2025 Volume 50 Issue 9 Pages 471-481
Pharmaceutical manufacturing and storage processes pose the potential risk of chemicals migrating from the packaging materials into pharmaceuticals. These migrants, known as extractables and leachables (E&Ls), consist of various chemicals that may pose a risk to patients during therapeutic use. Although exposure to E&Ls via the intravenous route is of greater concern, there is almost no toxicity information for these chemicals to determine the Permitted Daily Exposure (PDE). The purpose of this study was to establish the Threshold of Toxicological Concern for intravenous route exposure (TTCiv) for risk management of E&Ls contained in pharmaceuticals. First, we derived the oral PDEs of 287 chemicals from a list of 923 known E&Ls. Then, the modifying factor (α) for estimating the intravenous PDE from each oral PDE was calculated based on the ratio of the predicted blood concentrations (area under the curve (AUC) and maximum blood concentration (Cmax)) after oral and intravenous administration using the Integrated Chemical Environment (ICE) PBPK model. Additionally, due to uncertainty of the predictions without bioavailability information, the intravenous PDE was calculated using modifying factor 3 based on the maximum value of the root mean square error (RMSE) reported in varification of the High-Throughput Toxicokinetics (HTTK) model. In conclusion, by analyzing the distribution of the intravenous PDE for 287 chemicals, we propose a TTCiv of 27 µg/day/human, based on the ratio of Cmax. Our route extrapolation approach could contribute to the establishment of scientifically valid TTCs for not only E&Ls, but also for other impurities without toxicity information.
During the manufacturing and storage processes of pharmaceuticals, there is the risk of chemicals migrating from the packaging materials into pharmaceuticals. Impurities that may leach thus into pharmaceuticals are called extractables and leachables (E&Ls). Risk management for E&Ls is essential because exposure to E&Ls can pose unexpected risks to patients. In recent years, the advances in analytical technologies have made it possible to detect even trace amounts of E&Ls in pharmaceuticals. E&Ls have a wide variety of origins, including the raw materials used, additives used, reaction products, and chemicals migrating from the packaging materials, and thus they include widely diverse types of chemicals.
As for the global harmonized method for the risk assessment of E&Ls, the International Council for Harmonisation of Technical Requirements for Pharmaceuticals for Human Use (ICH) Q3E is currently formulating a guideline. The ICH has identified the need to develop harmonized thresholds for reporting and identifying E&Ls and qualifying leachables in the context of the route of administration, drug indication and patient exposure, with an emphasis on science- and risk-based approaches (ICH Q3E Concept Paper, 2020). The Product Quality Research Institute (PQRI) has suggested that Permitted Daily Exposure (PDE) needs to be derived for E&Ls detected in amounts above the Analytical Evaluation Thresholds (AET) (PQRI, 2021). Although exposure to E&Ls entering the body via the intravenous route is of greater concern, there is almost no toxicity information for E&Ls entering the body via the intravenous route to derive the corresponding PDE. Therefore, derivation of the PDE for the intravenous route remains a challenge.
The Threshold of Toxicological Concern (TTC) approach is effective for the risk management of trace amounts of chemicals for which toxicity information is not available. For the mutagenic impurities in pharmaceuticals, a TTC value of 1.5 µg/day has been adopted as specified in the ICH M7 guidelines (ICH M7 (R2), 2023). In regard to TTC for substances other than genotoxic carcinogens, the TTC approach based on the Cramer classification has been used, and further improvements have been considered for foods and fragrances (Munro et al., 1996; Patlewicz et al., 2022; Yang et al., 2017; Patel et al., 2020; Dolan et al., 2005). However, most of the toxicity information available to develop the TTC is based on the oral route toxicity studies. Therefore, the existing TTC cannot be directly applied to the management of E&Ls to which patients are exposed via parenteral routes, including the intravenous route. Parenteral exposure to E&Ls is considered riskier than exposure to these substances entering the body via the oral route (ICH Q3E Concept Paper, 2020). Accordingly, it is of particularly significance to derive TTC for intravenous exposure. Derivation of an internal exposure TTC (iTTC) based on the plasma concentration has been proposed (Arnot et al., 2022; Ellison et al., 2021; Ellison et al., 2019). Estimation of the internal concentrations (whole-body and blood) of chemicals from the no observed effect level (NOEL) via the oral route used by Munro et al. was attempted using toxicokinetic data and physiology-based pharmacokinetic (PBPK) prediction models, and the iTTC was calculated by applying an uncertainty factor of 100 to the 5th percentile value of these concentrations (Arnot et al., 2022). However, permitted exposure limits are necessary for the management of pharmaceutical impurities such as E&Ls. In addition, the chemical space of E&Ls appears to be quite different from that used for the iTTC derivation by Aron et al., and the validity of deriving TTC values for E&Ls needs to be verified. Masuda et al. addressed the differences in the chemical structure group by using a list of E&Ls and proposed TTC values for the exposure route (Masuda-Herrera et al., 2022). However, the TTC values proposed by Masuda et al. raised concerns about the availability of bioavailability calculations that rely on commercial modeling software.
In this study, we used reliable toxicity data to determine the oral PDE and estimated the TTCiv using the PBPK model, which is publicly available and transparent, as the input parameters and formulas needed for calculating the blood concentrations used to predict the bioavailability are available. We developed a method for extrapolating oral PDEs to intravenous PDEs using a PBPK model. In particular, the modifying factor (α) for estimating the intravenous PDE from the oral PDE was calculated for each chemical based on the ratio of the predicted blood concentrations (AUC and Cmax) after oral and intravenous administration.
The list of chemicals comprising E&Ls was collected from the ELSIE database (https://elsiedata.org/elsie-database/) noted by Masuda et al., 2022, the chemicals provided by the Japan Pharmaceutical Manufacturers Association (JPMA) as E&Ls, and the chemicals recognized as E&Ls from a review of the literature. Chemicals for which the structure or CAS Registry Number® could not be identified were excluded from the analysis. While the cohort of concern (COC) in ICH M7 were excluded, other chemicals were analyzed regardless of their genotoxicity. After eliminating duplicates, we finally included the 923 chemicals listed as E&Ls in the analysis in this study.
Derivation of oral PDEWe reviewed information on the repeated-dose toxicity, reproductive toxicity, and carcinogenicity for the 923 E&Ls from the following high-priority sources: Organisation for Economic Co-operation and Development (OECD) Screening Information Data Set (SIDS) report (SIDS Initial Assessment Report) (https://hpvchemicals.oecd.org/ui/Search.aspx), assessment reports from national assessment bodies, Japan Existing Chemical Data Base (JECDB) (https://dra4.nihs.go.jp/mhlw_data/jsp/SearchPage.jsp), and Environmental Risk Initial Assessment of Chemicals (MOE, Japan) (https://www.env.go.jp/chemi/risk/chemi_list/index.html). When multiple evaluation reports were available, the most conservative evaluation was adopted. In cases where the information from of high-priority sources was not available, the NOEL information was evaluated and adopted from the toxicity information data in the European Chemicals Agency (ECHA) chemicals database (https://chem.echa.europa.eu/). As a result, 287 chemicals with toxicity information from which oral PDEs could be derived were obtained.
The POD (point of departure; NOEL or the lowest-observed effect level (LOEL)) for oral administration was determined for each E&L. Conservative evaluation to the extent possible was performed based on the available information. The read-across approach was not adopted, and the results of direct exposure to the target chemical or its salt and information that could confirm dose-response relationships were used for derivation of the oral PDE. Changes resulting from administration of the chemicals were basically considered as the effects, including findings that could not be clearly identified as oral route-specific toxic effects (e.g., direct gastrointestinal damage). Dose-dependent changes in organ weights and blood biochemistry were used as evidence for NOEL (or LOEL) even in the absence of pathological findings. However, transient body weight changes at the start of administration of the chemical or male-rat-specific alpha2u globulin-induced nephropathy were not considered as toxic effects. In this study, we did not use the toxicity information of chemicals for which the details of the test method (species, route of administration, duration of administration, measurement items, etc.) required for the PDE derivation were not available. Information on the toxic endpoints in the most sensitive animal species available was used for setting the PDEs for each E&L. These endpoints included the study results in mice and dogs, as well as rats. The current ICE PBPK model cannot handle data other than from rats and humans, but almost all the toxicity studies on which the PDEs were conducted in rats, with few exceptions. In this study, chemicals for which the study results were for animals other than rats were excluded from the analysis.
The oral PDE were basically calculated from the NOEL, according to the ICH Q3D guideline (ICH Q3D (R2), 2022) (Formula 1), except in cases for which information on NOEL was lacking, where the LOEL was used instead.
The modifying factors used for the oral PDE derivation are shown in Table 1. As most of the detailed toxicity information was not available to calculate the benchmark doses, a modifying factor of 10 was applied to all LOEL data (F5=10).
| F1 | A factor to account for extrapolation between humans and rats | 5 |
| F2 | A factor to account for variability between individuals | 10 |
| F3 | A variable factor to account for toxicity studies of short-term exposure1) | 1-10 |
| F4 | A factor that may be applied in cases of severe toxicity2) | 1-10 |
| F5 | A variable factor that may be applied if the no-effect level was not established. | 1 or 10 |
1) F3 = 1 for studies that last at least one-half lifetime (1 year for rodents) or for reproductive studies in which the whole period of organogenesis was covered, F3 = 2 for a 6-month study in rodents, F3 = 5 for a 3-month study in rodents, F3 = 10 for studies of a shorter duration. F3=10 for studies of a shorter duration than 3 months.
2) F4 = 1 for fetal toxicity associated with maternal toxicity, F4 = 5 for fetal toxicity without maternal toxicity or for a teratogenic effect with maternal toxicity, F4 = 10 for a teratogenic effect without maternal toxicity.
As described in the ICH Q3D guideline, bioavailability can be used as a modifying factor (α) for establishing the PDE for different routes of exposure.
Based on the assumption that the ratio of the AUC is equivalent to the ratio of doses, the modifying factor (α) used for route-to-route extrapolation from the oral PDE to the intravenous PDE was determined. The PDEs for intravenous exposure were derived according to the procedure shown in Formula 2. The methods for calculating the modifying factor (α) based on the predicted blood concentration are shown below (Fig. 1).
Calculation of the modifying factor (α) based on the predicted blood concentration (AUC or Cmax). The ratio of the blood concentrations following intravenous and oral administration was calculated as the modifying factor (α). AUC, area under the curve, Cmax, maximum blood concentration, IV, intravenous.
The modifying factor (α) was calculated for each chemical based on the ratio of the predicted blood concentrations (AUC and Cmax) after oral and intravenous administration of the same dose.
The PBPK model of the ICE system, which provides free access to new approach methodologies (NAMs), data and tools for the evaluation and interpretation of chemical bioactivity data, was used to predict blood concentrations to modify the differences in bioavailability between the oral and intravenous routes (Pearce et al., 2017; Bell et al., 2020; Abedini et al., 2021). Using a PBPK model on the ICE web site (ICE v4.0.2) (https://ice.ntp.niehs.nih.gov/Tools?tool=pbpk), the AUC and Cmax of chemicals after oral administration of each oral NOEL (or LOEL/10), which is the POD to set the oral PDE, was calculated. The PBPK model was also used to estimate AUC and Cmax when oral NOEL (or LOEL/10) doses were administered via the intravenous route.
The parameters entered in the Solve_pbpk model of the PBPK model in the ICE system are shown in Table 2. The Solve_pbpk model is a multi-compartment PBPK model from EPA’s HTTK R package that includes gut, artery, vein, lung, liver, kidney, and the rest of the body compartments. The model can predict the distribution of chemicals in humans or rats following administration via the oral or intravenous route (ICE Chemical Characterization Tool User Guide, 2024). The PDE for intravenous exposure was calculated using the modifying factors (α) according to Formula 2.
| Items required to set in PBPK model | Input parameters |
| User Chemical Identifiers | CAS Registry Number® |
| Species | Rat |
| Body Weight | 0.25 |
| ADME Source | Default |
| Exposure Dose | NOEL (or LOEL/10) of each E&L |
| Model | Solve_pbpk |
| Exposure Route | Oral or iv |
| Exposure Interval, Hours | 24 (as default setting) |
| Simulation Length, Days | Administration period of each E&L |
| Output Conc. Units | mg/L |
The TTCiv was determined as the 5th percentile of the PDE distribution for intravenous exposure obtained by route extrapolation from oral exposure. We calculated the TTC based on the PDE for the oral route and then extrapolated it to the PDE for intravenous exposure using the modifying factor (α) based on the ratio of AUC or Cmax. In addition, to account for the lack of information on the bioavailability as a source of prediction uncertainty, the TTCiv was also calculated using a modifying factor based on the root mean square error (RMSE) previously reported in HTTK model validations (Breen et al., 2021; Wang, 2010; Linakis et al., 2020; Wambaugh et al., 2018).
Chemical properties profilingTo confirm the validity of the 287 chemicals for which PDEs were derived as a dataset for estimating the TTCiv for E&L, the chemical properties of the 287 chemicals and the other E&Ls were compared using the Chemical Characterization Tool in the ICE system (ICE version 4.1: revised September 2024). Of the 814 chemicals whose chemical properties were analyzed by ICE from 923 E&Ls, PCA analysis was performed on 287 chemicals for which PDE was derived and the remaining 527 chemicals. This tool uses principal component analysis (PCA) and plots the dynamic PCA results to compare the properties between two sets of chemicals based on their physicochemical properties and molecular descriptors. Dynamic PCA plots were generated for the physicochemical properties or molecular descriptors listed below. The PCA analysis was conducted based on the principal component with the highest variability (PC1) and the principal component with the second highest variability (PC2). The distance between the points in the PCA plots corresponds to the similarity levels between physicochemical properties and molecular descriptors (ICE Chemical Characterization Tool User Guide, 2024).
Chemical properties: BP_pred (predicted boiling point), LogHL_pred (predicted Henry’s Law), LogWS_pred (predicted water solubility at 25°C), LogKOA_pred (predicted log octanal/air partition coefficient), MP_pred (predicted melting point), LogVP_pred (predicted log vapor pressure), MW_total (total molecular weight).
Molecular descriptors: Zagreb, XLogP, WPOL, MW, VAdjMat, VABC, TopoPSA, LipinskiFailures, nRotB, topoShape, MLogP, nAtomLAC, nAtomP, nAtomLC, HybRatio, nHBDon, nHBAcc, bpol, nB, nBase, nAtom, nAromBond, naAromAtom, apol, ALogP, ALogp2, AMR, nAcid, nHeavyAtom, nH, nC, nN, nO, nS, nP, nF, nCl, nBr, nI, nX, nBonds, nBonds2, nBondsS, nBondsS2, nBondsS3, nBondsD, nBondsD2, nBondsT, nBondsM, Sv, Sse, Spe, Sare, Sp, Si, Mv, Mse, Mpe, Mare, Mp, Mi, nRing, n3Ring, n4Ring, n5Ring, n6Ring, n7Ring, n8Ring, n9Ring, n10Ring, n11Ring, RotBFrac, nRotBt, RotBtFrac, topoRadius, topoDiameter (ICE Chemical Characterization Tool User Guide, 2024).
Using the OECD QSAR Toolbox (ver. 4.6), we compared the toxicity profiles of the 923 chemicals in the E&Ls list and 287 chemicals for which PDEs were derived. The endpoint-specific profilers related to human health effects and the toxicological profiler (HESS) within the OECD QSAR Toolbox were compared between two datasets. For this comparison, the CAS Registry Number® from each list was entered into the OECD QSAR Toolbox and the profilers were applied. In cases where several structures were registered under a single CAS Registry Number®, we included all the structures in the analysis.
Endpoint-specific profilers related to human health effects include: repeated dose toxicity (HESS), in vivo mutagenicity (Micronucleus) alerts by ISS, in vitro mutagenicity (Ames test) alerts by ISS, DNA alerts for AMES, CA and MNT by OASIS, Protein binding alerts for Chromosomal aberration by OSIS, Carcinogenicity (genotox and nongeno tox) alerts by ISS, Oncologic Primary Classification, DART scheme, rtER Expert System - USEPA, Acute Oral Toxicity, Eye irritation/corrosion Inclusion rules by BfR, Skin irritation/corrosion Inclusion rules by BfR, Protein binding alerts for skin sensitization according to GHS, Protein Binding Potency h-CLAT, Protein binding alerts for skin sensitization by OASIS, Respiratory sensitization.
Information on the POD used to calculate the PDE was obtained for 287 chemicals out of the 923 E&Ls. In most cases where the lowest NOEL was obtained from toxicity studies in dogs or mice, the results were from short-term tests, so data from the longer-term rat toxicity studies in the ECHA chemicals database were adopted, resulting in the data from rats being adopted as the basis for the PDE determination. In cases where the final PDE was obtained from POD in animals other than rats, the chemicals were excluded from the analysis. The study types for the POD data obtained are shown in Table 3. Toxicity studies used to derive the PDE for the oral route were subacute studies (n=56), subchronic studies (n=110), chronic studies (n=35), combined repeated dose toxicity studies with the reproductive/developmental (repro/dev) toxicity screening studies (n=55), various types of repro/dev toxicity studies including multigenerational repro/dev toxicity studies and repro/dev screening studies (n=24), or carcinogenicity studies (n=7) (Table. 3).
| Study type | N |
|---|---|
| Subacute study | 56 |
| Subchronic study | 110 |
| Chronic study | 35 |
| Combined repeated dose toxicity studies with the reproductive/developmental toxicity screening study | 55 |
| Reproductive/developmental toxicity study | 24 |
| Carcinogenic study | 7 |
| Total | 287 |
The PDEs for the oral route calculated for the 287 chemicals ranged from 5 µg/day to 2,250 mg/day. The distribution of the PDEs is shown in Fig. 2.
Distribution of the oral PDEs for the oral administration route. This figure shows the distribution of the log10 values of the PDEs for the oral administration route of the 287 chemicals.
The modifying factor (α), calculated as the ratio of AUC and Cmax following intravenous and oral administration calculated using the PBPK model ranged from 0.0656 to 1.000 (Ave. 0.814, Med. 0.924, SD 0.242) and 0.0234 to 0.999 (Ave. 0.623, Med. 0.659, SD 0.290), respectively.
Derivation of the PDEs for intravenous exposureThe distribution of the estimated PDEs for intravenous exposure is shown in Fig. 3.
Distribution of the PDEs for the intravenous administration route. This figure shows the distribution of log10 for intravenous PDEs of 287 chemicals. (a) PDE calculated using the modified factor (α) based on the ratio of AUC. (b) PDE calculated using the modified factor (α) based on the ratio of Cmax.
The PDEs calculated based on the ratios of the AUC and Cmax following intravenous and oral administration ranged from 0.99 µg/day to 2,185 mg/day and 0.34 µg/day to 1,435 mg/day, respectively.
Estimation of TTCivThe TTC for oral administration (TTCoral) and the TTCiv calculated from the PDEs for the intravenous route using modifying factor (α) are shown in Table 4. The TTCoral was 168 µg/day. The TTCiv derived in this study was 116 µg/day for AUC and 80.3 µg/day for Cmax.
| Calculation based on AUC | Calculation based on Cmax | |
|---|---|---|
| modifying factor (α) | 116 μg/day | 80 μg/day |
| modifying factor (α) adjusted for prediction uncertainty |
39 μg/day | 27 μg/day |
Regarding the uncertainty in predicting the bioavailability, the TTCiv was adjusted using the modifying factor 3 based on the maximum value of the RMSE reported in the verification of the HTTK model (Breen et al., 2021; Wang, 2010; Linakis et al., 2020; Wambaugh et al., 2018). When route extrapolation was performed using α adjusted for prediction uncertainty, the TTCiv was 39 µg/day for AUC and 27 µg/day for Cmax.
Chemical properties profilingThe chemical properties of 287 chemicals for which the TTCiv was estimated and other E&Ls were characterized using the ICE Chemical Characterization Tool (Fig. 4). Chemical properties profiling was performed for all the 287 chemicals and for 527 chemicals of the E&Ls. As shown in Fig. 4(a), the PCA plots based on physicochemical properties were similar for both the 527 E&Ls and 287 chemicals, indicating that the 287 chemicals used to estimate the TTCiv cover the chemical space of all the 527 E&Ls. As shown in Fig. 4(b), the PCA plot based on molecular descriptors shows that, although a few chemicals fall outside the domain of the 287 chemicals, the 287 chemicals used to estimate TTCiv mostly cover the chemical space of the 527 E&Ls. Therefore, we concluded that the 287 chemicals represented an appropriate dataset for the TTCiv analysis for E&Ls.
PCA plots of the 527 chemicals in list of E&Ls and 287 chemicals used for the TTCiv derivation. (a) shows the result of a dynamic PCA based on the chemical properties; (b) shows the result of a dynamic PCA based on the molecular descriptors. The dynamic PCA plot shows the similarities in chemical properties among the members of each list of chemicals. Dynamic PCA plots are among the best of ways to compare the properties of two lists of chemicals. The chemicals in the E&Ls list of 527 chemicals are shown in blue and those in the list of 287 chemicals used for calculation of the PDE for intravenous exposure are shown in red. Each axis in the dynamic PCA plot represents the principal component (PC) describing a percentage of the data variability. The first PC (PC1) describes the highest percentage of variability within the data and the second PC (PC2) describes the second highest percentage of variability.
Figure 5 shows the percentage of chemicals with endpoint-specific profilers related to human health effects and HESS profilers in the OECD QSAR Toolbox. Analysis of the chemical characteristics using the OECD QSAR Toolbox indicated that the 923 E&Ls and 287 chemicals have almost the same profiles and the list of 287 chemicals contained a higher percentage of chemicals that matched profilers than the list of 923 E&Ls. Therefore, using the 287 chemicals was conservative for the TTC analysis.
Toxicity Profiles.
The need for development of assessment methods for chemicals lacking toxicity information for parenteral (e.g., intravenous route) exposure remains one of the key challenges in E&L risk management. In this study, oral PDEs were derived for 287 chemicals. However, for about 70% of the 923 chemicals in the list of E&Ls in this study, toxicity information to derive the oral PDE was lacking, which made it impossible to derive the PDEs based on chemical-specific toxicity data. The TTC approach is effective as the method of first choice for evaluating chemicals for which toxicity information is not available in E&L risk assessments. In this study, we used reliable toxicity data to determine the oral PDE and used transparent methods to derive the intravenous PDE. We then estimated the TTCiv from the intravenous PDEs.
We derived the oral PDEs from the toxicity data as the first step in estimating the TTCiv after determining the POD for each chemical, including chemicals of concern for genotoxicity. The oral PDEs for the 287 chemicals ranged from 5 µg/day to 2,250 mg/day, indicating that the range of toxicity of the E&Ls was quite wide, the spectrum extending from high toxicity to low toxicity.
For route extrapolation from the oral to the intravenous route, we attempted to apply a correction based on the bioavailability using blood concentration (AUC or Cmax) predictions from the ICE PBPK model. The ratio of the blood concentrations following intravenous and oral administration was calculated and used as a modifying factor (α) for extrapolation to the intravenous dose required to achieve a blood concentration equivalent to that following the oral dose.
The Solve_pbtk model in the ICE PBPK model is a multi-compartment model included in the U.S. Environmental Protection Agency’s (EPA’s) HTTK R package (Pearce et al., 2017). The ICE PBPK model output provides the concentration of the chemical over time in the plasma and each tissue compartment, the half-life, AUC, and Cmax in the plasma and each tissue compartment, and the steady state concentration (Css) in the plasma for some models (ICE PBPK Tool User Guide, ICE version 4.1: revised Aug 2024). Masuda et al. conducted route extrapolation using the PBPK model in the commercial model, ADMET Predictor, and derived TTC values for the parenteral route for the assessment of E&Ls (Masuda-Herrera, 2022). One of the features of the HTTK model in the ICE PBPK model used in this study is that it is a publicly available transparent model whose model design, input parameters, and formulas for calculating blood concentrations have been disclosed. Additionally, it is linked to chemical-specific data. If experimental data are unavailable, physicochemical properties can be predicted by in silico models within ICE, so that high throughput toxicokinetic predictions are available with minimal data for a wide range of chemicals (Abedini et al., 2021).
The primary concern in this analysis is the uncertainty in AUC and Cmax because these are based on estimated values, that is, the absence of authentic data that could be utilized for validation. Several validation results have been reported regarding the discrepancy between predicted and observed values in the HTTK model (Breen et al., 2021; Wang, 2010; Linakis et al., 2020; Wambaugh et al., 2018). Among these, the RMSE for peak concentration and AUC were shown to be around 2-3 times higher at the maximum. Validation involving both pharmaceutical and non-pharmaceutical chemicals indicated that the RMSE for peak concentration was 2.2 times, and that for AUC was 1.64 times (Breen, et al., 2021). Additionally, the World Health Organization (WHO) has indicated that in the validation of PBPK models used for risk assessment, simulations that are on average within a factor of 2 of the experimental data could be considered adequate (WHO, 2010).
Regarding the uncertainty in predictions, the TTCiv was calculated using the value obtained after dividing by 3, based on the RMSE reported in the verification of the HTTK model. The TTCiv derived in this study, when route extrapolation was performed using the modifying factor (α) without adjustment, was 116 µg/day for AUC and 80 µg/day for Cmax. When route extrapolation was performed using modifying factor (α) adjusted for prediction uncertainty, the TTCiv was was 39 µg/day for AUC and 27 µg/day for Cmax. In conclusion, we propose a TTCiv of 27 µg/day/human for E&Ls, on the basis of the ratio of Cmax.
In general, pharmaceuticals that cause concentration-dependent effects are considered as being dependent on Cmax, whereas those that exhibit time-dependent effects are considered as being dependent on the AUC and/or Cmax. In the toxicity evaluation of chemicals, including E&Ls, it is often difficult to obtain information on the pharmacokinetics, making it challenging to determine whether the observed toxic effects are time-dependent or concentration-dependent. Therefore, in this study, we adopted modifying factors (α) based on the Cmax because the Cmax-based approach provides a more conservative correction between the PDEs than the AUC-based approach. A TTC value of 35 μg/day for parenteral routes over a 10-year period to a lifetime has been proposed by the ELSIE consortium (Masuda-Herrera, 2022). However, the PQRI guidance does not provide a recommended threshold for chemicals without genotoxicity, sensitization, or irritation concerns (PQRI, 2021). The extrapolation methods and datasets used for the TTC of 35 μg/day for parenteral routes proposed by Masuda et al. for exposures over 10 years were different, yet it is close to the TTCiv (AUC based: 39 µg/day, Cmax based: 27 µg/day) derived by us. Chemicals with intravenous PDE below the TTCiv and low oral absorption rates may require individual assessment rather than estimation of the toxicity values by route extrapolation.
In the management of chemicals such as pharmaceutical impurities, chemicals that are genotoxic or identified with genotoxic alerts are managed with a TTC of 1.5 µg/day based on the cancer risk. For chemicals with short-term exposure, the less than lifetime (LTL) consideration is applied, and they are managed with a higher TTC (ICH M7 guidelines). In the assessment of general toxicity, there are no established methods for adjusting the exposure period from long-term to short-term, so similar methods to those for are not applied. In this study, we calculated the TTC values for E&Ls including chemicals of potential genotoxic concern for long-term exposure (>10year). Adjustments for short-term exposure will need to be considered in the future.
The TTC values are affected by the chemical space of the dataset. Analysis of the chemical characteristics using the ICE Characterization tool indicated that the 287 chemicals used to estimate the TTCiv covered a chemical space of other E&Ls. Analysis using the human health effects profiler of the OECD QSAR toolbox showed that the 287 chemicals had a high proportion of chemicals with profilers, so that we considered that using the 287 chemicals for the TTC analysis was conservative, and that therefore, the TTCiv based on the toxicity data of the 287 chemicals is appropriate as a threshold for E&Ls.
Our approach for route extrapolation using the PBPK model could contribute to the establishment of scientifically reliable TTCs for evaluation of chemicals without toxicity information, which is required for the risk assessment of E&Ls with intravenous exposure.
DisclaimerThe NOELs derived in this study are based on the opinion of the authors and do not reflect the views of any organization. When using data from the ECHA chemicals database, it is necessary to consider the various rights of the registrants and/or their organizations.
I would like to express my sincere gratitude to my advisor, Dr. Masahiro Takayoshi, for his invaluable guidance and support throughout this research. This research was supported by AMED under Grant Number JP22mk0101233.
Conflict of interestThe authors declare that there is no conflict of interest.
