Explore the Dosimetric Relationship between the Intake of Chemical Contaminants and Their Occurrence in Blood and Urine

Abstract

The dosimetric relationship between the human intake dose of a chemical contaminant (an “external dose”) and its concentrations in bodily fluids such as blood and urine (related to an “internal dose”), often characterized by a dose-to-concentration ratio, has critical applications in exposure science, toxicology, and risk assessment, especially in the “new approach methods” era. However, there is a lack of a mechanistic, systematic understanding of how such a dosimetric relationship depends on fundamental chemical properties, such as partition coefficients and biotransformation half-lives. Here, we investigate this issue using a well-evaluated toxicokinetic model, which links external and internal doses by quantifying the absorption and elimination of chemicals. Results are visualized in a series of chemical partitioning space plots, whereby a chemical’s dose-to-concentration ratio can be approximately predicted based on its partitioning between air, water, and octanol phases. Our results indicate that when taken in equal doses, chemicals with low volatility and moderate to high hydrophobicity exhibit the highest concentrations in the blood, and chemicals undergoing significant biotransformation tend to exhibit lower concentrations in comparison to their counterparts undergoing negligible biotransformation but possessing similar partitioning properties. Chemicals with high hydrophilicity have the highest concentrations in urine. Such revealed property dependence is similar for both adults and children and for individuals with normal body weights and with obesity. Overall, insights gained from this study are important in predicting blood and urinary concentrations from exposure information and in determining the exposure rate that produces the blood or urinary concentrations observed in biomonitoring studies.

Short abstract

A chemical’s partitioning and biotransformation properties control how its occurrence in blood or urine responds to the external intake dose.

1. Introduction

Every day, humans around the world are exposed to a variety of chemicals emitted from chemical production, use, and waste disposal.¹ Under chronic exposure, the levels of these chemicals in bodily fluids, such as blood and urine, are likely to be correlated to the levels at the site of target organs and hormone receptors, which govern the possibility and severity of health impacts. For this reason, it is vital to understand the dosimetric relationship between the human intake dose of a chemical (an “external dose”) and its resulting concentrations within bodily fluids (related to an “internal dose”). For example, biomonitoring studies measure chemical levels present in the human body, whereas toxicological benchmarks derived from in vivo animal studies and adjusted for various uncertainty factors, such as interspecies extrapolation, are often expressed as external doses representing safe levels of human exposure to chemicals. Examples of such toxicological benchmarks include the reference dose recommended by the U.S. Environmental Protection Agency and the minimal risk level used by the U.S. Agency for Toxic Substances and Disease Registry. In such a case, quantitative knowledge of the dosimetric relationship enables converting toxicological benchmarks to corresponding concentrations in the human body, often termed “biomonitoring equivalents” (BEs), bridging the gap between toxicological and biomonitoring data to inform health risks.^4,5 Additionally, such a dosimetric relationship can be used in reverse dosimetry to back calculate the initial dose of a compound that supports generating chemical concentrations in blood and urine, notably concentrations that can produce adverse bioactivity.²⁻⁷ A prominent example is the U.S. Environmental Protection Agency’s ExpoCast program, wherein urinary concentrations of chemical metabolites reported in the National Health and Nutrition Examination Survey (NHANES) biomonitoring program were used to predict the average daily intake doses of their original parent compounds.⁷ In addition, the dosimetric relationship allows converting route-specific (such as ingestion or inhalation) toxicological potency (e.g., the threshold of toxicological concern) into blood concentrations.⁸⁻¹⁰ With a widely accepted notion that doses from different routes are equivalent if they yield the same blood concentration, this conversion renders doses from different exposure routes comparable and extrapolatable.^9,11 This lays a quantitative basis for route-to-route extrapolation between route-specific toxicological benchmark doses.^9,11

When assessing the dosimetric relationship, it is important to consider the chemical properties of different compounds, including partition coefficients that describe a chemical’s hydrophobicity and volatility as well as biotransformation half-lives that describe a chemical’s resistance to metabolism in the human body. These properties are important because they affect how each compound interacts with internal absorption barriers and different mechanisms of elimination. Earlier studies on cows¹²⁻¹⁵ have elucidated the dependence of the dosimetric relationship, characterized by a “biotransfer factor” defined as the ratio of chemical concentration in beef or milk to intake dose through feeding, on the octanol–water partition coefficient (K_OW) that quantifies hydrophobicity.¹²⁻¹⁵ These studies showed that the biotransfer factors in beef and milk peaked at around a log K_OW of 6, owing to the efficient elimination of highly hydrophilic chemicals and minimal absorption of highly hydrophobic chemicals. That being said, those studies considered only partitioning properties and ignored biotransformation and other toxicokinetic processes. As such, their results apply only to those chemicals that undergo negligible biotransformation, such as persistent organic pollutants, and hence have limited applicability to chemicals with diverse biotransformation capabilities.

The contribution of fundamental chemical properties to the dosimetric relationship and the possible impacts of physiological factors (often age- and body weight-dependent) are still unclear. Since these chemical properties can often be obtained through laboratory testing or computational theory at an economical cost, revealing their quantitative connection to the dosimetric relationship enables us to forecast the bioaccumulative tendencies of different compounds. This helps in screening and prioritizing the vast number of chemicals on the market for their potential for human exposure and body accumulation. Earlier animal-based studies^12,16 used empirical statistical approaches, such as linear regression, to account for the impacts of one individual variable at a time, leading to a lack of inclusion of the combined effects of chemical properties and physiological factors. Essentially, this limitation can be addressed by mechanistic modeling, i.e., using physiologically based toxicokinetic models building on chemical and physiological processes by considering interactions among multiple variables. In most recent studies, toxicokinetic models were adopted to calculate the dosimetric relationship individually for several chemicals.¹⁷⁻¹⁹ A remarkable variability of at least 3 orders of magnitude in the calculated dosimetric relationships in these discrete case studies warrants a mechanistic, systematic understanding of how chemical properties govern the dosimetric relationship.²

The objective of this study is to mechanistically and systematically understand (i) the variability in the dosimetric relationship among chemicals with vastly diverse properties and (ii) the possible impacts of physiological factors on the variability among subpopulations with different physiological conditions (e.g., obese vs. lean individuals, children vs. adults). The following study quantifies the dosimetric relationship by calculating a ratio of the daily chemical intake dose to the concentration in the human blood or urine, based on outputs from a well-evaluated, mechanistic toxicokinetic model.²⁰⁻²² Such a ratio can gain broad applications in exposure science, toxicology, and risk assessment, especially in the era of “new approach methods”, e.g., determining equivalent doses for biomonitoring data comparison, estimating route-to-route extrapolation values, and back calculating the steady-state daily intake dose of a chemical. We identify partition coefficients and biotransformation half-lives whereby chemicals can be found at high concentrations when the daily intake dose is equal. Partition coefficients and biotransformation half-lives are selected here due to their frequent use in environmental exposure science²³ and the ability of prediction tools to reliably compute these values for most environmental chemicals.²⁴ Results are visualized in a series of chemical partitioning plots from which the order of magnitude of a chemical’s dose-to-concentration ratio (DCR) can be quickly read based on its partition coefficients between air, water, and octanol phases. We also explore whether such property dependence is similar for both adults and children and for individuals with normal body weights and those with obesity.

The novelty of this work lies in the fact that it reveals a series of general trends and principles describing the impacts of chemical properties on the route-specific dosimetric relationship. Complementing existing chemical-by-chemical estimates, e.g., those cited by ref (2), our work provides mechanistic insights into chemical properties that contribute to high accumulation in the human body. Such insights can inform the risk estimation and assessment of the vast number of chemicals and can be applied proactively, even before the production or commercialization of chemicals. Therefore, our gained knowledge has the potential to support decision-making regarding the safe and responsible use of chemicals. Furthermore, the knowledge can also inform biomonitoring studies that typically measure overall chemical levels in the human body without distinguishing between multiple exposure routes and multiple elimination processes.

2. Methods

2.1. Defining the Dose-to-Concentration Ratio (DCR)

Here, we quantified the dosimetric relationship using the ratio of daily intake dose (in μg/d) to chemical concentration in biological fluids (blood or urine; in μg/L), defined as DCR with a unit of L/d. It links the external dose with the internal occurrence of a chemical. Numerically, the DCR also represents the volume of blood or urine where contamination of a chemical can be contaminated over a unit of time. The DCR is assumed to be independent of the intake dose. This assumption is sound when metabolic biotransformation follows the first-order reaction, which is often the case at the typical exposure levels anticipated for industrial and environmental chemicals.¹⁸

We limit the scope of this work to the following. First, we focus on parent compounds only and do not include corresponding metabolites. That is, we consider the removal of parent compounds from the body as a route of elimination rather than considering the generation of bioactive metabolites. We also assume that exposure to chemicals and associated metabolites does not initiate irreversible effects that, in turn, adversely impact absorption and elimination processes. Second, we focus on DCRs at the steady state given that the steady state represents conservative estimates of reasonable worst situations in health risk and impact assessments.²⁵ Third, we use the “total” chemical concentrations in blood and urine, for example, combining both the freely dissolved and cell- or protein-bound chemicals in the blood. For detailed discussions on these assumptions, please see Section 4 “Implications and Limitations”.

2.2. Simulations of DCRs on the PROTEX Model

We simulated the dose-to-blood concentration (DCR_blood) and dose-to-urinary concentration (DCR_urine) using the human exposure and toxicokinetic module of a comprehensive chemical assessment model named PROduction-To-EXposure (PROTEX).²⁰⁻²²Figure 1 presents a schematic overview of the core structure of PROTEX’s human exposure and toxicokinetic module and its dependence on chemical properties. Briefly, the module predicts the age-dependent level of chemicals in biological media (e.g., blood, urine, and lipid) by mechanistically quantifying mass flows via multiple routes of absorption and elimination of chemicals by the human body (for details on route-specific considerations and quantification, see Supporting Information Texts S1 and S2). To do so, the module considers the intestinal absorption of chemicals through ingestion of food and non-food items (e.g., dust, soil particles, and surface residuals), the respiratory absorption of chemicals through inhalation of air and airborne particles, as well as the dermal absorption of chemicals on the human skin (not discussed in this paper). For a given chemical, PROTEX predicts an “absorption efficiency”, i.e., the fraction passing the absorption barrier, based on its chemical properties.²⁶ On the other hand, it considers the elimination of chemicals through biotransformation and non-biotransformation processes (exhalation, egestion, urination, and percutaneous excretion). PROTEX quantifies elimination using a total elimination half-life.²⁷ The total elimination half-life should not be confused with the biotransformation half-life because the difference between these two half-lives reflects the excretion of chemicals through processes other than biotransformation. For women of childbearing age, PROTEX additionally characterizes the loss of chemicals through reproduction and lactation,²¹ which are beyond the scope of this work. PROTEX characterizes the rates of chemical absorption and elimination as functions of age-, sex-, and bodyweight-dependent physiological factors (e.g., the rates of respiration, glomerular filtration, and skin desquamation) and chemical properties (partition coefficients and biotransformation half-lives). For detailed information, please see refs 20 and 21 Supporting Information Texts S1 through S4 brief the calculation of chemical elimination processes, urinary concentrations of chemicals, and absorption efficiencies.

Figure 1.

Schematic overview of PROTEX’s human exposure and toxicokinetic module. It calculates chemical concentrations in blood and urine resulting from a single unit dose of chemical intake by mechanistically quantifying absorption and elimination processes as functions of chemical properties (partition coefficients between octanol, air, and water and biotransformation half-life) and physiological factors (such as functions of age, sex, and bodyweight). For illustration purposes, the model is parameterized to represent a 25 year old archetypal female American (Results for Sections 3.1 through 3.3). Additional computations are conducted for a 3 year old female child and an obese individual with a body weight twice that of the American average (Results for Section 3.4).

The parameterization of PROTEX requires input information on both the modeled individual and the chemicals of interest. We parameterized the module to represent an archetypal female American whose anthropometric, physiological, and behavioral characteristics were representative of the medians (or averages if medians are not available) of U.S. females. A detailed description of this modeled individual can be found in ref (28); the parameterization is consistent with up-to-date generic toxicokinetic models such as the generic physiologically based toxicokinetic (G-PBTK) model²⁹ and the high-throughput toxicokinetic (HTTK) model.³⁰ The simulation was done for the entire lifetime of this modeled individual. In this paper, we discuss the results of age 25 in detail while presenting the results of age 3 as well. For comparison, we also ran PROTEX for obese individuals by doubling the default bodyweight (with the body fat percent being the same) and hence changing the bodyweight-dependent physiological parameters in the model.

PROTEX’s expression of absorption and elimination processes requires inputs of a chemical’s partition coefficients and biotransformation half-lives. For instance, PROTEX’s algorithm utilizes partition coefficients between water, air, and biologically relevant phases, such as neutral lipids, phospholipids, and proteins. In this work, these partition coefficients are further quantified from partition coefficients between air and water (K_AW), between octanol and air (K_OA), and between octanol and water (K_OW), using single-parameter linear free energy relationships (spLFERs).^24,31,32 These three partition coefficients follow a thermodynamic triangular relationship (log K_OW = log K_OA + log K_AW), allowing the calculation of each based on the other two. For each chemical, internal energies for adjusting the partition coefficients at 25 °C to their corresponding values at the human body temperature are calculated according to MacLeod et al.³³ PROTEX also requires the input of the whole-body biotransformation half-life (HL_human) to characterize a chemical’s tendency to resist metabolic biotransformation. PROTEX calculates the overall elimination half-life by considering biotransformation and non-biotransformation elimination processes.

2.3. Investigation into the Impacts of Chemical Properties

To investigate the impacts of chemical properties on DCRs, we repeatedly ran PROTEX to simulate blood and urinary concentrations (in μg/L) in response to a “unit” dose through ingestion or inhalation (1 μg/d) for a series of combinations of K_AW and K_OA ranging from 10^–7 to 10⁴ and from 10² to 10¹³, respectively, with an infinite biotransformation half-life (i.e., “perfectly persistent”) and a biotransformation half-life of 2 days (close to the central tendency of more than 12,000 organics on chemical lists by the Organization for Economic Co-operation and Development).³⁴ These combinations of partition coefficients cover the properties of a wide spectrum of environmental chemicals.²² We calculated DCR_blood and DCR_urine by dividing the unit dose of intake by the modeled concentrations. In addition, we simulated the efficiencies of intestinal and respiratory absorption and the relative importance of different routes of elimination as functions of chemical partitioning properties.

2.4. Calculation of Chemical Properties for Model Performance Evaluation

To evaluate PROTEX’s performance, we ran the model with partitioning and biotransformation properties of chemicals based on literature-reported measurements. Following the recommended best practices,²⁴ the partition coefficients of the non-ionizable chemicals and the neutral form of ionizable chemicals were the consensus values of predictions by the OPEn structure–activity/property Relationship App (OPERA),³⁵ poly parameter linear free energy relationships (pp-LFERs) (with Abraham solute descriptors computed by the Iterative Fragment Selection (IFS-QSAR)),³⁶ and EPI Suite.³⁷ The acid dissociation constants of ionizable chemicals were computed using OPERA and used to calculate distribution ratios at the body pH. The biotransformation half-lives were consensus values of predictions by IFS-QSAR and QSAR-INSubria-Chem (QSARINS-Chem). According to the IFS-QSAR and QSARINS-Chem training sets,²⁷ these predictions represent the overall biotransformation half-lives, including biotransformation before (i.e., the first pass effect)³⁸ and after intestinal absorption.

3. Results and Discussion

3.1. Evaluation of the Model Performance

Before interpreting and analyzing the relationship between chemical intake and concentrations, we first thoroughly evaluate PROTEX’s performance in predicting human exposure and toxicokinetics to gain confidence in the model’s fidelity. The evaluation is performed from two aspects: first, we evaluate the model’s capability to capture the quantitative relationship between (i) toxicological benchmark doses and “biomonitoring equivalents” (BEs) for six chemicals with available data from the literature^39,40 and (ii) steady-state blood concentration normalized by 1 unit dose of exposure for four chemicals with available data in the literature.^4,5,45 Second, since DCR depends on completing the absorption and elimination processes, we additionally evaluate the model’s performance in predicting the intestinal and respiratory absorption efficiencies (41 chemicals collected from the literature) as well as the observed overall elimination half-lives (1100 chemicals collected from the literature²⁷ ). Future studies may benefit from the model’s encouraging performance of predicting the overall elimination half-life, which is critical for quantifying the connection between chemical intake and body concentrations but has been recognized as an “important scientific challenge” because oftentimes “current literature half-lives are not consistent”.⁴¹

BEs are representative of the steady-state concentration of chemicals in a biological medium (mainly blood and urine) when an average human’s chemical exposure equals a toxicological benchmark (e.g., a reference dose).^39,40 In this sense, our calculated DCR can be seen as the ratio of the toxicological benchmark to the corresponding BE. Supporting Information Table S1 compares our calculated DCRs with ratios derived from the toxicological benchmark doses and BEs for ten environmental pollutants.⁴²⁻⁴⁸ The results demonstrate that eight of the ten chemicals tested were within 2 orders of magnitude of each other, with the predicted values being higher than the “real” measured values. This variation may be due to differing initial methods of deriving BE across the literature. For example, our predicted DCRs are closest to those measured using interspecies extrapolation (hexachlorobenzene, 2,4-D, and DDT) and PBTK modeling (benzene, bisphenol A, triclosan, and parathion), presumably due to increased accuracy within these methods. In contrast, our predicted DCRs differ most from BE values derived from the simple first-order elimination model (e.g., 2,2′,4,4′,5-PentaBDE), which may be a result of inaccurate assumptions made about variables such as absorption rate and half-life. For a more detailed discussion, see Supporting Information Table S1.

Figure 2a,b visualize the predicted absorption efficiencies for different combinations of log K_OA and log K_AW, representing various possible chemicals in the partitioning space for intestinal and respiratory routes, respectively. Diagonals from the top left to the bottom right of these two-dimensional contour diagrams represent chemicals with the same log K_OW. PROTEX predicts that intestinal absorption efficiencies are the highest for chemicals with a log K_OW between 1 and 9 (Figure 2a), whereas respiratory absorption is most efficient for water-soluble chemicals with a log K_AW lower than 1 (Figure 2b). Figure 2a,b show that these predicted trends were generally confirmed by the medians of measurements for human intestinal (23 chemicals)⁴⁹ and respiratory (18 chemicals)^50,51 absorption efficiencies. These predicted trends are also consistent with literature-reported findings for other mammals. For instance, intestinal absorption has been shown to be rather limited for super hydrophilic (log K_OW below −0.5) or hydrophobic chemicals (log K_OW above 7).⁵² The penetration of these chemicals is hindered by the lining of the intestinal tract, which combines hydrophilic (the unstirred water layer and epithelium) and hydrophobic (the epithelium) membranes.⁵³ For inhalation, however, less water-soluble chemicals, that is, those with a log K_AW greater than −1, are less likely to be absorbed because the water-rich mucous resists their penetration through the respiratory epithelia.⁵³ Nevertheless, measurements are currently limited and often divergent from each other. For instance, PROTEX predicts a respiratory absorption efficiency of 86% for nicotine; however, the measurements vary from nearly 70% to almost 100% depending on the respiratory pattern and the composition of the inhaled mixtures.^50,51

Figure 2.

PROTEX-predicted efficiencies through intestinal (panel a) and respiratory (panel b) absorption (contour diagrams) as a function of log K_AW and log K_OA in comparison with literature-reported measurements (colored dots). Comparison between the modeled and literature-reported total elimination half-lives (panel c). PROTEX-predicted half-lives for combined non-biotransformation elimination processes (exhalation, egestion, urination, and percutaneous excretion considered in this work) and the dominant processes of elimination (panel d).

Lastly, we use PROTEX’s human toxicokinetic module to simulate the total elimination half-lives of 1100 chemicals and compare the predictions with measurements collected by Arnot et al. (Figure 2c) from various accredited sources within the literature.²⁷ This collection process was done by searching through pharmacokinetic and toxicokinetic sources containing measured half-lives for organic chemicals in human adults. These 1100 chemicals are diverse in physicochemical properties, ranging from hydrophobic to hydrophilic (log K_OW from 5 to 11), volatile to minimally volatile (log K_OA from 2 to 13), and recalcitrant to labile (HL_human from 1 to 10⁶ h). Figure 2c indicates that the predicted and measured total elimination half-lives show a strong correlation. The root-mean-square deviation (RMSE) between the predicted and measured total half-lives is 0.58 log units, indicating that the model overall succeeds in predicting the elimination processes for a wide range of chemicals with a discrepancy within an order of magnitude. The PROTEX’s algorithm explains 81% of the variation (R² = 0.81). Supporting Information Figure S1 indicates the residual (namely the difference between predicted and measured half-lives) seems to be random, independent of chemical partitioning (K_OA and K_AW) and biotransformation (HL_human) properties. With there being no evident pattern to the outlying chemicals, we anticipate no systematic bias present within the half-life predictions. Additionally, we identify 93 outliers for which the predicted and observed half-lives diverge by over an order of magnitude. This discrepancy can be attributed to varying levels of uncertainty and possible errors in measurements. For instance, our predicted total elimination half-life of PCB-156 (2 years) was 13 times shorter than the measurement documented in the dataset (26 years).²⁷ Such a discrepancy is smaller than the variability in PCB-156's total elimination half-lives observed among several biomonitoring studies for adults, ranging from 1.6 to >100 years.⁵⁴ According to Shirai and Kissel,⁵⁵ half-lives either shorter than 1 year or longer than 10 years may be very unlikely for PCBs. Also, when measurements are taken, they may not necessarily represent the steady-state level of a chemical.²⁷ This discrepancy is due to the difficulty in telling if steady-state was reached at the time of the collection of the data.

3.2. Occurrence of Perfectly Persistent Organic Chemicals in the Blood

After gaining confidence in the model’s performance, we begin our discussion with the simplest case of perfectly persistent chemicals, namely, recalcitrant chemicals not subject to metabolic biotransformation. Figure 3 visualizes the simulated DCR_blood for different combinations of log K_OA and log K_AW, representing various possible hypothetical perfectly persistent chemicals as results of exposure through ingestion (Figure 3a) and inhalation (Figure 3b). Overall, the DCR_blood varies by more than 10 orders of magnitude among chemicals. In other words, we can expect the blood concentrations of various chemicals to differ by 10 orders of magnitude even if their intake doses are similar. This range is not unexpected because of the range (11 orders of magnitude for partition coefficients, as shown in Figure 3) of vastly diverse properties being analyzed for various chemicals and the effects these properties have on their absorption and elimination rates. Note that a lower DCR_blood (e.g., dark red areas in Figure 3a,b) means a higher blood concentration, with the intake dose being equal. Interestingly, the U.S. Environmental Protection Agency’s ExpoCast project predicted that for general Americans, the average population median intake rate varies by 8 orders of magnitude among ∼480,000 chemicals.⁵⁶ Combining the variabilities in DCR_blood, we would expect chemical concentrations to vary by up to 18 orders of magnitude in Americans.

Figure 3.

Chemical partitioning space illustrating the ratio of intake dose to blood concentration (DCR_blood) for ingestion (panel a) and inhalation (panel b) as well as the ratio of intake dose to urinary concentration (DCR_urine) for ingestion (panel c) and inhalation (panel d) of hypothetical perfectly persistent chemicals (molar mass = 300 g/mol) by an archetypal 25 year old female American. Chemicals are arranged by log K_AW and log K_OA; diagonals represent chemicals with equal log K_OW. Illustrative chemicals are marked based on their log K_AW and log K_OA: PCB = polychlorinated biphenyl; PBDE = polybrominated diphenyl ether, D5 = decamethylcyclopentasiloxane (cyclic volatile methyl siloxane), DTDP = ditridecyl phthalate, TCE = trichloroethylene, and TMP = tris(methyl) phosphate.

For both ingestion and inhalation, persistent chemicals with a high log K_OA and low log K_AW (lowly volatile and moderately to highly hydrophilic chemicals on the bottom right) were predicted to have the highest chemical concentration in the blood if they possess the same intake doses as other chemicals. When this is paired with the dependence of absorption efficiencies for ingestion (Figure 2a) and inhalation (Figure 2b) on partitioning properties, these identified chemicals are within the range of highest absorption efficiency. Furthermore, when ingestion and inhalation are directly compared, ingestion leads to higher DCR_blood at lower log K_OA and higher log K_AW (highly volatile but lowly hydrophilic chemicals on the upper left corner) than inhalation does because these chemicals are subject to efficient intestinal absorption (Figure 2a) but less efficient at diffusing through alveolar mucus, although they can freely move within the human airway.⁵³

However, a high absorption efficiency does not necessarily lead to a high concentration in the blood if the chemical is readily eliminated from the human body. For facilitating the interpretation, Figure 2d displays the dependence of the total non-biotransformation elimination half-life (in hours; resulting from combined routes of elimination other than biotransformation) on partitioning properties, as well as the PROTEX-predicted dominant routes of elimination (exhalation, urination, and fecal egestion). Figure 3a,b show that chemicals with a small log K_OA (highly volatile chemicals on the left) and a small log K_OW (highly hydrophilic chemicals in the bottom left corner) have high DCR_blood. This is mostly because these chemicals are efficiently eliminated from the human body through exhalation or urination, respectively (Figure 2d). In contrast, chemicals with the lowest DCR_blood are those with the longest elimination half-life (Figure 2d). Since these chemicals are less volatile (with a high log K_OA) and less water-soluble (with a high log K_OW), elimination through exhalation and urination is rather slow. Given that fecal egestion is usually not an efficient route of elimination, these chemicals may stay in the human body for a long time before reaching a heightened level.

Taken together, Figure 3a,b underscore that DCR_blood reflects a competition between the absorption and elimination processes. Figure 3a,b display four main mechanisms that lead to different DCR_blood: (I) high absorption and low elimination (chemicals on the bottom right), with typical examples including polychlorinated biphenyls and polybrominated diphenyl ethers, (II) high absorption and high elimination (chemicals on the upper left for ingestion and chemicals on the bottom left for inhalation), with typical examples including cyclic volatile methyl siloxanes, (III) low absorption and low elimination (chemicals on the upper right), with typical examples including ditridecyl phthalate, and (IV) low absorption and high elimination (chemicals on the bottom left for ingestion and chemicals on the upper left for inhalation), with typical examples including organic solvents, such as trichloroethylene, and water-soluble organophosphates, such as tris(methyl) phosphate.

We also predicted the DCR_urine for different combinations of log K_OA and log K_AW, representing hypothetical perfectly persistent chemicals (Figure 3c,d). For both ingestion (Figure 3c) and inhalation (Figure 3d), chemicals with a low log K_AW and a low log K_OW (highly hydrophilic chemicals on the bottom left) share the lowest DCR_urine. In other words, compared to other chemicals, these chemicals are more abundant in the urine even if they share the same levels of intake dose. Recalling Figure 2d, we can find that chemicals within this range tend to be fairly water-soluble, so inevitably, they would be eliminated through urination.

Interestingly, Figure 3 reveals that DCR_blood is governed by both absorption and elimination, whereas DCR_urine is to a greater extent governed by elimination. In other words, chemicals that tend to be eliminated through urination are more likely to achieve a high urinary level, with the intake dose remaining equal. Nevertheless, for the ingestion route, the less efficient intestinal absorption still limits the DCR_urine of chemicals with a log K_OW lower than −2, as evidenced by the bottom left side corner of Figure 3c.

3.3. Occurrence of Labile Organic Chemicals in Blood and Urine

We additionally consider how the biotransformation of a chemical impacts the DCR, given that organic chemicals can be subject to different extents of hepatic and intestinal biotransformation.²⁷ As compared to the perfectly persistent chemicals shown in Figure 3, chemicals possessing the same partitioning properties but undergoing rapid biotransformation (with a whole-body biotransformation half-life of 2 days) are selected in Figure 4 as an illustrative example. Note that the PROTEX model can also be parameterized to represent more persistent or labile chemicals if desired.

Figure 4.

Chemical partitioning space illustrating the ratio of intake dose to blood concentration (DCR_blood) for ingestion (panel a) and inhalation (panel b) as well as the ratio of intake dose to urinary concentration (DCR_urine) for ingestion (panel c) and inhalation (panel d) of chemicals (molar mass = 300 g/mol) with a human whole-body biotransformation half-life of 2 days by an archetypal 25 year old female American. Chemicals are arranged by log K_AW and log K_OA; diagonals represent chemicals with equal log K_OW.

Overall, compared with perfectly persistent chemicals, chemicals that undergo negligible biotransformation have increased DCRs. Such an increase is more pronounced for perfectly persistent chemicals already with low DCRs, namely chemicals with a high log K_OA and a low log K_AW (chemicals on the bottom right) for DCR_blood and chemicals with a low log K_AW and a low log K_OW (chemicals on the bottom left) for DCR_urine. Notably, for chemicals with a log K_OW between 4 and 10 and a log K_OA greater than 7, DCR_blood increased by 3 orders of magnitude from the case of perfect persistence (Figure 3a) to that with a biotransformation half-life of 2 days (Figure 4a), regardless of exposure routes. This is because these chemicals are associated with long non-biotransformation elimination half-lives (corresponding to elimination processes other than biotransformation, including exhalation, egestion, urination, and percutaneous excretion considered in this work; see Section 2.2), typically longer than 10,000 h (Figure 2d); therefore, the inclusion of a fairly short biotransformation half-life of 2 days substantially lowers the overall elimination half-life and greatly reduces the chemical level in the human body. In contrast, since the non-biotransformation elimination half-lives are already short for volatile, hydrophilic chemicals (less than 1000 h; Figure 2d), the inclusion of a biotransformation half-life of 2 days only slightly alters the total elimination half-life and poses minimal impact on DCRs.

3.4. Impacts of Physiological Factors on the Dose-to-Concentration Ratio

In previous sections, we based our modeling on an archetypal female American aged 25. One may wonder whether, and to what extent, the above results hold for individuals other than a 25 year old archetypal adult, such as children, obese individuals, and other vulnerable subpopulations in environmental health. Underlying this is the question of whether physiological factors have a more significant impact on the DCR than chemical properties.

Since the mechanistic nature of the PROTEX model enables exploring the interaction of chemical properties and physiological factors, we additionally calculated DCR_blood and DCR_urine of the hypothetical perfectly persistent chemicals for a child three years of age (Supporting Information Figure S2) and an adult with a higher than average body weight (Supporting Information Figure S3) through ingestion and inhalation. In comparison with those found earlier for an archetypal 25 year old American woman (Figures 3 and 4), despite a difference in the levels of DCR_blood and DCR_urine, the resulting distribution of hot spot areas in the outcomes of each simulation was similar. Notably, the DCR_blood is almost unchanged for chemicals with a log K_OA lower than 6.5 or chemicals with a log K_OW lower than 3. These chemicals are readily eliminated via exhalation and urination, respectively, and their distribution is less dependent on physiological parameters.

The differences seen between children and adults are due to children’s more efficient elimination of chemicals from their bodies.⁵⁷ This has been observed in extensive biomonitoring studies.^54,58 PROTEX predicts that the non-biotransformation elimination half-life of a given chemical is 1.8 to 5.5 times longer in adults than in children. Such a discrepancy is more pronounced for chemicals that are eliminated mainly through fecal egestion because children eat and defecate more relative to their body weight than adults do. For this reason, the DCR_blood of these chemicals are more responsive to age (Figure S2). It should be noted that PROTEX here calculates the “true” non-biotransformation elimination half-life, which excludes the dilution effect of the fast growth of children; the adult-child difference in the “apparent” non-biotransformation elimination half-life would become more remarkable should this effect be considered.⁵⁹

Furthermore, for those whose body weight is above average (Figure S3), the outcome was comparable to those of average weight. The increased fat volume found in an overweight adult increases the storage of hydrophobic chemicals and reduces their entry into systemic circulation for possible elimination.^60,61 Therefore, though overweight adults would in fact have higher burdens (total masses) of hydrophobic chemicals found within their entire body, the levels of these chemicals found specifically in their blood are similar to the levels of hydrophobic chemicals found within the blood of an average adult, leading to similar outcomes within the DCR.

These similarities between the modeled scenarios go to show that both age- and weight-dependent physiological variabilities play a limited role in governing a chemical’s DCR_blood and DCR_urine. That being said, if a hypothetical perfectly persistent chemical falls along the outside region of changing zones, there may be some numeric differences. While it is important to note this, the differences are small enough to still have comparable results among the modeled scenarios.

Based on these findings, it is believed that the conclusions in this work are generic and may also be applied to people from other countries and yield similar results. For obtaining more accurate population- and region-specific results, one can also run the PROTEX model, as long as the needed information about the population being observed is entered into the PROTEX model. If there are specific chemicals of interest, then the properties of these chemicals may also need to be entered. As shown in the simulations above, chemicals that undergo both substantial and negligible biotransformation within a population can be analyzed via this method.

4. Implications and Limitations

This work offers a pioneering systematic understanding of the quantitative relationship between the human intake dose and concentrations in bodily fluids for chemicals, taking into account the diversity of chemical partitioning and biotransformation properties. It discloses the dependence of the calculated DCR on the interaction between chemical properties and the physiological properties. This systematic understanding is especially useful in the “new approach methods” era because it enables a rapid preliminary evaluation of a chemical’s accumulation potential using only its readily available properties and general rules and principles outlined in this study, saving both time and resources. It also allows for comparability between chemicals to identify the concerning properties based on which production can be regulated. Such information parsimony allows for high-throughput exposure assessments for numerous chemicals already on the market, as well as for data-poor premanufactured and designed chemicals. Our chemical partitioning plots also facilitate quickly estimating the order of magnitude of the DCR_blood and DCR_urine for large numbers of chemicals based on only the information on partition coefficients and biotransformation half-lives. The PROTEX model itself can still provide more precise estimates of DCRs. Both can promote the applications of the DCR concept in exposure science, toxicology, and risk assessment. In environmental and health risk assessments, DCR can help convert estimated daily intake doses to concentrations in biological specimens, facilitating the use of biomonitoring data, e.g., biomarker concentrations measured in the U.S. NHANES and Canadian Health Measures Survey. For example, DCR can be used to derive BEs for comparisons with biomonitoring data to inform the risks posed to the general population.⁴²⁻⁴⁸

DCR can also be used to calculate route-to-route extrapolation factors between benchmark doses derived following ingestion or inhalation exposures.^9,11 The extrapolation factors should be chemical specific because they reflect the combined effects of route-specific absorption and elimination. However, the absence of chemical specific factors results in the use of conservative estimates, for example, the assumed 50% for intestinal absorption and 100% for respiratory absorption recommended by the European Guidance on Information Requirements and Chemical Safety Assessment.⁶² Noticing that the ratio of intestinal to respiratory absorption efficiencies is mathematically equivalent to the ratios between DCR_blood for ingestion and inhalation routes, we thus recommend using PROTEX-predicted DCRs based on chemical-specific data for more reasonable estimates in route-to-route extrapolation.

DCR also aids in screening-level reverse dosimetry to back-calculate the steady-state daily intake dose that is not likely to cause observable adverse biological impacts based on in vitro bioactivity tests or other new approach methods.^2,3,63 Sharing a similar meaning with the commonly used guidance or reference toxicological benchmark doses, such back-calculated intake doses complement and expand existing databases of toxicological points of departure.⁶⁴ It also enables investigating the consistency between the toxicity characterized by in vivo toxicological data and in vitro bioactivity data.^65,66

In pharmacy, DCR_blood can help determine the dosing of a given compound and how often to instruct taking it.⁶⁷ DCR_urine can also be used to predict the occurrence of pharmaceuticals in urine and, hence, the urinary excretion of unmetabolized pharmaceuticals. Such information helps determine the environmental releases of pharmaceuticals for ecological risk assessments, which aim to promote sustainable pharmacy by decreasing pharmaceutical waste and safeguarding ecological integrity.⁶⁸

Despite the wide implications of DCRs and mechanistic insights gained from this study, we must acknowledge the limitations of this work. First, we investigated the steady-state DCR because it represents the case of long-term, quasi-constant exposure to environmental chemicals. Although such a conservative scenario is desired for risk assessment,²⁵ it is important to note that this ratio may not be reflective of true chemical levels if chemicals have a long residence time within the human body or if chemicals whose production and emissions are subject to substantial temporal changes. Second, the current work builds largely on the mechanistic understanding of neutral organic chemicals. While PROTEX’s algorithm accommodates ionizable organic chemicals and performs satisfactorily for both neutral and ionizable organic chemicals, we should note that the mechanistic knowledge and quantitative characterization of the toxicokinetics of ionizable organic chemicals, e.g., the electrostatic interactions between charged chemicals and the body tissues, are still rather inadequate. Third, the current work focuses on the ingested and inhaled parent compounds and excludes their metabolites from the discussion. Compared to parent compounds, metabolites are typically more hydrophilic, hence, occupying different locations in the chemical partitioning space (typically moving downward in the diagrams).⁶⁹ Also, the use of the DCR approach limited to the parent compound when working with more bioactive metabolites would not lead to the intended protective human health estimates. Fourth, in future work, dermal absorption should also be considered when assessing the DCRs for chemicals often found in consumer products, such as lotions, sunscreens, and make-ups.

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acs.est.2c08470.

• Exposure and toxicokinetic module used in this work and elimination and absorption processes therein; literature reported and model predicted dose-to-concentration ratios; independence of prediction residuals and chemical properties; and predicted dose-to-concentration ratios for children and obese individuals (PDF)

The authors declare no competing financial interest.