Registration Dossier

Environmental fate & pathways

Endpoint summary

Administrative data

Description of key information

Environmental data for tungsten metal and sodium tungstate are presented in this section. The soluble species released by tungsten oxide are expected to be similar for each of the substances, and are thus expected to behave similarly in the environment. However, the amount of soluble species resulting from tungsten metal and sodium tungstate is different, with sodium tungstate being much more soluble.

Aquatic Bioaccumulation:

No information on the bioconcentration or accumulation in aquatic organisms was found for tungsten oxide.

Bioconcentration is the tendency of materials to concentrate directly from water in a living organism over time. There is no testing performed according to standard methodology in the published literature regarding bioconcentration of tungsten compounds in general or tungsten metal specifically, in aquatic organisms. However, in a static renewal, toxicity test onPoecilia reticulatetesting sodium tungstate, Strigul et al. (2010) measured tungsten uptake in 5 fish- 2 controls and 3 exposed to 7.5 g/L (nominal sodium tungstate concentration). The fish from the test group had died within the first 24 hours of exposure. The BCF was calculated as the ratio of tungsten concentration in fish tissue (in mg W per kg wet or dry) to tungsten concentration in water (in mg/L). BCF was calculated on both wet and dry weight of fish. Wet weight BCF for the test substance was calculated as 0.29 +/- 0.94 L/kg. Dry weight BCF for the test substance was calculated as 1.57 +/- 0.5 L/kg. These BCFs are low, indicating little to no immediate accumulation even at toxic exposure levels.

The most prevalent bioavailable form of tungsten is the soluble tungstate ion. However, because tungsten has a significant affinity for adsorption onto soils and stream or river sediments, levels in proximal natural waters are relatively much lower than the surrounding sediment and soil (see section 4.2.1 for more information). The extent to which tungsten compounds would release bioavailable tungstate ions into the aquatic environment is furthermore dependent on many factors including dissolved organic carbon (DOC), pH, and water hardness (Bednar et al., 2009). These data indicate that more alkaline waters will potentially possess much higher levels of bioavailable tungsten when exposed to the same amounts of tungsten than more acidic waters. A test performed using tungsten metal powder, according to the Transformation/Dissolution Protocol (UN GHS, 2007) showed that, under simulated natural conditions, after seven days, and at a loading rate of 100 mg/L, approximately 16509µg/L of tungsten ion is released at a pH of 8.5 (CANMET-MMSL, 2010). Thus, even at a relatively high pH, the magnitude of release would relatively low at environmentally relevant loadings. Furthermore, the median calculated tungsten partition coefficient for water-sediment of 140000 L/kg (Salminen (Ed) et al., 2005) indicates that upon reaching the water compartment, much tungsten is removed via adsorption to the sediment. Overall, it is unlikely that substantial exposure, and consequent uptake, would result from environmentally-relevant loadings.

Another important concern for the bioaccumulation/bioconcentration of metals is methylation. Methylation of metals (i.e. mercury) can allow metals to passively cross membranes and accumulate without homeostatic regulation. There is currently no evidence of methylated species of tungsten in the natural environment.

It is also important to consider active uptake of bioavailable tungsten. According to Adams and Chapman (2007) “Most metal species that form in aquatic solutions are hydrophilic and do not permeate the membranes (typically gills) by passive diffusion….uptake of metals is dependent on the presence of transport systems that provide biological gateways for the metals to cross the membrane.” Therefore, most metals enter organisms through active transport via transport proteins specific to that particular metal, as occurs with essential metals. Though tungsten is a non-essential metal, it is possible for metals such as tungsten, which mimic essential metals such as molybdenum, to be taken up. This has been demonstrated in studies examining chicks and rats fed sodium tungstate supplemented diets, which have demonstrated that tungsten may act as a competitive inhibitor of molybdenum uptake (Higgins et al., 1956). This phenomenon has not been studied in aquatic organisms; however, organisms such as fish have metabolic mechanisms to eliminate metals that are taken up or even to acclimate to metal exposure by decreasing metal uptake (McDonald and Wood, 1993 in Adams and Chapman, 2007).

Terrestrial Bioaccumulation:

Relatively low bioaccumulation of tungsten is observed in sunflower leaves at soil concentrations of 3900 mg W/kg soil, with calculated concentration factors plateauing at approximately 0.05 (Johnson et al., 2009). Tungsten concentrations factors calculated for ryegrass were higher and ranged from 56.1-0.202 (Strigul et al., 2005). However, it should be noted that background levels of tungsten in the collected soils used for testing were not determined prior to testing. Tungsten concentrations measured in earthworm tissue ranged from 1.52-193.2 mg/kg wet weight in soils with tungsten concentrations of 10-10000 mg/kg soil, respectively (non-aged soil) (Strigul et al., 2005). Additionally, tungsten concentrations of 10 and 10000 mg/kg soil yielded earthworm tissue concentrations of 3.45 and 25.9 mg/kg wet weight, respectively (Strigul et al., 2005). Using these paired concentration data the BCFs for earthworms in non-aged soils ranged 0.152-0.019 and BCFs for aged soils ranged 0.345-0.00259. However, it should be noted that background levels of tungsten in the collected soils used for testing were not determined prior to testing. Tungsten is not expected to bioaccumulate in terrestrial organisms.

Mobility in soil:

Two tungsten partitioning and mobility studies (Griggs et al., 2009 and Bednar et al. 2008) were selected to derive soil partition coefficients for tungsten compounds. These studies provide Kd estimates with various soil and solution characteristics and therefore provide a range of values that may be found in the environment. In addition they met the quality characteristics specified in the REACH guidance 7.13-2 (pg. 16) and followed methods equivalent to the suggested test method (OECD 106 batch equilibrium method). They are briefly summarized below.

Griggs et al., 2009: The authors measured Kd values using the batch sorption method. Initial batch slurries were prepared with 10 g of soil and 100 ml of sodium tungstate solution. Soil-metal systems were prepared in triplicate using each of three natural silty sand soils and each of 10 stock concentrations. Supernatant was sampled at 24 hours and at 100 days. The Freundlich model was used. The authors observed dynamic sorptive behavior in tungsten and suggest that in Kd studies, a longer equilibration time may provide a more accurate reflection of tungsten mobility in subsurface environments.

Bednar et al. 2008: Geochemical parameters are determined for tungstate species in a model soil that describe the potential for tungsten mobility. The Freundlich model was used. Soluble tungsten leached from a metallic tungsten-spiked soil after six to twelve months aging reached an equilibrium concentration >150 mg/L within 4 h of extraction with deionized water. Partition coefficients (Kds) determined for various tungstate and polytungstate compounds in the model soil suggest a dynamic system in which speciation changes over time affect tungsten geochemical behavior. Partition coefficients for tungstate and some poly-species have been observed to increase by a factor of 3 to 6 over a four month period, indicating decreased mobility with soil aging.

The Kd values from these studies along with information about the characteristics of the soils and solutions used are presented in Table 8. In accordance with the REACH guidance (pg. 17), a probability distribution was fit to the data using statistical software (SYSTAT, version 12). The best fitting distribution was a lognormal (5.160,1.076). The goodness of fit statistic for this distribution (Shapiro Wilks) 0.941 (p= 0.277) indicating that a lognormal distribution cannot be rejected. Non-parametric estimates were also calculated for the sake of comparison; however, they were not carried onto the risk assessment. The lognormal results of Median=174, 10%tile=44, 90%tile=692 will be passed forward to the risk assessment.

Mobility in sediment:

Sediment partition coefficients (Kd) were derived for tungsten using paired field-measured concentrations in sediment and water from locations throughout Europe. Stream sediment and water concentrations were obtained from the FOREGS Geochemical Baseline Database (Salminen et al, 2005) (http://www.gsf.fi/foregs/geochem/). The sediment and water samples were collected simultaneously allowing the measurements to be paired by the location identifiers in the database. This resulted in a dataset of 800 paired sediment and water concentrations. Sediment partitioning coefficients were calculated for each data pair as follows:

Kd = Cs / Caq

where Cs = total concentration of test substance in the solid phase (mg/kg) and Caq = concentration of test substance in aqueous phase (mg/L).

The detection limit for tungsten in water is reported in the database as 0.002 ug/L and the accompanying documentation states that non-detected concentrations were reported at half the detection limit (0.001 ug/L). Of the 800 data pairs, 106 included non-detected water concentrations. Being in the denominator of the equation, small differences in the reported water concentrations have a large effect on the Kd estimates. Therefore, assessment of the distribution of the Kd values was conducted both including and excluding the 106 pairs where 1/2 DL (0.001 ug/L) was used for the water concentration.

The distribution of the 800 pairs could not be readily fit to any of the commonly used continuous probability distributions (normal, lognormal, etc.) however due to the large number of data pairs, it is very well characterized and therefore it is appropriate to report empirically derived percentiles for this distribution: Median=140,000, 10%tile= 28,395, 90%tile=700,000 will be passed forward to the risk assessment.

Additional information