Background: In recent years, new strategies for diagnosing and treating diabetes mellitus have been developed, with novel diabetes stratifications emerging from the perspective of predicting metabolic disorders and complications to facilitate a personalized treatment approach. One particularly interesting approach is a classification based on five parameters (glycated hemoglobin - HbA1c, BMI, age at diagnosis, HOMA-IR and HOMA-β indices), which identifies five clusters. Three of these clusters include patients with type 2 diabetes: Severe Insulin-Resistant Diabetes (SIRD); Mild Age-Related Diabetes (MARD); and Mild Obesity-Related Diabetes (MOD) [1]. However, the practicality and clinical applicability of this logical classification system require further investigation.

Objectives: To conduct a comparative analysis of laboratory parameters characterizing carbohydrate and lipid metabolism in patients with different diabetes phenotypes, categorized according to cluster analysis criteria.

Materials and Methods: A retrospective analysis was conducted on data from 83 patients with type 2 diabetes mellitus (T2DM) who were hospitalized at the endocrinology department of the Federal State Budgetary Institution "Central Clinical Hospital of Civil Aviation" between 2024 and 2025. The characteristics of the course of diabetes (age of onset, combination of complications, BMI, etc.), glycemic and lipid profile indicators, insulin, and C-peptide in patients of clusters were assessed. SIRD (39 people), MARD (18 people) and MOD (26 people).

The study was approved by the Ethics Committee of the Patrice Lumumba Peoples' Friendship University of Russia. – protocol No 8 of 14.10.2025

Results: Postprandial blood glucose levels were significantly higher in the SIRD cluster (8.44 ± 2.93 mmol/L) compared to the MOD cluster (6.66 ± 1.78 mmol/L). Target HbA1c levels were achieved by 66.7% of patients in the MARD cluster, compared to 25.6% in the SIRD cluster and 26.9% in the MOD cluster. HDL cholesterol levels were significantly higher in the MARD cluster (1.29 ± 0.37 mmol/L) than in the MOD cluster (1.08 ± 0.24 mmol/L; p<0.01). 

The timing of type 2 diabetes manifestation differed across clusters. Earlier disease onset was observed in the MOD cluster, with a mean age of 47.62 ± 7.14 years, compared to 50.15 ± 9.79 years in the SIRD cluster and 66.53 ± 5.49 years in the MARD cluster. Diabetes duration was longest in the SIRD cluster at 16.70 ± 8.50 years, compared to 6.92 ± 4.85 years in the MARD cluster and 12.96 ± 7.57 years in the MOD cluster.

Conclusions: Cluster-based classification shows promise for implementing a personalized treatment approach. The study established that glycemic control was least effective in the SIRD cluster, which also had the longest overall disease duration. Patients in the MARD cluster achieved diabetes compensation more frequently and demonstrated lower cardiometabolic risks. An earlier disease onset (MOD) did not consistently correlate with a more severe disease course. Cluster differentiation enables the personalization of treatment strategies.