Association of dietary intake of copper and zinc with systemic inflammation: A cross-sectional study integrating machine learning.

OBJECTIVES

To examine the association of individual and multi-nutrient intake with inflammation indicators, developing an interpretable machine learning model to assess nutrient impacts on systemic inflammation in a Chinese population.

METHODS

A total of 874 participants were recruited in this cross-sectional study. Information of dietary intake was obtained via face-to-face interviews on 24-hour food intake and dietary consumption frequency using validated questionnaires. Systemic Inflammatory indices were assessed based on the peripheral blood count. The associations were estimated by linear regression models or restricted cubic splines (RCS) approach. Six machine learning models were developed to evaluate the contribution of nutrient on systemic inflammation, with the optimal model interpreted via SHapley additive explanation (SHAP).

RESULTS

Regression analysis showed higher intake of Cu and Zn was positively associated with higher value of lymphocyte-to-monocyte ratio (LMR), an anti-inflammatory index (β = 0.06. 95% CI: 0.02-0.10; β = 0.07. 95% CI: 0.01-0.13, respectively). A dose-dependent manner was observed for the Cu-LMR association. The random forest model demonstrated superior performance for LMR prediction (R²=0.32). SHAP interpretation identified sex, Cu intake, and age as the top predictors, with higher intake of Cu, Zn, carbohydrates, vitamin B3, dietary fiber, and iron correlating with elevated LMR predictions.

CONCLUSIONS

Higher dietary intake of Cu and Zn were significantly associated with lower levels of systemic inflammation. The findings integrating machine learning highlights the potential benefits of optimizing dietary Cu and Zn intake to mitigate systemic inflammation.