Uncertainty analysis and updated user guidance for interspecies correlation estimation models and low toxicity...

Area of Science:

• Environmental Chemistry
• Ecotoxicology
• Computational Toxicology

Background:

• Interspecies Correlation Estimation (ICE) models predict species' acute toxicity using log-linear relationships.
• Increased global use necessitates standardized guidance for consistent application across diverse chemicals.
• Existing models require expansion for low-toxicity compounds with high acute values, often outside the original domain.

Purpose of the Study:

• To expand ICE uncertainty analyses and user guidance for low-toxicity compounds.
• To evaluate prediction accuracy when surrogate values exceed the model domain.
• To provide updated guidance for applying and interpreting ICE models.

Main Methods:

• Utilized the standardized acute toxicity database underpinning ICE models.
• Created "truncated" models from the lower 75th percentile of surrogate toxicity data.
• Predicted toxicity for chemicals in the upper 25th percentile (μg/L and scaled mg/L) and compared to cross-validation and measured values.

Main Results:

• Prediction accuracy for scaled values did not differ from μg/L data when model slopes were 0.66-1.33.
• An uncertainty analysis determined a two-orders-of-magnitude interval for ICE confidence intervals.
• Updated ICE user guidance was developed based on these analyses.

Conclusions:

• ICE models can be reliably applied to low-toxicity compounds by scaling data beyond the original domain.
• The updated guidance and uncertainty analysis enhance the reliability of ICE predictions.
• This work advances the application and interpretation of ICE models in environmental risk assessment.