Skip to main page content
Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
, 147 (1), 55-67

Toxicokinetic Triage for Environmental Chemicals

Affiliations

Toxicokinetic Triage for Environmental Chemicals

John F Wambaugh et al. Toxicol Sci.

Abstract

Toxicokinetic (TK) models link administered doses to plasma, blood, and tissue concentrations. High-throughput TK (HTTK) performs in vitro to in vivo extrapolation to predict TK from rapid in vitro measurements and chemical structure-based properties. A significant toxicological application of HTTK has been "reverse dosimetry," in which bioactive concentrations from in vitro screening studies are converted into in vivo doses (mg/kg BW/day). These doses are predicted to produce steady-state plasma concentrations that are equivalent to in vitro bioactive concentrations. In this study, we evaluate the impact of the approximations and assumptions necessary for reverse dosimetry and develop methods to determine whether HTTK tools are appropriate or may lead to false conclusions for a particular chemical. Based on literature in vivo data for 87 chemicals, we identified specific properties (eg, in vitro HTTK data, physico-chemical descriptors, and predicted transporter affinities) that correlate with poor HTTK predictive ability. For 271 chemicals we developed a generic HT physiologically based TK (HTPBTK) model that predicts non-steady-state chemical concentration time-courses for a variety of exposure scenarios. We used this HTPBTK model to find that assumptions previously used for reverse dosimetry are usually appropriate, except most notably for highly bioaccumulative compounds. For the thousands of man-made chemicals in the environment that currently have no TK data, we propose a 4-element framework for chemical TK triage that can group chemicals into 7 different categories associated with varying levels of confidence in HTTK predictions. For 349 chemicals with literature HTTK data, we differentiated those chemicals for which HTTK approaches are likely to be sufficient, from those that may require additional data.

Keywords: IVIVE; environmental chemicals; high throughput; toxicokinetics.

Figures

FIG. 1.
FIG. 1.
Comparison of the median, upper, and lower 95th healthy adult human percentiles of steady-state plasma concentration (Css) predicted on the y-axis using SimCyp (Wetmore et al., 2012) and on the x-axis using assumptions refined to better reflect screening for environmental chemicals. The Css values were estimated assuming a steady-state infusion of 1 mg/kg BW/day. Each chemical’s median and upper and lower 95th percentiles are connected by a line. The dashed line indicates the identity (perfect predictor) line.
FIG. 2.
FIG. 2.
Evaluation of the predictive ability of HTPBTK models by comparing predictions with in vivo measurements in the rat for the AUC (A) and Cmax (B). Each chemical may have more than one study with potentially different doses and routes. Dose route is indicated by symbols (oral, po—triangle; intravenous, iv—circle; subcutaneous, sc – square). The upper dotted line is the linear regression for intravenous doses, while the lower dotted line is for the oral doses. The dashed line indicates the identity (perfect predictor) line.
FIG. 3.
FIG. 3.
HTPBTK simulation results evaluating the steady-state infusion assumption to predict Css. A, Histogram of number of chemicals versus the days needed to effectively reach steady-state assuming 3 daily doses (every 8 h). B, Comparison of peak plasma concentrations due to 3 daily does predicted with the HTPBTK model with the Css due to a constant infusion exposure as used in Figure 1 and previous reverse dosimetry studies. The dashed line indicates the identity (perfect predictor) line.
FIG. 4.
FIG. 4.
A, Comparison of HTTK predicted Css values with measured in vivo Css values from literature studies. Two PFCs are plotted on top of each other. The dashed line indicates the identity (perfect predictor) line. Chemicals plotted below the identity line have predictions greater than literature values, and vice versa. The thick grey lines indicate the discrepancy between measured and predicted values (ie, the residuals). Comparison of the actual residuals calculated from (A) and the residuals predicted using a random forest analysis based on chemical descriptors is shown in (B). The most important factors in the random forest model, as calculated using the method of decrease in node impurities, are shown in (C). These factors are described in detail in Tables 2 and 3.
FIG. 5.
FIG. 5.
A recursive partitioning regression tree was used to classify the discrepancy between the Css predicted from in vitro data and the in vivo Css (Obach et al., 2008; Wetmore et al., 2012). Each “leaf” of the tree shows a group of chemicals for which HTTK either overestimates Css (making conservative predictions) or underestimates Css. For all but 3 groups, the predictions are on the order of the observed Css (approximately within a factor of 3.2× greater or lesser). For the other 3 groups, the Css is 5.2×, 7.7×, and 120× overestimated. The dashed line indicates the identity (perfect predictor) line.
FIG. 6.
FIG. 6.
Distribution of the 349 chemicals with IVIVE data among the triage elements in Table 4. Css is not meaningful for the chemicals that do not reach steady-state. Chemicals where the protein binding assay failed cannot easily be categorized since fup is an important predictor (Table 3) of discrepancy between HTTK Css predictions and literature values for Css. The remaining chemicals are categorized as likely having HTTK predictions for Css that would be overestimated, underestimated, or on the order (±3.2× the in vivo value).

Similar articles

See all similar articles

Cited by 32 PubMed Central articles

See all "Cited by" articles

Publication types

Feedback