GRAPHICAL ABSTRACT

Highlights

• This analysis of AID systems in T1D includes data from 65 RCTs with 3,623 patients.

• AID systems improved TIR by 11.74% and GRI by 3.74 compared to other therapies.

• Dual-hormone FCL systems showed greater TIR improvement than HCL systems.

• The benefits for TBR were more pronounced in patients with longer diabetes duration.

INTRODUCTION

Type 1 diabetes mellitus (T1DM), resulting from insulin deficiency caused by autoimmunity mediated destruction of β-cells, requires lifelong treatment with exogenous insulin. Adjustments in daily insulin dosing based on carbohydrate counting and frequent self-monitoring of blood glucose are arduous and challenging for both patients with T1DM and their caregivers [1]. Therefore, many patients with T1DM fail to meet the glycemic target goals in all age groups, e.g., target goals were achieved in only 17% for youth and by only 21% for adults [2] and life expectancy remains significantly 24 years shorter than in the non-diabetic population [1]. To improve this situation, insulin therapy has been improved by multiple daily injection (MDI), continuous subcutaneous insulin infusion (CSII), sensor-augmented pump (SAP), SAP with low-glucose suspend feature (SAP+LGS) and SAP with predictive low-glucose suspend feature (SAP+PLGS) in recent years, with consequent glucose improvement in patients with T1DM [3-5].

Automated insulin delivery (AID) system, is a closed-loop system that can deliver insulin automatically according to the glucose level by combining an insulin pump, continuous glucose monitoring (CGM), and a control algorithm, including full closed-loop (FCL) systems and hybrid closed-loop (HCL) systems; AID systems are emerging as a promising therapy for T1DM by providing appropriate insulin dosing in real-time, to limit both hyperglycemia and hypoglycemia. Therefore, AID systems could reduce the burden of glucose management compared with traditional insulin therapy [6,7].

AID systems in T1DM have attracted increasing attention in different populations of all age groups [8]. Several meta-analyses of AID systems uncovered promising efficacy [9-12]. Two large pooled analyses, including 24 studies with 585 participants in 2017 and 40 studies with 1,027 participants in 2018 [11,12], verified its favorable effect with increasing time in range (TIR) and decreasing time in either hyperglycemia or hypoglycemia in all age groups. But the maximum sample sizes for these analyses were only 54 and 75, respectively, and the longest follow-up duration was only 3 months. Since then there has been an upsurge in larger AID systems clinical trials, up to 326 cases, of longer duration, up to 24 months [13,14]. Recently, a meta-analysis conducted in 2023, including 25 studies with 1,345 participants, demonstrated the long-term effectiveness of AID systems in improving TIR, and the favorable effect on time below range (TBR) and time above range (TAR) [10]. However, that meta-analysis focused on children and adolescents and had potential methodological limitation, such as the handing of median values [15]. Moreover, updated artificial intelligence and science and technology have only since become available to allow FCL systems [16-19]. It is, therefore, timely we believe, to provide a comprehensive re-evaluation of both efficacy and safety of AID systems in patients with T1DM.

METHODS

The current study was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement [20] and registered with PROSPERO (CRD42023475763).

Data sources and searches

The literature published on PubMed, Embase, Web of Science, and Cochrane Library, and the grey literature from the ClinicalTrials.gov website was searched from database inception to August 31, 2023, using the keywords including “type 1 diabetes mellitus,” “artificial pancreas,” “automated insulin delivery,” and “randomized controlled trial.” The detailed search strategy is shown in the Supplementary Methods (Search strategy). Our search was restricted to studies published in English.

Study selection

We included randomized controlled trial (RCT) that compared AID systems with insulin-based standard care in T1DM patients, irrespective of age, trial design (parallel or crossover), setting (outpatient or inpatient), or intervention duration. Studies involving non-T1DM participants and pregnant women were excluded. In addition, editorials, case reports, conference papers, and guidelines were excluded. The treatments in the control group included MDI, CSII, SAP, and SAP with LGS or PLGS.

Data extraction and quality assessment

Reference management software (Endnote X9, Clarivate, Philadelphia, PA, USA) was used to duplicate the identified literature. Three researchers (W.F., C.D., and R.X.) initially independently screened the title and abstract of the literature, then the full text, extracted data from each study using a standardized data extraction form in Excel as shown in the Supplementary Methods (Data extraction form) and evaluated the quality of each clinical trial using the Cochrane risk of bias tool that each study was classified as either high, low, or unclear risk of bias [21] and the certainty of evidence for all outcomes following the Grading of Recommendations Assessment, Development, and Evaluation (GRADE) framework with GRADEpro GDT software (https://www.gradepro.org) [22], and disagreements were resolved by consensus or arbitrated by a fourth reviewer (X.L.).

The primary outcome was percentage of TIR (3.9 to 10.0 mmol/L [70 to 180 mg/dL] or 4.0 to 10.0 mmol/L [72 to 180 mg/dL] or 4.0 to 9.9 mmol/L [72 to 178 mg/dL], depending on the study). The secondary outcomes were other glucose metrics including the percentage of TAR (>10 mmol/L [180 mg/dL]) or TBR (<3.9 mmol/L [70 mg/dL] or <4.0 mmol/L [72 mg/dL], depending on the study), coefficient of variation (CV) of glucose, glycosylated hemoglobin (HbA1c) and insulin dose. In addition, the glycemia risk index (GRI) [23], a new composite CGM metric of glycemic risk, was also selected as a second outcome. Safety outcomes include diabetic ketoacidosis (DKA) and severe hypoglycemia.

Data synthesis and analysis

Review Manager version 5.3 (Cochrane, London, UK) and STATA version 14.0 (StataCorp., College Station, TX, USA) were used to perform statistical analyses. The 95% confidence intervals (CIs) were calculated for all results. Binary adverse outcomes were compared with the relative risk (RR). Continuous outcomes were compared with weighted mean differences, except for the insulin dose, which was compared with standardized mean differences (SMDs) owing to the different units in different studies. When the outcome medians and ranges were reported, appropriate formulae were selected to calculate the mean and variance [24]. We combined the study data from both parallel-group and crossover designs. Crossover trials were analyzed using group means and standard deviations as if they were parallel-group trials. Based on this assumption, the analysis was generally conservative [25]. In addition, we conducted a priori-decided subgroup analysis according to the timing of the intervention (24 hours vs. overnight), type of hormone (single vs. dual), age (adults vs. adolescents vs. mixed population), baseline HbA1c level (<7.5% vs. ≥7.5%), and remote monitoring (yes vs. no). We defined the Inreda artificial pancreas (AP), CamAPS HX, and bionic pancreas as the FCL and the other AID systems as the HCL, based on the consensus recommendations which defining no carbohydrate counting and manually initiated prandial boluses as FCL systems [26]. We also conducted a post hoc subgroup analysis based on type of closed-loop (dual-hormone FCL, single-hormone FCL, and HCL), time of follow-up (<3 months vs. ≥3 months), disease duration (<20 years vs. ≥20 years), type of algorithm and type of comparator.

The heterogeneity of the analysis was assessed by the I² and the chi-squared test based on the Cochran Q test. If I² ≥50% or P<0.1 for the chi-squared test, the random effects model was used by the DerSimonian and Laird estimation method; otherwise, a fixed-effects model was used. Sensitivity analysis was performed to estimate the robustness of the meta-analysis results, which was conducted using the leave-one-out strategy and repeating the meta-analyses, including studies with a parallel design. Publication bias was assessed using a funnel plot and Egger’s test, where a P value less than 0.05 was considered the presence of publication bias. For all analyses, except for the Q test, statistical significance was set at P<0.05.

Data and resource availability

The data sets generated during or analyzed in the current study are available from the corresponding author upon reasonable request.

RESULTS

Study selection

A total of 14,392 records were identified from the database search, and 79 records were identified from ClinicalTrials.gov. Two hundred and forty-one studies were assessed through a full-text review for eligibility and 65 studies were included in this analysis (Supplementary Fig. 1).

Characteristics of included studies

The baseline characteristics of the studies are summarized in Supplementary Table 1. Given the different populations, settings, and amounts of intervention hormones, each of eight studies were entered as two separate comparisons in the meta-analysis [27-34]. Thus, a total of 73 comparisons from 65 studies with 3,623 patients with T1DM (44 crossover and 21 parallel designs) were included in the analysis. There are 34 studies published in the last 5 years and the remaining 31 studies were published between 2013 and 2017. Seven studies used FCL systems and the remainder used HCL systems, with a wide variation in follow-up time ranging from one night to 24 months and in baseline HbA1c level ranging from 6.7% to 10.6%.

Risk of bias assessment

The Cochrane risk of bias tool was used to assess the methodological quality and bias of all included studies, as shown in Supplementary Fig. 2. Random sequence generation presented a low risk of bias in all studies because of the randomized control study design. Given the nature of the intervention, blinding of participants and personnel presented a high risk of bias in all studies, and blinding of outcome assessors presented a high risk of bias in most of the included studies except for three studies [35-37]. Allocation concealment presented a low risk of bias in 15 studies and an unclear risk of bias in other studies.

Primary outcomes

Sixty-two comparisons from 56 studies with 3,306 patients were pooled to analyze TIR. TIR was 11.74% (95% CI, 9.37 to 14.12; P<0.001; I²=94.6%, moderate certainty) of time per day (2 hours 49 minutes) higher with AID systems compared with other control treatments (Table 1, Fig. 1). The favorable effect of TIR was consistent in all subgroups (Table 2). Of note, AID systems had a greater improvement in TIR in dual-hormone FCL (19.64% of time per day [4 hours 43 minutes]; 95% CI, 13.74 to 25.55) compared with HCL (10.87% of time per day [2 hours 37 minutes]; 95% CI, 8.31 to 13.44) (Fig. 2). Additionally, the model predictive control (MPC)-based AID systems system revealed a similar improvement in TIR (11.41% of time per day [2 hours 44 minutes]; 95% CI, 8.74 to 14.09) compared with proportional integral derivative (PID)-based AID systems (11.60% of time per day [2 hours 47 minutes]; 95% CI, 7.65 to 15.54), and the patients with baseline HbA1c level of more than 7.5% showed a greater improvement in TIR (13.25% of time per day [3 hours 11 minutes]; 95% CI, 10.27 to 16.22) compare with the patients with baseline HbA1c level of less than 7.5% (8.84% of time per day [2 hours 7 minutes]; 95% CI, 5.91 to 11.78) with a significant subgroup difference (P=0.039). In addition, there was no difference in the subgroup of follow-up time (11.82% of time per day [2 hours 50 minutes] vs. 11.31% of time per day [2 hours 43 minutes], P value for subgroup difference=0.810).

Secondary outcome and safety outcome

Sixty comparisons with 3,253 patients were pooled for TBR. Compared to the control treatment, the period using AID systems was shortened by approximately 17 minutes per day (–1.20%; 95% CI, –1.62 to –0.79; P<0.001; I²=91.7%, moderate certainty) (Table 1, Supplementary Fig. 3). The favorable effects were consistent for all subgroups. However, differences in reduction in TBR were higher in studies with diabetes duration of more than 20 years (–1.80% of time per day [26 minutes]; 95% CI, –2.45 to –1.15) compared with diabetes duration of less than 20 years (–0.86% of time per day [12 minutes]; 95% CI, –1.41 to –0.31) with a significant subgroup difference (P=0.031). Use of AID systems was also associated with a significant reduction in studies with baseline HbA1c less than 7.5% compared with baseline HbA1c more than 7.5% (–1.93% of time per day [28 minutes] vs. –0.87% of time per day [13 minutes], P=0.033). There was no significant statistical difference in the subgroup of follow-up time (–1.57% of time per day [27 minutes] vs. –0.81% of time per day [12 minutes], P=0.057).

In total, 54 comparisons with 2,769 patients were pooled for TAR. TAR was 10.17% of time per day (95% CI, –13.18 to –7.16; P<0.001; I²=96.0%, moderate certainty) lower for AID systems compared to control treatment, equivalent to 2 hours 26 minutes per day (Table 1, Supplementary Fig. 4). Similarly, TAR improved in all subgroups. Results were more significantly favorable for dual-hormone FCL systems (–16.88% of time per day [4 hours 3 minutes]; 95% CI, –22.41 to –11.35) compared with HCL systems (–9.22% of time per day [2 hours 13 minutes]; 95% CI, –12.56 to –5.89). AID systems showed a reduction in TAR in MPC-based systems (–10.96% of time per day [2 hours 39 minutes]; 95% CI, –14.09 to –7.83; P<0.001). There was almost no difference in the follow-up of less than 3 months compared to more than 3 months.

In terms of glucose variability, compared to control, treatment with AID systems demonstrated a favorable effect on CV (–1.31%; 95% CI, –2.16 to –0.47; P=0.002; I²=83.1%, moderate certainty) (Table 1, Supplementary Fig. 5). The favorable effect was greater in dual-hormone FCL systems than HCL systems both in CV (–4.34% vs. –1.11%). Significantly greater reductions were seen in the dual-hormone systems, in adult, with remote monitoring, and with diabetes duration of more than 20 years group for CV.

Considering the risk of hypoglycemia and hyperglycemia, we performed an additional pooled analysis of GRI. Use of AID systems was associated with the reduction in GRI (–3.74; 95% CI, –6.34 to –1.14; P=0.005; I²=99.3%, moderate certainty) (Table 1, Supplementary Fig. 6).

Further pooled analyses of 25 comparisons with 2,375 patients for the long-term effect of AID systems on glycemic control exhibited that the HbA1c level was 0.36% (95% CI, –0.44 to –0.28; P<0.001; I²=28.0%, high certainty) lower in AID systems than control treatment. Finally, no difference between AID systems and other insulin-based treatments was seen in the insulin dose (SMD, 0.03; 95% CI, –0.08 to 0.15; P=0.577; I²=57.2%, moderate certainty) (Table 1, Supplementary Figs. 7 and 8).

For the safety outcomes, episodes of severe hypoglycemia were reported in 19 trials: 53 patients occurred during AID systems treatment and 38 patients occurred during control use. Episodes of DKA were reported in nine trials: eight patients occurred during AID systems and five patients occurred in control therapy. Pooled effects were RR 0.92 (95% CI, 0.65 to 1.31; P=0.616; I²=18.7, high certainty) for severe hypoglycemia and RR 1.15 (95% CI, 0.47 to 2.80; P=0.766; I²=0.0, high certainty), indicating no differences between AID systems and control therapy (Table 1, Supplementary Figs. 9 and 10).

Sensitivity analyses

We further performed a sensitivity analysis for all outcomes to detect whether any single study could affect the reliability of the included studies by omitting studies one by one (Supplementary Figs. 11-19). When any single study was excluded, all outcomes showed that the point estimate of pooled effects still stayed within the 95% CI, indicating that our analysis results were stable.

We also repeated the meta-analysis for all outcomes using studies with a parallel design (Supplementary Table 2). The result also was consistent for TIR (12.29%; 95% CI, 9.44 to 15.14; P<0.001). Similar results were observed for all other secondary outcomes. In addition, despite the washout period in randomized crossover trials, considering possible within-person differences, we performed another sensitivity analysis for the primary outcome, adjusting for within-person differences in studies that did not report the mean and standard error of paired differences [25]. The estimate for TIR was unchanged at 11.67% (95% CI, 9.85 to 13.50; P<0.001).

Publication bias and GRADE assessment

Publication bias was assessed for all outcomes by using a funnel plot (Supplementary Fig. 20) and Egger’s test (Supplementary Table 3). We find symmetrical funnel plots and non-significant Egger’s test results for TIR (P=0.204), TBR (P=0.488), TAR (P=0.418), CV (P=0.178), GRI (P=0.373), HbA1c (P=0.128), insulin dose (P=0.264), severe hypoglycemia (P=0.072), and DKA (P=0.334). Overall, the results for all outcomes were robust.

Considering that the nature of the intervention precludes the blinding of patients and personnel, the certainty of the evidence was not rated down for risk of bias. There was substantial heterogeneity for primary and secondary outcomes except HbA1c, severe hypoglycemia, and DKA, resulting in a downgraded of the items of inconsistency. Therefore, the certainty of evidence was high for HbA1c, severe hypoglycemia and DKA, and moderate for other outcomes (Supplementary Table 4).

DISCUSSION

In this comprehensively systematic review and meta-analysis comparing AID systems with other conventional insulin-based treatments (non-AID systems), our data demonstrated that the application of AID systems therapy resulted in significant improvement in glycemic outcomes for individuals living with T1DM. Greater treatment effects favoring AID systems concerning GRI was observed in adults and patients with poorer glycemia control and longer diabetes duration. Importantly, use of dual-hormone FCL systems yielded superior glycemic control compared with HCL systems including increased time of TIR throughout the day. More benefits of hypoglycemia were also seen in patients on AID systems who were within target glycemic control and with longer diabetes duration. Also, using AID systems beyond 3 months up to 24 months showed better glycemic control in all age groups, supporting the recommendation of long-term use of AID systems even though in patients within target glycemic control.

Currently, landmark progress in insulin replacement therapy has shifted from SAP to HCL systems [38], and then to FCL. FCL system is a big step for lightening the burden of glycemia control, which relieving the hard work of carbohydrate counting. Despite FCL systems being developed recently there is no pooled analysis comparing the effects of FCL and HCL. In our study, users of dual-hormone FCL, not single-hormone FCL, showed greater benefits concerning glycemic control (increased TIR) potentially due to the joint result of insulin and glucagon. Regarding baseline HbA1c level that may affect the outcome, 88% (7/8) of comparisons in the FCL groups and 70% (45/65) of comparisons in the HCL groups were above 7.5%. Therefore, the high percentage of hyperglycemia in FCL groups at baseline may contribute to the benefit of FCL systems in terms of TIR and TAR. Indeed, a RCT found that both a new advanced HCL system, RocketAP, and an HCL system, Unified Safety System Virginia, were able to achieve greater glycemic control with announced meals compared to RocketAP with unannounced meals [39]. More head-to-head trials comparing FCL with HCL systems in all age groups should be performed.

Another major concern with insulin therapy is the high risk of hypoglycemia, especially life-threatening nocturnal hypoglycemia [40]. To screen who would benefit the most from hypoglycemia risk, we performed subgroups analysis. Our data demonstrated that the beneficial effect of AID systems for hypoglycemia was more pronounced in patients within target glycemia control and longer diabetes duration, with AID systems reducing TBR by 1.93% and 1.80%. It is interesting that TBR reduced by less in children (down 0.91%) than in adolescents (down 1.55%), which is congruent with recent results that the benefits of TBR were less in children than adults [10]. Difference in improving TBR between different ages can be attributed, at least partially, to small insulin doses, unpredictable food intake, unscheduled exercise activities and rapid growth in children and adolescents [41]. This data adds to the need to pay more attention to young people when employing AID systems.

GRI, as a novel composite derived from CGM metrics, is recognized to be a promising tool for assessing glycemic quality in clinical practice [42] and has been proposed as a comprehensive index in the latest consensus on CGM metrics [43]. Recent real-world data revealed that HCL systems provided a significant improvement in GRI [44]. However, GRI was only available in 6 of our pooled comparisons, suggesting that this index is not widely used in clinical trials. In our analysis, the favorable effects of AID systems were reflected in the better index of GRI, underscoring the feasibility of using GRI as a measurable parameter to evaluate glycemic risk. In this sense, future studies might be recommended to provide the results of GRI, enabling a pooled analysis included more studies.

Encouragingly, these beneficial effects of AID systems appeared to be consistent over the long-term follow-up period, with an effect evident in the first 3 months. This result is in line with a recent meta-analysis focused on children and adolescents [10], though it included only nine trials of long-term intervention. Our larger meta-analysis (of 23 trials) included adults as well as children and adolescents, incorporated more evidence and estimated a broader variety of outcomes, supporting the robust extension of those previous results.

Another interesting finding is that beneficial glycemic effects were particularly prominent in the dual-hormone group compared with single-hormone group, findings suggested in previous meta-analysis [11]. Our present analysis further confirmed and expanded the generalizability of the results across various populations and settings. Beyond the benefits of reduced hypoglycemia offered by glucagon in dual-hormone AID systems, glycemic variability and hyperglycemia were also significantly lower in dual-hormone systems, suggesting the value of glucagon in mimicking physiologic glucose regulation. Additionally, with regard to algorithm, in congruent with previous meta-analysis [45] and a head-to-head comparison of MPC versus PID algorithms [46], our meta-analysis showed that both MPC- and PID-based AID systems performed well in glycemic control. However, in contrast to previous results supporting the superior performance of MPC algorithms in TIR whether meal was announced or not [45,46], we found that MPC-based AID systems did not present a significantly greater improvement in maintaining blood glucose in the target range over PID-based AID systems. In brief, the results from this meta-analysis could guide optimal use and selection of the AID systems to facilitate improvements in safe glycemic control.

Our study has several strengths. To our knowledge, this is the largest and the most comprehensive analysis of pooled RCTs, without restrictions of setting and age, ensuring more generalizability. Half of the studies included were published in the last 5 years when the technology of AID systems has shown rapid development. In particular, we compared the different effects of FCL and HCL, and found that dual-hormone FCL systems could achieve greater glycemic control, emphasizing the superiority of multi-hormonal combination [16,47]. Moreover, the value of GRI as the evaluation indicator for glycemic control in AID systems application as proposed illustrated a more favorable effect in AID systems for adult patients or with baseline HbA1c ≥7.5%, or with diabetes duration ≥20 years. Finally, we have demonstrated that improved glycemic control which occurs in the short-term can persist.

Nevertheless, we acknowledge several limitations in the meta-analysis. First, even though there are 11 studies (17%) with a sample size of more than 100, the sample size was relatively small in most studies, which may affect the accuracy of the effect estimated. Second, heterogeneity was high for glucose outcomes, which may be explained by different types and times of intervention or different baseline characteristics. Third, all studies were performed in an open-label manner due to the nature of the intervention. Lack of blinding leads to a high risk of performance bias. Fourth, we did not assess the impact of the AID systems intervention on sleep quality, quality of life, and self-management burden for T1DM patients or the medical burden for the entire country.

In conclusion, AID systems significantly improve glucose control in patients with T1DM compared to other insulin-based treatments. Greater benefits of hypoglycemia were observed in patients within target glycemic control and longer diabetes duration. The favorable effect of TIR was more significant in the dual-hormone FCL systems than in the HCL systems.

SUPPLEMENTARY MATERIALS

NOTES

CONFLICTS OF INTEREST

No potential conflict of interest relevant to this article was reported.

AUTHOR CONTRIBUTIONS

Conception or design: C.D., Z.Z., X.L.

Acquisition, analysis, or interpretation of data: W.F., C.D., R.X.

Drafting the work or revising: W.F., C.D., Z.L., R.D.L., Z.Z., X.L.

Final approval of the manuscript: W.F., C.D., R.X., Z.L., R.D.L., Z.Z., X.L.

FUNDING

This work was supported by the Unveiling and Leading Program of Hunan Province (2021JC0003), the Science and Technology Innovation Program of Hunan Province (2020RC4044), and the Sinocare Diabetes Foundation (2021SD06, LYF2022039).

Acknowledgements

The authors thank Editage (www.editage.cn) for English language editing.

Fig. 1.

Forest plot for time in range comparing automated insulin delivery systems with other insulin-based treatment. SD, standard deviation; CI, confidence interval.

Fig. 2.

Prespecified subgroup analyses for primary and secondary outcomes by the type of automated insulin delivery systems (dual-hormone full closed-loop systems, single-hormone full closed-loop, or hybrid closed-loop systems). CI, confidence interval; HbA1c, glycosylated hemoglobin.

REFERENCES

• 1. Gregory GA, Robinson TI, Linklater SE, Wang F, Colagiuri S, de Beaufort C, et al. Global incidence, prevalence, and mortality of type 1 diabetes in 2021 with projection to 2040: a modelling study. Lancet Diabetes Endocrinol 2022;10:741-60.ArticlePubMed
• 2. Foster NC, Beck RW, Miller KM, Clements MA, Rickels MR, DiMeglio LA, et al. State of type 1 diabetes management and outcomes from the T1D exchange in 2016-2018. Diabetes Technol Ther 2019;21:66-72.ArticlePubMedPMC
• 3. REPOSE Study Group. Relative effectiveness of insulin pump treatment over multiple daily injections and structured education during flexible intensive insulin treatment for type 1 diabetes: cluster randomised trial (REPOSE). BMJ 2017;356:j1285.ArticlePubMed
• 4. Alotaibi A, Al Khalifah R, McAssey K. The efficacy and safety of insulin pump therapy with predictive low glucose suspend feature in decreasing hypoglycemia in children with type 1 diabetes mellitus: a systematic review and meta-analysis. Pediatr Diabetes 2020;21:1256-67.ArticlePubMedPDF
• 5. Bosi E, Choudhary P, de Valk HW, Lablanche S, Castaneda J, de Portu S, et al. Efficacy and safety of suspend-before-low insulin pump technology in hypoglycaemia-prone adults with type 1 diabetes (SMILE): an open-label randomised controlled trial. Lancet Diabetes Endocrinol 2019;7:462-72.ArticlePubMed
• 6. Nwokolo M, Hovorka R. The artificial pancreas and type 1 diabetes. J Clin Endocrinol Metab 2023;108:1614-23.ArticlePubMedPMCPDF
• 7. Wilson LM, Jacobs PG, Riddell MC, Zaharieva DP, Castle JR. Opportunities and challenges in closed-loop systems in type 1 diabetes. Lancet Diabetes Endocrinol 2022;10:6-8.ArticlePubMed
• 8. Templer S. Closed-loop insulin delivery systems: past, present, and future directions. Front Endocrinol (Lausanne) 2022;13:919942.ArticlePubMedPMC
• 9. Fang Z, Liu M, Tao J, Li C, Zou F, Zhang W. Efficacy and safety of closed-loop insulin delivery versus sensor-augmented pump in the treatment of adults with type 1 diabetes: a systematic review and meta-analysis of randomized-controlled trials. J Endocrinol Invest 2022;45:471-81.ArticlePubMedPDF
• 10. Zeng B, Gao L, Yang Q, Jia H, Sun F. Automated insulin delivery systems in children and adolescents with type 1 diabetes: a systematic review and meta-analysis of outpatient randomized controlled trials. Diabetes Care 2023;46:2300-7.ArticlePubMedPMCPDF
• 11. Weisman A, Bai JW, Cardinez M, Kramer CK, Perkins BA. Effect of artificial pancreas systems on glycaemic control in patients with type 1 diabetes: a systematic review and meta-analysis of outpatient randomised controlled trials. Lancet Diabetes Endocrinol 2017;5:501-12.ArticlePubMed
• 12. Bekiari E, Kitsios K, Thabit H, Tauschmann M, Athanasiadou E, Karagiannis T, et al. Artificial pancreas treatment for outpatients with type 1 diabetes: systematic review and meta-analysis. BMJ 2018;361:k1310.ArticlePubMedPMC
• 13. Boughton CK, Allen JM, Ware J, Wilinska ME, Hartnell S, Thankamony A, et al. Closed-loop therapy and preservation of C-peptide secretion in type 1 diabetes. N Engl J Med 2022;387:882-93.ArticlePubMed
• 14. Bionic Pancreas Research Group, Russell SJ, Beck RW, Damiano ER, El-Khatib FH, Ruedy KJ, et al. Multicenter, randomized trial of a bionic pancreas in type 1 diabetes. N Engl J Med 2022;387:1161-72.ArticlePubMedPMC
• 15. Bekiari E, Karagiannis T, Haidich AB, Tsapas A. Meta-analysis of artificial pancreas trials: methodological considerations. Lancet Diabetes Endocrinol 2017;5:685.Article
• 16. Boughton CK, Hartnell S, Lakshman R, Nwokolo M, Wilinska ME, Ware J, et al. Fully closed-loop glucose control compared with insulin pump therapy with continuous glucose monitoring in adults with type 1 diabetes and suboptimal glycemic control: a single-center, randomized, crossover study. Diabetes Care 2023;46:1916-22.ArticlePubMedPDF
• 17. Mosquera-Lopez C, Wilson LM, El Youssef J, Hilts W, Leitschuh J, Branigan D, et al. Enabling fully automated insulin delivery through meal detection and size estimation using artificial intelligence. NPJ Digit Med 2023;6:39.ArticlePubMedPMCPDF
• 18. Tsoukas MA, Majdpour D, Yale JF, Fathi AE, Garfield N, Rutkowski J, et al. A fully artificial pancreas versus a hybrid artificial pancreas for type 1 diabetes: a single-centre, open-label, randomized controlled, crossover, non-inferiority trial. Lancet Digit Health 2021;3:e723-32.PubMed
• 19. Blauw H, Onvlee AJ, Klaassen M, van Bon AC, DeVries JH. Fully closed loop glucose control with a bihormonal artificial pancreas in adults with type 1 diabetes: an outpatient, randomized, crossover trial. Diabetes Care 2021;44:836-8.ArticlePubMedPDF
• 20. Page MJ, McKenzie JE, Bossuyt PM, Boutron I, Hoffmann TC, Mulrow CD, et al. The PRISMA 2020 statement: an updated guideline for reporting systematic reviews. BMJ 2021;372:n71.ArticlePubMedPMC
• 21. Higgins JP, Altman DG, Gotzsche PC, Juni P, Moher D, Oxman AD, et al. The Cochrane Collaboration’s tool for assessing risk of bias in randomised trials. BMJ 2011;343:d5928.ArticlePubMedPMC
• 22. Guyatt G, Oxman AD, Akl EA, Kunz R, Vist G, Brozek J, et al. GRADE guidelines: 1. introduction-GRADE evidence profiles and summary of findings tables. J Clin Epidemiol 2011;64:383-94.ArticlePubMed
• 23. Klonoff DC, Wang J, Rodbard D, Kohn MA, Li C, Liepmann D, et al. A glycemia risk index (GRI) of hypoglycemia and hyperglycemia for continuous glucose monitoring validated by clinician ratings. J Diabetes Sci Technol 2023;17:1226-42.PubMed
• 24. Wan X, Wang W, Liu J, Tong T. Estimating the sample mean and standard deviation from the sample size, median, range and/or interquartile range. BMC Med Res Methodol 2014;14:135.ArticlePubMedPMCPDF
• 25. Elbourne DR, Altman DG, Higgins JP, Curtin F, Worthington HV, Vail A. Meta-analyses involving cross-over trials: methodological issues. Int J Epidemiol 2002;31:140-9.ArticlePubMed
• 26. Phillip M, Nimri R, Bergenstal RM, Barnard-Kelly K, Danne T, Hovorka R, et al. Consensus recommendations for the use of automated insulin delivery technologies in clinical practice. Endocr Rev 2023;44:254-80.PubMed
• 27. Burnside MJ, Lewis DM, Crocket HR, Meier RA, Williman JA, Sanders OJ, et al. Open-source automated insulin delivery in type 1 diabetes. N Engl J Med 2022;387:869-81.ArticlePubMed
• 28. Haidar A, Legault L, Matteau-Pelletier L, Messier V, Dallaire M, Ladouceur M, et al. Outpatient overnight glucose control with dual-hormone artificial pancreas, single-hormone artificial pancreas, or conventional insulin pump therapy in children and adolescents with type 1 diabetes: an open-label, randomised controlled trial. Lancet Diabetes Endocrinol 2015;3:595-604.ArticlePubMed
• 29. Haidar A, Messier V, Legault L, Ladouceur M, Rabasa-Lhoret R. Outpatient 60-hour day-and-night glucose control with dual-hormone artificial pancreas, single-hormone artificial pancreas, or sensor-augmented pump therapy in adults with type 1 diabetes: an open-label, randomised, crossover, controlled trial. Diabetes Obes Metab 2017;19:713-20.PubMed
• 30. Haidar A, Rabasa-Lhoret R, Legault L, Lovblom LE, Rakheja R, Messier V, et al. Single- and dual-hormone artificial pancreas for overnight glucose control in type 1 diabetes. J Clin Endocrinol Metab 2016;101:214-23.PubMed
• 31. Kariyawasam D, Morin C, Casteels K, Le Tallec C, Sfez A, Godot C, et al. Hybrid closed-loop insulin delivery versus sensor-augmented pump therapy in children aged 6-12 years: a randomised, controlled, cross-over, non-inferiority trial. Lancet Digit Health 2022;4:e158-68.ArticlePubMed
• 32. Russell SJ, El-Khatib FH, Sinha M, Magyar KL, McKeon K, Goergen LG, et al. Outpatient glycemic control with a bionic pancreas in type 1 diabetes. N Engl J Med 2014;371:313-25.ArticlePubMedPMC
• 33. Thabit H, Tauschmann M, Allen JM, Leelarathna L, Hartnell S, Wilinska ME, et al. Home use of an artificial beta cell in type 1 diabetes. N Engl J Med 2015;373:2129-40.ArticlePubMedPMC
• 34. Wilson LM, Jacobs PG, Ramsey KL, Resalat N, Reddy R, Branigan D, et al. Dual-hormone closed-loop system using a liquid stable glucagon formulation versus insulin-only closed-loop system compared with a predictive low glucose suspend system: an open-label, outpatient, single-center, crossover, randomized controlled trial. Diabetes Care 2020;43:2721-9.ArticlePubMedPDF
• 35. Abraham MB, de Bock M, Smith GJ, Dart J, Fairchild JM, King BR, et al. Effect of a hybrid closed-loop system on glycemic and psychosocial outcomes in children and adolescents with type 1 diabetes: a randomized clinical trial. JAMA Pediatr 2021;175:1227-35.PubMed
• 36. Choudhary P, Kolassa R, Keuthage W, Kroeger J, Thivolet C, Evans M, et al. Advanced hybrid closed loop therapy versus conventional treatment in adults with type 1 diabetes (ADAPT): a randomised controlled study. Lancet Diabetes Endocrinol 2022;10:720-31.ArticlePubMed
• 37. McAuley SA, Lee MH, Paldus B, Vogrin S, de Bock MI, Abraham MB, et al. Six months of hybrid closed-loop versus manual insulin delivery with fingerprick blood glucose monitoring in adults with type 1 diabetes: a randomized, controlled trial. Diabetes Care 2020;43:3024-33.ArticlePubMed
• 38. Boscari F, Avogaro A. Current treatment options and challenges in patients with type 1 diabetes: pharmacological, technical advances and future perspectives. Rev Endocr Metab Disord 2021;22:217-40.ArticlePubMedPMCPDF
• 39. Garcia-Tirado J, Diaz JL, Esquivel-Zuniga R, Koravi CL, Corbett JP, Dawson M, et al. Advanced closed-loop control system improves postprandial glycemic control compared with a hybrid closed-loop system following unannounced meal. Diabetes Care 2021;dc210932.ArticlePubMed
• 40. Kaur J, Seaquist ER. Hypoglycaemia in type 1 diabetes mellitus: risks and practical prevention strategies. Nat Rev Endocrinol 2023;19:177-86.ArticlePubMedPDF
• 41. Urakami T. The advanced diabetes technologies for reduction of the frequency of hypoglycemia and minimizing the occurrence of severe hypoglycemia in children and adolescents with type 1 diabetes. J Clin Med 2023;12:781.ArticlePubMedPMC
• 42. Kim JY, Yoo JH, Kim JH. Comparison of glycemia risk index with time in range for assessing glycemic quality. Diabetes Technol Ther 2023;25:883-92.ArticlePubMed
• 43. Battelino T, Alexander CM, Amiel SA, Arreaza-Rubin G, Beck RW, Bergenstal RM, et al. Continuous glucose monitoring and metrics for clinical trials: an international consensus statement. Lancet Diabetes Endocrinol 2023;11:42-57.PubMed
• 44. Eviz E, Yesiltepe Mutlu G, Karakus KE, Can E, Gokce T, Muradoglu S, et al. The advanced hybrid closed loop improves glycemia risk index, continuous glucose monitoring index, and time in range in children with type 1 diabetes: real-world data from a single center study. Diabetes Technol Ther 2023;25:689-96.ArticlePubMed
• 45. Kang SL, Hwang YN, Kwon JY, Kim SM. Effectiveness and safety of a model predictive control (MPC) algorithm for an artificial pancreas system in outpatients with type 1 diabetes (T1D): systematic review and meta-analysis. Diabetol Metab Syndr 2022;14:187.ArticlePubMedPMCPDF
• 46. Pinsker JE, Lee JB, Dassau E, Seborg DE, Bradley PK, Gondhalekar R, et al. Randomized crossover comparison of personalized MPC and PID control algorithms for the artificial pancreas. Diabetes Care 2016;39:1135-42.ArticlePubMedPMCPDF
• 47. Lal RA, Maikawa CL, Lewis D, Baker SW, Smith AA, Roth GA, et al. Full closed loop open-source algorithm performance comparison in pigs with diabetes. Clin Transl Med 2021;11:e387.ArticlePubMedPMCPDF

Figure & Data

References

Citations

Citations to this article as recorded by