Registration Dossier

Diss Factsheets

Ecotoxicological information

Short-term toxicity to fish

Currently viewing:

Administrative data

Link to relevant study record(s)

short-term toxicity to fish
Type of information:
read-across from supporting substance (structural analogue or surrogate)
Adequacy of study:
key study
Justification for type of information:
The REACH registration of silver (powder and massive forms of zero-valent, elemental, silver) is underpinned, in common with other metals, by a read-across (or analogue) approach from the properties of the free ion. This principle of read-across from the free ion has been extended to also include nanosilver.
The scientific validity of read-across from the hazard properties of ionic silver (source) to nanosilver (target) under REACH is underpinned by both theoretical and empirical considerations.
The theoretical basis for the use of ionic silver data as the foundation of the risk assessment of nanosilver is based on the premise that the free metal ion (Me+) is the most toxic metal form/species (Starodub et al. 1987). This consideration was implicit in the development of the Free Ion Activity Model [FIAM] (Morel 1983, Paquin et al. 2002, Campbell 1985, Brown and Markich 2000) and, more recently, the Biotic Ligand Model [BLM] (Paquin et al 2002, Niyogi and Wood 2004) that has underpinned the risk assessment of several metals (e.g. Cu, Ni, Zn) under the Existing Substance Regulations and REACH; and most recently the development of the Environmental Quality Standard (EQS) for nickel and nickel substances under the Water Framework Directive (WFD). When considered on an equal mass basis ionic silver would therefore be expected to have greater toxicity than nanosilver simply on the basis that silver ions are released over time from the surface of particles (via oxidative dissolution). As the properties of nanosilver are read-across directly from ionic silver (not just to the fraction of silver ions released from nanosilver), this read-across is also expected to introduce considerable precaution into the hazard component of the risk assessment of nanosilver as all nanosilver, irrespective of coating, morphology or particle size distribution is assumed to behave similarly to the free ion.
This theoretical consideration has been tested by conducting a comprehensive review of the available scientific literature for nanosilver, with particular emphasis on the comparative effects on REACH relevant biotic systems (REACH information requirement) of ionic silver and nanosilver. This review is described for each endpoint in subsequent sections of the report ‘Nanosilver: read-across justification for environmental information requirements’ (attached in IULCID Section 13) and is summarised below. Furthermore, this theoretical consideration was confirmed by a specific ecotoxicity testing programme undertaken by the EPMF following the silver substance evaluation and designed to compare the effects of the smallest nanosilver form registered under REACH (‘Nano 8.1’ or ‘Silberpulver Typ 300-30’) and silver nitrate to algae, Daphnia (long-term) and soil microorganisms.

In terms of ecotoxicity, nanosilver on an equal mass basis has been found to be significantly less toxic than ionic silver for the majority of REACH endpoints and of equivalent toxicity for some others. None of the empirical information available suggests that nanosilver is consistently more hazardous than ionic silver on an equivalent mass basis, or that “nano-specific” effects that would prejudice the validity of read across from ionic silver to nanosilver occur. In addition, with very limited exceptions which are described further in the report ‘Nanosilver: read-across justification for environmental information requirements’, none of the available data suggested consistent relationships between particle morphology, size, particle size distribution (raw or agglomerated) or coating (surface treatment) and effects.
Notter et al. (2014) presents a meta-analysis of published EC50 values for ionic silver and nanosilver. The authors demonstrate that almost 94% of acute toxicity values assessed for freshwater, seawater and terrestrial systems using algae, annelid, arthropoda, bacteria, crustacea, fish, nematoda, plant, protozoa and rotatoria show that the nanoform of silver is less toxic than the dissolved metal (when normalised for total metal concentration).
In addition, a specific ecotoxicity testing programme designed to compare the effects of the smallest nanosilver form registered under REACH (‘Nano 8.1’ or ‘Silberpulver Typ 300-30’) and silver nitrate to algae, Daphnia (long-term) and soil microorganisms was undertaken following the silver substance evaluation. This demonstrated that the nanoform of silver is less toxic than ionic silver (based on EC10 and EC50 values for total, ‘conventional’ dissolved (<0.45μm) and ‘truly’ dissolved (<3kDa) silver). Therefore, taking the full body of evidence into account, the read-across use of toxicity values from ionic to nanosilver as a ‘worst case’ approach is justified and scientifically defensible for environmental endpoints. The Substance Evaluation Conclusion document for silver agreed with this conclusion for the nanosilver forms covered in the REACH dossier (RIVM 2018).

Short-term toxicity to fish
Published data from several short-term toxicity to fish studies using various sizes of nanoparticles and coating types are included in the REACH dossier as Endpoint Study Records. All of the nanosilver LC50 values from these studies are at least an order of magnitude higher than the 96h LC50 of 1.2 μg/L for Pimephales promelas (Bielmyer et al. 2007) for ionic silver. Together with the theoretical basis for read-across based on the free-ion, this supports the use of ionic silver as the ‘worst case’ basis to read across properties to nanosilver, irrespective of any data available describing morphology, size, size distribution or coating. A summary of these supporting studies is available under Section 4.1.3 of the report ‘Nanosilver: read-across justification for environmental information requirements’ (attached in IUCLID Section 13).

For further information and data matrix see 'CSR Annex 9 - Read Across Justification Nanosilver ENV_SUMMARY_200706' attached in IUCLID section 13.
Reason / purpose for cross-reference:
read-across source
96 h
Dose descriptor:
Effect conc.:
1.2 µg/L
Nominal / measured:
not specified
Conc. based on:
Basis for effect:
mortality (fish)
Remarks on result:
other: 95% CI 0.72 - 2.07 µg dissolved Ag/L. Fish aged 1-4 day at start of test.
Details on results:
Other LC50 values: 3.37 (7 d old), 5.9 (27 d old), 10.4 (41 d old) µg/L dissolved Ag.
Results with reference substance (positive control):
Not applicable
Reported statistics and error estimates:
Trimmed Spearman Karber analysis was used to estimate 48- and 96-h LC50s.
Validity criteria fulfilled:
not applicable
The 96 hour LC50 for fish aged 1-4 days was 1.2 µg dissolved Ag/L. Other LC50 values were: 3.37 (7 d old), 5.9 (27 d old), 10.4 (41 d old) µg dissolved Ag/L. Silver toxicity varied with age and size of P. promelas. As fish aged and mass increased, 96 hour LC50 values increased (organism sensitivity decreased).
Executive summary:

The toxicity of silver toP. promelaswas assessed in a flow through test. For fish aged 1 -4 days, the 96 hour LC50 is 1.5 µg/L dissolved Ag. Other LC50 values were: 3.37 (7 d old), 5.9 (27 d old), 10.4 (41 d old) µg dissolved Ag/L. Silver toxicity varied with age and size of fish. As fish aged and mass increased 96 h LC50 values increased (organism sensitivity decreased). The GLP status of this study is unknown and it is unclear which guideline was followed. However, there is sufficient information for an assessment to be made and the study is considered suitable for use for this endpoint.

Description of key information

Key value for chemical safety assessment

Additional information

Summary of available data for uncoated and coated nanosilver

Reliable and relevant data on the short-term toxicity of uncoated and coated nanosilver to fish are available from 10 studies (Griffitt et al. 2008, Kennedy et al. 2010, Bilberg et al. 2012, Farmen et al. 2012, Kashiwada et al. 2012, George et al. 2012, Cunningham et al. 2013, Hoheisel et al. 2012, Kim et al. 2013, Wang et al. 2012). These studies comprise effects assessment on four species (Danio rerio, Pimephales promelas, Salmo salar and Oryzias latipes) after exposure to various sizes of nanoparticles and coating types. Across these studies a total of 46 LC50 values are available, from exposures ranging from 48 hours to 7 days. Available studies are focussed on the effects of spherical nanoparticles, although the characterisation of several of the test materials used by Kennedy et al. (2010), Bilberg et al. (2012) and Cunningham et al. (2013) describe the presence of small rods or triangular particles. All studies were conducted in freshwater media.

The particle size of raw nanomaterials across the studies ranged from 3.6 to 225.3 nm. However, the majority of studies report effects associated with primary particle sizes between 20 and 149 nm (10th to 90th percentile, respectively). It is worth noting that some of these materials would not be considered as nanomaterials according to the EU definition as their median primary particle size was greater than 100 nm.

The extent of aggregation/agglomeration by primary particles in experimental media or stock solutions differed between studies but in almost all instances authors report that some degree of aggregation/agglomeration in test systems occurs, increasing the size of particles in environmental media. The characterisation results for nanosilver in test systems report particle sizes after aggregation/agglomeration of between 3.8 and 3,078 nm, with the majority of studies reporting particle sizes between 29.6 and 310 nm (10th to 90th percentile, respectively). Kennedy et al. (2010), report that PVP and citrate coated nanosilver particles appear to form smaller aggregates in test systems than uncoated or EDTA coated nanosilver materials.

LC50 values range from 9.0 µg/L for ~30 nm particles in P. promelas after 48 hours exposure (Kennedy et al. 2012) to 44,780 µg/L for similar size nanoparticles in Danio rerio after a 120 hour exposure (Kim et al. 2012). The 10th to 90th percentile range of LC50 values for all forms of nanosilver tested for acute fish toxicity is 38 to 42,490 µg/L. All of the results from nanosilver studies are at least an order of magnitude less sensitive than the 96 hour LC50 of 1.2 µg/L used as the key data for this endpoint, based on exposure of one day old Pimephales promelas to ionic silver in dechlorinated tap water (Bielmyer et al. 2007).

In addition, where studies undertook a comparative assessment of nanosilver and ionic silver within their own study designs (Griffitt et al. 2008, Kennedy et al. 2010, Bilberg et al. 2012, Farmen et al. 2012, Cunnigham et al. 2012, Hoheisel et al. 2013, Kim et al. 2013, Wang et al. 2012) nanosilver was less sensitive than ionic silver on six occasions and equally as sensitive, at least for a single type of nanosilver, on two occasions.

There is no statistically significant correlation between smaller particle size and increased toxicity (Kendall test, p>0.05) between either raw or media particle size and LC50. An additional analysis excluding LC50 values from studies with materials with primary particle size greater than 100 nm also resulted in no statistically significant correlation (Kendall test, p>0.05) between either raw or media particle size and LC50.

When the LC50 values from materials with different coatings were compared using a non-parametric ANOVA procedure (Kruskal-Wallis, p<0.05) a statistically significant difference between coating material and LC50 value was identified. PVP and citrate coated materials appear to have lower toxicity than other coating materials and uncoated materials. However, this cannot be confirmed using a statistical post-hoc test as the variance and normality of the different groups is homogenous.

This observation is supported by data from Kevin Kwok (pers comm) on the acute toxicity (96 hour LC50) of nanosilver particles of various sizes and coatings (gum arabic, PVP, citrate-tannic acid, citrate-glutathione and citrate) to Japanese medaka (Oryzias latipes). These data describe differential toxicity on the basis of coating material, but with limited influence of particle size. The relative toxicity of nanomaterials, from most to least toxic, was as follows: gum arabic < PVP < citrate-tannic acid < citrate-glutathione < citrate < PVP (nanoamor). The toxicity of PVP coated nanoparticles was variable, but appeared to be dependent on the manufacturer of the nanomaterial, possibly due to differences in the chain length of PVP polymer molecules used in the coating, although additional information would be required to confirm this.

Critically, irrespective of coating of particle size, all forms of nanosilver were significantly less toxic than ionic forms of silver. Of particular significance are the conclusions for Hoheisel et al. (2012) (all authors from the US-EPA) who state that, based on their findings, regulatory approaches based on the toxicity of ionic silver to aquatic life would not be under protective for environmental releases of nanosilver.

Exposure duration is considered as a critical factor in ecotoxicity tests with metals as sensitivity is known to be influenced by exposure duration (greater toxicity after longer exposure). Some of the available acute data for nanosilver reported here are from exposure durations of 48 hours, which is significantly shorter than the standard 96 hour exposure exposure duration for acute fish toxicity. However, these data have not been excluded from this analysis on the basis of exposure duration as they are otherwise “well documented and scientifically acceptable” (Klimisch et al. 1996). As the majority of data for nanosilver from this endpoint are from studies of at least 96 hours exposure, with many of longer duration, this is not considered to have adversely affected the integrity of the comparison between the relative sensitivity of ionic and nanosilver for this endpoint.