Specific notes on ecotoxicity data

Dealing with freshwater and marine ecotoxicity data 7.3.1

According to the EQS-guidance the treatment of freshwater and marine toxicity data (i.e. species living and tested in water with salinity >0.5 ‰) will be changed. Previously, these datasets were kept separated and

the freshwater QSfw, eco was based on freshwater species only. The approach of the EQS-guidance is also

adopted for the drainage ditch risk assessment and already briefly presented in Section 6.2.

Additional data can be available for a variety of species, being either freshwater or marine species. The presence of marine data is generally less relevant for PPPs, but the option to include literature data will probably generate more data. Furthermore, studies on marine species are part of the standard dossier for registration in the USA, and will thus sometimes be available in the EU-dossier too. For the purpose of this report, marine species are defined as living and tested in brackish or saltwater (salinity >0.5 ‰). The question how to deal with these data has been subject of discussion within the framework of the WFD, but the

considerations made within that context are applicable in general. The following is taken largely from a document that was prepared by the Netherlands as a background document to the WFD-guidance (EC, 2011).

Both the TGD (EC, 2003) and its revision under REACH (ECHA, 2008), recommend that for plant protection products (PPPs) toxicity data for freshwater and saltwater organisms should not be pooled for PNEC derivation. The reasoning in both documents is equal: 'within trophic levels differences larger than a factor of 10 were shown for several metals and pesticides indicating that for these compounds fresh water and saltwater data should not be pooled for hazard assessment and PNEC derivation'. ECETOC (2000) is given here as a single reference to a background document. The methodological choice made in the two guidance documents (TGD, REACH) is not clearly underpinned. The a priori separating of aquatic toxicity data for PPPs has been adopted by Lepper (2005). Attempts to retrieve the ECETOC (2000) publication failed, although part of this work was probably published in ECETOC Technical report 82 (2001), and by Hutchinson et al. (1998), Leung et al. (2001) and Wheeler et al. (2002).

Maltby et al. (2005) and Brock et al. (2008) pointed out that for pesticides with a specific mode-of-action, it is rather the taxonomic group than the place in the food chain or food web (trophic level) that determines the sensitivity. In their study, of the ten insecticides of which SSDs were compared based on acute data, no

significant differences between HC5 values for freshwater and saltwater taxa of the same sensitive taxonomic

group were found. For two compounds (out of ten; permethrin and chlorpyrifos) differences in HC5 could be

established when arthropods were compared, but this difference could not be demonstrated anymore after selecting crustaceans as sensitive taxonomic group. Maltby et al. (2005) conclude that freshwater and saltwater toxicity data can be combined, but that it is important to be aware of differences in taxonomic position and consequences for threshold concentrations.

Solomon et al. (2001) showed that differences (fresh vs. marine) were observed when comparing acute data for permethrin. For fenvalerate a difference in sensitivity was only observed when data for arthropods (insects and crustaceans) and fish were compared. When comparing complete datasets (including e.g. algae,

Mollusca), the 10th _{or 5}th _{percentile of the freshwater and marine datasets were similar. Note that Solomon}

et al. (2001) did not make a distinction between crustaceans and insects in comparing marine and freshwater toxicity data for arthropods. Leung et al. (2001) showed a difference in acute sensitivity to chlordane. However, the authors pointed out that 'there is considerable potential for freshwater to saltwater prediction'. They state that differences between the taxonomic compositions of the data sets should be considered. Wheeler et al. (2002) have compared SSDs for pesticides based on acute toxicity data and reported differences in HC5 values ranging from a factor of 2 to 12 for five (out of seven) of the compounds. They

concluded that for pesticides, freshwater data could be used for saltwater risk assessments, but with - possibly - an additional 'modest' safety factor depending on how the sensitive taxonomic groups are represented in the saltwater data set.

A draft report on the SETAC 2006 workshop on quality standards setting (Anonymous, 2007) reports on this topic that: 'Overall the lack of data hampers a sound and definitive comparison, but current scientific opinion is that there is no systematic bias in sensitivity between freshwater and marine species, provided similar tests and endpoints are involved.' Also: 'if there is no indication of differential sensitivity to a particular substance between freshwater and marine organisms, it may be appropriate to combine both datasets in a single SSD, although any resulting quality standard should be regarded as tentative.' Please note that construction of SSDs in quality standard derivation occurs only for very data rich compounds.

Based on the above presented information from the literature, the EQS-guidance states that a statistical evaluation should be performed to test whether or not data from freshwater and marine species should be treated separately. Where there are sufficient toxicity data in both the freshwater and marine datasets to enable a statistical comparison, the following procedure should be followed. The null hypothesis is that freshwater and saltwater organisms do not differ in their sensitivity to the compound of interest; i.e. they belong to the same statistical population:

1. All freshwater data are collected and tabulated (note: this data set contains one toxicity value per species, see Footnote 1 in Section 6.2 for an explanation). Next, a logarithmic transformation of each of these toxicity values is performed.

2. All marine data are collected and tabulated (note: this data set contains one toxicity value per species). Next, a logarithmic transformation of each of these toxicity values is performed.

3. Using an F-test, determine whether the two log-transformed data sets have equal or unequal variances. Perform the test at a significance level (α) of 0.05.

4. A test for differences between the data sets e.g. a two tailed t-test where the data are normally distributed (with or without correction for unequal variances, depending on the results of step 3), is performed. Perform the test at a significance level (α) of 0.05 .

5. Especially for compounds with a specific mode-of-action, it is important to identify particularly sensitive taxonomic groups and perform a separate statistical analysis for this specific group. If enough data are available to make a comparison for individual or related taxonomic groups (e.g. crustaceans, arthropods, fish, vertebrates), this may help to determine if there are differences between saltwater and marine species. Note that there are only few marine insects.

In those cases where there are too few data (either freshwater or marine) to perform a meaningful statistical comparison and there are no further indications (spread of the data, read-across, expert judgement) of a

difference in sensitivity between freshwater vs. marine organisms, the data sets may be combined for QSfw, eco

derivation. The notes given in Section 6.2 on the use of marine mesocosms also apply to QSfw, eco derivation. In

general, it is proposed to use marine mesocosm data only in addition to freshwater data. In practice, this means that a single marine mesocosm without any equivalent freshwater studies will only be used as supportive evidence, but not as the sole basis for the QSfw, eco.

Special considerations on micro-organisms 7.3.2

According to the EQS guidance (EC, 2011), data for bacteria representing a further taxonomic group may only be used if non-adapted pure cultures were tested. Furthermore, studies with bacteria (e.g. growth tests) are

regarded as short-term tests. Consequently, unlike for algae, NOECs or EC10 values derived from bacterial

studies may not be used in the derivation of the AA-EQS using assessment factors. EC50 values from bacterial

tests may be used as additional acute data.

The EQS-guidance probably refers to bacteria tests with a short contact time in which a generic parameter such as CO2 evolution is measured. If, however, a reliable bacteria test is available that is comparable to

an algae test in terms of duration and endpoint (i.e. 72 hours and specific growth rate), there is scientific evidence to include the endpoint in the dataset. The same principle applies to toxicity data using protozoans. For the purpose of EQS-derivation for PPPs within the context of the present report, it is therefore proposed to accept NOECs for bacteria and protozoans as chronic endpoints, if obtained in a comparable way as those for algae.

The EQS guidance does not make reference to fungi as a specific taxonomic group. As pointed out previously (see Section 6.4.3.3), data on fungi are considered relevant for fungicide risk assessments and may become available in the (near) future. If growth tests with fungi are present, it is advised for the time being to treat the data similarly to algae, i.e. include the EC50 for the acute dataset and the NOEC in the chronic dataset. It was also noted in Section 6.4.3.3 that the kingdom of fungi is diverse. The selection of relevant species for which standardised ecotoxicity tests may be developed is therefore identified as a further research need. In addition, more research is needed into the life-span and generation time of aquatic fungi, to determine whether or not short-term tests can be used to derive chronic endpoints. These points should be considered when updating the EQS-guidance, and are therefore taken forward to Chapter 9.

Endocrine disruptors 7.3.3

When there are indications that a substance may cause adverse effects via disruption of the endocrine system of mammals, birds, aquatic or other wildlife species, the assessor should consider whether the AF that is normally applied for a certain combination of data (see 8.2) would be sufficient to protect against effects caused by such a mode-of-action, or whether a larger AF is needed. Since PPPs with endocrine disrupting properties will not be authorised, this is less relevant for the present report, although the way in which endocrine disruption should be evaluated under the PPP-regulation is still under discussion.

Use of non-testing methods to reduce uncertainty 7.3.4

Emphasis is placed on experimental toxicity data for deriving a QSfw, eco. However, non-testing methods (e.g.

QSARs, read-across methods) are also available which can be used to predict toxicity of certain organic chemicals and endpoints. They should not be used to generate critical data to derive a QSfw, eco, but predicted

data can play a role in reducing uncertainty and thereby influence the size of AF chosen for extrapolation. In principle, the PPP dossier already contains enough data to derive a QSfw, eco by any of the methods described

below. However, in case there is uncertainty as to whether the potentially most sensitive taxonomic group is included in the dataset, or when deciding on the applicability of SSDs, non-testing methods can be considered. Reference to this is made in the following sections where relevant. It should be noted that most QSARs have been derived for those organisms which are already included in the PPP-dossier. Furthermore, care should be taken in the application of QSARs for substances with a specific mode of action.