Nocturnal Glucose Patterns with and without Hypoglycemia in People with Type 1 Diabetes Managed with Multiple Daily Insulin Injections

Abstract

Nocturnal hypoglycemia (NH) is a potentially dangerous and underestimated complication of insulin therapy. In this study, we aimed to determine which patterns of nocturnal glucose profiles are associated with NH in patients with type 1 diabetes (T1D) managed with multiple daily insulin injections. A dataset of continuous glucose monitoring (CGM) recordings obtained from 395 adult subjects with T1D was used for modeling. The clustering of CGM data was performed using a hierarchical clustering algorithm. Ten clusters without hypoglycemia and six clusters with NH episode(s) were identified. The differences among the clusters included initial and final glucose levels, glucose change during the night, and the presence of uptrends or downtrends. Post-midnight hyperglycemia was revealed in 5 out of 10 clusters without NH; in patterns with downtrends, initially elevated glucose prevented NH episodes. In clusters with initially near-normal glucose levels and downtrends, most episodes of NH were observed from midnight to 4 a.m.; if glucose was initially elevated, the episodes occurred at 2–4 a.m. or 4–6 a.m., depending on the time of the start of the downtrend. The results demonstrate the diversity of nocturnal glucose profiles in patients with T1D, which highlights the need for a differentiated approach to therapy adjustment.

Machine Learning and Deep Learning Models for Nocturnal High- and Low-Glucose Prediction in Adults with Type 1 Diabetes

“…There are a variety of nocturnal glucose profiles in these subjects. The patterns differ in bedtime and early morning glucose levels, the presence of downward and upward trends, and episodes of hypoglycemia; a lot of patients experience both hyperglycemia and hypoglycemia during the night [6]. Therefore, new technological solutions are urgently needed to predict glucose in patients managed with MDIs.…”