EU’s next generation risk assessment: hurdles and opportunities for new approach methodologies

The EU’s decision to ban animal testing for toxicity testing, has positively influenced the pace of developing New Approach Methodologies (NAMs). This development also supports replacing animal methods in other forms of risk assessment (RA), such as for oral-toxicity testing. This study aims to identify the hurdles and opportunities for validation and implementation of NAMs in the current EU’s chemical RA. Through conducting semi-structured interviews with 14 stakeholders, experiences and perspectives about the validation and implementation of NAMs in RA for orally ingested chemicals were analyzed. Stakeholders considered the use of NAMs for RA processes both a cultural and generational issue. Both were perceived as hurdles for reaching the next generation RA approach. The differing views on NAMs originated from experience and stakeholder positions, but communication and collaboration on developing future RA approaches could support overcoming this skepticism. Irrespectively of their background, all interviewees were generally optimistic that NAMs will support the development of more accurate and sustainable RA. This research highlights the need for the EU to adjust legislation and guidance documents to shift in testing requirements from the traditional overexposure approach to more predictive, mechanistic testing in RA, which will take time. This study, however, shows that—when all stakeholders engage in communication and confidence building—NAMs can already play an important role in reducing and refining animal testing.


Introduction
Since the establishment of RA procedures for chemicals in the 1980s, the safety of food, medicinal, and chemical products as well as cosmetics has to be established before trading on the global market (IPCS 2010). Since then, different legislative procedures have been put in place in the European Union (EU), that require business operators to request pre-market authorization to allow products to be placed on the market (Knight et al. 2021;Vrolijk et al. 2020). RA is key in these pre-market authorization procedures. Animal testing has long served as the gold standard within the RA procedure for all types of chemical products; whilst at the same time, since 1993 efforts are being made to ban animal testing in cosmetic products (Akbarsha et al. 2019). This ban was supposed to come into effect in 1998 to allow for sufficient time to develop animal-free methods, but a lack of animal-free methods led to a delay until finally, the entry into force of the Cosmetics Regulation in 2009 (Regulation (EU) 1223/2009) (Akbarsha et al. 2019).
However, nowadays with the increased call to reduce, replace and refine, in all chemical sectors that use experimental animals for toxicity testing, major developments in NAMs are made at a higher pace (Fischer et al. 2020). This allows for a swifter transition from using animal models to animal-free methods in establishing safety of regulated products, e.g., for oral toxicity testing of pesticides [as regulated under Regulation [EU] 283/2013] or chemicals [Regulation (EC) 440/2008]. These and other regulatory acts and their associated guidelines define dossier requirements, under which registering business representatives should substantiate the safety of their products (de Boer et al. , 2022Vrolijk et al. 2020). The dossiers should include i.e. safety 1 3 and kinetic information about the substance, generated from in silico, in vitro, or in vivo experimental animal studies (Vrolijk et al. 2020). The applicability and acceptance of NAMs as an alternative for e.g., 90-day toxicity studies is only described to a limited extent in EU legislation and guidance documents (Vrolijk et al. 2020). The limited application of NAMs for RA procedures can be explained by the endpoints of oral toxicity testing covering complex interactions and effects, which are difficult to predict by NAMs (Benfenati et al. 2016;ECHA 2017). Furthermore, research shows that stakeholders are uncertain about the reliable use of NAMs for RA EFSA et al. 2016;Schiffelers et al. 2012;Tarazona 2020;Zuang et al. 2021). Consequently, different factors are believed to negatively influence the validation and implementation of NAMs in EU's RA.

EU practices to validate and implement NAMs
In the EU, 3 scientific agencies are involved in evaluating the safety of new substances based on the dossiers submitted by registrants. Depending on the intended use of such substances, dossiers are submitted to the European Chemical Agency (ECHA), who deals with chemicals, the European Food Safety Authority (EFSA) who assesses foods, and the European Medicines Agency (EMA) who addresses medicinal products (European Parliament and Council of the European Union 2002Union , 2004. These agencies are responsible for both assessing the dossiers submitted to the European Commission on their scientific remit, as well as communicating the dossiers' requirements. The agencies' assessments are subsequently used by the Commission to decide upon market authorization of the product (European Parliament and Council of the European Union 2002).
Regarding the cross-cutting topic of animals used for experimental procedures, Directive 2010/63/EU has been implemented in the EU that applies to all regulatory acts and guidance documents that touch upon the use of animals for scientific testing. The objective of this Directive is to protect animals used for experimental procedures, by i.e., fostering the replacement and reduction of the use of experimental animals, and refining breeding, accommodation, care, and use of animals [Article 1(1[a])]. Together, these efforts for replacement, reduction and refinement of experimental animals are known as the principles of the 3Rs. More recently, the Commission has re-emphasized their aim to stimulating research, innovation, and promotion of animalfree methods such as NAMs in the 2020 EU Green Deal (EC 2020a, b). Directive 2010/63/EU appoints the Joint Research Centre (JRC) as the responsible institution for coordinating and promoting the 3Rs principles, but many different organizations are involved in activities, research projects and initiatives to reduce, refine and replace animal testing for RA. ECHA, EFSA, and EMA for example support the implementation of the 3Rs principles in chemicals, food and feed, as well as veterinary and medicinal products (EMA 2016). The EU's scientific agencies collaborate with experts on various NAM-related projects such as EU-ToxRisk or Toxtree ).

NAM implementation and validation
Whilst many stakeholders in the chemical, food and medicinal product sectors have endorsed the need to move away from animal testing for RA, currently one on one replacement methods are not possible, as highlighted in ECHA's perspective about replacing oral toxicity studies (ECHA 2017). Specifically for method replacements, in vitro and in silico test methods have been designed that fall under the umbrella of NAMs. Similarly, NAMs can support reduction and refinement of toxicity testing, both via the development of test methods and by using approaches that facilitate data integration into the regulatory structure, such as Adverse Outcome Pathways (AOPs), physiological based pharmacokinetic modeling (PBPK), read-across and Integrated Approaches to Testing and Assessment (IATA) (Kass 2019;Tarazona 2020). Therefore, NAMs include both newly developed test methods as well as data integration methods.
For regulators, adopting modified or new approaches for RA is hampered due to scientific obstacles with NAMs, e.g., the prediction of endpoints for oral toxicity testing (Benfenati et al. 2016;Schiffelers et al. 2012;Zuang et al. 2021). Using new approaches requires detailed insights into what data the methods provide, and how these data support the safety profile of a product. Regulators, for instance, need to rely on expert judgements, which are used in developing IATA as a science-based approach for hazard identification that integrates existing data with targeted data generation via combinations of methods, including NAMs (Benfenati et al. 2016;OECD 2021;Zuang et al. 2021). Similarly, AOPs offer a conceptual framework using existing data to link a certain effect of a compound, the molecular initiating event, and a potential adverse outcome, which serve i.e. as a framework to develop IATA (OECD 2021). The reliability of NAMs increases predominantly with the amount of used data, integrated or generated. However, regulatory bodies need tools that collect, curate, store and analyze these large amounts of data (Berggren et al. 2015). Whereas IATA and AOPs may seem as promising approaches and methods to reduce animal testing for RA, these examples demonstrate the complexity of the integration and validation of NAMs in RA.
The limited acceptance of NAMs is linked to the significant doubts that governmental and non-governmental stakeholders have about the reliable use of NAMs in the RA (de Boer et al. 2020EFSA 2016;Tarazona 2020;Zuang et al. 2021). Their uncertainty can partially be explained by the risk aversive attitude of society, which negatively influences the acceptance of risks as well as the acceptance of using alternative methods for risk analysis (Schiffelers et al. 2012). Whilst many developments in NAMs have been published in literature, very little insight is available into the applicability of NAMs for RA. Whereas animal testing allows to study adverse effects that are triggered after exposure, NAMs provide insights into the action mechanism of a compound, with a subsequent proof whether this causes an adverse event. Therefore, applying NAMs requires a transformation in toxicity testing by understanding key biological pathways that cause adverse outcomes (Blaauboer et al. 2016;National Research Council et al. 2007).
As described above, uncertainties regarding validity and implementation have strongly impacted the ban of animal testing for cosmetics in the 1990s and 2000s (Akbarsha et al. 2019). In 2013, the ban was imposed in Regulation (EC) 1223/2009 on using experimental animals. However, in the authorization of cosmetics this resulted in the development of various new approaches to analyse toxicity of cosmetics (Desprez et al. 2019;Desprez et al. 2018). And they have been shown to get accepted for establishing various safety parameters (Almeida et al. 2017). However, even though new methods were developed e.g., for skin sensitization, this has not immediately resulted in changes of EUs RA practices for orally ingested chemicals.
To ensure that NAMs development and validation can lead to swift incorporation in EU legislation dealing with the safety of orally ingested chemical products, the development of strategies and measures needs to be more targeted. Today, even with stakeholders collaborating in ongoing projects and activities, many stakeholders actively contribute to the debate of validating and implementing NAMs in the RA system for orally ingested chemical products by publishing academic papers on specific methods or procedures from their isolated perspective. So far, Schiffelers et al. (2012) investigated the rationale behind the regulatory acceptance of the 3Rs principles, whereas this is the first qualitative study aimed to generate insights into the stakeholders perspective of the implementation of NAMs in the full RA analysis cycle of the EU. To facilitate the development of more targeted strategies and measures and their validation and implementation in the RA of the EU welcomed by all stakeholders, this paper explores stakeholders' perception of NAMs and their possible implementation in the RA of chemical products in the EU.

Material and methods
To study the perspective of stakeholders who either influence, have potential interest in, or are affected by implementing NAMs in the RA of the EU, a qualitative research approach was chosen. Qualitative studies offer the opportunity for detailed in-depth understanding of processes and views on particular topics (Döring et al. 2016). Therefore, semi-structured, in-depth interviews with 14 stakeholders were conducted.
A stakeholder of toxicity studies in the EU's regulatory system dealing with safety assessments (in this study termed 'practices') is defined as a group or person responsible for the implementation and validation of animal-free methods for RA, and/or having an interest in its success (Table 1) (Freeman 1984). Primarily, this entails risk assessors and risk managers. Secondarily, organizations and individuals who are directly affected by RA requirements are involved in RA, such as those involved in developing and conducting tests for the safety of chemical products: academics, business representatives, and civil society representatives. As displayed in Table 1, the described stakeholder groups were further divided into stakeholder sub-groups, such as industrial or university research for academics. When their responsibilities and operations overlap, stakeholders fall into multiple sub-categories. These stakeholders were identified through a mapping exercise to identify all ongoing activities in the EU that relate to NAM development and implementation, in which EU agencies or EU institutions are involved (available upon request).
All stakeholders were invited for participation via e-mail, using purposive sampling from the mapping exercise, to receive insights from stakeholders of each category (Flick 2014). Additionally, snowball sampling was used to acquire further insightful information from newly identified interviewees. After 14 interviews, a dataset with rich, in-depth, and complex data was collected. No new findings emerged from the later interviews, suggesting that data saturation was reached (Morse 1995). However, due to the study approach, data saturation was not aimed to be achieved: information power of the data set was deemed most relevant. As described by Malterud et al. (2016), the following 5 items influenced the determination of the sample size for this study: 1. relatively narrow study aim 2. sample with highly specific experience and knowledge that was sampled purposively 3. existing theory could be used in the analysis set-up of the work 4. strong dialogues were held with participants 5. data was analyzed as one case.
This resulted in an appropriate sample size of n = 14 for the aims and purpose of this study.

Researcher reflexivity
To support unbiased data generation and analysis by the researchers and decrease the risk of researcher bias, reflexivity is key. Such reflexivity can support keeping the interview free from any personal preconceptions of the world. Researcher reflexivity was a continuous process throughout the study period that started with the study design and required a continues self-monitoring on the decisions made and actions taken by the first author (SO), with the support of the principal researchers involved in the project (authors MV and AdB) (Dodgson 2019).
This reflection was conducted through bracketing, a method to identify the most important perspectives and preconceptions of the authors involved in the study (Dodgson 2019). This allowed for critically reviewing the personal perspectives on the importance of food safety (and legislation dealing with this), animal testing and animal welfare. With their background in nutritional sciences and professional experience in the food industry, the first author specifically believes that food safety is key in increasing the quality of EU citizens' lives. Animal testing for food safety is considered a convenient whilst expensive method Table 1 Stakeholder groups of toxicity studies in the RA of the EU (EC 2017(EC , 2020a(EC , b, 2021aEFSA 2016EFSA , 2019 *n = 14 interviewed stakeholders. In total, 38 stakeholders were approached: 14 did not respond to the invitation nor reminder e-mail; 10 potential interviewees declined the invitation or referred to other stakeholders to analyze food safety hazards, even though the perspective of the first author is that animal welfare is insufficiently addressed in EU legislation dealing with food safety and chemical safety in general.

Data collection
Data were collected via semi-structured interviews held in English via Zoom. All interviews lasted 30-60 min and were recorded and transcribed verbatim. An interview guide which consisted of themes to be addressed in the interviews was developed by the research team. Through literature research, 3 topics were identified as relevant for answering the research question: I. "EU's RA" to receive information on stakeholders' experience in the process of RA for orally ingested chemical regulated products; II. "Animal-free approaches and NAMs" to discuss the extent of interviewees to which they were familiar with NAMs, how they were informed about the use of NAMs in RA procedures and how legislation and relevant stakeholders dealing with NAMs influenced their work; III. and "personal perspective" to receive stakeholders' personal attitude towards and needs in the context of NAMs in the RA. Two pilot interviews supported the development of a coherent interview guide.

Data analysis
All interviews were audio-recorded for verbatim transcription. Even though pre-identified themes had informed the interview topics, an open coding process was used to analyze the transcriptions. Transcription followed the methods described by Flick (2014) and Drehsing and Pehl (2015), ensuring that transcriptions were coherent whilst offering detailed insights into the content and context of the interview.
To guarantee privacy and anonymity of stakeholders, names and identifiers were replaced by overarching terms, such as "risk assessor" instead of stakeholders' names. All interviewees factually reviewed their transcripts before coding started.
Transcripts were coded using the qualitative data analysis software ATLAS.ti (version 8.4.5). The content-coding started with the first interview and continued simultaneously with the following ones. All interviews were screened for relevant information. After that, the pre-coding phase allowed for screening the interviews for extreme statements, theoretical interests, empirical quantity, and the final coding of assigned codes to the relevant information (Mayring 2015). The codes were generated in vivo by one author, using statements and formulations made by the interviewee to reduce the risk of researcher bias. To reduce the risk of interpretation bias, coding was reviewed by the two other authors. A code book (available upon request) was developed to allow for grouping codes into broader categories and finally themes, which was used as the basis for the findings presented in the results section (Flick 2014).

Ethical approval
The study protocol (ERCIC_250_01_04) was reviewed and approved by the Ethical Review Committee Inner City Faculties of Maastricht University. All interviewed stakeholders gave their oral and written informed consent.

Results and discussion
Desk research, performed prior to conducting stakeholder interviews, showed an early stage of NAM implementation into RA, and RA stakeholders seemed uncertain about NAM use for RA purposes (de Boer et al. , 2022EFSA et al. 2016;Schiffelers et al. 2012;Tarazona 2020;Zuang et al. 2021). However, through conducting semi-structured interviews with 14 stakeholders, the results gained deeper insights into the perceived hurdles and opportunities that stakeholders face when implementing NAMs for regulatory assessments. Stakeholders' opinions on the development and implementation of potential of NAMs were concentrated on 5 interconnected themes that were identified following in vivo coding of the interview transcripts: (1) Risk assessment approach, (2) Prerequisites for NAMs implementation, (3) Stakeholder's tasks and perceptions (4) Communication, (5) Collaboration.
The interview findings were presented and discussed per theme.

Risk assessment procedure
Academics, risk assessors and business representatives label the current RA approach a "paradigm […] developed in 60 s, 70 s" (risk assessor). RA is considered to focus on finding adversities in overexposure experiments rather than identifying hazardous substances: "Yes, if you fill up a rat with some chemical, and it does not die immediately, but it gets cancer in a couple of months or years. Does that mean anything for us? So, it is a dose response question. You go very high up [dosages] in an animal and it causes cancer. Oh, dear me. We are exposed to this chemical as well. Oh, dear me, do we all get cancer? And the answer is no, we do not" (academic).
According to interviewees, focusing on identifying adversities has led to the formation of 'box-ticking exercises' for RA in the EU in specific legislative procedures, such as when seeking authorization for pesticides. Multiple interviewees believe that this observational overexposure approach needs to change to reflect real-word conditions, which additionally creates opportunities for the implementation of NAMs: Rather than searching for the minimum exposure leading to adversity, real-world consumption and exposure data should be used to investigate their effect on human bodies. Interviewees describe this as a fundamental change for RA approaches, since exposure on consumption level induces different pathways in the body than consuming the substance on a toxic level. This was expected leading to relevant different endpoints in RA testing.
The critical attitude of various interviewees towards the current RA approach coincides with a long-lasting perspective of science, policy, and public: 30 years ago Ames and Gold (1990) described that the observational overexposure approach is not suitable for carcinogenicity assessments. Scientific literature has since highlighted that the current approach is debatable, because of its questionable relevance for ensuring human safety (Bhattacharya et al. 2011;Bridges 2003). Already in 2007, the US National Research Centre developed recommendations to use a new toxicity-testing system evaluating on key toxicity pathways, which steers away from using the observational overexposure approach (Bhattacharya, et al. 2011;National Research Council et al. 2007). At the same time, for more than 20 years NGOs have been working on campaigns against animal testing (Cruelty Free International 2021). The awareness about the need for change does not seem to be a vital factor for implementing a different RA approach: this awareness seems already embedded in both the scientific and the regulatory science community.

Prerequisites for NAMs implementation
The interviewed stakeholders did not expect that NAMs will be integrated into RA as one-on-one replacements for animal studies. Rather, a more holistic approach to RA was suggested to be the best way to benefit from NAMs. In such a holistic approach, in vitro assays form different "testing batteries, […] which are targeting different aspects of an endpoint. So, basically like an AOP" (risk manager).
To build a holistic and approximated predictive approach, different interviewees referred to the potential of NAM such as AOPs for understanding the effects of substances. One interviewed academic describes it as "(…) a map of all of the systemic toxicity, which allows us to ensure that we have AOPs that essentially cover this space, this biology that we are interested in". Insights into systemic toxicity facilitate the more detailed understanding of complex biology. Currently, NAMs can only reflect pieces of these mechanisms and, consequently, are considered primarily suitable for establishing short-term effects, according to interviewees.
For the investigation of long-term effects, animal studies are the preferred choice as they are living organisms holding all biological mechanisms, even though it is clear there are limitations to these tests: "It is not 100% sure to what extent animal studies reflect the biology in humans" (risk assessor). Fundamental research is essential to understand the complex biology of humans and animal studies can facilitate the understanding of specific biological factors. Academics and risk assessors describe that new fundamental research findings, such as specific pathways of adverse events being identified, are crucial for NAMs development. Subsequently, applied science is needed to verify whether novel findings would promote method development for RA. Stakeholders stated that science is moving fast towards the development of reliable NAMs, and they are confident that the necessary level of understanding NAMs mechanisms can be reached.
Until that point of having more reliable NAMs, the mainstream toxicological approach used in RA offers limited opportunities for using and implementing NAMs. A few methods and approaches are already partially recognized and implemented in RA, such as read-across. A risk assessor declares NAMs could be used for the first-layer screening of substances (Tier 1) and, by this, already reduce the need for animal studies for follow-up investigations by identifying chemicals with potential concerns. By correlating insights generated through NAMs in Tier 1 with data from subsequent higher tier in vivo studies, NAMs understanding and calibration will be supported, which can foster RA solely based on NAMs. Only then, toxicological testing strategies could facilitate the replacement of legally required in vivo animal testing by NAMs. Stakeholders described to built these strategies under the assumption that NAMs offer a better methodology than traditional methods for RA: "There is a strong argument, that the new approach methodologies will provide a superior approach in the long run, because it is based much more on human biology and not animal biology" (academic). Interviewed stakeholders believe that NAMs can replace animal studies if they address limitations that are intrinsically linked to animal testing strategies. Their results must be reliable, NAMs should have a predictive capacity, should facilitate the understanding of how chemicals react, and have to "recapitulate […] advanced knowledge on human biology, not recapitulating again, the advantages and limitations of animal methods" (risk assessor).
Interviewees attribute the increasing complexity of RA to an increasing number of stakeholders in RA and risk management, as well as the adoption of the EU Green Deal. The latter expands the information requirements for chemicals and can thereby push for increased numbers of animal studies to provide this. Risk managers state that the increased complexity is slowing down the legislative process. However, the expanding information requirements also offer an opportunity for NAMs to be implemented: "It means that we have to do more tests to satisfy those requirements. And you cannot do those with animal testing, it is just too expensive" (risk manager).
Previous research already identified the need to establish the scientific and regulatory validity of NAMs as one of the key issues, before these methods can be used to significantly reduce the need for animal studies (Knight et al. 2021;Mondou et al. 2021). Research projects into NAMs at the same time have been described as mostly shorterterm endeavours, which cannot sufficiently support largerscale scientific developments and regulatory validation of methods (Deluyker et al. 2016;Knight et al. 2021). Because when methods are considered relevant and reliable, their subsequent validation is often still a lengthy and complex process: test method validation requirements such as those established at OECD level may not always be well aligned with NAMs development and implementation (Mondou et al. 2021).
Whereas some scholars active in RA have been shown to believe that NAMs will mainly play a role in addressing gaps in testing strategies using in vivo experiments (Avila et al. 2020), others believe that, in the future, NAMs will completely replace animal testing through for example integrating different NAMs for RA purposes (Blaauboer et al. 2016;Graepel et al. 2019). The role of NAMs will highly depend on the output required for safety assessments: will regulatory testing remain focused on identifying apical endpoints of toxicity (as currently studied in animal models), or will the transition towards exposure driven approaches result in requiring more mechanistic insights into toxicity in humans, as studied with NAMs (Blaauboer et al. 2016;de Boer et al. 2020de Boer et al. , 2022Graepel et al. 2019). In any case, stimulating the uptake of currently validated NAMs in the RA process will be essential for all involved stakeholders to get experience with NAMs and the data that they produce (Schiffelers et al. 2012). As further described in the sections below, such confidence building is key for the acceptance of NAMs in RA for regulated products (Basketter 2020; Mahony 2019).

Stakeholders' tasks and perceptions
All stakeholders have their specific role in the process of getting regulated products authorized for placement on the EU market. As displayed in Table 2, the interviews provided insights into the official roles of these stakeholder groups, as well as what determined their behaviour and actions, next to the needs that these stakeholders have. One academic stated that the EU's risk managers are known to be extremely busy because of their multistakeholder tasks. For instance, they have to keep contact with all stakeholders (including the public) and to ensure that everyone's need is represented in decisions, whilst at the same time, prioritizing these needs: "[It is a] trade-off between offering better protection or the perception of better protection, and increasing animal testing, increasing the burden on industry and economic operators" (risk manager).
Because risk assessors inform risk managers about the scientific evidence supporting the safety of innovative products and procedures, the tasks and perceptions of both stakeholders are highly interconnected. However, risk assessor's focus more on the technical aspects of complying with legislation currently in force. An interviewed risk assessor stated that one of the key needs for RA is raw data, which published research papers often lack. Risk assessors need these raw data to ensure reliability and reproducibility of data: "60% of those papers are not reproducible even by the researchers themselves" (risk assessor). An interviewed risk assessor describes that risk assessors try to minimize irrelevant studies on both the regulatory and the industrial side. However, a business representative described the EFSA as "very unwilling to get involved in any activities which involve industry. They get very concerned that if they seem to be working with an industry that can be accused of getting too close to industry" (business representative).
Business operators strive for international harmonization of RA, while risk managers claim that a 100% international harmonization is impossible because of the large diversity in policies, legislation, and related requirements. In general, business representatives do not feel sufficiently represented in the EU's approach to RA. According to an interviewed academic, this has resulted in the set-up of additional industrial research programs that were undertaken simultaneously to equivalent EU programs. Risk managers and risk assessors stated that companies only participate in actions and projects which can save them money. At the same time, other interviewees perceived business representatives as " […] very open to the development of alternatives" (NGO representative). One specific driver for the development of animal-free RA methods mentioned by a business representative, is the increased consumer demand for vegan foods. This drives industrial innovations of new components and ingredients, for which safety cannot be assessed through animal testing because of the vegan positioning. This increases the difficulty for businesses, specifically for small and medium sized enterprises, to gather all data needed for RA in and outside the EU.
In contrast to the industry, "it is not [academics] primary work to meet regulatory standards" (risk assessor). They are seen to concentrate predominantly on high-quality science, which is basically not the applied and swift science and data that the industry needs. Academics build close communities and exchange information within their community through i.e. conferences and publications. They are considered independent and can flexibly refocus their research. This enables science to develop at a higher speed than legislation.
Recurrently, interviewees from all stakeholder groups mention perspectives which influence the actions of other stakeholders. They come to the same conclusion that the comfortability of doing things as they were done before "protects you from blunders and mistakes" (academic). Relying on traditional methods such as in vivo studies is more comfortable than trying to understand the underlying logic of innovations and NAMs: "So, we must ask ourselves the question whether the reassurance we get from animals, from the in vivo studies, is a real reassurance, or [whether it] is a false reassurance. We should avoid by any means anything that gives us only a false sense of reassurance, in the sense that it adds very much nothing to our level of knowledge. And that is critical thinking and continuous rethinking of how we assess drug development" (risk assessor).
The interviews highlight that every stakeholder has a specific role to play in the process of developing and taking up NAMs for RA, whilst being highly dependent on each other for an successful implementation. The stakeholders had specific expectations from each other and from collaborative projects, even though not all expectations may be realistic. However, the need to get all stakeholders aligned is not new: previous research has highlighted the need for 'concerted and coordinated efforts with clear goals' (Knight et al. 2021) and that an 'approach to communication is required between stakeholders' (Schiffelers et al. 2012).

Communication between stakeholders
Linked to the described roles and responsibilities of stakeholders is the communication between these different players active in RA. All interviewees explicitly state that the quality of communication is a vital factor to understand the needs in RA for each involved stakeholder. This is considered essential for ensuring a well-functioning and successful RA, which is key for implementing NAMs. However, the quality of current communication efforts was perceived as highly different between the interviewees: some stakeholders saw a good communication between parties, others called for improvements especially between risk assessors and industry applicants. These perceptions seemed to depend on stakeholder's background, their contact with other stakeholders and the ongoing interactions between stakeholders. The next section differentiates the stakeholder positions further.
Besides providing funding for NAMs research, risk managers and some risk assessors perceive themselves as facilitators for communication between the different stakeholders as well as promoters of validated NAMs: "A lot of the communication [between the stakeholders in the risk management and assessment] is facilitated by Commission staff. So, obviously we make every effort to make sure everybody has a say in what is going on" (risk manager).
Risk managers organize public consultations to get insights into the needs of all stakeholders. One risk manager described the communication with other risk managers and risk assessors as two-sided: formal communication procedures, such as staff working documents and frequent official interactions and bilateral interactions, which have developed from for example initially working together, and after changing positions, staying connected on a personal level. "So that all strengthens the ecosystem" (risk manager).
For risk managers, strong communication with risk assessors is essential because of the strict responsibility division in the EU when it comes to RA and risk management. Risk managers are focused on their responsibility related to legal and policy aspects and rely on the agencies' communicated information about technological aspects, to inform their decision-making process. Consequently, other stakeholders described risk managers as "(…) customers of the results or output […]. So, they are looking at what we are doing, what the results are, rather than having a deep discussion" (academic). In the interviews, the keyword "regulator" became an indicator for understanding the responsibility division. While risk managers clearly distinguish responsibilities between risk managers and risk assessors, this seemed more blurry for risk assessors. Business representatives and NGOs make no or only minor distinctions between risk managers and risk assessors and assign them similar responsibilities.
One academic perceives the quality of communication between risk managers and risk assessors as insufficient for EU programs such as Horizon 2020. Risk assessors cannot be actively involved because this would be a double funding of the programs where "the Commission is paying for the project, and EFSA is funded by European money" (academic). Risk managers already should represent the needs of risk assessors in the current funding programs, even though the extent to which this is and can be done is questioned by interviewees. Additionally, the quality of communication between academics and risk managers was considered unsatisfactory: initial questions driving research programs and requests to not always be clearly formulated, and academics described that they have to make assumptions about what risk managers expect when drafting research proposals.
The communication between the involved EU agencies is highly dependent on their mandates. The mandates of EFSA and ECHA allow for collaborative projects, leading to closer communication between them than with EMA "(…) because of the similarities between pesticides and biocides" (risk assessor). However, organizations communicate predominantly by publishing their results and seem to assume that other agencies will read them.
An additional reason for differences between EFSA and ECHA, when compared to EMA, is that EMA's approach to RA is considered more intervened with risk management than the RA of EFSA and ECHA. Officials representing EMA perceived themselves as facilitators for the communication between industry and science to encourage the development and use of NAMs. EFSA aims to be open to everyone and organizes stakeholder meetings when drafting essential documents, which is considered a response to the experienced general lack of trust from i.e. NGOs. One risk assessor describes especially the communication with animal welfare NGOs as "going in circles", like an exchange of arguments to defend personal opinions instead of finding solutions for the issues discussed. Risk managers recognize that there "tend[s] to be a lack of communication between the two parties, […] researchers and government agencies. I think the 3Rs centres can be a very good neutral arena, where these two parts can actually start communicating" (risk manager). Different expectations between risk assessors and academia-and their (lack of) communication about it-hamper the implementation of potentially useful NAMs into the RA. While risk assessors need NAMs that can support RA under existing regulatory procedures, academia follows recent scientific developments and focuses on peer recognition and may thereby answer questions that do not immediately fit within current RA requirements.
In the industry, RA dossiers are developed by researchers, sometimes in collaboration with academia. Interviewed business representatives often compare the communication of ECHA and EFSA with the US FDA and perceive ECHA and EFSA as repudiating. Communication in the EFSA stakeholder discussion group (StaDGER) is considered insufficient. In contrast, business representatives experience the communication with the US FDA generally more open, and specifically more encouraging when it comes to using NAMs. They mention to not understand why direct communication with EMA is possible, whereas no comparable counterparts exist in ECHA and EFSA. "[In EMA businesses] often ask questions on whether there is a need for a specific type of animal study, mandated by guideline. They very often request a waiver for some studies, very often the applicants request general advice" (risk assessor).
This close communication between risk assessor and applicants is believed to support the reduction of unnecessary animal studies. The timing of communication with EU agencies, often taking place at the end of the RA process, is considered problematic by business representatives. This is believed to lead to a higher amount of conducted animal studies, especially for safety dossiers from inexperienced companies. "[Having a] dialogue earlier on with the right regulator, is really key to the overall success and overall understanding to make sure that both parties are clear about where their product is going through" (business representative). Due to the experienced lack of communication with EFSA, one business representative expects opportunities at international level to implement NAMs outside of the EU, or through international guidance documents. And consequently, this could positively impact EU's RA. Like for all stakeholders, communication was a crucial factor for the NGO representative, who claimed to have a vast network of players active in NAM development and implementation. NGOs are facilitators for research by providing research material for animal-free studies and financial funding, by supporting both collaborations with international partners and registrants who are active in the grey area of regulated products (e.g., REACH and cosmetic regulations which require animal testing for workers safety whilst mandating non-animal methods for cosmetics, respectively). Also, NGOs are involved in promoting animalfree methods by providing training for using animal-free methods. In their roles, NGOs try to identify the needs of other stakeholders and aim for finding consensus in collaborative projects or convince specific stakeholders to keep collaborations going. In contrast, academics and risk assessors described animal welfare driven stakeholders as inflexible, a claim that NGOs agreed regarding specific points, because the communication merely consists of an exchange of arguments without getting anywhere.
All stakeholders believed that communication is key for successfully collaborating with each other to develop and implement NAMs in RA procedures. They mentioned slight communicative discrepancies within primary stakeholders and conceded difficulties between primary and secondary stakeholders. Similarly to the roles and responsibilities of stakeholders (see Sect. 3.3), the views differed regarding the quality of current communication efforts and opportunities. EFSA's risk communication may be optimized through the adoption of the Transparency Regulation [Regulation (EU) 2019/1381], but also other stakeholders could support fostering communication between and amongst stakeholders in RA. The call for transparency in decision making requires something that was not a priority in the beginning of the RA: risk assessors need to explain in a clear language what they are doing and how they have reached their conclusions, which in turn may bring better communication between the stakeholders (Bridges 2003). As shown in previous research, constant dialogues and feedback between all involved partners are essential for NAM acceptance-from development to implementation (Avila et al. 2020). Our results suggest that this would benefit further optimization.

Collaboration
All stakeholders agreed that workshops and collaborative projects are efficient platforms for communication and collaboration, because it brings a variety of stakeholders together to discuss and share their needs. Additionally, getting new perspectives from various stakeholders can be integrated into their work, supporting the implementation of NAMs. However, an NGO representative stated that in order to make a collaboration successful, the objectives of all stakeholders should be well aligned. With different stakeholders having their own agenda and priorities, collaborations were described as "herding the cats" (NGO representative). Especially with potentially changing priorities over time, long-term collaborations can be challenging.
Currently on EU level, many ongoing different collaborative projects on NAMs with a wide range of expertise from academia, research institutes, member state, and risk assessment bodies are brought together. These projects focus on the technicalities of NAM development for regulatory purposes. However, business representatives called for stronger collaborations: "[We need] more opportunities to work as a tripartite group of interested parties. So, we need industry, we need regulators, and we need academics coming together, and working on projects together to solve issues around animal-free approaches for food safety, but in many of the regulatory areas as well". The interviewed business representatives said that US risk assessors are both easier to approach and to collaborate with and thereby, collaborations are seen as more fruitful to integrate NAMs in RA.
In contrast, risk assessors perceived their collaboration with the industry as sufficient and mentioned that contacts with the industry where made through representative organizations such as European Partnership for Alternative Approaches to Animal Testing (EPAA). All interviewed business representatives, academic scholars, and a risk assessor saw U.S. institutions as essential collaboration partners when it comes to NAMs development and implementation: interviewees describe that in the U.S., a solid vision has been developed to facilitate the development and implementation of NAMs, something that is currently missing in the EU.
One reoccurring mentioned factor important for both communication and collaboration was the effective use of existing data. Many types of data are already available, such as data and findings that provide information for read-across, the development of other NAMs as well as their implementation in RA. Academics and risk assessors suggested that an increased transparency of academic reviews by establishing an open peer-review as a standard for publications may stimulate the accessibility of more high-quality data. One academic said that open-source publications like Zenodo, 1 an online academic repository, would make data accessible to everyone. Since many academic publications do not offer insights into raw data, transparency of academic publications is considered low, and the data can often not be used for regulatory purposes. Accessing data that is already available from RA or business operators is challenging, because of intellectual property and other legal restrictions. This leads to replicating studies and thereby, increased use of animal studies. One interviewed academic put forward that there is a general need for an encrypted data-sharing platform between EU agencies and companies, without infringing legal confidentiality. An interviewed risk assessor went even further and proposed a database in which all data from human cohort studies are collected for scientific use, to support the improvement of EUs RA. Additionally, interviewees emphasized that the legal framework and associated guidance documents determine which data can be used for RA. Business representatives said they cannot use innovative methods based on existing data as much as they would like: "If you look at EFSA, I have got guidance documentation […]. But […] all it talks about is animal testing".
Even though most risk assessors recognize the need for increased transparency for NAMs acceptability, one risk assessor mentioned that guidance documents provide sufficient freedom of interpretation for the use of NAMs: "With proper scientific thinking, actually, most of our European guidelines of non-clinical studies can be interpreted in a 3Rs compliant manner".
Collaboration between stakeholders to support the development and uptake of NAMs was a key topic in the conducted interviews. Every interviewee mentioned the need to further optimize collaborations to foster the development, validation, and implementation of NAMs in EU's RA for chemicals. Such collaborations are highly dependent on stakeholders' roles and responsibilities (Sect. 3.3) and communication between parties (Sect. 3.4). Dedicated collaborative workshops have previously been described in literature as important steps to share data and insights (Basketter 2020;Graepel et al. 2019;Mahony 2019;Mahony et al. 2020;Mone et al. 2020). Both are essential for building confidence regarding NAMs, by showing what data can be generated through NAMs, and how this data should be interpreted and how it can be aligned with RA requirements, which may not necessarily optimally reflect scientific insights. In general, all stakeholder groups perceived that legislation cannot keep up with the pace of progression in science. A recently EURL ECVAM commissioned AOP framework study made a similar conclusion (Zuang et al. 2021).

Moving forward
The use of NAMs for RA processes is considered both a cultural and a generational issue. The interviews highlighted that all stakeholders considering and are willing to use animal-free approaches in analyzing the safety of orally ingested chemicals used as chemical ingredients, foods and medicinal products. However, risk managers and risk assessors were insufficiently confident to implement them, because of missing reliability and reproducibility of data generated with these methods. Business representatives and the NGO representative feel strongly restricted by current legislation, also by the actions taken by risk managers and risk assessors. Patterns, behaviour, dependencies and different requirements are influencing stakeholders' perceptions and their approach to innovation, and this in turn affects the RA. The culture in RA is a mix of different stakeholder's cultures, roles, and responsibilities. Understanding the culture of stakeholders is vital for implementing NAMs "(…) Instead of just focusing on the technicalities of RA" (academic).
From this study, it became clear that stakeholders had differing views on the benefits and suitability of NAMs, ranging from their ideas about costs to their reliability and relevance. These perspectives seemed to be attributable to their background, experiences, and stakeholder positions. This is again a hurdle for their implementation into the RA (Zuang et al. 2021). NAMs innovate an RA that is currently based on traditional methods, and they should be slowly implemented alongside traditional methods to overcome the bias induced scepticism. Sufficient reliable and reproducible data are needed to increase confidence of and reduce uncertainties around NAMs based RA. All stakeholders want to ensure the safety of products. However, the pursuit for securing safety leads to uncertainties in the approach of safety assessments. These need to be resolved internationally-preferably at OECD level-in order to overcome them at EU level (Wolf et al. 2020). With the EU considering international standards in legislation, and scientific agencies contributing to international harmonization initiatives (Röttger-Wirtz 2017), international collaborations are vital for NAMs advancement and uptake.
In general and independently of their background, all stakeholders were optimistic that NAMs will support the development of more accurate and sustainable RA practices. However, most interviewees were critical about the time frame for implementing NAMs in the current RA system of toxicological testing, because the current observing overexposure approach is a time-consuming hurdle for a more accurate RA and the implementation of NAMs. They suggested that current RA should move away from 'box ticking exercises' that impede using NAMs in order to fulfil regulatory requirements to an approach that focuses on real-life conditions. But already prior to establishing a new approach, NAMs could be implemented alongside with traditional methods as Tier 1 screening to reduce animal testing, which builds confidence and understanding for NAM data. An ontology framework could facilitate NAM development for regulatory purposes and to understand human biology, such as the MoA ontology model for safety evaluation of chemicals, the RISK21-framework, and other compartments (IATAs, AOPs, defined approaches, case studies etc.) (Turley et al. 2019). Detailed understanding of human biology could further change the current emphasis in RA on testing adversity endpoints for analysing pathways.
Stakeholders perceived the culture of stakeholder groups as a hurdle for reaching the next generation RA approach. Discrepancies in the (perceived) communication between primary and secondary stakeholders hinder to fully understand each other's needs and to implement those requirements in their work. For instance, academics and risk assessors have different focus' leading to academically scientific outputs that risk assessors cannot use for regulatory purposes. Legal restrictions on EU and industry level impede data sharing for further collaborations, development, and validation of NAMs. This requires open data agreements without exceeding political and industrial restrictions, which the European Open Science Cloud (EOSC) could provide in future. To reduce potential biases affecting the validation and implementation of NAMs in RA, all stakeholder groups need to share a common understanding and build confidence. EURL ECVAM's knowledge management strategy aims to reach this for AOPs on international level (Zuang et al. 2021). This can be seen as a first important step towards better implementation of NAMs in RA for chemicals.

Strengths and limitations
Whereas previous research and workshops have addressed the development and uptake of NAMs, this is the first qualitative study aimed to generate insights into the perspective of stakeholders into the implementation of NAMs in the full risk analysis cycle in the EU. This approach allowed for generating detailed insights into the perspective and experiences of stakeholders, providing a comprehensive overview with nuanced insights into the implementation and validation of NAMs in RA, in particular in the combination of risk assessment and risk management wishes and needs. There are however various limitations linked to the study design and conduction. Firstly, the large complexity of players active in NAMs development and uptake in the EU's RA may have resulted in some stakeholders being insufficiently represented in this study. Still, the high information power of the interview sample, that has resulted in obtaining complex, in-depth and rich data on the perspectives of stakeholders representing different views, suggests that key issues influencing the acceptance and uptake of NAMs were highlighted well, allowing for generalizing the findings to the broader categories of NAMs and RA stakeholders. Secondly, various interviewees made no or only minor distinctions between risk managers and risk assessors and assigned them similar responsibilities under the term "regulator", which made distinctions between the perceived risk managers' and assessors' roles and responsibilities for certain points difficult. Finally, to reduce the possibility of researcher bias at data collection and data analysis, we used bracketing to identify personal perspectives potentially relevant to the study. Researcher reflexivity, together with coding of the first author being continuously discussed with both other authors, should reduce the risk for informed bias.

Conclusion
By exploring the perceptions of active stakeholders for RA of chemical products in the EU, this study highlights the hurdles and opportunities for implementing NAMs for current RA of the EU. Previous research shows that the current approach of RA, being highly dependent on animal models that are used for testing adverse effects of orally ingested compounds, needs to be updated by stimulating the use of NAMs (de Boer 2019;de Boer et al. 2020;Knight et al. 2021;Vinken et al. 2020;Vrolijk et al. 2020).
However, as previously described by Avila et al. (2020) andSchiffelers et al. (2012), NAM acceptance requires having dialogues and providing feedback amongst all involved partners. For example, Schiffelers et al. (2012) showed that both intensified communication between all stakeholders and sharing test data can increase certainty regarding NAMs, leading to an increased use for RA.
Our research brought together perceptions of stakeholders from different parts of the process (risk assessors, risk managers, business representatives, academics, and an NGO representative), which showed that all stakeholders have the goal to increase uptake of NAMs in RA, but different views on how NAM development and uptake can and should be stimulated. Also, the perceived roles, responsibilities, and needs were not always well-aligned between stakeholders. To move forward with the development, validation, and implementation of NAMs, it is important to build a roadmap to evolve risk assessment by stimulating stakeholders to improve communication rather than trying to revolutionize it (Schiffelers et al. 2012). Therefore, it is essential to consider to what extent the legislative framework stimulates the uptake of NAMs, and to foster research collaborations and confidence building by making data accessible and insightful. A shift in testing requirements, from the traditional overexposure approach to more predictive, mechanistic testing in RA may still require adjustments in legislation and therefore, will take time. However, this study shows that if all stakeholders are willing and are capable to engage in communication and confidence building, NAMs can already play an important role in reducing, refining, and replacing animal testing.

Conflict of interest
The authors declare that they have no conflict of interest.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.