208
p.F. Landrum and S.w. Fisher
nonlinearity was accounted for by molecular size as stated above. Steric properties (e.g., molecular size, surface area, or configuration) of even relatively smal1
molecules such as polychlorinated biphenyls (PCBs) can influence the relative
bioaccumulation of contaminants compared with traditional log BCF - log Kow
relationship (Shaw and Connel1, 1984). In addition, there is evidence that J-octanol becomes a less ideal solvent for larger molecules. Thus, accounting for the
relative solubility in octanol removes some of the observed nonlinearity in BCF
prediction (Banerjee and Baughman, 1991).
Because the contaminant lipophilicity as measured by Kow and extent of contaminant accumulation are so wel1 related, a convention developed to normalize
the contaminant concentrations to the lipid contents of organisms. This technique
reduced the variability between organism species and resulted in improved predictions for accumulation from water (Barron, 1990; Connell, 1988). However,
cases remain that demonstrate the limitations of lipid normalization to totally
account for the variation in contaminant accumulation among species. For instance, in Lake Baikal lipid-normalized BCF-K ow relationships had different
slopes for two fish species (Kucklick et aI.. 1994). Similarly, the BCFs for lake
trout and white fish from Siskiwit Lake on Isle Royle exhibit significant variability even with lipid normalization; the regression with log Kow was weak for
pesticides, and the correlation for PCBs was even more variable (Swackhamer
and Hites, 1988). In some cases, the absence of improved relationships despite
lipid normalization may be due to inclusion of multiple contaminant classes in the
regression (Axelman et aI., 1995; Connell, 1988). Where contaminant characteristics change, the interaction with lipids also changes. Thus, even in a single
species, the slopes of the relationships between lipid-normalized BCF and log Kow
are different for different contaminant classes (Axelman et aI., 1995). Thus,
predictability will depend on both the composition of the lipids and the characteristics of the contaminant. Both characteristics will contribute to an interaction
that will determine the relative contaminant solubility in the organism's lipids and
the ability of log Kow to predict that solubility interaction. This predictability wil1
generally be good within a species and class of contaminants (Axelman et aI.,
1995; Connell, 1988). However, the variance in the predicted BCF may be substantial if attempts are made to predict across species and contaminant classes
(Connell, 1988). For instance, the intercepts for regressions of log Kow against the
lipid-normalized BCF range over an order of magnitude among organisms
whereas the slopes vary from 0.844 to 1.0. (Connell, 1988). Despite the abovementioned caveats, lipid normalization of contaminant concentrations remains the
most viable and useful method of predicting BCFs and serves an important role in
screening new contaminants for potential BCFs.
9.2.2. Bioaccumulation
The variance in the predicting contaminant accumulation, using such predictors as
log Kow' increases substantially compared with predictions from aqueous exposures when attempting to predict the thermodynamic limits for contaminated
p.F. Landrum and S.w. Fisher
nonlinearity was accounted for by molecular size as stated above. Steric properties (e.g., molecular size, surface area, or configuration) of even relatively smal1
molecules such as polychlorinated biphenyls (PCBs) can influence the relative
bioaccumulation of contaminants compared with traditional log BCF - log Kow
relationship (Shaw and Connel1, 1984). In addition, there is evidence that J-octanol becomes a less ideal solvent for larger molecules. Thus, accounting for the
relative solubility in octanol removes some of the observed nonlinearity in BCF
prediction (Banerjee and Baughman, 1991).
Because the contaminant lipophilicity as measured by Kow and extent of contaminant accumulation are so wel1 related, a convention developed to normalize
the contaminant concentrations to the lipid contents of organisms. This technique
reduced the variability between organism species and resulted in improved predictions for accumulation from water (Barron, 1990; Connell, 1988). However,
cases remain that demonstrate the limitations of lipid normalization to totally
account for the variation in contaminant accumulation among species. For instance, in Lake Baikal lipid-normalized BCF-K ow relationships had different
slopes for two fish species (Kucklick et aI.. 1994). Similarly, the BCFs for lake
trout and white fish from Siskiwit Lake on Isle Royle exhibit significant variability even with lipid normalization; the regression with log Kow was weak for
pesticides, and the correlation for PCBs was even more variable (Swackhamer
and Hites, 1988). In some cases, the absence of improved relationships despite
lipid normalization may be due to inclusion of multiple contaminant classes in the
regression (Axelman et aI., 1995; Connell, 1988). Where contaminant characteristics change, the interaction with lipids also changes. Thus, even in a single
species, the slopes of the relationships between lipid-normalized BCF and log Kow
are different for different contaminant classes (Axelman et aI., 1995). Thus,
predictability will depend on both the composition of the lipids and the characteristics of the contaminant. Both characteristics will contribute to an interaction
that will determine the relative contaminant solubility in the organism's lipids and
the ability of log Kow to predict that solubility interaction. This predictability wil1
generally be good within a species and class of contaminants (Axelman et aI.,
1995; Connell, 1988). However, the variance in the predicted BCF may be substantial if attempts are made to predict across species and contaminant classes
(Connell, 1988). For instance, the intercepts for regressions of log Kow against the
lipid-normalized BCF range over an order of magnitude among organisms
whereas the slopes vary from 0.844 to 1.0. (Connell, 1988). Despite the abovementioned caveats, lipid normalization of contaminant concentrations remains the
most viable and useful method of predicting BCFs and serves an important role in
screening new contaminants for potential BCFs.
9.2.2. Bioaccumulation
The variance in the predicting contaminant accumulation, using such predictors as
log Kow' increases substantially compared with predictions from aqueous exposures when attempting to predict the thermodynamic limits for contaminated