Workshop Report 28

Assessment factors for deriving PNECs: Food for thought

Ad M.J. Ragas
Radboud University, the Netherlands

Within regulatory contexts such as REACH and the European Water Framework Directive (WFD), assessment factors are used to derive safe exposure levels for aquatic and terrestrial ecosystems from single species toxicity data. These safe exposure levels are also referred to as PNECs. If toxicity data are available for a limited set of aquatic species – e.g. an alga, a daphnia and a fish – the lowest value is typically divided by an assessment factor to arrive at the PNEC. The value of this assessment factor varies between 10 and 1000, depending on the number and the nature of the available data. If chronic NOECs are available for an extensive set of aquatic species (i.e. > 15 species covering at least 10 different taxonomic groups), the 5th percentile of the species sensitivity distribution is determined and an assessment factor of 1-5 is subsequently applied to arrive at the PNEC. The main aim of the current contribution is to formulate recommendations for improving the use of assessment factors in deriving PNECs. These recommendations are based on a statistical analysis of a large set of chronic toxicity data resulting from aquatic single species tests and mesocosm experiments.

A database with chronic single species NOECs on 20 different chemicals was compiled based on data reported in the open literature. Chronic mesocosm data were found for 6 of these substances and were also included in the database. For each of the substances in the database, the 5th percentile of the SSD (HC5) was determined. This HC5 was then compared with:

  • the PNEC reported in the mesocosm experiments (if available);
  • PNECs derived by applying a safety factor of 10 to the lowest value of a limited dataset of 3, 6 or 9 NOECs. These datasets were generated by parametric bootstrapping of the available single-species NOECs.

Mesocosm PNECs were generally lower than the HC5, with 2 notable exceptions, i.e. lindane and dimethoate, which can be explained by the limited set of species in the mesocosm. The HC5 is on average a factor of 2.0 lower than the PNEC derived from a set of 3 chronic NOECs. This difference increases to a factor of 4.5 and 7.2 for datasets with 6 and 9 chronic NOECs, respectively. Based on these results 2 general recommendations are formulated:

  • The assessment factor of 10 that is currently being applied to the lowest value of small datasets (i.e. alga, daphnid and fish) should be differentiated depending on the number of available data, e.g. a factor of 20 if one value is available for each taxonomic group, but a value of 5 when 3 or more values are available for each taxonomic group.
  • The default assessment factor of 2 is suggested for the HC5 of the SSD. This default value can be further refined based on the specific characteristics of the available toxicity data, i.e. representativeness, mode of action, interspecies variability and uncertainty.