Neutrophil Percentage to Albumin Ratio is Associated With Diabetic Kidney Disease in Community Cohorts: A More In-Depth Insight Incorporating Machine Learning.

Aim: The study aimed to investigate the relationship between neutrophil-percentage-to-albumin ratio (NPAR) and diabetic kidney disease (DKD), and to evaluate its potential for predicting DKD progression.

Methods: This cross-sectional study utilised data from the National Health and Nutrition Examination Survey (NHANES). We employed weighted multivariable logistic regression and restricted cubic splines (RCS) to examine nonlinear relationships between NPAR levels and DKD. Additionally, subgroup analyses were performed to assess heterogeneity across demographic strata. Machine learning (ML) algorithms were employed to develop DKD prediction models and receiver operating characteristic (ROC) curves for each model were plotted on the test set to evaluate predictive performance. Finally, the study applied Shapley Additive Explanations (SHAP) to interpret feature contributions to predictions.

Results: A total of 10,526 participants were included. After full covariate adjustment, the continuous variable NPAR was positively associated with the prevalence of DKD (OR = 1.16, 95% CI: 1.11-1.22, p < 0.001). The results of the RCS showed a significant nonlinear trend in the correlation between NPAR and DKD (P-non-linear < 0.0001). Subgroup analysis discovered that NPAR was generally associated with an increased possibility of developing DKD, but the subgroup differences were not statistically significant. Predictive modelling revealed NPAR had a good performance in assessing the risk of DKD incidence.

Conclusion: In the general population, high NPAR is positively associated with the development of DKD, and predictive modelling of DKD that includes NPAR has shown excellent performance. These findings provide a rationale for NPAR as a potential non-invasive biomarker for early detection of DKD.