User Guide to Thresholds and Classification - Environmental ...

344User Guide for Thresholds and Classificationsrespects to the one for which a prediction is needed. If one data set contains such an appropriate analogueused to derive the model, the measured value in the database for that compound vs model prediction shouldbe tested. If the results fit well with the overall model, it is likely the most reliable one to use. Likewise, if noneof the models contain test data for such an analogue, testing of the chemical in question is recommended.Octanol-water-partition coefficient (K OW )Computerised methods such as CLOGP (USEPA, 1999), LOGKOW (USEPA, 2000a) and SPARC (USEPA,2000b) are available to calculate log K OW directly from chemical structure. CLOGP and LOGKOW are basedupon the addition of group contributions, while SPARC is based upon a more fundamental chemicalstructure algorithm. Caution should be used in using calculated values for compounds that can undergohydrolysis in water or some other reaction, since these transformations need to be considered in theinterpretation of aquatic toxicity test data for such reactive chemicals. Only SPARC can be employed in ageneral way for inorganic or organometallic compounds. Special methods are needed in making estimates oflog KOW or aquatic toxicity for surface-active compounds, chelating compounds, and mixtures.Values of log KOW can be calculated for pentachlorophenol and similar compounds, both for the ionised andunionised (neutral) forms. These values can potentially be calculated for certain reactive molecules (forexample, benzotrichloride), but the reactivity and subsequent hydrolysis also need to be considered. Also,for such ionisable phenols, pKa is a second parameter. Specific models can be used to calculate log KOWvalues for organometallic compounds, but they need to be applied with caution since some of thesecompounds really exist in the form of ion pairs in water.For compounds of extremely high lipophilicity, measurements up to about 6 to 6.5 can be made by shakeflask, and can be extended up to about log KOW of 8 using the slow stirring approach (De Bruijn et al, 1989).Calculations are considered useful even in extrapolating beyond what can be measured by either of thesemethods. Of course, it should be kept in mind that if the QSAR models for toxicity, etc. are based onchemicals with lower log K OW values, the prediction itself will also be an extrapolation; in fact, it is known thatin the case of bioconcentration, the relationship with log K OW becomes non-linear at higher values. Forcompounds with low log K OW values, the group contribution can also be applied, but this is not very useful forhazard purposes since for such substances, particularly with negative log K OW values, little if any partitioningcan take place into lipophilic sites and as Overton reported, these substances produce toxicity throughosmotic effects (Lipnick, 1986).Bioconcentration factorIf experimentally determined BCF values are available, these values should be used for classification.Bioconcentration measurements must be performed using pure samples at test concentrations within watersolubility, and for an adequate test duration to achieve steady state equilibrium between the aqueousconcentration and that in the fish tissue. Moreover, with bioconcentration tests of extended duration, thecorrelation with log KOW levels off and ultimately decreases. Under environmental conditions,bioconcentration of highly lipophilic chemicals takes place by a combination of uptake from food and water,with the switch to food taking place at log KOW ≈ 6. Otherwise log KOW values can be used with a QSARJanuary 2012 EPA0109