Rethinking Nutrition Science: A Call for a Causal Framework
A recent opinion paper in Clinical Nutrition argues for a fundamental methodological shift in how evidence on diet and health is synthesized. The authors propose adopting a counterfactual framework from causal inference to address what they see as persistent issues of context-dependent and inconsistent conclusions in the field.
The core problem, according to the paper, is that nutrition studies often treat foods as having intrinsic health effects independent of context. This is problematic because diets are compositional: increasing the intake of one food typically requires decreasing the intake of another.
The Counterfactual Argument
The paper’s central thesis is that the health effect of a food corresponds to what it replaces in the diet, not to an inherent property of the food itself. To properly assess this, the authors advocate for a counterfactual framework, where causal effects are defined relative to specific interventions and their alternatives.
A key requirement in this framework is the consistency assumption, which states that an exposure must represent a well-defined intervention. Different versions of that exposure (e.g., the same food item in different dietary contexts) should not be treated as interchangeable.
The Challenge of Dietary Substitution
The authors emphasize that dietary interventions are inherently about substitution. They distinguish between:
- Effects that allow for broader dietary changes.
- Substitution effects, which reflect replacing one food with another while keeping overall intake constant.
They illustrate this with an example from a randomized trial comparing dry-cured ham to cooked ham, which showed favorable metabolic changes. The interpretation of such benefits, the paper states, depends entirely on the nature of the specific substitution being tested.
A major challenge identified is that current meta-analyses often pool results from studies that, while assessing the same foods, may reflect different causal contrasts based on distinct contexts and comparators. This pooling of heterogeneous contrasts can obscure true diet-health relationships.
A Proposed Tool: Network Meta-Analysis
As a potential solution, the authors suggest that network meta-analysis (NMA) provides a methodological framework that can address some limitations. NMA incorporates multiple comparators simultaneously, which can better model the relational nature of dietary choices by analyzing competing alternatives.
For NMA to support a valid causal interpretation, the paper notes three key assumptions must be met:
- Consistency
- Transitivity
- Clinical Comparability
Implications for Future Research
The authors conclude that new tools alone are insufficient. Lasting improvement requires a reframing of fundamental research questions. They call for:
- Reframing questions from "Is this food healthy?" to "Compared with what is this food healthy?"
- Clearer definitions of dietary exposures in study designs.
- More transparent reporting of the substitution context and energy balance in published research.
This shift, the paper argues, could improve the translational relevance, coherence, and interpretability of nutrition science by aligning evidence synthesis more closely with the true causal structure of dietary exposures.