Statistical methods for Continuous Glucose Monitoring (CGM) data

Diabetes is a chronic disease resulting from the pancreatic inability to produce insulin in reaction to blood glucose levels. There are two main types of diabetes. Type 1 diabetes is characterized by absolute insulin deficiency, whereas Type 2 is characterized by insulin resistance. In 2014, approximately $422$ million people worldwide were living with Type 1 or 2 diabetes, with $30$ million people being affected in the US alone. High levels of blood glucose observed in diabetes lead to increased risk of adverse health effects including retinopathy, cardiovascular disease, lower extremity amputations cognitive dysfunction, and premature morbidity and mortality. Thus, the primary treatment goal in diabetes is glucose control. However, a normal glucose profile is non-constant, with typical normal values varying between $70$-$120$ mg/dL and peaks associated with meal intakes. The highly non-linear and non-stationary nature of glucose profiles is due to a wide range of environmental factors including time, quantity and composition of meals, physical activity time, intensity and type, stress, and sleep quality.

Continuous Glucose Monitors (CGMs) are small wearable devices that measure the glucose levels continuously throughout the day, with some monitors taking measurements as often as every 5 minutes. Data from these monitors provide a detailed quantification of the variation in blood glucose levels during the course of the day, and thus CGMs play an increasing role in clinical practice. For more on CGMs, see Rodbard (2016) “Continuous Glucose Monitoring: A Review of Successes, Challenges, and Opportunities.”.

Out group focuses on developing statistical methods to aid analyses and interpretation of CGM data. Some selected research products are below.

CGM datasets

Our team has released a list of public Continuous Glucose Monitoring (CGM) datasets. Thank you to an amazing team of undergraduate researchers: Mary Martin, Elizabeth Chun, David Buchanan, Eric Wang and Sangaman Senthil who assembled this collection as part of their Aggie Research Project. To cite the collection:

• Mary Martin, Elizabeth Chun, David Buchanan, Eric Wang, Sangaman Senthil & Irina Gaynanova. (2020, June 15). irinagain/Awesome-CGM: List of public CGM datasets (Version v1.0.0). Zenodo.

Software products:

• R package iglu for calculating various metrics from CGM glucose profiles, and visualizing the data. Github repository, CRAN version, accompanying website and the paper. The package has graphical user interface via shiny app, which can be accessed locally after installing the package, or directly from the website (see below). To cite the current version of the package:

  • Broll S*, Buchanan D*, Chun E*, Muschelli J*, Fernandes N*, Seo J*, Shih J*, Urbanek J, Schwenck J*, Gaynanova I (2021). iglu: Interpreting Glucose Data from Continuous Glucose Monitors. R package version 3.1.0

  • Chun E*, Fernandes JN* and Gaynanova I (2024+) An Update on the iglu Software for Interpreting Continuous Glucose Monitoring Data. Diabetes Technology and Therapeutics

  • Broll S*, Urbanek J, Buchanan D*, Chun E*, Muschelli J, Punjabi N and Gaynanova I (2021).Interpreting blood glucose data with R package iglu. PLoS One, Vol. 16, No. 4, e0248560.

• Shiny app for iglu. R package iglu functionality directly via graphical user interface.

Relevant publications:

• Coulter A*, Aurora RN, Punjabi N and Gaynanova I (2024+). Fast variable selection for distributional regression with application to continuous glucose monitoring data arXiv

• Chun E*, Fernandes JN* and Gaynanova I (2024+) An Update on the iglu Software for Interpreting Continuous Glucose Monitoring Data. Diabetes Technology and Therapeutics [CRAN R package] [GitHub repository]

• Chun E*, Gaynanova I, Melanson E and Lyden, K (2024) Pre- Versus Postmeal Sedentary Duration; Impact on Postprandial Glucose in Older Adults with Overweight or Obesity. Journal for the Measurement of Physical Behaviour, Vol. 7, No. 1.

• Sergazinov R*, Chun E*, Rogovchenko V*, Fernandes N*, Kasman N* and Gaynanova I (2024). GlucoBench: Curated List of Continuous Glucose Monitoring Datasets with Prediction Benchmarks. International Conference on Learning Representations (ICLR). [GitHub repository]

• Sergazinov R*, Leroux A, Cui E, Crainiceanu C, Aurora RN, Punjabi N and Gaynanova I (2023). A case study of glucose levels during sleep using fast function on scalar regression inference. Biometrics, Vol. 79, No. 4, 3873-3882.

• Sergazinov R*, Armandpour M and Gaynanova I (2023). Gluformer: Transformer-Based Personalized Glucose Forecasting with Uncertainty Quantification. IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP) [Python code]

• Gaynanova I (2022). Digital biomarkers of glucose control - reproducibility challenges and opportunities. ASA Biopharmaceutical Report, Vol. 29, No. 1, 21-26.

• Aurora RN, Gaynanova I, Patel P*, and Punjabi N (2022). Glucose profiles in obstructive sleep apnea and type 2 diabetes mellitus. Sleep Medicine, Vol. 95, 105-111

• Gaynanova I, Punjabi N and Crainiceanu C (2022). Modeling continuous glucose monitoring (CGM) data during sleep. Biostatistics, Vol. 23, No. 1, 223-239. [R code]

• Fernandes N*, Nguyen N*, Chun E*, Punjabi N and Gaynanova I (2022). Open-Source Algorithm to Calculate Mean Amplitude of Glycemic Excursions Using Short and Long Moving Averages. Journal of Diabetes Science and Technology, Vol. 16, No. 2, 576-577. [Github page to reproduce] [R package with implementation]

• Broll S*, Urbanek J, Buchanan D*, Chun E*, Muschelli J, Punjabi N and Gaynanova I (2021). Interpreting blood glucose data with R package iglu. PLoS One, Vol. 16, No. 4, e0248560. [R package]

• Gaynanova I, Urbanek J and Punjabi N (2018). Letter to the Editor: Corrections of equations on glycemic variability and quality of glycemic control. Diabetes Technology & Therapeutics, Vol. 20, No. 4, 317.