Analyses of case examples: Other cases

Case 4: Identification of Isobutylparaben as a Substance of Very High Concern (SVHC)

Purpose of assessment

The aim of this assessment was to identify isobutylparaben as a substance of very high concern (SVHC) due to its endocrine disruptive properties relevant to human health. However, a low number of studies on endocrine effects of isobutylparaben prevented the assessment of isobutylparaben as a SVHC based solely on the evidence from this substance. Therefore, data gap filling by use of read-across from butylparaben as the source substance was applied. Butylparaben was already characterized as a SVHC due to endocrine disruption relevant to human health and has a very similar structure to isobutylparaben.

Description of methodology applied

This evaluation of isobutylparaben was based on read-across from butylparaben.

The first step of the read-across process was the justification of similarity between the two substances. The evaluation based the structural similarity on similarities in chemical structure, physico-chemical properties, toxicological effects and estrogenic properties. The two substances also have similar major metabolites. In addition to isobutylparaben and butylparaben, 4 other parabens (methyl, ethyl, propyl and isopropyl paraben) were included in the analysis, and trends across the category were analysed.

Following the identification and justification of the source substance, a thorough literature search was conducted, including studies from the open literature and data from ToxCast and ComTox chemicals dashboard as well as results from the Danish (Q)SAR database.

Mechanistic in vitro and in vivo data showed similar endocrine activity between the two substances (estrogenicity). Read-across from butylparaben to isobutylparaben was conducted to show adversity and a biologically plausible link between the two.

Identification of included substance

↓

Justification of similarity (6 parabens)

↓

Literature search and data acquisition (6 parabens)

↓

Read-across between 2 parabens

Workflow 4 Isobutylparaben. SVHC identification (ECHA 2022b)

Comparison with procedures described in the OECD GD 194

The assessment was an analogue approach, read-across from a source substance (butylparaben to a target substance (isobutylparaben), aiming to fill data gaps for isobutylparaben. For a comparison with the stepwise procedure recommended for analogue approaches in the OECD GD 194, each step in the procedure was outlined and assessed in table 9.

Table 9 Analysis of case 4 (Identification of Isobutylparaben as a Substance of Very High Concern (SVHC)) according to the approach recommended in the OECD GD 194

Step	Approach recommended in the OECD GD 194	Analysis of case 4 (Identification of Isobutylparaben as a Substance of Very High Concern (SVHC))
0	Check whether the substance is a member of an existing category	Not performed.
1	Identification of potential analogues	It was argued that IBP shared close structural similarity with butylparaben (BP), the only difference being that IBP had an isopropyl group at the end of the alkyl chain while BP had a butyl group. Four other parabens with shorter alkyl chain lengths were also identified, methyl-, ethyl-, propyl- and isopropylparaben. This is in line with the recommendations in the OECD GD 194.
2	Data gathering for the analogues	Data was gathered for the source substance BP. Data was also gathered for the four other structurally similar parabens identified, which in addition to BP could provide trend information relevant to the read-across. This is in line with the recommendations in the OECD GD 194.
3	Evaluation of available data for adequacy	Available data used to substantiate the read-across was evaluated to be reliable and adequate, as originating from either OECD TG studies or adequately reliable studies from the open literature. This is in line with the recommendations in the OECD GD 194.
4	Construct a matrix of data availability	Qualitative and to the extent possible, quantitative data matrixes were constructed for in vitro and in vivo endocrine activity and adverse effects for the group of six parabens, showing trends across the group with a particular emphasis on multiple lines of evidence for IBP and BP. This is in line with the recommendations in the OECD GD 194.
5	Assess the adequacy of the analogue approach and fill the data gap	The adequacy of the analogue approach was assessed as acceptable due to the demonstrated similarity in mechanistic information from a robust trend analysis including both the source substance and supported by 4 additional structural analogue substances in a congeneric series, data availability and results. It was argued that the lines of evidence showed correspondence between the results of in vitro and in vivo endocrine activity and available adversity endpoints in BP and IBP, justifying the read-across from the source substance BP to the target IBP for the data gap on adversity relating to effects on sperm quality in perinatally exposed rats. This is in line with the recommendations in the OECD GD 194. The data gap was filled in accordance with the recommendations in the OECD GD 194.
6	Document the analogue approach	The identification of IBP as an endocrine disruptor, including the information from the read-across in the analogue approach, was justified in accordance with the template for proposals for identifying substances of very high concern under the REACH Regulation (Annex VX Dossier template). This is in line with the recommendations of the OECD GD 194.

Examination according to the ECHA RAAF

In the RAAF, the analogue approach common AEs are also part of the category approach common AEs. For description of the common AEs, see table 10. The methodology for the case example of isobutyl paraben fits into both scenario 1 and 2; An analogue approach supported by biotransformation to common substances (scenario 1) and different substances having the same type of effects (scenario 2). The scenario-specific assessment elements are included in this case example, as a data matrix is constructed, and the relevant data gap(s) is filled by use of read-across.

The SVHC proposal for isobutylparaben describes that parabens are commonly metabolised (hydrolysed) by esterases with formation of a common metabolite parahydroxybenzoic acid (PHBA). Studies investigating the endocrine activity of the major metabolite PHBA show inconsistent results. With regards to metabolism of parabens in developing animals, limited capacity to metabolise BP has been found in offspring relative to dams during lactation. This suggests higher internal exposure levels of the parent molecule during the developmental phase (ECHA 2022b). In the examination according to the ECHA RAAF we therefore include scenario 2 (an analogue approach with different compounds having the same type of effect(s) as the most appropriate scenario as the read-across applied is for the data gap on adversity relating to effects on sperm quality in perinatally exposed rats.

The assessment elements (AE) specific for these scenarios are listed and assessed in table 10.

Table 10 Analysis of case 4 (Identification of Isobutylparaben as a Substance of Very High Concern (SVHC)) according to the approach recommended in the ECHA Read-across Assessment Framework (RAAF)

Code	Approach recommended in the RAAF	Analysis of case 4 (Identification of Isobutylparaben as a Substance of Very High Concern (SVHC))
AE A.1	Identity and characterization of source substance Chemical identity and impurity profile of each category member are sufficiently detailed for assessment of the category approach.	This assessment element is judged to be acceptable with high confidence. Chemical identity was reported by EC numbers and CAS numbers. Names, synonyms and structures were also reported for all substances. Both source and target substances are monoconstituent substances.
AE A.2	Link of structural similarities and differences with the proposed prediction A category hypothesis has been provided and whether it applies to all category members.	This assessment element is judged to be acceptable with high confidence due to the clear substance identification available for both source and target substances and the data availability in vitro and in silico predictions.
AE A.3	Reliability and adequacy of the source study(ies) Study design of source substance(s) fulfills the information requirement and the test material(s) correctly represent source substance(s) in terms of purity and impurities.	This assessment element is judged to be acceptable with high confidence as the source substance is well described in the literature and already identified as an SVHC based on its endocrine disrupting properties for human health.
AE A.4	Bias that influences the prediction Is inclusion of other structurally similar substances in the category possible or would they change the prediction of properties for the target substance(s)? The source substance(s) used for the predictions corresponds to the reliable study(ies) giving rise to the highest concern for the properties under consideration.	This assessment element is judged to be acceptable with high confidence as other parabens were assessed as potential source substances, and the final approach selected BP as the best suited source substance with supporting evidence from four other bisphenols providing information to a trend analysis.
	Scenario 2-specific AEs
AE 2.1	Substances the test organism is exposed to Have the substances the test organism is exposed to after administration of the source and target substances been identified?	This assessment element is judged to be acceptable with medium confidence (minor reservations). The level of mother substances and relevant metabolites, respectively, are not measured in all available in vivo studies. However, the major metabolite is identified to be similar between the target and source substances.
AE 2.2	Common underlying mechanism, qualitative aspects Has the common underlying mechanism for the observed effects been established and does it allow a prediction of qualitative similar effects?	This assessment element is judged to be acceptable with high confidence as a common underlying mechanisms has been identified. Qualitative aspects of the mechanisms of the source and target substances were shown in the data matrix.
AE 2.3	Common underlying mechanism, quantitative aspects Has it been established that the common underlying mechanism leads to the same quantitative outcome for the source and target substances?	This assessment element is judged to be acceptable with high confidence as potency of the mechanism is judged to be similar between the two substances. The quantitative aspects of the underlying mechanisms of the source and target substances were included in the data matrix, when available.
AE 2.4	Exposure to other substances than those linked to the prediction Has the possibility of other compounds than those linked to the prediction being present (impurities) or formed (intermediates, metabolites) been considered and if so, what their influence on the prediction is?	This assessment element is judged to be acceptable with high confidence as source and target substance were shown to have similar major metabolites, and their potential contribution to the effects observed was taken into consideration.
AE 2.5	Occurrence of other effects than covered by the hypothesis and justification Can other acting mechanisms of the source and/or target substance than the hypothesized one be present and contribute to the observed toxicity and can they impact the prediction?	This assessment element is judged to be acceptable with high confidence as mechanistic and toxicological information of source and target substances was described in detail.

Discussion

In this case example, a category was formed, a data matrix was developed and data gap filling by read-across is conducted. The outcome of the case was identification of isobutylparaben as a Substance of Very High Concern (SVHC) due to its endocrine disruptive properties relevant to human health under the REACH Regulation.

Advantages

The case example illustrated the value of developing a solid data matrix with all relevant data included. Although it was an analogue approach read-across between two substances, the case was supported by a trend analysis for four other structurally similar parabens with similar effects. The case is a solid example of a read-across that involves available in silico, in vitro, in vivo and ADME data to support the case. The approach was also supported by the fact that a lot of data was available for the included substances.

Challenges

When performing data gap filling by read-across, interpolation is preferred as it leads to a more robust read-across prediction, but extrapolation is often used in analogue approaches due to the limited number of substances involved, unless the hypothesis is that the source and target substances are transformed to common metabolites. In this case the read-across was an extrapolation performed between the two members of the paraben group with the longest carbon-backbone.

However, the read-across hypothesis was supported by the trend analysis and although this was not a category approach the available data allowed interpolation across the carbon-backbone lengths of all 6 parabens, ranging from IBP (3 carbons) to BP (4 carbons) as the source substance and the four other members (1–3 carbons).

Uncertainties

Not many uncertainties were identified in this case.

Regulatory perspective

This case is an example of a successfully applied read-across analogue approach in a regulatory context.

Learnings

A solid data matrix (with ADME, in silico, in vitro, in vivo information) is a key tool for data gap filling.

Learnings specific for endocrine disruptors

Similar structures combined with ADME, in silico, in vitro (and if available in vivo) data on endocrine activity can be used as a basis for read-across of adversity info for successful identification of endocrine disruptors based on data gap filling by read-across.

Case 5: Brominated flame retardants

Purpose of assessment

The aim of the project was to make preliminary structural groupings of brominated flame retardants (BFRs) for possible future work to address these by grouping approaches rather than as individual substances, as the regulation of a single substance is very time demanding and the process needs to start over if the molecular structure changes even slightly to have a new substance, possibly leading to regrettable substitution. One group was chosen for further work to explore identification of a critical health effect based on literature search and apply information from the experimental literature data and (Q)SAR predictions generated for all group members to make a preliminary justification for read-across for the critical effect in a category approach (Danish EPA, 2016).

Description of methodology applied

The grouping was conducted in several steps as illustrated in workflow 5.

Initially, 85 BFRs were identified. Following the identification of the BFRs, structural information was gathered in the format of SMILES (Simplified Molecular-Input Line-Entry System) strings where possible, for use in (Q)SAR model predictions.

Substances incompatible with (Q)SAR analyses were excluded leaving 67 substances for further analysis. The 67 BFRs were predicted in >150 (Q)SAR models. In a next step, in vitro experimental data from the US programmes ToxCast and Tox21 were retrieved . The generated (Q)SAR predictions along with the ToxCast/Tox21 information was submitted to the Leadscope Predictive Data Miner software to group the substances. The grouping (termed clustering in Leadscope) was based on structural similarity alone or structural similarity plus (Q)SAR predictions. Clustering based on Tox21/ToxCast data did not yield meaningful groups as only 6 substances had such data.

The Leadscope clustering led to 15 groups and 7 singletons that could not be attributed to any of the groups:

Aromatic substances – 10 substances
Dibromo-(2,3-dibromopropoxy)benzene derivatives – 5 substances
Cycloalkanes – 5 substances
Phthalate acid and its anhydride – 2 substances
Phenols and bisphenols – 5 substances
Benzyl ethyl oxygen bridge derivates – 2 substances
Methoxy dibromobenzene derivatives – 5 substances
Phthalates/Benzoate – 4 substances
TBBPA ethers – 5 substances
Small linear and branched brominated alkyl alcohols – 4 substances
TBBPA esters/acrylate – 3 substances
Phosphates – 2 substances
Triazines – 3 substances
Biphenyles – 2 substances
Diphenyl ethers – 4 substances

The group “small linear and branched brominated alkyl alcohols” was chosen for further analysis due to positive (Q)SAR predictions of genotoxicity and cancer for all 4 members, REACH registrations for 2 members and Tox21/ToxCast data availability for 3 out of the 4 members. Based on the 4 members of this group, structural boundaries for a “category” were defined and all theoretical BFRs with structures living up to the category definition were determined, resulting in a category with 61 members. The 61 members of this new category were subjected to further (Q)SAR analyses using human health related (Q)SAR models and a literature / database search for the 25 members with a CAS RN was performed to attempt determining the critical human health effects. The literature / database search retrieved relevant results for 3 of the category members. The critical effect of these three members was identified to be multiple-organ carcinogenic effect, most probably exerted by a genotoxic mode of action either by the parent or metabolite(s). Read-across for the critical effect from the three category members with experimental data and the one member with a classification for the identified critical effect to the remaining 57 structurally similar target analogues in the category was not performed but observations were gathered to support possible follow-up work.

Identification of brutto list of 85 substances

↓

(Q)SAR predictions possible for 67 substances

↓

ToxCast and Tox21 data for 6 substances

↓

67 substances in 15 groups based on Clustering in Leadscope

↓

1 group selected: QSAR, REACH registrations ToxCast data

↓

Further work on selected group of 4 substances

↓

New category with 61 members formed

↓

(Q)SAR on group structural similarity

↓

Literature search for 25 members with CAS RN

↓

Final group of 61 substances with genotoxic/carcinogenic potential

Workflow 5 Brominated flame retardants (Danish EPA, 2016)

Comparison with procedures described in the OECD GD 194

The grouping of brominated flame retardants led to the formation of a preliminary category of small linear and branched brominated alkyl alcohols with suspected genotoxic and carcinogenic potential. For a comparison with the stepwise procedure recommended for category approaches in the OECD GD 194, each step in the procedure was outlined and assessed in table 7.

Table 7 Analysis of case 5 (Brominated flame retardants) according to the approach recommended in the OECD GD 194

Step	Approach recommended in the OECD GD 194	Analysis of case 5 (Brominated flame retardants)
0	Check whether the substance is a member of an existing category	For the generated preliminary category: “small linear and branched brominated alkyl alcohols”, the OECD (Q)SAR Application Toolbox was used to check whether any of the 61 identified category members were associated with an existing category. This approach is in line with the recommendations in the OECD GD 194.
1	Develop category hypothesis and definition and identify category members	The generated group “small linear and branched brominated alkyl alcohols” consisted of 4 substances and was chosen for further analysis due to positive (Q)SAR predictions of genotoxicity and cancer as well as REACH registrations for 2 members and ToxCast data availability for multiple members. A working definition for a preliminary category was made and all theoretical structural members were identified (61 including the 4 members of the original group). Based on the 4 substances from the original group, the boundaries for a preliminary category were defined as: Small linear and branched brominated alkyl alcohols having 3–5 carbons, 2–3 bromine atoms and 1–2 alcohol groups. This approach is in line with the recommendations in the OECD GD 194. Impurity profiles for the substances for which experimental data was found, were not mapped and taken into consideration. This is not in line with the OECD GD 194.
2	Gather data for each category member	In the generated preliminary category: “small linear and branched brominated alkyl alcohols”, (Q)SARs predictions were generated for the 61 members of the category and a literature search for experimental data on human health effects was performed for the 25 substances which had a CAS RN. This approach is in line with the recommendations in the OECD GD 194.
3	Evaluate available data for adequacy	Literature searches were conducted and all retrieved data were checked in order to find relevant data on human health useful for performing a category approach and read-across. This is in line with the recommendations in the OECD GD 194.
4	Construct a matrix of data availability For all category members vs. endpoints, arranged in suitable order to reflect trends or progression actors the category. Supporting data may be better represented graphically, e.g.usingaheatmap.	For the generated preliminary category: “small linear and branched brominated alkyl alcohols”, a data matrix was constructed for the 4 source substances, including (Q)SAR predictions and experimental data, including in vitro data. Heat maps for (Q)SAR predictions were constructed for all 61 members. Data on absorption, distribution, metabolism and excreting (ADME) were available for one of the three members of the category identified in the generated preliminary category. This approach is in line with the recommendations in the OECD GD 194.
5	Perform a preliminary evaluation of the category and fill data gaps	The generated preliminary category: “small linear and branched brominated alkyl alcohols” was evaluated based on available experimental toxicological information for three group members and (Q)SAR predictions for all group members. This approach is in line with the recommendations in the OECD GD 194. For the generated preliminary category: “small linear and branched brominated alkyl alcohols”, (Q)SAR predictions for all group members were generated for both mechanistic and AO endpoints within the group boundaries. However, data gap filling was prepared for further work but not finalized. Results were communicated in a transparent way. This is in accordance with the recommendations of the OECD GD 194.
6	Perform and/or propose testing	Not performed.
7	Perform a further evaluation of the category	The adequacy of the preliminary category approach was documented in the report with preliminary hypothesis/justification and recommendations for further work. This is in accordance with the recommendations of the OECD GD 194.
8	Document the finalized category and refine the category rationale	The preliminary category approach and proposals for further work to complete read-across was documented in the report in accordance with the recommendations of the OECD GD 194.

Examination according to the ECHA RAAF

In the ECHA RAAF, the methodology in the category approach fits into scenario 6; A category approach with no variation among the category members, different substances having the same type of effects. The assessment elements (AE) specific for this scenario are listed and assessed in table 8. The scenario-specific assessment elements are included in this case example, as a data matrix is constructed, and the relevant data gap(s) is filled by use of read-across. The case example fits into scenario 4 or 6, targeting categories of different substances with qualitatively similar properties. Since it in the case example was not evaluated whether there are variations in properties observed among the source substances, it could not be further evaluated whether the case example fits into scenario 4 or 6. The assessment elements are also similar, so a distinction is not necessary, and Assessment Element (AE) “4.1/6.1” to “4.5/6.5” are included in the table.

Table 8 Analysis of case 5 (Brominated flame retardants) according to the approach recommended in the ECHA Read-across Assessment Framework (RAAF)

Code	Approach recommended in the RAAF	Analysis of case 5 (Brominated flame retardants)
AE C.1	Substance characterization Chemical identity and impurity profile of each category member are sufficiently detailed for assessment of the category approach.	For the generated preliminary category: “small linear and branched brominated alkyl alcohols” this assessment element is judged to be acceptable with medium confidence as impurity profiles for the tested source substances were not assessed. However, chemical identity was reported by CAS RN and EC numbers, names, structures and SMILES for all members for the preliminary category. The preliminary category was defined as having 3–5 carbons, 2–3 bromine atoms and 1–2 alcohol groups. The working definition of the category was made based on the span of the four source substances in the preliminary structural grouping exercise. The final group had 61 members.
AE C.2	Structural similarity and differences within category Structural similarities among all members are identified and structural differences allowed within the category are described.	For the generated preliminary category: “small linear and branched brominated alkyl alcohols”, this assessment element is judged to be acceptable with medium confidence. Investigation of the relevance of sub-categorization based on further mechanistic analyses was attempted but not finalized but is relevant for further work. Structural substance information was described and assessed in detail throughout the report.
AE C.3	Link of structural similarities and structural differences with the proposed regular pattern A category hypothesis has been provided and whether it applies to all category members.	For the generated preliminary category: “small linear and branched brominated alkyl alcohols”, this assessment element is judged to be acceptable with medium confidence, as it was identified that there may be differences in the genotoxic mechanism (direct acting versus active metabolite(s) and this needs further investigation. Structural substance information was described and assessed in detail throughout the report. The identified sources of experimental information could give promise of possible successful read-across and although all three substances did not cluster together in the (Q)SAR based clusterings for genotoxicity, carcinogenicity, manual inspection of the predictions showed good consistency in the (Q)SAR predictions between the members.
AE C.4	Consistency of effects in data matrix Construct a data matrix for all category members vs. existing experimental data, arranged in suitable order to reflect trends or progression across the category.	For the generated preliminary category: “small linear and branched brominated alkyl alcohols”, this assessment element is judged to be acceptable with medium confidence, as it was identified but not fully investigated that there may be different mechanisms of the genotoxic cancer for different members. Heat maps based on the (Q)SAR predictions were presented in the report as well as experimental data for the 3 substances identified in the initial group identification plus one additional substance. No experimental data for the remaining 21 substances were found in the literature.
AE C.5	Reliability and adequacy of the source study(ies) Study design of source substance(s) fulfills the information requirement and the test material(s) correctly represent source substance(s) in terms of purity and impurities.	For the generated preliminary category: “small linear and branched brominated alkyl alcohols”, this assessment element is judged to be acceptable with high confidence. All three members were included in the training set for the DTU (Q)SAR Ames model with positive experimental results and 2,3- DBPA was included with positive experimental results in the DTU (QSAR) models for chromosomal aberrations in CHL cells and SHE cell transformation in vitro. All three had positive predicted indications in all included models for carcinogenicity and had harmonized classifications for carcinogenicity.
AE C.6	Bias that influences the prediction Is inclusion of other structurally similar substances in the category possible or would they change the prediction of properties for the target substance(s)? The source substance(s) used for the predictions corresponds to the reliable study(ies) giving rise to the highest concern for the properties under consideration.	For the generated preliminary category: “small linear and branched brominated alkyl alcohols”, this assessment element is judged to be acceptable with minor reservations as impurity profiles for the substances for which experimental data was found, were not taken into consideration.
	Scenario 6-specific AEs
AE 4.1/6.1	Substances the test organism is exposed to Have the substances the test organism is exposed to after administration of the source and target substances been identified?	For the generated preliminary category: “small linear and branched brominated alkyl alcohols”, this assessment element is judged to be acceptable with medium confidence (minor reservations) as experimental data contained data from REACH registrations, harmonized and notified classifications and regulatory evaluation processes. ADME information was available for one substance.
AE 4.2/6.2	Common underlying mechanism, qualitative aspects Has the common underlying mechanism for the observed effects been established and does it allow a prediction of qualitative similar effects?	For the generated preliminary category: “small linear and branched brominated alkyl alcohols”, this assessment element is judged to be acceptable with high confidence. Metabolism profilers were run on the category members identifying multiple metabolites. (Q)SAR predictions were run for category members for bioavailability, metabolism/transformation, ED endpoints, repro/developmental, organ and repeated dose toxicity and genotoxicity and carcinogenicity. Several (Q)SAR models were run for genotoxicity and carcinogenicity for the 61 members in the category of small linear and branched brominated alkyl alcohols. All were predicted to be positive for carcinogenic and genotoxic properties indicating that they may have a carcinogenic potential with a possible mutagenic/genotoxic mode of action. The estimated specificities of the models as established by leave-many-out cross-validations are between 85.9% and 95.1%, i.e. the overall false positive rates of the models are around 5%- 14%. No ‘one single mechanistic interpretation’ in relation to mutagenicity and cancer could be established. The structural alerts indicated that all members shared the same mutagenic/genotoxic MoA with some variations in their possible mechanisms of action.
AE 4.3/6.3	Common underlying mechanism, quantitative aspects Has it been established that the common underlying mechanism leads to the same quantitative outcome for the source and target substances?	This assessment element is judged to be acceptable as although the applied (Q)SAR models did not predict potency but rather gave binary positive/negative predictions, quantitative measures may not necessarily be needed for the critical effect of genotoxic carcinogenicity.
AE 4.4/6.4	Exposure to other substances than those linked to the prediction Has the possibility of other compounds than those linked to the prediction being present (impurities) or formed (intermediates, metabolites) been considered and if so, what their influence on the prediction is?	For the generated preliminary category: “small linear and branched brominated alkyl alcohols”, this assessment element is judged to be acceptable with high confidence. Metabolites were predicted for all of the 61 category members in two Toolbox rat metabolism simulators. Profiler-predicted alerts were communicated in heat maps. The OECD toolbox contains information about metabolites of one substance.
AE 4.5/6.5	Occurrence of other effects than covered by the hypothesis and justification Can other acting mechanisms of the source and/or target substance than the hypothesized one be present and contribute to the observed toxicity and can they impact the prediction?	For the generated preliminary category: “small linear and branched brominated alkyl alcohols”, this assessment element is judged to be acceptable with medium confidence. From comprehensive analyses of experimental data for the source substances and (Q)SAR predictions for all category members the category substances seemed to share the same mutagenic/genotoxic mode of action, but with variations in their mechanisms of action. So sub-categorization was found to possibly be relevant but would require further analysis, which was, however, outside the scope of the project.

Code	Approach recommended in the RAAF	Analysis of case 5 (Brominated flame retardants)
AE C.1	Substance characterization Chemical identity and impurity profile of each category member are sufficiently detailed for assessment of the category approach.	For the generated preliminary category: “small linear and branched brominated alkyl alcohols” this assessment element is judged to be acceptable with medium confidence as impurity profiles for the tested source substances were not assessed. However, chemical identity was reported by CAS RN and EC numbers, names, structures and SMILES for all members for the preliminary category. The preliminary category was defined as having 3–5 carbons, 2–3 bromine atoms and 1–2 alcohol groups. The working definition of the category was made based on the span of the four source substances in the preliminary structural grouping exercise. The final group had 61 members.
AE C.2	Structural similarity and differences within category Structural similarities among all members are identified and structural differences allowed within the category are described.	For the generated preliminary category: “small linear and branched brominated alkyl alcohols”, this assessment element is judged to be acceptable with medium confidence. Investigation of the relevance of sub-categorization based on further mechanistic analyses was attempted but not finalized but is relevant for further work. Structural substance information was described and assessed in detail throughout the report.

Code	Approach recommended in the RAAF	Analysis of case 5 (Brominated flame retardants)
AE C.3	Link of structural similarities and structural differences with the proposed regular pattern A category hypothesis has been provided and whether it applies to all category members.	For the generated preliminary category: “small linear and branched brominated alkyl alcohols”, this assessment element is judged to be acceptable with medium confidence, as it was identified that there may be differences in the genotoxic mechanism (direct acting versus active metabolite(s) and this needs further investigation. Structural substance information was described and assessed in detail throughout the report. The identified sources of experimental information could give promise of possible successful read-across and although all three substances did not cluster together in the (Q)SAR based clusterings for genotoxicity, carcinogenicity, manual inspection of the predictions showed good consistency in the (Q)SAR predictions between the members.
AE C.4	Consistency of effects in data matrix Construct a data matrix for all category members vs. existing experimental data, arranged in suitable order to reflect trends or progression across the category.	For the generated preliminary category: “small linear and branched brominated alkyl alcohols”, this assessment element is judged to be acceptable with medium confidence, as it was identified but not fully investigated that there may be different mechanisms of the genotoxic cancer for different members. Heat maps based on the (Q)SAR predictions were presented in the report as well as experimental data for the 3 substances identified in the initial group identification plus one additional substance. No experimental data for the remaining 21 substances were found in the literature.
AE C.5	Reliability and adequacy of the source study(ies) Study design of source substance(s) fulfills the information requirement and the test material(s) correctly represent source substance(s) in terms of purity and impurities.	For the generated preliminary category: “small linear and branched brominated alkyl alcohols”, this assessment element is judged to be acceptable with high confidence. All three members were included in the training set for the DTU (Q)SAR Ames model with positive experimental results and 2,3- DBPA was included with positive experimental results in the DTU (QSAR) models for chromosomal aberrations in CHL cells and SHE cell transformation in vitro. All three had positive predicted indications in all included models for carcinogenicity and had harmonized classifications for carcinogenicity.

Code	Approach recommended in the RAAF	Analysis of case 5 (Brominated flame retardants)
AE C.6	Bias that influences the prediction Is inclusion of other structurally similar substances in the category possible or would they change the prediction of properties for the target substance(s)? The source substance(s) used for the predictions corresponds to the reliable study(ies) giving rise to the highest concern for the properties under consideration.	For the generated preliminary category: “small linear and branched brominated alkyl alcohols”, this assessment element is judged to be acceptable with minor reservations as impurity profiles for the substances for which experimental data was found, were not taken into consideration.
	Scenario 6-specific AEs
AE 4.1/6.1	Substances the test organism is exposed to Have the substances the test organism is exposed to after administration of the source and target substances been identified?	For the generated preliminary category: “small linear and branched brominated alkyl alcohols”, this assessment element is judged to be acceptable with medium confidence (minor reservations) as experimental data contained data from REACH registrations, harmonized and notified classifications and regulatory evaluation processes. ADME information was available for one substance.
AE 4.2/6.2	Common underlying mechanism, qualitative aspects Has the common underlying mechanism for the observed effects been established and does it allow a prediction of qualitative similar effects?	For the generated preliminary category: “small linear and branched brominated alkyl alcohols”, this assessment element is judged to be acceptable with high confidence. Metabolism profilers were run on the category members identifying multiple metabolites. (Q)SAR predictions were run for category members for bioavailability, metabolism/transformation, ED endpoints, repro/developmental, organ and repeated dose toxicity and genotoxicity and carcinogenicity. Several (Q)SAR models were run for genotoxicity and carcinogenicity for the 61 members in the category of small linear and branched brominated alkyl alcohols. All were predicted to be positive for carcinogenic and genotoxic properties indicating that they may have a carcinogenic potential with a possible mutagenic/genotoxic mode of action. The estimated specificities of the models as established by leave-many-out cross-validations are between 85.9% and 95.1%, i.e. the overall false positive rates of the models are around 5%- 14%. No ‘one single mechanistic interpretation’ in relation to mutagenicity and cancer could be established. The structural alerts indicated that all members shared the same mutagenic/genotoxic MoA with some variations in their possible mechanisms of action.

Code	Approach recommended in the RAAF	Analysis of case 5 (Brominated flame retardants)
	Scenario 6-specific AEs
AE 4.3/6.3	Common underlying mechanism, quantitative aspects Has it been established that the common underlying mechanism leads to the same quantitative outcome for the source and target substances?	This assessment element is judged to be acceptable as although the applied (Q)SAR models did not predict potency but rather gave binary positive/negative predictions, quantitative measures may not necessarily be needed for the critical effect of genotoxic carcinogenicity.
AE 4.4/6.4	Exposure to other substances than those linked to the prediction Has the possibility of other compounds than those linked to the prediction being present (impurities) or formed (intermediates, metabolites) been considered and if so, what their influence on the prediction is?	For the generated preliminary category: “small linear and branched brominated alkyl alcohols”, this assessment element is judged to be acceptable with high confidence. Metabolites were predicted for all of the 61 category members in two Toolbox rat metabolism simulators. Profiler-predicted alerts were communicated in heat maps. The OECD toolbox contains information about metabolites of one substance.
AE 4.5/6.5	Occurrence of other effects than covered by the hypothesis and justification Can other acting mechanisms of the source and/or target substance than the hypothesized one be present and contribute to the observed toxicity and can they impact the prediction?	For the generated preliminary category: “small linear and branched brominated alkyl alcohols”, this assessment element is judged to be acceptable with medium confidence. From comprehensive analyses of experimental data for the source substances and (Q)SAR predictions for all category members the category substances seemed to share the same mutagenic/genotoxic mode of action, but with variations in their mechanisms of action. So sub-categorization was found to possibly be relevant but would require further analysis, which was, however, outside the scope of the project.

Discussion of the case example: Brominated flame retardants

The initial (Q)SAR analyses of 85 identified BFRs resulted in 15 groups of BFRs with relatively few members in each. For the initial structurally similar group of small alkyl alcohols in focus (4 substances), the subsequent category definition resulted in a large group of structurally similar substances category members (61 substances) including the 4 from the initial list of BFRs.

Advantages

(Q)SARs were performed in a structured and transparent manner with hypotheses and assessments illustrating the role modelling can have in assessing – also large - categories of structurally similar substances. For the 61 substances in the final group, the structural alerts indicated that all members shared the same mutagenic/genotoxic MoA with some variations in their possible mechanisms of action.

Based on reliable (Q)SAR predictions, the few category members (for which experimental data existed, and which shared comparable toxicological effects), could be used as source substances for data gap filling by read-across to the remaining group members. The substances with experimental data were thus central for data gap filling within the group.

Challenges

With 15 groups formed from the initially identified 85 BFRs, the example shows that despite the same primary use as flame retardants and the presence of similar structural components, including the same halogen, it is not possible to address BFRs homogeneously from a grouping and regulatory point of view.

(Q)SAR predictions for many different toxicological endpoints were obtained. However, as experimental data in general was scarce and available only for a very limited number of group members, the occurrence of these additional structural alerts was not possible to evaluate further.

Uncertainties

Not many uncertainties were identified for this study. It is noted that impurity profiles for the substances for which experimental data was found, were not taken into consideration.

Regulatory perspectives

Following up on this grouping exercise, the brominated flame retardant BMP received a harmonized classification as carcinogenic (Carc. 1B) and mutagenic (Muta 1B) in 2018 based on available experimental data (ECHA 2018). In 2020, the structurally similar substance TBNPA got a harmonized classification as carcinogenic (Carc. 1B) and mutagenic (Muta 1B) based on read-across from both BMP and 2,3-DBPA, which also is a structurally similar brominated flame retardant with a harmonized classification, due to lack of relevant experimental data for TBNPA (ECHA 2020a).

Further, in 2023 ECHA published a regulatory strategy to identify substances used as flame retardants, assess the needs for regulatory risk management and through grouping aim at faster regulatory action for these substances (ECHA 2023a).

ECHA included in its assessment information on use and exposure of the flame retardants and suggested that information was adequate for the two reactive alcohols (BMP and TBNPA) to initiate a restriction based on their carcinogenic properties.

ECHA further identified aromatic brominated flame retardants as of general concern due to their known or potential PBT/vPvB properties (ECHA 2023a). Aromatic brominated structures were included in several different subcategories in the category approach by DK EPA 2016 and thus suggested a potential for further work on grouping based on other endpoints than carcinogenicity and mutagenicity/genotoxicity. This could be further explored by using (Q)SARs. Several new ED-related QSARs have been developed since the work of the DK EPA 2016 report and could also be included in future follow-up work.

The use of combinations of reliable (Q)SAR predictions and in vitro data to support data gap filling of adverse effects by read-across could be applied more extensively in the future to regulate larger groups of chemicals.

Learnings

It may be valuable to start with broad groups of substances, also including substances not in current use, to have as wide a structural domain and as many potential source substances as possible.
Structural similarity can be a good starting point for formation of groups
(Q)SAR predictions can be used to search for similar effect profiles across (sub-)groups
A data matrix is a useful tool for analysing potential data gaps to possibly be filled by read-across.

Learnings specific for endocrine disruptors

None.