Automated Insulin Delivery in Elderly with Type 1 Diabetes: A Prespecified Analysis of the Extension Phase

Introduction

Type 1 diabetes (T1D) is characterized by severe insulin deficiency and is currently best managed in most people with a hybrid closed loop (HCL) insulin pump incorporating continuous glucose monitoring (CGM). ¹ Management of T1D in older adults poses several challenges, including increased consequences of hypoglycemia, concern of dexterity limiting the use of devices, cognitive impairment and decreased executive functioning making disease self-management difficult. The Automated Insulin Delivery in Elderly with T1D (AIDE T1D) crossover randomized controlled trial (RCT) comparing HCL, predictive low glucose suspend (PLGS), and sensor-augmented pump (SAP) therapy using a Tandem t:slim X2 insulin pump, with Control-IQ (HCL), Basal-IQ (PLGS) (Tandem Diabetes Care, San Diego, CA), and Dexcom G6 (Dexcom, San Diego, CA) demonstrated that both HCL and PLGS significantly decreased hypoglycemia compared with SAP, with HCL also significantly decreasing hyperglycemia. ² Following completion of the trial, participants were invited to join a 3-month extension phase in which they were given the opportunity to choose between HCL, PLGS, and SAP.

Methods

The AIDE protocol (NCT04016662) was approved by the institutional review board (IRB) of the JAEB Center for Health Research. Principal eligibility criteria for this trial included age ≥65 years, clinical diagnosis of T1D >1 year, HbA1c <10.0%, and no use of an automated insulin delivery system within 1 month prior to enrollment. Severe cognitive impairment was an exclusion criterion. Additional eligibility criteria have previously been published (reference). The study was conducted at four academic endocrinology practices in the United States and included the 3-period RCT followed by a 12-week extension phase in which participants chose their preferred method of insulin delivery (HCL, PLGS, SAP, or CGM with injections).

At the end of the 12-week-extension period, a study visit was completed to collect vital signs, insulin pump and CGM data uploads, cognitive assessments, patient-reported outcomes (PROs), and a blood sample to measure HbA1c at a central laboratory. PROs included the Gold Survey score, Hypoglycemia Fear Survey, Hypoglycemia Confidence Scale, and Diabetes Distress scale.

Data are presented as mean ± SD unless otherwise noted. ²

Results

Seventy-five of the 78 participants who completed the RCT opted to continue in the extension phase. Among these 75 participants, age was 71 ± 4 years (range 65–86 years), age at T1D diagnosis was 29 ± 17 years (range 4–66 years) and duration of T1D was 42 ± 17 years (range 1–68 years). Almost all of the participants were non-Hispanic white (97%). A total of 85% had cardiovascular disease at study start, 27% had renal disease, and 49% had rheumatological disease (Supplementary Table S1).

Three participants opted not to participate in the extension phase due to the following reasons: two started study-prohibited antihyperglycemic medications, and one preferred to not use the study pump.

Of the 75 patients who initiated the extension phase, HCL was selected for use in the extension phase by 68 (91%), PLGS, by 6 (8%), and CGM only by 1 (1%). The six participants who chose to use PLGS reported their rationale of having better glycemic control and more individual control than afforded with HCL. Four of the 75 patients used a system other than the one they initially selected at some point during the extension phase. These participants were excluded from the analysis of glycemic outcomes and PROs.

Among the 65 participants who exclusively used HCL during the extension phase, median percentage of CGM use in the extension phase was 98% (IQR 97, 99%) and HCL use was 96% (IQR 94, 97%). Participants on HCL maintained glycemic control during the extension phase with a high time-in-range and low time <70 mg/dL (Table 1, Fig. 1, Supplementary Fig. S1). The proportion of HCL participants meeting glycemic targets is provided in Supplementary Table S2.

OPEN IN VIEWER

FIG. 1.

Time below range and time-in-range among participants who used HCL during the extension phase (N = 65). Panel A shows boxplots of percent time-below-range (<70 mg/dL and <54 mg/dL) and Panel B shows boxplots of percent time-in-range (70–180 mg/dL and 70–140 mg/dL). In each box, the black dot represents the mean, the horizontal line inside each box represents the median, and the bottom and top of each box represent the 25th and 75th percentiles, respectively.

The 5 participants completing the extension phase with PLGS also maintained glycemic outcomes with time <70 mg/dL of 1.85% ± 1.24% and time <54 mg/dL of 0.28% ± 0.23%. CGM-measured hypoglycemic events were 0.5 ± 0.5 per week. Mean glucose was 151 ± 12 mg/dL and time-in-range 70–180 mg/dL 74% ± 7%.

For the 65 patients completing the extension phase with HCL, PROs at the end of the extension phase were distributed similar to baseline: Gold Score Survey changed from 2.8 ± 1.7 to 2.5 ± 1.5, Hypoglycemia Fear Survey from 1.1 ± 0.6 to 0.9 ± 0.5, Hypoglycemia confidence scale from 3.3 ± 0.6 to 3.5 ± 0.5, and Diabetes Distress Scale from 1.7 ± 0.6 to 1.5 ± 0.5. For the 5 patients completing the extension phase with PLGS, PROs at the end of the extension phase were also distributed similar to baseline: Gold Score Survey changed from 3.6 ± 1.8 to 3.8 ± 2.4, Hypoglycemia Fear Survey from 0.9 ± 0.6 to 1.2 ± 1.0, Hypoglycemia confidence scale from 3.5 ± 0.5 to 3.4 ± 0.9, and Diabetes Distress Scale from 1.8 ± 0.7 to 1.5 ± 0.6.

One severe hypoglycemic event with loss of consciousness occurred during the extension phase in an HCL user. The occurrence of the severe hypoglycemic event was anticipated by the patient prior to her losing consciousness for about 4 min. CGM at the time of the loss of consciousness was about 60 mg/dL. Subsequently, she spontaneously regained consciousness and was provided caloric fluid. Emergency medical services (EMS) had been called by the family when the patient lost consciousness and came to the home to provide assistance. CGM signal indicated glucose in the 80s by the time EMS arrived. No further help was needed. No events of diabetic ketoacidosis were reported during the extension period.

Discussion

In this prespecified 3-month extension to the AIDE trial in older adults with T1D, the vast majority (91%) of participants chose to continue using the HCL system. Notably, only 8% of participants who participated in the extension chose the PLGS system. During the extension phase, glycemic control was similar to what had been achieved in the main RCT HCL phase and was also maintained in those who chose PLGS. Collectively, these results demonstrate that older adults with T1D can successfully use advanced automated insulin delivery to durably improve glycemic control and, when given a choice, strongly prefer HCL over other approaches.

There has been concern that older adults with T1D could find HCL insulin delivery overwhelming and therefore might prefer simpler approaches. Our results dispel this notion, at least in the patient population represented in this trial. Although we observed that participants needed somewhat more time to adopt safe and effective CGM and insulin pump and HCL use during the RCT, once they were familiar with the technology they strongly preferred the HCL system during the extension phase. This may be because they appreciated the better control and usability of the technology once they had developed confidence in it. During the extension phase, participants were provided with support comparable to what they might receive in clinical practice. In spite of this, they were able to effectively use the technology and maintain improved glycemic control. This suggests that our results are generalizable to clinical practice.

Current clinical practice guidelines suggest applying less stringent glycemic targets, including higher HbA1c levels, in older adults with T1D to minimize risk for hypoglycemia. ³ Our results demonstrate that HCL can enable older adults to achieve standard targets for glycemic control while reducing risk for hypoglycemia. To our knowledge, in the AIDE trial, HbA1c improved to the lowest that has been achieved safely in older adults with T1D. The current study extends the results of smaller prior studies and demonstrates the effectiveness and safety for a longer period of time in this vulnerable population. ^4,5 We are encouraged by the fact that we achieved excellent glycemic control safely in older adults with T1D with the current HCL system.

Strengths of our study include the largest sample size of any trial to date of older adults with T1D. Moreover, the intensity of our support and monitoring was similar to clinical practice. Limitations include relatively short duration of observation. This may have limited our ability to detect any chances in cognitive function or PROs. Although our patients had comorbidities, in general, they were in reasonable health, without major cognitive impairment and independently functioning. Whether similar glycemic control and a lower risk of hypoglycemia can be achieved in a more impaired population of older adults with T1D remains to be determined. Studies of AID need to be undertaken in older adults with T1D and more advanced comorbidities, physical impairments and cognitive dysfunction.

In conclusion, we conducted a 3-month extension phase to the AIDE RCT, which demonstrated that older adults with T1D strongly preferred an HCL insulin delivery system when given a choice. Moreover, they were able to use the system to maintain improved glycemic control with minimal support after initial education and use of the system in the RCT phase. These results demonstrate that advanced automated insulin delivery is an effective and safe approach for the management of T1D in older adults and should be among the options offered to this group that has often been excluded from newer technology-based solutions.