Hazard Assessment Based on Chemical Categories

Read-across and in silico techniques are used extensively in the chemical and pharma world to fill data gaps for structurally similar substances. Extensive experience in applying these techniques was gained under two voluntary high production volume (HPV) chemical programs – the International Council of Chemical Associations' (ICCA) Cooperative Chemicals Assessment Program (with the cooperation of the Organization of Economic Cooperation and Development) and the U.S. Environmental Protection Agency's HPV Challenge Program. These programs resulted in a compilation of publicly available baseline sets of health and environmental effects data for thousands of chemicals. The American Cleaning Institute's (ACI) contribution to these national and global efforts included the compilation of these datasets for 261 substances. Chemicals that have structural similarities are likely to have similar environmental fate, physical-chemical and toxicological properties, which was confirmed by examining available data from across the range of substances within categories of structurally similar HPV chemicals. These similarities allowed the utilization of read-across, trend analysis techniques and qualitative structure activity relationship ((Q)SAR) tools to fill data gaps. 

A.What is a Chemical Category?

A chemical category is a group of chemicals whose physicochemical and toxicological properties are likely to be similar or follow a regular pattern as a result of structural similarity. The similarities may be based on: a common functional group; the likelihood of common precursors and/or breakdown products, via physical or biological processes, which result in structurally similar chemicals; and an incremental and constant change across the category. OECD expands the category approach to also include: a common mode or mechanism of action or adverse outcome pathway; and common constituents or chemical classes, similar carbon range numbers. This is frequently the case with complex substances often known as "substances of unknown or variable composition, complex reaction products or biological material".

B. Generic Process The ?rst step in making a hazard assessment of a chemical is to compile and assess the adequacy of existing information on each of the environmental and toxicological endpoints that constitute a base set of data. When adequate information is not available for that chemical, then options for completing the dataset are evaluated. In following the principle of avoiding or minimizing the use of animals in tests, alternative methods such as using data from like-substances (read-across) and quantitative structure-activity relationship analysis ((Q)SAR) 

Hazard Assessment for Chemical Categories tools are highly useful methods to ful?ll data requirements for meeting various regulatory programs, and for regulatory programs that allow for these methods to ?ll data gaps. 

C. Regulatory Guidance's Many guidance documents present approaches for grouping chemicals into categories (OECD; European Chemical Agency (ECHA); European Centre for Ecotoxicology and Toxicology of Chemicals (ECETOC). OECD Guidance encompasses different possibilities and interpretations. For the purpose of developing screening level hazard assessments for chemicals undertaken within the OECD Cooperative Chemicals Assessment Program (CoCAP, formerly the OECD HPV Chemicals Program) and the US HPV Challenge Program, American Cleaning Institute (ACI) used a category approach to ?ll hazard data gaps where they existed across 261 substances. The Screening Information Data Set (SIDS), a set of 18 hazard endpoints consisting of: 10 physical-chemical properties and environmental fate parameters; ?ve human health endpoints comprised of acute, repeated-dose, reproductive, developmental and genetic toxicities; and three ecological endpoints including acute ?sh, invertebrate and plant toxicities, were used as the baseline dataset for the two programs. In Addition to applying read-across and in silico methods using existing data for these 261 chemicals, existing data from an additional 46 substances, or "supporting substances", with similar structures were used to help ?ll out the 261 datasets. The use of in silico chemical pro?ling is an integral part of EPA's chemical assessments, as one of the known bene?ts was a reduction in testing resulting in fewer test animals needed to ?ll out datasets, as opposed to conducting tests to ?ll out data gaps for each individual chemical. An additional advantage of a chemical category assessment approach is that identi?cation of consistent patterns of effects within a category in itself increases con?dence in the reliability of the endpoint

Physical-chemical properties Melting Point Boiling Point Water Solubility Partition Coef?cient Vapor Pressure Environmental fate Photodegradation Fugacity Biodegradation Hydrolysis Environmental toxicity Acute Toxicity to Fish Acute Toxicity to Aquatic Invertebrates Acute Toxicity to Aquatic Plants Chronic Toxicity to Aquatic Invertebratesa Mammalian toxicity Acute Toxicity Repeated Dose Toxicity Reproductive Toxicity Developmental Effects Bacterial Mutagenicity Mammalian Mutagenicity values for all the individual chemicals in the category, compared to evaluation of data purely on a chemical-by-chemical basis. The use of alternative methods such as read-across, trend analysis, and (Q)SARs can enhance the understanding of the behavior of chemicals without the use of animal testing.

D. Read-across Method Read-across and trend analysis are also methods used to deduce the physicalchemical and (eco-) toxicological properties of chemicals lacking speci?c test data. Read-across is used across regulatory jurisdictions, including, but not limited to the UK's Health and Safety Executive (HSE), the UK Environmental Agency, Environment and Health Canada, U.S. EPA, and the European Chemicals Agency (ECHA), particularly, as a means to ?ll data gaps for information requirements under speci?c regulations. 1. The Principle of the Read-Across Technique is that endpoint or test information for one chemical is used to predict the same endpoint or test outcome for another chemical considered to be similar by scienti?c justi?cation. There are four approaches for read-across: a.One-To-One (one analogue used to make an estimation of a single chemical); Many-to-one (two or more analogues used to make an estimation for a single chemical); b. One-To-Many (one analogue used to make estimations for two or more chemicals); c. Many-To-Many (two or more analogues used to make the estimation for two or more chemicals). Read-across can be qualitative or quantitative. i. In Qualitative Read-Across, the presence (or absence) of a property/activity for a data-poor chemical is inferred from the presence (or absence) of the same property/activity for one or more data-rich chemicals. ii. In Quantitative Read-Across, the known value(s) of a property for one or more data-rich chemicals is used to estimate the unknown value of the same property for the target chemical. Read-across can be done via interpolation or extrapolation. Interpolation can be performed where trends in toxicity or factors in?uencing toxicity have been identi?ed and the category members arranged in line with the trend (for example in the order of increasing carbon chain lengths), data from category members on either side of a data-poor category member can be used to predict its hazards. Extrapolation methods involve identifying trends in toxicity or factors that in?uence toxicity and, again arranging the category members in line with the trend, predicting the hazards of data-poor category members through the use of data from category members at the other end of the category. E. Category Endpoint For a given category endpoint, the category members are related by a trend such that the properties change in a predictable manner and there is a pattern in the changing potency of the properties across the category. For example, a category with increasing chain length, with a common functional group, will affect solubility/log Kow, which in turn may affect bioavailability and hence toxicity, both mammalian and aquatic. Analysis of these changes is referred to as trend analysis. A demonstration of consistent trends in the behavior of a category of chemicals is one of the desirable attributes of a chemical category and one of the indicators that a common mechanism for all chemicals may be involved. When some chemicals in a category have measured values and a consistent trend is observed, missing values can be estimated. For some endpoints, external (Q)SAR models or expert systems exist and can be used to ?ll data gaps. These models are different from the internal (Q)SAR models established within a trend analysis in that these systems were not developed as part of the category formation process. The data gap ?lling is recommended to only be done for compounds that ?t into the applicability domain of the selected (Q)SAR model/expert system. Computer programs and platforms such as EPI Suite?, SPARC, and the OECD QSAR Toolbox are examples of these expert systems. a. Estimation Programs Interface (EPI Suite)? is a U.S. EPA modeling program that estimates physical-chemical properties and environmental fate properties of chemicals. SPARC Performs Automated Reasoning in Chemistry (SPARC) is also a predictive modeling system that calculates a large number of physical-chemical properties from molecular structure across all classes of industrial organic chemicals. SPARC execution involves the classi?cation of molecular structures and the selection and execution of appropriate mechanistic models, such as induction, resonance, and ?eld effects to quantify reactivity. b. The OECD QSAR Toolbox is a platform that incorporates various modules and databases from other sources to group chemicals into categories and ?lls (eco) toxicity data gaps to assist in the assessment of the hazards of chemicals. The category approach used in the Toolbox: i. Focuses on Intrinsic Properties of Chemicals (mechanism or mode of action, (eco-) toxicological effects ii. Allows for Entire Categories of Chemicals to be assessed when only a few category members are tested, saving costs and the need for testing on animals, and; iii. Enables Robust Hazard Assessment through mechanistic comparisons without testing. To facilitate practical application of (Q)SAR approaches in regulatory contexts by governments and industry and to improve their regulatory acceptance, OECD initiated the (Q)SAR Project to develop principles for the validation of (Q)SAR models, guidance documents as well as the (Q)SAR Toolbox. In 2007, the European Regulation EC 1907/2006, better known as Registration, Evaluation, Authorization, and Restriction of Chemical substances (REACH), included in silico models, namely (Q)SARs, as an alternative strategy to evaluating the safety of substances in the regulation. There are a number of reasons that category assessments are used. For reasons of ef?cient resource utilization and animal welfare, it is important to reduce as much as possible the number of in vivo tests to be conducted where scienti?cally justi?able. 

• F. Category Formation and Examples of The Use of Read-Across and In Silico Methods The chemical categories included in this analysis are: aliphatic acids; aluminum alkoxides; alkyl sulfates, alkane sulfonates and ole?ns; amine oxides; fatty acid methyl esters; glycerides; hydrotropes; linear and branched alkylbenzene sulfonates (Linear Alkylate Sulfonate (LAS)/Alkyl Benzene Sulfonate (ABS); and long chain alcohols (C6-22 primary aliphatic alcohols). These chemicals are found predominately in down-the-drain industrial, institutional, household and personal care products.

Firstly, an expert evaluation of potential analogues is usually undertaken. For the 9 chemical categories described herein, the most common practice was to group substances based on the common functional group(s) and an incremental and constant change across the category (e.g., a chain-length category), often observed in physical-chemical properties. In some cases, an effect can be present or follow a trend for some but not all members of the category. For these instances, subcategories are usually established within a category. This improves the practicality and ?exibility of the category approach without altering the scienti?c basis of the approach. The largest grouping, aliphatic acids (78 substances), was subcategorized into 14 subgroups based on the degree of saturation, single or multi-constituency, and presence and type of salt. For the long chain alcohols (C622 primary aliphatic alcohols), an intrinsic subcategorization was recognized as decreasing water solubility and increasing lipophilicity was observed with increasing chain length, leading to a cut-off for acute aquatic toxicity effects at C13 to C14 and around C15 for chronic effects. At C > 18, biodegradability was reduced. Sub categories were also established for hydrotropes; alkyl sulfates, alkane sulfonates and a-ole?n sulfonates; and glycerides. While not subcategorized, structural differences were noted in the amine oxides, aluminum alkoxides, LAS/ ABS, and fatty acid methyl esters categories. By way of definition, A hydrotrope is a compound that solubilizes hydrophobic compounds in aqueous solutions. Typically, hydrotropes consist of a hydrophilic part and a hydrophobic part (like surfactants) but the hydrophobic part is generally too small to cause spontaneous self-aggregation. Once all the data were assembled, a hazard assessment is generated for each chemical category. Within the assessment, the justi?cation of the category and the uses of read-across, trend analysis and (Q)SARs to ?ll data gaps are usually presented. a. Category Description - Hydrotropes Category Compounds known as hydrotropes are amphiphilic substances composed of both a hydrophilic and a hydrophobic functional group. The hydrophobic part of the molecule is a benzene substituted (i.e., methyl [common name: toluene], dimethyl [common name: xylene] or methyl ethyl [common name: cumene]) apolar segment. The hydrophilic, polar segment is an anionic sulfonate group accompanied by a counter ion (e.g., sodium and ammonium). This segment is a comparatively short side-chain. The hydrotropes category may be initially considered as three sub-groups: the methyl, dimethyl, and methyl ethyl benzene sulfonates (or the toluene, xylene and cumene sulfonates). Although, the counter ion will also determine the physical and chemical behavior of the compounds, the chemical reactivity and classi?cation for this purpose is not expected to be affected by the difference in counter ion (i.e., Na+, NH+, Ca++, or K+). In general, the presence of one or two methyl groups or a methyl-ethyl group on the benzene ring is not expected to have a signi?cant in?uence on chemical reactivity. Alkyl substituents are known to be weak ortho- and para-directing activators and the difference between methyl and methyl-ethyl are negligible. Going from methylbenzene to dimethylbenzene and to methyl-ethyl-benzene, the number of carbon atoms - and thus the organic character - increases. This progression improves solubility in a polar solvent and reduces solubility in polar solvents like water. Hence, reactivity in aqueous solutions may differ somewhat for the hydrotropes. However, the decisive factor in determining water solubility of these compounds is an ionic character, not the number and identity of the alkyl substituents on the benzene ring.

It is, therefore, safe to assume that the three sub-groups are expected to be generally comparable and predictable in their chemical behavior (as such or in solution) and that members from one sub-group may be useful for reading across to other sub-groups and to the hydrotropes category as a whole.

b. Application of (Q)SAR - Hydrotropes Category As illustrated by the hydrotropes category, endpoint values for many of the physical-chemical and environmental fate endpoints can be ?lled through the use of external (Q)SAR methods (EPI Suite?, https://www.epa.gov/tsca-screeningtools/epi-suitetm-estimation-program-interface). Forty-four of 56 physicalchemical and environmental fate endpoints were modeled using the EPI Suite. 

c. Examples of Read-Across - Hydrotropes Category The hydrotropes category has been assessed for mutagenic potential in a variety of in vivo and in vitro assays. Speci?cally, mouse micronucleus cytogenic assays with calcium xylene sulfonate and sodium cumene sulfonate, Ames assay with calcium xylene sulfonate, sodium cumene sulphonate and sodium xylene sulphonate and mouse lymphoma, sister chromatid exchange and chromosome aberration assays with sodium xylene sulfonate. No positive results were seen in vitro or in vivo in any of the studies. Therefore, the available data indicate that the chemicals in the hydrotropes category do not have a genotoxic potential. Read-across was also applied to the reproductive toxicity endpoint. No reproductive toxicity studies are reported for the hydrotropes category. However, a 91-day oral rat feeding study with sodium cumene sulfonate, a 90-day feeding study with sodium xylene sulfonate and a 90-day and 2-year dermal studies with sodium xylene sulfonate included the examination of sex organs such as the prostate, testes, and ovaries. There is no evidence from these repeat dose studies to suggest that these chemicals would have an adverse effect on reproductive organs and thus these negative results were applied to the category. One substance (calcium xylene sulfonate) has been evaluated for the potential to cause developmental toxicity in rats. Following US EPA TSCA Guideline 1985, no treatment-related effects were observed at 0, 150, 1500, or 3000 mg/kg body weight (bw)/ day. Therefore, the NOAEL for maternal and fetal toxicity was the highest dose tested at 3000 mg/kg bw/day (corresponding to 936 mg active ingredient/kg bw/day) across the category.

d . Quanti?cation of Bene?ts It is important to understand the process of building a well- constructed category because of the major impact the use of read-across and in silico methods can have on ful?lling hazard endpoints and thereby eliminating the need for testing. The number of tests not needed after applying read- across and in silico techniques was calculated by adding the number of substances without data with a Klimisch score of "1" or "2" (modeled or measured) for physical-chemical and environmental fate endpoints. Incidentally, the Klimisch score is a method of assessing the reliability of toxicological studies, mainly for regulatory purposes, that was proposed by H.J. Klimisch, M. Andreae and U. Tillmann of the chemical company BASF in 1997. For the toxicity endpoints, only measured data with Klimisch scores of "1" or "2" (reliable category data) were acceptable. A total of 8 physical-chemical tests and four non-SIDS endpoint invertebrate chronic aquatic toxicity tests were needed to ful?ll the data requirements for all substances. Seven physical-chemical tests were performed on two members of the glycerides category (vapor pressure and partition coef?cient for CAS 538-23-8; melting and boiling points, water solubility, vapor pressure and partition coef?cient for CAS 7360-38-5) and one vapor pressure test (for CAS 1300-72-7) was needed to ful?ll requirements for the hydrotropes category. For the long chain alcohols category, reproduction tests were performed with C10, C12, C14 and C15 alcohols in accordance to OECD Test Guideline (TG) 211, D. Magna Reproduction Test and under full Good Laboratory Practices to establish a SAR for chronic aquatic effects. 

G. Conclusions Read-across, trend analysis and (Q) SAR methods can be used to address information requirements under various regulatory and voluntary programs, the acceptance of alternative approaches to testing hinges on the validity of the category. The nine categories of analysis were assembled following OECD guidelines and were accepted by OECD member country regulatory experts, and/or by the U.S. EPA. Additionally, the hydrotropes and amine oxides categories are included in the compendium of case studies that helped shape REACH guidance on chemical category assembly and read-across.  The analysis herein is speci?c to the SIDS hazard endpoints gathered for the purposes of the HPV voluntary programs under U.S. EPA and ICCA for 261 chemicals. The analysis assumes universal testing using OECD tests for ?lling data gaps. While the use of read-across and in silico techniques to ?ll data gaps for the 18 hazard endpoints across the 261 substances of this assessment are considered by the U.S. The use of read-across and in silico methods for ?lling gaps where hazard data are lacking for speci?c chemicals within categories of chemicals has many bene?ts. First and foremost, ACI consortia demonstrated a ?rm adherence to the Three R's (Replacement, Reduction, Re?nement) as guiding principles for the more ethical use of animals in testing. By eliminating the need for animal testing through the use of read-across and in silico methods where scienti?cally justi?ed by avoiding the use of hundreds of thousands of animals. Among the SIDS endpoints that require vertebrate animals, testing was completely avoided by using read-across and in silico methods. The substantial reductions in testing costs was another signi?cant bene?t achieved in the ACI program through the application of these techniques. There are two major aspects of any read-across exercise, namely assessing similarity and uncertainty. While there can be an over-arching rationale for grouping organic substances based on molecular structure and chemical properties, these similarities alone are generally not sufficient to justify a readacross prediction. Further scientific justification is normally required to justify the chemical grouping, typically including considerations of bioavailability, metabolism and biological/mechanistic plausibility. Sources of uncertainty include a variety of elements which are typically divided into two main issues: the uncertainty associated firstly with the similarity justification