Accelerometer Use in the Prevention of Exercise-Associated Hypoglycemia in Type 1 Diabetes: Outpatient Exercise Protocol

Regular aerobic exercise confers a plethora of health benefits to all individuals and is considered an essential component of the management of type 1 diabetes (T1D) [1]. However, in contrast to non-diabetic subjects - in whom the increased muscle energy requirement during exercise leads to suppression of endogenous insulin secretion - patients with T1D are dependent upon exogenous insulin and are thus at risk for exercise-associated hypoglycemia [1]. Exercise-associated hypoglycemia is the most frequently reported adverse event related to exercise in diabetes [2] and hypoglycemia can occur during exercise or several hours afterwards [3,4]. Although previous research has shown that pre-meal dose reduction of subcutaneous insulin can be effective at decreasing the incidence of exercise-associated hypoglycemia [5], patients with T1D often do not perform such adjustments [6,7].

In contrast to subcutaneous insulin injections, which are reliant upon the patient or caretaker to determine dosage, the insulin pump provides a unique opportunity to avoid hypoglycemia via user-independent, computer-based algorithms for determining insulin delivery. Previous research conducted here at Stanford has demonstrated that algorithms based on continuous glucose monitor (CGM) data can prevent hypoglycemia in the sedentary setting by inducing insulin pump suspension [8-10]. In addition, a study of children and adolescents conducted at Stanford (as a center in the DirecNet group) demonstrated that suspending an insulin pump at the beginning of a period of moderate aerobic exercise reduces the risk of hypoglycemia during that exercise period and subsequently overnight [11]. Thus, by utilizing exercise-detecting accelerometers and an algorithm to initiate pump suspension during exercise, it is likely possible that people with diabetes could avoid exercise-associated hypoglycemia even if they failed to manually alter their pump settings. However, to date, no published studies have utilized accelerometer-derived data in an insulin pump suspension algorithm during exercise.

Accelerometers are light-weight motion-sensing devices that can be worn to provide information about the intensity and duration of physical activity [12]. They are small, inexpensive, and could easily be incorporated into current sensors and "patch" pumps. They can also be used independently or combined with a heart rate monitor (HRM) [13], although most commercially available HRMs currently require a chest strap that can be uncomfortable to wear. Previous studies evaluating the effect of physical activity on insulin sensitivity have utilized accelerometers (worn on a belt at the small of the back, the right side of the trunk in the mid-axillary line, or the left side of the chest) with and without HRMs for activity recognition during subjects' everyday lives. These data were used to classify activity as sedentary, light, moderate, or vigorous based on acceleration signal counts measured over one-minute intervals [13-17]. One study investigated four different accelerometers in a clinical research setting and found each to be very accurate in assessing the intensity of physical activity, regardless of subjects' body habitus [18]. Thus, these devices can provide a reliable means by which the onset, duration, and intensity of exercise can be recognized and reported in real-time to the other components of an artificial pancreas. When combined with CGM and insulin delivery data, this exercise information is a valuable tool in designing an algorithm to decrease or stop insulin delivery in order to decrease the risk of exercise-associated hypoglycemia.

In the first phase of this study (in press), 22 subjects with type 1 diabetes went about their everyday lives while wearing an insulin pump, CGM, and accelerometer/heart rate monitor. After the monitoring period, the devices were downloaded and the data were used to augment an existing predictive low glucose suspend (PLGS) algorithm to incorporate activity. In a computer simulator, the PLGS algorithm reduced hypoglycemia by 64%, compared to 73% and 76% reductions for the accelerometer-augmented and HRM-augmented algorithms, respectively.

In the next phase of this study, we seek to test the newly developed algorithm in a real-life setting in the form of a structured sports (soccer) camp to further see if modifications to the algorithm are required.