Abstract
Background: Hyperuricemia frequently coexists with type 2 diabetes mellitus (T2DM), contributing to a heterogeneous patient population. While previous analyses compared the overall longitudinal effects of allopurinol and SGLT2 inhibitors in this cohort, it remains unclear whether baseline metabolic heterogeneity modifies treatment response. This study aimed to determine whether data-driven metabolic clustering identifies phenotypic subgroups with prognostic or predictive relevance in hyperuricemic T2DM. Methods: In a retrospective cohort of 224 patients with T2DM and hyperuricemia, model-based clustering was applied to age, diabetes duration, body mass index (BMI), serum uric acid (sUA), HbA1c, eGFR, and sex. A sensitivity analysis excluded outliers, yielding 207 patients. Longitudinal trajectories of eGFR and sUA were assessed using linear mixed-effects models and individual slopes. Effect modification by cluster was tested via three-way interactions and analysis of covariance. Results: Clustering identified two groups with weak separation: an adipose-metabolic cluster (n = 116; exclusively male, BMI 33.1 ± 5.7 kg/m(2), sUA 478 ± 62 µmol/L) and a lean-metabolic cluster (n = 91; exclusively female, BMI 31.3 ± 6.0 kg/m(2), sUA 426 ± 67 µmol/L). Treatment-agnostic analyses showed no differences in eGFR and sUA slopes or in all-cause mortality across clusters. In both clusters, SGLT2 inhibitors yielded significantly more favourable eGFR slopes than allopurinol, while sUA reductions were comparable across treatments. No significant three-way interactions were detected. Conclusions: In this cohort, although baseline metabolic characteristics differ among patients, using the selected baseline variables, no clinically actionable treatment-relevant phenotypes were identified.