Aqueous and tissue residue-based interspecies correlation estimation models provide conservative hazard estimates for...

Area of Science:

• Environmental Chemistry
• Toxicology
• Computational Chemistry

Background:

• Accurate prediction of chemical toxicity is crucial for environmental risk assessment.
• Interspecies correlation estimation (ICE) models offer a statistical approach to predict toxicity across species.
• Data compilation strategies can influence the accuracy and uncertainty of predictive models.

Purpose of the Study:

• To develop and compare Interspecies Correlation Estimation (ICE) models for 30 nonpolar aromatic compounds.
• To evaluate the impact of two data compilation approaches (across-study vs. within-study) on model prediction accuracy and uncertainty.
• To assess the reliability of ICE-derived hazard concentrations compared to empirical species sensitivity distributions (SSDs).

Main Methods:

• Developed ICE models for 30 nonpolar aromatic compounds using two data compilation approaches: Type 1 (across studies) and Type 2 (within studies).
• Developed Target Lipid (TLM) ICE models (Type 2-TLM) using target lipid concentrations from the Type 2 dataset.
• Assessed model prediction uncertainty and compared ICE-based hazard concentrations with empirical SSDs.

Main Results:

• Most statistically significant models (90%) demonstrated good performance (MSE < 0.27, adj R² > 0.59).
• Type 2-TLM and Type 2 models exhibited the lowest variation in mean square errors.
• Cross-validation confirmed agreement between predicted and observed values for 86% of models.
• Most predicted values were within a 2-fold difference of observed values.
• No significant differences were found between ICE-based and empirical SSDs.
• ICE-based hazard concentrations increased significantly with increasing log KOW.

Conclusions:

• ICE models provide a statistically sound and conservative approach for deriving hazard estimates for aquatic life protection.
• The choice of data compilation strategy has a limited impact on the overall reliability of ICE-derived hazard concentrations.
• ICE models are valuable tools for environmental risk assessment, particularly when empirical data is scarce.