Statistical aspects of the estimation of human risks

Area of Science:

• Toxicology
• Risk Assessment
• Pharmacokinetics

Background:

• High to low dose extrapolation is crucial for estimating low-level toxic agent exposure effects.
• Accurate estimation of low-dose effects (response rates of 10^-6) is limited by study sample sizes.
• Mathematical dose-response models are essential for extrapolating toxicological data.

Purpose of the Study:

• To highlight the challenges and limitations in high to low dose extrapolation for risk assessment.
• To emphasize the critical role of mathematical dose-response models in low-dose extrapolation.

Main Methods:

• Analysis of various mathematical dose-response models used for extrapolation.
• Comparison of model behavior in observable versus unobservable low-dose response ranges.

Main Results:

• Mathematical models appear similar at observable response rates but diverge significantly at unobservable low doses.
• Risk estimates at low doses are highly dependent on the chosen mathematical model form.
• Differences in risk estimates can span 3-4 orders of magnitude due to model selection.

Conclusions:

• The choice of mathematical dose-response model is the most significant limitation in high to low dose extrapolation.
• Uncertainties in dose-response models, pharmacokinetics, thresholds, and exposure patterns contribute to risk estimation variability.
• Substantial uncertainty exists in risk estimates derived from high to low dose extrapolation methodologies.