BIC

Malfunction of blood-glucose-level regulation causes microvascular and metabolic dysfunctions that result in progressive failure of many organ systems in the body, leading to complications, such as blindness, neuropathy, kidney failure, lower-limb amputations, and cardiovascular disease. Diabetes mellitus, characterized by an above-normal glucose level, is a result of inadequate insulin production, insulin action, or both. Both instances of glucose dysregulation have significant impacts on the medical health care systems.

Excursions of glucose levels from desired range can be mitigated by closely monitoring the blood glucose values and accordingly administer insulin or glucose or drugs to alter tissue glucose metabolism. Importantly, as glucose levels are adjusted artificially, both hyper- and hypo-glycemic states must be avoided. For example, an aggressive control of glycemic state involves the risk of entering hypoglycemia episodes, which can result in loss of consciousness, seizure, coma, and even death.

Recent developments and expected near-future improvements in continuous glucose monitoring (CGM) devices open new opportunities for glycemia management of diabetes mellitus patients. Some of these opportunities include the application of open-loop/manual control in situations where ad-hoc regulation of glucose levels is necessary, and, eventually, in concert with advanced infusion pumps for closed-loop/automatic control applications. However, in its current configuration, CGM only provides information about a patient's current glycemic state, resulting in reactive glucose regulatory interventions (i.e., the glucose level may already be at an unacceptably high or low level) rather than proactive interventions.

We at the BIC are focusing our efforts on predictive monitoring to provide the capability to indicate the need for early, proactive intervention to adjust therapy before glucose levels drift from the desired range. In doing so, we couple CGM devices with mathematical models, so that current and a set of previous glucose measurements can be used to predict glucose levels in the future. Practical application of such predictive monitoring requires mathematical models that are both highly predictive and individualized -- capable of estimating significantly different individual responses to insulin, meals, and daily activities -- and portable from individual to individual -- requiring minimum, if any, manual model tuning for each individual.

Employing CGM data from type 1 diabetic patients collected over a continuous five-day period, we have developed autoregressive models capable of predicting glucose values up to 30-min ahead, with an accuracy within ±15% of actual measured value (top panel in figure). The clinical accuracy of such predictions was evaluated using the Clarke-error grid, which reveals that over 98% of the prediction points lie in the clinically-acceptable zones A and B of the grid.

Our future efforts include: (1) developing individual-specific autoregressive with exogenous input (ARX) models, which can explicitly take into account the individual's body mass index (BMI), age, weight, height, and glycemic index (GI) and/or calories of the diet intake, and (2) developing biomathematical first-principles models based on glucose metabolism and regulatory mechanisms.