Exploring Glycaemic Variability relationships and patterns using rolling window time series

Glycaemic Variability (GV) is a widely used indicator in the management of type 1 diabetes mellitus. This study analysed the OhioT1DM dataset consisting of 12 subjects with 8 weeks of Continuous Glucose Monitoring (CGM) data for evaluating the GV metrics. Various GV metrics were explored, including the Glucose Management Indicator (GMI), Mean Amplitude of Glycaemic Excursions (MAGE), Average Daily Risk Range (ADRR), High Blood Glucose Index (HBGI), and J-index through a rolling window method to uncover their intricate relationships. These metrics were calculated using a 14-day rolling window methodology, which shifted the window daily, thereby capturing dynamic GV metric patterns and trends over time providing useful insights and potentially becoming an aid for diabetes self-management. Pearson’s correlation analysis was performed to investigate relationships between these measures. Most subjects had strong J-index and ADRR correlations (0.70 to 0.94), but two subjects demonstrated weak correlations (0.17 and 0.18). Across all subjects, the GMI and J-index were highly correlated (0.69 to 0.99), while HBGI had high correlations with both J-index (0.83-1) and GMI (0.87-1). MAGE demonstrated varying correlations with other metrics. In conclusion, the study reveals the trending capability of the rolling window and provides valuable insights into the GV metrics’ relationships, suggesting potential interchangeability between J-index, GMI, and HBGI. Finally, the absence of a significant correlation between MAGE and the other metrics suggests it has a unique role in capturing a different aspect of GV. These findings contribute to a better understanding of GV and have implications for personalised diabetes management of GV and have implications for personalised diabetes management.