Association between the TyG-SIRI index and diabetic kidney disease: a cross-sectional study with external validation

Objective Insulin resistance and systemic inflammation are key drivers of DKD, but whether a composite index capturing both pathways improves risk stratification remains unclear. This study aimed to investigate the association between TyG-SIRI and DKD in diabetic patients. Methods This study included participants with diabetes from the NHANES and an independent validation cohort. Logistic regression and ROC analysis evaluated associations and discriminative ability for three renal outcomes. In NHANES, survey-weighted logistic regression was performed. We assessed the incremental value of TyG-SIRI over a base model in terms of model performance (NRI and IDI), discrimination (ΔAUC with DeLong’s test), calibration (calibration curves and Brier score), and clinical net benefit (DCA). Quartile analysis and restricted cubic spline (RCS) regression were employed to characterize the dose–response relationship. ROC analysis was used to assess the discriminative ability of TyG-SIRI for DKD. The same analytical framework was applied to the validation cohort for exploratory external validation. Results In NHANES, logistic regression showed TyG, eGDR-BMI, SIRI, and NLR were associated with DKD (all P < 0.05). Adjusted ln (1+TyG-SIRI) was significantly associated with DKD (OR = 1.50, 95% CI: 1.28–1.77), albuminuria (OR = 1.53, 95% CI: 1.28–1.84), and low-eGFR (OR = 1.36, 95% CI: 1.10–1.68). Adding TyG-SIRI to the baseline model significantly improved AUC from 0.702 to 0.722 (DeLong P = 0.003), with NRI of 0.321 and IDI of 0.024 (both P < 0.001). RCS revealed a nonlinear relationship between TyG-SIRI and DKD ( P overall < 0.001, P nonlinear = 0.021). Quartile analysis showed a graded association (Q4 vs Q1: OR = 2.83, 95% CI: 1.83–4.37 , P < 0.001). Calibration improved (Brier score: 0.186 to 0.181), and DCA indicated net clinical benefit across threshold probabilities of 9%–60%. In the validation cohort, TyG-SIRI remained independently associated with DKD (OR = 1.46, 95% CI: 1.20–1.78), with improved risk reclassification, and demonstrated a nonlinear dose–response relationship, with concurrent calibration enhancement over the baseline model. Conclusion TyG-SIRI was independently associated with DKD in diabetic patients. In both cohorts, TyG-SIRI offers modest gains in risk stratification, calibration, and net benefit, suggesting potential utility, yet moderate improvement requires prospective validation.