Weight of evidence approaches for deriving HC5s
Sandrine Andres
INERIS, France
Experience gained in developing Quality Standards (e.g. PNEC) in the framework of the REACH Technical Guidance Document (ECHA, 2011) shows that only a few substances can benefit from the use of an species sensitivity distributions, even if substances appear after an initial assessment as data rich substances. The main drawback is the lack of validated studies for the additional taxa. Indeed, the level of standardisation for testing these additional taxa (such as mayfly, dragonfly, amphibian, rotifer, molluscs, etc...) is usually lower than for the regular algae/daphnid/fish simplified trophic chain. As a consequence, this additional information is often not used for the assessment.
In order to make the best use of all the information available, a Multi-Criteria Decision Analysis (MCDA) tool was developed in the framework of the research project AMORE (Multi-Criteria Analysis for the Development of a Decision Support Tool for the prevention of Environmental Risks). This tool is based on weight of evidence (WoE) methodology, which aims to improve the evaluation of ecotoxicological data, through the assessment of their relevance and reliability for the definition of SSDs. The methodology allows us to rank the acceptability of ecotoxicological data for further use in the risk assessment process and therefore optimise their influence in the production of reliable SSDs, through a weighted bootstrap modelling procedure for data resampling.
In this project, it was hypothesised that the SSD can be based on all available ecotoxicity data, which can be heterogeneous and often non comparable. These data can be obtained through different approaches (e.g. experimental or even modelling) and conditions, e.g. the protocol can be standardised or not; time duration can vary among experiments, leading to chronic or acute data; different physiological endpoints can be observed, e.g. mortality, growth, reproduction; statistics used for interpreting data can differ, e.g. leading to NOEC or ECx.
The methodology is based on the assessment of a hierarchically structured set of 57 criteria, which is used for assigning a quantitative score to every ecotoxicological datum and was created based on the review of the state of the art frameworks for the assessment of ecotoxicological data. The different endpoints are analysed based on their production method and specifically on 3 main aspects: the ‘Experimental Reliability’, the ‘Statistical Reliability’ and the ‘Biological Relevance’ of the experimental or modelling protocol used. This assessment has been developed with the contribution of an expert panel of scientists on ecotoxicology. Knowledge and preferences of experts have been gathered through a participatory process, and is used for the calculation of the aggregated reliability scores of data. The nature of the process mandates the use of Fuzzy Logic during the aggregation phase, for handling the inherent uncertainty which appears in the form of unreported information, as well as possible lack of knowledge of the experts.
This approach allows for a weighed use of the available information available in a weight of evidence perspective.