A new paper argues that foods do not have fixed health effects in isolation. What matters is what they replace on the plate, and that shift could change how nutrition evidence is interpreted.
Opinion: Is this food healthy? Reframing nutrition evidence through counterfactual comparisons. Image Credit: Anna Puzatykh / Shutterstock
A recent opinion paper published in the journal Clinical Nutrition advocates reframing nutrition research through counterfactual comparisons.
Nutrition science continues to generate conclusions that are often perceived as context-dependent or inconsistent, despite decades of research. Nutrition science often relies on proxy outcomes, such as intermediate physiological measures and biomarkers, and their interpretation is contingent on the underlying causal contrasts. Dietary interventions are inherently compositional, meaning that increasing intake of one food necessitates a decrease in intake of another.
As such, studies assessing the same food(s) may reflect heterogeneous causal contrasts defined by distinct contexts and comparators. Pooling these heterogeneous contrasts in meta-analyses would produce estimates that might obscure the relationship between health and diet. In this paper, the authors argue that improvements in nutrition evidence synthesis warrant a meta-analytical reframing through a causal inference lens that integrates comparator context.
Counterfactual Framework in Nutrition Science
Moving beyond associative or descriptive interpretations toward causal inference requires accounting for the counterfactual framework. Modern causal inference highlights that causal effects are defined relative to specific interventions and alternatives. As such, meaningful causal interpretation is contingent on exposure specification and counterfactual comparisons.
The consistency assumption is a key requirement: it states that, for a given exposure, an observed outcome reflects the potential outcome associated with that intervention. It requires that the exposure represent a well-defined intervention, such that different treatment versions should not yield systematically distinct outcomes. If not, the estimated effect would become ambiguous.
For instance, red meat intake may refer to unprocessed lean meat, processed meat, or meat consumed along with refined carbohydrate-rich foods or vegetables. While these scenarios have the same exposure label, they are distinct interventions with varying health effects and biological mechanisms. Therefore, treating such different versions as interchangeable risks undermines causal inference.
Dietary Substitution and Relational Health Effects
Many studies treat foods as having intrinsic effects independent of the dietary context; nonetheless, the compositional nature of diets challenges this assumption. Modifying a single component of the diet does not occur in isolation; instead, it corresponds to specific scenarios based on how other dietary components are allowed to change.
This characteristic has implications for the definition of estimands and the interpretation of results. The paper distinguishes between effects that allow broader dietary changes and substitution effects that reflect replacing one food with another under constant intake. The health effect of a given food corresponds to the specific substitution rather than an intrinsic property of the food per se. For instance, a randomized controlled trial (RCT) compared the intake of dry-cured ham with that of cooked ham (control).
While the intervention appeared to yield favorable changes in metabolic markers compared with the control, interpretation depends critically on the nature of the tested substitution. If the comparator has less favorable effects, the observed benefit would imply a relative improvement over the alternative food rather than the intervention's inherent cardioprotective properties.
Network Meta-Analysis and Causal Inference
Meta-analyses of RCTs often represent the best evidence, provided that the underlying studies examine the same causal question. However, many nutrition meta-analyses aggregate effect estimates from different dietary contrasts without equivalent counterfactual comparisons. Consequently, pooled estimates lack explicit causal interpretation.
These challenges do not imply that evidence synthesis is intrinsically flawed. Instead, they suggest that traditional meta-analytical methods may be insufficient for compositional exposures, such as diet. In contrast, network meta-analysis (NMA) provides a methodological framework addressing some of these limitations by incorporating multiple comparators. NMA can preserve the relational nature of dietary interventions by modeling competing alternatives.
In the examples discussed by the authors, NMA reveals differences specific to comparators, whereas conventional meta-analyses may report minimal or neutral effects. Notably, NMAs do not eliminate all challenges, as valid causal interpretation requires meeting three key assumptions: consistency (indirect and direct evidence are consistent), transitivity (studies remain comparable across treatment contrasts), and clinical comparability (interventions are adequately homogeneous).
Counterfactual Nutrition Research Implications
Taken together, the limitations described herein do not indicate a failure of meta-analyses per se. Instead, they reflect a mismatch between the causal structure of dietary exposures and traditional evidence synthesis. Approaches like the NMA, which preserve the comparator structure, are better aligned with causal inference.
Nevertheless, methodological tools alone are not sufficient; therefore, improvements in nutrition science require reframing research questions to reflect well-defined counterfactual contrasts. The authors also call for clearer exposure definitions and more transparent reporting of substitution context and energy balance. Moving from “is the food healthy?” to “compared with what is this food healthy?” could improve the translational relevance, coherence, and interpretability of nutrition science.
Heavily processed foods may raise heart disease risk beyond poor nutrition alone