Factors associated with chronic kidney disease in patients with gout: a two-centre study incorporating...

Chronic kidney disease (CKD) is a significant comorbidity in patients with gout, yet the specific phenotypes associated with its prevalence remain insufficiently characterised. This study aimed to evaluate the association between clinical indicators and ultrasonographic features with CKD using a transparent, interpretable machine learning approach.

This cross-sectional bicentric study analysed a cohort of 666 patients with gout, who were partitioned into a derivation cohort (n=500) for model development and internal validation, and an independent validation cohort (n=166). Three architectures (GLM, Random Forest, SVM) modelled associations between biomarkers and CKD status. SHapley Additive exPlanations (SHAP) analysis quantified individual variable contributions, and a dynamic nomogram visualised cumulative relevance.

Six key correlates were identified, comprising two clinical indicators (age and hypertension) and four ultrasonographic findings (bone erosion, high-burden (≥3 joints) tophi, synovial hyperplasia, and double contour sign). The GLM demonstrated the strongest associative performance, with an AUC of 0.842 (internal) and 0.835 (independent cohort), indicating high discriminative power. SHAP analysis identified advanced age and ultrasound-detected tophi as the most influential contributors to the observed associations. DCA and calibration plots confirmed that the framework accurately reflected the underlying probability distribution.

We established an interpretable analytical framework that characterises the clinical and ultrasonographic profile associated with CKD in gout. By integrating imaging markers with clinical indicators, this approach identifies associated phenotypes, providing a transparent reference for individualised assessment and enhanced clinical vigilance regarding renal status.