Association of the dietary index for gut microbiota with metabolic syndrome and its components combining interpretable machine learning algorithms.

backgroundPrevious studies have emphasized the critical role of diet and gut microbiome in Metabolic syndrome (MetS). The dietary index for gut microbiota (DI-GM) represents a novel dietary index that effectively reflects the diversity of gut microbiota; nevertheless, its applicability to MetS and its components remains unknown.

methodsFor this study, we enrolled 19,702 individuals from NHANES 2007-2020. DI-GM comprises dietary information of 14 dietary components, including 10 beneficial and 4 unfavorable ones. Weighted logistic regressions evaluated associations of DI-GM with MetS and its components, whereas weighted linear regression analyzed its association with 6 MetS-related biochemical indicators. Modified Poisson regression, sensitivity analyses after multiple imputation and subgroup analyses ensured robustness. Restricted cubic spline (RCS) analysis explored whether a non-linear relationship exists. Nine machine-learning models were developed for MetS prediction, and six discrimination characteristics selected the optimal model. SHapley Additive exPlanations (SHAP) was utilized to interpret the contributions of variables for model decision-making capacity.

resultsAfter fully adjusting for confounders, the DI-GM score exhibited a noticeable negative correlation with the prevalence of MetS (OR: 0.95, 95% CI: 0.93-0.97, P-value < 0.001), along with elevated waist circumference (OR: 0.91, 95% CI: 0.88-0.94, P-value < 0.001), elevated blood pressure (OR: 0.95, 95% CI: 0.93-0.98, P-value < 0.001), reduced high-density lipoprotein (OR: 0.95, 95% CI: 0.93-0.98, P-value < 0.001) and elevated fasting blood glucose (OR: 0.94, 95% CI: 0.90-0.98, P-value = 0.002). RCS exhibited a significant inverse association of DI-GM with MetS for non-linear relationship when met the score of 5. Subgroup analysis demonstrated that the association remained stable and consistent across the majority of the subgroups. XGboost presented superior performance and SHAP analysis revealed that higher DI-GM exhibited considerable inverse influence, ranking after "BMI ≥ 30", "Age", "Race-Non-Hispanic Black" and "CKD-Yes".

conclusionsOur study presents compelling evidence that higher scores of the DI-GM are associated with a lower prevalence of Mets and its components. Dietary strategies that incorporate the DI-GM score could contribute to the harmonious ecological state of the gut microbiome and be crucial in the prevention of MetS.