Bayesian methods for uncertainty factor application for derivation of reference values.

Area of Science:

• Environmental Health Risk Assessment
• Toxicology
• Bayesian Statistics

Background:

• The EPA's Integrated Risk Information System (IRIS) process uses Reference Dose (RfD) and Reference Concentration (RfC) for non-carcinogen toxicity criteria.
• The National Research Council (NRC) recommended Bayesian methods for applying uncertainty factors (UFs) to adjust toxicity criteria.
• UFs account for uncertainties in extrapolating animal data to human populations and other variability.

Purpose of the Study:

• To explore and refine the NRC's suggested Bayesian methodology for applying uncertainty factors in risk assessment.
• To represent individual UFs as probability distributions rather than single values.
• To improve the accuracy and transparency of RfD and RfC derivations.

Main Methods:

• Applied Bayesian statistical methods to model individual uncertainty factors (UFs) as probability distributions.
• Refined the NRC's approach by considering each UF separately within a Bayesian framework.
• Examined 24 EPA IRIS program evaluations to test the refined methodology.

Main Results:

• Representing individual UFs as distributions significantly increased the geometric mean fold change in RfD/RfC values, ranging from 3 to over 30.
• The magnitude of change depended on the number of UFs applied and the assessment's complexity.
• Demonstrated a more nuanced quantification of uncertainty compared to traditional methods.

Conclusions:

• The refined Bayesian methodology provides a more robust approach to incorporating uncertainty in toxicity assessments.
• This method enhances the precision of RfD and RfC values by treating UFs as distributions.
• Recommendations are provided for implementing this refined methodology in regulatory risk assessment.