Quantifying temporal glucose variability in diabetes via continuous glucose monitoring: mathematical methods and clinical application

Diabetes Technol Ther. 2005 Dec;7(6):849-62. doi: 10.1089/dia.2005.7.849.


Background: Continuous glucose monitors (CGMs) collect detailed blood glucose (BG) time series, which carry significant information about the dynamics of BG fluctuations. In contrast, the methods for analysis of CGM data remain those developed for infrequent BG self-monitoring. As a result, important information about the temporal structure of the data is lost during the translation of raw sensor readings into clinically interpretable statistics and images.

Methods: The following mathematical methods are introduced into the field of CGM data interpretation: (1) analysis of BG rate of change; (2) risk analysis using previously reported Low/High BG Indices and Poincare (lag) plot of risk associated with temporal BG variability; and (3) spatial aggregation of the process of BG fluctuations and its Markov chain visualization. The clinical application of these methods is illustrated by analysis of data of a patient with Type 1 diabetes mellitus who underwent islet transplantation and with data from clinical trials.

Results: Normative data [12,025 reference (YSI device, Yellow Springs Instruments, Yellow Springs, OH) BG determinations] in patients with Type 1 diabetes mellitus who underwent insulin and glucose challenges suggest that the 90%, 95%, and 99% confidence intervals of BG rate of change that could be maximally sustained over 15-30 min are [-2,2], [-3,3], and [-4,4] mg/dL/min, respectively. BG dynamics and risk parameters clearly differentiated the stages of transplantation and the effects of medication. Aspects of treatment were clearly visualized by graphs of BG rate of change and Low/High BG Indices, by a Poincare plot of risk for rapid BG fluctuations, and by a plot of the aggregated Markov process.

Conclusions: Advanced analysis and visualization of CGM data allow for evaluation of dynamical characteristics of diabetes and reveal clinical information that is inaccessible via standard statistics, which do not take into account the temporal structure of the data. The use of such methods improves the assessment of patients' glycemic control.

Publication types

  • Case Reports
  • Research Support, N.I.H., Extramural

MeSH terms

  • Adult
  • Aged
  • Aged, 80 and over
  • Amyloid / therapeutic use
  • Blood Glucose / analysis*
  • Blood Glucose Self-Monitoring / methods
  • Data Interpretation, Statistical
  • Diabetes Mellitus, Type 1 / blood*
  • Diabetes Mellitus, Type 1 / drug therapy
  • Diabetes Mellitus, Type 1 / surgery
  • Female
  • Humans
  • Hypoglycemic Agents / therapeutic use
  • Islet Amyloid Polypeptide
  • Islets of Langerhans Transplantation
  • Male
  • Markov Chains
  • Middle Aged
  • Monitoring, Ambulatory / instrumentation
  • Monitoring, Ambulatory / methods*
  • Randomized Controlled Trials as Topic


  • Amyloid
  • Blood Glucose
  • Hypoglycemic Agents
  • Islet Amyloid Polypeptide
  • pramlintide