Identification of distinct metabolic subtypes among Japanese men with prediabetes: a data-driven cluster analysis...

To identify distinct metabolic subtypes among Japanese men with prediabetes and evaluate their association with the future risk of type 2 diabetes.

A data-driven cluster k-means cluster analysis was conducted in 2,172 men with prediabetes, defined as a fasting plasma glucose (FPG) concentration of 100-125 mg/dL (5.6-6.9 mmol/L) from the JDOIT-1 cohort. Four biochemical indicators-body mass index (BMI), fasting plasma glucose (FPG), alanine aminotransferase (ALT), and non-HDL cholesterol-were standardized and used for clustering. The optimal number of clusters (k = 3) was determined using the elbow method. The association between cluster membership and 5-year diabetes-free survival was examined.

Three metabolic phenotypes were identified: Cluster 1 (MRP, n = 1,149), metabolically resilient; Cluster 2 (OFIP, n = 581), obesity- and insulin-resistant; and Cluster 3 (NOHP, n = 442), non-obese hyperglycemic. Annual diabetes incidence rates were 0.70%, 2.73%, and 6.73% in the MRP, OFIP, and NOHP groups, respectively (p < 0.001). Compared with MRP, adjusted hazard ratios were 3.95 (95% CI 2.64-5.92) for OFIP and 9.81 (95% CI 6.75-14.26) for NOHP. Lifestyle interventions significantly reduced diabetes risk only in the OFIP group (HR 0.59, 95% CI 0.36-0.97, p = 0.037).

Unsupervised clustering identified distinct metabolic subtypes predictive of diabetes onset. Diabetes risk increased progressively from metabolically resilient to insulin-resistant and hyperglycemic phenotypes. Targeted lifestyle interventions may be particularly effective in individuals with obesity- and dyslipidemia-related insulin resistance.

University hospital Medical Information Network (UMIN) Center UMIN000000662)、Approval date 11 December 2006.

The online version contains supplementary material available at 10.1007/s13340-026-00917-2.