Uncovering non-linear dietary predictors of cardiovascular disease risk in older adults with periodontitis: a cross-sectional analysis.

Objective: This study investigates whether the chronic inflammatory state associated with periodontitis alters the relationship between habitual diet and cardiovascular disease (CVD) risk in older adults. Methods: Data from three NHANES cycles (2009-2014) and the MyPyramid Equivalents Database (MPED) were integrated. To validate the generalizability of the findings, 183 older adults with periodontitis were included in an external validation study recruited from an independent hospital-based cohort. Feature selection was performed using correlation coefficients and the BORUTA algorithm. Six machine learning models were constructed, and SHAP and LIME algorithms were applied to interpret the associations between dietary trace elements, habitual food intake, and CVD risk. Results: The XGBoost model demonstrated superior predictive performance (Validation AUC-ROC: 0.854 for NHANES, 0.889 for MPED). SHAP analysis identified key protective factors against CVD in older adults with periodontitis. In the exploratory NHANES model, the top predictors included theobromine, lycopene, total sugar, food folate, beta-cryptoxanthin, and magnesium. The MPED model identified meat, whole grains, cured meat, tomatoes, eggs, added sugars, and total fruits as strong protective factors. Crucially, multivariable logistic regression analysis of the external cohort confirmed that higher consumption of red meat (OR = 0.46, Conclusion: Periodontitis-induced inflammation may partially invert dietary CVD risks, rendering higher intake of meat and sugars protective in this specific context. Interpretable XGBoost models reveal these non-linear effects, enabling more precise clinical nutrition guidance for older adults with periodontitis.