Continuous glucose monitoring for optimising glycaemic performance in individuals without diabetes.
Thesis DisciplineMechanical Engineering
Degree GrantorUniversity of Canterbury
Degree NameDoctor of Philosophy
Continuous glucose monitoring (CGM) devices are becoming ubiquitous in the of care individuals with type 1 and type 2 diabetes. However, to a lesser extent, these devices could have further benefit in optimising blood glucose (BG) levels and related aspects in individuals without diabetes in physiologically stressful situations. This research centres on the metabolic effects of systemic inflammation and the use of these emerging glucose sensors to detect metabolic changes for use in computer aided monitoring and decision support. Specifically, this work investigates the use of CGM sensors in critically ill patients and endurance athletes to improve performance of glycaemic control and optimise nutrition delivery.
Critically ill patients often experience high levels of insulin resistance and stress-induced hyperglycaemia, which can negatively impact outcomes. Studies have shown glycaemic control (GC) can reduce intensive care unit (ICU) patient mortality. However, there is a significant difficulty in creating protocols that produce GC without excessive hypoglycaemia. Thus, continuous glucose monitors with their 2-5 minute measurement provide the opportunity to better monitor BG levels and thus aviod hypoglycaemia.
Recently, a CGM device specifically designed for the ICU became available, The Sentrino (Medtronic, MiniMed, Northridge, California). Hence, a large scale study was designed for implementation in Christchurch Hospital mixed medical ICU to gather further information about the feasibility of CGM in ICU. First, the performance of the Sentrino CGM system in a mixed medical ICU environment was analysed. CGM and BG data were gathered from 13 patients recruited to Phase 1 of the trial. The Sentrino device achieved a mean absolute relative difference (MARD) of 14.7%. Overall, the Sentrino performance was acceptable, but required calibration measurements every 4 hours on average. Hence, questions still remain if this CGM system can reduce the time and workload cost of glycaemic control.
Phase IIA was an observational study of the performance and integration of hyper- and hypo- glycaemic guardrail alarms with an existing, proven GC protocol. This study analysed sensor glucose (SG) and BG data, and alarm data from 8 patients recruited to this phase of the trial. Overall, sensor and alarm performance was good, only 32/93 total alarms required calibration due to a significant difference between SG and BG. However, there was a high rate of false positive alarms, 27/29 hypoglycaemic alarms were false positives.
An analysis was undertaken to better understand the impact of sepsis, oedema, and some medications on CGM results, as well as to assess the nurse compliance and feedback to gain insight of the clinical impact of using CGM to guide GC. This analysis used data from 21 patients in Phase I and Phase IIA. Approximately 50% of all enrolled subjects were severely oedematous and/or septic by design. For any CGM to be successful in reducing nurse workload and increasing patient safety, the following recommendations are made: Avoid severely oedematous patients where fluid is likely to leak from ruptured skin
Waterproof CGM sensors
Reconsider insertion technique to lessen the risk of capillary damage
Wireless transmission between sensor and monitor unit for ease of patient mobility
Before CGM can be used to guide glycaemic control protocols, the impact of suboptimal accuracy resulting from CGM error must first be characterised. The impact of CGM sensor error on the Stochastic TARgeted (STAR) GC protocol was then investigated using the simple CGM error model generated from the Sentrino data. Currently, the sensor technology trialled here is not accurate enough and if it was used to guide glycaemic control there would be a large increase in hypoglycaemic events with a median time below 4.4 mmol/L of 2.25%. In addition, the STAR protocol achieves very good and safe control already. Hence, improving the “safety” of the protocol is difficult with only 1.35% time below 4.4 mmol/L. Potentially, a less successful glycaemic control protocol would display the greater benefit of the Sentrino guardrails and CGM in general.
The glucose metabolism of athletes is not fully defined in current literature and nor are the effects of exercise on the overall glucose metabolism of athletes. There are only a few studies that investigate how metabolic parameters, such as endogenous glucose production change with exercise, and none that attempt to quantify endogenous insulin secretion or sensitivity during exercise. Additionally, an individual’s tolerance of carbohydrate is highly variable and is related to a number of factors including age and genetics. CGM devices have the potential to personalise nutrition based on glucose response. Such research using continuous glucose monitors has not been undertaken in athletic subjects before.
In a study of 10 sub-elite athletes, it was found hyperglycaemia persists >60 mins post exercise after race simulation exercise test. Plasma insulin and insulin secretion both peaked 60 mins post intense exercise to median cohort values of 256 pmol/L and 1150 pmol/min, respectively. These median peak values of plasma insulin and insulin secretion were approximately 5 and 9 times higher than the median fasting levels in this cohort, respectively. In general, this response is greater and more prolonged than reported in previous studies. The most likely reason for this outcome is subjects received two glucose boluses, one during and immediately one post exercise, as per recommended nutritional guidelines for competition while in other studies athletes remained fasted. From the same 10 athletes, a simple 1-dimensional model of endogenous insulin secretion was created, with an R2 = 0.53 and a glucose coefficient (a1) of 2559 mU.l/mmol.hr. The proposed model of endogenous insulin secretion, based on physiological measurements, provides a simple estimate of insulin secretion with comparable physiological parameters to existing literature.
The successful Intensive Control Insulin-Nutrition-Glucose model was adapted to allow insulin sensitivity to be identified during and after exercise in a well-trained cohort. The model appeared to be best able to identify insulin sensitivity during steady state periods of exercise as insulin sensitivity (SI) trends in these periods match known physiology. However, when boluses were delivered non- physiological jumps in SI occur as the model does not capture the highly patient-specific transient effects of glucose boluses, and non-constant rates of endogenous glucose production and/or non- insulin mediated glucose uptake.
The performance of CGM during exercise was investigated by comparing reference measurements to CGM data collected form the 10 subjects during the exercise test. During steady state exercise, all sensors performed better than results reported for diabetes cohorts with median MARD of 9.7%, 9.6% and 11.1% for each sensor analysed. Sensors agreed very well with each other with zero-lag cross- correlation coefficients between 0.88 and 0.97 for the different sensor pairings. Overall, these results demonstrate the good accuracy and performance of CGM devices in active athletes while exercising, confirming the applicability of these monitors for use in this new domain.
When the sensor glucose profiles of 10 athletes over a 6 day monitoring period 4/10 athletes studied spent more than 70% of the total monitoring time above 6 mmol/L even with the 2 hour period after meals removed. Only one participant spent substantial time below 4 mmol/L and this was largely due to a significantly lower overall calorie intake compared to recommendations. This study provides a unique insight in to the day to day glucose levels of athletes that could only be achieved through the use of CGM devices highlighting the need for further investigation on the recommend diets of athletes to better determine the causes and impact of the hyperglycaemia seen on health and performance.
Overall this research delineates the potential and pitfalls of using CGM to optimise blood glucose levels in ICU and athletes. In particular, it highlights areas that need to be improved before they can be relied upon to guide GC protocols in ICU. In addition, the ability to examine blood glucose trends over a longer period highlighted several aspects of the athlete metabolism that are contradictory to current literature. The results presented are promising for these devices in both fields. However, improvement in CGM sensor technology and further research is needed before CGM can be used to optimise glycaemic performance in ICU patients and athletes.