Causal models in epidemiology: past inheritance and genetic future

Area of Science:

• Epidemiology
• Genetics
• Causal Inference

Background:

• Genetic research offers new opportunities for epidemiology to identify disease causes.
• Genomics and proteomics are increasingly used to study gene-environment interactions.
• Understanding theoretical models of causality is essential for interpreting these studies.

Purpose of the Study:

• To explore different causality models relevant to gene-environment interactions in disease epidemiology.
• To clarify the concepts of interaction (effect modification) and etiologic fraction.
• To highlight limitations in current epidemiologic approaches to causality.

Main Methods:

• Review of theoretical models of causality, starting with Rothman's 'pie' model.
• Discussion of newer approaches like directed acyclic graphs and structural equation models.
• Analysis of limitations in assessing interaction and etiologic fractions.

Main Results:

• Interaction is a fundamental aspect of causal processes, not a secondary finding.
• Current methods for assessing interaction often neglect life course and temporal dynamics.
• A clear distinction between individual-based and population-level causality models is needed.
• Population-level analyses face uncertainty in quantifying interaction and etiologic fractions.

Conclusions:

• Refined theoretical models are necessary for robust gene-environment interaction studies in epidemiology.
• Careful consideration of causality, interaction, and etiologic fractions is vital for accurate interpretation.
• Addressing limitations in current methods will improve epidemiologic causal inference.