Nutrition researchers said on Apr. 15 that the health effects of foods depend largely on what they replace in a person's diet, according to a new opinion paper published in Clinical Nutrition.
The topic is important because nutrition science often produces findings that appear inconsistent or difficult to interpret. The authors argue that focusing on what foods are substituted for others could help make nutrition research more meaningful and relevant.
The paper explains that dietary interventions are inherently compositional, meaning an increase in one food requires a decrease in another. This makes it challenging to study foods as if they have fixed effects independent of context. "What matters is what they replace on the plate, and that shift could change how nutrition evidence is interpreted," the authors write.
According to the paper, studies sometimes treat different versions of a food—such as unprocessed lean meat versus processed meat—as interchangeable exposures, even though these can have distinct health outcomes. The authors suggest this approach risks undermining causal inference and clarity about which substitutions are actually being tested.
The article also highlights limitations with traditional meta-analyses, which often pool results from studies examining different dietary contrasts without clear counterfactual comparisons. The researchers recommend network meta-analysis (NMA) as an alternative method because it can model multiple comparators and better reflect the relational nature of dietary changes. However, they caution that NMAs require careful attention to assumptions such as consistency and clinical comparability among interventions.
In conclusion, the authors call for improvements in nutrition science by reframing research questions around well-defined counterfactual contrasts and providing clearer definitions of exposures and substitution contexts. They argue this shift—from asking whether a food is healthy to asking compared with what it is healthy—could improve both coherence and practical relevance in future research.
