Implications of nonlinearity, confounding, and interactions for estimating exposure concentration-response functions...

Area of Science:

• Environmental Health Sciences
• Toxicology
• Biostatistics

Background:

• Biological mechanisms underlying exposure-related diseases often exhibit thresholds and nonlinearities.
• Traditional dose-response modeling struggles with complex concentration-response (C-R) functions involving nonlinearities.

Purpose of the Study:

• To review challenges in statistical regression modeling of nonlinear C-R functions.
• To advocate for advanced statistical approaches for causal interpretation and risk assessment.

Main Methods:

• Review of statistical regression modeling limitations for nonlinear C-R functions.
• Discussion of methods to control for non-causal sources of exposure-response coefficients.
• Advocacy for causal Bayesian networks and dynamic simulation models.

Main Results:

• Nonlinear C-R functions with thresholds are common in biological mechanisms.
• Traditional regression models face challenges in accurately capturing these nonlinearities.
• Non-causal factors can lead to misleading regression coefficients.

Conclusions:

• Advanced statistical models like causal Bayesian networks are crucial for interpreting nonlinear C-R functions.
• These methods can improve the causal effectiveness of health regulations.
• Accurate modeling of exposure-response is vital for human health protection.