The following questions / concerns were discussed:
- Review current tools and key (statistical) methodology, including assumptions about distributions of sensitivity, use of hierarchical models, interspecies correlations. Identify where there are important differences and what the implications of these could be.
Statistical aspects considered by the group included SSD-based extrapolation methods as well as alternative approaches. 2 SSD-based extrapolation tools were discussed: U.S. EPA’s Web-ICE (http://www.epa.gov/ceampubl/fchain/webice/) and a recent tool developed by Peter Craig (Craig, 2013). Both tools were limited to acute toxicity data, incorporated taxonomic distance, and provided similar outputs including HC5 estimates using limited data. Web-ICE was considered to have more defined user rules, but was a less statistically rigorous SSD generator than hSSD. Also, the level of confidence with Web-ICE is lower when extrapolating over large taxonomic distance. Outputs of the hSSD tool were considered to be user dependent, which could result in substantially varying results between users. Of additional concern was that a high degree of ecological community expertise or knowledge was necessary to provide reliable estimates. Both tools were considered better alternatives to the use of generic safety factors. Both tools also require the availability of appropriate datasets, including standardised toxicity values and relevant exposure metrics (e.g. dissolved metals).
A variety of alternative approaches to SSD development were discussed, including trait-based SSDs, chemoinformatic methods (e.g. QSAR, read across), and determination of protective levels by just focusing on sensitive species. While trait-based SSDs were considered to have potential utility, questions on what traits should be considered (e.g. ecological, physiological, etc.) remained. It was unclear if there was sufficient knowledge for a sensitive species approach that would ensure protection of multiple aquatic systems. Overall, there was no consensus recommendation for clear alternatives to the SSD-based extrapolation methods above, and additional research would be needed.
2. As sensitivity to chemical stress seems to be related to taxonomic closeness, how could this be used in the construction and interpretation of SSDs?
There was general consensus that taxonomic closeness can be important in extrapolation of sensitivity across species. Species sensitivity may be considered highly correlated at the Family level. Understanding sensitive taxa, such as to Family level, could be used to ensure SSDs are representative of aquatic communities. Uncertainties remain regarding the need to alter SSD composition for different aquatic communities, including fresh versus saltwater species, large versus small assemblages, and sensitive versus robust systems. The proportion of invertebrates and fish in the SSD was noted as probably important in the estimation of HC5 of compounds that can have large differences in species sensitivity such as insecticides. The need to integrate water column species with other compartments such as sediment and terrestrial systems was also noted.
3. Do models based on prior knowledge provide advantages over other methods?
Prior knowledge can provide significant advantages to both constructing and interpreting SSDs. Important aspects include knowledge of MOA, taxon sensitivity, composition of the aquatic community being assessed, physiological and ecological species traits, and physico-chemical properties of the chemical. Relationships between taxon sensitivity and MOA are important because some chemicals will show large taxon specific differences in toxicity, such as herbicide sensitivity of plants versus fish, and acetylcholinesterase inhibitor toxicity to invertebrates versus fish. SSD development should consider this information, such as including plant species in a herbicide SSD and consider the proportion of invertebrates in constructing insecticide SSDs. Knowledge of chemical properties such as solubility are important in understanding maximum values to include in SSDs (i.e. should not exceed the solubility cut off).
4. Are current modelling success criteria, such as those identified in the REACH TGD, sufficient, overly prescriptive or insufficient?
The current REACH criteria for SSD composition includes 10 species of 8 taxonomic groups. Overall, these requirements seem reasonable and are consistent with the 8 family minimum data requirement in U.S. EPA guidelines for developing U.S. Ambient Water Quality Criteria for aquatic life. However, these criteria may be hard to meet because of limited data for the number of substances. There was general consensus that additional research was needed on minimum datasets and taxa diversity requirements, and the use of extrapolation methods to fill species sensitivity data gaps. The question whether current modelling success criteria are sufficient should also be considered relative to alternative approaches. If the alternative is the use of assessment factors (AF), it has clearly been shown that they provide an inconsistent method, i.e. the method is more conservative for large data sets (n > 6) than for small data sets (n < 6). In this context, optimising the method by making the AF dependent of the number of available toxicity data or allowing the application of the SSD for smaller sample sizes should be considered. There was general consensus that the current modelling criteria of REACH TGD are a guideline and that motivated deviation of these guidelines, based on a solid scientific justification, should always be possible on a case-by-case basis.
5. What are the research needs?
A variety of research and development needs were discussed, including methods validation, developing alternative estimation approaches, incorporating knowledge of chemical properties and exposure, and peer review and engaging with stakeholders. One identified research need was to compare trait-based SSDs with traditional strictly taxonomic-based SSDs, and to define what traits are most relevant to SSD generation. Alternative approaches should be explored, including focusing on sensitive taxa rather than broadly populating an SSD. However, there is uncertainty of what the sensitive taxa will be for many substances. A sensitive species approach may require novel methods development, including integrating chemical structure, genomic, traits and MOA information. An additional research question was whether critical body residue (CBR)-based SSDs could be developed by incorporating bioconcentration factors into the SSD generation. There was general consensus that extrapolation approaches and minimal dataset SSDs are better alternatives to generic safety factors, but validation against field and mesocosm data is required. MOA was considered to be an important determinant of species sensitivity and research is needed to determine linkages between MOA and SSD composition requirements. SSD research and development has been focused on acute toxicity data for water column organisms. The development and validation of chronic toxicity extrapolation methods and approaches applicable to other environmental compartments (such as sediment, soil and air) both remain significant research needs. How to best leverage knowledge of the compounds chemical properties, behaviour and environmental exposure scenarios should be explored. Finally, stakeholders and others in the scientific community should be engaged to assist in peer review, validation and tool improvement, and to facilitate communication of uncertainties and the value of research investment.