An evaluation of some methods for fitting dose-response models to quantal-response developmental toxicology data

Area of Science:

• Developmental Toxicology
• Statistical Modeling
• Quantitative Risk Assessment

Background:

• Accurate dose-response modeling is crucial for developmental toxicology studies.
• Traditional methods often rely on strict distributional assumptions that may not hold.
• Alternative statistical approaches are needed to handle complex toxicological data.

Purpose of the Study:

• To discuss dose-response modeling for quantal-response developmental toxicology data.
• To emphasize methods that minimize reliance on exact distributional assumptions.
• To contrast these flexible methods with traditional beta-binomial distribution analyses.

Main Methods:

• Employed quasi-likelihood methods, bootstrapping, and jackknifing.
• Fit dose-response models using a binomial likelihood for resampling procedures.
• Investigated quasi-likelihood with assumptions on intralitter correlation structure.

Main Results:

• Resampling procedures (bootstrapping, jackknifing) provide robust estimation under binomial likelihood.
• Quasi-likelihood methods demonstrate theoretical robustness to misspecified intralitter correlation.
• Simulation study evaluated the practical performance of these asymptotic results.

Conclusions:

• Methods avoiding exact distributional assumptions, such as quasi-likelihood and resampling, are valuable for quantal-response developmental toxicology.
• These approaches offer greater flexibility and robustness compared to traditional beta-binomial models.
• Further evaluation through simulation confirms the practical utility of these statistical techniques.