Methods for Deriving Pesticide Aquatic Life Criteria - Reviews of Environmental Contamination and Toxicology - page 46

methods (Jaworska et al. 2003). QSARs are used by several existing criteria deri-
vation methodologies to fill data gaps. That is, if little or no toxicity data are
available for criteria derivation, toxicity can be estimated using QSARs for some
compounds and species.
The most commonly used chemical parameter in QSARs is the K ow . QSARs are
developed for classes of chemicals, such as inert, less inert, reactive, and specifi-
cally acting chemicals (Verhaar et al. 1992). These classes were described by Vaal
et al. (1997b) as nonpolar narcotics, polar narcotics, reactive compounds, and
specifically acting compounds. For fathead minnows, Russom et al. (1997) further
separated the specifically acting compounds into oxidative phosphorylation
uncouplers, acetylcholinesterase inhibitors, respiratory inhibitors, electrophiles/
proelectrophiles, and central nervous system seizure agents. Using K ow data alone,
QSARs with good predictive power can be developed for narcotic chemicals.
However, for chemicals with a specific mode of toxic action, additional physical-
chemical data are needed, such as reactivity or p K a , and the predictive models become
more complex (Auer et al. 1990). Ramos et al. (1998) suggest that models based on
real phospholipid membrane-water partitioning, rather than K ow s, would more
accurately predict the toxicity of polar and nonpolar narcotics. The recent
“Workshop on Regulatory Use of (Q)SARs for Human Health and Environmental
Endpoints,” (summarized in Jaworska et al. 2003) produced a series of papers that
provide guidance on assessing reliability, uncertainty, and applicability of QSARs
(Eriksson et al. 2003). The papers from this workshop also effectively review the
use of QSARs in international decision-making frameworks for prediction of eco-
logical effects and environmental fate of chemicals (Cronin et al. 2003).
When insufficient data are available, several water quality criteria derivation
methodologies allow for the use of QSARs to estimate aquatic toxicity (discussed
below). When assessing hazards of chemicals for which little or no ecotoxicity data
are available, the USEPA Office of Pollution Prevention and Toxics (OPPT) uses
QSARs, under the Toxic Substances Control Act (TSCA), to estimate toxicity
(Nabholz 1991). Toxicity values, calculated from QSARs, are used in statistical
extrapolation or AF methods to derive criteria. In contrast, neither the national
(USEPA) nor the newer Great Lakes criteria derivation methodologies allow the
use of QSARs in criteria derivation (USEPA 1985, 2003a).
Although recognizing that QSARs exist for many modes of toxic action, the
Dutch guidelines allow the use of QSARs only for substances that have a nonspe-
cific mode of action (i.e., those acting by narcosis; RIVM 2001). The guidelines
provide 19 QSARs for aquatic species that represent nine different taxa. NOECs
estimated from QSARs may be used as inputs into extrapolation models for deriva-
tion of ERLs. In the UK (Zabel and Cole 1999) QSARs, or other models may be
used to predict toxicity in the absence of other data, but such data are not used to
derive EQSs (used only for support).
The OECD guidelines (1995) offer two QSAR approaches. First, is borrowed
from the USEPA's OPPT, and is based on the classification of chemicals by their
structure, without considering mode of toxic action. The specifics of this approach
are described by Nabholz (2003). Second, is a method that classifies chemicals,