Beneficial Effects of Control-IQ Automated Insulin Delivery in Basal-Bolus and Basal-Only Insulin Users With Type 2 Diabetes Free

The study was conducted at three diabetes centers in the United States. The protocol was approved by a central institutional review board, and informed consent was obtained from each participant. The FDA approved an investigational device exemption for the conduct of the trial. The trial is registered at ClinicalTrials.gov (NCT05111301), and a full list of study site personnel is provided (Supplementary Material).

To be eligible for the trial, potential participants had to have type 2 diabetes treated with either basal-bolus insulin therapy (MDI or via an insulin pump) or basal-only insulin (without bolus insulin) for at least 3 months, have a total daily insulin dose (TDD) ≤200 units/day, and have an A1C of 7.5–12.0% (58–108 mmol/mol). Glucose-lowering medications in addition to insulin were permitted provided they were started >3 months before enrollment and their dose was stable. A complete list of inclusion and exclusion criteria is provided in Supplementary Table S1. A list of noninsulin glucose- lowering medications and dosages for each participant is provided in Supplementary Table S2.

Eligible participants initiated use of an unblinded continuous glucose monitoring (CGM) system with their existing insulin therapy and adjunctive therapies for 2–4 weeks. Participants using a Dexcom G6 CGM system with ≥85% of possible glucose values captured during the 14 days before enrollment could skip the CGM run-in period. A successful CGM run-in (≥85% use) was followed by open-loop use of the study pump (i.e., without Control-IQ AID functionality operational) and CGM for 2–4 weeks. Insulin therapy adjustments were made to optimize open-loop glycemic control.

Participants who successfully completed the open-loop pump run-in (defined as using the pump for the specified period with infusion set changes, meal bolus announcements, and no safety concerns) were trained on the use of the study pump in closed-loop mode (i.e., with Control-IQ functionality activated). Instructions were provided for management of hypoglycemia, prolonged hyperglycemia, and sick days/illnesses. Participants were asked to use the system’s programmed sleep mode each night and asked to perform an exercise session with the exercise mode enabled for ≥30 minutes at least once per week. In addition to the study CGM system, AID pump, and related supplies, participants were provided with a blood glucose meter (Contour Next One, Ascensia Diabetes Care, Parsippany, NJ) and a blood ketone meter (Precision Xtra, Abbott Diabetes Care, Alameda, CA). After training, the 6-week period of Control-IQ use was initiated.

Phone contacts or clinic visits occurred 3 days (phone), 1 week (phone), 2 weeks (clinic), 3 weeks (phone), and 6 weeks (clinic) after initiation of Control-IQ. The occurrence of device issues or adverse events were solicited at each visit and contact. Weight and height were measured at baseline, with a follow-up weight measured at 6 weeks. A1C and random C-peptide (Cobas e 801; Roche Diagnostics, Basel, Switzerland) were measured at baseline at a central laboratory. Patient- reported outcome (PRO) questionnaires were completed at screening, at the start of Control-IQ use, and at 6 weeks. Details about testing at each visit are provided in Supplementary Table S3.

Safety outcomes included severe hypoglycemia (defined as hypoglycemia requiring assistance because of an altered cognitive state), diabetic ketoacidosis (DKA; defined by the criteria established in the Diabetes Control and Complications Trial) (8), hyperosmolar hyperglycemia syndrome (HHS), and other adverse events. CGM outcomes measured over 6 weeks included time in range (TIR; 70–180 mg/dL), time >180 mg/dL, mean glucose, time <54 mg/dL, and additional hyperglycemia and hypoglycemia metrics. TDD and body weight were also assessed. PRO measures included the DAWN Impact of Diabetes Profile (DIDP), the Diabetes Impact and Device Satisfaction (DIDS) measure, the PROMIS Sleep-Related Impairment Questionnaire, and the System Usability Scale. A detailed list of all questionnaires administered in this study is provided in Supplementary Table S4.

Analyses were conducted overall and separately for basal-bolus insulin users, basal-only insulin users, and users of sodium–glucose cotransporter 2 (SGLT2) inhibitor or glucagon-like peptide 1 (GLP-1) receptor agonist medications. Baseline CGM metrics were calculated from the last 7 days of the CGM run-in period and 14 days of the open-loop insulin pump period. Follow-up metrics were calculated over the 6 weeks of Control-IQ use. Analyses were conducted overall, for daytime (6:00 a.m. to 11:59 p.m.) and nighttime (midnight to 5:59 a.m.), and for the periods during and after each at-home exercise session.

Summary statistics appropriate to the distribution (mean ± SD or median [interquartile range (IQR)]) were calculated for each CGM metric at baseline, during Control-IQ use, and for change. CGM metrics during Control-IQ use were compared with baseline values (obtained during the unblinded CGM run-in period) using paired t tests when the distribution was normally distributed and a robust regression when the distribution was skewed. The false discovery rate was used to adjust for multiple comparisons (9). Analyses were conducted using SAS, v. 9.4, statistical software (SAS Institute, Cary, NC).

Of the 42 adults who were enrolled into screening, 10 did not meet eligibility criteria and 2 were withdrawn by the site because of nonadherence during the pump run-in phase. Among the 30 who initiated use of Control-IQ, the mean age was 54 ± 12 years, mean A1C was 8.6 ± 1.2% (70 ± 13.1 mmol/mol), and mean BMI was 31 ± 6 kg/m². Prior insulin delivery was with an MDI regimen in 15 participants, an insulin pump in 2, and basal insulin only in 13. Metformin was being used by 21 (70%), and 22 (73%) were using an SGLT2 inhibitor or GLP-1 receptor agonist (Table 1).

All 30 participants completed the 6-week trial, with 100% of scheduled visits and contacts completed. There were four unscheduled visits and 84 unscheduled contacts among the 30 participants during the 6 weeks of Control-IQ use. Common reasons for these contacts included exercise session logistics (28%), reminders about device time changes for Daylight Saving Time (18%), diabetes management issues (18%), study supplies logistics (13%), device training (9%), and protocol/procedures training (6%).

Control-IQ was active a median of 96% (IQR 91–97%, range 57–99%) of possible time during the 6 weeks, with 23 participants (77%) in closed-loop functionality ≥90% of the time and 27 (90%) in closed-loop functionality ≥80% of the time. The three participants with <80% of time in closed-loop functionality had Control-IQ active for 78%, 71% (reported running out of study CGM supplies), and 57% (reported running out of study insulin) of the 6-week period, respectively.

Mean TIR increased from 56% at baseline to 71% with Control-IQ (mean change 15%, 95% CI 5–24%, P = 0.007), representing an increase of 3.6 hours/day (Table 2 and Figure 1A and B). The beneficial effect was evident over the entire 24 hours of the day (Figure 1C and Supplementary Table S5). Improvement in TIR occurred rapidly, starting with the first day of Control-IQ use, and mean TIR was stable thereafter (Figure 2). The higher the baseline A1C was, the greater the treatment effect was (P = 0.05) (Supplementary Figure S1). In the 17 participants with a baseline A1C ≥8.0% (64 mmol/mol), mean TIR increased by 19 ± 32% compared with 9 ± 13% among the 13 participants with a baseline A1C <8.0% (64 mmol/mol) (Supplementary Figure S2).

Mean time >180 mg/dL decreased by 15% (95% CI −24 to −5%, P = 0.007), median time >250 mg/dL decreased by 3.7% (95% CI −7.0 to −0.3%, P = 0.03), and mean glucose decreased by 22 mg/dL (95% CI −37 to −6 mg/dL, P = 0.01). Median time <54 mg/dL was 0.00% at baseline and 0.02% during Control-IQ use (mean change = 0.00%, 95% CI = −0.02 to 0.02%).

Both prior MDI regimen/insulin pump users and prior basal-only insulin users showed improvement in CGM metrics (Table 2 and Supplementary Figure S3). Mean TIR increased by 17% in MDI/pump users and by 13% in basal-only insulin users. Results were similar in the 22 participants using an SGLT2 inhibitor or GLP-1 receptor agonist in addition to insulin, for whom TIR increased by 16% and median percentage of time <54 mg/dL changed by 0.00% with Control-IQ (Supplementary Table S6).

There were 252 at-home exercise sessions recorded by 28 of the 30 participants. During exercise and for 2 hours after exercise sessions, mean TIR was 76 ± 32% and median time <70 mg/dL was 0.0% (IQR 0.0–0.0%). From the end of exercise until 6:00 a.m. the next morning, mean TIR was 73 ± 24% and median time <70 mg/dL was 0.0% (IQR 0.0–0.0%).

Mean insulin TDD was 0.62 ± 0.40 units/kg/day at baseline and 0.67 ± 0.36 units/kg/day during Control-IQ use. In prior MDI/pump users, insulin TDD was 0.76 ± 0.41 units/kg/day at baseline and 0.81 ± 0.40 units/kg/day using Control-IQ, whereas in prior basal-only insulin users, insulin TDD was 0.44 ± 0.31 and 0.47 ± 0.17 units/kg/day, respectively. Among the average of 342 boluses delivered per participant by previous MDI users, 52% were autoboluses with the remainder manual boluses, whereas among the average of 327 boluses delivered per participant by prior basal-only insulin users, 53% were autoboluses.

Body weight increased from a median of 81.9 kg (IQR 72.2–107.2 kg) at baseline to 83.2 kg (IQR 74.4–106.5 kg) at 6 weeks (P = 0.003), with the increase observed only in the basal-only insulin users (from a median 79.7 kg at baseline to 81.2 kg at 6 weeks) and not in the MDI/pump users (median 88.5 kg at baseline and 87.1 kg at 6 weeks).

On the PRO surveys, the DIDS satisfaction score increased from 6.9 ± 1.6 to 8.6 ± 1.7 from screening to 6 weeks (P = 0.006), indicating a significant improvement in satisfaction from prior therapy. There were no changes on the DIDS diabetes impact score or the DIDP score. The PROMIS sleep t-score changed from 55.6 ± 7.3 at screening to 52.7 ± 6.4 at 6 weeks (P = 0.19) (Supplementary Table S7). Participants rated the system highly on the System Usability Scale (81.8 of 100), indicating “excellent” usability (10).

There were no episodes of severe hypoglycemia, DKA, HHS, or other serious adverse events.

In this study of adults with type 2 diabetes using basal-bolus or basal-only insulin therapy, use of Control-IQ was safe and associated with a substantial, statistically significant decrease in hyperglycemia resulting in an increase in TIR of 3.6 hours/day (25.2 hours/week). PRO survey results indicated a high degree of satisfaction with Control-IQ compared with prior therapy, an important finding because patient-reported satisfaction is a leading indicator of adherence to treatment interventions (11). The amount of CGM-measured hypoglycemia was low at baseline and remained low during use of Control-IQ. No serious adverse events occurred, and no safety concerns were observed. Although some participants, particularly those not using prandial insulin at the time of enrollment, had weight gain with Control-IQ use, such a finding is not uncommon with glycemic management intensification (12), especially when A1C levels are high (mean A1C was 8.6% in our study cohort), which is indicative of insufficient insulin TDD at baseline. Formal guidance regarding food choices focusing on weight management was not part of this protocol.

The mean TIR during Control-IQ use of 71% is similar to what has been reported in adults with type 1 diabetes (4). Additionally, the median percentage of time participants used the pump with closed-loop functionality was 96%, which is at least as high as what has been observed with type 1 diabetes (4).

The findings in this study are also similar to what has been reported for real-world off-label use of Control-IQ for individuals with type 2 diabetes from the Tandem t:connect Web application. In one analysis, TIR was 76% after 180 days using Control-IQ in 134 prior pump users and 74% in 173 prior MDI users (13). In a second analysis of Control-IQ use for up to 1 year, 378 individuals with type 2 diabetes who switched from Basal-IQ (predictive low-glucose suspension functionality in a t:slim X2 pump) to Control-IQ showed an improvement in TIR from 69% at baseline to 78% after 1 year (14). In an analysis of Medicare and Medicaid recipients who were Control-IQ users, 500 individuals with type 2 diabetes improved from a TIR of 64% at baseline to 72% after using Control-IQ for at least 30 days (15).

Few clinical trials have evaluated an AID system in at-home studies involving people with type 2 diabetes. The fully closed-loop CamAPS HX system has been evaluated in two two-period crossover trials. In 26 adults with type 2 diabetes, Daly et al. (16) reported an increase in TIR from 32 to 66% using the CamAPS HX system for 8 weeks compared with standard insulin therapy without real-time CGM. In 26 adults with type 2 diabetes who were undergoing dialysis, Boughton et al. (17) reported an increase in TIR from 43 to 57% using the CamAPS HX system with fast- acting insulin aspart for 20 days compared with standard insulin therapy without real-time CGM.

Slightly more than half of boluses for prior users of either an MDI regimen or basal-only insulin were given as autoboluses, a unique feature of Control-IQ technology that allows for up to a 6-unit automatic bolus to be given by the system up to once per hour. These autoboluses occur in addition to automated changes in the AID basal rate. In a meta-analysis of randomized trials of Control-IQ users with type 1 diabetes who were 2–72 years of age, Beck et al. (18) showed that the percentage of daily boluses that were autoboluses was 26% for participants with an A1C <7.0% compared with 52% for those with an A1C ≥8.5%. The latter percentage is similar to what was observed in this trial of insulin users with type 2 diabetes. As in the type 1 diabetes analysis, in the current study, the higher the baseline A1C was, the greater the TIR improvement using Control-IQ was. This suggests that autobolusing as a feature of the Control-IQ algorithm has considerable impact in improving A1C levels in individuals with a high A1C before initiating Control-IQ. We speculate that, before the study, missed boluses with an MDI therapy regimen or failure to advance to adding bolus therapy when needed for basal-only insulin users, may explain the elevated A1C and low TIR levels at baseline. Even without changing behavior to initiate the majority of their boluses on their own, participants in the study were able to receive bolus insulin and significantly improve their glycemic outcomes.

The trial was intended to provide a preliminary assessment of Control-IQ use in type 2 diabetes and not definitive results. The short duration and single-arm design of this trial prevented us from assessing whether use of the closed-loop system would lead to persistent weight gain or its longer-term impacts. The sample size was too small to perform any analyses according to C-peptide level. The two participants withdrawn by their site during the pump run-in phase both exhibited a lack of responsiveness to required study contacts and failure to adhere to their diabetes-related medications. It was beyond the scope of this study to optimize the baseline treatment plans of enrolled participants, which may have been affected by medication costs, patient preferences, side effects, and insurance coverage.

The results of this trial demonstrate that, during 6 weeks of use, Control-IQ was safe, with no serious adverse events and no increase in CGM-measured hypoglycemia, with the majority of participants using concurrent GLP-1 receptor agonist and/or SGLT2 inhibitor therapy. There was a substantial improvement in TIR and mean glucose related to a reduction in hyperglycemia. Participant satisfaction improved with Control-IQ use, and the system was rated highly by most participants on the System Usability Scale. The results of this preliminary study support proceeding to conduct a pivotal trial of Control-IQ in adults with type 2 diabetes treated with insulin.

This study was funded by Tandem Diabetes Care.

C.J.L.’s institution has received research support from Abbott Diabetes Care, Dexcom, Insulet, and Tandem Diabetes Care, and she has been a paid consultant to Dexcom. D.R.’s institution has received funding and study supplies from Dexcom and Tandem Diabetes Care. K.P. has received research support from Altimmune, Dexcom, Eli Lilly, Medtronic, Novo Nordisk, Oramed, and Tandem Diabetes Care. T.B. has received research support from Abbott Diabetes Care, Dexcom, Eli Lilly, MannKind, Medtronic, Novo Nordisk, Tandem Diabetes Care, and Viacyte and speaker fees from AstraZeneca, Boehringer Ingelheim, Dexcom, Eli Lilly, and Novo Nordisk. C.M.L.’s institution has received research support from Abbott Diabetes Care, Dexcom, Insulet, and Tandem Diabetes Care. G.O. has received research support, including salary and product support, from Abbott Diabetes Care, Dexcom, Omnipod, and Tandem. J.L.’s, C.K.’s, and R.W.B.’s institutions have received funding and study supplies from Dexcom and Tandem Diabetes Care. No other potential conflicts of interest relevant to this article were reported.

C.J.L. researched and interpreted the data and wrote the manuscript. D.R. performed statistical analysis and contributed to writing and reviewing the manuscript. Y.C.K., K.P., T.B., L.C., D.D., C.M.L., G.O., C.R., J.L., C.K., and R.W.B. researched and interpreted the data, contributed to discussion, and reviewed/edited the manuscript. C.J.L. 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.