Registration Dossier

Ecotoxicological information

Endpoint summary

Administrative data

Description of key information

Additional information

Discussion on methodology

In undertaking environmental risk assessments the best use of all available ecotoxicity data should be made. However, the assessment of ecotoxicity data for many petroleumproducts is complicated since several different test methods and procedures have been used. As petroleum products contain a mixture of substances with a range of solubilities a critical aspect with respect to interpreting the validity of ecotoxicity tests is how the test media is prepared. Although not always explicitly stated most of the data generated in the period up to the early 1990s originated from experiments in which a "water soluble fraction" (WSF) was tested. WSFs are prepared by mixing the petroleum product with the aqueous test medium (e. g. 25 to 50 mL product with 1 L of medium). After mixing the test solutions are then allowed to stand, the aqueous phase is separated and dilutions of this medium are used in testing the species under study. The results are expressed either as (a) the dilution, or % WSF, or (b) the concentration of dissolved hydrocarbons expressed in mg/L (CONCAWE, 1992). A disadvantage with these WSF studies is that it is not possible to convert the quoted result to the amount of product that must be added to a given volume of aqueous medium to produce the effect.

The problems of preparing test media for oil products were recognised in the early 1990s. As a consequence the recommended method, which enables ecotoxicity assessments of petroleum products to be interpreted, was to determine the amount of test substance that must be equilibrated with the test medium to produce a specified level of effect. This is the so-called "loading rate" or water accommodated fraction (WAF) methodology as developed by Girling et al. (1992) and reported in CONCAWE (1992). Even with these laboratory based studies there are doubts about their value in the context of risk assessment owing to the fact that once a petroleum product is released to the environment its constituent substances will partition to the various compartments (water, sediment, soil and air) in accordance with their physico-chemical properties. The assumption being that in the receiving environment the substances will be degraded and transformed in accordance with their individual susceptibilities to physical, chemical and biological degradation processes and will exhibit effects in accordance with their individual toxic potencies.

Discussion on mechanisms of toxicity and PNEC derivation
In an attempt to better understand the potential for adverse effects of a product, the effects of a product's constituent substances (hydrocarbon blocks) can be integrated in such a way that an overall assessment of their combined effects can be made. For the assessment of toxic effects it is important that the method of integration meets the assertion that effects can only be integrated for substances that share the same mode of toxic action. All components of petroleum products exhibit non-polar narcosis effects on organisms.

Under ideal circumstances a PNEC for a hydrocarbon block would be derived from ecotoxicological test data for one or more components that are representative of that block. The TGD sets out how this can be done either by applying an Assessment Factor to the lowest acute ecotoxicological effect or chronic no effect concentration or by applying statistical extrapolation methods to a number of data points. For petroleum products this was not a practical option since the majority of its mass is comprised of chemical components that cannot be accurately described by a chemical structure (and which may not have a unique CAS number) and for which there is an absence of ecotoxicological data. Under such circumstances the only practical option is to estimate a PNEC using a relationship between physico-chemical descriptors of a component or a hydrocarbon block and concentrations resulting in ecotoxicological effects or absence of an effect. This is the hypothesis encompassed by the Target Lipid Model (TLM) described by McCarthy et al. (1991).

The theory underpinning the TLM is that the concentration of a substance in a lipid that is responsible for the onset of a non-polar narcosis effect is the same when expressed on a molar basis for a range of taxonomic groups e. g. fish, invertebrates and algae. Consequently the toxic potency of a substance depends upon its capacity to achieve the threshold concentration within an organism. There are a number of variables that determine this capacity, key of which are the solubility of the substance in water and lipid and its molecular size. In an application of the theory, DiToro et al. (2000) have published a non-polar narcosis-based QSAR for predicting the aqueous concentration of a hydrocarbon substance that induces a specified level of biological effect. The QSAR relates biological effect to the log Kow of the substance. Log Kow is a function of the solubility of a substance in water and lipid (octanol) but is limited by molecular size because large molecules cannot pass through biological membranes.

In the absence of measured ecotoxicity data for a substance the TLM and associated QSARs provide a theoretical basis for predicting the ecotoxicity of a substance. By extension of the theory it should also be possible to evaluate the toxicity of a mixture of substances provided that they have the same mode of toxic action. McGrath et al. (2004) have validated the theory by characterising the aquatic toxicity of six gasoline blending streams and have showed that predicted and measured toxicity were in good agreement.

Having established procedures that enable the toxicity of a mixture of hydrocarbons to be predicted, McGrath et al. (2004) have also utilised statistical theory developed by a number of workers to define an acute species sensitivity distribution for narcotic chemicals. A relationship has been established enabling the concentration of a hydrocarbon substance to be determined that will affect a specified proportion of the species present in a community. By setting the proportion to a notional low level (e. g. 5%), a hazard concentration (HCx where x is the proportion that might be affected i. e. 5%) is obtained. The HCx has similarities with a hazard concentration derived by applying statistical extrapolation procedures described in the TGD to a set of test substance data. It can also be considered analogous to, and used for risk assessment in the same way as, a PNEC derived by applying an Assessment Factor (AF) specified in the TGD to a lowest acute EC50 or LC50 value in a data set.

Short term toxicity to fish:

Information on short term toxicity to fish was available for vacuum gas oils, hydrocracked gas oils and distillate fuels. The 96h LL50 for freshwater fish is 21 mg/L. This data is from Shell report 6304 (1996). This is the most conservative value from the two key studies. This study is considered reliable (2), it is a GLP compliant, near guideline study, with minor restrictions in design and/or reporting but otherwise adequate for assessment.

A wide range of values have been reported for the supporting studies which cannot be compared directly because the studies vary in design, test species and different endpoints have been investigated. Although the Exxon 142958 (1998) study reported a lower LL50, this was not selected as a key study because insufficient data was available to confirm that the test validity criteria were met.

Long term toxicity to fish:

The aquatic toxicity was estimated using the PETROTOX computer model, which combines a partitioning model (used to calculate the aqueous concentration of hydrocarbon components as a function of substance loading) with the Target Lipid Model (used to calculate acute and chronic toxicity of non-polar narcotic chemicals). PETROTOX computes toxicity based on the summation of the aqueous-phase concentrations of hydrocarbon block(s) that represent a petroleum substance and membrane-water partition coefficients (KMW) that describe the partitioning of the hydrocarbons between the water and organism. The estimated freshwater fish NOEL (No Observed Effect Level) value is 0.083 mg/l based on mortality.

Short term toxicity to aquatic invertebrates:

Information on short term toxicity to aquatic invertebrates was available for vacuum gas oils, hydrocracked gas oils and distillate fuels. The 48 h EL50 for Daphnia was 68 mg/L. This data is from Shell report 6304 (1996). This is the most conservative value from the two key studies. This study is a GLP compliant, guideline study (reliability 2).

A wide range of values have been reported for the supporting studies, they have not been considered for use as key studies because insufficient information was supplied to determine whether they meet the validity criteria of the test methods.

Long term toxicity to aquatic invertebrates:

The aquatic toxicity was estimated using the PETROTOX computer model, which combines a partitioning model (used to calculate the aqueous concentration of hydrocarbon components as a function of substance loading) with the Target Lipid Model (used to calculate acute and chronic toxicity of non-polar narcotic chemicals). PETROTOX computes toxicity based on the summation of the aqueous-phase concentrations of hydrocarbon block(s) that represent a petroleum substance and membrane-water partition coefficients (KMW) that describe the partitioning of the hydrocarbons between the water and organism. The estimated freshwater invertebrate NOEL (No Observed Effect Level) value is 0.2 mg/l based on immobility and numbers of live young produced per adult by Day 21.

Toxicity to aquatic algae and cyanobacteria:

Information on short term toxicity to aquatic algae and cyanobacteria was available for vacuum gas oils, hydrocracked gas oils and distillate fuels.

The Shell (6304) study reported a 72 h ErL50 value of 22 mg/L for diesel fuel. Clark (2003) reported the same 72 h value when ultra low sulphur diesel was tested (EL50 = 22 mg/L). The Shell (6304) study has been chosen to represent this endpoint. It is a reliable (2), GLP compliant, near guideline study with no restrictions, it is fully adequate for assessment.

Toxicity to microorganisms:

The aquatic toxicity was estimated using the PETROTOX computer model, which combines a partitioning model (used to calculate the aqueous concentration of hydrocarbon components as a function of substance loading) with the Target Lipid Model (used to calculate acute and chronic toxicity of non-polar narcotic chemicals). PETROTOX computes toxicity based on the summation of the aqueous-phase concentrations of hydrocarbon block(s) that represent a petroleum substance and membrane-water partition coefficients (KMW) that describe the partitioning of the hydrocarbons between the water and organism. The estimated 40-hr EL50 value forTetrahymena pyriformis, one of the most sensitive microorganism species, is >1000mg/L and the estimated NOEL is 3.217 mg/L.

Some information for this category has been generated using the models PETROTOX and/or SPARC. The QMRFs for PETROTOX and SPARC are attached in IUCLID Section 13, with the associated QPRF.