To the Editor,

We read the accepted manuscript by Berlivet et al. that examines the links between dietary species richness (DSR) and various aspects of dietary sustainability, including environmental impact, diet quality and cost within the NutriNet-Santé cohort. The study adds to the growing field of nutritional epidemiology that looks at food biodiversity as a possible indicator of sustainable eating patterns. However, several methodological and interpretative issues deserve further discussion before applying these findings to policy.

First, we should recognise concerns about external validity. The NutriNet-Santé cohort includes a volunteer-based population in France with a relatively high socio-economic status and strong health motivation. This could limit how well the findings apply to other groups. In low-and middle-income countries, dietary diversity is often affected by structural issues like food insecurity, affordability and seasonal availability, instead of individual choices. Global data show that socio-economic inequalities greatly influence dietary quality and diversity across populations^{(Reference Ruel and Alderman1)}, and that nutritional outcomes connect with overlapping issues of undernutrition, obesity and climate change^{(Reference Swinburn, Kraak and Allender2)}. Therefore, we should be cautious in applying DSR findings to global dietary policy.

Second, while the use of a 264-item FFQ linked to species-level classification is innovative, there are significant methodological limitations. FFQ can be affected by recall bias and measurement error, especially for mixed dishes and processed foods. Additionally, DSR does not consider within-species variation, food processing or nutrient density, all of which impact diet quality and environmental effect. These limitations may lead to exposure misclassification and weaken observed associations.

Third, residual confounding is a major limitation in observational nutritional epidemiology. Dietary biodiversity strongly correlates with socio-economic status, education, health awareness and organic food consumption, all of which independently affect dietary quality and environmental impact^{(Reference Springmann, Wiebe and Mason-D’Croz3)}. Although the authors applied multivariable adjustments, there may still be unmeasured confounding, which restricts causal conclusions. Similar concerns have arisen in large-scale studies of diet and environmental outcomes.

Fourth, we should interpret dietary biodiversity as not always uniformly beneficial. The authors note different associations by food source. Higher plant-based DSR is linked to lower environmental impact and better diet quality, while higher animal-based DSR is associated with increased environmental pressure and worse diet quality. This supports global evidence that plant-heavy diets are more sustainable in both health and environmental aspects^{(Reference Hirvonen, Bai and Headey4)}. Thus, combining biodiversity into a single measure may mask important differences in dietary quality.

Lastly, the economic aspect needs careful thought. Although the overall dietary cost remained stable, higher plant-based DSR was related to increased costs, which raises equity concerns. Evidence shows that healthier and eco-friendly diets are not always the cheapest, especially when they focus on nutrient-rich plant foods^{(Reference Willett, Rockström and Loken5)}. Without addressing cost issues, recommendations promoting dietary biodiversity might unintentionally increase nutritional inequalities.

In conclusion, while Berlivet et al. offer valuable insights into DSR and sustainability outcomes, we must consider limitations in generalisability, measurement error, residual confounding and equity concerns. More research across varied food systems and socio-economic contexts is necessary before we can treat dietary biodiversity as a public health goal.