Work

Hyperuricaemia risk and its association with kidney failure and mortality in newly diagnosed type 2 diabetes with preserved renal function: a multicentre machine learning study

Year: 2026

Type: article

Abstract: Introduction Hyperuricaemia is prevalent in type 2 diabetes mellitus (T2DM) and associated with adverse renal and mortality outcomes; however, existing predictive models have primarily targeted populations with diabetic nephropathy or reduced renal function, limiting their applicability to early-stage T2DM. We developed a machine learning (ML)-based model to predict incident hyperuricaemia in newly diagnosed patients with T2DM with preserved kidney function and evaluated its association with long-term cardiorenal and mortality outcomes. Research design and methods Using multicentre electronic health record databases in South Korea, ML models were developed to predict incident hyperuricaemia within 1 year of T2DM diagnosis. Model performance was externally validated across three independent databases. Based on model-derived risk stratification, 5-year risks of kidney failure, cardiovascular outcomes and all-cause mortality were evaluated. Kaplan-Meier survival analysis was used to compare cumulative incidence between risk groups in the development cohort. HRs were estimated using Cox regression and pooled using random-effects meta-analysis, with adjustment for age, sex, hypertension, insulin use and diuretic use. Results The Extreme Gradient Boosting model achieved the highest discrimination (area under the receiver operating characteristic curve: 0.769 in the development cohort (n=5263); 0.709–0.750 in external validation). In fully adjusted pooled analyses, high-risk patients exhibited significantly elevated 5-year risks of kidney failure (HR=2.47; 95% CI 1.18 to 5.18) and all-cause mortality (HR=2.56; 95% CI 1.81 to 3.62). The association with cardiovascular outcomes was attenuated after full adjustment (HR=1.12; 95% CI 0.96 to 1.30). Conclusions In newly diagnosed patients with T2DM with preserved kidney function, individuals classified as high-risk for incident hyperuricaemia by ML model were found to have higher long-term risks of kidney failure and all-cause mortality, independent of traditional risk factors. These findings suggest that model-derived hyperuricaemia risk may serve as a prognostic marker reflecting underlying renal and systemic vulnerability before overt organ damage occurs.

Source: BMJ Open

DOI: 10.1136/bmjopen-2026-121518