User Guide to Thresholds and Classification - Environmental ...

343User Guide for Thresholds and Classificationsanother should be carefully studied, and attempts made to document these from the chemical literature(Lipnick, 1991b).Given these limitations in using QSARs for making predictions, it is best employed as a means ofestablishing testing priorities, rather than as a means of substituting for testing, unless some mechanisticinformation is available on the untested compound itself. In fact, the inability to make a prediction along withknown environmental release and exposure may in itself be adequate to trigger testing or the development ofa new QSAR for a class of chemicals for which such decisions are needed. A QSAR model can be derivedby statistical analysis, for example, regression analysis, from such a data set. The most commonly employedmolecular descriptor, log KOW, may be tried as a first attempt.By contrast, derivation of a mechanism based QSAR model requires an understanding or workinghypothesis of molecular mechanism and what parameter or parameters would appropriately model theseactions. It is important to keep in mind that this is different from a hypothesis regarding mode of action, whichrelates to biological/physiological response, but not molecular mechanism.Use of Quantitative Structure Activity Relationships in aquatic classificationThe inherent properties of substances relevant for classification purposes concerning the aquaticenvironment are:the partition coefficient n-octanol-water (log KOW);the BCF;degradability – abiotic and biodegradation;acute aquatic toxicity for fish, Daphnia, and algae; andprolonged toxicity for fish and Daphnia.Test data always take precedence over QSAR predications, providing the test data are valid, with QSARsused for filling data gaps for purposes of classification. Since the available QSARs are of varying reliabilityand application range, different restrictions apply for the prediction of each of these endpoints. Nevertheless,if a tested compound belongs to a chemical class or structure type (see above) for which there is someconfidence in the predictive utility of the QSAR model, it is worthwhile to compare this prediction with theexperimental data, as it is not unusual to use this approach to detect some of the experimental artefacts(volatilisation, insufficient test duration to achieve equilibrium, and water solubility cut-off) in the measureddata, which would mostly result in classifying substances as lower than actual toxicity.When two or more QSARs are applicable or appear to be applicable, it is useful to compare the predictionsof these various models in the same way that predicted data should be compared with measured (asdiscussed above). If there is no discrepancy between these models, the result provides encouragement ofthe validity of the predictions. Of course, it may also mean that the models were all developed using data onsimilar compounds and statistical methods. On the other hand, if the predictions are quite different, this resultneeds to be examined further. There is always the possibility that none of the models used provides a validprediction. As a first step, the structures and properties of the chemicals used to derive each of the predictivemodels should be examined to determine if any models are based upon chemicals similar in both of theseJanuary 2012 EPA0109