Novel nomogram accurately predicts gout risk in diabetic kidney disease patients using five key factors

Patients with diabetic kidney disease (DKD) face a significantly elevated risk of developing gout, a comorbidity that severely exacerbates health outcomes. This coexistence creates a complex clinical challenge, as both conditions are linked to metabolic dysfunction and inflammation. Current clinical practice often lacks a standardized, reliable tool for early and individualized gout risk assessment within this vulnerable patient population. Identifying specific risk factors and developing a predictive model is crucial to enable proactive management and improve patient care.

Researchers conducted a two-center, retrospective analysis of 5,670 DKD patients managed at two hospitals between September 2017 and February 2025. The cohort was split into training, internal validation, and external validation groups. Key risk factors for gout were identified using an integrated approach combining Best Subset Regression and Least Absolute Shrinkage and Selection Operator (LASSO) analysis, followed by multivariate logistic regression. A risk nomogram was then developed, and its performance was rigorously assessed using Area Under the Receiver Operating Characteristic Curve (AUC) values, calibration plots, Hosmer-Lemeshow goodness-of-fit tests, and decision curve analysis (DCA).

Overall, 5,670 DKD patients were analyzed, with 4,617 forming the training and internal validation cohorts, and 1,088 serving as an external validation cohort. The final predictive model incorporated five significant variables: sex, body mass index (BMI), fasting C-peptide, estimated glomerular filtration rate (eGFR), and serum urate levels. The developed nomogram demonstrated strong discriminatory power across all cohorts.

The nomogram achieved impressive AUCs of 0.784 in the training cohort, 0.782 in the internal validation cohort, and 0.770 in the external validation cohort, indicating robust predictive accuracy. Calibration plots confirmed close alignment between predicted and observed outcomes, with Hosmer-Lemeshow p-values of 0.769, 0.332, and 0.520 for the respective cohorts, suggesting excellent model fit. Furthermore, DCA revealed net benefits across a range of clinically relevant threshold probabilities, underscoring the model's practical utility.