Introduction

Children coping with diabetes mellitus (DM) has been a crisis globally, which is a cluster of metabolic diseases, rather than a single illness, that are characterized by chronic hyperglycemia1. Type 1 Diabetes Mellitus (T1DM), an autoimmune disease destructs the pancreatic islet cells due to the inability in producing insulin, hormone essential for metabolism of glucose2. The reported cases of children with T1DM is rising staggeringly, and about 96,000 children under the age of fifteen are diagnosed with T1DM annually3. There are variability in the prevalence of adolescent T1DM with the highest incidence of T1DM reported in the United States (US), India and Brazil4. There is a huge economic burden from the chronic nature of the disease due to its multiple short and long-term complications, presenting it as a global health crisis to handle5.

Due to β-cell destruction and absolute insulin deficiency nature of T1DM6, the therapeutic goal for adolescents with type 1 diabetes is to reach an optimal glycemic control to avoid acute and chronic complications without compromising the wellbeing of children, and social outcomes their families7 The complicated nature of management of adolescent T1DM is aimed to prevent the development of complications such as cardiovascular disease, retinopathy, nephropathy, and neuropathy by achieving optimal glycaemia, avoiding hypoglycemic events8.

Glycemic targets for children with T1DM have become more rigorous over time, and the average blood glucose (sugar) level goal is now 70 to 120 mg/dL (4 to 7 mmol/L) for all children regardless of age9,10. The handling of hypoglycemic remains a challenge with adolescents due to their varied physical activities and requirements. Only less than 70% the youth affected with T1DM achieve glycemic targets despite advances in insulin therapies, educational awareness, insulin pumps, and ‘continuous glucose monitoring systems11. The daily tasks of dietary intake control, strict medical control and continuous monitoring makes it difficult for young youth to control their T1DM situations.

The management of T1DM has been revolutionized by the recent advances in technology, like the insulin pumps and continuous glucose monitoring sensors12. The Sensor-Augmented Insulin Pump (SAP), which mimics the insulin production by the pancreas via continuous subcutaneous insulin injection (CSII), is a conventional treatment13. This traditional approach of SAP therapy does not functionally prevent the chances of occurrence of hypoglycemia. This was a hindrance to the guarantee that the mean blood glucose value will meet therapeutic expectations after using the SAP therapy14. The most current development is the integration of the insulin pumps and continuous glucose monitoring sensors, to modify and administer insulin based on the values detected by the sensor15. This paved way for the artificial pancreas or called as the closed-loop system for diabetes management. The Closed-Loop Control System (CLC) insulin delivery systems is characterized by real-time glucose-responsive insulin administration and combines glucose-sensing and insulin- delivery components16.

Though these advances in diabetes technology are widely used in clinical practice17, clinical evidence for the practice of use of CLC insulin delivery as an alternative to SAP therapy has not yet done or available18. Studies on inpatient and outpatient adolescent patients with T1DM have been done on closed-loop systems with improved glycemic outcomes and reduction in hypoglycemia in children, especially in the overnight period. CLC has been associated with fewer adverse effects than other insulin therapies in the treatment of adolescents with T1DM19,20,21,22,23,24,25,26. Though the effectiveness SAP insulin delivery was stated to be effective in maintaining the daytime glycemic outcomes, it needs to be compared with contemporary interventions27,28,29. There is a serious gap in the availability of evidence to support clinical decisions for the right insulin therapy for adolescent T1DM. It is very crucial to identify if Closed-loop insulin delivery can be a potential replacement for SAP insulin therapy of adolescent T1DM management by comparing the glycemic outcomes and hypoglycemic events to measure the adverse effects of the therapies. This meta-analysis is therefore aimed to compare the efficacy and safety of CLC to SAP for adolescents with type 1 diabetes (T1DM) by evaluating the glycemic outcomes and hypoglycemic events of the two therapies.

The participants, intervention, comparisons, outcomes, and type of studies (PICO) of the current review were as follows: Participants (P): Patients younger than 19 years, Intervention (I): Closed-Loop Control (CLC) in T1DM adolescent patients. Comparisons (C): Sensor-augmented insulin pump (SAP) in T1DM adolescent patients. Outcomes (O): Efficacy of the intervention assessed by average blood glucose value measured by continuous glucose measurement, time in range (TIR), and standard deviation (SD) of glucose variability. Safety was evaluated according to reported hypoglycemic events. Hence this review raises the research question “Does Closed-Loop Control System have a better efficacy and safety compared to Sensor-Augmented Insulin Pump for managing type one diabetes in adolescents?”

Methodology

This study adheres to The Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines30 and complies all the steps advised in Cochrane handbook of systematic reviews. This study has been registered with Prospero with ID CRD42022333310.

Search strategy

A through bibliographic search of electronic databases of PubMed, MEDLINE, EMBASE, and the Cochrane Library was undertaken in this study. Boolean operators were used for the keywords designed, for searching and identifying the relevant literature. Table 1 is explanatory of the combination of keywords used in this study for the assimilation of articles to review. Grey literature was obtained by searching Web of Science, ProQuest Dissertations and clinicaltrials.gov. The reference sections of retrieved original articles and reviews were scanned for studies that might have been missed in the primary searches. Studies were filtered with regard to study design, methodological features, the reported glycemic outcomes, and the adverse effects evaluated under each study.

Table 1 Keyword strategy used in the database search.
Full size table

The participants, intervention, comparisons, and outcomes (PICO) of the current meta-analysis were as follows.

Participants (P): Patients younger than 19 years as per the definition of an adolescent as a person “10 to 19 years inclusive” and a child “a person 19 years or younger”31. Intervention (I): Closed-Loop Control (CLC) in T1DM adolescent patients. Comparisons (C): Sensor-Augmented Insulin Pump (SAP) in T1DM adolescent patients. Outcomes (O): Efficacy of the intervention assessed by average blood glucose value measured by continuous glucose measurement, time in range (TIR), and standard deviation (SD) of glucose variability. Safety was evaluated according to reported hypoglycemic events.

Data extraction

Relevant full text articles were assimilated after review of the titles and abstracts. The eligibility criteria including the inclusion and exclusion criteria for this study are described in Table 2. Figure 1 illustrates the PRISMA flow diagram for the studies selected in the search process and eligibility appraisal. Review manager 5.4.1 (Revman, Cochrane Collaboration, and Oxford, UK) was used to manage, analyze, and synthesize the included study data. The institutional research board and ethics committee ruled out that approval was not required for this study being a review study.

Table 2 Eligibility criteria for study selection in this study.
Full size table
Figure 1
figure 1

PRISMA flow chart descriptive of the study selection30,32.

Full size image

Data extraction of the related was done using a custom made data-extraction form in excel. All relevant information on the included studies was extracted into an electronic database, including participant and intervention characteristics, relevant glycemic outcomes and hypoglycemic events to evaluate the adverse effects of treatments, type of insulin delivery technology used and industrial funding or influence on the study.

Quality assessment

The risk of bias method from the Cochrane Collaboration was used33 to appraise the quality of the included studies. The studies were graded as low, high, or unclear risk of bias for each of the following items using this method. The domains included in this grading of risk of bias were the random sequence generation and allocation concealment, (both items relate to selection bias), masking of participants and personnel (detection bias), incomplete outcome data (attrition bias), selective reporting (reporting bias), and other biases. Figure 2 is descriptive of the risk of bias tool used in this study.

Figure 2
figure 2

Risk of bias summary about the methodological quality of studies included using the Cochrane risk of bias tool. Symbols show low risk of bias (+), and high risk of bias (−).

Full size image

Outcome evaluation

Efficacy: glycemic outcomes-day, night and during strenuous physical activities

The primary endpoints were the day, night and during strenuous physical activities monitoring of the mean (1) Blood Glucose (BG) level from continuous glucose monitoring, (2) Time In range (TIR) for the percentage of time spent in normoglycemia, 70–180 mg/dL34, and (3) Standard Deviation (SD) of glucose variability.

Safety: adverse effects-day, night and during strenuous physical activities

The adverse effects outcomes were analyzed from the day, night and during strenuous physical activities monitoring of the time spent while in hypoglycemia < 70 mg/dL, and in hyperglycemia > 250 mg/dL34.

Data analysis

Statistical analysis and the assessment of heterogeneity was done for each reported outcomes in the included studies. All the aggregated outcome measurements has been unified in units to meaningfully analyze the data. The weighted mean difference (WMD) with 95 % confidence interval (CI) was calculated for all the continuous outcomes. The I2 statistic and χ2 test was used to evaluate the heterogeneity of the analysis results. If I2 > 50% or p < 0.1 for the χ2 test, the random-effects model was adopted; otherwise, the fixed- effects model was used35. P value less than 0.05 was considered statistically significant. All statistical analyses were performed by RevMan software (version 5.0, Oxford, United Kingdom). When the level of heterogeneity was less than 50%, a fixed-effect model was used36. The meta-analysis was performed with Review manager 5.4.1 (RevMan, Cochrane Collaboration, Oxford, UK).

Informed consent

For this type of study, formal consent is not required.

Patient and public involvement

This study being a meta-analysis systematic literature review, the patients selected were recruited by the researchers of the included studies. All the patient and families related aspects involved in design and implementation of the interventions were priori addressed by the authors of the selected studies.

Results

Study search and data extraction

An initial search gave 869 articles from the keyword combinations. The included trials were published between 2015 and 2022. After strict screening and quality assessment 1137,38,39,40,41,42,43,44,45,46,47 Randomized Controlled Trials (RCT) were selected for the review which were relevant to the search terms and criteria. A total of 570 adolescent patients were included in this study from the selected studies. This included 298 patients who took CLC insulin therapy and 272 patients who had SAP as their insulin therapy. Only 1037,38,39,40,41,42,43,44,46,47 articles were included in the meta-analysis after reviewing and accounting for heterogeneity. (Fig. 1). Table 3 provides the summary of the data extracted from the attributes of the included studies.

Table 3 The summary of the attributes included studies.
Full size table

Characteristics and quality of trials

In relation to the masking of participants and personnel, almost all of the trials were rated at ‘‘low risk of bias’’ (9 of 11 trials, 81.81%); as for attrition bias and reporting bias, almost all the trials were rated at ‘‘low risk of bias,’’ because they reported the complete outcome data (10 out of 11 trials, 90.90%). There were no studies at ‘‘high risk of bias’’ with any issues relating to random sequence generation, allocation concealment and masking of outcome assessment (11 out of 11 trials, 100%). Figure 2 shows the risk of bias summary based on review quality appraisal judgements about each risk of bias item for each included study.

Efficacy: glycemic outcomes during day, night and during strenuous physical activities

The results from the included studies were pooled by unifying the measurement units to mg/dL. Hence all the pooled comparison results in this meta-analysis are in mg/dL.

Mean blood glucose (BG) level: day, night and during strenuous physical activities

The average BG was compared in 9 studies37,39,40,41,42,43,44,46. The day monitoring comparison of the BG level showed [Mean Difference (IV, Random, 95% CI) − 4.33 [− 6.70, − 1.96]]. Pooled studies show [Heterogeneity: Tau2 = 6.24; Chi2 = 77.30, df = 8 (P < 0.00001); I2 = 90%. Test for overall effect: Z = 3.58 (P = 0.0003)]. The night monitoring level of BG was reported by 5 studies37,39,42,46,47, and it was compared. The results showed (Mean Difference (IV, Random, 95% CI) − 16.61 [− 31.68, − 1.54]). Pooled studies show [Heterogeneity: Tau2 = 215.07; Chi2 = 75.06, df = 4 (P < 0.00001); I2 = 95%. Test for overall effect: Z = 2.16 (P = 0.03)]. The Forest plot in Fig. 3a and b are illustrative of these results. Only two studies37,39 showed the results for glycemic outcome during extreme physical activities like winter sports. The physical activity monitoring comparison of the BG level demonstrates [Mean Difference (IV, Random, 95% CI) − 8.27 [− 19.52, 2.99]]. Pooled studies show [Heterogeneity: Chi2 = 2.95, df = 1 (P = 0.09); I2 = 66%. Test for overall effect: Z = 1.44 (P = 0.15)]. Figure 3c depicts this observation.

Figure 3
figure 3

(a) Forest plot of comparison: Mean BG-Day. (b) Forest plot of comparison: Mean BG-Night. (c) Forest plot of comparison: Mean BG-Physical activity.

Full size image

Time in range (TIR): day, night and during strenuous physical activities

Time in range (TIR) for the percentage of time spent in normoglycemia, during the day 70–180 mg/dL was compared in all the 10 studies37,38,39,40,41,42,43,44,46,47. The results showed (Mean Difference (IV, Random, 95% CI) − 13.18 [− 19.18, − 7.17]. Pooled daytime study aggregate show [Heterogeneity: Tau2 = 60.11; Chi2 = 67.63, df = 7 (P < 0.00001); I2 = 90%. Test for overall effect: Z = 4.30 (P < 0.0001)] Excluding 4 studies40,43,44,46 to due to statistical heterogeneity only four studies37,38,39,46 which reported the night time monitoring results of TIR were pooled. The comparison meta results show [Mean Difference (IV, Random, 95% CI) − 15.36 [− 26.81, − 3.92]]. Pooled studies for nighttime data shows [Heterogeneity: Tau2 = 94.12; Chi2 = 16.83, df = 3 (P = 0.0008); I2 = 82%.Test for overall effect: Z = 2.63 (P = 0.009)]. The Forest plot in the Fig. 4a and b elaborates these findings. Two studies37,39 recorded the TIR during strenuous physical activities and the results are as follows show [Mean Difference (IV, Random, 95% CI) − 7.39 [− 17.65, 2.87]]. Pooled studies for nighttime data shows [Heterogeneity: Chi2 = 3.53, df = 1 (P = 0.06); I2 = 72%. Test for overall effect: Z = 1.41 (P = 0.16)]. Figure 4a–c are illustrative of these results.

Figure 4
figure 4

(a) Forest plot of comparison: Time in Range TIR-Day. (b) Forest plot of comparison: Time in Range TIR-Night. (c) Forest plot of comparison: Time in Range TIR-Physical activity.

Full size image

Standard deviation (SD) of glucose variability: day and night

The SD of glucose variability for day time monitoring was compared all the 10 studies37,38,39,40,41,42,43,44,46,47 (Mean Difference (IV, Random, 95% CI) − 0.40 [− 0.79, − 0.00]]. Pooled studies for daytime measurements were homogeneous with results as shown [Heterogeneity: Tau2 = 0.05; Chi2 = 8.09, df = 7 (P = 0.32); I2 = 13%. Test for overall effect: Z = 1.98 (P = 0.05)]. The night SD of glucose variability was compared in 5 studies38,39,43,45,46 as reported. The results of the pooled studies were (Mean Difference (IV, Random, 95% CI) -0.86 [− 2.67, 0.95] Pooled studies were homogeneous with results as shown [Heterogeneity: Tau2 = 1.46; Chi2 = 6.28, df = 3 (P = 0.10); I2 = 52%. Test for overall effect: Z = 0.93 (P = 0.35)]. Figure 5a and b shows the Forest plot of this stated results. Figure 5a and b are explaining these results. Two studies39,50 reported the variability as the coefficient of variation and it was converted to standard deviation and pooled. It can be noted that no studies directly reported the standard deviation of glucose variability during physical activities.

Figure 5
figure 5

(a) Forest plot of comparison: SD of glucose variability-Day. 5(b) Forest plot of comparison: SD of glucose variability-Night.

Full size image

Safety: adverse effects (AE) outcomes

Hypoglycemic events: day, night and during strenuous physical activities

AEs were compared in 10 studies37,38,39,40,41,42,43,44,46,47, including a total of 570 subjects. After excluding studies to account for statistical heterogeneity, the day time reporting results for hypoglycemic events was pooled in from 937,38,39,41,42,43,44,46,47 of the included studies are (Mean Difference (IV, Random, 95% CI) − 0.54 [− 1.86, 0.79]]. Pooled studies exhibits heterogeneity, which is shown as [Heterogeneity: Tau2 = 2.78; Chi2 = 58.72, df = 8 (P < 0.00001); I2 = 86%. Test for overall effect: Test for overall effect: Z = 0.80 (P = 0.43)]. The 7 studies37,38,39,40,43,44,46,47 were pooled in after accounting for heterogeneity, reported the night comparison results for hypoglycemia and the results shows (Mean Difference (IV, Random, 95% CI) 0.04 [− 0.20, 0.27]). Pooled studies show homogenous results which were, [Heterogeneity: Tau2 = 0.00; Chi2 = 3.29, df = 6 (P = 0.77); I2 = 0%. Test for overall effect: Z = 0.30 (P = 0.77)]. Forest plot in Fig. 6a and b are descriptive of these meta results. The pooled results from the two studies37,39 which included results for hypoglycemia during physical activities are as follows, (Mean Difference (IV, Random, 95% CI) 0.00 [− 0.25, 0.25]). Pooled studies show no heterogeneity, which is shown as [Heterogeneity: Chi2 = 0.00, df = 1 (P = 1.00); I2 = 0%. Test for overall effect: Z = 0.00 (P = 1.00)]. Figure 6a–c are descriptive of these results.

Figure 6
figure 6

(a) Forest plot of comparison: Hypoglycemia-Day. (b) Forest plot of comparison: Hypoglycemia-Night. (c) Forest plot of comparison: Hypoglycemia-Physical activity.

Full size image

Hyperglycemic events: day, night and during strenuous physical activities

Hyperglycemic events were monitored to assess the AEs of both the insulin delivery systems under comparison. Six studies37,38,39,40,41,43 reported day time results for hyperglycemic events and the pooled results are [Mean Difference (IV, Random, 95% CI) − 0.48 [− 2.62, 1.65]]. Pooled studies show [Heterogeneity: Tau2 = 4.16; Chi2 = 31.45, df = 5 (P < 0.00001); I2 = 84%. Test for overall effect: Z = 0.44 (P = 0.66)]. From the six studies37,38,39,40,41,43 which reported the night comparison results for hyperglycemia were pooled and the results shows [Mean Difference (IV, Random, 95% CI) − 7.11 [− 12.77, − 1.45]]. Pooled studies show [Heterogeneity: Tau2 = 36.21; Chi2 = 83.74, df = 5 (P < 0.00001); I2 = 94%. Test for overall effect: Z = 2.46 (P = 0.01)]. Hyperglycemia during physical activities were reported by two studies37,39 which were pooled in this meta-analysis and the results are as follows, [Mean Difference (IV, Random, 95% CI) − 0.00 [− 0.10, 0.10]] Pooled studies show [Heterogeneity: Chi2 = 3.45, df = 1 (P = 0.06); I2 = 71%. Test for overall effect: Z = 0.04 (P = 0.97)]. Forest plot in Fig. 7a–c are illustrative of these pooled results.

Figure 7
figure 7

(a) Forest plot of comparison: Hyperglycemia-Day. (b) Forest plot of comparison: Hyperglycemia-Night. (c) Forest plot of comparison: Hyperglycemia-Physical activity.

Full size image

Discussion

A significant morbidity and mortality have been associated with T1DM among adolescents due to its poor prognosis. Before, SAP therapy was a crucial advancement in diabetes treatment but, currently the development of CLC insulin delivery systems has been dramatically gaining clinical importance. Till now, there has been no comprehensive meta-analyses comparing the day and night time efficacy and safety features between traditional SAP therapy and the currently used CLC insulin delivery for pediatric and adolescents with type 1 diabetes. This study is the first meta-analysis to examine the day and night time efficacy and safety monitoring comparison of CLC insulin delivery systems versus SAP therapy in the treatment of adolescents with T1DM. The results from this study shows the supreme authority of the CLC insulin delivery systems in maintaining all the glycemic outcomes than SAP therapy for day and night values of mean BG, TIR, and SD of glucose variability. This study considered the monitoring results during daily routine activities along with strenuous physical activities like skiing and camp activities. Thus, this study, clearly updates and upholds well-defined evidence of the CLC insulin delivery’s efficacy to be clinically used for young patients with T1DM. The safety comparison results from this studies identifies that, CLC insulin delivery was associated with fewer AEs, especially hypoglycemic and hyperglycemic-related events during day and night, than SAP which deems CLC to be an ideal treatment of choice for adolescents with type 1 diabetes.

Limitations

The limitation of this study is that only 11 RCTs were included, and the number of included patients were limited. One45 among the included studies though it participated adolescent patients, failed to separate the adolescent result reports from the adults. Many studies and RCTs are concentrated around the insulin therapy of adults and pediatric T1DM based studies are not given enough attention. Although in this study, CLC insulin delivery was seen to have decreased risk of AEs, hypoglycemia is still the dominant AE. There is an imminent need of extensive high-quality RCTs to ensure the reliability of this conclusion. The included studies were mostly conducted in Europe and United States, which may cause a regional bias in identifying with race related outcomes. There was significant heterogeneity when all the included studies were pooled, which could decrease the reliability of the results. The included studies used different equipments, which possibly increased the heterogeneity of the results. This led to the exclusion of studies with conflicting results from the majority of included studies and had more confidence results. For this reason, random-effects meta-analysis was utilized to incorporate heterogeneity among studies. Finally, all the 1137,38,39,40,41,42,43,44,45,46,47 included studies had the industrial support on their research which can have potential influence results reported by them, favoring the equipment or technology used in the experiment.

Conclusion

CLC insulin delivery exhibits significantly better day and night efficacy and safety than SAP therapy in adolescents with type 1 diabetes. Closed-loop safely and significantly improves glycemic control, maintains time in range, reduces hypoglycemia and hyperglycemia in adolescent populations with T1DM. Tedious and continuous technical improvement of closed-loop systems is required to further improve safety and efficacy, likely through the development of open-frame, personalized, cloud-based ecosystems.