Clusters of Nocturnal Glucose Dynamics in Patients with Type 1 Diabetes Managed by Continuous Subcutaneous Insulin Infusion

Background and Aim: Continuous subcutaneous insulin infusion (CSII) with sensor-augmented pumps is widely used in the treatment of type 1 diabetes. Patterning of glucose profiles using mathematical methods may be the basis for personalizing glycemic management in these patients. In this study, we performed clustering of nocturnal glucose dynamics in adult patients with type 1 diabetes managed with CSII. Materials and Methods: A dataset of continuous glucose monitoring recordings obtained from 111 patients with type 1 diabetes was used for modeling. Time series from 0 to 6 a.m. were analyzed. We applied a hierarchical clustering algorithm for identifying clusters of glucose dynamics. Monte Carlo method and the Silhouette Score were used for quality assessment of the clustering. Results: Six non-hypoglycemic clusters and four clusters with low glucose episodes were recognized. The most common patterns were characterized by close to normal glucose levels with or without a slight downtrend. Three other non-hypoglycemic clusters were distinguished by apparent hyperglycemia after midnight and marked downtrends. Two clusters represented stable, but increased to varying degrees, glucose concentrations. The most common pattern with low glucose episode(s) was characterized by an initial glucose just above the hypoglycemic threshold, and a slow decrease to low values by 2 to 4 a.m. N ext cluster showed a permanent downtrend with most episodes of low glucose between 4 and 6 a.m. Two rare patterns represented initial downward trends, a mean time from midnight to the first low glucose episode in these clusters was 0.6 and 2.2 hours. Conclusion: The results of this study demonstrate different patterns of nocturnal glucose dynamics in patients with type 1 diabetes managed with CSII. Patterning glucose dynamics may be a useful tool for personalized adjustment of basal insulin infusion rate in these patients.