Abstract:
The dawn phenomenon is characterized by a rise in blood glucose levels in the early morning, usually between 4:00 AM and 8:00 AM. It is very common in individuals with diabetes. The rise has been speculated to be due to hormonal regulation that promotes glucose production and reduces insulin sensitivity. Continuous Glucose Monitoring (CGM) is a tool that can be used to assess these trends and provide information to better understand and detect this phenomenon at the individual level. This study focuses on CGM data from a single adult with diabetes over 180 days to evaluate patterns related to the dawn phenomenon and to train an AI model to readily detect these trends and use the patient's past data to predict future results, specifically accounting for the dawn phenomenon. Glucose levels were measured between 3:00 AM and 8:00 AM, and trends were recorded to fine-tune the model to the study and the patient’s interests. A personalized model was developed to predict early-morning glucose increases based on nocturnal glucose trends. Results showed consistent, significant increases of 20-200mg/dL above the nocturnal nadir. The model effectively predicted glucose elevation with minimal error after repeated data inputs and manual fine-tuning. Application of the model was associated with a 2 percentage-point reduction in hemoglobin A1C, suggesting meaningful impacts on duration. CGM enables detailed analysis of the dawn phenomenon on an individual basis, which can be further refined depending on the amount of data given. Further research is needed to generalize the study's findings and to validate them with other clinical cases.
