Grouping of nanomaterials to read-across hazard endpoints: a review

Figures & data

Table 1. List of frameworks proposed for NM grouping.

Approach (reference)	Assessment goal(s)	Basis for grouping	Predefined groups	Testing strategy supported	Practicality (standard methods identified)	Applicability	Comments on applicability within REACH
DF4nanoGrouping (Arts et al. 2015)	Human health hazard assessment (inhalation exposure)	Assignment to one of the four pre-defined categories is based on •water solubility, •particle morphology and composition, •dissolution rate, •surface reactivity, •dispersability, • in vitro reactivity Confirmation of the categorisation of a NM comes through information from in vivo studies (biopersistence, biodistribution, and genotoxicity). Use release and exposure information are applied as qualifiers to support grouping or appropriate testing	Four categories are identified: Soluble NMs biopersistent high aspect ratio NMs Passive NMs Active NMs	Exposure-based test waiving is supported; toxicological tests are identified for the inhalation route	Preferred methods, protocol, and existing test guidelines are identified (for measurement of PC properties)	The approach is applied to carbonaceous, metal sulphate NMs, metal oxides, amorphous silica, and pigments. Belonging of a NM to a pre-defined class determined testing strategy and risk management	The approach takes into account REACH requirements, and aims at test waiving. The proposed testing strategy focuses on inhalation route of exposure (oral and dermal exposure are briefly addressed)
MARINA framework (Oomen et al. 2015)	Human health and environmental hazard and risk assessment	A group includes NMs with low variability in PC, exposure, (eco)toxicological kinetic, or fate properties	Suggested predefined categories: quickly dissolving NMs passive NMs active NMs NMs with high aspect ratio	Yes: the Marina Risk Assessment Strategy is supported	Lack of standard methods	No case studies presented	Generic framework suitable for application within REACH
RIVM grouping approach (Sellers et al. 2015)	Human health and environmental hazard and risk assessment	Similarity may be supported by information about chemical identity, particle characteristics, fundamental transport and behavior, activity and reactivity (the so called Tier 0 testing)No threshold to determine similarity has been developed	No	Yes, read-across is considered for each endpoint to waive tests on animals	Standard methods are mostly not available	The approach is applied to TiO2 and Ag NMs	Read-across is supported when PC properties of the NM under evaluation are similar to the ones reported in the reference studies (then test waiving is proposed)
US-Canada Regulatory Cooperation Council (RCC, 2013a, 2013b)	Human health and environmental safety	NMs are classified on their chemical composition in order to identify (dis)similarities with bulk chemicals and for supporting future read-across. PC properties for each group are identified to support grouping and read-across: size shape surface chemistry solubility composition crystal structure	Seven classes are identified: CNTs inorganic carbon metal and metalloid oxides metals, metal salts, and metalloids; semiconductor quantum dots organics other classes	Yes. a flowchart is provided where, according to available information on exposure and PC properties, testing is suggested	Lack of standard methods	No case studies	A testing strategy is identified according to the exposure route PC properties for the definition of similarity for read-across are identified for each class
US National Institute for Occupational Safety and Health (NIOSH) (Kuempel et al. 2012)	Human health risk assessment in occupational settings (inhalation)	Particle size shape density surface area reactivity solubility	Four categories are identified: higher solubility particles that can reach systemic tissues poorly soluble particles for which toxicity is related to the SSA poorly soluble toxic NMs where reactive particle specific surface area dose determines toxicity fibrous particles for which the toxicity is related to bioperistence and genotoxicity	A testing strategy is supported but notproposed	Specific standard methods are not identified	The approach is applied to fine and ultrafine particles (diesel exhaust particulate, carbon black in the ultrafine range). One NM falling in a predefined class would be compared to the identified reference materials and the risk estimate would be made according to identified PC properties	Occupational inhalation exposure is considered; the approach may support the selection of PC properties for hazard grouping

Table 2. List of occupational banding tools containing a hazard module (mainly from Liguori et al. 2016).

Approach (reference)	Assessment goal(s)	PC properties considered for hazard classification^a	Hazard classes	Availability of a case study
Quantitative clustering framework (Drew et al. 2017)	Estimation of potency-based groups for pulmonary inflammation	Density, surface area, and diameter were most predictive of the potency group	Group 1–4 with one corresponding to highest potency	The framework is developed from a dataset of 18 particles (NMs and bulk) and 25 rodent studies; four potency groups are identified according to dose-response. The model is then tested on a separate dataset of six materials, using only the physicochemical information to predict the hazard potency group
ENM safety Classifier (Fortino and Grevo, 2017)	Group NMs according to toxicity	Intrinsic properties (nanospecific properties are not mentioned)	Three classes are identified as low, medium, and high toxicity	The tool is developed for 31 NMs starting from the NANOSOLUTIONS dataset and it is validated with data from MARINA project
(Oosterwijk et al. 2016)	Risk banding framework (occupational settings)	Size, net charge, hydrophobicity, solubility (at controlled pH), and ion toxicity, reactivity	Four hazard classes defined according to local and systemic toxicity, depending on net charge, solubility and reactivity and on net charge, size and hydrophobicity respectively	Conceptual framework
NanoSafer (Jensen et al. 2014)	Occupational safety (e.g. at SMEs); precautionary risk assessment	PC properties: size, shape, water solubility, and surface coating. Materials OEL; risk	Four control banding classes are identified based on toxicity	No case study available
Swiss precautionary matrix (Höck et al. 2013)	Employers and employees prioritize health risks and implement control measures	Stability, redox activity, catalytic activity, and ROS formation potential; induction potential for inflammation	Three classes of potential effects (low medium and high)	No case study available
Stoffenmanager (Van Duuren-Stuurman et al. 2012)	Human health risk assessment in occupational settings (inhalation)	Shape (fibre length). Inhalation hazard; water solubility; biopersistence	Five classes are identified (lowest to highest hazard levels are identified)	Iron powder. Falling in a risk band identifies the level of hazard control
IVAM^b guidance (Cornelissen et al. 2011)	Workers and occupational exposure	Shape (fibrous particle) and water solubility	Three hazard categories: (water) soluble nanoparticles, synthetic, persistent nanomaterials (non-fibrous) fibrous, nonsoluble nanomaterials	NA
ANSES CB tool (ANSES, 2010)	Exposure prevention; for small to large enterprises	Reactivity, solubility (in water or in lung lining fluid) rate, shape, and biopersistence. Preliminary hazard band of the bulk material or of an analogue material: if available, the hazard band for the NM addressed is derived from that	Five hazard bands are identified (from very low to very high)	No case study available in the guidance
CB Nanotool (Zalk et al. 2009)	Oriented for nanotechnology researchers risk assessment and management	Chemical form, size, shape, surface reactivity, solubility; Information on parent material or NM: Toxicity (lowest OEL), LD50, mutagenicity, carcinogenicity, reproductive toxicology, dermal toxicity, and asthmagenicity	Four bands for identified as severity scores (taking into account both exposure and hazard information by summing the identified factors)	The tool was applied to several activities and NMs (Paik Zalk, and Swuste, 2008)

^aToxicity data, if used.

^bIVAM: environmental research agency and consultancy which had its origins in the University of Amsterdam’s Environmental Science Department.

ROS: reactive oxygen species; SME: small and medium enterprises; OEL: occupational exposure limit; LD50: lethal dose killing the 50% of sample.

Table 3. Grouping for read-across: case studies identified in the literature and computational applications that aim at ranking NMs.

Approach	Objective of the study	Methods	Dataset	Properties for grouping	Result	Comments
(Lamon et al. 2018)	Read-across hazard endpoints to fill data gaps for in vitro genotoxicity (comet assay)	Chemoinformatic tools such as hierarchical clustering, principal component analysis, and decision forest are applied to a dataset of TiO2 NM to predict the outcome of the in vitro comet assay	Six source nanomaterials (TiO2); read-across is made to two target TiO2 nanoforms	Properties considered in the analysis are: total non-TiO2 content including coating and impurities surface coating organic matter crystal type and size shape primary particle diameter specific surface area isoelectric point density pore volume dustiness biodurability redox potential	Presence of coating was predicted to affect the outcome of the in vitro comet assay	This study represents an application of read-across to an in vitro endpoint and is presented following the ECHA proposed framework (ECHA 2017)
Sizochenko et al. (2018)	Read-across of NMs to fill data gaps for in vitro cytotoxicity (EC50/IC50) for bacteria, algae, protozoa, human keratinocyte cell, Balb/c 3T3	Generation of a self-organising map model is combined with interspecies correlation and is applied for data mining of the nanoparticle toxicity library to identify groups based on toxicity	30 silica and metal oxides	Enthalpy of cation formation	Four groups are identified (group 1–3 of increasing toxicity, and one corresponding to unknown toxicity)	The endpoint is not relevant for REACH. the model does not predict all the NMs taken into consideration (group 4 has unclear toxicity value)
Gajewicz et al. (2015)	Read-across in dataset of NMs to fill data gaps for the cytotoxicity endpoint (reduction of cells viability, EC50) for E. coli and HaCaT cell line (human keratinocytes)	One calculated molecular descriptor is identified in each case study (endpoint) and is used to group NMs in the Euclidean space; NMs from the validation set are (qualitatively) predicted according to the Euclidean distance	Two datasets with 17 and 18 metal oxide NMs	Euclidean distance is the similarity metric; properties considered are: enthalpy of formation of a gaseous cation having the same oxidation state as the one in the metal oxide structure (ΔHMe+) Mulliken’s electronegativity (χ^C)	Three groups of NMs with increasing toxicity properties were identified in both case studies	This study represents an application of a qualitative read-across that could be relevant according to REACH Annex XI but the endpoint is not required by REACH
Liu et al. (2015)	Consider dosimetry modeling in in vitro toxicity ranking	Hazard ranking was based on the EC50 and slope of the dose-response curves. Sedimentation of NMs was calculated via a fate model considering Brownian motion and gravitational settling	Seven metal oxide NMs	Hazard ranking considering delivered dose was based on dose-response analyses and compared with administered dose ranking	The comparative ranking between administered and delivered dose did not show any difference	The approach would be useful if there was an validation to in vivo studies
Chen et al. (2014)	Prediction of ADME by biological surface adsorption index(BSAI)	Calculation of the adsorption coefficient k as a function of five variables (describing molecular interactions); principal component analysis (PCA) for clustering	23 NMs (metal oxides, Ag, and organic NMs)	Clustering using the five identified variables	The prediction of the adsorption index was improved compared to the previous model	The approach could be applied by using the five variables as the basis for ranking and read-across to predict ADME
Zhang et al. (2012)	Validate hazard ranking based on high-throughput screening (HTS) output and on in vivo tests	Regression tree analysis for the effect of conduction energy band and dissolution on metal oxides NMs toxicity (cell viability)	24 metal oxides	Cellular toxic NMs were identified depending on band gap. Dissolution was then addressed to assess the toxicological impact of NMs	Dissolution and conduction energy band could predict the toxicity of 7 out of 8 NMs that were predicted (and tested) as toxic and 1 out of 8 NMs that was predicted (and tested) as non-toxic	This application identifies a group of non-toxic NMs and two groups of toxic NMs depending on dissolution and conduction energy band. The authors show correlation between in vitro results and in vivo acute pulmonary inflammation
George et al. (2011)	Hazard ranking of a set of metal oxide NMs	Information on NM type, dose, duration of exposure, cellular targets, and cytotoxicity events was extracted from HTS data and visualised in self-organising maps and 4 groups of NMs were identified according to the observed effect	Two cell lines and four cytotoxicity responses for seven metal oxide NMs	NMs were ranked according to similarity between their lethal response outcome or the cytotoxic response profile of each cell line (HTS data)	NMs were ranked in four groups according to the cytotoxicity endpoints. They were ranked in five groups according to the in vivo tests	Hazard-based ranking of NMs. A more extensive dataset on PC properties may help to identify a group of similar NMs (structural similarity)

ADME: absorption, distribution, metabolism, and excretion; ICPMS: inductively coupled plasma mass spectrometry; SEM: scanning electron microscope; TEM: transmission electron microscope

Arts, J., M. Hadi, M.A. Irfan, A. M. Keene, R. Kreiling, D. Lyon, M. Maier, et al., 2015. “A Decision-Making Framework for the Grouping and Testing of Nanomaterials (DF4nanoGrouping).” Regulatory Toxicology and Pharmacology 71 (2): S1–S27. doi:10.1016/j.yrtph.2015.03.007.

 PubMed Web of Science ®Google Scholar

Oomen, A., E. Bleeker, P. Bos, F. Broekhuizen, van, S. Gottardo, M. Groenewold, D. Hristozov, et al., 2015. “Grouping and Read-across Approaches for Risk Assessment of Nanomaterials.” International Journal of Environmental Research and Public Health 12 (10): 13415. doi:10.3390/ijerph121013415.

 PubMed Web of Science ®Google Scholar

Sellers, K., N. Deleebeek, M. Messiaen, M. Jackson, E. Bleeker, D. Sijm, and F. Broekhuizen. 2015. Grouping Nanomaterials A Strategy Towards Grouping and Read-across (No. 2015-0061), RIVM Report. Bilthoven, Netherlands: RIVM.

 Google Scholar

RCC. 2013a. Nanotechnology Initiative Work Element 2, Priority Setting: Development of a Joint Nanomaterials Classification Scheme, Nanotechnology Initiative. Ottawa, Canada: Regulatory Cooperation Council (RCC). Innovation, Science and Economic Development Canada.

 Google Scholar

RCC. 2013b. Nanotechnology Initiative Work Element 3: Risk Assessment/Risk Management, Nanotechnology Initiative. Ottawa, Canada: Regulatory Cooperation Council (RCC). Innovation, Science and Economic Development Canada.

 Google Scholar

Kuempel, E. D., V. Castranova, C. L. Geraci, and P. A Schulte. 2012. “Development of Risk-Based Nanomaterial Groups for Occupational Exposure Control.” Journal of Nanoparticle Research 14 (9):1029. doi:10.1007/s11051-012-1029-8.

 PubMed Web of Science ®Google Scholar

Liguori, B., S. F. Hansen, A. Baun, and K. A. Jensen. 2016. “Control Banding Tools for Occupational Exposure Assessment of Nanomaterials - Ready for Use in a Regulatory Context?” NanoImpact 2: 1–17. doi:10.1016/j.impact.2016.04.002.

 Web of Science ®Google Scholar

Drew, N. M., E. D. Kuempel, Y. Pei, and F. Yang. 2017. “A Quantitative Framework to Group Nanoscale and Microscale Particles by Hazard Potency to Derive Occupational Exposure Limits: Proof of Concept Evaluation.” Regulatory Toxicology and Pharmacology 89: 253–267. doi:10.1016/j.yrtph.2017.08.003.

 PubMed Web of Science ®Google Scholar

Fortino, V., and D. Grevo. 2017. Molecular Signatures and the Computational Classifier Capable to Predict ENM Safety (No. Nanosolutions D11.3), NANOSOLUTIONS. Helsinki.

 Google Scholar

Oosterwijk, M. T., M. L. Feber, and E. Burello. 2016. “Proposal for a Risk Banding Framework for Inhaled Low Aspect Ratio Nanoparticles Based on Physicochemical Properties.” Nanotoxicology 10 (6): 780–793. doi:10.3109/17435390.2015.1132344.

 PubMed Web of Science ®Google Scholar

Jensen, K. A, A. T. Saber, H. V. Kristensen, B. Liguori, A. CØ. Jensen, I. K. Koponen, and H. Wallin. 2014. NanoSafer vs. 1. 1 Nanomaterial Risk Assessment Using First Order Modeling. Japan: Nagoya.

 Google Scholar

Höck, J., T. Epprecht, H. Hofmann, K. Höhner, H. Krug, C. Lorenz, L. Limbach, et al., 2013. Guidelines on the Precautionary Matrix for Synthetic Nanomaterials, Version 3. Köniz, Switzerland: Federal Office of Public Health (FOPH).

 Google Scholar

Van Duuren-Stuurman, B., S. R. Vink, K. J. M. Verbist, H. G. A. Heussen, D. H. Brouwer, D. E. D. Kroese, M. F. J. Van Niftrik, E. Tielemans, W. Fransman, 2012. “Stoffenmanager Nano Version 1.0: A Web-Based Tool for Risk Prioritization of Airborne Manufactured Nano Objects.” The Annals of Occupational Hygiene 56, 1–17. doi:10.1093/annhyg/mer113.

 PubMedGoogle Scholar

Cornelissen, I. R., F. Jongeneelen, P. Broekhuizen, and F. V. Van, Broekhuizen. 2011. Guidance working safely with nanomaterials and nanoproducts, the guide for employers and employees [WWW Document]. http://www.etui.org/content/download/3553/39999/file/Guidance+on+safe+handling+nanomats&products.pdf

 Google Scholar

ANSES. 2010. Development of a control banding tool for nanomaterials (No. 2008-SA-0407), Expert Committee (CES) on Physical Agents, Control Banding.

 Google Scholar

Zalk, D. M., S. Y. Paik, and P. Swuste. 2009. “Evaluating the Control Banding Nanotool: A Qualitative Risk Assessment Method for Controlling Nanoparticle Exposures.” Journal of Nanoparticle Research 11 (7): 1685–704. doi:10.1007/s11051-009-9678-y.

 Web of Science ®Google Scholar

OECD. 2018. Case Study on the Use of Integrated Approaches for Testing and Assessment on Grouping and Read-across for Nanomaterials - genotoxicity of Nano-TiO2 (No. ENV/JM/MONO(2018)), Series on Testing & Assessment. Paris: OECD.

 Google Scholar

ECHA. 2017. Appendix R. 6–1: Recommendations for Nanomaterials Applicable to the Guidance on QSARs and Grouping, Version 01. Helsinki. https://doi.org/10.2823/884050.

 Google Scholar

Sizochenko, N., A. Mikolajczyk, K. Jagiello, T. Puzyn, J. Leszczynski, and B. Rasulev. 2018. “How the Toxicity of Nanomaterials towards Different Species Could Be Simultaneously Evaluated: A Novel Multi-Nano-Read-across Approach.” Nanoscale 10 (2): 582–591. doi:10.1039/C7NR05618D.

 PubMed Web of Science ®Google Scholar

Gajewicz, A., M. T. Cronin, B. Rasulev, J. Leszczynski, and T. Puzyn. 2015. “Novel Approach for Efficient Predictions Properties of Large Pool of Nanomaterials Based on Limited Set of Species: Nano-Read-across.” Nanotechnology 26 (1): 015701. doi:10.1088/0957-4484/26/1/015701.

 PubMed Web of Science ®Google Scholar

Liu, R., H. H. Liu, Z. Ji, C. H. Chang, T. Xia, A. E. Nel, and Y. Cohen. 2015. “Evaluation of Toxicity Ranking for Metal Oxide Nanoparticles via an in Vitro Dosimetry Model.” ACS Nano 9 (9): 9303–9313. doi:10.1021/acsnano.5b04420.

 PubMed Web of Science ®Google Scholar

Chen, R., Y. Zhang, F. D. Sahneh, C. M. Scoglio, W. Wohlleben, A. Haase, N. A. Monteiro-Riviere, and J. E. Riviere. 2014. “Nanoparticle Surface Characterization and Clustering through Concentration-Dependent Surface Adsorption Modeling.” ACS Nano 8 (9): 9446–9456. doi:10.1021/nn503573s.

 PubMed Web of Science ®Google Scholar

Zhang, H., Z. Ji, T. Xia, H. Meng, C. Low-Kam, R. Liu, S. Pokhrel, et al., 2012. “Use of Metal Oxide Nanoparticle Band Gap to Develop a Predictive Paradigm for Oxidative Stress and Acute Pulmonary Inflammation.” ACS Nano 6 (5): 4349–4368. doi:10.1021/nn3010087.

 PubMed Web of Science ®Google Scholar

George, S., T. Xia, R. Rallo, Y. Zhao, Z. Ji, S. Lin, X. Wang, et al., 2011. “Use of a High-Throughput Screening Approach Coupled with in Vivo Zebrafish Embryo Screening to Develop Hazard Ranking for Engineered Nanomaterials.” ACS Nano 5 (3): 1805–1817. doi:10.1021/nn102734s.

 PubMed Web of Science ®Google Scholar