Development of adverse outcome pathways relevant for the identification of substances having endocrine disruption properties Uterine adenocarcinoma as adverse outcome.

Annex B.1_ Methodological Approach Annex B.2_ Reduced availability of GnRH description and quantification Annex B.3_ Dose and Temporal concordance Annex B.4_Plausible MIEs Annex B.1_ Methodological Approach Annex B.2_ Reduced availability of GnRH description and quantification Annex B.3_ Dose and Temporal concordance Annex B.4_Plausible MIEs

The PPR Panel recommends submitting the AOPs developed to the OECD AOP program to further support regulatory uptake. The PPR Panel recommends that the uncertainties identified in the AOPs and putative AOP network should be further explored and possibly resolved including additional data. Further development would include fine‐tuning and harmonising the methodologies for quantifying the certainty of the KERs (e.g. weighing system for the three criteria) and of the overall certainty of AOPs and AOP network (e.g. using Bayesian network). Quantitative understanding of the proposed KERs and of the overall AOPs, and putative AOP network is recommended to enhance the usability of AOPs in regulatory risk assessment. This can be achieved for instance using sequence of response–response models with dedicated experiments. It is highly likely that more AOPs from the AOP knowledge base (AOP‐KB) may link into this putative AOP network and that multiple layers, such as different taxonomies and life stages, or branches linking other MIEs, KEs and/or AOs to this network are conceivable. Specifically, feedback loops, which are well‐established in, e.g. the HPG axis, are clearly lacking. Adding such a layer would greatly enhance the applicability of the putative AOP network in this Scientific Opinion. Since AOPs are chemically agnostic, the developed AOPs could also be useful for chemicals other than PPPs for which ED properties have to be assessed. Particularly, for data‐poor substances, the developed AOPs could help to initiate AOP‐informed IATA. Oestrous cycle monitoring and biomarkers of E2 actions, e.g. vaginal cornification, changes in (circulating) hormone levels, could be included in chronic toxicity studies to indirectly identify increased E2 availability, associated with uterine adenocarcinoma and/or disruption of the hypothalamic‐pituitary‐ovarian (HPO) axis. Investigation of NAMs (e.g. standardised tests exploring estrogenic activity; in vitro assays; PBPK modelling exploring correlation between blood and tissue E2 levels) application in the context of some of the MIEs/Kes currently part of the AOP network is recommended. Further guidance, from the OECD, on the structuring of AOP networks would be useful to support further representation of the complexity of biological response. In light of the need for methodological development and harmonisation across EFSA activities, the PPR Panel recommends EFSA to organise an internal workshop to collect best practices and share experiences in the context of the AOP development. The PPR Panel recommends submitting the AOPs developed to the OECD AOP program to further support regulatory uptake. The PPR Panel recommends that the uncertainties identified in the AOPs and putative AOP network should be further explored and possibly resolved including additional data. Further development would include fine‐tuning and harmonising the methodologies for quantifying the certainty of the KERs (e.g. weighing system for the three criteria) and of the overall certainty of AOPs and AOP network (e.g. using Bayesian network). Quantitative understanding of the proposed KERs and of the overall AOPs, and putative AOP network is recommended to enhance the usability of AOPs in regulatory risk assessment. This can be achieved for instance using sequence of response–response models with dedicated experiments. It is highly likely that more AOPs from the AOP knowledge base (AOP‐KB) may link into this putative AOP network and that multiple layers, such as different taxonomies and life stages, or branches linking other MIEs, KEs and/or AOs to this network are conceivable. Specifically, feedback loops, which are well‐established in, e.g. the HPG axis, are clearly lacking. Adding such a layer would greatly enhance the applicability of the putative AOP network in this Scientific Opinion. Since AOPs are chemically agnostic, the developed AOPs could also be useful for chemicals other than PPPs for which ED properties have to be assessed. Particularly, for data‐poor substances, the developed AOPs could help to initiate AOP‐informed IATA. Oestrous cycle monitoring and biomarkers of E2 actions, e.g. vaginal cornification, changes in (circulating) hormone levels, could be included in chronic toxicity studies to indirectly identify increased E2 availability, associated with uterine adenocarcinoma and/or disruption of the hypothalamic‐pituitary‐ovarian (HPO) axis. Investigation of NAMs (e.g. standardised tests exploring estrogenic activity; in vitro assays; PBPK modelling exploring correlation between blood and tissue E2 levels) application in the context of some of the MIEs/Kes currently part of the AOP network is recommended. Further guidance, from the OECD, on the structuring of AOP networks would be useful to support further representation of the complexity of biological response. In light of the need for methodological development and harmonisation across EFSA activities, the PPR Panel recommends EFSA to organise an internal workshop to collect best practices and share experiences in the context of the AOP development.

To address the assigned task, i.e. to develop AOP(s) relevant for the identification of substances having endocrine‐disrupting properties and leading to uterine adenocarcinoma as AO, EFSA established an ad hoc expert Working Group of the PPR Panel (from this point onwards referred as working group (WG)), that met regularly to address this PPR Panel self‐task. 3 The WG based its activity on the AOP OECD methodology, that is well suited to support the implementation of the EFSA‐ECHA ED guidance. The AOP OECD frame, and associated methodology, aims at describing a sequence of measurable mechanistic KEs linked by defined key event relationships (KERs), which are expected to be triggered by the activation of a MIE. This cascade of KEs finally results in an AO of regulatory relevance. This approach is in line with the criteria of the EFSA‐ECHA ED guidance and offers the advantage of being ‘chemically agnostic’, being therefore applicable to any chemical substance under assessment. The WG carried out a preparatory exercise to gather elements useful for the AOPs development. First, AOPs related to uterine neoplasms were collected from AOP‐Wiki and information on known stressors causing uterine neoplasms as an apical effect were retrieved from the literature. Second, to consolidate the meaning and the interpretation of the AO, the WG assessed the scientific information on uterine neoplasms in women and animal models from a clinicopathological perspective (see Appendix  A ). Based on this revision, it was concluded that the selected AO in rodent models (uterine adenocarcinoma) corresponds to type I (endometrioid) endometrial carcinoma (EC) in women. Type I EC derives from a multistep process preceded by atypical hyperplasia, and such progression has been described in several models in rodents, sharing many similarities with lesions described in women. Then, AOPs were postulated through collection of data (i.e. National Toxicology Program (NTP) studies, literature) on chemicals stressors, using the AO uterine adenocarcinoma as apical adverse effect and for which putative MoAs have been proposed. The stressors inducing uterine adenocarcinoma and other neoplasms via a genotoxic MoA have not been considered further since genotoxicity is not an endocrine MoA and it is already addressed as a stand‐alone endpoint in regulatory assessment. From this initial evaluation, multiple MIE and pathways were identified. Of relevance was the identification of the preliminary common blocks (critical nodes): (1) estrogen dominance/hormonal measurements and (2) increased proliferation in estrogen‐dependent tissue/organs; these suggested the possibility to define an AOP network (Figure  1 ). Postulated AOPs and AOP network. The increase in proliferation and the hormonal levels (e.g. including estrogen dominance and increase of estrogen levels) were identified as dominant/critical KEs and defined the possibility to develop an AOP network. The green (last) AOP refers to exposure to estrogenic chemicals (e.g. diethylstilbestrol (DES)) during early development (Suen et al.,  2018 ) A dedicated protocol was elaborated (Viviani et al.,  2023 – Annex A ) to provide a plan detailing the strategy for the development of the postulated AOPs and AOP network. The protocol was based on EFSA ( 2020 ) and articulated in various phases some of which iterative. An overview of the phases is provided in Figure  2 and are summarised below: – definition of the scope of the assessment and identification of problem formulation questions (based on preliminary problem formulation set by the WG, see Section  1.2 ) and of initial postulated AOPs (conceptual model) (see Figure  1 ); – refinement of the structure of the postulated AOPs based on an iterative process; this consisted in (a) exploration of the available scientific literature and expert judgement for the identification of the most plausible MIEs/KEs/AOPs; (b) assessment of the biological plausibility of KERs and justification of the sequence and selection and inclusion of new KEs if needed; – definition of the final AOP structure; – prioritisation of KEs and KERs whose knowledge is less advanced and requires a structured, systematic and comprehensive approach in the evidence retrieval, appraisal and analysis; – definition of the methods for conducting the systematic retrieval, screening for relevance and appraisal of the evidence available in the literature; – overall evidence synthesis and integration (WoE) based on biological plausibility, essentiality and empirical support; – assessment of the overall AOPs certainty. definition of the scope of the assessment and identification of problem formulation questions (based on preliminary problem formulation set by the WG, see Section  1.2 ) and of initial postulated AOPs (conceptual model) (see Figure  1 ); refinement of the structure of the postulated AOPs based on an iterative process; this consisted in (a) exploration of the available scientific literature and expert judgement for the identification of the most plausible MIEs/KEs/AOPs; (b) assessment of the biological plausibility of KERs and justification of the sequence and selection and inclusion of new KEs if needed; definition of the final AOP structure; prioritisation of KEs and KERs whose knowledge is less advanced and requires a structured, systematic and comprehensive approach in the evidence retrieval, appraisal and analysis; definition of the methods for conducting the systematic retrieval, screening for relevance and appraisal of the evidence available in the literature; overall evidence synthesis and integration (WoE) based on biological plausibility, essentiality and empirical support; assessment of the overall AOPs certainty. Adapted from Annex A Viviani et al.,  2023 – Summary of the phases for the scientific assessment process as reflected in the protocol (Phase 1, Phase 2a, Phase 2b, Phase 2c, Phase 3a, Phase 4, Phase 5) (Annex A, from Viviani et al.,  2023 ). The split in subgroups (box with dashed boards) and Phase 3b were not defined a priori The protocol did not anticipate the decision of splitting the WG in subgroups (dashed grey box in Figure  2 ) that was made later on in the process and represents an amendment to the plan. An evidence‐based approach was adopted in the development of the AOPs. The gradient of comprehensiveness and systematicity varied across subgroups though reflecting the level of certainty and consolidated knowledge available for the various components of the AOP. This was considered appropriate also in light of the discussions taking place in EFSA and more generally in the scientific community on the best way to adapt the use of systematic approaches in the AOP framework (Hoffmann et al., 2022 ; Svingen et al.,  2021 ). Detailed description of the phases is provided in the following paragraphs. The postulated AOP(s) and AOP network showed in Figure  1 were used as a conceptual model to develop the AOPs and address the ToRs. To refine/integrate the postulated AOPs and AOP network, an extensive mapping of the available scientific literature was performed to identify MIEs and KEs. This phase was carried out by the contractor with the contribution of the WG; the methodology and results are described in detail in Viviani et al. ( 2023 ) – Annexes A and B . Briefly, the search strategy to map the evidence for the postulated AOPs from ‘Estrogenic activity’ (MIE) to ‘Uterine adenocarcinoma’ (AO) included targeted combinations of search terms with relevance to the field of the MIE and AO. The collected records have been scanned by Topic modelling, a machine learning technique that clusters papers according to their semantic similarity and provides sets of words (topics) describing the body of evidence using a probabilistic model that assigns likelihood to each paper to be related to a specific topic. The method allows to explore a large set of scientific papers in an automatic way helping the mapping towards KEs possibly relevant for the AOP in a fast way. In accordance with their biological domains, relevant topics were clustered in MIE, KE, stressors, general molecular mechanisms, extra‐uterine events (do not refer to the uterus) (Figure  3 ). Adapted from Viviani et al. ( 2023 ). General pathways and effectors (MIEs and KEs) identified by topic modelling. Each box refers to the topics identified as part of the iterative process using topic modelling. Extra‐uterine: collectively, these topics tag human pathological conditions (i.e. obesity, anovulation, polycystic ovary) leading to unopposed estrogen exposure (Passarello et al.,  2019 ) and physiological pathways (i.e. oestrous cyclicity, neuronal control of sexual behaviour, ovary and hormonal regulation) with a control on estrogen production. Intra‐uterine: ‘MIEs’, ‘KEs’ and ‘General mechanisms’ have been clustered in ten subgroups. These topics provide more downstream events that are potentially leading to uterine adenocarcinoma. Details are reported in Annex D in Viviani et al. ( 2023 ) Considering the scope of the procurement, the contractor focused on MIEs (and consequent KEs) that are expected to be activated at the target site (uterine endometrium) and for which prototypical stressors are available (i.e. tamoxifen and estradiol (E2) (Figure  3 intra‐uterine factors – topics 1, 5 and 6)). The WG considered other topics (i.e. estrogen metabolism, induced imbalance in sex steroid hormones) related to MIEs/KEs that are expected to be activated before (‘upstream’) the events at the target site. Some of these (Figure  3 – extra uterine factors topics 3 and 4, intra‐uterine factor topic 2) were selected on the basis of the preliminary analysis conducted (see Sections  2.1 and 2.2 ). At this stage, it was decided to develop in dedicated subWGs the AOPs including the identified upstream topics, i.e. – senescence or chemical‐induced imbalance in sex steroid hormones (subsequently modified into hypothalamic–pituitary–gonadal (HPG) axis perturbation‐induced imbalance in sex steroid hormones); – and estrogen metabolism. Consistently, the postulated AOP network was refined (Figure  4 ). Refined AOP network after topic modelling. In bold upstream events occurring before the activation of estrogen receptor (ER) in the uterine mucosa Based on the outcome of Phase 2, it was decided to differentiate the approaches to retrieve, extract and appraise the evidence for KEs and KERs included in the postulated AOPs. This choice was driven at first by the definition of well‐established vs less‐established knowledge on the biological plausibility of KERs in the scientific community pertaining to the different AOPs. The AOP including ER activation in the uterus and downstream KEs leading to the AO was carried out by the contractor. Methods were planned in advance with few exceptions (i.e. methods for synthesis and integration only generically described in the protocol) and are extensively described in Viviani et al. ( 2023 ). The AOPs dealing with upstream KEs occurring before ER activation in the uterus were assigned to two subgroups of the WG (subWGs). A less systematic, yet structured approach (at least for evidence retrieval and selection) was adopted for the development of these AOPs. The methodologies used by the two sub‐groups included expert knowledge and structured literature searches and are described in Annexes A.1 and B.1 , respectively. The AOP including ER activation in the uterus and downstream KEs leading to the AO was carried out by the contractor. Methods were planned in advance with few exceptions (i.e. methods for synthesis and integration only generically described in the protocol) and are extensively described in Viviani et al. ( 2023 ). The AOPs dealing with upstream KEs occurring before ER activation in the uterus were assigned to two subgroups of the WG (subWGs). A less systematic, yet structured approach (at least for evidence retrieval and selection) was adopted for the development of these AOPs. The methodologies used by the two sub‐groups included expert knowledge and structured literature searches and are described in Annexes A.1 and B.1 , respectively. For all the KERs included in some of the putative AOPs, an uncertainty analysis was carried out to identify the possible inconsistencies or limitations in the evidence supporting a particular linkage. The AOP handbook mentions the ‘confidence in a KER’ in relation to the uncertainty assessment. In this opinion, the concept of confidence was used as a starting point for the qualitative evaluation (e.g. ‘low’, ‘moderate’ or ‘high’) of the three criteria biological plausibility, empirical support and KEs essentiality (OECD, 2018d ). KEs essentiality was addressed as part of the KER. It is acknowledged that the OECD AOP users’ handbook provides the possibility to evaluate the essentiality as element of the AOP as a whole, i.e. essentiality of the KE for the activation of the downstream pathway, and therefore not specific of each individual KER, or indeed as part of the specific KER description, i.e. essentiality of the KE up for the activation of the adjacent KE down. The final aim of the exercise was to quantify the certainty in each individual KER and express it using the ‘conditional probability that a downstream KE would occur when an adjacent upstream KE occurs’. The conditional probability of occurrence was framed expressing the certainty rather than the uncertainty considering the fact that the KERs in the AOP are considered to have at least a minimum level of biological plausibility. It was agreed that a probability distribution was the best way to express a KER certainty. A single probability value was considered insufficient to express the possible divergencies across experts. The KER certainty quantification was conducted in two steps. In step 1, a qualitative judgement (low, moderate, high) was elicited by each individual expert for each of the three criteria (biological plausibility, empirical evidence and essentiality) using an expert knowledge elicitation (EKE) approach (EFSA, 2014) based on the evidence collected by each subgroup on the respective AOP and KERs (Annexes A.2 , B.2 and C ) and the collegial discussion in the WG. The main divergencies in the individual judgements were discussed and whenever possible the purpose was to achieve a consensus. In cases when a consensus could not be achieved, divergent judgements were kept. In step 2, each qualitative individual expert judgement was translated into quantitative scores. It was agreed to use equally sized ranges as domains of the probability distributions (i.e. truncated normals) expressing quantitatively the qualitative judgements (‘low’, ‘moderate’ or ‘high’). The central value of the domain was adopted as the means of the distributions. The certainty on the three criteria were combined using weights reflecting the influence of the three criteria on the overall certainty of a KER. A probability distribution is the result of the combination. The biological plausibility was assigned a weight of 0.60, whereas 0.20 was the weight given to the empirical evidence and the KE essentiality. The rationale for giving a larger weight to the biological plausibility is that it represents a guiding principle in the AOP framework and in the ECHA‐EFSA ED guidance. The other two criteria are relying on evidence that is related to specific stressors and therefore are inherently contributing to increase certainty in the relationship between KEs with lower strength compared to biological plausibility. The set of weights and the shape of the distribution was agreed before starting the elicitation. It was acknowledged that the choice on the weighing system and the shape of the distribution remains subjective and could vary in future EFSA's mandates (see Section  5 , Future Perspective and Recommendation). More details on the methodology used to quantify the KER certainty are provided in Annex D . The KERs included in some of the identified putative AOPs were assessed for their biological plausibility, empirical support and the KEs essentiality (see Section  2.3.5 ). The assessment has been performed for each individual KER but not for the AOPs or for the AOP network as a whole. A qualitative evaluation was then carried out for these AOPs (see Section  3.2.2 ). A quantitative assessment was not carried out; consistently, a quantification of the certainty of the putative AOP network was not conducted. In previous AOP developments (EFSA PPR Panel, 2021 ) the quantification of the overall certainty for each AOP, and for the entire AOP network was provided. The quantification was achieved combining the conditional probabilities for each KER certainty using a probabilistic graphical model (i.e. Bayesian network). This exercise would have implied estimate of additional conditional probabilities for each KE given all its upstream KEs (more than one in a AOP network) translating in different conditional probabilities for the individual AOP's KER. This activity was considered not achievable within the available time and resources allocated to this mandate. Therefore, it was not pursued in this Scientific Opinion (see Section  5 , Future Perspective and Recommendation).

protein kinase B adverse outcome adverse outcome pathway AT‐rich interaction domain 1A body mass index bisphenol A bromodeoxyuridine E‐cadherin cyclin‐dependent kinase 4 and 6 cyclin dependent kinase inhibitor 2A catenin β1 aromatase diethylstilbestrol deoxyribonucleic acid estrone estradiol/17β‐estradiol estradiol/progesterone estriol estrogenic, androgenic, thyroidal, and steroidogenic endometrial carcinoma European Chemical Agency endocrine disruption European Food Safety Authority epidermal growth factor receptor enzyme‐linked immunosorbent assay estrogen receptor Erb‐B2 receptor tyrosine kinase 2 extracellular signal‐regulated kinase 1 and 2 estrogen receptor α Fischer 344 follicle‐stimulating hormone gamma‐aminobutyric acid‐ergic genetically engineered mice genistein gonadotropin‐releasing hormone human mutL homolog 1 hereditary nonpolyposis colorectal carcinoma hypothalamus pituitary gonadal hypothalamic-pituitary-ovarian hydroxysteroid dehydrogenase enzymes integrated approaches to testing and assessment International Programme on Chemical Safety Joint Research Center knowledge base key event key event relationship Kirsten rat sarcoma virus luteinising hormone liquid–liquid extraction regulator of the tumour suppressor p53 microsatellite instability molecular initiating event mitogen inducible gene 6 N ‐methyl‐ N ‐nitrosourea mode of action messenger ribonucleic acid neuromedin U nonspecific molecular profile National Toxicology Programme Organisation for Economic Co‐operation and Development Organisation for Economic Co‐operation and Development Conceptual Framework ovariectomised p16 cyclin dependent kinase inhibitor 2A progesterone physiologically based pharmacokinetic proliferating cell nuclear antigen progesterone receptor gene phosphatidylinositol‐3‐kinase phosphatidylinositol‐4,5‐bisphosphate 3‐kinase catalytic subunit alpha postnatal day polymerase ɛ protection of telomeres protein 1A Plant Protection Products and their Residues plant protection products progesterone receptor phosphatase and tensin homolog persistent vaginal cornification Registration, Evaluation, Authorisation and Restriction of Chemicals radioimmunoassay reverse transcription polymerase chain reaction steroid hormone binding globulin somatic copy‐number alterations Sprague‐Dawley sine oculis–related homeobox 1 solid phase extraction steroid sulfatase sulfotransferase sulfotransferase family 1E tetrabromobisphenol A standardised test guidelines term of reference tumour protein P53 working group World Health Organization weight of evidence

The EU regulatory frame for plant protection products (PPP) and biocidal products set specific provisions aiming at phasing out endocrine disruptors (EDs); once it is proven that a substance is an ED, in principle it cannot be authorised for use. Type I endometrial carcinomas (EC), the most common uterine cancers in women, are known to be hormonally driven by unopposed estrogen signalling. As such chemicals interfering with estrogen‐receptor mediated signalling pathways could potentially be causal to EC. In regulatory settings, establishing the biologically plausible link between uterine adenocarcinoma observed in rodent models corresponding to type 1 EC and (altered) estrogen signalling may be challenging in the absence of a good mechanistic understanding, and this can preclude a firm conclusion on ED potential of a substance. For pesticide substances for which there is evidence that they may have endocrine disruptive properties, it is required to elucidate the mode/mechanism of action. Development of adverse outcome pathways (AOPs) for EC can provide a practical tool for the scientific implementation of the EFSA‐ECHA ED guidance. AOPs will permit to systematically organise available data and knowledge that describes scientifically plausible relationships between molecular initiating events (MIEs), key events (KEs) and an apical adverse outcome (AO). In this scientific opinion, the PPR Panel explored the development of AOPs for the (AO) uterine adenocarcinoma as useful tools informing about the relationship between the AO and chemicals affecting the estrogen signalling pathway. In addition, by detecting and/or identifying data gaps and/or research needs, the developed AOPs could serve to give guidance for further works. An evidence‐based approach methodology was applied, and literature reviews were produced using a structured framework assuring transparency, objectivity, and comprehensiveness. Of note, a new strategy in the categorisation of information was implemented through the implementation of a machine learning – Topic Modelling approach to the systematic review. In addition, a probabilistic quantification of the weight of evidence has been carried out. Several AOPs were developed; these converged to a common critical node, that is increased estradiol (E2) availability in the uterus followed by estrogen receptor (ER) activation in the endometrium; therefore, a putative AOP network was considered. Upstream KEs and downstream KEs (i.e. expected to be activated respectively before and after the events at critical node) of the AOPs were identified. Whenever possible, KEs from AOPs in AOP‐Wiki were used. Using the element KE, ‘ER activation’ from the critical node as MIE, downstream KEs and key event relationships (KERs) within the uterus were focused on epigenetics‐related pathways, where the knowledge is less‐established. For these, a systematic literature search was applied and a risk of bias process to the selected papers was carried out (Viviani et al.,  2023 ). MIE and the KEs and KERs upstream to the critical node were developed in individual AOPs. The knowledge on the biological plausibility of KERs for upstream components of these AOPs was considered more advanced (well‐established) and consequently a systematic literature review was not conducted, instead dedicated approaches have been developed. Expert knowledge elicitation (EKE) was the selected methodology to conduct the uncertainty analysis and weight of evidence (WoE) for some AOPs, i.e. the extra‐uterine sulfotransferase inhibition and GnRH pathways and the uterine AOPs. The KERs included in these AOPs were assessed for their biological plausibility, empirical support and essentiality of the KEs. The assessment was conducted in the context of each set of MIEs/KEs/KERs included in individual AOPs. The collected data on the AOP network were therefore evaluated qualitatively whereas a quantitative uncertainty analysis for weight of the AOP network certainty has not been performed. The uncertainties identified in the AOPs and putative AOP network should be further explored and further methodological developments for quantifying the certainty of the KERs and of the overall AOPs and AOP network are recommended. Inclusion of endpoints informing on estrogen signalling in regulatory studies could be considered, and investigation of new approach methodologies (NAMs) application in the context of some of the MIEs/KEs currently part of the putative AOP network developed is recommended.

Developing AOP should be accompanied by a search of existing content in the AOP‐Wiki to determine whether analogous AOPs and/or KEs or KERs already exist in the knowledgebase. This prevents duplicated effort and help to ensure that KEs and KERs are shared among AOPs, allowing for de facto creation of AOP networks. In the following list, existing AOPs, KEs and KERs possibly relevant for this scientific opinion. are presented. It is noted that not all the events and pathways herein described are including data and the list below aims at providing an overview of the existing titles. AOPs AOP 63: Cyclooxygenase inhibition leading to reproductive dysfunction AOP 73: Xenobiotic Inhibition of Dopamine‐beta‐Hydroxylase and subsequent reduced fecundity AOP 102: Cyclooxygenase inhibition leading to reproductive dysfunction via interference with meiotic prophase I/metaphase I transition AOP 103: Cyclooxygenase inhibition leading to reproductive dysfunction via interference with spindle assembly checkpoint AOP 112: Increased dopaminergic activity leading to endometrial adenocarcinomas (in Wistar rat) AOP 126: Alpha‐noradrenergic antagonism leads to reduced fecundity via delayed ovulation AOP 153: Aromatase Inhibition leading to Ovulation Inhibition and Decreased Fertility in Female Rats AOP 167: Early‐life estrogen receptor activity leading to endometrial carcinoma in the mouse. AOP 168: GnRH pulse disruption leading to mammary adenomas and carcinomas in the SD rat. AOP 169: GnRH pulse disruption leading to pituitary adenomas and carcinomas in the SD rat. AOP 309: Luteinising hormone receptor antagonism leading to reproductive dysfunction AOP 445: Estrogen Receptor Alpha Agonism leads to Impaired Reproduction AOP 7: Aromatase reduction leading to impaired fertility KEs/KERs KEs related to GnRH and LH KE 530 decreased, GnRH pulsatility/release in AOP 73 xenobiotic inhibition of dopamine‐beta‐hydroxylase and subsequent reduced fecundity KE 530 decreased, GnRH pulsatility/release in AOP 126 alpha‐noradrenergic antagonism leads to reduced fecundity via delayed ovulation KE 969 decreased GnRH release, decreased kisspeptin stimulation of gnrh neurons in AOP 153 aromatase inhibition leading to ovulation inhibition and decreased fertility in female rats KE 1071 decreased, GnRH pulsatility/release in hypothalamus in AOP 169 GnRH pulse disruption leading to pituitary adenomas and carcinomas in the sd rat. KE 1071 decreased, GnRH pulsatility/release in hypothalamus in AOP 168 GnRH pulse disruption leading to mammary adenomas and carcinomas in the sd rat. KE 690 Reduced, Luteinising hormone (LH), plasma KE 971 Ovulation of oocytes reduced, delayed or blocked, Decrease or delay in LH surge required for ovulation KE 531 Decreased, LH Surge in AOP 126 Alpha‐noradrenergic antagonism leads to reduced fecundity via delayed ovulation KE 1072 Decreased, LH Surge from anterior pituitary in AOP 168 GnRH pulse disruption leading to mammary adenomas and carcinomas in the SD rat. KE 970 Decreased LH release from Anterior Pituitary, Decreased GnRH stimulation of Anterior Pituitary Gonadotrophs in AOP 153 Aromatase Inhibition leading to Ovulation Inhibition and Decreased Fertility in Female Rats KE 129 Reduction, Gonadotropins, circulating concentrations KEs and KERs related to ovulation and oestrus cyclicity KE 1073 anovulation, KE 488 Decrease, Ovulation KE 532 Delayed, Ovulation KE 1695 Impaired ovulation KE 1989 Impaired, Oocyte maturation and ovulation KE 1074 Interruption, Ovulation KER 1110‐Decreased, LH Surge from anterior pituitary leads to prolonged, estrus KER: 1111 prolonged, oestrus leads to Increased, circulating estrogen levels KEs related to estrogen receptor KE 1181 Activation, Estrogen receptor KE 1065 Activation, estrogen receptor alpha KE 111 Agonism, Estrogen receptor KE 1984 Agonism, Estrogen Receptor alpha KE 112 Antagonism, Estrogen receptor KE1710 Binding to estrogen receptor (ER)‐α in immune cells KE 1180 Estrogen receptor activation, KE 748 Increased, Estrogen receptor (ER) activity KE 1064 prepubertal increase, Estrogen receptor (ER) activity KE 1046 Suppression, Estrogen receptor (ER) activity AOP 63: Cyclooxygenase inhibition leading to reproductive dysfunction AOP 73: Xenobiotic Inhibition of Dopamine‐beta‐Hydroxylase and subsequent reduced fecundity AOP 102: Cyclooxygenase inhibition leading to reproductive dysfunction via interference with meiotic prophase I/metaphase I transition AOP 103: Cyclooxygenase inhibition leading to reproductive dysfunction via interference with spindle assembly checkpoint AOP 112: Increased dopaminergic activity leading to endometrial adenocarcinomas (in Wistar rat) AOP 126: Alpha‐noradrenergic antagonism leads to reduced fecundity via delayed ovulation AOP 153: Aromatase Inhibition leading to Ovulation Inhibition and Decreased Fertility in Female Rats AOP 167: Early‐life estrogen receptor activity leading to endometrial carcinoma in the mouse. AOP 168: GnRH pulse disruption leading to mammary adenomas and carcinomas in the SD rat. AOP 169: GnRH pulse disruption leading to pituitary adenomas and carcinomas in the SD rat. AOP 309: Luteinising hormone receptor antagonism leading to reproductive dysfunction AOP 445: Estrogen Receptor Alpha Agonism leads to Impaired Reproduction AOP 7: Aromatase reduction leading to impaired fertility KEs related to GnRH and LH KE 530 decreased, GnRH pulsatility/release in AOP 73 xenobiotic inhibition of dopamine‐beta‐hydroxylase and subsequent reduced fecundity KE 530 decreased, GnRH pulsatility/release in AOP 126 alpha‐noradrenergic antagonism leads to reduced fecundity via delayed ovulation KE 969 decreased GnRH release, decreased kisspeptin stimulation of gnrh neurons in AOP 153 aromatase inhibition leading to ovulation inhibition and decreased fertility in female rats KE 1071 decreased, GnRH pulsatility/release in hypothalamus in AOP 169 GnRH pulse disruption leading to pituitary adenomas and carcinomas in the sd rat. KE 1071 decreased, GnRH pulsatility/release in hypothalamus in AOP 168 GnRH pulse disruption leading to mammary adenomas and carcinomas in the sd rat. KE 690 Reduced, Luteinising hormone (LH), plasma KE 971 Ovulation of oocytes reduced, delayed or blocked, Decrease or delay in LH surge required for ovulation KE 531 Decreased, LH Surge in AOP 126 Alpha‐noradrenergic antagonism leads to reduced fecundity via delayed ovulation KE 1072 Decreased, LH Surge from anterior pituitary in AOP 168 GnRH pulse disruption leading to mammary adenomas and carcinomas in the SD rat. KE 970 Decreased LH release from Anterior Pituitary, Decreased GnRH stimulation of Anterior Pituitary Gonadotrophs in AOP 153 Aromatase Inhibition leading to Ovulation Inhibition and Decreased Fertility in Female Rats KE 129 Reduction, Gonadotropins, circulating concentrations KEs and KERs related to ovulation and oestrus cyclicity KE 1073 anovulation, KE 488 Decrease, Ovulation KE 532 Delayed, Ovulation KE 1695 Impaired ovulation KE 1989 Impaired, Oocyte maturation and ovulation KE 1074 Interruption, Ovulation KER 1110‐Decreased, LH Surge from anterior pituitary leads to prolonged, estrus KER: 1111 prolonged, oestrus leads to Increased, circulating estrogen levels KEs related to estrogen receptor KE 1181 Activation, Estrogen receptor KE 1065 Activation, estrogen receptor alpha KE 111 Agonism, Estrogen receptor KE 1984 Agonism, Estrogen Receptor alpha KE 112 Antagonism, Estrogen receptor KE1710 Binding to estrogen receptor (ER)‐α in immune cells KE 1180 Estrogen receptor activation, KE 748 Increased, Estrogen receptor (ER) activity KE 1064 prepubertal increase, Estrogen receptor (ER) activity KE 1046 Suppression, Estrogen receptor (ER) activity

Uterine cancers represent the fourth most common malignancy in women in terms of estimated age‐standardised incidence rates in Europe (IARC, 2012 , available at: http://gco.iarc.fr// ). 4 Among uterine cancers in women, EC is the most common and among these Type I EC (also known as endometrioid adenocarcinoma) accounts for approximately 80% of non‐genetic cases. Type I EC is referred to as endometrioid adenocarcinoma since it is usually well differentiated and has a glandular growth pattern resembling normal endometrial epithelium. Since endometrium is a hormone‐dependent tissue, hormones play a fundamental role in the development of endometrial tumours. Type I EC is estrogen dependent, is hormone‐receptor‐positive, develops mainly in postmenopausal women, is low grade and usually indolent. Type I EC develops in the setting of unopposed estrogen signalling, resulting from absolute excess of estrogen or relative deficiencies of P4 (Sherman,  2000 ; Rodriguez et al.,  2019 ). It arises in a background of endometrial hyperplasia, in fact women with atypical hyperplasia have a 40% risk of concurrent cancer (Trimble et al.,  2006 ). Similar to other tumours, development of type I EC involves the stepwise acquisition of several genetic alterations in tumour suppressor genes and oncogenes, e.g. phosphatase and tensin homolog (PTEN), catenin β1 (CTNNB1), phosphatidylinositol‐4,5‐bisphosphate 3‐kinase catalytic subunit alpha (PIK3CA), AT‐Rich Interaction Domain 1A (ARID1A), Kirsten rat sarcoma virus (K‐RAS). Appendix  A contains a thorough description of EC in women, including clinicopathological and molecular classifications, aetiology, pathogenesis and morphological features. Animals represent important models for human type I EC. Rodents in particular are used to evaluate the risk of chemicals to cause EC, and to shed light into the pathogenesis of this tumour. There are numerous models that can be used for these purposes, reviewed in detail in Appendix  A . In rodents, progression from hyperplasia to uterine adenocarcinoma has been described, and the continuum of lesions detected in rodents share many similarities with lesions described in women. With regard to possible factors affecting uterine carcinogenesis in rodents, the reproductive senescence, including the age‐related increase in prolactin in rats, could be a confounding factor in the assessment of human relevance of the development of uterine adenocarcinoma and should be considered during risk assessment (Appendix  A ). Based on the evidence that type I EC is hormone‐driven by unopposed estrogen signalling, resulting from absolute excess of estrogens or relative deficiencies in P4 (Rodriguez et al.,  2019 ), therefore chemicals interfering with estrogen signalling could potentially be causal to uterine cancers. Yoshida et al. ( 2015 ) identified seven pesticides associated with treatment‐related increases in the incidence of endometrial adenocarcinomas in combined rat chronic toxicity and carcinogenicity studies. The rat strains used in these studies were Wistar rats, with the exception of two studies of pyriminobac‐methyl and benthiavalicarb‐isopropyl, where Fischer rats were used. Notably, no increase in incidence of endometrial adenocarcinoma was observed in mouse carcinogenicity studies (Yoshida et al.,  2015 ). The authors were able to predict the MoA of four compounds based on the endpoints of mechanistic studies, including modulation of estrogen metabolism (for cyenopyrafen and benthiavalicarb‐isopropyl) and increased E2 to P4 ratio (for pyriminobac‐methyl and spirodiclofen). Recent mechanistic studies of Tetrabromobisphenol A (TBBPA) have indicated the possibility that chemically induced inhibition of E2 excretion (through SULT1E1 inhibition) from the body results in increased serum E2, thereby promoting the development of uterine adenocarcinoma in rats. The relevance of this pathway in humans has not yet been established, and clarification of possible species differences in the toxicokinetics of these chemicals is considered crucial for confirmation of these results (Wikoff et al.,  2016 ). In the EU, different pieces of legislation have specific provisions aimed at phasing out EDs (i.e. chemicals in general (REACH) (Regulation (EC) No 1907/2006), PPP, biocidal products, water framework directive and medical devices). In the context of PPP under Regulation (EC) No 1107/2009, a hazard‐based approach is taken: once it is proven that a substance (i.e. an active substance, safener or synergist) is an ED, 5 the substance in principle cannot be authorised for use (COM/2018/734 final). Consequently, for pesticide substances for which there is evidence that they may have ED properties, it is required 6 to integrate the standard data set with additional information or specific studies to elucidate the MoA/mechanism of action of the substance and/or to provide sufficient evidence for relevant adverse effects. Establishing the biologically plausible link between uterine adenocarcinoma and (altered) estrogen signalling may be straightforward in case a substance induces uterine adenocarcinoma in routine rodent carcinogenicity studies (i.e. carcinogenicity studies or combined chronic toxicity/carcinogenicity studies following OECD TG 451 (OECD, 2018b) or OECD TG 453 (OECD, 2018c) respectively), affects other estrogen‐mediated parameters in the core data package and show positive results in additional studies investigating estrogen activity. In such a case, conclusion on the ED properties of the substance can be drawn according to the EFSA‐ECHA ED guidance. On the other hand, the implementation of EFSA‐ECHA ED guidance could be far more challenging for a substance inducing uterine tumours in carcinogenicity studies not accompanied by a pattern of complementary estrogen‐mediated effects in the dataset, and/or evidence of EATS endocrine activity is lacking or even negative. In such situations, postulating and developing endocrine‐mediated, as well as non‐endocrine mediated MoAs is not straightforward in the absence of a good mechanistic understanding, which precludes a firm conclusion on ED potential of the substance. Development of AOPs for uterine adenocarcinoma can provide a practical tool for the scientific implementation of the EFSA‐ECHA ED guidance also helping in discriminating between endocrine versus non‐endocrine mediated pathways. This will permit to systematically organise available data and knowledge describing scientifically plausible relationships between MIEs, KEs and uterine adenocarcinoma as apical AO. Furthermore, this could help to identify bioassay methods for investigating relevant KEs. Eventually, AOPs could be organised in an AOP network. These AOPs would be highly beneficial for the assessment of potential ED pesticide substances as a basis for postulating plausible MoAs, to decide whether available data are sufficient or which further data would be necessary, and finally to conclude on ED potential in accordance to Commission Delegated Regulation (EU) 2017/2100. Since AOPs are chemically agnostic, the developed AOPs could also be useful for any other chemical for which ED properties have to be assessed. Particularly, for data‐poor substances, the developed AOPs could help to initiate AOP‐informed integrated approaches to testing and assessment (IATA) (see Section  5 , Future perspective, and Recommendation). Several AOPs were developed. The AOP from uterine (endometrial) estrogen receptor alpha activation (MIE) to uterine adenocarcinoma (AO) was developed by the contractor (Viviani et al., 2023 ). This AOP represents the more downstream component of the AOP network. AOPs were developed covering pathways upstream to the activation of the ERα in the uterine mucosa, as follows: estrogen sulfotransferase (SULT1E1) inhibition; 17β‐hydroxysteroid dehydrogenase enzymes (HSD17B) inhibition; aromatase (CYP19A1) induction: these cover MIEs/KEs/KERs related to disruption of estrogen metabolism; gonadotropin‐releasing hormone (GnRH) reduction covering KEs/KERs related to perturbation of HPG axis. estrogen sulfotransferase (SULT1E1) inhibition; 17β‐hydroxysteroid dehydrogenase enzymes (HSD17B) inhibition; aromatase (CYP19A1) induction: these cover MIEs/KEs/KERs related to disruption of estrogen metabolism; gonadotropin‐releasing hormone (GnRH) reduction covering KEs/KERs related to perturbation of HPG axis. KER certainty quantification has been conducted in the context of this scientific opinion and is described below. The AOP from SULT1E1 inhibition to ER activation in the uterus was developed in full: KEs and KERs were described and KERs certainty was quantified (see Section  3.2.3 ). The AOPs from HSD17B inhibition and aromatase induction were described but the KERs were not subjected to KERs certainty quantification (putative). A summary of these AOPs is presented below, details are available in Annex A . As for other steroids, the circulatory transport and action of estrogens in target tissues are regulated by a complex interplay between sulfation and desulfation pathways. Non‐sulfated estrogens may exert their biological effect by binding to the cognate nuclear receptor or may be downstream converted to more active steroids. Sulfated steroids are highly soluble, and this facilitates not only their renal excretion, but also their circulatory transit fuelling peripheral steroid metabolism. Sulfated steroids, as such, are inert and unable to bind and activate their nuclear receptor; active transportation into the cell and intracellular desulfation are required (Foster and Mueller,  2018 ). The expression and activities of enzymes involved in these pathways is regulated by hormonal factors, and it shows a cyclic pattern in the human endometrium. Sulfation and desulfation pathways dramatically alter the levels of available active steroids and in disease, such as steroid dependent cancer, where the SULT pathway is down‐regulated (decreased enzyme expression and activity), while steroid sulfatase (STS) activity is elevated (reviewed in Mueller et al.,  2015 ). Potent inhibitors of estrogen sulfotransferases have the potential to increase the bioavailability of E2 in target tissues (e.g. uterus), thereby causing an estrogenic effect. HSDs are integral parts of systemic (endocrine) and local (intracrine) mechanisms. In target tissues, they convert inactive steroid hormones into their corresponding active forms and vice versa; their interplay modulates the transactivation of steroid hormone receptors (or other elements of the non‐genomic signal transduction pathways), acting as pre‐receptor molecular switches (Marchais‐Oberwinkler et al.,  2011 ; Salah et al.,  2019 ). In detail, HSD17Bs catalyse the NAD(P)(H)‐dependent oxidoreduction of hydroxyl/keto groups at position C17 of androgens and estrogens; and estrogen receptors transactivate their target genes by binding the 17β‐hydroxylated steroids with much higher affinity than the 17‐oxo steroids. In the uterus, estrone (E1) is converted into the more potent E2 by HSD17B1, and the reverse reaction leading to the production of estrone from E2 is catalysed mainly by 17β‐hydroxysteroid dehydrogenase type 2 (HSD17B2). Derangements in this interplay can contribute to local imbalance in estrogen ‘activation’ and ‘deactivation’ and proliferative pathological conditions in the uterus (e.g. Mori et al.,  2015 ; Hashimoto et al.,  2018 ). High levels of HSD17B1 mRNA and increased E2/E1 ratio are described in uterine proliferative conditions such as endometriosis, endometrial hyperplasia and uterine leiomyoma and inhibitors of this enzyme are investigated as potential therapeutics for these conditions (Marchais‐Oberwinkler et al.,  2011 ; Salah et al.,  2019 ). The role of HSD17B2 in pathologies of estrogen‐sensitive tissues is emerging. Downregulation of HSD17B2 gene was noted in the distal uterus of TBBPA‐treated rats (Sanders et al., 2016). Utsunomiya et al. ( 2001 ) demonstrated HSD17B2 immunoreactivity was increased in the cytoplasm of endometroid adenocarcinoma cell. Hashimoto et al. ( 2018 ) demonstrated in HEC‐1B cells that androgen‐mediated increase in HSD17B2 mRNA levels is associated with significantly reduced E2 induced cell proliferation. Supporting the hypothesis of a role of HSD17B2 to limit E2 availability in EC, the authors demonstrated that a positive HSD17B2 immunoreactive status of endometroid carcinoma cells in patients was inversely associated with the histological grade, proliferation index and clinical stage, and patients tended to have better prognosis than those negative for HSD17B2 immunoreactivity. High levels of HSD17B2 mRNA have been demonstrated in the epithelial cell component of whole endometrial tissue upon exposure to P4, both in vivo and in vitro . Estrogen is synthesised in the gonads and in several extragonadal organs, such as skin, adipose tissues, liver, heart and brain, by aromatase, the enzyme responsible for the conversion of androgens to estrogens E1 and, to a lesser extent, E2 (Simpson,  2003 ; Bulun et al.,  2005 ; Bulun,  2009 ). In premenopausal women, the ovary is the primary source of estrogens (primarily in the granulosa cells and corpus luteum), and the cyclic expression of estrogen by the ovaries drives endometrial proliferation (Mihm et al.,  2011 ). Disease‐free endometrium lacks aromatase and thus does not produce estrogen locally (Bulun,  2009 ; Zhao et al.,  2016 ). In postmenopausal women, peripheral tissues, especially adipose tissue, become the main site of estrogen synthesis (Davis et al.,  2015 ). The estrogen precursor androstenedione is primarily secreted by the adrenal glands (Zhao et al.,  2016 ), and estrogens (E1, E2) are produced in many extragonadal organs (skin, adipose tissues, liver, heart and brain) (Bulun et al.,  2005 ). After menopause, adipocytes, preadipocytes and mesenchymal stem cells within fat tissue are the predominant sources of aromatase. Aromatase levels and activity increase as a function of age and adiposity (Simpson and Mendelson,  1987 ; Bulun and Simpson,  1994 ) and, therefore, contribute to estrogen‐induced endometrial proliferation in the postmenopausal woman (Blakemore and Naftolin,  2016 ; Zhao et al.,  2016 ). Consistent with the role of adipose tissue in estrogen synthesis, obesity is more strongly associated with the development of endometrial cancer than any other cancer type in women (Reeves et al.,  2007 ). This association has been well established and follows a dose–response relationship, with the incidence of endometrial cancer increasing as body mass index (BMI) increases (Onstad et al.,  2016 ). Due to its fundamental role in the synthesis of estrogens, aromatase has been proposed as an important molecular target for many environmental ED chemicals (Laville et al. 2006 ). The local concentration of E2 in endometrial cancer tissues is reported to be higher than in blood or in the endometrium of cancer‐free women (Potischman et al.,  1996 ; Sherman et al.,  1997 ; Berstein et al.,  2002 ). It is well documented that E2 is synthesised in EC in situ , thus contributing to cancer progression. Aromatase protein and mRNA were detected in EC using immunohistochemistry and reverse transcription polymerase chain reaction (RT‐PCR), whereas aromatase expression is low or undetectable in endometrial hyperplasia (a precursor lesion of endometrial cancer) (Bulun and Simpson 1994 ; Watanabe et al.,  1995 ). Increased aromatase in the uterus was previously proposed as putative MoA for ED‐induced uterine carcinogenesis in rodents (Yoshida et al.,  2015 ), but since the normal endometrium in women does not express aromatase, the WG focused on the role of peripheral aromatase as a potential target for ED. The evidence collected around this AOP was overall poor and considered preliminary and no further assessment of the AOP has been conducted by the group. The WG recommended the development of the evidence‐based AOP associated with this MIE. A summary is presented below, full details are available in Annex B . Due to the multiplicity of possible MIEs (Kisspeptin decrease, gamma‐aminobutyric acid‐ergic (GABAergic) modulation, neuropeptides and vasopressin role, etc.), it was decided to develop this AOP starting from the common relevant KE, the reduced availability of GnRH at pituitary level. However, discussion on plausible MIEs is included under Annex B.4 . Ovarian hormones regulate normal human endometrial cell proliferation, regeneration and function and therefore they are implicated in endometrial carcinogenesis directly or via influencing other hormones and metabolic pathways. The role of unopposed estrogen in the pathogenesis of EC has received considerable attention, together with other hormones, such as androgens and GnRH. One of the key homeostatic hormonal loops in this system is provided by the ovarian hormones, E2 and P4, that modulate the activity of the neuronal network controlling the release of GnRH. The hypothalamic GnRH neurons release GnRH in an episodic manner into the pituitary portal circulation to generate distinct pulses of luteinising hormone (LH) and follicle‐stimulating hormone (FSH) throughout the ovarian cycle. Thus, the brain and pituitary produce an on‐going pulsatile pattern of gonadotropin secretion that slows on oestrous to allow appropriate follicular development and a surge pattern of secretion at mid‐cycle to initiate ovulation. Numerous studies have reported that the oestrous‐stage decline in LH pulse frequency results from the post‐ovulatory secretion of P4 (Soules et al.,  1984 ; Smith et al.,  1989 ; Goodman 2015 ) and the administration of P4 was found to dramatically slow GnRH pulse generator activity in the mouse (McQuillan et al.,  2019 ). Thus, it seems very likely that P4 is the key gonadal hormone exerting a negative feedback influence upon the pulse generator during the cycle and does so to bring about the post‐ovulatory slowing of pulsatility. As follicles grow, estrogen synthesis increases in the female ovary. This in turn promotes GnRH pulses in the hypothalamus. GnRH binds to its receptor expressed by pituitary gonadotropic cells and induces the release of 2 gonadotropins, LH and FSH. In turn, LH and FSH stimulate gametogenesis and steroidogenesis in the gonads (Duffy et al.,  2018 ). An LH surge is needed and responsible for the downstream pathways that induce ovulation; this includes resumption of meiosis in the oocyte and cellular changes that allow rupture of the follicle to release the egg for fertilisation. It increases intrafollicular proteolytic enzymes, weakening the wall of the ovary and allowing for passage of the mature follicle (Robker et al.,  2018 ). The suppression of GnRH availability, due to the impairment of regulatory systems or destruction of the peptide, results in a failure of response to pre‐ovulatory level of estrogen to produce LH surges. Without the LH surge, the downstream pathways are not able to function and as a result ovulation does not occur. If the LH surge is delayed, then ovulation may be delayed as well and fails to occur within the correct time window. This can have a negative impact on the reproductive health of females and perturb the oestrous cycle. In most cases, if ovulation is blocked or delayed, the ratio of estradiol/progesterone (E2/P4) remains high due to lack of P4 increase that is initiated after ovulation. As a result, ovarian and circulating steroid hormone levels remain in the ‘pre‐ovulatory’ state, i.e. high E2, and low P4. In addition, with ovulation disruption, formation of corpus lutea is delayed or inhibited. This overall disrupts the cycle and can lead to persistent oestrous. Persistent oestrous is characterised by the lack of corpus lutea formation, and observation of cysts and antral follicles. Morphologically, it is demonstrated by persistent vaginal cornification (PVC). It is considered persistent if at least two cycles were perturbed with the appearance of PVC (Finch,  2014 ; Stewart et al.,  2022 ). A prolonged increased circulating E2/P4 ratio leads to an increase of E2 bioavailability in a variety of estrogenic‐responsive organs, including the uterus due to insufficient counterbalance by P4. However, compensatory mechanisms (e.g. intracrine networks) may differ across different tissues. The degree to which E2/P4 ratio should increase to overwhelm these compensatory responses has not been established. The AOP covering MIE/Kes/KERs related to activation of uterine estrogen receptor‐alpha leading to endometrial adenocarcinoma via epigenetic modulation was developed by the contractor (Viviani et al., 2023 ) and will be published as EFSA external report. In line with the AOP handbook, the goal of this overall assessment is to provide a high‐level synthesis and overview of the certainty in each KER included in the AOP (i.e. probability that a downstream KE would occur when an upstream adjacent KE occurs) based on the available evidence and the knowledge gaps or weaknesses identified in the uncertainty analysis. Here, a summary of the outcome of the KER certainty quantification is given (detailed description in Annex D ). All KERs included in AOPs related to SULT1E1 inhibition, perturbation of HPG axis and related to AOP from uterine (endometrial) estrogen receptor alpha activation to uterine adenocarcinoma were evaluated in terms of biological plausibility, empirical support and KEs essentiality and these elements were used for the quantification of the WoE throughout the uncertainty analysis. It is noted that systematic and harmonised approaches for the quantification of the certainties in KERs should be considered for future development in EFSA (see Section  5 , Future Perspective and Recommendation). Mean and standard deviation of the probability distributions used to express certainty in the 3 criteria for each KER, summarising the EKE performed on each of them, are reported in Table  1 . Details on the results of the EKE and the way they have been combined across experts are provided in Annex D . Mean and standard deviation of the average expert judgement per each AOP, KER and criterion. BP, biological plausibility; EE, empirical evidence; E, essentiality ERα activation led to uterine adenocarcinoma via epigenetic modulation It has been noted that for the KERs included in AOP on SULT1E1 inhibition and perturbation of the HPG axis the biological plausibility is above 0.80. This reflects the well‐established scientific knowledge around events describing those physiological processes in mammals. Regarding the AOP from uterine (endometrial) estrogen receptor alpha activation (MIE) to uterine adenocarcinoma (AO) the biological plausibility of KER1 (i.e. era activation leading to Epigenetic modulation) was found to be below 0.40 reflecting the fact that this biological relationship is still not completely established. This is however expected considering the complexity of epigenetic mechanisms and is reflected also in the difficulties in identifying specific biological markers of potential relevance for uterine adenocarcinoma. For the other KERs in the AOP, the certainty ranges between 0.56 and 0.85. Inclusion of available data on chemical stressors was considered for empirical support of the KERs. – namely dose–response concordance and temporal relationships between and across multiple KEs. For events included in SULT1E1 inhibition pathway, it has been noted that experiments are currently missing where the SULT1E1 inhibition and E2 availability in the uterus (i.e. KER 1) are measured in the same experiment whereas some evidence exists for increased E2 availability in the uterus and ER activation (i.e. KER2). This reflects the lack of proper/standardised analytical methods to estimate the level of E2 in target tissues. However, gene expression is used as a preferential methodology and specific gene activation is considered a suitable marker of estrogenic activation in the uterus; in this case evidence is available. The partial lack of knowledge is reflected by the certainty in the KERs that is 0.14 and 0.59 for KER1 and KER2, respectively. For the events included in the AOP from uterine (endometrial) estrogen receptor alpha activation (MIE) to uterine adenocarcinoma (AO), there is lack of empirical evidence for the downstream relationships (i.e. KER 4, increased proliferation leading to genetic instability; and KER 5, genetic instability, leading to uterine adenocarcinoma). This reflects a gap in the knowledge on how to measure genomic instability in the case of hormone dependent tumours. The certainty in the KERs never achieves levels above 0.48 with the only exception of KER3 (Expression of factors ruling proliferation leading to an increased proliferation (hyperplasia) (certainty 0.81)). The empirical support of the relationships included in the perturbation of the HPG axis pathway, is strong (above 0.80) with only one exception (see KER 3, from delayed ovulation leading to estrogen dominance) for which the certainty is 0.46. The limitation in the empirical support is one of the reasons making difficult to perform a quantitative understanding of the KERs (e.g. response–response relationship) included in the AOP network (see Section  5 , Future perspective, and Recommendation). For SULT1E1 inhibition, uncertainties exist regarding the essentiality of the KEs with certainty quantification achieving 0.50 and 0.74 for KER1 and KER2, respectively. Evidence was in some cases limited to indirect evidence (KER1) while and in others (KER2) the limitation could have been due to the use of not fully appropriate key words in the search. Indeed, it was acknowledged that the ovariectomised (OVX) animal model is likely already representing a model for essentiality (i.e. interruption of the HPG axis would allow the isolation of the uterus, therefore proving that availability of E2 and its perturbation, is essential for the downstream activation of the ER at uterine level) but the model was not included in the search. The evidence on essentiality is strong (above 0.80) for the perturbation of HPG axis, except for KER3, i.e. delayed ovulation leads to estrogen dominance (certainty 0.45). Instead, it is weaker for the downstream part from the ER activation leading to uterine adenocarcinoma (certainty around 0.15 for KER4 and KER5, 0.42 for KER1 and above 0.8 for KER2 and KER3). This is likely depending upon the complexity of the KE considered (related to epigenetic mechanisms) and gap in the knowledge on how to measure genomic instability. The certainty in the three criteria has been combined to derive the final certainty for KERs using probability distribution whose means and standard deviations are provided in Table  2 . KER certainties (probability distribution means and standard deviations) For further details on how certainties in each KER in the putative AOP network were quantified, please refer to Section  2.3.5 and Annex D . The use of AOPs in understanding and describing mechanisms of endocrine disruption as well as supporting risk assessment and regulatory decision‐making is generally accepted. Since the conceptual design of AOPs in the late 1980s and subsequent establishment into a formalised framework, further evolvement of the AOP conceptual framework is still ongoing. More specifically, the ongoing efforts to develop AOP networks to better capture the complexity of biological systems is considered to greatly enhance the applicability of AOPs as such. An AOP network is defined as an assembly of two or more AOPs that share one or more KEs, including specialised KEs, such as MIEs and AOs (Knapen et al.,  2018 ; Svingen et al.,  2021 ). One example of a well‐characterised AOP network is the one for the assessment of thyroid hormone disruption (Knapen et al.,  2020 ). Here, we have described AOPs focusing on uterine adenocarcinoma. These individual AOPs logically converge into a common KE (critical node), namely ‘increased E2 availability’ (Figure  5 ). Graphical representation of the putative AOP network. Dark green = described and evaluated in the context of the current Scientific Opinion; Light green = partially described; grey solid = not described in the current AOP; grey dotted = not described in the current AOP and biological link not explored/not known; blue = critical KEs for the AOP network A few studies have investigated estrogen levels in human uterine tissues levels (Alsbach et al.,  1983 ; Thijssen et al.,  1987 ; Thijssen and Blankenstein,  1989 ). E2 concentrations in the endometrium follow partly the pattern seen in circulating levels, though tissue/plasma ratios change extensively from around 25 in the proliferative phase to less than 10 around and after ovulation. These changes in ratio are the result of larger changes in uterine tissue than in plasma E2 levels. In contrast, the concentrations of E1 in the endometrium and myometrium of premenopausal women show only slight variations during the menstrual cycle. Moreover, tissue/plasma ratios of E1 appear relatively stable at 4‐7 along the menstrual cycle. After menopause, E2 remains the most important estrogen at cellular level in the uterus, considering that E2 levels in uterine tissue are higher than that of E1. Conversely, circulating levels of E1 are higher than E2 in postmenopausal women (Simpson,  2003 ; Rezvanpour and Don‐Wauchope,  2016 ). Factors that may increase uterine E2 tissue levels include local increased conversion of the less potent E1 to E2 by HSD17B1, decreased SULT1E1 activity leading to decreased sulfoconjugation and inactivation of E2, or increased extra‐uterine CYP19 (aromatase) activity leading to increased conversion of androgens into estrogens (described in Annex A : A.2 , A.3 , A.4 , A.5 ). Moreover, a decline in E2 levels and ER expression in the secretory phase is preceded by an LH surge leading to increased progesterone receptor (PR) expression and progesterone levels (described in Annex B ). PR actions are considered anti‐estrogenic, among others through increase of HSD17B2 in the endometrium (Yang et al.,  2001 ), which is involved in the conversion of E2 to E1. A delay in activation of the progesterone signalling will lead to increased uterine availability of E2. E2 can be measured by radioimmunoassay (RIA), enzyme‐linked immunosorbent assay (ELISA) or by analytical chemistry techniques, e.g. using liquid chromatography–tandem mass spectrometry (LC–MS/MS). Typically, steroid hormones are extracted from a matrix using liquid‐liquid extraction (LLE) or solid‐phase extraction (SPE). For tissue levels, endometrial tissues can be homogenised in a buffer or snap‐frozen and pulverised, after which steroid hormones can be extracted. In in vitro studies with endometrial cells, often E2 secretion in the cell culture medium is used as proxy for E2 availability in the cell, but cell homogenates can also be prepared to determine intracellular levels. Both for tissue and cells, cytosolic and nuclear fractions can be separated after homogenisation via ultracentrifugation to determine E2 localisation in cellular compartments, or even further separated into receptor‐bound and free fractions. It should be noted that E2 levels are often poorly defined in experimental studies, and it is not always clear whether total or free estrogen levels are being investigated. Estrogens can be bound to albumin and/or steroid hormone binding globulin (SBHG) or occur in conjugated form as sulfates and glucuronides. It is not always well‐described in studies which form of estrogen is assessed and how sample preparation methods may influence the levels measured. An additional challenge in measuring and interpreting (changes in) hormone levels in females of reproductive age is the natural dynamic in hormone levels due to menstrual/oestrous cycles. Consequently, a single time‐point measurement of E2 may provide little information, especially when additional parameters are lacking such as cycle phase, or levels of other steroid hormones or gonadotropins. Why putative AOP network The relative level of confidence of the overall AOP network presented in Figure  5 has not been assessed as part of this Scientific Opinion. Although the AOP conceptual framework is intended to simplify complex system biology pathways while assuming linearity between the MIE and the AO, it is possible that a MIE may lead to multiple AOs, or that a stressor targets multiple MIEs, which may lead to one or more AOs. A clear example is the stressor atrazine, which is a known aromatase inducer as well as GnRH modulator (see Annex B ). Moreover, what is considered an MIE in one AOP can be a KE in another AOP, as is the case for ‘ER activation’. It is highly likely that more AOPs from the AOP knowledge base (AOP‐KB) may link into the putative AOP network proposed in the current Scientific Opinion and that multiple layers, such as different taxonomies and life stages, or branches linking other MIEs, KEs and/or AOs to this network are conceivable. Specifically, feedback loops, which are well established in, e.g. the HPG axis, are clearly lacking in the currently discussed AOPs. Adding such a layer would greatly enhance the applicability of this putative AOP network (see Section  5 , Future perspective and Recommendations). Regulatory implications This putative AOP network could help to postulate and investigate several other endocrine MoAs, particularly when direct ER binding and activation is not leading the effect (e.g. negative in vitro estrogen binding and transactivation assays, ToxCast ER model and uterotrophic bioassay in rodents). Indeed, the AOPs interacting with estrogen metabolism by triggering inhibition of SULT1E1 or inhibition of HSD17B2 would not be identified in the in vitro ER assays as well as in the ToxCast ER Bioactivity Model but a positive outcome would be likely in the uterotrophic bioassays; moreover, the uterotrophic assay is not requested as in vivo follow‐up study when output from the ToxCast ER Bioactivity Model is available for the substance (EFSA‐ECHA ED guidance). Substances potentially triggering a decrease in GnRH availability will likely not be identified in studies addressing estrogenic activity including the uterotrophic bioassay when performed in ovariectomised adult females, lacking an intact HPG axis. Exploring the different MIEs postulated in this AOP network seems therefore a useful tool to address the potential underpinning MoA of increased incidence of uterine adenocarcinoma as can be observed in carcinogenicity studies (see Section  5 , Future Perspective and Recommendation). Sexual steroid hormones and gonadotropins levels are not routinely investigated in OECD TG dedicated to repeated dose toxicity and reproduction. Indeed, due to physiological changes of hormonal levels during oestrous cycle and the limited standard number of animals per group, the average number of animals in each stage of the cycle is generally too low to permit any conclusions (Stanislaus et al.,  2012 ). Implementation of an in vitro test battery exploring several MIEs could allow a bespoken follow up testing strategy specifically designed and statistically powered (e.g. with appropriate animal numbers and duration of exposure, synchronised oestrus cycle, etc.) in order to generate reliable mechanistic data. As mentioned above, hormones levels are not routinely performed in regulatory studies. However, oestrous cycle monitoring could allow to identify indirectly increased E2 bioavailability in the uterus, tonic levels of E2 being reflected in another E2 sensitive tissues (i.e. vagina). Furthermore, it allows to measure the integrity of the HPG axis. Monitoring of oestrous cyclicity is included in OECD GD studies dedicated to reproductive toxicity but is not required in chronic studies. Inclusion of this parameter in chronic toxicity studies could help to pick‐up potential shift of cyclicity according to time (see Section  5 , Future perspective, and Recommendations). Above considerations apply not only to pesticide substances but also to any other chemicals for which the assessment of the ED properties is requested. Eventually, the proposed AOP network could serve as a basis to implement AOP‐informed IATA for data‐poor substances or to answer specific regulatory problem formulations.

This Scientific Opinion aimed at developing AOPs to support the identification of substances having endocrine‐disrupting properties leading to uterine adenocarcinoma. The PPR Panel considered that for the problem formulation, as expressed in the ToRs, AOPs can be used to indicate the association between a specified pathway and a human adverse health outcome, which can be triggered by chemicals, including pesticides. Several AOPs were developed, converging into common critical nodes, i.e. the increased E2 availability in the uterus (endometrium) followed by ER activation. The results obtained allowed the formation of an AOP network, which may be further expanded through connections with already existing KEs and/or AOs for female reproduction. AOPs are increasingly recognised as important tools in the regulatory field. In this regard, the AOPs proposed in this Scientific Opinion identified gaps and new endpoints for which further development is prudent (see Section  5 Future perspective and Recommendations). Moreover, this exercise also allowed to test a new strategy in the categorisation of information through the implementation of a machine learning – topic modelling approach to the systematic review, that allowed clustering papers according to the semantic similarity in an automated way, saving time and resources.

In the EU, specific provisions aiming at phasing out endocrine disruptors (EDs) are present in different regulatory frameworks, including chemicals (REACH) (Regulation (EC) No 1907/2006), plant protection products (PPP) and biocidal products frames. For both pesticides and biocides, once it is proven that a substance is an endocrine disruptor, in principle it cannot be authorized for use (COM/2018/734 final). 1 The scientific criteria for the determination of endocrine‐disrupting properties of a substance (ED criteria) are based on the WHO‐IPCS definition of an ED (WHO‐IPCS,  2002 ) and are set by the Commission Regulation (EU) 2018/605 (amending Annex II to Regulation (EC) 1107/2009) and by the Commission Delegated Regulation (EU) 2017/2100 for Plant Protection Products (PPPs) and Biocidal Products, respectively. 2 A guidance document for the implementation of ED criteria for substances pursuant to the PPP Regulation (EC) No 1107/2009 and the Biocidal Products Regulation (EU) No 528/2012 was developed by the European Food Safety Authority (EFSA) and the European Chemicals Agency (ECHA) with the technical support of the Joint research Center (JRC) in 2018 (EFSA‐ECHA et al.,  2018 , thereafter indicated as EFSA‐ECHA ED guidance). The EFSA‐ECHA ED guidance presents the steps necessary to identify a substance as an ED, in line with the ED criteria laid down in the above regulations. Although the ED criteria are intended to cover all endocrine disrupting modes of action (MoAs), the EFSA‐ECHA ED guidance mainly addresses estrogenic, androgenic, thyroidal, and steroidogenic (EATS) modalities. This is because there is a relatively good mechanistic understanding and testing of the EATS modalities and of the pathways leading to adverse effects linked to perturbation of the endocrine activity. Indeed, at present, standardized test guidelines (TGs) for in vivo and in vitro testing are available for the EATS modalities only, coupled with scientific agreement on the interpretation of the effects observed. These are compiled in the OECD Guidance Document on Standardized Test Guidelines for Evaluating Chemicals for Endocrine Disruption (OECD GD 150, OECD, 2018a), which includes the ‘OECD Conceptual Framework (OECD CF) for Testing and Assessment of Endocrine Disrupters’. Nevertheless, the general principles outlined in the EFSA‐ECHA ED guidance are also applicable to other endocrine (non‐EATS) modalities. The EFSA‐ECHA ED guidance has been in force since November 2018 and is having an important impact on the risk assessment procedure for the approval of pesticide active substances. It is recognized that the process to establish a biologically plausible link between an adverse effect and the endocrine activity implies an extensive weight of evidence (WoE) and MoA analysis from the available dataset of studies. In this context, the PPR Panel considered that the Adverse Outcome Pathway (AOP) framework is an optimal tool to streamline the implementation of the EFSA‐ECHA ED guidance and to strengthen the regulatory ED assessment. The development of dedicated AOPs should be considered for both EATS mediated and other endocrine mediated adverse outcomes (non‐EATS) and could be carried out using a top‐down approach i.e., from an adverse outcome (AO) backwards. Valuable tool e.g., as described in OECD GD 150, are available to recognize what is considered EATS mediated adverse outcome for which, the biological and toxicological knowledge is considered sufficient to define the observed adversity as likely endocrine mediated. However, there are limitations in the approach taken in the OECD GD 150, and development of AOPs using a top‐down approach is expected to facilitate the overall WoE analysis to conclude on adversities linked to perturbation of the endocrine activity. For this Scientific Opinion, the PPR Panel proposed to explore this approach by postulating AOP(s) for adverse effects (apical outcomes) on the uterus as evidence collected in the studies used for the risk assessment of PPP active substances, followed by the identification of the intermediate key events (KEs), including perturbation of the endocrine activity, and of the molecular initiating event (MIE). The outcome of this scientific opinion could serve other areas where ED properties are assessed on a case‐by‐case basis as part of the risk assessment. The PPR Panel self‐tasked to: Develop, in line with the guidance for the AOP development, four Adverse Outcome Pathways (AOPs) for which the selected adverse outcome is plausibly linked to an endocrine activity, considering the uterus as a target organ. Collect the developed AOPs in one scientific opinion. Develop, in line with the guidance for the AOP development, four Adverse Outcome Pathways (AOPs) for which the selected adverse outcome is plausibly linked to an endocrine activity, considering the uterus as a target organ. Collect the developed AOPs in one scientific opinion. Available information in the scientific literature will be collected using a systematic review, which will be used for determining the exact AOPs to be developed. Regarding ToR1, the uterus was identified by the PPR Panel as relevant for AOPs development since it is a non‐endocrine organ, but it is listed in the OECD GD 150 as a target of EAS‐mediated activity and routinely evaluated in pesticide risk assessment. Therefore, dedicated AOPs would represent a tool for risk assessors in the analysis and postulation of MoAs where the endocrine activity is an intermediate KE leading to uterine adverse outcome/s. Uterine neoplasms can be observed in standard regulatory studies for pesticides in rodent models; among these, uterine adenocarcinoma is identified as the AO of relevance in this Scientific Opinion, considering its potential relevance to humans and its possible relation to ED properties of the substance. Uterine endometrial carcinomas (EC) represent the most common uterine cancers in women, and Type I EC (or endometrioid adenocarcinoma) makes up the majority of endometrial cancer cases (~ 85%). Clinical and epidemiological observations suggest that these neoplasms are hormone‐driven and many of the identified risk factors involve excess estrogens, or estrogen receptor‐mediated signalling unopposed by progesterone signalling (Rodriguez et al.,  2019 ). In rodent models, ‘uterine adenocarcinoma’ (also described as uterine endometrial carcinoma; or endometrial adenocarcinoma, Dixon et al.,  2014 ) shares many similarities with the lesion observed in women. In regulatory studies, most of the time, it is the only information available, making a retrospective evaluation of the MoA complex. Further information on the human relevance of this AO is provided in Section  3 (Assessment) and in Appendix  A . (Comparative pathology of endometrial carcinoma and reproductive senescence). The intrinsic complexity in developing AOPs associated with the uterus as a target organ made the PPR Panel consider developing four AOPs, as a preliminary estimate. It was further considered that uterine adenocarcinoma can originate from various MIEs, thereby further increasing the complexity and limiting the possibility to identify at priori the (number of) AOPs to be developed. Therefore, it was acknowledged that the exact number of AOPs was to be defined after the completion of the exercise requested in this scientific opinion. Regarding ToR2, for efficiency purposes, it has been chosen to contract out part of the activity (e.g. protocol development, systematic review and development of a part of the AOP) (Viviani et al.,  2023 ). Consequently, the developed AOPs are collected in this Scientific Opinion and in an EFSA external report. To address the ToRs, a preliminary problem formulation was carried out, tailored to the identified AO, and articulated in the following questions and sub‐questions: Q1. What is(are) the AOP(s) relevant for the identification of substances having endocrine‐disrupting properties and leading to uterine adenocarcinoma as an AO based on the available evidence assessed in light of biological plausibility, essentiality and empirical evidence? Q2. What is the overall certainty in the AOPs relevant for the identification of substances having endocrine‐disrupting properties and leading to uterine adenocarcinoma as an AO? – What is the level of certainty in the relationship between the MIE (upstream KEs) and the downstream KE(s)? – What is the level of certainty in the relationship between the upstream KE(s) and the AO(s)? What is the level of certainty in the relationship between the MIE (upstream KEs) and the downstream KE(s)? What is the level of certainty in the relationship between the upstream KE(s) and the AO(s)?

Methodological Approach Click here for additional data file. SULT1E1_INHIBITION description and quantification Click here for additional data file. SULT1E1_INHIBITION_Dose and Temporal concordance Click here for additional data file. HSD17B2_INHIBITION description Click here for additional data file. AROMATASE INDUCTION_description Click here for additional data file. Methodological Approach Click here for additional data file. Reduced availability of GnRH description and quantification Click here for additional data file. Dose and Temporal concordance Click here for additional data file. Plausible MIEs Click here for additional data file. AOP related to activation of uterine estrogen receptor‐alpha leading to endometrial adenocarcinoma Click here for additional data file. Results_KER_certainty Click here for additional data file.