Periodontitis and Advanced Cardiovascular-Kidney-Metabolic Syndrome: Mediation and Machine Learning-Based...

This study investigated variables that statistically account for a portion of the periodontitis-advanced cardiovascular-kidney-metabolic (CKM) syndrome association in cross-sectional analyses, and developed interpretable machine learning risk stratification models in individuals with periodontitis, to provide preliminary cross-sectional evidence for future hypothesis testing.

A total of 4184 participants were enrolled from the 2009-2014 National Health and Nutrition Examination Survey (NHANES). After feature selection, variables that statistically accounted for a portion of the periodontitis-advanced CKM association were explored through mediation analysis. Three common machine learning models were subsequently constructed, and their discriminative performance was evaluated based on indicators including the area under the receiver operating characteristic curve (AUC), followed by model interpretation using Shapley Additive Explanations (SHAP).

Periodontitis was independently associated with higher odds of advanced CKM syndrome after full adjustment (OR = 1.423, 95% CI: 1.108-1.828, P < .01). NPAR statistically accounted for 7.490% of the total association between periodontitis and advanced CKM syndrome. The AUC values on the test set of the logistic regression, SVM, and XGBoost models were 0.824, 0.827, and 0.803, respectively. The models consistently showed favorable discriminative performance in both the general population and 5 random splits of the periodontitis cohort.

Mediation analysis suggested markers of systemic chronic inflammation and nutritional status that statistically accounted for a portion of the periodontitis-advanced CKM syndrome association. Exploratory risk stratification models showed favorable discriminative performance for advanced CKM syndrome in individuals with periodontitis. These findings provide preliminary evidence that may generate hypotheses for future prospective studies on oral-cardiovascular-kidney-metabolic interactions.