Variability of in vivo fish acute toxicity data.

The variability of toxicity data contained within databases was investigated using the widely used US EPA ECOTOX database as an example. Fish acute lethality (LC50) values for 44 compounds (for which at least 10 data entries existed) were extracted from the ECOTOX database yielding a total of 4654 test records. Significant variability of LC50 test results was observed, exceeding several orders of magnitude. In an attempt to systematically explore potential causes of the data variability, the influence of biological factors (such as test species or life stages) and physical factors (such as water temperature, pH or water hardness) were examined. Even after eliminating the influence of these inherent factors, considerable data variability remained, suggesting an important role of factors relating to technical and measurement procedures. The analysis, however, was limited by pronounced gaps in the test documentation. Of the 4654 extracted test reports, 66.5% provided no information on the fish life stage used for testing. Likewise, water temperature, hardness or pH were not recorded in 19.6%, 48.2% and 41.2% of the data entries, respectively. From these findings, we recommend the rigorous control of data entries ensuring complete recording of testing conditions. A more consistent database will help to better discriminate between technical and natural variability of the test data, which is of importance in ecological risk assessment for extrapolation from laboratory tests to the field, and also might help to develop correction factors that account for systematic differences in test results caused by species, life stage or test conditions.