Preparing for CHANGE: How systems thinking supports an effective transition from animal studies to NAMs

Abstract

None

1. Introduction

Substantial time, money and effort have been invested in shifting the toxicology testing paradigm away from animal studies to a system based on New Approach Methods (NAMs), yet animal data remain the default in most regulatory contexts. Moving away from the default requires a thoughtful transition to ensure that the potential benefits of NAMs are realised while undesirable consequences, such as compromising the level of health protection afforded by chemical regulations, are avoided.

To help address this issue, we established the CHANGE project, a three-year, multi-disciplinary, international initiative to promote effective use of NAMs in regulatory and other decision-making relating to the use of chemical substances (Mathisen et al., 2024). CHANGE is funded by the European Food Safety Authority “Tailor-Made Activities” programme, the Norwegian Scientific Committee for Food and Environment, and is supported in-kind by the Evidence-Based Toxicology Collaboration.

The fundamental insights driving our approach throughout CHANGE are that the processes of NAM development, evaluation, and application in decision-making happen in a system, that the shift away from animal studies towards effective use of NAMs can be achieved more rapidly if strategies for achieving the shift are directly targeted at system behaviour, and that designing such strategies requires the development of a comprehensive model of the complex dynamics of the regulatory toxicology system.

In this commentary we summarise why, according to the model we developed in CHANGE, the regulatory toxicology system is inherently resistant to the uptake of NAMs. We present where system-level interventions could tackle this resistance, and we describe the core features of our methodological approach that distinguish CHANGE from other systems-oriented initiatives.

CHANGE represents the first effort to model the regulatory toxicology system as an interconnected dynamic system. The present manuscript therefore serves as a high-level guide to methodological information and modelling outputs presented elsewhere and summarised in Table 1. This manuscript also presents the foundations of a forthcoming white paper of strategic, system-level interventions that will be the final output of CHANGE.

Table 1.

Overview of principal CHANGE publications.

CHANGE publications	Summary
Time for CHANGE: system-level interventions for bringing forward the date of effective use of NAMs in regulatory toxicology (Mathisen et al., 2024)	Commentary introducing the CHANGE project, its rationale, objectives, and three-phase “explore, reflect, design” methodological approach.
Exploring experiences of the regulatory toxicology system: system-level promoters and inhibitors of new approach methodologies (Bearth et al., 2025)	Research article describing the “explore” phase of CHANGE, eliciting from people working in or around the regulatory toxicology system their experience of barriers and enablers of use of NAM data.
Improving how we use workshops when solving complex research problems: reflections from the CHANGE project (Jones et al., 2025)	Methods article presenting recommendations for conduct of participatory research workshops, emphasising project team structures, effective social interaction, and facilitation strategies.
How to organise a successful toxicology workshop? A participant perspective on the Collaboration to Harmonise the Assessment of Next Generation Evidence (CHANGE) workshop in Oslo, 18–20 June 2024 (Diemar et al., 2025)	Commentary and companion piece to Jones et al. (2025), presenting participant perspectives on the first in-person CHANGE workshop, reflecting on the strengths and challenges of the approaches used.
Modelling the Regulatory Toxicology System to Promote the Effective Use of New Approach Methodologies: A System Thinking Approach (Bearth et al., submitted)	Research article describing the “reflect” phase of CHANGE, presenting the CHANGE system model and the methods used to derive the model from participants’ experience of working in or around the regulatory toxicology system.
Report of Collaboration to Harmonise the Assessment of Next Generation Evidence (CHANGE) 2nd workshop in Oslo, June 24 - 26, 2025 – participants perspectives on the CHANGE working methodology (Diemar et al., submitted)	Provides participants' reflections on the second ("reflect") in-person workshop, including the collaborative approach used to refine system understanding, and the value of the workshop for identifying ways to support the uptake of NAMs.
A protocol for organising co-design research workshops (title to be decided, in preparation)	A repeatable, step-by-step procedure to support the planning, design and delivery of successful co-creation or co-design workshops.
CHANGE white paper (title to be decided, in preparation)	The main intended deliverable of CHANGE, a policy white paper for decision-makers working in or around regulatory toxicology, presenting a series of system-level interventions that can be expected to improve the effectiveness of use of NAMs in regulatory toxicology.
Designing system-level interventions for improving the effectiveness of use of NAMs in regulatory toxicology (title to be decided, forthcoming)	Final research article describing the methods and results of the “design” phase of CHANGE, as the detailed work underpinning the White Paper.

2. Why the regulatory toxicology system is resistant to NAMs

In our model, developed using a causal loop diagram approach to system modelling (Kim, 1999; Meadows, 2008), the regulatory toxicology system is shown to be resistant to NAMs due to 16 feedback loops which combine to preserve the status quo despite appetite for change (Bearth et al., submitted). These feedback loops can be summarised in terms of two primary explanations as to why the regulatory toxicology system behaves in a way that makes it resistant to more effective use of NAMs.

The first system-based explanation is that using NAMs instead of animal studies for decision-making is associated with higher actual and expected risks that a decision based on NAM data will “fail” through some combination of being subjected to successful legal challenge, insufficiently protective of health, excessively costly, or inhibitive of innovation in chemical use. In this context, animal testing remains dominant not because it is always better, but because it is the incumbent, it is legally ingrained, supported by decades of infrastructure, guidance, expertise, and institutional familiarity, and generally regarded as being able to protect humans and the environment. By contrast, NAMs are of a more recent nature, less well-established, and many actors are less familiar with how to apply them; consequently, additional evidence is often needed to demonstrate reliability and predictability for human health endpoints compared to established animal-based approaches. Therefore, the use of NAMs for decision-making may be associated with higher expected or actual uncertainty and/or risks, and as a result, resources, confidence, and credibility remain focused on animal testing, keeping the system tied to familiar and established practices.

The second system-based explanation is that the system favours short-term reassurance, procedural and market certainty even while recognising the potential value of making greater use of NAMs. To reduce uncertainty about the consequences of NAM use, or to speed up decision-making, organisations often rely on familiar and established approaches that are embedded in regulatory and legal frameworks (which in this case is requiring or conducting animal tests) or due to lack of data may not assess a chemical at all. These “quick fixes” provide short-term reassurance or procedural certainty at the cost of gaining experience and familiarity with NAMs, which maintains a status quo of limited trust and lack of experience in using NAMs, discourages early adopters to move forward in using NAMs, and fosters continued reliance on animal data.

3. The drivers of systemic resistance to NAMs

In CHANGE, we have identified three main drivers of systemic resistance to effective use of NAMs. The drivers are a “disconnected system” of siloed actors, a “bottleneck” of a high rate of NAM creation and a lower rate of evaluation for application, and reinforcement of “early mover hesitancy” by the actual and expected legal, financial, and reputational risks to decision-makers of applying NAMs outweighing their benefits. Designing interventions that target these drivers, reversing or breaking the feedback loops that result in systemic resistance to NAMs while avoiding undesirable consequences, should lead to more effective use of NAMs.

3.1. Driver 1: A disconnected system

Regulatory toxicology is a complex system in which actors from academia, industry, risk assessment, and risk management have distinct roles, responsibilities, processes, and culture, and operate largely within separate silos. This siloing creates inconsistent expectations about what regulatory and other decision-making actors need and what scientific researchers can provide, impedes learning about what is and is not working in relation to NAM uptake and why, leads to duplication of effort, inconsistent interpretation of NAM data, and erodes trust between actors.

Interventions to de-silo the system could focus on encouraging existing actors or creating new ones to take on mediating and connecting roles between other actors in the system. Such mediating roles could include connecting and translating NAM innovation into regulatory applications, identifying and further developing NAMs that are promising for decision-making, and ensuring that NAMs are usable in decision-making for specific contexts.

Without improved connection between silos, the funders, academics, contract research organisations, and other actors involved in developing NAMs will not understand the needs of regulators and other actors using NAMs to make decisions, while the actors involved in using NAMs to make decisions will not be able to make their demands for NAM developers realistic or sufficiently well-understood.

3.2. Driver 2: A NAM innovation-application bottleneck

An imbalance exists within the NAM ecosystem, whereby a productive component of the system feeds into a downstream component that lacks the capacity to absorb this productivity.

NAM innovation is primarily conducted within research institutions that are culturally and infrastructurally optimised for generating novel scientific knowledge, resulting in rapid technological development. This innovation is driven by incentives such as competitive funding, publication pressure, and recognition for scientific novelty.

In contrast, activities essential for applying NAMs in decision-making - including legislative processes, method validation and assessment of limitations, and regulatory guideline updates - tend to be slower, and receive less funding, and result in fewer publications. Application-oriented activities prioritise robustness, standardisation, and fitness-for-purpose rather than novelty and academic outputs. As a result, the validation and implementation part of the system cannot keep pace with the rate of innovation.

Addressing this bottleneck could focus on incentivising more evaluation and validation of NAMs, increasing the pace and volume of evaluation activity, integrating decision-maker needs into the NAM creation process, and gaining experience in using NAMs and sharing information about when and why NAMs have been successfully used or not.

Without interventions to overcome the bottleneck, the opportunity cost associated with underutilised NAM technologies is likely to increase, resulting in continued investment in methods that are not translated into practice, along with increasing overall uncertainty about when and how any given NAM might be successfully applied in decision-making.

3.3. Driver 3: Reinforcement of early mover hesitancy

Using NAMs to make decisions - be it on the industry side when developing and bringing products to market, or in regulatory contexts when setting restrictions on chemical substances - carries with it a collection of legal, financial, and reputational risks. Some of these are real and some are perceived, but either way they can make decision-making organisations cautious about diverging from established practices and defaulting to animal data instead.

Until NAM data is established in a decision-making process, there is insufficient incentive to use such data because the anticipated risk of doing so outweighs potential benefits for the user. However, without use the NAM data will not become established. This creates a self-reinforcing dynamic of early-mover hesitancy where the anticipated risk of using NAM data prevents the data being used, which prevents the accumulation of experience needed to understand, validate, and appropriately manage those risks.

Early-mover hesitancy is reinforced by interdependent expectations between industry and regulators. Industry actors may be reluctant to rely on NAM data due to uncertainty regarding regulatory acceptance and the potential financial consequences of using NAMs, for example if additional animal studies are still required. At the same time, regulatory actors with the obligation to protect health and the environment may default to established approaches due to limited experience with NAMs, legal caution, and uncertainty regarding NAM reliability in specific contexts.

Intervening to break this dynamic involves a complex balancing act of de-risking early adoption of NAM data in decision-making while ensuring that adoption of NAM data does not translate into increased health risks for the populations exposed to chemical substances. Where risks are anticipated but not real, building up experience and improving communication so decision-makers more accurately anticipate risks may be considered; where risks are real, ways of reducing legal and financial exposure of either regulators or regulated parties could be considered.

4. Using a system model to design interventions that will improve the effectiveness of use of NAMs

Systems change in response to interventions that reshape system dynamics (Meadows, 2008). Intervening to successfully address the three drivers of systemic resistance to NAMs, and to increase the effectiveness of use of NAMs in decision-making, requires a detailed and comprehensive system model. The model we have developed in CHANGE consists of more than 80 system variables that interact in 16 dynamic feedback loops (Bearth et al., submitted).

To translate our understanding of system dynamics into improving the effective use of NAMs, we incorporated a stock-and-flow component (the “NAM maturation chain”) into our overall model. This component describes the process of NAM development, evaluation, and eventual use in decision-making, and is designed to model how interventions influence the progression of NAMs from development to use, retaining NAMs that are promising and eliminating NAMs that are not. This component enables users of our model to evaluate the extent to which interventions are likely to represent a system-level net gain or net loss in effectiveness of use of NAMs.

Throughout this commentary, we have been emphasising effective use of NAMs. By effective use of NAMs, we mean that the potential benefits of NAMs are realised while undesirable consequences (such as compromising protection of health and the environment) are avoided. The CHANGE system model supports the design and prioritisation of interventions that can be expected to shift system dynamics in a desirable direction, towards a system in which NAMs providing actionable evidence are effectively translated from innovation into routine use in decision-making.

Callout 1 shows in simplified form how the feedback loops modelled in CHANGE can be used to design interventions for improving the effectiveness of use of NAMs. The NAM maturation chain and its connection to feedback loops is explained in detail in Bearth et al. (submitted).

5. Methods for creating the CHANGE system model

System modelling presents inherent challenges, as the drivers of system behaviour are often not directly observable. A system can be conceptualised as an iceberg (Fig. 1): while events in a system are visible at the surface, and patterns behind events are somewhat observable, the underlying structures and mechanisms responsible for events are not directly observable and need to be extrapolated from observed events and patterns.

Fig. 1.

The iceberg model of system observability.

Based on the premise that the underlying structures of a system can be revealed through the observations and experiences of persons working within it, we employed a novel, rigorous, and inclusive co-creation methodology involving almost 100 international participants from industry, government, NGOs, and academia, reflecting the fact that NAM use spans multiple interdependent regulatory frameworks across and within countries and impacts a wide range of people across society.

Our approach in CHANGE has consisted of three phases, each involving data collection in workshops: an “Explore” phase (2024) where we collected participant experience in working with or around NAMs, which we interpreted into a series of 51 hypotheses about system function; a “Reflect” phase (2025) where we challenged and refined with participants the individual hypotheses about system function into a coherent, integrated system model; and finally a “Design” phase (2026) that uses our system model as the basis for developing with participants strategic interventions, the results of which will be presented as a White Paper (see Fig. 2).

Fig. 2.

The CHANGE approach to designing interventions for improving the effectiveness of the use of NAMs.

6. Conclusion

Limited progress in the use of NAMs for regulatory decision-making, despite considerable investment of time, money, and effort, suggests a need for novel approaches to designing strategic, system-level interventions that will improve the effectiveness of use of NAMs.

CHANGE offers the first systems-based response to this challenge, pursuing a lengthy, cross-disciplinary, multi-stakeholder effort to provide a valid, comprehensive model of system dynamics, grounding the design of strategic interventions in solid theory that explains how the interventions can be expected to work.

We look forward to seeing our ideas and approach being extended, challenged, and refined by groups that share our goal of ensuring NAMs make their full contribution to timely, efficient, health-protective decision-making around chemical use.

Disclaimers

The views expressed in this manuscript are solely those of the authors.

The author Kristina A. Thayer is employed at California Environmental Protection Agency’s Office of Environmental Health Hazard Assessment (OEHHA). The views are those of the author and do not necessarily reflect the views or policies of OEHHA or California Environmental Protection Agency.

The authors Helena T. Hogberg and Andrew A. Rooney are employed at the U.S. NIH, National Institute of Environmental Health Sciences. The contributions of the NIH author(s) are considered Works of the United States Government. The views expressed in this manuscript are those of the authors and do not necessarily represent the views or policies of the views of the NIH or the U.S. Department of Health and Human Services.

CHANGE is funded by The European Food Safety Authority (EFSA). The views expressed in this manuscript are those of the authors and do not necessarily represent the views or policies of EFSA.

Callout 1. Simple feedback loop abstracted from the CHANGE system model, to explain the role of the model in designing interventions for improving the effectiveness of use of NAMs in regulatory toxicology.

In this callout, we show part of the “accumulating experience in regulation” feedback loop in regulatory toxicology, as modelled in CHANGE under the “early mover hesitancy” driver of resistance to change (see Section 3.3 of this manuscript). This is to illustrate how system modelling using causal loop diagrams can support intervention design for improving the effectiveness of use of NAMs. Readers should note the loop has been abstracted, redrawn, and simplified for the purpose of illustration. The full model is much more complex, with each variable connected to many other variables and feedback loops. The full model is shown and explained in Bearth et al. (submitted).

The yellow circles (nodes) in the diagram show variables in the system model. The arrows in the diagram show how one variable increases or decreases a connected variable. A solid arrow means two connected variables change in the same direction (e.g. more stress means more headaches, or less stress means less headaches). A dashed arrow means connected variables change in opposing directions (e.g. more stress means less sleep, or less stress means more sleep).

The circles and arrows are connected in a reinforcing feedback loop, whereby (starting at the bottom node) if regulatory actors had more experience in using NAM data for decision-making, there would be less uncertainty and perceived and actual risk for decision-makers, which would result in less early mover hesitancy among regulatory actors (i.e. reluctance to use NAM data in decision-making), which would mean more openness of the regulatory system towards NAMs, which would mean less reliance on old routines in regulatory toxicology, which would mean there would be more NAM familiarity, knowledge and skills among regulatory actors, which would lead to more experience of regulatory actors using NAM data for decision-making, and so on around the loop.

To help ensure this loop results in more positive outcomes in relation to NAM use, one intervention that could be implemented would be the addition of simulation training to the model (red node), whereby realistic decision-making scenarios are created within which regulators can practice using NAM data. More simulation training should increase experience of regulatory actors in using NAM data for decision-making, which should reduce uncertainty and perceived and actual legal risk for regulatory actors, which should reduce early mover hesitancy, and so forth.

Of course, in a complex system simulation training may help, but it may not make a huge difference on its own. There may also be system factors that work against it, such as reluctance of prospective participants to spend time on training in routines they do not consider to be useful for their jobs, and possible negative consequences such as over-confidence with NAM data. More complex interventions that target multiple nodes and manage undesirable consequences would therefore be likely to make bigger differences - but more complex interventions may also require actors with more power and influence, such as international funders or governments, and may therefore be more challenging to implement.

The point of the system model is to support a comprehensive approach to identifying and reflecting on such issues, to help design potentially effective interventions that are grounded in a shared understanding of how these interventions can be expected to work. The variables and connections in the model are intended to be intuitive; it is how they are made explicit and connected into a whole system that has value for intervention design.

Illustration of part of the "accumulating experience in regulation" feedback loop from the CHANGE system model, including an intervention (simulation training) that could support more effective use of NAMs.

Alt-text: Unlabelled box dummy alt text

Funding

This article originated from Collaboration to Harmonise the Assessment of Next Generation Evidence (CHANGE), which is funded by the European Food Safety Authority (EFSA), The Norwegian Scientific Committee for Food and Environment (VKM) and the Evidence-Based Toxicology Collaboration (EBTC).

CRediT authorship contribution statement

Gro H. Mathisen: Conceptualization, Funding acquisition, Investigation, Project administration, Visualization, Writing – original draft, Writing – review & editing. Paul Whaley: Conceptualization, Funding acquisition, Investigation, Visualization, Writing – original draft, Writing – review & editing. Birgit Kopainsky: Conceptualization, Investigation, Writing – review & editing. Heather M. Ames: Conceptualization, Funding acquisition, Investigation, Writing – review & editing. Sebastian Hoffmann: Conceptualization, Funding acquisition, Investigation, Visualization, Writing – review & editing. Trine Husøy: Conceptualization, Funding acquisition, Investigation, Writing – review & editing. Lowenna B. Jones: Conceptualization, Investigation, Writing – review & editing. Camilla Svendsen: Conceptualization, Investigation, Writing – review & editing. Gunn E. Vist: Conceptualization, Funding acquisition, Investigation, Writing – review & editing. Denise Bloch: Writing – review & editing. Aleksandra Cavoski: Writing – review & editing. Weihsueh A. Chiu: Writing – review & editing. Holly G. Davies: Writing – review & editing. Arianna Giusti: Writing – review & editing. Thomas Hartung: Writing – review & editing. Helena T. Hogberg: Writing – review & editing. Seok Kwon: Writing – review & editing. Olivia J. Osborne: Writing – review & editing. Andrew A. Rooney: Writing – review & editing. Christophe Rousselle: Writing – review & editing. Jennifer B. Sass: Writing – review & editing. Ulla Simanainen: Writing – review & editing. Kristina A. Thayer: Writing – review & editing. Daniele Wikoff: Writing – review & editing. Angela Bearth: Conceptualization, Investigation, Writing – review & editing.

Declaration of competing interest

The authors declare the following financial interests/personal relationships which may be considered as potential competing interests: Gro H Mathisen reports financial support was provided by European Food Safety Authority. If there are other authors, they declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Data availability

The data that has been used is confidential.