Improved Satisfaction While Maintaining Safety and High Time in Range (TIR) With a Medtronic Investigational Enhanced Advanced Hybrid Closed-Loop (e-AHCL) System

The first commercially available automated insulin delivery (AID) system, the MiniMed 670G System (670G), was initially approved by the U.S. Food and Drug Administration in September 2017 for use in people with type 1 diabetes ≥14 years of age (1). Data from randomized controlled studies provided evidence of improvement in glycemia in adults and children (2,3). However, the real-world experience derived from clinical audits in some centers with this first-in-class device suggested a significant discontinuation rate likely due to the high number of alarms, the workload imposed upon the user to maintain the system in closed loop, and dissatisfaction with continuous glucose monitoring (CGM) sensor performance (4–6).

Since 2017, additional AID systems have become commercially available, including the MiniMed 780G system (780G). The 780G incorporates refinements implemented in the light of clinical experience with the 670G system. These refinements include a choice of glucose targets, an automated correction bolus, reduced alarms, and reduced requirements by the system for a confirmatory finger-stick glucose reading and greater persistence in AID mode. In addition, the commercially available 700 series pumps have Bluetooth connectivity facilitating the automatic transmission of device data seamlessly and automatically to the cloud, without user input, as was previously required (7).

Glucose control reflected by %CGM time in target (70–180 mg/dL) glucose range (%TIR), and hypoglycemia reduction as assessed by %CGM time below range (TBR) of <70 mg/dL with currently available AID systems represent a substantial improvement over manual insulin dosing (8–14). Both clinical trial and real-world data indicate that these systems are, on an average, able to achieve the international consensus recommended %TIR targets of >70%, while simultaneously reducing %TBR of <70 mg/dL to <4.0% (8–14).

Consequently, because of the greater emphasis that is now being placed on the user experience, such as that related to satisfaction with the technology, and psychological outcomes, additional refinements have been made to 780G to further enhance the user experience. These changes have resulted in an investigational enhanced advanced hybrid closed-loop system (e-AHCL) system that incorporates an updated algorithm; a new, easy to use, no calibration, all-in-one sensor; and a 7-day infusion set with extended durability.

We hypothesized that e-AHCL is feasible for managing insulin delivery and will provide an excellent user experience without any compromise in glucose control. Our aim was to determine feasibility and treatment satisfaction of the e-AHCL system in adults with type 1 diabetes in comparison with earlier iterations of 780G.

We conducted a single-center nonrandomized three-stage exploratory study (Supplementary Fig. 1) evaluating user perceptions and glycemic outcomes associated with Medtronic’s e-AHCL system compared with the earlier Medtronic AID systems (clinical trial no. ACTRN12622000017729). The trial protocol was approved by a central Human Research Ethics Committee (St Vincent’s Hospital Melbourne HREC Reference: HREC 277/21). Written informed consent was provided by all participants.

The study comprised a run-in period of 12 weeks duration with the participants using their current AID system, which was a non-Bluetooth 780G system (670G V4.0 insulin pump) requiring manual upload of data and lacking the target of 110 mg/dL, with Guardian Sensor 3 (GS3) requiring calibration and with a commercially available insulin infusion set approved for 3 days use. Following run-in, sequentially in three stages of 12 weeks each, all participants were educated in the operation of and provided with the following: 1) a Bluetooth-enabled 780G insulin pump that automatically uploaded pump data, in conjunction with GS3, and a 3-day insulin infusion set (stage 1) followed by 2) a Bluetooth enabled Medtronic 780G insulin pump, a calibration-free Guardian sensor 4 (GS4), and a 3-day insulin infusion set (stage 2) and, finally, 3) the e-AHCL system incorporating a Bluetooth-enabled 780G insulin pump with an updated AHCL algorithm that included modifications to the calculation of the autocorrection boluses and daily user adaptations, a new easy-to-use calibration-free all-in-one sensor, and a 7-day extended insulin infusion set with a 3-mL reservoir (stage 3) (Medtronic, Northridge, CA). AID system characteristics are summarized in Supplementary Table 1.

There were five study visits. Following consent, a clinical history and physical examination were performed, with blood collected for HbA_1c and electrolytes. All study participants were provided with and instructed in the use of a bespoke phone app (KeyLead Health, Melbourne, Australia) on which, for the study duration, they entered data in real time regarding time spent uploading the pump, changing the sensor, and changing the insulin set. When the participant had all the relevant materials in front of them, they pressed a “button” on their mobile phone when the activity commenced, and, when the activity was completed, they pressed it once more. The information was transmitted via the cloud in real time to a study-specific platform where the information was recorded under the participant’s unique study identification number. Blood was collected for HbA_1c, and questionnaires were administered following run-in and at the end of each study stage.

Participants eligible for inclusion in the study had a clinical diagnosis of type 1 diabetes of >1 year duration, were stable on sensor-augmented insulin pump therapy for >3 months, were proficient in carbohydrate counting, and had an HbA_1c of <10.0% (<86 mmol/mol). Excluded from the study were women who were pregnant or desiring pregnancy, those with a history of diabetic ketoacidosis (DKA) within the last 3 months, an eGFR (estimated glomerular filtration rate) of <40 mL/min/1.73 m², gastroparesis, or any major medical or psychiatric illness potentially impacting the participant’s ability to complete the study.

The primary outcome was treatment satisfaction measured by the Difference in the Diabetes Technology Questionnaire (DTQ) current score (15,16). DTQ is a 30-item validated measure of satisfaction with diabetes technological tools used in type 1 diabetes management. It yields separate scores for “current” satisfaction (DTQ-current: How much is this a problem now?) and for “change” in satisfaction (DTQ-change: How has it changed compared with before the study?), with a range of 5 for each score and a total score range of 30 to 150. It has been used in other studies evaluating novel diabetes technologies (17,18).

Secondary participant-reported outcomes included the System Usability Scale (SUS) (19), Diabetes Medication System Rating Questionnaire short form (DMSRQ-SF) (20), User Experience Questionnaire (UEQ) (21), Problem Areas in Diabetes (22), Pittsburgh Sleep Quality Index (23), and the Fear Survey short form (24). Hypoglycemia awareness was assessed by the Gold questionnaire with a score of ≥4 suggesting impaired awareness of hypoglycemia (25). Outcomes, collected continuously in real time using the KeyLead Health phone app, were time to uploaded device data for review, time to change insulin sets, and time to change glucose sensors.

Secondary glycemic outcomes included CGM metrics standardized according to convention (26) and HbA_1c at baseline and at the end of the 12-week intervention.

Safety outcomes included severe hypoglycemia (requiring third-party assistance) and DKA.

Systems performance outcomes included percent time CGM was used and percent time in closed loop, the number of finger-stick glucose measurements, number of alarms, number of closed-loop exits, and average duration of wear of the sensor and of the insulin set.

The exploratory nature of the study precludes a power calculation because there is no prior randomized controlled study using the DTQ-current score to assess treatment satisfaction comparing different iterations of insulin pump devices. The sample size was chosen to correspond to previous exploratory closed-loop studies, and determined as per time and resources. The statistical analysis plan was agreed to by investigators in advance with analyses performed on an intention-to-treat basis. We anticipate that 22 participants with an estimated dropout of 10%, based upon prior experience with insulin pump studies, will provide exploratory data for the assessment of differences between device iterations.

All outcomes for each stage are presented as mean (SD) or median (interquartile range [IQR]). Outcomes were compared between stage 3 and all of the other stages (run-in, stage 1, and stage 2). No other between-stage comparisons were made. Comparisons were made using linear mixed-effects linear regression with unstructured covariance. Participants were included as random effect, and stage was included as fixed effect. Results are expressed as mean change with 95% CI. In some cases, the outcome required a transformation using natural logarithm (line, sensor, and insulin set change time), and results are expressed as mean fold change with 95% CI. Because of poor model fit, some outcomes were compared using Wilcoxon sign rank test, and results are expressed as median change with 95% CI. A P value of <0.05 was considered statistically significant. All analyses were performed using Stata 17.

All primary and secondary outcome results were reported with no adjustment for multiple comparisons.

The Index of Relative Socio-economic Advantage and Disadvantage was used to determine socioeconomic status derived from post codes. A low score indicates relatively greater disadvantage and a lack of advantage in general. A high score indicates a relative lack of disadvantage and greater advantage in general (27).

Twenty-two adults (11 women) with type 1 diabetes, all of whom were non-Hispanic White in ethnicity, mean (SD) age of 49 years (13) and diabetes duration of 29 years with a baseline HbA_1c of 6.6 (0.4)%/50 (4) mmol/mol, BMI of 29.2 (6.2) kg/m², and a median (IQR) Index of Relative Socio-economic Advantage and Disadvantage core of 8 (6,9), were enrolled, of which 21 participants completed the study. Two of the 22 participants had an HbA_1c of >7.0% (53 mmol/mol). All were established on insulin pump therapy (12 [5] years) and experienced in CGM use (7 [4] years). Six (29%) had impaired hypoglycemia awareness. Twenty participants were using faster insulin aspart and two used aspart to manage their glucose levels (Supplementary Table 2). Nineteen had phones compatible with 780G Bluetooth function. The study was conducted between December 2021 and January 2023.

DTQ-current scores for e-AHCL (stage 3) were significantly more favorable when compared with run-in (non-Bluetooth 780G/GS3/3-day insulin infusion set) and stage 1 (780G/GS3/3-day insulin infusion set) but did not differ significantly from stage 2 (780G/GS4/3-day insulin infusion set) (Table 1). Fig. 1 provides an overview of participant responses to DTQ-current clustered according to question domains: “hypoglycemia behavior,” “device acceptance and satisfaction,” “diabetes distress and burden,” and “other.” There was a progressive improvement in treatment satisfaction with each iteration of the technology, and e-AHCL appears to have benefited all domains. A corresponding pattern of responses was observed with DTQ-change (Table 1). Details regarding responses to individual DTQ questions are provided in Supplementary Table 3.

Comparing e-AHCL with run-in and stage 1, there were significant improvements in UEQ (Table 1). Compared with the standardized UEQ questionnaire benchmark, the e-AHCL scored “excellent” in all categories except “novelty,” where it was deemed “good” (Fig. 2).

DMSRQ-SF scores indicated that e-AHCL was more convenient, and that efficacy was greater compared with all prior AID iterations. There was also less interference with their daily lives and less self-monitoring of glucose burden compared with run-in and stage 1. Social burden was lower compared with run-in. There were no differences in perceptions regarding negative events, psychological well-being, treatment satisfaction, or treatment preference (Table 1).

Scores for e-AHCL were better for all SUS metrics compared with run-in but did not differ for stage 1 and stage 2 (Table 1).

e-AHCL compared with stage 1 was significantly better regarding sleep quality, with other comparisons for sleep of borderline significance. There were no differences in diabetes distress as assessed by Problem Areas in Diabetes comparing e-AHCL with any of the prior iterations. Fear of hypoglycemia improved using e-AHCL compared with run-in and stage 1 but did not differ from stage 2. These differences were predominantly attributable to changes in the worry subscale. Gold scores from run-in (2.8 [1.5]) to study end 2.6 [1.6]) did not change (P = 0.33).

For e-AHCL, %TIR was 84.9 (4.4)%. Differences in %TIR favored e-AHCL compared with run-in and stage 2. Percent TBR (<70 mg/dL) for e-AHCL was 1.5 (1.0)%, time in clinical low range (<54 mg/dL) was 0.1 (0.0, 0.2)%, and coefficient of variation was 29.8 (2.5)%, all of which were better than run-in, stage 1, and stage 2. Minor differences in mean glucose were observed, favoring e-AHCL over the system used in stage 2. There was a minor reduction in HbA_1c, favoring e-AHCL in comparison with stage 2 (Table 2). However, the proportion of participants with an HbA_1c of <7.0% (53 mmol/mol) did not differ according to study stage, and there were no other differences of significance.

There were no episodes of DKA or severe hypoglycemia. Thirteen participants (two during run-in, two during stage 1, five during stage 2, and four during stage 3) contracted coronavirus disease 2019 (COVID-19) during the study. None required hospitalization.

The percentage time of CGM wear was >99% for all stages. The percent of CGM time in closed loop was 100% for all stages. The number of alarms were less, and fewer finger-stick glucose readings were performed for e-AHCL than for all prior stages (Table 2).

The number of insulin set changes with e-AHCL did not differ significantly compared with prior stages and run-in. The time spent per set change was lower for e-AHCL compared with all prior stages and run-in. The total time spent on set changes was lower for e-AHCL compared with stage 2 and run-in. The number of sensor changes did not differ for e-AHCL compared with prior stages and run-in, although the total time and time spent per sensor change with e-AHCL was less than in each of the prior stages and run-in. The number of device uploads performed manually was significantly less with e-AHCL compared with run-in. The transition from manual device upload in run-in to automated device upload with e-AHCL saved 40 min over 12 weeks (Supplementary Table 4).

Our study followed, for a year, the user experience of a cohort of adults living with type 1 diabetes who received in-series AID systems with incremental refinements culminating in an investigational e-AHCL. Refinements included improvements to the device menu for a more user-friendly interface, permitting simplified access to key actions, and an additional glucose target (stages 1, 2, and 3), a calibration-free CGM (stages 2 and 3), and modifications to the closed-loop algorithm, a 7-day insulin infusion set, and a less intrusive all-in-one CGM platform (stage 3). These refinements incorporated as part of e-AHCL resulted in a requirement for fewer finger-stick capillary glucose readings, fewer alarms, less time spent on uploading device data to the cloud, and less time spent on insulin set changes and on glucose sensor changes. A panel of tools assessing self-reported outcomes favored e-AHCL compared with prior iterations of the AID system. These included an improved user experience and device satisfaction, reductions in burden and discomfort associated with the use of technology to manage their glucose levels, and a reduction in fear of hypoglycemia. A general assessment of e-AHCL usability by SUS deemed it excellent. These nonglycemic benefits were achieved without sacrifices in glucose levels. Indeed, small but important benefits were observed in reductions in %TBR despite excellent participant glucose levels at baseline as reflected by both HbA_1c and CGM metrics.

AID systems are now part of mainstream therapy for diabetes. They have resulted in major improvements in glucose control, with an increasing body of real-world evidence indicating that consensus glucose targets are being met by a substantial proportion of AID users with average TIRs of >70% with simultaneous reductions in hypoglycemia (9,13). The initial emphasis on glucose-related AID outcomes, and the drive to increase time in euglycemia (time in tight range) is understandable. However, given that the effectiveness of these devices in managing glucose levels has now been firmly established, greater emphasis in the further development of AID systems will be on optimizing the interface with the user.

The Medtronic 670G was the first commercial AID system to market. Data from the first randomized controlled trial revealed that, while participants allocated HCL felt more positive about their diabetes and experienced a greater sense of satisfaction from managing the condition compared with those who continued with multiple daily injections or insulin pump using finger-stick glucose monitoring, there were no objective increases in diabetes treatment satisfaction or reductions in diabetes distress. This may be explained by the counterbalancing of glycemic benefits with the burden associated with using this early system (2). These findings were consistent with other earlier reports of qualitative outcomes with hybrid AID systems (28).

In addition, while real-world reports corroborated trial evidence supporting significant improvements in glucose control, these reports also indicated that there were patients attending clinics who did not attain the %TIR and %TBR improvement reported in clinical trials evaluating 670G (2,3), with significant discontinuation rates observed at 1 year (4–6). Potential limitations contributing to user discontinuation of the first-generation device were an excessive number of alerts, inconvenient requests for finger-stick readings and numerous closed-loop exits by the system, sensor performance that did not meet the users’ expectations, and a fixed AID glucose target of 120 mg/dL. Therefore, based upon clinical experience and user feedback, refinements to the first-to-market Medtronic system were implemented that included an improved menu, a selection of glucose targets for the user, an automated correction bolus, and software changes permitting greater persistence in closed loop with fewer alarms, and a reduced number of requests for finger-stick glucose measurements, resulting in the 780G system that is currently available commercially (29,30). Further improvements to 780G have resulted in the e-AHCL system.

The major increment in use satisfaction as assessed by the primary study end point, the change in DTQ-current, occurred with the transition from stage 1 to stage 2. The increment was maintained with e-AHCL. The e-AHCL sensor had a simplified insertion process compared with those used in earlier stages. More importantly, while GS3, used in run-in and stage 1, required at least twice daily finger-stick calibrations, GS4 and the e-AHCL sensors were calibration free. These improvements were reflected in the total number of capillary finger-stick glucose readings performed, which fell dramatically from 338 and 299 over the 84 days of run-in and stage 1, respectively, to 47 and 18 during stages 2 and 3, respectively. This substantial reduction in the need for finger-stick glucose measurements would have had a significant beneficial impact upon the user experience. It would also have favorable health economic implications. Unlike the AID systems used in the antecedent stages, the e-AHCL system incorporated an extended-wear infusion set with a life of up to 7 days. The infusion set change frequency (21 vs. 25) was not significantly different for the 7-day insulin infusion set when compared with the commercially available 3-day set. While we did not collect detailed data as to why infusion sets were changed, possible causes include the loss of site viability caused by line occlusions due to sensitization and an inflammatory response with increased duration of wear, the increasing difficulty in keeping a set affixed to the skin over an extended period (31), and a habit of simultaneously changing the reservoir and the infusion sets.

The person living with type 1 diabetes experiences multiple health care–related demands upon their time, so any saving is likely of value. The time spent operating a device may be expected to contribute significantly to their burden of care. Time savings with e-AHCL may have contributed to the positive user experience with the device. The total time savings associated with device upload, sensor changes, and insulin set changes amounting to over 6.5 h per year, on average, with e-AHCL would be expected to significantly reduce the burden of care imposed by the device on users.

We acknowledge several limitations to this study, the most significant being that it was not randomized, and the results of stage 3 may have been impacted by the prior stages. There may have been a learning effect, although all participants were established insulin pump users and experienced in CGM for an average for 12 and 7 years, respectively, and each stage was of 3 months duration to minimize the confounding effect of the previous stage. Nor was there a power calculation performed, because of a lack of prior data pertaining to person-related outcomes and AID system comparisons. Therefore, this was a pilot study. Tools assessing the user experience were restricted to questionnaires. Qualitative data with semistructured interviews may have also helped with further delineating benefits and the overall thoughts of participants, and this will be a consideration in future studies.

We also recognize that the study cohort had excellent glucose levels and that this may limit the translation of our findings to people living with type 1 diabetes with suboptimal glycemia. However, AID systems have transformed glucose management, and an HbA_1c of <7.0% (53 mmol/mol) in the absence of significant hypoglycemia is by no means unusual for those using these technologies (9,11). All participants were of non-Hispanic White ethnicity, reflecting the type I diabetes demographic who use insulin pumps in Australia, and we suggest caution in extending these findings to other ethnic groups.

The study was conducted during the COVID-19 pandemic, and it is possible that infection with the virus may have impacted glucose control and user perceptions. However, there were a greater number of COVID-19 cases in stages 2 and 3 compared with run-in and stage 1, and, if there was a bias associated with infection that increased negative perceptions, it would be expected to have a greater impact on the later stages. Finally, while CGM metrics with e-AHCL compared favorably with earlier iterations, we would suggest caution, as the differing sensor configurations, while accurate in themselves, may not have provided identical data.

Study strengths include the emphasis on person-reported outcomes using multiple validated tools. Novel aspects of the study include the assessment of these person-reported outcomes in people with type 1 diabetes through iterative improvements of an AID device and the collection of information in real time regarding the burden of time in operating the device.

To conclude, while AID systems have transformed the glucose management of people with type 1 diabetes, we have not replaced like for like, and there remains a burden of care. Our findings show that incremental changes to these systems, including the automated upload of device data to the cloud, fewer alarms, removal of a requirement to calibrate the sensor with finger-stick measurements, improvements in the glucose sensor form factor, a more responsive algorithm, and an extended-wear insulin set, have had a positive impact on the user experience. Ongoing research into the interface of technology with the person living with diabetes is important, and the insights provided will benefit those with this, at present, lifelong and incurable condition who choose to use AID devices.

Acknowledgments. The authors thank the trial participants for the time devoted to this study.

Funding. The study was supported by a grant and materials from Medtronic Diabetes.

Duality of Interest. Y.W.K. has received honoraria from Insulet for speaker fees to present data on this study. S.T. reports nonfinancial support from Abbott Diabetes Care. A.R., B.G., and N.K. are employees of Medtronic. D.N.O. has received honoraria from Medtronic, Insulet, Abbott Diabetes Care, Novo Nordisk, and Sanofi, and research support from Medtronic, Insulet, Dexcom, Roche, GlySens, BioCapillary, and Endogenex, and is on advisory boards for Medtronic, Insulet, Abbott Diabetes Care, Ypsomed, Novo Nordisk, and Sanofi. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. D.N.O. led the study and wrote the first draft of the manuscript. C.Y.Y., Y.W.K., B.G., N.K., A.R., S.T., and D.N.O. co-designed the study. C.Y.Y., Y.W.K., T.A., H.J., E.N., L.R., and D.N.O. screened and consented participants and implemented the protocol. S.T., S.V., and D.N.O. analyzed study data. All co-authors critically reviewed the manuscript and approved its submission for publication. D.N.O. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.