Novel metabolic syndrome subgroups and their association with cardiovascular outcomes: A data-driven cluster analysis of the UK Biobank

Abstract

Aims Metabolic syndrome (MS) is a heterogeneous condition associated with increased cardiovascular disease (CVD) risk. This study aimed to identify subgroups of individuals with MS using cluster analysis and to evaluate their differential risks for incident CVD events. Materials and Methods Using UK Biobank data, we conducted a data-driven cluster analysis of drug-na & iuml;ve individuals with MS (n = 62 776) and a control group without MS (n = 230 999). Separate clustering was performed for men and women using the following six variables: waist circumference; systolic and diastolic blood pressure (BP) values; triglyceride, high-density lipoprotein (HDL)-cholesterol, and fasting blood glucose values. Cox and logistic regression models were used to analyse the clusters in terms of their association with 3 point major adverse cardiovascular events (3P-MACE) (myocardial infarction, stroke, and CVD-related mortality). Results We identified three distinct MS subgroups with different risk profiles for 3P-MACE compared to the non-MS group. Cluster 1, characterised by the lowest HDL-cholesterol and highest triglyceride levels, had the lowest CVD risk among the MS clusters (adjusted hazard ratio [aHR] 1.37; 95% confidence interval [CI] 1.30-1.46). Cluster 2 exhibited intermediate clinical and CVD risks (aHR 1.61; 95% CI 1.53-1.70). Cluster 3, characterised by the highest BP and fasting glucose levels, showed the highest CVD risk (aHR 1.87; 95% CI 1.74-2.00). This data-driven subtyping uncovered significant risk heterogeneity even within groups with the same number of MS criteria and was substantially more sensitive in identifying individuals with the highest-risk profile than the traditional counting method. Conclusion This clustering reveals clinical heterogeneity within the MS population. Identifying these subgroups provides a more nuanced approach to risk assessment than traditional criteria counting. This framework may help clinicians to better identify high-risk individuals who could benefit from more intensive monitoring and targeted management strategies.