Unraveling the Impact of Serum Zinc Levels on Chronic Kidney Disease: Machine Learning and SHAP Value Interpretation.

Zinc is an essential trace element involved in antioxidant defense, immune regulation, and metabolic homeostasis, but its association with chronic kidney disease (CKD) in the general population remains unclear. We investigated the association between serum zinc levels and prevalent CKD among U.S. adults and explored the predictive relevance of serum zinc using machine-learning approaches. We conducted a cross-sectional analysis of adults from NHANES 2011-2016. Prevalent CKD was defined as estimated glomerular filtration rate < 60 mL/min/1.73 m2 and/or urinary albumin-to-creatinine ratio ≥ 30 mg/g. Survey-weighted logistic regression models, restricted cubic spline analysis, and prespecified subgroup analyses were performed. Exploratory machine-learning analyses were conducted using a temporal split, and SHAP was used for model interpretation. A total of 4192 participants were included, of whom 685 had prevalent CKD. In the fully adjusted model, each 1 μmol/L increase in serum zinc was associated with lower odds of prevalent CKD (OR = 0.93, 95% CI: 0.87-0.99). Compared with the lowest quartile, the fully adjusted ORs were 0.65, 0.69, and 0.56 for the second, third, and highest quartiles, respectively (p for trend = 0.012). A nonlinear association was observed (p for nonlinear = 0.023), with an apparent turning point at approximately 12.29 μmol/L. In exploratory machine-learning analyses, random forest showed the highest temporal test-set AUC, while SHAP analysis suggested that serum zinc contributed meaningful information within the predictive framework. Higher serum zinc levels were associated with a lower prevalence of CKD in U.S. adults. Serum zinc may be a relevant biomarker associated with kidney health, although temporality and causality cannot be inferred.