Bayesian model-averaged benchmark dose analysis via reparameterized quantal-response models.

Area of Science:

• Biostatistics
• Environmental Health
• Toxicology
• Risk Assessment

Background:

• Estimating minimum exposure levels causing adverse effects (benchmark doses, BMDs) is crucial for risk assessment.
• Parametric Bayesian estimation is increasingly used for BMD analysis, with various dose-response models available.
• Existing models may have different parametric interpretations, posing challenges for analysis.

Purpose of the Study:

• To present reparameterized dose-response models for explicitly incorporating prior information on BMDs.
• To enhance Bayesian estimation techniques for BMD analysis using Bayesian model averaging.
• To address model uncertainty and improve the reliability of BMD estimates.

Main Methods:

• Developed reparameterized dose-response models to integrate prior knowledge about the BMD.
• Applied Bayesian model averaging to BMD estimation to account for multimodel uncertainty.
• Utilized a carcinogenicity testing example to demonstrate the proposed methodology.

Main Results:

• The reparameterized models facilitate the direct use of prior information on the BMD.
• Bayesian model averaging provides robust point estimates and credible intervals for BMDs.
• The approach effectively handles model uncertainty in dose-response analyses.

Conclusions:

• The proposed Bayesian methods offer improved estimation of benchmark doses.
• Incorporating prior information and using model averaging enhance the reliability of risk assessment.
• These advancements are valuable for biomedical and environmental risk assessment applications.