Guidance on risk assessment of the application of nanoscience and nanotechnologies in the food and feed chain: Part 1, human and animal health

1.2.1 Interpretation of the Terms of Reference

Dermal and inhalation exposure were added to cover the main routes of exposure to nanopesticides and feed additives.

Environmental risk assessment will be addressed in a separate document (Part 2) as requested in the Terms of Reference provided by EFSA.

1.2.2 Definition of nanomaterial

The International Organization for Standardization (ISO) has defined nanomaterial as a material with any external dimension on the nanoscale (‘nano-object’) or having an internal or surface structure in the nanoscale (‘nanostructured material’) (ISO, 2015). In particular, a nano-object is defined as a discrete piece of material with one, two or three external dimensions on the nanoscale. ‘Nanoparticles’ are nano-objects with all external dimensions on the nanoscale, where the lengths of the longest and shortest axes do not differ significantly. If the dimensions differ significantly, typically by more than a factor of three, other terms, such as ‘nanofibre’ (two external dimensions in the nanoscale) or ‘nanoplate’ (one external dimension on the nanoscale) may be preferred to the term nanoparticle. In turn, a ‘nanostructured material’ is defined as a material having internal or surface nanostructure, i.e. a composition of interrelated constituent parts in which one or more of those parts is a nanoscale region. ‘Nanoscale’ is defined as ranging from approximately 1 to 100 nm (ISO, 2015).

According to the ISO nanotechnologies vocabulary, which can be freely consulted on www.iso.org/obp, ‘engineered nano-object’ is defined as a nano-object designed for a specific purpose or function, ‘manufactured nano-object’ as a nano-object intentionally produced to have selected properties or composition and ‘incidental nano-object’ as a nano-object generated as an unintentional by-product of a process (ISO, 2015).

Size is the key parameter for the identification of a nanomaterial. All nanomaterials occur with a size distribution, i.e. the constituting entities do not all have the same size. Often particulate materials comprise particles with lengths both below and above 100 nm. Owing to the reactivity of nanoparticles, mainly related to their high surface free energy, larger clusters (‘secondary particles’) often result from agglomeration and/or aggregation of constituting primary particles. In some cases, the size distribution of manufactured nanomaterials covers a rather wide length range.

The European Commission issued a Recommendation for a definition of a nanomaterial in 2011 to provide a common basis for regulatory purposes across most areas of EU policy (currently under review4). The provisions of the recommended definition include a requirement for review in the light of experience and of scientific and technological developments. The European Commission is expected to conclude this review in 2018. If this recommended definition (or any update of it) were to be embedded in the food law, it would provide further information on whether or not a material should be regarded as a nanomaterial in the context of any of the food regulations. According to that recommended definition, ‘nanomaterial’ means a natural, incidental or manufactured material containing particles in an unbound state or as an aggregate or as an agglomerate and where, for 50% or more of the particles in the number–size distribution, one or more external dimensions is in the size range 1–100 nm. In specific cases and where warranted by concerns for the environment, health, safety or competitiveness the number–size distribution threshold of 50% may be replaced by a threshold between 1 and 50%.

For the purposes of the recommended definition, ‘particle’ means a minute piece of matter with defined physical boundaries, ‘agglomerate’ means a collection of weakly bound particles or aggregates where the resulting external surface area is similar to the sum of the surface areas of the individual components, and ‘aggregate’ means a particle comprising of strongly bound or fused particles. In addition, it is specified that fullerenes, graphene flakes and single wall carbon nanotubes should be considered as nanomaterials even though one or more external dimensions are below 1 nm.

The current recommended definition is used as the reference for determining whether a material should be considered as a ‘nanomaterial’ for legislative and policy purposes in the EU. According to the recommended definition the criterion is solely the size of the constituent particles of the material, without regard to hazard, toxicokinetics or risk. Although this Recommendation is currently under review, and has not yet been adopted under the relevant regulatory frameworks, the SC advises taking this and any future reviews into consideration when assessing the safety of materials consisting of small particles.

From a risk assessment perspective, it is essential to point out that size-dependent properties and biological effects that are of potential concern for human health, specifically toxicokinetic behaviour and particle–cell interactions, are not rigidly related to specific size thresholds. They depend on dose and may continue to occur even when the particles constituting the nanomaterial have a size well above 100 nm. Furthermore, whereas physical, chemical and biological properties of materials may change with size, there is no scientific justification for a single size limit associated with these changes that can be applied to all nanomaterials (SCENIHR, 2010). Therefore, potential risks arising from specific properties related to the nanoscale have to be assessed focusing on such properties and potentially related hazards, which may be independent of the proportion of particles constituting the material with a size below 100 nm. In line with the conclusions of SCENIHR (2010)1 and the EFSA Scientific Committee (2011a), the EFSA Scientific Committee reiterates that not all nanomaterials have new hazard properties compared with larger sized counterparts and that therefore a case-by-case assessment is necessary.

This Guidance emphasises size-related properties associated with specific hazards and the need to provide relevant nanospecific information in order to perform a risk assessment. In addition to the size of the material, a number of other properties that may be associated with adverse health effects such as chemical composition, morphology (in particular, aspect ratio), surface properties, crystallinity, solubility and others) should also be taken into account.6 Particle properties that significantly alter levels/pathways of uptake and/or affect the mobility and persistence in the body are of high relevance. Increased bioavailability of nanomaterial compared to the corresponding conventional form related to particular surface properties resulting from, e.g. use of specific coatings and encapsulation, should be flagged for risk assessment according to the present Guidance.

1.2.3 Definition of engineered nanomaterial

According to the Novel Food regulation, ‘For consistency and coherence purposes, it is important to ensure a single definition of engineered nanomaterial in the area of food law’.

The use nanomaterial as a pesticide and the use of the term ‘nanopesticide’ herein is explained under the sector specific information of Appendix 96.

4.2.1 Parameters

The characterisation of the material under investigation is essential to unambiguously define its identity. Similar to conventional chemicals (e.g. food additives), names, identifiers and a number of physicochemical parameters need to be measured. In addition, a broader range of parameters needs to be addressed for nanomaterials, relating on the one hand to material identity, and to properties that may be of biological/toxicological relevance on the other.

Owing to the current gaps in knowledge relating to properties, behaviour and effects of nanomaterials, it is difficult to identify a definitive shortlist of those parameters that can adequately describe a nanomaterial in terms of both physicochemical and toxicological aspects. Different international expert committees and working groups have considered certain parameters important for safety assessment of nanomaterials. These are presented as a list of parameters to be reported in Table 1. This list is not definitive, however, and has to be changed in future to include more, less or different parameters that might be added with the advancement of scientific insights as well as legislative developments.

The parameters in Table 1 have been derived from the reports published by the Scientific Committee on Emerging and Newly Identified Health Risks (SCENIHR, 2009); the OECD Working Party on Manufactured Nanomaterials in its exploratory project on ‘Safety testing of a representative set of nanomaterials’ and the revised version of its ‘Guidance manual for the testing of manufactured nanomaterials’ (OECD WPMN, 2009a, 2010); the International Organization for Standardization; the EU's Scientific Committee on Consumer Safety (SCCS, 2012); the ProSafe16 project (European Union H2020 project ProSafe, 2015); the ECHA Guidance on the preparation of registration dossiers that cover nanoforms (ECHA, 2017b); the ECHA Appendix R.6-1 for nanomaterials applicable to the Guidance on QSARs and Grouping of Chemicals (ECHA, 2016); and a recent publication by DeLoid et al. (2017a).

In some instances, not all of the parameters listed in Table 1 (and Table B.1 in Appendix 87) may be relevant for a given material as determined by its composition, function, purpose and/or intended use. In such cases, justification should be provided for the characteristics that are not determined or provided, or to explain why they were not deemed applicable to a particular nanomaterial.

Currently, no generally accepted systematic nomenclature exists for nanomaterials. ISO TC 229 has drafted a series on vocabulary and terminology of NMs (ISO 80004 series). The CODATA-VAMAS Working Group on the Description of Nanomaterials has published a ‘Uniform Description System for Materials on the Nanoscale’ (CODATA-VAMAS Working Group on the Description of Nanomaterials, 2016) that proposes in detail the information that should be supplied to describe a nanomaterial in the best possible and most unambiguous way. The SC suggests that applicants follow the schemes proposed by ISO and the CODATA-VAMAS Working Group when naming a nanomaterial.

In some instances, however, the material may be too complex to define precisely in terms of chemical composition and stoichiometry. Examples could be complex iron oxide hydroxides or polymers. Other examples include materials already authorised for use in FCMs such as a ‘butadiene, ethyl acrylate, methyl methacrylate, styrene copolymer (either not crosslinked or crosslinked with divinylbenzene or 1,3-butanediol dimethacrylate) in nanosized particles, (FCM substance Nos 998, 859 and 1043)’. In simpler materials (e.g. metal oxides), the stoichiometry in the surface layer may also differ from the core of the particle. In these cases, the material should be described as exactly as possible. In any instance, the elemental composition must be given (e.g. the empirical formula) and additional information on the starting material, the reaction process(es) and the intended composition should be provided.

For nanomaterials consisting of multicomponent particles, the overall material should be described together with the individual components. In the case of a nanomaterial consisting of a mixture of different types of particles, each component should be described individually according to Table 1, and the ratio of all components in the mixture should be provided. The structure of the particles should also be described as exactly as possible. This includes information on the distribution of individual components in the particle, e.g. homogeneous mixture, core/shell and coatings.17 Coating is a thin layer of a component that covers completely or partially the surface of a particle and is strongly bound (either chemically or physically) to the surface. Stabilisers (or dispersants) are substances that are added to a dispersion of nanomaterial to prevent agglomeration, aggregation or sedimentation. They are not seen as a part of the particle and should be reported under ‘formulation’. Substances strongly bound to the particle surface for stabilisation purposes should be reported under ‘Surface (chemical) composition’ or as coating (when covering the entire particle).

Changes in manufacturing process(es) can not only lead to significant differences in the physicochemical and morphological characteristics of nanomaterials between different batches, but may also introduce new/different impurities and residual materials. Furthermore, for some materials, fundamentally different production processes are in place (e.g. for pyrogenic vs. precipitated silica as described in Fruijtier-Pölloth, 2012; sulfate or chloride process for converting titanium ores into TiO₂) that largely define the surface and crystallographic structure, and thus the particle properties. It is therefore important to provide a description of the manufacturing process.

Table 1 is also meant to be applicable for multicomponent materials (e.g. core-shell or coated particles). Table 1 is therefore structured into a Section for general and ensemble information on the overall material and a Section on detailed chemical and physical information for its individual components. In the case of a monocomponent particle (e.g. uncoated TiO₂), information has to be provided for component 1 only. Examples for component 2 information are details in Table 1B of Appendix 87. Some of the general parameters of Table 1 (and Table B.1 in Appendix 87) might already be required under the sectorial legislations. Appendix 89 provides a list of corresponding techniques for each parameter.

Table 1: Descriptors and parameters on what data are to be provided for characterisation of pristine material, together with hypothetical examples (not food related, not consistent, data for illustrative purposes only). For clarity, the Table is divided in the different Sections: Table 1 for Information on the overall material, Table 1B for Information on the chemical components, and Table 1 for extrinsic properties

Table 1A. Information on the overall material

Item Parameters (incl. specification ranges)	Explanation	Example
Name	The name used in the submitted application. This could for example be the name of the nanomaterial	Ti-Max
Description Short description of the material	Provide a brief description of the material	Nano grade titania coated with a protective silica layer
Intended use	Describe the foreseen use and function of the material	UV protection to be incorporated in food contact materials
Material composition and purity Relative amounts of components and impurities (in mass %)	Relative amount of the constituents in mass %, as well as chemical identity of any impurities and their relative amounts in mass % should be provided	TiO2 97.1% ± 0.3%a SiO2 2.8% ± 0.1% purity: 99.9% impurities: Fe2O3 0.1% ± 0.02% Specifications composition: TiO2 97.0% ± 0.5% SiO2 3.0% ± 0.2% Purity ≥ 99.7%
Elemental composition Empirical formula of the complete material or relative amounts of element (in mass %)	The relative elemental composition of the particle should be provided as the simplest positive integer ratio of atoms present in the material. Alternatively, the relative mass amounts of the contained elements may be provided	Ti26SiO54 Ti 58.23% (m/m) O 40.32% (m/m) Si 1.31% (m/m)
Particle size Agglomeration or aggregation state Mean and median diameter [nm] graphical diagrams of size distribution	Data on primary and secondary (agglomerates and aggregates) particle size, number-based size distribution and mass-based size distribution of the material should be provided as measured by more than one independent technique, one being electron microscopy (EM) if the measurement is feasible (cfr. current publications on critical issues of EM, e.g. drying, artefacts). If EM cannot be applied, the use of a different imaging technique is suggested. Information on the used characterisation techniques and methods (e.g. which techniques, instruments, settings, SOPs, method performance characteristics, data conversion) should be provided. Data should be provided both as median particle diameter (x50 in nm), with an indication of the width of the distribution (e.g. standard deviation, in nm) and with an estimate of the uncertainty of the median diameter (± expanded uncertainty, confidence level 95%, in nm). Together with the size distribution information on the lower and upper cut-off limits for the calculation of the relative amount of particles has to be provided. Data obtained with a particle counting technique (such as EM) should be provided as number-based size distributions. Data obtained with other techniques (such as centrifugal liquid sedimentation (CLS) or dynamic light scattering (DLS)) should be provided in the original metrics as produced by the technique (e.g. intensity, volume- or mass-based). In these cases, a conversion to the number-based size distribution must also be provided, including information on the algorithms used for conversion and the associated uncertainty. Most light scattering based techniques (incl. CLS and DLS) provide light intensity-based distributions, which can be converted into their equivalent volume- or mass-based distributions using Mie light scattering theory. As this step can introduce considerable errors on the results due to the unknown refractive index of nanoparticles, the parameters used in the conversion must be reported in detail. Reporting of the original light intensity-based results can help to assess the reliability of the results. For each material, at least two graphs showing particle size distributions shall be shown: one with the relative number versus size (continuous graph or histogram) and one with number-weighted sum function (cumulative numbers)	Primary particles: TEM data: median diameter x50 = 85 nm (uncertainty = 5 nm, 95% confidence level), width of distribution: SD = 15 nm mean diameter: 89 nm (uncertainty = 6 nm, 95% confidence level), width of distribution: SD = 13 nmCLS data: median diameter x50 = 97 nm (uncertainty = 19 nm, 95% confidence level), width of distribution: SD = 8 nm) mass-based arithmetic mean diameter: 105 nm (uncertainty = 10 nm, 95% confidence level), width of distribution: SD = 13 nm) [plus diagrams as above] Aggregates: [provide size data in the same manner as for primary particles, see above] Specifications size: median diameter 85 nm ± 5 nm
Particle shape Description of the shape, porosity, aspect ratio, EM image of the nanomaterial	Information should be provided on the particle shape, aspect ratio and whether or not the material is porous. This should also include appropriate EM images to support the description. For powders, information on porosity can be obtained from gas adsorption measurements	Irregular particles, aspect ratio 1 to 3, non-porous
Structure Description of the structure, including (relative) thickness of structural elements	Spatial distribution of the components (e.g. homogeneous mixture, core–shell, surface coating) should be provided. A graphical sketch for non-homogeneous particles should be provided to demonstrate the schematic distribution if applicable. The sketch should reflect schematically the shape of the particles. Information should be provided on any surface coatings or shells in terms of coating or shell material and the proportion of the coating or shell material in relation to the mass of the nanomaterial	TiO2 particles with a surface coating of silica. Thickness of the coating 1.8 nm (± 16% (g/g))
Surface chemical composition Description of the composition of the groups or coatings on the particle surface	Information on chemical characteristics of the particle surface, e.g. the components bound to the surface, the presence of functional groups (e.g. carboxy, amino, hydroxy). Information should also be provided on any surface contamination	Hydrophilic acidic silica surface, free –OH groups
Production process Name of the production process of the material	The production process used to prepare the entire nanomaterial (i.e. not of the individual components in cases of multicomponent particles) should be described as it can have a significant effect on the properties of the nanomaterial	SiO2 precipitation on dispersion of wet-chemically synthesised TiO2 particles
Surface area MSSA [m2/g] VSSA [m2/cm3]	Where appropriate (for powder materials) data on mass and volume specific surface area of the material should be provided The conditions under which the measurements took place have to be reported	MSSA 15 m2/g (via BET according to ISO 9277) VSSA 65 m2/cm3 (via BET according to ISO 9277 and assuming a density value of 4.1 g/cm3)
Surface charge Zeta potential [mV]	Zeta potential values along with the conditions under which measurements were made (e.g. pH, ionic strength) should be provided	Zeta potential: −26 mV (deionised water, pH 8), Isoelectrical point: pH 2.2, method: electrophoretic light scattering according to ISO 13099-2
Appearance Description	Describe the appearance, e.g. ‘white powder’	White powder
Melting point m.p. [°C]	Provide the melting point of the nanomaterial	1,840°C
Boiling point b.p. [°C]	Provide the boiling point of the nanomaterial	2,900°C
Density Density [kg/m3]	Information on the density (specify type of density, e.g. bulk, pour) of nanomaterial should be provided	Bulk density: 4.1 g/cm3
Porosity fraction of the volume of voids over the total volume [%]	Information on the porosity of nanomaterial should be provided	Non-porous
Dustiness	Provide the dustiness for powder material (e.g. EN15051)	According to DIN EN 15051 B: Wr: 280 mg/kg Wi: 13,200 mg/kg
pH	The pH value of a dispersion of the nanomaterial should be provided along with description of the conditions under which the measurement was carried out	pH 5.8, 10 g/L, 20°C
Formulation Formulation medium Dispersing agents (stabilisers) Auxiliaries Concentration of nanomaterial in dispersion	Description should be provided to indicate the form in which a nanomaterial is present in a formulation, e.g. powder, dispersion. Information should also indicate other material(s) with which the nanomaterial may have been mixed/formulated. This should include information on any dispersants/stabilisers and other auxiliaries (e.g. preservatives, processing aids, etc.) used. The concentration of the nanomaterial in the mixture should be provided, in terms of both mass (g/kg) and particle number (n/kg), as well as the mass of the material as present in its ionic form	Dry powder

• ^a The measurement uncertainty should be reported as detailed in Section 29.

4.2.2 Specifications and representativeness of the test material

In view of the potential significant differences in the physicochemical characteristics of nominally the same nanomaterial resulting from variations in the manufacturing process, or from being produced by different manufacturers, or by ageing effects (e.g. agglomeration/aggregation, sedimentation) a detailed and comprehensive proposed specification for the pristine (as produced) nanomaterial intended to be used in food/feed should be provided by the applicant. The proposed specification should provide the acceptable range for each physicochemical parameter in view of the batch-to-batch variation and ageing effects. This information will be used by the risk assessor to decide whether or not the batch(es) used in the toxicity texting could be considered representative for risk assessment of the use in food/feed. More specific guidance on the number of batches and batch-to-batch variation is provided in the relevant guidance for conventional materials (e.g. EFSA ANS Panel, 2012). No more specific guidance on ageing (nor on homo/hetero agglomeration/aggregation) can be provided as little is known about complex transformation and the analytical tools are to be developed.

As mentioned in Section 8, this Guidance also applies to materials that are not engineered as nanomaterial but contain a fraction of particles, less than 50% in the number–size distribution, with one or more external dimensions in the size range 1–100 nm. This is expected to be the case of manufacturing processes for powdered or particulate food chemicals that typically result in materials with a range of sizes. Even where the median size of the particles is generally significantly greater than 100 nm, a small fraction is always expected to be present with at least one dimension below 100 nm. For re-evaluations of authorised materials (i.e. food and feed additives), the test requirements stipulated in current EFSA guidance documents and European Commission guidelines for the intended use in the food/feed area apply in principle to food chemicals containing a fraction of particles with at least one dimension below 100 nm and adequately conducted toxicity tests should detect hazards associated with such food chemicals, including their nanoparticulate fraction. In such cases of re-evaluation, however, it is also essential that thorough information on the size distribution of the material is provided and the assessment should consider whether the material as a whole retains properties that require risk assessment according to this Guidance. For example, EU specifications for TiO₂ (E 171) should include a characterisation of particle size distribution present in TiO₂ (E 171) used as a food additive. The measuring methodology applied should comply with this Guidance.

4.2.3 Techniques and methods

Care should be taken in the selection of characterisation techniques, the evaluation of results and their documentation. It is known that the results obtained from different particle size measurement techniques may differ because they address slightly different physical parameters, e.g. hydrodynamic diameter vs. geometric diameter (Domingos et al., 2009). In other words, particle size analysis methods produce method-defined or procedurally defined size values. As a result, the best suited technique depends on the physical and chemical properties of the nanomaterial, as well as on the intended use of the size values. Moreover, particle size distribution data are usually reduced to an average value. There are differences in the types of averaging between methods, which can amplify the already existing differences between methods.

In a comparison of most currently available techniques, Babick et al. (2016) demonstrated that significant differences were observed in the results for a number of industrial materials. The observed differences were mostly related to the technique-specific way of size determination since all other influences on the result (e.g. data handling, sample preparation, differences in test materials) were minimised by the study design. As an example, the use of some analytical methods, such DLS, may not be optimal for measuring nanomaterials that have a low refractive index (such as nanosilica, polymer encapsulates, etc.), because that can impact the intensity with which the light is scattered. Therefore, although nanoparticles of less refractive materials would be detected by DLS, the limit of detection (LOD) in terms of particle size would be high. This means that size measurements of such materials by DLS will likely be skewed towards measuring larger sized particles, and particles and agglomerates/aggregates in the lower range of the nanoscale might not be measured accurately. For these reasons, it is required that the size parameter should always be measured by at least two independent techniques, one being electron microscopy. If electron microscopy is not applicable (e.g. for some organic nanomaterials), it is recommended to use another imaging technique instead of electron microscopy.

Keeping the existing recommended definition of nanomaterial in view, it should be straightforward to regard a particulate material as nanomaterial when it has been intentionally produced to have the particle size distribution in the nanoscale (1–100 nm). However, it may not be easy to decide the nanomaterial nature of a material if it has not been produced as such as a nanomaterial but contains a fraction of the particles in the nanoscale. A typical example of this can be a micronised material produced to have particle sizes in the micrometer range, but also contains a fraction of the particles in the nanoscale. In the absence of an upper size cut-off threshold for particles to be included in the determination of 50% in numbers in the current recommended definition, it is difficult to decide whether or not the whole material should be regarded a nanomaterial. This is where other (supporting) criteria, such the use of VSSA, or confirmatory analytical tests, have been proposed in the decision scheme developed by NanoDefine project. The scheme provides a structured way for deciding whether or not a material should be regarded as nanomaterial, and under what conditions a suitable analytical technique may be needed to confirm or exclude it as a nanomaterial. In all borderline cases, the use of imaging techniques based on electron microscopy has been recommended.

Several techniques are available for determination of the various parameters listed in Table 1. In many instances, there is more than one suitable technique available, each with advantages and disadvantages for specific materials and size ranges. In the literature, there are reviews assessing the suitability of different techniques for a range of nanomaterials (Bowen, 2002; Hassellöv et al., 2008; Domingos et al., 2009; Linsinger et al., 2012; Peters et al., 2014b; Babick et al., 2016). An overview of techniques commonly used for the characterisation of nanomaterial is given in Appendix 89. The selection of the appropriate technique is the responsibility of the assessor/applicant in charge and depends on the parameter and the chemical nature of the material. For the size-related parameters, a technique selection support is provided by the NanoDefine e-tool and method manual (available from http://www.nanodefine.eu/index.php/downloads/nanodefine-technical-reports).

Sampling and sample preparation are often crucial steps in the overall analytical process. They usually contribute the largest uncertainty to the result. A critical issue in the sample preparation of nanomaterial is the proper dispersion of particles. This issue is addressed in detail in Section 23. General guidance for sampling also applies to the characterisation of nanomaterial. Special attention has to be paid to sampling, e.g. minimum sample size because of the particulate nature of the analytes (Ersbøll et al., 2010)) and possible segregation and stratification effects (Brüning, 2017).

4.3.1 Characterisation in agri/food/feed products

It is currently difficult to distinguish an intentionally added nanomaterial from background levels of the same materials/substances in nanosized or non-nanosized particle form that may be present in agri/food/feed products, especially when they are present at low levels. Appropriate methods (e.g. stable isotope analysis, elemental fingerprinting) can be applied to distinguish the intentionally added nanomaterial from background levels of the same or similar materials of geogenic, biogenic or anthropogenic origin.

When characterisation of nanomaterial in food/feed matrices is difficult owing to the current limited availability of analytical methods, possible food/feed matrix interactions of the nanomaterial may be determined using food simulants (e.g. water, oil, ethanol, acetic acid or simulants representing the characteristic composition of the target food, e.g. starch for carbohydrate-rich foods). However, the use of a simulant creates an uncertainty, as extrapolation from the results obtained with the simulant may not fully reflect the nanomaterial properties in a real food. With method development and availability, such characterisation of nanomaterial can be expected to shift from food simulants to real food/feed matrices.

4.3.2 Characterisation in test media for in vitro and in vivo testing and in biological matrices

For the toxicological assessment of nanomaterial, it is essential to know in which form the nanomaterial is presented to the test systems. In addition, characterisation of nanomaterial in the test system is relevant to determining the effect of the test medium/formulation (and its constituents) on the characteristics and properties of the nanomaterial so that the validity of the toxicity test outcome may be determined and to allow comparison with the nanomaterial in the food/feed matrix to which exposure takes place.

For in vitro testing as well as for administration of nanomaterial in in vivo studies it is essential that the nanomaterial is properly dispersed in the medium. Dispersion protocols have been developed and published for a number of nanomaterial and purposes (Bihari et al., 2008; Jacobsen et al., 2010; Jensen et al., 2011; Taurozzi et al., a–2012e; Hartmann et al., 2015; Mast and De Temmerman, 2016; OECD TG318 (OECD, 2017b)). More dispersion protocols are available via the websites of international organisations (e.g. OECD – http://www.oecd.org/science/nanosafety/; European Commission-JRC – https://ec.europa.eu/jrc/en/scientific-tool/jrc-nanomaterials-repository; US-FDA – https://www.fda.gov/scienceresearch/specialtopics/nanotechnology/default.htm) and of their respective research projects (e.g. NanoGenoTox – http://www.nanogenotox.eu; Nanopartikel – http://www.nanopartikel.info; NanoDefine – http://www.nanodefine.eu/; NANoREG – www.nanoreg.eu).

In the absence of standardised dispersion protocols, the dispersion efficiency of the applied protocol and the stability of the dispersion should be tested and documented. Apart from their tendency to agglomerate and aggregate (which should be addressed by a proper dispersion protocol and use of dispersants that are compatible with the biological test system), nanomaterial may adhere to the wall of glass ware, tubing, pipette tips, vials etc. see Section 67). Appropriate analytical techniques depend on the type of nanomaterial and medium.

For in vitro studies, the nanomaterials may have to be characterised in the exposure medium at the start and end of the experiment to confirm actual presence in the test system and to observe potential changes that the materials may undergo (Section 65). Characterisation in these cases should include the number-based particle size distribution and concentration. Owing to the possible presence of other particulate materials in the test medium (e.g. proteins) it is mandatory to use a chemically specific method (e.g. single particle inductively coupled plasma mass spectrometry (spICP-MS) or transmission electron microscopy–energy-dispersive X-ray spectroscopy (TEM–EDX)) for these measurements. Non-specific methods such as DLS or CLS are not suited unless it can be demonstrated that the material under investigation is the only (nano)particulate material in the test medium.

ADME studies (as described in Section 43) require the measurement of nanomaterial in body fluids, tissues and excreta. It is not only relevant to quantify the amount of nanomaterial present, but also to specify in which form the nanomaterial is present in these compartments. This includes chemical composition, size and shape, but may also refer to surface modifications and other parameters relevant to the nanomaterial properties. Since many methods for nanomaterial analysis in biological matrices are rather complex and laborious, a tiered approach can be considered for specific cases. For inorganic nanomaterial that contains elements with very low background levels in the matrix the samples can first be screened by non-nanospecific methods for the total content of the respective elements, by e.g. inductively coupled plasma mass spectrometry (ICP-MS), atomic emission spectrometry (AES), atomic absorption spectroscopy (AAS), X-ray fluorescence (XRF). Only positive samples have to be measured with a nanospecific method, e.g. spICP-MS, scanning electron microscopy–energy-dispersive X-ray spectroscopy (SEM–EDX), which may reduce the effort considerably. Recent studies have shown the potential of some nanomaterials to accumulate very specifically, resulting in high concentrations in specific cell types (Sadauskas et al., 2007; Powell et al., 2010; Loeschner et al., 2011; Landsiedel et al., 2012; Kermanizadeh et al., 2015b). Homogenisation of the entire compartment (e.g. liver) may lead to a dilution of the particles below the detection limit. In these cases, mapping techniques can be applied, e.g. time-of-flight secondary ion mass spectrometry (ToF-SIMS), laser ablation inductively coupled mass spectrometry (LA-ICP-MS), chemical force microscopy (CFM), hyperspectral imaging, etc.

4.3.3 Solubility and degradation/dissolution rate

Information on solubility and degradation rate of the pristine material is requested as described in Table 1, Section 16. In this guidance, degradation is considered a general term for the disintegration of a nanomaterial, e.g. due to dissolution, enzymatic or chemical degradation. In addition, the degradation rate in conditions representative of the human gastrointestinal tract and lysosomal fluid is considered key information in the present nanospecific Guidance because this is where nanomaterials generally distribute to and where degradation can occur due to the acidic conditions and presence of enzymes (see Figure 2 and more details in Sections 38 and 40). It is therefore important to understand the fundamental differences between solubility and degradation/dissolution rate.

Solubility is the proportion of solute in solvent under equilibrium conditions (i.e. in a saturated state, see also glossary). It is important to note the difference between dissolution (materials are solubilised into their individual ionic or molecular species) and dispersion (colloidal suspension of particles). Solubility is determined as the concentration of the dissolved material in a saturated solution (i.e. undissolved material present as solid phase). The solubility is dependent on external parameters such as solvent, temperature, pressure and pH. Care has to be taken when the concentration of the dissolved species in the liquid phase is measured to distinguish between dissolved species and dispersed particles. A separation of those species may be achieved by suitable filtration or centrifugation techniques. Limitations for very small particles and particles of a density similar to the solvent have to be taken into account. Protocols and guidelines for the determination of the solubility of nanomaterials have been proposed (Tantra et al., 2016).

High solubility is commonly understood if more than 1 mol/L solvent is dissolved.

The degradation/dissolution rate refers to the kinetics of dissolution. Nanomaterials may degrade/dissolve faster than their bulk counterparts because of their high surface-to-volume ratio. The dissolution rate is influenced by various factors, including solvent, temperature, pH, concentration, and presence of substances interacting with the particle's surface. It can be determined by kinetic measurements such as time- dependent concentration changes (of either the nanoparticles or the dissolved species) or changes in the particle size distribution (to smaller sizes). Dissolution is addressed in detail in Section 38 (In vitro degradation tests).

4.3.4 Characterisation and quantification of nanomaterial in FCM and after transfer from FCM

Various applications of nanomaterial for use in FCM are described in published literature. In some applications, the nanomaterial is applied into surface layers (e.g. in coatings); in others they are embedded in the full FCM matrix (composites) or incorporated in active materials. Nanomaterials when incorporated into an FCM matrix may structurally differ from the pristine nanomaterial. For instance, mineral clays may exfoliate in the polymer matrix under the processing conditions. Therefore, in addition to the characterisation of the nanomaterial used for manufacture of a FCM, the need arises for characterisation of the nanomaterial when present in the FCM (on the surface, in the matrix) and possibly when migrating from the FCM.

Nanomaterials on the surface or in the host polymer matrix can be characterised by their size and shape (see Table 1), typically by using microscopy techniques (SEM, TEM). Other applied techniques for this characterisation include Fourier transform infrared spectroscopy (FTIR) and X-ray diffraction (XRD).

To assess the exposure of the consumer to a nanomaterial from FCM, it is essential to determine the potential migration of the nanomaterial from the FCM into the food matrix. This can be achieved by direct measurements on the nanomaterial in the food matrix, or in the food simulant used in migration testing, or by migration modelling of the nanomaterial in the polymer matrix (Duncan and Pillai, 2014; Noonan et al., 2014; Franz and Welle, 2017; Stormer et al., 2017).

Additionally, consideration should be given to potential release of the nanomaterial from the FCM through mechanical stress or physical disintegration of a FCM polymer matrix. This can be achieved by abrasion testing applying appropriate FCM material stress conditions (such as bending, stretching, thermal stress) and suitable food simulants (able to disperse the nanomaterial) or abrasives (solids generating friction with the FCM surface as used in scratch or tribological tests). It should be noted that abrasion is not covered by ‘conventional’ migration testing or modelling and therefore case specific testing is recommended. The migration patterns of nanomaterials from biodegradable polymer nanocomposites (e.g. polylactic acid (PLA)) may be different from those in conventional (plastic) polymers when official EU food simulants such as 95% ethanol are used. In these cases, the aggressiveness of the food simulant towards the polymer may affect its integrity and the polymer chain-size distribution, and cause physical release of the nanomaterial into the food simulant. In such cases, migration modelling does not apply and testing may need to be performed. Migration/abrasion testing is also needed when nanomaterials have been applied in a coating on an FCM surface (Golja et al., 2017).

To determine the amount of nanomaterial in food simulants after migration, it is possible to use a tiered approach and first apply a total-elemental-analysis method in conjunction with migration modelling estimation (taking into account the concentrations, sizes and shapes of the nanomaterials in the FCM). In this case, an appropriately sensitive detection technique (e.g. ICP-MS) should be selected to minimise the possibility of missing (very) low levels of particles. If the test results and the models estimates indicate the possibility of migration/release of the nanomaterial, more nanospecific technique(s) should be employed to ascertain whether the migrating entities are in nanoparticle or in a solubilised (non-nanomaterial) form.

Further information specific to the evaluation of substances used to manufacture FCM are available in the EFSA CEF Panel opinion on ‘Recent developments in the risk assessment of chemicals in food and their potential impact on the safety assessment of substances used in food contact materials’ (EFSA CEF Panel, 2016a). In general, it is recommended to check and consider the most recent version of the EFSA Guidance(s) specific to FCM.

4.4.1 Standardised methods

Preference should be given to standardised methods where available and applicable. While a number of standard methods are available for particulate materials in pure solid state (e.g. powders), there are hardly any standard methods available for the characterisation of nanomaterial in complex matrices. Appendix 89 provides an overview of currently available standard methods at the time of issuing this Guidance. It is recommended to search the ISO and CEN databases18 for the most up-to-date and appropriate methods.

In cases where no standard methods are available, the applicant is responsible for providing methods for the physicochemical characterisation and quantification of the nanomaterial for which approval is sought that are appropriate both for the pristine state and in matrices. The respective methods have to have standard operation procedures (SOPs) as well as validation reports that are provided with the dossier.

4.4.2 Method validation and performance criteria

As in other analytical fields, it has to be demonstrated that the methods used for the characterisation of nanomaterials in their pristine form (as manufactured) and in commercial formulations, food/feed matrices and in toxicity test systems are fit for purpose and deliver reliable results. The ideal methods will have gone through proper validation (both intra- and inter-laboratory) following existing international guidelines (e.g. IUPAC, 2002); Commission Decision 2002/657/EC19), with adaptation if necessary. The use of any validation protocols differing from internationally agreed protocols would need justifying. The validation would also include determination of the method performance parameters, such as specificity; selectivity; robustness/ruggedness; recovery/trueness; repeatability, and reproducibility; detection/quantification limits for size, number and mass concentration; and measurement uncertainties. Assay robustness for a nanomaterial could be established using similar principles for assessing assay robustness for a non-nanomaterial. It would be worthwhile to include appropriate non-nanoscale controls for the tested nanomaterials. Methods addressing the determination of particle number-weighted size distribution or external dimension of the constituent particles in the nanoscale and beyond should be assessed against general performance requirements as developed by NanoDefine (Rauscher and Mech, 2018). Guidance for the validation of methods for the detection and quantification of engineered nanoparticles in food has been published (Linsinger et al., 2013) and is also applicable to other matrices. The validation report documenting the results on these parameters should be part of the characterisation report. The performance characteristics (including detection limit) should be within reasonable limits that reflect the current state of the art and should be provided in a justification with references to similar techniques in this area. It has to be shown that the performance meets the requirement, e.g. in terms of sensitivity (detection limits) and precision.

4.4.3 Reference materials

Reference materials are essential for controlling and comparing the performance of analytical methods used for nanomaterial characterisation and in their validation. Only a few certified reference materials are currently available, however, for which certification usually covers only one property (e.g. particle size, surface area). The Federal Institute for Materials Research and Testing (BAM) online, www.nanorefmat.bam.de/en/ has inventories of the currently available nanomaterial reference materials.

More reference materials are currently under development and can be expected to become available over time. In addition to the certified reference materials, the European Commission JRC has recently made available a repository of industrial nanomaterials to be used as representative test materials for safety testing (https://ec.europa.eu/jrc/en/scientific-tool/jrc-nanomaterials-repository). These nanomaterials were used for testing by several EU-funded projects (e.g. MARINA,20 NANoREG21) as well as the OECD WPMN and can be used as test materials by any research laboratories to generate comparable toxicological results (Totaro et al., 2016). In the absence of certified reference materials, self-generated and properly characterised and documented test materials may also be used. The ISO has issued a technical specification for the preparation of reference nanomaterials that should be taken into account for these cases (ISO, 2013).

6.2.1 In vitro gastrointestinal digestion

Assessment of the degradation rate of nanomaterials in conditions representative of the human gastrointestinal tract is considered a key first step in the stepwise approach (Figure 2). If a high degradation rate can be demonstrated, as detailed later this paragraph, the standard safety assessment procedure for conventional materials can be followed.

A suite of in vitro digestion models have been described in the literature that assess the release or degradation/dissolution of non-nanomaterials (Dressman et al., 1998; Krul et al., 2000; Oomen et al., 2002; Brandon et al., 2006; Minekus et al., 2014; Lichtenstein et al., 2015; Kästner et al., 2016). In vitro digestion models have been applied to determine the release of various orally ingested compounds, e.g. contaminants from soil (Oomen et al., 2003; Van de Wiele et al., 2007), food contaminants (Versantvoort et al., 2005; Dall'Asta et al., 2010), food mutagens (Krul et al., 2000), food components (Blanquet-Diot et al., 2009; Tydeman et al., 2010), contaminants in toys (Brandon et al., 2006) and drugs (Dressman et al., 1998; Kostewicz et al., 2002; Blanquet et al., 2004). These models simulate the conditions of the gastrointestinal tract (including mouth, stomach and gut). The differences between these models relate to the extent to which physiology is simulated, e.g. from very simple to rather sophisticated by using static or dynamic conditions, and with or without enzymes, bile salts etc. In addition, the physiology that is simulated may vary between models: fasted versus fed conditions, baby versus adult.

An in vitro digestion method suitable for food under fed conditions (as opposed to fasted) has been described by Minekus et al. (2014) that is harmonised by the COST Infogest network.24 The effects of differences in pH, mineral type, ionic strength, digestion time and, enzyme activity were also discussed. The method consists of a short simulation of mouth conditions, followed by a gastric phase at pH 3 for 2 h and an intestinal phase at pH 7 for 2 h. The composition of the digestion fluids was exactly described. While this method is not specified for nanomaterials, nor is it an officially standardised method, this it is considered a key approach also to be used for nanomaterial in food, i.e. for simulating physiological conditions in the gastrointestinal tract after food consumption.

For fasted conditions, several in vitro digestion methods have been described and compared by Koch et al. (2013). These methods can be explored further for applicability in this context.

Recently, some experience has been gained with the application of in vitro digestion models to nanomaterials. NANoREG D5.0225 shows that the degree of aggregation/agglomeration (see Table 1) of several nanomaterials (Ag, SiO₂ and ZnO) in artificial saliva, gastric juice and intestinal juice varies over these stages. Nanoparticles were still present in the intestinal stage, although considerable degradation was observed for Ag and ZnO (up to 45% was degraded after mouth, stomach and 2 h of intestinal digestion under the specified conditions). Degradation measurement in such complex matrices was challenging, and a single, robust and rapid test method for all types of materials in all types of matrices could not be developed. Possible techniques include spICP-MS, and ICP-MS/AES based methods in combination with a separation technique like ultrafiltration (NANoREG D5.02). Furthermore, Peters et al. (2012) and Walczak et al. (2013) investigated respectively SiO₂ and Ag particle distribution in artificial mouth, gastric and intestinal conditions. Here also, the degree of aggregation/agglomeration varied between the different compartments of mouth, stomach and intestine. Sieg et al. (2017) describe that the aluminium nanoparticles remained unchanged in saliva, and strongly agglomerated in the gastric phase showing also an increased ion release. The levels of aluminium ions decrease in the intestinal fluid and particles de-agglomerated. Altogether, dissolution of nanoparticles was limited. DeLoid et al. (2017b) found that iron(III)oxide (Fe₂O₃) nanoparticles in an oil-in-water emulsion showed different size distribution in an in vitro digestion model with or without Fe₂O₃, in the mouth, stomach and intestinal phase. This size distribution comprised all particles, unspecified to chemical composition. There appeared to be minimal dissolution of the Fe₂O₃ particles in all three stages of digestion.

Remarkably, intestinal digestion of soluble silver ions (from AgNO₃) and ionic aluminium also resulted in the formation of particles (of 20–30 nm) composed of silver, sulfur and chlorine (Walczak et al., 2013) and aluminium (Sieg et al., 2017).

There is a lack of validation and standardisation of in vitro digestion models for nanomaterials. At present, no comparison for nanomaterials has been made between in vitro degradation/dissolution data from such digestion models and in vivo data. Lefebvre et al. (2015) concluded that in vitro digestion models are generally applicable to nanomaterials, as the basis of the models is mimicking the conditions of the gastrointestinal tract. Important properties affecting degradation/dissolution are expected to include physical forces, temperature, pH, presence of enzymes, salts and bile, and presence of food (Bellmann et al., 2015). Therefore, a rationale for the used in vitro digestion model used in the stepwise approach should be provided in view of the representativeness of the physiologic state for exposure, and whether the model is expected to represent worst-case, realistic or favourable conditions for in vitro degradation of the specific nanomaterial. For example, fasted conditions may be less representative of the use of nanomaterials in food products, whereas low pH conditions in the stomach – as may occur in fasted conditions – may promote the degradation of most metals and metal oxides. Therefore, a careful choice between fed or fasted conditions should be made for the test system in view of anticipated conditions and the worst-case situation, depending on the characteristics of the nanomaterial and the application. In some cases, it may be relevant to investigate the degradation rate under both physiological conditions. Further scientific research on the impact of food components on the degradation of nanomaterials is recommended (Lichtenstein et al., 2015). The in vitro digestion model should also be critically assessed for its reliability and reproducibility. To increase the reliability of the degradation information from the model, the degradation rate should be determined by including different time points (at least four time points in duplicate at about 5, 15, 30 and 60 min) in the intestinal phase. The study should be performed at three different concentrations as this may affect the degradation, and the middle concentration should be representative of human exposure. This can be calculated by the estimated daily intake that, depending on the anticipated use, is ingested at once or throughout the day assuming that the daily volume of secretions into the gastrointestinal tract is 4–5 L. Furthermore, the particle number–size distribution and concentration should be analytically determined with a chemically specific method (i.e. verifying the chemical identity of the measured particles, e.g. spICP-MS or TEM-EDX). The concentration of the solute and, if present, other degradation products, should also be determined.

Some materials may degrade completely in the conditions of the stomach and then precipitate in the intestinal conditions as salts or nano- or microsized particles (Walczak et al., 2013). The dissolved fraction should not be separated from the rest during in vitro digestion as this may promote the dissolution. It has also been shown that SiO₂ particles can form large non-nanosized agglomerates in the conditions of the stomach that may disagglomerate in the intestine stage (Peters et al., 2012). This indicates that the absence of small particles in stomach conditions is insufficient to conclude that there will be no exposure to the nanoparticles. Therefore, only information on the dissolution rate in the intestinal conditions is considered relevant and should be provided. This means that the simulation of the stomach conditions still need to precede the intestinal conditions, but no data on the dissolution rate in the stomach phase are required.

The measured concentrations of solute, degradation products and particles should be compared with the start situation at the beginning of the in vitro digestion, in saliva or in the matrix as introduced into the in vitro digestion model. Analytical limitations such as detection limits should be taken into consideration (see Section 21).

A nanomaterial is considered to degrade quickly/have a high degradation rate if the degradation rate profile in the intestinal phase shows a clear decrease in the presence of particles over time (no plateau), and that 12% or less of the material (mass-based26) – compared with the particulate concentration at the beginning of the in vitro digestion – is present as particles after 30 min of intestinal digestion. This is indicative that the rest of the material should be fully degraded to non-nanomaterial (e.g. ionic) under gastrointestinal conditions (as mentioned in Figure 3). Details of the rationale and discussion of the uncertainty for this cut-off value can be found in Appendix 91. The cut-off value assumes a first-order half-life in the intestinal phase of 10 min. It is considered feasible to measure this value analytically, and the time required to reach the intestinal epithelium and be taken up by cells is of the same order of magnitude. In such cases, a nanospecific risk assessment would not always be required. However, in case of complete digestion in gastrointestinal fluids, localised exposure (e.g. in the upper gastrointestinal tract) needs to be considered (Holpuch et al., 2010). When applicable to the material, a comparison to an ionic control should be included to identify de novo particle formation from these ions.

It should be noted that the cut-off for a high degradation rate is based a pragmatic choice based on limited science (see Appendix 91) and may need modification with increasing knowledge.

If it cannot be demonstrated that the material quickly degrades, one should continue to Step 1 of Figure 3.

6.2.2 Stability in lysosomal fluid

Once in the body, some nanomaterials may not be easily cleared and may accumulate over time. Assessment of the stability in lysosomal conditions is important to screen the potential of nanomaterials for biopersistence and intracellular accumulation (Utembe et al., 2015). Lysosomal conditions are considered as model as this is where nanomaterials generally distribute to and where degradation can occur due to the acidic conditions and presence of enzymes.

Release of ions due to degradation in lysosomal fluid can be an indication of toxicity due to these ions and should be considered in further testing.

Artificial lysosomal fluid simulates the inorganic environment within lysosomes (hydrolytic enzymes are typically not included) and is buffered at pH 4.5–5.0 (see e.g. Stopford et al., 2003; Stefaniak et al., 2005; Henderson et al., 2014; Pelfrêne et al., 2017).

To assess the stability in lysosomal fluid, pristine materials should be submitted to in vitro simulated lysosomal degradation and the degradation rate in lysosomal fluid has to be determined by considering different time points (at least four) in duplicate at three different concentrations. Time points for sampling and concentrations have to be properly selected and justified; time points are normally expected to be in the range of hours and extend up to, e.g. 72 or 96 h. Degradation and particle size of nanomaterials after lysosomal treatment should be characterised using the same approach described for in vitro gastrointestinal digestion, where a half-life of about 24 h is considered indicative of a high degradation rate given that degradation in lysosomal conditions should provide a predictive estimate for potential accumulation in cases of where frequent exposure is expected. This half-life would result in 12% or less of the material (mass-based) being present at 72 h compared to the particulate concentration at the beginning of the degradation test (first-order kinetics). As with in vitro gastrointestinal digestion, no evidence of a plateau should be visible.

6.8.1 Chronic toxicity and carcinogenicity

Chronic toxicity and carcinogenicity studies are performed in a single species, generally the rat. Either separate studies (OECD TGs 452 (2009d) and OECD TGs 451 (2009c), respectively) or preferably the combined study (OECD TG 453 (2009e)) can be carried out. Carcinogenicity study in a second species would only be triggered by the results in the preferred species (equivocal results or species specific findings) or by observations from specialised studies to investigate the mode of action or mechanism of toxicity or carcinogenicity observed.

6.8.2 Reproductive and developmental toxicity

There are indications in the literature suggesting that systemic exposure to specific nanomaterials may result in adverse effects on reproduction and development (Brohi et al., 2017). In this review, ROS generation and inflammation, direct interaction with the male and female reproductive systems, alterations of ovarian gene expression and steroidogenesis, adverse effects on fetal morphology, organogenesis and development have been related to reproductive and developmental toxicity induced by specific nanomaterials. Some nanomaterials, depending on their physicochemical properties, were reported to be able to penetrate the blood–testis barrier and the placental barrier (Brohi et al., 2017).

The data from Step 2 subchronic toxicity testing are relevant when considering the need for reproductive and developmental testing in Step 3.

The repeated dose 90-day oral toxicity study (OECD TG 408 (2017a)) offers only limited information on reproductive toxicity and none on developmental toxicity; it can inform about effects on the reproductive organs and, if assessed, the oestrous cycle, but it does not assess fertility and the whole reproductive cycle from in utero exposure onwards, through sexual maturity to conception, gestation, prenatal and postnatal development. Decisions on whether tests are necessary for reproductive and developmental toxicity need to be considered in light of the toxicity data and toxicokinetics information available. If the Step 2 toxicokinetic study shows that the test material is systemically available in the test species (normally rodents) or suspected to be systemically available in humans, Step 3 testing for reproductive and developmental toxicity is required. For materials that do not appear to be systemically available, indications of effects on reproductive organs or parameters in the 90-day oral toxicity will also trigger Step 3 testing for reproductive and developmental toxicity. In the case where absorption appears to be very low, Step 3 reproductive and developmental toxicity studies may still be needed if the tissue distribution data from the 90-day oral toxicity study indicate that the test material is able to reach reproductive organs and distribute there.

Step 3 testing for reproductive and developmental toxicity comprises a prenatal developmental toxicity study (OECD TG 414 (2001)) and an extended one-generation reproductive toxicity study (EOGRTS) (OECD TG 443 (2012a)). Cohorts for the preliminary assessment of additional more specific endpoints should be routinely incorporated in the EOGRTS. Where it already exists, a multigeneration study, instead of an EOGRTS, would be acceptable, provided that sufficient information on possible neurotoxicity and immunotoxicity is available (for example from an extended 90-day study, OECD TG 408 (2017a)).

With the additional parameters evaluated in the F1 generation in the EOGRTS, it is expected that the F2, with their limited parameter assessments, would seldom affect the hazard characterisation for risk assessment (Piersma et al., 2011). When predicted human exposures are considered adequately characterised, however, this may be factored into the decision to require the assessment of an F2 generation.

In devising appropriate Step 3 additional testing, a case-by-case approach should be adopted with careful consideration given to all available data as well as animal welfare issues. Step 3 testing is triggered by results in Step 2 studies and might be comprised of additional studies for. e.g. endocrine, developmental neurotoxicity (OECD TG 426 (2007)), and mode-of-action studies that could include both guideline studies and experimental studies designed on a case-by-case basis.

6.8.3 Immunotoxicity and Allergenicity

Immunotoxicity may manifest in the form of adverse effects on the structure and function of the immune system itself, or an adverse consequence (such as an allergic or autoimmune reaction) that may arise from the immune response. Most inorganic and small-molecule organic substances in conventional (non-nanomaterial) forms are not immunogenic, but may act as haptens and become immunogenic after binding to proteins. In addition, they may act as adjuvants for other immunogens. For example, aluminium compounds have been used as adjuvants in vaccines. Nanomaterial forms of these compounds may present different immunological behaviour compared with conventional materials.

In addition to being potentially immunogenic/antigenic themselves, nanoparticles may also act as ‘Trojan horses’ for other immunogens/antigens by binding them on to the particle surfaces and carrying them to immune cells (see Section 12). The state of current knowledge on the immunotoxicity of nanomaterials, and guidance on methods to evaluate this, will be described in a WHO publication, which has undergone public consultation in 2017 (DRAFT WHO, 2017 ‘Draft Environmental Health Criteria Document: Principles and Methods to Assess the Risk of Immunotoxicity Associated with Exposure to Nanomaterials’, once the final document is published, an additional reference will be provided as a corrigendum).

The tiered approach to testing outlined in this present Guidance includes, at Step 2, a 90-day study in rats (OECD TG 408 (2017a)). This involves investigation of the effect of the nanomaterial on a number of parameters that may be indicative of an immunotoxic or immunomodulatory effect. These include: changes in spleen and thymus weights relative to body weight in the absence of overt toxicity, histopathological changes in these and other organs of the immune system (e.g. bone marrow, lymph nodes, Peyer's patches); changes in total serum protein, albumin:globulin ratio and in the haematological profile of the animals, i.e. the total and differential white blood cell counts.

The effects may be extended or, alternatively, seen for the first time in Step 3 studies, notably the EOGRTS (OECD TG 443 (2012a)), but also in chronic toxicity/carcinogenicity studies conducted according to OECD TGs 452 (2009d), OECD 451 (2009c) or OECD 453 (2009e). In the EOGRTS, a cohort of animals is specifically dedicated to assessing the potential impact of exposure on the developing immune system. In subchronic and chronic studies, haematological and clinical chemistry data are generally provided, together with phenotypic analysis of spleen cells (T-, B-, NK cells) and bone marrow cellularity. The EOGRTS provides additional information on the primary immunoglobulin M (IgM) antibody response to a T-cell-dependent antigen such as sheep red blood cells (SRBC) or keyhole limpet hemocyanin (KLH).

Evaluation of the potential of a nanomaterial to adversely affect the immune system may be based on an integrated assessment of the results obtained from these toxicity studies (Steps 2 and 3). If these results indicate that the nanomaterial has such a potential, additional Step 4 studies should be considered, on a case-by-case basis. These will normally be designed to investigate the underlying mechanisms of the effects seen, and/or their biological significance. Step 4 studies may include specialised functional, mechanistic, and disease model studies. There are no OECD guidelines for these extended specialised studies, but these should be based on IPCS.

At present, there are no data or validated studies in laboratory animals that would allow assessment of the potential of a substance to cause allergic reactions in susceptible individuals following oral exposure. Where the nanomaterial is a potential allergen (e.g. a protein or a peptide) or contains residues of proteins or other known potential allergenic molecules, the principles discussed in the EFSA Guidance on the Allergenicity of GMOs should be followed in evaluating allergenic components. These principles for the determination of allergenicity include the investigation of structural aspects of the protein or peptide, in silico (or bioinformatics) approaches, immunoglobulin E (IgE) binding and cell-based methods, analytical profiling techniques and animal models (EFSA GMO Panel, 2010; EFSA NDA Panel, 2014). Since no single experimental method yields decisive evidence for allergenicity and allergic responses, a weight-of-evidence approach taking into account all the information obtained from various test methods is recommended.

6.8.4 Neurotoxicity

Indications of potential neurotoxic effects of a test substance are obtained from the modified 90-day toxicity study (Step 2b). This study involves investigation of the effect of the nanomaterial on a number of parameters that may be indicative of a neurotoxic effect. These include changes in clinical signs, functional observational battery and motor activity, brain weight relative to body weight in the absence of overt toxicity, and histopathological changes in this organ. Other information, such as read-across considerations or physicochemical properties that are indicative of neurotoxic potential should also be considered. Where indications of potential neurotoxicity are seen at Step 2, further neurotoxicity testing (OECD TG 424 (1997a)) should be considered. Such testing should be carried out on a case-by-case basis and is intended to confirm or further characterise (and quantify) the potential neurotoxic response induced by the nanomaterial. Information from other studies should also be considered in designing these studies to minimise confounding effects secondary to general toxicity. Further specialised studies can also be performed to elucidate mechanisms to improve extrapolation from animals to humans when completing the risk assessment.

6.8.5 Endocrine activity

Evidence exists that several types of nanoparticles may adversely affect the endocrine system, and in particular the male and female reproductive systems (Iavicoli et al., 2013; Larson et al., 2014). Hormone alterations also appear to play a role in other toxic effects of nanoparticles with sex-related patterns (Tassinari et al., 2014; Ammendolia et al., 2017). However, the role of endocrine-related modes of action in the toxic effects of nanomaterials is largely unknown and unexplored, and warrants further investigation.

The starting point for investigating endocrine disruptive properties is the design of the modified 90-day toxicity test (OECD TG 408 with extended parameters from the OECD TG 407 (2008)) (preliminary reference (OECD TG 408 (2017a)). The additional parameters place more emphasis on endocrine-related endpoints (e.g. determination of thyroid hormones, gross necropsy and histopathology of tissues that are indicators of endocrine-related effects) and (as an option) assessment of oestrous cycles.

For further testing, the EFSA Scientific Opinion of 2013 (EFSA Scientific Committee, 2013) describes scientific criteria for the identification of endocrine disruptors (EDs) and the appropriateness of existing test methods for assessing effects mediated by these substances on human health and the environment (mainly based on OECD TG 150 (2012b)). To distinguish between EDs and other groups of substances with different modes of action, it was concluded that an ED is defined by three criteria: the presence of (i) an adverse effect in an intact organism or a (sub)population; (ii) an endocrine activity; and (iii) a plausible causal relationship between the two.

6.8.6 Gut microbiome

Currently, there are limited data on the interaction of nanoparticles with the gut microbiome (e.g. Frohlich and Frohlich, 2016; Bouwmeester et al., 2018), but further research is ongoing. For nanomaterials with antibacterial properties (Hadrup et al., 2012), the possibility of interactions and on the composition of the microbiome and the mucus should be considered, especially for unabsorbed nanoparticles (Fröhlich and Roblegg, 2012; Mercier-Bonin et al., 2016).

The consequences of interactions with the gut microbiota on systemic or local tissues or physiological processes (whether direct or indirect) should be identifiable in the examinations that are part of the modified 90-day toxicity study. Currently, there are neither specific tests for this nor a clear understanding of the significance of changes in composition of the microbiome or their effects on health and further research would be valuable.

6.9.1 Specific issues for in vitro studies

Although ‘validated’ in vitro methods specifically for nanomaterials are not currently available, the results of ‘valid’ methods may be considered for hazard identification (See Glossary for both terms).

In vitro studies with nanomaterials require extra attention to the suitability of the test methods for the purpose, e.g. test reagents used for standardised toxicity tests might react with the nanomaterial or the read-out signal of the test.

Additional quality controls should include (where available), negative and positive reference nanomaterials, assay reagent controls (such as the dye) and the non-nanomaterial controls.

It also has to be shown that the target cells were exposed to the nanomaterial along with the determination of the number-based size distribution and concentration of nanomaterials at the start (and end if applicable) of in vitro testing. These should be measured in the exposure medium using an appropriate method (see Section 23). Models based on DLS and density have been proposed to estimate how much nanomaterial is reaching the cell system (DeLoid et al., 2017a). Other models exist and can also be used to assess if the cells are truly exposed (e.g. Hinderliter et al., 2010).

For nanomaterials, the nominal concentration/dose may not be representative of the concentration/dose reaching the cells. Therefore, the assessment of the dose delivered to the cell system (Rischitor et al., 2016) and the internalised dose (the fraction of nanomaterials internalised by the cells) is highly recommended to allow better interpretation of the results and for comparison with or extrapolation to in vivo situations. In in vitro tests a series of concentrations should be used, keeping in mind the response–concentration range, and their choice justified. Exclusive use of high concentrations that lead to extensive agglomeration/aggregation should be avoided as well as conditions leading to sedimentation of the material.

At least two independent in vitro assays (see Section 47) per individual endpoint need to be selected.

The possible interference of nanomaterials with in vitro test systems also has to be taken into account, e.g. with assay components (reagents, proteins, nutrients), and with optical read-out system (e.g. dyes) (Kroll et al., 2012; Guadagnini et al., 2015). Case-by-case background controls, e.g. cell-free medium, all the reagents and the nanomaterials, should also be included and processed in the result interpretation. These background corrections should also take into account any changes in the dispersion status of the nanomaterial during and after the test, because this might influence their degree of interference with a test system based on optical measurements.

Specific issues with in vitro digestion are addressed in Section 45.

6.9.2 Specific issues for in vivo studies

Nanomaterials are potentially unstable when dispersed. In oral toxicity studies, the test material can be administered by adding it to the animal feed, the drinking water or by gavage. For proper administration the nanomaterial should be homogeneously blended into the feed matrix or stably and uniformly dispersed in the drinking water or gavage vehicle. The stability and physicochemical characteristics of the nanomaterial in the vehicle should always be determined (see Section 21). Possible interactions with the administration vehicle, either the food matrix or water, need to be determined in advance before in vivo administration. This may require dispersions for testing to be prepared freshly and used immediately after preparation. Complete delivery of the dose should be checked because a nanomaterial may, for example, adsorb to the walls of the drinking vessel or the gavage syringe and therefore may no longer be available (i.e. there is no exposure) (Kreyling et al., 2017).

Application by gavage ensures that the nanomaterial is dispersed, characterised and administered under well-defined conditions. It is known that this method of administration can give a fairly precise dose of nanomaterial delivered to the animal and a well-characterised degree of dispersion (Kreyling et al., 2017).

On the other hand, application by gavage is not likely to be representative of the lower concentrations delivered over time when the nanomaterial is administered via drinking water or feed, two ways that more closely simulate human dietary exposure. Gavage provides a bolus of nanomaterial at a given time. Absorption kinetics following bolus gavage administration differs from kinetics following continuous administration leading to a greater likelihood of effects associated with the peak concentration rather than total exposure. In addition, when exposure to the nanomaterial is expected to happen via solid foods, the lack of co-ingestion of dietary components (with which a nanomaterial can interact) is another limitation of gavage. However, at the current state of knowledge, bolus gavage administration of the nanomaterial still might be the method of choice for identification and characterisation of hazards associated with the nanomaterial; this is because of the certainty of the administered dose and thus the dose–response relationship for possible adverse effects. In specific cases, and especially when exposure occurs mainly through solid and liquid foods, additional groups with dietary or drinking water administration have to be included to determine whether hazards associated with the nanomaterial are observed under more realistic exposure scenarios.