Benchmark Dose Analysis via Nonparametric Regression Modeling.

Area of Science:

• Toxicology
• Biostatistics
• Quantitative Risk Assessment

Background:

• Traditional benchmark dose (BMD) estimation relies on parametric modeling, which can lead to inaccurate low-dose inferences if the model is misspecified.
• Parametric model uncertainty is a significant challenge in quantitative risk assessment, potentially compromising the safety of low-dose estimations.

Purpose of the Study:

• To introduce and evaluate a nonparametric approach for estimating BMDs using quantal-response data.
• To assess the small-sample properties of nonparametric, bootstrap-based confidence limits for BMDs.
• To provide a robust alternative to parametric modeling in the presence of model uncertainty.

Main Methods:

• Utilized an isotonic regression method for nonparametric BMD estimation.
• Employed bootstrap-based methods to derive confidence limits for the nonparametric BMD estimates.
• Conducted a simulation study to investigate the small-sample performance of the confidence limits.
• Applied the method to a real-world example from cancer risk assessment.

Main Results:

• The nonparametric approach based on isotonic regression provides a viable alternative for BMD estimation.
• Nonparametric, bootstrap-based confidence limits demonstrate useful properties for BMD estimation.
• The method was successfully illustrated with a cancer risk assessment example, highlighting its practical applicability.

Conclusions:

• The nonparametric isotonic regression method offers a valuable alternative to parametric modeling for BMD estimation.
• This approach addresses concerns regarding parametric model uncertainty in quantitative risk assessment.
• The study supports the use of nonparametric methods for more reliable low-dose inference.