Advancing the Adverse Outcome Pathway Framework – An International Horizon Scanning Approach

Abstract

Our ability to conduct whole organism toxicity tests to understand chemical safety has been outpaced by the synthesis of new chemicals for a wide variety of commercial applications. As a result, scientists and risk assessors are turning to mechanistically-based studies to increase efficiencies in chemical risk assessment and are making greater use of in vitro and in silico methods to evaluate potential environmental and human health hazards. In this context, the adverse outcome pathway (AOP) framework has gained traction in regulatory science since it offers an efficient and effective means for capturing available knowledge describing the linkage between mechanistic data and the apical toxicity endpoints required for regulatory assessments. A number of international activities have focused on AOP development and various applications to regulatory decision-making. These initiatives have prompted dialog between research scientists and regulatory communities to consider how best to use the AOP framework. While expert-facilitated discussions and AOP development have been critical in moving the science of AOPs forward, it was recognized that a survey of the broader scientific and regulatory communities would aid in identifying current limitations while guiding future initiatives for the AOP framework. To that end, a global Horizon Scanning exercise was conducted to solicit questions concerning the challenges or limitations that must be addressed to realize the full potential of the AOP framework in research and regulatory decision making. The questions received fell into several broad topical areas: AOP networks, quantitative AOPs, collaboration on and communication of AOP knowledge, AOP discovery and development, chemical and cross-species extrapolation, exposure/toxicokinetics considerations, and AOP applications. Expert ranking was then used to prioritize questions for each category, where four broad themes emerged that could help inform and guide future AOP research and regulatory initiatives. In addition, frequently asked questions (FAQs) were identified and addressed by experts in the field. Answers to FAQs will aid in addressing common misperceptions and will allow for clarification of AOP topics. The need for this type of clarification was highlighted with surprising frequency by our question submitters, indicating that improvements are needed in communicating the AOP framework among the scientific and regulatory communities. Overall, Horizon Scanning engaged the global scientific community to help identify key questions surrounding the AOP framework and guide the direction of future initiatives.

Graphical Abstract

Going Global to Advance 21st Century Science

Status of the Adverse Outcome Pathway Framework

The evolution of the adverse outcome pathway (AOP) framework is a logical progression emanating from prior efforts focused on pathway-based approaches to support the use of mechanistic data in chemical risk assessment [1–4] (BOX1-Timeline). As discussed by Ankley et al. [1], the AOP framework incorporates and extends the pathway-based concepts of mode and/or mechanism of action commonly used by scientists assessing chemical hazards in human health. The AOP framework, at its simplest, portrays causal linkages between a molecular initiating event (MIE; the initial interaction between a molecule and a biomolecule or biosystem that can be linked to an outcome via a pathway[5]) and subsequent measurable responses (termed key events or KEs) across biological levels of organization that culminate in an adverse outcome—typically at the level of the individual or population—relevant to a given risk assessment scenario (Figure 1).

Box1.

FIGURE 1:

Adverse outcome pathway (AOP) Framework describing components and structure of a linear pathway. Adapted from OECD AOP Handbook

Although the AOP framework represents more of an evolution than a revolution in toxicology and risk assessment, there has been a surprisingly widespread interest in the framework by those involved in assessing chemical risks both to human health and the environment (both in the context of wildlife and ecosystems). Part of this stems from a much-needed advancement in precise and harmonized terminology, and practical demonstration(s) of this terminology to relevant case examples of pathway-based toxicology [1]. But, arguably, the more significant factor contributing to the attention given the AOP framework involves the present convergence between identified needs in the arena of chemical risk assessment/regulation, and the increasing availability/viability of scientific tools to potentially address these needs. For example, regulatory programs such as REACH (Registration, Evaluation, Authorization and Restriction of Chemicals; http://ec.europa.eu/environment/chemicals/reach/reach_en.htm) in Europe, or that being implemented under the 2016 Frank R. Lautenberg Chemical Safety for the 21st Century Act (https://www.epa.gov/assessing-and-managing-chemicals-under-tsca/frank-r-lautenberg-chemical-safety-21st-century-act) in the United States require assessment of the potential impacts of a greater number of chemicals than in the past while using fewer animals.

The data needed to address these legislated mandates cannot realistically be generated using traditional whole animal testing approaches. Rather, there needs to be an increased emphasis on cost-effective tools that enable robust prediction of potential chemical impacts when (as usually is the case) empirical data are lacking (e.g., [2, 3]). Concurrent with the realization that predictive methods will need to play an increasingly prominent role in regulatory toxicology has been the recent explosion in technology in the biological sciences including ‘omics, novel in vitro models, and high-throughput (robotic-based) in vitro testing, which enables the collection of types and amounts of mechanistic data that even a decade ago would have been unimaginable. Furthermore, increased computational power now allows for more efficient handling, storing, and analyzing of the resulting datasets. A pressing need stemming from these technological advances is the ability to effectively apply the generated data to support regulatory decision-making and risk assessment. As such, the AOP framework provides the basic translation and communication tool needed to facilitate the application of predictive toxicology techniques to chemical risk assessment.

This potentially critical role of AOPs has been recognized by different organizations involved in regulatory activities concerning chemicals such as the Organisation for Economic Cooperation and Development (OECD; www.oecd.org). Therefore, as the AOP framework gains international traction, critical evaluation and thoughtful discussion of the strengths and limitations of the framework are necessary to ensure consistent and proper application.

A Horizon Scanning Approach for Global Input

Although the AOP framework has attracted global attention since the publication of the foundational paper in 2010 [1], there remains a relatively small community, primarily in North America and the European Union, involved in guiding the direction of this framework for both research and regulatory decision-making. Horizon Scanning, which is a method of systematically searching for and identifying emerging trends, opportunities, and limitations that might impact the future directions of a defined subject, provides an opportunity to expand input from few contributors to many [6]. At a finer scale, research identification and prioritization exercises [e.g.,[7]] have been used to identify important and actionable research questions that can help align scientific effort with policy priorities [8].

A Horizon Scanning-like exercise was initially described in 1981 by Fahey et al. [9], where a multifaceted survey, focused on environmental scanning and forecasting, was used to identify future (short-term or futuristic) organizational directions. Horizon Scanning approaches have evolved over the years and have been used to elucidate future directions and develop recommendations for advancing science in a variety of fields including invasive species, global conservation, threats to biodiversity, genomic testing, research and policy prioritization [7, 10–15]. A recent example of a research identification and prioritization exercise is the work of Boxall et al. [7], which identified key questions pertaining to the exposure, effects, and risks of pharmaceuticals and personal care products in the environment, in order to prioritize the most critical questions to guide future research initiatives [7]. Questions were solicited globally across representative research and regulatory sectors and discussed during an expert workshop culminating with a top 20 list of questions [7]. The description and outcome of the Boxall et al. [7] exercise have been cited in over 330 publications providing an indicator of the impact of the Horizon Scanning exercise in identifying future research priorities (Google Scholar, accessed February 9th, 2017).

Objectives of AOP-Focused Horizon Scanning Exercise

The main objectives of Horizon Scanning and the subsequent expert ranking exercise described herein were to gauge current perceptions regarding the AOP framework and its application and collect ideas and input from the broader scientific community that can help guide and direct further the development of the framework (BOX2-Horizon Scanning/Rankingxy). Specifically, questions were solicited globally from individuals or groups representing government, academia, industry/business, and non-governmental organizations, via an online survey. The survey asked participants to propose questions that consider key outstanding challenges or limitations that must be addressed in order to realize the full potential of the AOP framework (Supplemental Material, S1). All questions submitted were considered during this exercise, regardless of the level of previous involvement in, or knowledge of the AOP framework by the participant. This provided the opportunity to identify and answer commonly asked questions about AOPs, as well as bring forth new thinking on the science needed for advancement of the framework.

Box2.

Clarifying What is Known

In order to enhance the utility of the AOP framework, it is critical to ensure that basic knowledge about the development structure, terminology, and current (or envisioned) uses are clearly described and understood by those who intend to develop, critically evaluate, or utilize AOPs. From the Horizon Scanning exercise, a number of questions that often emerge during AOP training sessions and/or discussions among scientists and colleagues involved in risk assessment/regulation were identified as “frequently asked questions (FAQs).” These FAQs could readily be answered by one of more of the co-authors of this paper who have been closely involved in the evolution of the AOP framework (BOX3-FAQ; Supplemental Material, S2). The answers to these FAQs provide a common starting point for those interested in the AOP framework to initiate discussions around new, forward-thinking approaches. Further, identification and clarification of common misperceptions will facilitate better communication when considering current and future directions of the AOP framework.

Box3.

A Path Forward

The Horizon Scanning survey was followed by a ranking exercise that identified priority questions regarding key needs and current limitations that may be considered for future initiatives aimed at advancing the science underlying the AOP framework and its regulatory applications. Briefly, the 340 questions collected from Horizon Scanning were initially binned into broad topic areas. Subsequently, an expert panel (co-authors) completed a series of best-worst scaling (BWS) comparisons (e.g., [16]) for each topic area, thus allowing the development of each expert’s rank-ordered list of candidate questions for each topic. The rankings facilitated the quick identification of issues that experts uniformly viewed as relatively important (or unimportant). Where there was a divergence in rankings, the panel was able to explore various opinions in depth and arrive at a consensus of the question’s importance. Consensus among experts regarding important questions suggested that four themes, if addressed, could enhance the utility of the AOP framework for research and regulatory decision-making: (1) AOP networks and their applications, (2) quantitative AOPs (qAOPs) and their applications, (3) regulatory use of the AOP framework, and (4) expanding awareness of, involvement in, and acceptance of AOPs to support aspects of predictive toxicology and regulatory decision-making (BOX4-Key questions). Below we briefly discuss priority questions and issues in each of these areas.

Box4.

AOP Networks and Their Applications

Horizon Scanning Participant Question, Quotation1 “What are the guiding principles and best practices for developing AOP networks and how do they differ (or remain the same) from those developed for linear AOP development?”

Individual AOPs are represented by discrete, pragmatic, and sequential units that begin with a single MIE and assume adequate perturbation with an individual stressor to “drive” the pathway all the way to a defined adverse outcome [17]. This construct can be useful on its own for a variety of applications (e.g., assessing the toxicity of single chemicals including read across, hazard screening, and predictive toxicological applications); however, chemicals may affect more than one MIE/KE and assessment scenarios typically involve complex mixtures including chemical and non-chemical stressors causing pathway perturbations that interact with multiple MIEs and/or shared KEs and key event relationships (KERs) that may culminate in one or many adverse outcome(s). As intended by the developers, the current principles of AOP development within the AOP-Wiki (www.aopwiki.org) support construction of AOP networks from simpler units of development. Horizon Scanning participants (BOX4A) emphasized that examples and guidance on how to develop, evaluate, and analyze those networks to predict the joint actions of multiple stressors or individual stressors impacting multiple AOPs are needed. (HorizonParticipants also asked for information about available tools and technologies for the construction and evaluation of networks. For example, there are potentially useful approaches from network theory that can facilitate development and aid in understanding the complexity of AOP networks [18–20], which could be explored further. Also, it may be that capturing additional types of information in KE and KER descriptions are needed for more effective use of AOP networks. For example, ontological descriptions of key events in the AOP-Wiki are now being integrated to link together common biology, even in cases where AOP developers may have used different terms to describe analogous events [21]. Additionally, in the case of multiple interacting pathways, the question of whether or not there are unique experimental designs or different structural augmentations that need to be considered for network development is a critical element to address. While these challenges have not been formally addressed to date, the growing number of published examples of AOP networks (e.g., [22–24]) make it tractable to begin probing these issues in a serious and deliberative manner. Examples and established approaches for analyzing and using AOP networks become particularly relevant when considering application of AOP networks in certain regulatory contexts.

Quantitative AOPs and Their Applications

Horizon Scanning Participant Question, Quotation2 “What are the key principles that can guide the development of qAOPs?”

Qualitative AOP descriptions provide a scientifically credible basis to link apical hazards of regulatory concern to specific pathway perturbations or biological activities (often measured in vitro). However, many risk assessment and regulatory activities also require the ability to define the exposure conditions (in terms of dose, duration, frequency, etc.) under which an adverse outcome will be observed and/or with what probability. Information captured in the “quantitative understanding of the linkage” section of the KER descriptions within the AOP framework provide the foundational information for addressing this desire for quantification. However, the descriptions alone rarely provide for the implementable prediction that a decision-maker may need. Quantitative AOPs are considered as a likely solution for addressing these needs.

Horizon Scanning Participant Question, Quotation3 “How do we create quantitative AOPs for use in risk assessment, particularly where multiple species need to be considered?”

Quantitative AOPs can be described in various ways, ranging from expert judgement-based scoring, requiring limited information, where elements of the AOP are weighted using expert opinion, to more probabilistic approaches, where statistical relationships exist between MIE/KE and the adverse outcome, to mechanistic approaches [25]. The more mechanistic approaches employ mathematical models or relationships of MIE, KE and KER (e.g., response-response relationships between KERs) to quantitatively predict the risk of an adverse effect given specified initial conditions (e.g., a set of exposure conditions) [26]. As with AOP networks, clarity of what constitutes a qAOP would provide value to an expanding audience of developers and users of these approaches and was an area highlighted by the questions submitted to the Horizon Scanning survey. Examples of well-developed qAOPs currently are somewhat limited, but scientifically-sound development of such actionable tools, anchored to AOP knowledge, is widely regarded as a critical step needed to unlock the full potential the AOP framework holds for predictive risk assessment [25–27]. In considering the development of qAOPs, guidance on how better to incorporate delayed toxicity, epigenetics, repeat exposures, and contributing factors such as organism life stage, sex, or other taxa specific physiological considerations, when applicable, are also needed. Such attributes may affect the magnitude or temporality of the biological responses, including the adverse outcome, and are therefore important to capture. Providing guidance is particularly important when considering the potential application of qAOPs.

The number and scope of questions falling within the theme of qAOP development and application reflect the importance of the topic to the long-term impact of the framework on risk assessment and regulation. Consistently, in the context of qAOP application, there was an identified need to systematically address and reduce uncertainties in extrapolating across species, and from individual to the population to community/ecosystem levels. Horizon Scanning participants were particularly interested in identifying what tools or resources are available or need to be developed to evaluate these challenges.

Regulatory Use of the AOP Framework

Horizon Scanning Participant Question, Quotation4 “How will it be determined if an AOP is fit for the purpose for which it has been developed?”

From questions raised during the Horizon Scanning exercise, respondents clearly recognized the need for guidance on how AOPs at different stages of development (e.g., putative, formal, quantitative, networks) and all degrees of peer review (e.g., no review, published, or OECD endorsed) should be applied to different research and regulatory applications. A common opinion derived from many previous AOP-focused workshops is that utility of the AOP framework will ultimately depend on the level of development and corresponding weight of evidence for a given AOP [28], including the degree of quantitative understanding of the relationships linking KEs. It also has been broadly accepted that the determination of whether an AOP is suitable for a particular application should be evaluated on a case by case basis, which has been termed as an evaluation of “fit-for-purpose.” However, what is lacking are transparent examples demonstrating how fit-for-purpose should be strategically and systematically evaluated for a particular research need or regulatory/risk assessment application. It is envisioned that AOP knowledge could be applied to a wide range of regulatory scenarios, including water quality criteria derivation, chemical-specific risk assessments, site-specific risk assessments, wastewater effluent and mixture assessments, and prioritization and screening. Therefore, examination of how the AOP framework would be pragmatically applied to such diverse regulatory scenarios would be highly desirable.

Questions of how an anticipated regulatory application may determine the level of AOP development and documentation needed, and how to best define acceptable uncertainty for a given application, also are issues that need to be considered when evaluating fit-for-purpose. In this context, and as highlighted by multiple survey questions, it is necessary to demonstrate how the practical synthesis of knowledge using AOPs aids decision making in present and future regulatory scenarios. In particular, case examples of how AOP constructs ranging from putative AOPs to AOP networks or qAOPs, and associated weight of evidence evaluations that can advance and/or improve decision-making for risk assessors and risk managers are needed.

Expand Awareness of, Involvement in, and Application of the AOP Framework

Horizon Scanning Participant Question, Quotation5 “What is the origin of the AOP framework?”

As exemplified by the FAQs and demographic information collected during Horizon Scanning, improved communication and technical exchanges surrounding the topic of AOPs is a priority across all sectors and geographic locales. Accurate and timely communication is particularly challenging, as guiding principles and best practices for AOP development continue to evolve as experience is gained and pragmatic approaches are identified for addressing challenges highlighted by AOP practitioners [29]. For example, recently it was recognized that indirect KERs were incorrectly being used to represent weak evidence or uncertainty in the relationship between upstream and downstream KEs, whereas the intended use was to use the term “indirect” to identify KERs with non-adjacent KEs (as opposed to “direct” or adjacent KEs). Therefore, AOP guidance documents coordinated through the OECD [30] currently are being revised to eliminate the confusion surrounding KER terminology. Timely communication of changes and clarifications in guidance would ensure consistency among AOP developers. Ongoing AOP development efforts, including training with relevant tools, AOP-focused courses (at universities or professional meetings and for regulators) and workshops, as well as other methods of outreach through digital or printed media must continue to be developed and distributed with expert input to inform, correct or clarify misperceptions, identify common challenges, and define evolving AOP, and associated weight of evidence, terminology and guidance (BOX5-In the Know). The Horizon Scanning exercise demonstrated that there is cross-sector interest in the development and use of the AOP framework, as well as a need to address key scientific challenges moving forward. Therefore, engagement of global stakeholders through outreach and communication is needed, and development of methods to do so effectively and efficiently were identified as a priority for the AOP concept to realize its full potential in research and regulatory realms.

Box5.

Horizon Scanning Participant Question, Quotation6 “How can the broader toxicological community be effectively engaged in the development of AOPs when there may be little professional motivation to do so?”

Coordinated development of AOPs through collaboration across sectors is essential. Priority questions of how to engage a broad range of communities and expertise, whether the most effective collaboration tools are available or could be improved upon, and how AOP development and review can be further streamlined and accelerated (perhaps computationally) were raised through the Horizon Scanning effort. The idea of crowd-sourcing AOP development as a means for collaboration and collection of information in a consistent format were motivations in developing the AOP knowledgebase. This knowledgebase is a combination of four independently developed platforms for capturing, reviewing, browsing, and communicating publically available AOPs (http://aopkb.org/background.html). As a component of the AOP knowledgebase, the AOP-Wiki (https://aopwiki.org) was designed to encourage cooperative AOP development with the intent that multiple experts could communicate about, comment on, and/or contribute to development at all stages, although it is difficult to quantify how much this collaborative approach has been used to date. However, this type of cooperation among scientists is needed to accelerate the pace and quality of AOP development, especially when considering the daunting task of assembling complex AOP networks or qAOPs. This is consistent with survey questions pertaining to how to best engage and incentivize participation in AOP development across sectors and areas of expertise. Finally, another key consideration for routine use of AOPs for regulatory applications or in research is aimed at increasing confidence in the framework which can be supported through increased participation in development and review. Addressing the challenges identified through the Horizon Scanning exercise should be considered a starting point in increasing the support for and confidence in the AOP framework.

Continued Evolution of the AOP Framework

The AOP-focused Horizon Scanning and question prioritization exercise identified global questions and key challenges regarding the applicability and readiness of the AOP framework to address current regulatory and scientific needs. Additionally, it identified FAQs that allowed for the clarification of common misperceptions, which ultimately will improve understanding of the AOP framework and discussions surrounding its domains of applicability. The broad themes identified above were used to develop workgroup themes and specific charge questions to be addressed at an upcoming (April 2017) Society of Environmental Toxicology and Chemistry (SETAC) Pellston workshop titled “Advancing the Adverse Outcome Pathway Framework – An International Horizon Scanning Approach.” It is further anticipated that the questions identified and prioritized by this Horizon Scanning activity also will help inform other workshops and ultimately be utilized by the global scientific community as they anticipate research and regulatory needs (Supplemental Material, S3).

Supplementary Material

S1.

Screen-shots of the Horizon Scanning survey used to collect participant information and solicit questions.

S2.

Complete Answers to all Frequently Asked Questions

S3.

Data collected from Horizon Scanning Survey, including all submitted questions with rationale.