The use of new approach methodologies for the environmental risk assessment of food and feed chemicals

Abstract

New Approach Methodologies (NAMs) provide tools for supporting both human and environmental risk assessment (HRA and ERA). This short review provides recent insights regarding the use of NAMs in ERA of food and feed chemicals. We highlight the usefulness of tiered methods supporting weight-of-evidence approaches in relation to problem formulation (i.e., data availability, time, and resource availability). In silico models, including quantitative structure activity relationship models, support filling data gaps when no chemical property or ecotoxicological data are available, and biologically-based models (e.g., toxicokinetic-toxicodynamic models, dynamic energy models, physiologically-based models and species sensitivity distributions) are applicable in more data rich situations, including landscape-based modelling approaches. Particular attention is given to provide practical examples to apply the approaches described in real-world settings. We conclude with future perspectives, with regards to the need for addressing complex challenges such as chemical mixtures and multiple stressors in a wide range of organisms and ecosystems.

Introduction

Modern environmental risk assessment (ERA) for food and feed chemicals aims to assess risks resulting from exposure to chemicals, physical or biological hazards on a range of relevant organisms at different levels of biological organisation (e.g., individuals, populations, communities, ecosystem, landscape) [1]. The first step is problem formulation, most often provided by risk managers, that sets the scene to tackle the ERA question. This includes hazard(s) and organism(s) to be assessed, protection goals and the definition of an action plan. The latter considers available data, resources and timelines to carry out the ERA through several steps and the application of tiered approaches [2-4]:

• Exposure assessment investigates sources, extent and frequency of exposure, environmental fate and occurrence to quantify environmental concentrations and exposure metrics for a given taxon.

• Hazard identification and characterisation aim to derive hazard metrics, as environmental standards, from ecotoxicity data (i.e., reference points). Reference points include no-observed-effect-concentration, effect or lethal concentrations/dose at individual, population or ecosystem levels.

• Risk characterisation brings exposure and hazard metrics together to conclude as to whether protection goals are reached and sufficient protection for the organism(s) is achieved. In the case protection goals are not met, higher tiers may be applied, or risk managers may consider risk mitigation measures.

Recently, the vision and goals of the European Green Deal have placed the ERA process at the forefront as a key step to achieve environmental protection and sustainability. Regulatory and advisory bodies worldwide contribute actively through prospective and retrospective ERA, monitoring, development of methods and research activities [5,6]. At the European Food Safety Authority (EFSA), ERA aims to provide scientific advice to risk managers for regulatory and EU market approval of pesticides, feed additives and genetically modified organisms. In addition, a goal of ERA is to contribute to the development of harmonised methods accounting for biodiversity and ecosystem services, protection goals, recovery, non-target organisms and endangered species [7-9].

Recently, New Approach methodologies (NAMs) have come into the ERA arena to integrate mechanistic understanding of toxicity, exposure and risk using predictive in silico and in vitro models, OMICs technologies, integrated exposure-effect models at individual/population level, as well as landscape models. This concise overview focuses on the use of such NAMs in ERA of food and feed chemicals, highlights the use of a systems-based approach integrating information from different levels of biological organisation and concludes with the future challenges for next generation ERA (NGERA).

Exposure tools for ERA: from environmental fate to environmental concentrations

Identification and characterisation of chemical properties, environmental fate and concentrations of substances as well as food or feed consumption patterns are the key pillars for environmental exposure assessment at the taxon, population and landscape level. First, use patterns defining potential sources and fate properties are combined with environmental characteristics to assess the behaviour of the chemical [1]. In this context, the Organisation for Economic Co-operation and Development (OECD)’s Harmonised Templates (OHTs) provide means to standardise data formats for physico-chemical properties and environmental fate in standard data formats including OHTs 24 to 40 and 401. Further details are available under: https://www.oecd.org/chemicalsafety/testing/oecdguidelinesforthetestingofchemicals.htm.NAM.

In addition, in silico NAMs, such as quantitative structure—activity relationship (QSAR) models, allow robust and swift predictions of physicochemical properties (e.g., water solubility and partitioning parameters, degradation half-lives) in addition to contributing to fate models for environmental distribution. The latter, in turn, can be integrated with multi-pathway exposure models describing chemical transfer into soil, water and food ingested by organisms [10,11].

Environmental concentrations/occurrence are measured using a wide range of analytical methods including mass spectrometry (MS), gas Chromatography—MS/MS, and non-targeted analytical techniques, all of which are highly relevant to pesticides, feed additives and contaminant environmental monitoring [12,13]. In silico NAM models are also available to predict environmental concentrations such as the equilibrium criterion model as an evaluative multimedia mass balance model [11].

For a given aquatic or terrestrial organisms, data for environmental fate, occurrence and consumption patterns are combined to determine exposure (external dose) in a given medium. These, in turn, contribute to lead to internal dose estimation through bioavailable fractions considering the significant inter-species differences that affect both toxicokinetic (TK) and toxicodynamic (TD) of chemicals.

Toxicokinetics and toxicodynamics: moving from environmental concentrations to internal dose and effects at different levels of biological organisation

Available open-access databases and in silico modelling platforms

A number of scientific advisory bodies have integrated large (eco)toxicological datasets into open access databases for key terrestrial and aquatic test species as well as non-targeted species. Examples are highlighted in Table 1 [14-17].

Table 1.

Examples of open access databases with (eco)toxicological datasets.

Database	Brief description	Website
OECD e-chem portal	Global searchable portal for information on chemical substances including hazard and risk information from national, regional and international agencies	https://www.echemportal.org/echemportal/index.action
US-EPA CompTox Chemicals Dashboard	Large database (~1.2M substances) containing data on physicochemical properties, fate and transport, hazard (TK and TD), exposure and in vitro bioactivity data for over. It integrates data from the ECOTOXicology Knowledgebase (ECOTOX) with data for aquatic and terrestrial organisms as well as wildlife for thousands of chemicals.	https://comptox.epa.gov/dashboard https://cfpub.epa.gov/ecotox/
EFSA’s OpenFoodTox	Database providing data for over 5600 food and feed chemicals (). Such hazard data are available for active substances in pesticides and feed additives in key terrestrial and aquatic test species.	https://zenodo.org/record/1252752#.Xd_5G-hKjD4
ECHA’s database	Database from industry-submitted Registration, Evaluation, Authorisation and Restriction of Chemicals Regulation (REACH) dossiers. Covering a range of hazardous properties, classification and labelling, and information on safe use for chemicals manufactured and imported in Europe.	https://echa.europa.eu/information-on-chemicals

These substance databases and their associated structures have been the basis to develop many in silico models and NAM modelling platforms predicting multiple TK and TD endpoints, particularly for data poor compounds, and allow for combining of the results from multiple models using a weight of evidence approach [18]. Recent examples include the OECD QSAR Toolbox, the National Institute of Environmental Health Sciences (NIEHS)’s OPERA and Integrated Chemical Environment suites as well as the VEGA Hub [10,19,20].

Toxicokinetics and toxicodynamics: bridging exposure towards internal dose and toxicological impacts on individual organisms, populations and ecosystems

TK data allow for the bridging of exposure as external dose to internal dose using TD data of relevance to ERA such as mortality, reproduction or sub-lethal effects. For a given compound, TK parameters can reflect acute exposure, such as maximum plasma concentration, or chronic elimination and persistence (i.e., half-life, bioaccumulation factor (BAF) or bio-concentration factor (BCF) and clearance). Since the persistence of chemicals has an impact on toxicity, internal dose predictions provide useful metrics to bridge internal dose with TD data [21,22]. NAM tools to predict TK parameters include biologically-based models of chemical or generic nature such as TK models or physiologically-based kinetic models (PB—K) through so-called forward dosimetry [1,21-23]. In addition, quantitative in vitro-to-in vivo (QIVIVE) extrapolations models and PB-K models can also allow recalculation of exposure from in vitro experiments and monitoring of data, respectively, through the so-called reverse dosimetry [16,24-26]. Recently, an OECD guidance document has illustrated the use of PB-K models in ERA with a number of case studies using generic models for farm animals and fish [16,25-27].

Sound TK-TD modelling aims to depict the relationships between internal exposure and dose—response relationships, with tiered approaches being applied depending on regulatory frameworks, available data, time and resources. Broad applications include ERA for pesticides, feed additives, environmental contaminants and industrial chemicals [28]. At low tiers, ERA may use simple NAMs such as QSARs or dose—response models. In contrast, more refined taxon-specific NAM-based models are available for higher tier assessments and range from general unified threshold (GUTs) and dynamic energy budget (DEB) models and population-based models for given taxa to species sensitivity distributions (SSDs) to compare toxicity across species for the whole ecosystem [29]. These models have been applied particularly in aquatic ERA [30-32]. These NAM-based models have the major advantages of guaranteeing a better statistical foundation to derive parameter estimates while taking into account variability and uncertainty as illustrated in the TK tool MOSAIC [22,33,34].

Moving towards landscape modelling for a systems-based approach in environmental risk assessment

Advances in gathering and using environmental and geospatial data, including “big data” approaches, offer possibilities for next-generation landscape-based ERA. These models introduce spatial representations and consider landscape structure and characteristics of land use. As illustrated in Figure 1, landscape modelling is based on the geo-location of emission sources and receptors, and the adaptation of the exposure and effect assessments to local landscape characteristics. For food and feed chemicals including agrochemicals, this approach offers two key improvements [35]:

Figure 1.

A schematic representation of how landscape modelling can introduce spatial representation in risk assessment. Key factors are geo-located (emission sources, exposed receptors). Adaptations to the site conditions of both exposure and effects can be considered.Vectors downloaded from Vecteezy.com.

1. Estimation of aggregate (one chemical) and combined (multiple chemicals) exposures related to uses in a range of crops and periods.

2. Consideration of agricultural management practices and local relevance of ecosystem functions/services in ERA.

Landscape modelling has been mostly applied for ERA of pesticides to terrestrial organisms [28]. Some models are spatially explicit, based on actual characteristics of one or several selected areas [36]. Other models allow the user to select real or hypothetical landscapes [37]. While current regulatory use is limited [31], developments in this area are paving the way for more informative ERA [38] and integrated system approaches for pesticides [39]. In addition to terrestrial vertebrates [36,37], significant developments are on-going in relation to the bee health area and risk for pollinators [40].

Putting it together using NAM-based tools: a systems-based approach for environmental risk assessment

For each step of the ERA process, in vivo (eco)toxicological data and NAM-based data offer valuable contributions to introduce systems-based approaches for data-poor or data-rich chemicals as proposed in Figure 2.

Figure 2.

Systems-based approach for Environmental Risk Assessment highlighting the steps and role of NAM-based data and at different levels of biological organisation (modified from Astuto et al., 2022). Legend: TK: Toxicokinetics, TD: Toxicodynamics, -Standard Methods, -NAM-based methods.

Exposure assessment for a given taxon requires use patterns, environmental fate properties and occurrence of chemicals using available data or in silico predictions. These can then be used to characterise landscape exposure per ecotype when combined with landscape characteristics. For hazard identification and characterisation in target and non-target species, reference points as environmental standards can be derived on an internal dose basis using exposure, TK and TD data based in vivo or NAM-based data, at the individual, population or ecosystem and landscape level for a given crop. Interspecies differences and other drivers such as coexposures and biological hazards can also be integrated in the scheme. Finally, risk characterisation metrics are then expressed as population effects and recovery on a crop and landscape basis, resulting in quantification of landscape impact on ecosystem services.

In practice, for data-poor chemicals, this can be achieved through the use of historical data from data-rich chemicals and NAMs, including in silico models. This approach can already be applied, with some concerted efforts, to single chemicals in a regulatory context such as pre-market authorisation of pesticides or feed additives or post-market monitoring of regulated compounds and environmental contaminants.

Future perspectives: addressing complex challenges in the modern world

The use of NAMs in ERA for food and feed chemicals has been highlighted for NGERA with a systems-based approach allowing the integration of chemical-specific information at different levels of biological organisation encompassing single-organism at the individual level to multiple-species population level all the way to ecosystem services using landscape modelling [37,41]. For population and communities, a challenge is how to incorporate ecological interactions including trophic relationships, potential species competition for food resources or habitat occupancy as well as space and time dimensions to assess risk at the landscape level [42]. Another major challenge is the consideration of exposure to multiple chemicals and multiple stressors in species of ecological relevance including nutrition, varying environmental factors and habitats and pathogens. The systems-based approach highlighted here can support the development of relevant approaches and NAM-based models to move towards exposome and multi-stressor approaches in ERA (Figure 3). These approaches have recently been illustrated for the RA of multiple stressors, in bees and are currently under development for farm animals, wild birds, reptiles and amphibians [40,43,44]. Ultimately, addressing such challenges will be particularly relevant for depicting the balance between chemical exposure, agriculture, food production, security and sustainability to match the requirements of the Green Deal and support the One-Health paradigm [6,45]. Finally, cooperation and training between regulatory bodies, risk managers, academia, industry and non-governmental bodies is becoming critical as a means to share data, avoiding effort duplication and supporting harmonisation and implementation of NAM-based tools and guidance into ERA [1,23].

Figure 3.

Modelling multiple chemicals and multiple stressors in Environmental Risk Assessment.

Data availability

No data was used for the research described in the article.