Glucose Control Using a Standard Versus an Enhanced Hybrid Closed Loop System: A Randomized Crossover Study

Background

The MiniMed 670G system (Medtronic, Northridge, CA) is the first commercially available hybrid closed loop (HCL) insulin delivery system. It is currently used by people living with type 1 diabetes (T1D) in the United States and recently has received Conformite Europeene, Canadian regulatory and Therapeutic Goods Administration approvals, paving the way for use in Europe, Canada, and in Australia, respectively. The Medtronic 670G utilizes an PID (proportional–integral–derivative) algorithm with insulin feedback. ¹ Previous studies demonstrated that the system was effective and safe in adolescents and adults with T1D, with HCL glucose control engaged for a median of 87% of the time. ²

Insights gained from trials and clinical use have informed iterative modifications to the standard HCL (s-HCL) system, resulting in a prototype enhanced HCL (e-HCL) system. Medtronic's automatic insulin delivery system has several tunable parameters that can be adjusted to improve glycemic efficacy, safety, and user experience. In the iteration of the system (e-HCL) investigated in this study, the modifications incorporated focused on improving the user experience with the aim of increasing the time spent in auto mode and reducing device-related alarms. Furthermore, it has been clinically observed that increased time in auto mode is associated with superior overall glycemic control.

To attain this goal, the system's time-out thresholds were increased for its personalized minimum and maximum delivery. This decision was based upon a large retrospective analysis with the MiniMed 670G system (Medtronic) and an in silico analysis that indicated that the system could tolerate longer periods of insulin suspension without the risk of ketosis. Conversely, the HCL insulin delivery system was also modified to safely tolerate longer insulin delivery, at its personalized maximum insulin deliver rates, without the risk of insulin stacking. The changes made to the e-HCL time-out thresholds for minimum and maximum insulin delivery rates were optimized to achieve longer time in HCL and to enhance user experience with fewer alarms and less distress. These changes required the retuning of the gains of the HCL's PID controller to maintain the expected safety and efficacy of the system

Methods

It was our objective to perform an exploratory two-stage randomized crossover study involving adults with T1D, comparing glucose control, closed loop (CL) exits, and alarm frequency with the s-HCL versus e-HCL systems (ACTRN12617001449325). Inclusion criteria were age >18 years, T1D of >1 year duration, stable on insulin pump therapy for >3 months, proficient in carbohydrate counting, continuous glucose monitoring (CGM) sensor experience, and HbA1c <10.0% [86 mmol/mol]. Exclusion criteria were pregnancy, eGFR (estimated Glomerular Fitration Rate) and the units are mL/min/1.73m2 <40, diabetic ketoacidosis or severe hypoglycemia in the past 3 months, diabetic gastroparesis, tape allergy, and major medical or psychiatric illness.

After Human Research and Ethics committee approval and individual written informed consent, participants were assigned s-HCL and e-HCL systems in random order (using sealed opaque envelopes) before run-in, for 1 week each (and 1 week washout between stages), undertaken in a supervised live-in (hotel) setting. For both s-HCL and e-HCL stages, insulin delivery was through a suitably configured Medtronic pump with identical interventions designed to challenge glucose control and persistence in CL. These challenges included two missed meal boluses, two high-glycemic index and two high-fat meals (breakfasts and dinners), and a 40-min bout of moderate-intensity exercise. Challenges were separated by at least 8 h. The primary outcome was CGM sensor time in target range (3.9–10.0 mmol/L). Secondary outcomes included additional CGM metrics and algorithm-related alerts and CL dropouts as detailed in Table 1. All participants answered a standardized questionnaire and were interviewed individually by an investigator regarding their experience with s-HCL and e-HCL at study conclusion. As this represented an exploratory study, data to determine sample size were not available. Statistical analysis was by paired t-test for normally distributed data and otherwise by Wilcoxon-signed rank test.

Results

The study was conducted between October 22, 2017 and November 16, 2017. All participants completed the protocol. All 11 pump-experienced T1D adults (9 women; mean [SD] age: 51 years [15 years]; diabetes duration 26.4 years [7.7 years]; HbA1c 7.5% [1.0 %] or 58 mmol/mol [7.7 mmol/mol]) completed both e-HCL and s-HCL stages. Insulin action times and carbohydrate-to-insulin ratios were identical when participants were using s-HCL and e-HCL. Results are summarized in Table 1. In comparison with s-HCL, e-HCL resulted in significantly fewer CL alerts (8.6 [5.8] per week vs. 3.9 [2.8] per week; P = 0.01) and CL exits (3.5 [3.1] per week vs. 0 [0.0] exits per week; P = 0.004). Differences between s-HCL and e-HCL, respectively, for time in target range (69% [11%] vs. 74% [10%]; P = 0.15), mean glucose (8.9 mmol/L [1.5 mmol/L] vs. 8.4 mmol/L [0.6 mmol/L]; P = 0.09), and time in CL (96.9% [5.1%] vs. 98.5 [0.8]; P = 0.23, equating to ∼3 h for 1 week) favored e-HCL but did not reach statistical significance. No episodes of severe hypoglycemia or ketoacidosis occurred. Participants underwent standardized interviews at study completion. All expressed a preference for the e-HCL over the s-HCL. Reasons given for the preference for e-HCL included a better user interface (n = 8), a perception of improved glucose control (n = 6), a requirement for fewer calibrations to remain in CL (n = 2), and reduced alerts and exits from CL (n = 8). The majority (9 of 11) of participants indicated that they were either “extremely satisfied” or “very satisfied” with their overall experience with the e-HCL system and 2 of 11 indicated that they were “somewhat satisfied.” When specifically requested to indicate their preference for further future improvements, the majority of responses from the participants at the conclusion of the study focused on glucose control and related to a reduction in the glucose level targeted by the system, the ability to adjust this target, or more aggressive corrections to reduce high-glucose levels.

Discussion

The 670G s HCL system has previously demonstrated safety among people living with T1D. ² Nevertheless, the 670G HCL system as it currently stands does not represent an endpoint, but rather should be seen as an important evolutionary step on the journey toward full automation of subcutaneous insulin delivery. Therefore, the system will be subject to refinements in light of clinical experience. Our data demonstrate that further refinement is possible and valued by users.

A limitation of this protocol is that with multiple glycemic challenges during 1 week in a supervised setting, the study does not reflect the everyday experience of a person with T1D. Therefore, the mean sensor glucose outcomes measured are not directly comparable with those of other CL studies.

Although mean sensor glucose was higher in this study than studies reported previously, due to multiple glycemic challenges, these preliminary data in this pilot study suggest safety, with significantly reduced CL exits and alerts, user preference for e-HCL over s-HCL, and a trend toward improved glucose control with e-HCL in an adult group. Further refinement of the CL system may be merited and larger longer term home-based studies more reflective of usual care are required to confirm and extend these findings.