Data-driven cluster analysis of lipids, inflammation, and aging in relation to new-onset type 2 diabetes mellitus

Abstract

Purpose: Early detection and intervention are vital for managing type 2 diabetes mellitus (T2DM) effectively. However, it's still unclear which risk factors for T2DM onset are most significant. This study aimed to use cluster analysis to categorize individuals based on six known risk factors, helping to identify high-risk groups requiring early intervention to prevent T2DM onset.

Methods: This study comprised 7402 Korean Genome and Epidemiology Study individuals aged 40 to 69 years. The hybrid hierarchical k-means clustering algorithm was employed on six variables normalized by Z-score-age, triglycerides, total cholesterol, non-high-density lipoprotein cholesterol, high-density lipoprotein cholesterol and C-reactive protein. Multivariable Cox proportional hazard regression analyses were conducted to assess T2DM incidence.

Results: Four distinct clusters with significantly different characteristics and varying risks of new-onset T2DM were identified. Cluster 4 (insulin resistance) had the highest T2DM incidence, followed by Cluster 3 (inflammation and aging). Clusters 3 and 4 exhibited significantly higher T2DM incidence rates compared to Clusters 1 (healthy metabolism) and 2 (young age), even after adjusting for covariates. However, no significant difference was found between Clusters 3 and 4 after covariate adjustment.

Conclusion: Clusters 3 and 4 showed notably higher T2DM incidence rates, emphasizing the distinct risks associated with insulin resistance and inflammation-aging clusters.