|
5G - Species sensitivities distribution (TU2/12) Improving SSDs for small data-sets. Scott-Fordsmand, Janeck1, Damgaard, Christian 1, 1 National Environmental Research Institute, Silkeborg, Jylland, Denmark ABSTRACT- Common for the SSD approaches are that they require an extensive number of input data, for example, it has been estimated that 15-20 NOEC-data is the minimum requirement for a reliable use of the SSD. When performing a risk assessment for a compound (or an element) there are often less than 10 data available making the assumption for using the SSD approach less reliable. Apart from the uncertainty on whether the data fits a selected model there is another kind of uncertainty, which becomes especially important with small data sets, which is the uncertainty of the underlying data. When using NOEC values each of these have per definition no variation. However, using the SSD approach one of the prerequisites is strictly speaking that the values employed should be comparable. In contrast to this the NOECs may reflect from zero up to a 50% or more effect. To eliminate this problem it is recently considered to use ECx values rather than NOECs, as these reflect comparable effect levels between species. Hence, the ECx approach eliminates some of the uncertainty of the NOEC approach. Using ECx values it becomes clear that the input to the SSD is statistics rather than data, and a mean and variance then define the input. Enabling the SSD approach to include this will provide even more clarification of the uncertainty involved in the PNEC derivation. In other words, the larger uncertainty reflected in the variance around a mean ECx estimate from poorly designed experiments will be included the PNEC derivation and this will increase the reliability of the PNEC estimate. The present paper reflect on the above posed problems related to SSD when only small data-sets are available, and suggest answers on how to solve this. Key words: ECx, SSD, NOEC, few data |
