification of nonlinearities in TK datasets.

Linear low-dose extrapolation for noncancer heath effects is the exception, not the rule.

The nature of the exposure-response relationship has a profound influence on risk analyses. Several arguments have been proffered as to why all exposure-response relationships for both cancer and noncarcinogenic endpoints should be assumed to be linear at low doses. We focused on three arguments that have been put forth for noncarcinogens. First, the general "additivity-to-background" argument proposes that if an agent enhances an already existing disease-causing process, then even small exposures increase disease incidence in a linear manner. This only holds if it is related to a specific mode of action that has nonuniversal properties-properties that would not be expected for most noncancer effects. Second, the "heterogeneity in the population" argument states that variations in sensitivity among members of the target population tend to "flatten out and linearize" the exposure-response curve, but this actually only tends to broaden, not linearize, the dose-response relationship. Third, it has been argued that a review of epidemiological evidence shows linear or no-threshold effects at low exposures in humans, despite nonlinear exposure-response in the experimental dose range in animal testing for similar endpoints. It is more likely that this is attributable to exposure measurement error rather than a true nonthreshold association. Assuming that every chemical is toxic at high exposures and linear at low exposures does not comport to modern-day scientific knowledge of biology. There is no compelling evidence-based justification for a general low-exposure linearity; rather, case-specific mechanistic arguments are needed.

Development of a source-to-outcome model for dietary exposures to insecticide residues: an example using chlorpyrifos.

Probabilistic models of interindividual variation in exposure and response were linked to create a source-to-outcome population model. This model was used to investigate cholinesterase inhibition from dietary exposures to an insecticide (chlorpyrifos) in populations of adults and 3 year old children. A physiologically based pharmacokinetic and pharmacodynamic (PBPK/PD) model was used to calculate the variation in sensitivity occurring from interindividual variability in physiology, metabolism, and physical activity levels. A dietary intake model characterizes the variation in dietary insecticide exposures and variation in anthropometry in the populations. Published equations were used to describe the necessary physiology for each simulated individual based on the anthropometry from the dietary intake model. The model of the interindividual variation in response to chlorpyrifos was developed by performing a sensitivity analysis on the PBPK/PD model to determine the parameters that drive variation in pharmacodynamics outcomes (brain and red blood cell acetylcholinesterase inhibition). Distributions of interindividual variation were developed for parameters with the largest impact; the probabilistic model sampled from these distributions. The impact of age and interindividual variation on sensitivity at the doses that occur from dietary exposures, typically orders of magnitude lower than exposures assessed in toxicological studies, was assessed using the source-to-outcome model. The resulting simulations demonstrated that metabolic detoxification capacity was sufficient to prevent significant brain and red blood cell acetylcholinesterase inhibition, even in individuals with the lowest detoxification potential. Age-specific pharmacokinetic and pharmacodynamic parameters did not predict differences in susceptibility between adults and children. In the future, the approach of this case study could be used to assess the risks from low level exposures to other chemicals.

Application of a source-to-outcome model for the assessment of health impacts from dietary exposures to insecticide residues.

The paper presents a case study of the application of a "source-to-outcome" model for the evaluation of the health outcomes from dietary exposures to an insecticide, chlorpyrifos, in populations of adults (age 30) and children (age 3). The model is based on publically-available software programs that characterize the longitudinal dietary exposure and anthropometry of exposed individuals. These predictions are applied to a validated PBPK/PD model to estimate interindividual and longitudinal variation in brain and RBC AChE inhibition (key events) and chlorpyrifos concentrations in blood and TCPy in urine (biomarkers of exposure). The predicted levels of chlorpyrifos and TCPy are compared to published measurements of the biomarkers. Predictions of RBC AChE are compared to levels of inhibition associated with reported exposure-related effects in humans to determine the potential for the occurrence of adverse cholinergic effects. The predicted distributions of chlorpyrifos in blood and TCPy in urine were found to be reasonably consistent with published values, supporting the predictive value of the exposure and PBPK portions of the source-to-outcome model. Key sources of uncertainty in predictions of dietary exposures were investigated and found to have a modest impact on the model predictions. Future versions of this source-to-outcome model can be developed that consider advances in our understanding of metabolism, to extend the approach to other age groups (infants), and address intakes from other routes of exposure.

The Residential Population Generator (RPGen): Parameterization of Residential, Demographic, and Physiological Data to Model Intraindividual Exposure, Dose, and Risk.

Exposure to chemicals is influenced by associations between the individual's location and activities as well as demographic and physiological characteristics. Currently, many exposure models simulate individuals by drawing distributions from population-level data or use exposure factors for single individuals. The Residential Population Generator (RPGen) binds US surveys of individuals and households and combines the population with physiological characteristics to create a synthetic population. In general, the model must be supported by internal consistency; i.e., values that could have come from a single individual. In addition, intraindividual variation must be representative of the variation present in the modeled population. This is performed by linking individuals and similar households across income, location, family type, and house type. Physiological data are generated by linking census data to National Health and Nutrition Examination Survey data with a model of interindividual variation of parameters used in toxicokinetic modeling. The final modeled population data parameters include characteristics of the individual's community (region, state, urban or rural), residence (size of property, size of home, number of rooms), demographics (age, ethnicity, income, gender), and physiology (body weight, skin surface area, breathing rate, cardiac output, blood volume, and volumes for body compartments and organs). RPGen output is used to support user-developed chemical exposure models that estimate intraindividual exposure in a desired population. By creating profiles and characteristics that determine exposure, synthetic populations produced by RPGen increases the ability of modelers to identify subgroups potentially vulnerable to chemical exposures. To demonstrate application, RPGen is used to estimate exposure to Toluene in an exposure modeling case example.

Calibrating an agent-based model of longitudinal human activity patterns using the Consolidated Human Activity Database.

Patterns of human behavior over extended periods of time are important for characterizing human exposure to hazardous chemicals. Because longitudinal behavior patterns for an individual are difficult to obtain, exposure-assessors have characterized such patterns by linking daily records from multiple individuals. In an earlier publication, we developed an alternative strategy that was based on agent-based simulation modeling. Specifically, we created a software program, Agent-Based Model of Human Activity Patterns (ABMHAP), that generates year-long longitudinal behavior patterns. In this paper, we both calibrate and evaluate ABMHAP using human behavior data from the U.S. Environmental Protection Agency's Consolidated Human Activity Database (CHAD). We use the longitudinal data (data on individuals' activities over multiple days) in CHAD to parameterize ABMHAP, and we use single-day behavior data from CHAD to evaluate ABMHAP predictions. We evaluate ABMHAP's ability to simulate sleeping, eating, commuting, and working (or attending school) for four populations: working adults, nonworking adults, school-age children, and preschool children. The results demonstrate that ABMHAP, when parameterized with empirical data, can capture both interindividual and intraindividual variation in behaviors in different types of individuals. We propose that simulating annual activity patterns via ABMHAP may allow exposure-assessors to characterize exposure-related behavior in ways not possible with traditional survey methods.

Organizing mechanism-related information on chemical interactions using a framework based on the aggregate exposure and adverse outcome pathways.

This paper presents a framework for organizing and accessing mechanistic data on chemical interactions. The framework is designed to support the assessment of risks from combined chemical exposures. The framework covers interactions between chemicals that occur over the entire source-to-outcome continuum including interactions that are studied in the fields of chemical transport, environmental fate, exposure assessment, dosimetry, and individual and population-based adverse outcomes. The framework proposes to organize data using a semantic triple of a chemical (subject), has impact (predicate), and a causal event on the source-to-outcome continuum of a second chemical (object). The location of the causal event on the source-to-outcome continuum and the nature of the impact are used as the basis for a taxonomy of interactions. The approach also builds on concepts from the Aggregate Exposure Pathway (AEP) and Adverse Outcome Pathway (AOP). The framework proposes the linking of AEPs of multiple chemicals and the AOP networks relevant to those chemicals to form AEP-AOP networks that describe chemical interactions that cannot be characterized using AOP networks alone. Such AEP-AOP networks will aid the construction of workflows for both experimental design and the systematic review or evaluation performed in risk assessments. Finally, the framework is used to link the constructs of existing component-based approaches for mixture toxicology to specific categories in the interaction taxonomy.

Establishing a system of consumer product use categories to support rapid modeling of human exposure.

Consumer product categorizations for use in predicting human chemical exposure provide a bridge between product composition data and consumer product use pattern information. Furthermore, the categories reflect other factors relevant to developing consumer product exposure scenarios, such as microenvironment of use (e.g., indoors or outdoors), method of application/form of release (e.g., spray versus liquid), release to various media, removal processes (e.g., rinse-off or wipe-off), and route-specific exposure factors (dermal surface areas of application, fraction of release in respirable form). While challenging, developing harmonized product categories can generalize the factors described above allowing for rapid parameterization of route-specific exposure scenario algorithms for new chemical/product applications and efficient utilization of new data on product use or composition. This can be accomplished via mapping product categories to likewise categorized release and use patterns or exposure factors. Here, hierarchical product use categories (PUCs) for consumer products that provide such mappings are presented and crosswalked with other internationally harmonized product categories for consumer exposure assessment. The PUCs were defined by applying use and exposure scenario information to the products in EPA's Chemical and Products Database (CPDat). This paper demonstrates how these PUCs are being used to rapidly parameterize algorithms for scenario-specific use, fate, and exposure in a probabilistic aggregate model of human exposure to chemicals used in consumer products. The PUCs provide a generic representation of consumer products for use in exposure assessment and provide an efficient framework for flexible and rapid data reporting and consumer exposure model parameterization.

Simulating exposure-related behaviors using agent-based models embedded with needs-based artificial intelligence.

Exposure to a chemical is a critical consideration in the assessment of risk, as it adds real-world context to toxicological information. Descriptions of where and how individuals spend their time are important for characterizing exposures to chemicals in consumer products and in indoor environments. Herein we create an agent-based model (ABM) that simulates longitudinal patterns in human behavior. By basing the ABM upon an artificial intelligence (AI) system, we create agents that mimic human decisions on performing behaviors relevant for determining exposures to chemicals and other stressors. We implement the ABM in a computer program called the Agent-Based Model of Human Activity Patterns (ABMHAP) that predicts the longitudinal patterns for sleeping, eating, commuting, and working. We then show that ABMHAP is capable of simulating behavior over extended periods of time. We propose that this framework, and models based on it, can generate longitudinal human behavior data for use in exposure assessments.

Characterizing the noncancer toxicity of mixtures using concepts from the TTC and quantitative models of uncertainty in mixture toxicity.

This article explores the use of an approach for setting default values for the noncancer toxicity, developed as part of the Threshold of Toxicological Concern (TTC), for the evaluation of the chronic noncarcinogenic effects of certain chemical mixtures. Individuals are exposed to many mixtures where there are little or no toxicological data on some or all of the mixture components. The approach developed in the TTC can provide a basis for conservative estimates of the toxicity of the mixture components when compound-specific data are not available. The application of this approach to multiple chemicals in a mixture, however, has implications for the statistical assumptions made in developing component-based estimates of mixtures. Specifically, conservative assumptions that are appropriate for one compound may become overly conservative when applied to all components of a mixture. This overestimation can be investigated by modeling the uncertainty in toxicity standards. In this article the approach is applied to both hypothetical and actual examples of chemical mixtures and the potential for overestimation is investigated. The results indicate that the use of the approach leads to conservative estimates of mixture toxicity and therefore its use is most appropriate for screening assessments of mixtures.