What Are We Estimating? Revisiting Standard Nutritional Models Through the Target Trial Framework.

Area of Science:

• Nutritional Epidemiology
• Causal Inference
• Biostatistics

Background:

• Standard nutritional models are often used with causal intent in epidemiological research.
• Clarifying causal questions is crucial for understanding dietary exposures and health outcomes.

Purpose of the Study:

• To apply the target trial framework to nutritional epidemiology.
• To connect common causal questions to causal inference tools.
• To guide the selection of valid models for dietary interventions.

Main Methods:

• Utilized the target trial framework and causal inference tools.
• Defined key causal estimands for dietary strategies (e.g., total effect of chicken, comparative effect of fish vs. chicken).
• Employed the g-formula to estimate average causal effects and analyze standard nutritional models.

Main Results:

• Demonstrated that several standard nutritional models are re-parameterizations of the conditional outcome model within the g-formula.
• Highlighted the need to update total energy intake in standard multivariate and residual models within g-formula implementations to avoid inaccurate predictions.
• Provided a unified framework linking traditional models to causal inference.

Conclusions:

• The target trial framework offers a robust approach for causal questions in nutritional epidemiology.
• Standard models can be validly used within causal inference frameworks like the g-formula with appropriate adjustments.
• Findings provide guidance for selecting models that yield valid estimates for actionable dietary interventions.