System Maintenance Notice

Please note that maintenance work is scheduled between 4 AM and 5 AM - CET. During this time, you may experience reduced performance. Thanks for your patience.

Publications

Journal Article Epublish

PubMed: 42865429 DOI: 10.3389/fendo.2026.1963929

Distinct insulin sensitivity profiles revealed by clustering based on continuous glucose monitoring data in people living with type 1 diabetes.

Front Endocrinol (Lausanne) 17:1963929

Authors: Svjatoslavs Kistkins, Anastasija Luganceva, Sergejs Lobanovs, Timurs Mihailovs, Aleksejs Fedulovs, Leonora Pahirko, Valdis Pīrāgs, Othmar Moser, Harald Sourij, Jelizaveta Sokolovska, Dmitrijs Bliznuks

Affiliations

  • Research Institute of Pauls Stradins Clinical University Hospital, Riga, Latvia.
  • Faculty of Medicine and Life Sciences, University of Latvia, Riga, Latvia.
  • Institute of Applied Computer Systems, Riga Technical University, Riga, Latvia.
  • Research Institute of Pauls Stradins Clinical University Hospital, Riga, Latvia. Faculty of Medicine and Life Sciences, University of Latvia, Riga, Latvia.
  • Institute of Human Movement Science, Sport & Health, Exercise Physiology, Training & Training Therapy Research Group, University of Graz, Graz, Austria.
  • Interdisciplinary Metabolic Medicine Trials Unit, Division of Endocrinology and Diabetology, Medical University of Graz, Graz, Austria.

Abstract

BACKGROUND

Continuous glucose monitoring (CGM) has improved glycemic management in Type 1 Diabetes Mellitus (T1DM). However, interpretation of CGM data remains challenging. CGM data clustering may support individualized management and identify insulin resistance.

AIM

This study aimed to cluster CGM data based on interstitial glucose levels (2.2-27.8 mmol/L) and glucose variability, and to compare the identified clusters with biochemical and anthropometric markers of insulin sensitivity.

METHODS

CGM data from 75 T1DM patients and 20 subjects without diabetes were analyzed using Uniform Manifold Approximation and Projection (UMAP) for dimension reduction, followed by K-means clustering. Estimated glucose disposal rate (eGDR) was calculated to assess insulin sensitivity.

RESULTS

Seven clinically distinct CGM phenotypes were identified: Normoglycemia-in-Diabetes (N), Normoglycemia-in-Non-Diabetes (No), Mild Hyperglycemia (MiH), Moderate Hyperglycemia (ModH), Severe Hyperglycemia (SeH), Hypoglycemia-Associated Normoglycemia (HAN), and Hypoglycemia-Associated Hyperglycemia (HAH). The SeH cluster demonstrated the highest median average glucose level (13.83 [13.07;14.12] mmol/L), while HAH exhibited the highest median coefficient of variation (50 [44;52]%). The lowest median average glucose level was observed in HAN (6.14 [5.95;6.17] mmol/L). A marginally significant difference in ALT levels was observed between SeH and No (p=0.07). eGDR differed significantly between SeH/HAN (p=0.001), ModH/HAN (p=0.019), and MiH/HAN (p=0.043), with marginally significant differences between HAN/HAH (p=0.063). Triglycerides differed significantly between HAN/SeH (p=0.029), HAN/MiH (p=0.024).

CONCLUSION

This exploratory CGM-based clustering approach identified phenotypically distinct T1DM subgroups with differing insulin sensitivity profiles. These preliminary findings require validation in larger independent cohorts. If confirmed, such phenotyping may support individualized diabetes management and identification of insulin resistance.

MeSH

Diabetes Mellitus, Insulin Resistance, Insulin, Cluster Analysis, Precision Medicine, Diabetes Mellitus, Type 1, Endocrine System Diseases, Continuous Glucose Monitoring