Sustained Effectiveness of an Advanced Hybrid Closed-Loop System in a Cohort of Children and Adolescents With Type 1 Diabetes: A 1-Year Real-World Study Free

Since their introduction, automated insulin delivery (AID) systems have increasingly been used in clinical practice. These devices are now recognized as the gold standard for the management of people with type 1 diabetes (T1D) across all age-groups (1). The Medtronic MiniMed 780G (Medtronic, Northridge, CA) was one of the earliest AID systems introduced in the European market. This device belongs to the second-generation of AID systems, also known as advanced hybrid closed loop (AHCL). The MiniMed 780G system received European Medicines Agency approval in June 2020 for use in individuals aged >7 years and is currently available in >40 countries worldwide. The core of the automatic function of MiniMed 780G, known as SmartGuard technology, relies on a proportional integrative derivative controller, enhanced by key elements of a fuzzy logic artificial pancreas algorithm (2). These features include the automatic delivery of correction boluses at 5-minute intervals and the flexibility for users to choose between three target glucose levels (100, 110, and 120 mg/dL). At the time of its commercialization, the MiniMed 780G was integrated with the Guardian Sensor 3 to work as an AID system. As of the end of 2021, the Guardian Sensor 3 has been gradually replaced by the more advanced Guardian 4 Sensor, which eliminated the need for glucose calibrations to keep the closed-loop system working properly.

The safety and effectiveness of the MiniMed 780G system have been reported in clinical trials and observational studies (3–8). However, there are still few real-world data on the 1-year performance of this device in children and adolescents living with diabetes.

This study investigated glycemic outcomes achieved by children and adolescents with T1D during their first 12 months of MiniMed 780G use. Additionally, the study aimed to identify any potential predictors of the achievement of therapeutic goals in our study cohort.

The study adopted a multicenter, longitudinal, observational real-world design and was conducted across 25 pediatric diabetes centers affiliated with the Italian Society for Pediatric Endocrinology and Diabetes (ISPED). The study focused on children and adolescents with T1D who started using the MiniMed 780G between June 2020 and June 2022. Inclusion criteria consisted of a diagnosis of T1D based on the latest International Society for Pediatric and Adolescent Diabetes Clinical Practice Consensus Guidelines (9), age between 7 and 18 years, children assent and informed consent from their parents to access continuous glucose monitor (CGM) data remotely. Exclusion criteria included partial clinical remission, defined as an insulin dose-adjusted glycated hemoglobin (HbA_1c) ≤9% (10), the presence of uncontrolled associated disorders, and the use of concomitant medications known to influence blood glucose levels. The study adhered to the ethical regulations outlined in the Declaration of Helsinki and received approval from the University of Messina Ethics Committee (no. 39-23). Written informed consent from at least one parent of each study participant involved in the research study was obtained before the start of study procedures.

Demographical and clinical data, including age, sex, diabetes duration, treatment type, and mean value of HbA_1c in the year prior to starting the AHCL system, were collected by reviewing medical records. Before starting AHCL therapy, all children and their caregivers received comprehensive training according to the standard clinical practice recommended by ISPED for the use of technology (11). CGM metrics from a 14-day run-in period (T0), wherein study participants used the device in manual mode, were collected. Over the 12-month study period, anthropometric parameters (height, weight, and BMI), insulin therapy details, HbA_1c measurements, and CGM metrics from the preceding 2-week period were gathered 15 days after the automatic mode activation (T1) and during each 3-month follow-up visit (T2, T3, T4, and T5). Participants who switched to alternative treatment strategies during the study period or missed follow-up appointments were excluded from the analysis of the entire study. Inconsistent use of the CGM system (mean daily use <70%) was also considered as an exclusion criterion from the analysis. Data from the AHCL system were obtained using CareLink Professional software. The following glycemic metrics were recorded: mean sensor glucose and its SD, percentage of time in range (TIR) between 70 and 180 mg/dL, percentage of time above range (TAR) >180 mg/dL, percentage of time between 180 and 250 mg/dL (TAR_{Level 1}), percentage of time >250 mg/dL (TAR_{Level 2}), percentage of time below range (TBR) 70 mg/dL, percentage of time between 54 and 70 mg/dL (TBR_{Level 1}), percentage of time <54 mg/dL (TBR_{Level 2}), percentage of time in tight range (TITR) between 70 and 140 mg/dL, glucose management indicator (GMI), and coefficient of variation (CV). Glycemia risk index (GRI) was calculated according to the following equation: (3.0 × TBR_{Level 2}) + (2.4 × TBR_{Level 1}) + (1.6 × TAR_{Level 2}) + (0.8 × TAR_{Level 1}) (12).

The analysis also considered the following AHCL settings: active insulin time (AIT), SmartGuard target, and number of carbohydrate (CHO)-to-insulin ratios per day. Information related to system engagement, including mean daily automatic mode use and sensor wear, number of exits from SmartGuard, total insulin daily dose and its distribution between basal and bolus delivery, amount of automatic correction boluses, number of user-initiated boluses per day, and carbohydrates (CHO) entered per day, were also captured.

Data were automatically transmitted to the cloud for participants with the “MiniMed Mobile” app. For others, glucose and insulin delivery data were manually downloaded during quarterly follow-up visits. To assess the effectiveness of AHCL systems according to HbA_1c levels, study participants were categorized into three groups: those with baseline HbA_1c <7% (<53 mmol/mol), those with HbA_1c between 7% (53 mmol/mol) and 8% (64 mmol/mol), and individuals with HbA_1c >8% (>64 mmol/mol). The relative change of TIR was calculated to assess the percentage of variation from baseline to the end of the study.

Numeric data are presented as mean, SD, median, and interquartile range, while categorical variables are expressed as absolute frequencies and percentages. These descriptive statistics were calculated for each observation time point (baseline T0, T1, T2, T3, T4, and T5). The nonparametric approach was used due to the nonnormal distribution of most numeric variables, as confirmed by the Kolmogorov-Smirnov test. To compare time points for all numeric variables, the Wilcoxon test was used. The McNemar test was applied to analyze changes in the variables “target SmartGuard” and “AIT” across different time points. The Bonferroni correction was applied, adjusting the α level for multiple two-by-two comparisons. To investigate potential differences among the four quartiles of TIR after 1 year of AHCL use, the Kruskal-Wallis test was applied for all numeric parameters. The Jonckheere-Terpstra test was used to assess any significant trends within these quartiles. The χ² test was used to compare the qualitative variables across the four TIR quartiles. Similarly, potential variations in glycemic metrics were evaluated among the four quartiles of automatic correction boluses. TIR changes throughout the study period in different subgroups according to baseline HbA_1c levels were evaluated by using the Wilcoxon test. A linear mixed model was applied to determine the association between TIR changes over time and variables, including daily insulin dose, number of automatic boluses, percentage of automatic mode use, SmartGuard exits, number of CHO-to-insulin ratios per day, number of user-initiated boluses, daily CHO intake, and AIT. Univariate and multivariate logistic regression models were applied to identify significant predictors for the simultaneous achievement of TIR >70%, GMI <7%, and TBR<4% at the 1-year follow-up. All statistical analyses were conducted using IBM SPSS Statistics 22 for Windows (IBM Corp., Armonk, NY), with statistical significance defined as a P value <0.05.

Of 430 individuals who were enrolled based on inclusion and exclusion criteria, 62 were subsequently excluded from the study analysis due to various reasons. Specifically, 28 individuals (45.2%) wore the glycemic sensor for an insufficient duration to obtain reliable glucose data, 19 participants (30.6%) had unavailable CGM data at one or more time points, and 15 individuals (24.2%) discontinued AHCL therapy during the study period. Among the latter group, 11 participants temporarily discontinued pump use during the summer months, while 4 individuals definitively suspended continuous subcutaneous insulin infusion (CSII) treatment, opting to revert to insulin multiple daily injections.

The final study cohort consisted of 368 children and adolescents with a slight predominance of boys (52.2%). The median age of participants was 12.9 years, and the median duration of diabetes was 5.2 years. In the year prior to the start the MiniMed 780G use, the median HbA_1c value was 7.3% (56.3 mmol/mol). Baseline HbA_1c levels ranged from 5.7% to 10.9%. Among the participants, 245 (66.6%) were already using an insulin pump at the time of enrollment, while 123 (33.4%) were previously on multiple daily injections. Regardless of their previous insulin regimen, 85.6% of children and adolescents were already using CGM before starting AHCL therapy. The median BMI of study participants was 20.1 kg/m² (+0.5 SD score [SDS]).

After 15 days of SmartGuard use, most glucose metrics significantly improved compared with T0. Particularly, TIR increased from 62.8% to 75.4% (P < 0.001), while TAR_{Level 1} and TAR_{Level 2} decreased from 24.7% to 17.7% and from 10% to 4.5%, respectively (P < 0.001 for both). GRI significantly decreased from 41.9 to 27.8 (P < 0.001). Mean sensor glucose also decreased from 164.3 mg/dL to 145.3 mg/dL (P < 0.001). TBR_{Level 1}, TBR_{Level 2}, and CV were similar to the previous observational period (P = 0.166, P = 0.602, and P = 0.330, respectively). Almost half of participants (47.6%) set the glucose target at 100 mg/dL. The percentage of individuals using AIT at 2 h was 40.4%.

Subsequent 3-month follow-up visits demonstrated that TIRs remained relatively stable compared with the first 15 days of AHCL use. Similarly, no substantial changes were detected in other glucose metrics, total daily insulin dose, distribution between basal and boluses, CHOs intake, and SmartGuard-related data. All the details about glycemic metrics, insulin data, and device settings for each observational time are presented in Table 1.

After 1 year of follow-up, TIR remained consistently higher than at baseline (P < 0.001). TAR_{Level 1}, TAR_{Level 2}, and mean sensor glucose and its SD showed significant reductions from baseline (P < 0.001 for all parameters). GRI significantly decreased from 41.9 to 27.6 (P < 0.001) (Fig. 1). Among other glycemic metrics, CV was 34.1 ± 5%, GMI was 6.8 ± 0.4%, and TITR was 51.1 ± 10.3%. The percentage of automatic corrections was significantly higher compared with the first 2 weeks of AHCL use (30.8% vs. 28.0%, P < 0.001). Time spent in automatic mode, the frequency of SmartGuard exits, the number of user-initiated boluses per day, and daily CHO intake remained consistent with the values observed at T1. The percentage of individuals setting the glucose target at 100 mg/dL and AIT at 2 h substantially increased to 57.4% and 55.1%, respectively. Notably, the mean HbA_1c value over the study duration was found to be significantly lower than the previous year (6.9 ± 0.6% vs. 7.4 ± 0.9%, P < 0.001). No differences were observed in BMI SDS between baseline and after 12 months of AHCL use (0.54 ± 1.1 vs. 0.57 ± 1.1, P = 0.162).

When considering participants categorized into four subgroups based on TIR interquartile ranges, we found that most of the other glucose metrics concomitantly improved with TIR. Interestingly, CV levels ranged from 36.7% in the first TIR quartile to 30.6% in the fourth TIR quartile. Higher TIR levels were associated with a reduced number of automatic correction boluses (P < 0.001), fewer SmartGuard exits (P = 0.021), and longer time spent in automatic mode (P = 0.030). The percentage of participants setting AIT at 2 h was significantly higher in children and adolescents with higher TIR levels. The Jonckheere–Terpstra test revealed an age-related increase in TIR levels (P = 0.015), while no differences in BMI were detected among the four TIR quartiles (Table 2). A strong association was also found between the percentage of automatic correction boluses and all glycemic metrics (Supplementary Table 1).

Although TIR improvement was found regardless HbA_1c levels prior to starting the AHCL system, the most relevant increase was observed in participants with higher baseline HbA_1c. Specifically, in this subgroup, TIR changed from 54.3% during the run-in period to 72.3% after 1 year of AHCL use, demonstrating a relative variation of 34.1%. In the other two subgroups, TIR showed relative increases of 21.1% in participants with baseline HbA_1c between 7 and 8%, and 13.4% in participants with baseline HbA_1c levels <7%.

There were no significant differences in glucose metrics between users of Guardian Sensor 3 and Guardian 4 Sensor, both after 15 days and 12 months of AHCL therapy, with the exception of lower TBR levels and CV observed in children and adolescents using Guardian 4 Sensor (Supplementary Table 2).

After 1 year of AHCL use, clinical targets in TIR, TBR, and GMI were achieved by 69.8%, 83.7%, and 68.2% of the study population, respectively. More than half of participants (56.6%) concomitantly met the aforementioned glycemic outcomes. Multivariate logistic regression analysis showed that older age (P = 0.014), a lower percentage of automatic correction boluses (P < 0.001), and a higher percentage of time spent in SmartGuard (P = 0.001) were significantly correlated with the simultaneous achievement of TIR, TBR, and GMI. No other demographic variables or factors dependent on users’ behavior were found to be statistically significant (Table 3). Linear mixed models revealed that TIR levels over time were associated with daily insulin dose (P = 0.008), the percentage of SmartGuard use (P < 0.001), the number of automatic correction boluses (P < 0.001), the frequency of exits from SmartGuard (P < 0.001), and daily CHO intake (P = 0.027) (Supplementary Table 3).

Our study showed that the AHCL system led to a prompt and sustained improvement of glycemic outcomes. Time spent in target glucose values fell within universally recommended ranges (13) from the first 2 weeks of AHCL use and remained stable during the entire study period. The mean HbA_1c value during the first year of AHCL use significantly decreased compared with the previous year. Additionally, the sustained enhancement of glucose control related to the use of AHCL system was also been supported by other emerging glycemic metrics such as TITR and GRI. Specifically, TITR levels were higher than the threshold of 50%, which is considered the target goal for this glucose metric (14,15). GRI is a composite metric aiming to assess the quality of glucose control by considering the risk for both hypoglycemia and hyperglycemia (12). In our analysis, GRI levels significantly decreased during the study period, falling within a range indicative of “low-risk” glycemic control. These results remark the effectiveness and safety of second-generation AID technology use in young people with T1D (16). To date, evidence regarding the efficacy of the MiniMed 780G over a 12-month mode in children and adolescents with T1D has been limited to few studies, conducted in relatively small cohorts (17–19). Our findings in terms of glycemic targets improvement are consistent with those studies.

Another encouraging finding from our analysis was that over half of participants simultaneously achieved targets recommended for TIR, GMI, and TBR at the end of the study. Predictors of this outcome included older age, longer time spent in automatic mode, and a lower number of automatic correction boluses. Other studies have reported a positive effect of age on glycemic targets, results that are consistent with our findings (20,21). In our study, TIR levels appeared to increase linearly with age increase, suggesting that AHCL performance may be influenced by different lifestyles between children and adolescents. Younger children often practice unplanned physical activities and may be more prone to consume unannounced extra meals, especially when they are not under parental supervision. Conversely, results obtained from older study participants are encouraging, particularly given the undeniable challenges of managing T1D during the tricky phase of adolescence. It is widely acknowledged that adolescents often face hindrances in diabetes self-management, resulting in suboptimal clinical outcomes (22).

Among pump-related factors, time spent in automatic mode emerged as a significant contributor to the achievement of favorable glycemic outcomes. Our analysis also revealed that individuals with the highest TIR levels had an average time with SmartGuard activated of 96.3%. The ability to spend more time in automatic mode is closely linked to the recent introduction of the Guardian 4 Sensor, which has gradually replaced the Guardian Sensor 3. This innovative CGM system brought some major updates, including the elimination of mandatory glucose calibrations. Studies evaluating clinical and psychological outcomes have already demonstrated the advantages of removing the hindrance of repeated finger-stick measurements both on glycemic outcomes and users’ burden (21,23–25). In our analysis, Guardian 4 Sensor users spent less time in hypoglycemia and achieved greater glycemic stability after 1 year of use.

The amount of automatic correction boluses has been identified as another predictor of the achievement of glycemic outcomes. The frequency of automated boluses offers intriguing insights for diabetes-care providers. These data can shed light on users’ behavioral patterns that hinder the achievement of glycemic outcomes, such as the failure to bolus before meals or the complete omission of meal boluses. The amount of automatic correction boluses in our study cohort was substantially higher than those reported by previous studies (3,4). Additionally, a significant increasing trend in the number of correction boluses was observed throughout the study period, suggesting a potential progressive decline in participants’ engagement to the proper AHCL system use. Some authors have proposed that the ideal percentage of automated boluses should be in the low- to mid-20% range (6). Our results corroborate this hypothesis, indicating that a percentage of automatic correction boluses <22% leads to more ambitious glycemic goals. However, a reciprocal cause-and-effect relationship between glycemic stability and the lack of need for repeated correction boluses cannot be ruled out.

Interestingly, the most relevant improvement in TIR was found in individuals who had suboptimal glycemic levels before using the AHCL system. This finding is consistent with results from other studies conducted in both adults and children with T1D (26–28).

Another noteworthy observation from our analysis was the stability of BMI throughout the study period. This result aligns with a recent Polish study, which similarly reported no significant changes in the BMI of children and adolescents during the first year of AHCL use (18). These consistent findings challenge the previously assumed association between CSII therapy and a progressive increase in BMI over time (29).

Limitations of our study include the absence of some factors that may interfere with the performance of the AHCL system, such as the pubertal stage, physical activity levels, and potential use of specific setting modes (e.g., temporary target). Additionally, the study involved Pediatric Diabetes Centers affiliated with ISPED, regardless of the number of children and adolescents with T1D followed, the type of institution (i.e., primary care, secondary care, tertiary care, or research faculty), and the expertise in technology applied to diabetes. Participants were not homogeneously distributed across sites. Owing to the multicenter study design, some differences in training methods for AHCL may exist. Specifically, CHO count training in smaller centers was performed by physicians rather than dietitians. Close follow-up visits after the application of the AHCL system occurred face-to-face in some centers and just virtually in others. However, all involved centers complied with standard clinical practices for the use of technology established by ISPED recommendations (11). Finally, our study cohort exclusively consisted of individuals living in Italy; thus, the presented data only reflect the implementation of this technology in one country and cannot be generalized.

Nonetheless, the strengths of our study lie in its moderate sample size, extended follow-up period, and comprehensive inclusion of both glucose metrics and clinical data.

Our study highlights the effectiveness of MiniMed 780G in rapidly and sustainably improving glycemic targets among youth with T1D. Our findings also suggest that even children and adolescents who may not fully adhere to their diabetes management can benefit from AHCL systems. The plateauing of glycemic outcomes observed beyond the initial weeks of use emphasizes the need for future investigations into potential factors hindering the achievement of more ambitious goals over time. Further research is also awaited to identify the optimal AHCL settings for younger individuals.

Duality of Interest. S.P. received speaking honoraria from Roche and Movi SpA. G.S. received speaking honoraria from Lilly. B.B. reports a grant from Abbott. R.B. is an advisory board member of Medtronic, Tandem Diabetes, Abbott, Novo Nordisk, Lilly, and Aventis. D.L.P. received travel grants from Lilly and Menarini. S.Z. reports grants from Sanofi Italy and Movi SpA. F.L. received speaking honoraria from Movi SpA and is an advisory board member of Sanofi. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. S.P. and G.S. conceptualized the study and wrote the first draft of the paper. B.B., N.M., M.B., R.B., F.S., E.M., F.D.C., S.M., V.G., C.M., C.A.P., C.A., D.T., B.F., R.R., A.Z., M.D., D.L.P., E.C., C.R., R.F., P.R., I.R., and G.M. collected data. A.A. realized statistical analysis. S.Z. and M.M. reviewed and edited the manuscript. F.L. contributed to discussion and reviewed and edited the manuscript. All authors approved the final version of the manuscript. S.P. 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.

Handling Editors. The journal editors responsible for overseeing the review of the manuscript were Steven E. Kahn and Jeremy Pettus.