A comprehensive systematic review of published systematic reviews investigating the effect of diabetes monitoring systems was performed. It was based on the following approach: (1) identification of a research question; (2) identification of inclusion and exclusion criteria; (3) data extraction; (4) reporting results; (5) assessment of risk of bias; (6) discussion and interpretation of the results. The research questions formulated during the study were focused on the available new technologies for diabetes monitoring; available systematic reviews and meta-analyses reporting the treatment outcomes; types of therapeutic outcomes reported and their variability.

The type of treatment outcomes that were observed were changes in the clinical parameters (HbA1c, glucose level, etc.); changes in the quality of life (QoL); number and frequency of hypo- and hyperglycemia episodes; change in the risk for complications; level of satisfaction from the intervention.For the identification of the studies, we performed an internet search of electronic databases PubMed, Cochrane library, Embase and proposed by them related articles. The key words for the search were “diabetes” AND “systematic review,” AND “CGMS” OR “SAPT” OR “CSII” OR “FGM” OR “Closed loop systems” OR “telemedicine.” For the presentation of the search results was used the principles of the PRISMA checklist with its four steps the search approach, as follows: identification, screening, eligibility, and inclusion. The search encompasses the period since the first systematic review identified till the end of 2019.

Specific inclusion and exclusion criteria were formulated. The criteria for inclusion were to be a systematic review of interventional or non-interventional (observational) studies about diabetes monitoring systems, to present clinical and/or economic effect of new technologies on diabetes patients, English language. The criteria for exclusion were lifestyle maintaining technologies, medicines, diet,exercise, algorithms’ decision supporting systems, mobile apps (especially those that relate to the lifestyle maintaining), alarms, and m-health. Our focus on excluding some studies was more on removing ones that rely solely on telemonitoring by physicians. If more than two studies were found from the same authors, the latest published article was taken into account.

Five authors (MK, MD, ZM, KT, GP) reviewed selected articles for duplication, relevance to the inclusion and exclusion criteria, and consolidate the evidences by systematizing them according to the technology described, type of the diabetes for which the technology is recommended, sources of information, clinical and/or economic results reported, recommendations for future application or improvement. Each author independently reviewed the articles for eligibility. Discrepancies between the authors were overcome through discussion, until reaching consensus. The main focus was on technologies aiming to improve diabetes control via improvement in glycated hemoglobin (HbA1c), hypo- or hyperglycemic episodes, or glucose secretion monitoring and not on improving the lifestyle habits and compliance.

The data extracted were summarized in tables including the following relevant information:

The upper and lower changes in clinical parameters were summarized by intervention, type of diabetes, and type of review [systematic review (SR) or systematic review and meta-analysis (SR/MA)]. The results were presented in tables and figures. We have not included detailed information about the sample size of primary studies and location of the study because we have focused mainly on the clinical outcomes described in the published systematic reviews and/or meta-analyses.

The methodology quality of systematic reviews and meta-analysis included in the current systematic review was assessed using the GRADE system (Table 3). GRADE has four levels of evidence—also known as certainty in evidence or quality of evidence: very low, low, moderate, and high. All included SRs and MAs were evaluated in five domains—Risk of Bias, Imprecision, Inconsistency, Indirectness, and Publication Bias. Publication bias in this context refers to the stated values of HbA1c in the reviewed papers and the conclusions drawn from the results, since unpublished and unreferenced studies are not available. All study conclusions were carefully reviewed along with the Supplemented Data. To minimize the risk of including predominantly “positive” analyses, we aimed to include also government-sanctioned assessments, which were available online, such as IQWiG reports, and reports by the US AHRQ (Agency for Health Research and Quality).

After the search in PubMed, and EMBASE with key words diabetes and new technologies 1,495 studies were screened based on abstracts. Then we performed the second search in PubMed and Cochrane library with the terms diabetes, systematic review, and new technologies and we identified 300 studies. We excluded 195 due to duplication and 1,216 due to non-correspondence with the inclusion criteria, and 84 studies were reviewed manually. After adding the term meta-analysis, we expand the list to 138 studies. Out of them 56 candidate full texts were reviewed a third time by an independent reviewer, and a further 24 were excluded. There are multiple reasons for excluding studies, which are all summarized in Supplementary Table 1, some of the reasons are systematic review of the application of the technologies, not purely on effectiveness, systematic reviews commissioned by regulatory bodies, which did not provide all evidence, but just a full working summary and others. Finally, we included 32 full systematic reviews and/or meta-analyses for which the research quеstion was clearly defined (12–38, 40–44). The included studies were of high quality, predominantly focusing on randomized controlled trials. Seven of the included systematic reviews included also observational studies assessing the effect of DMS in everyday clinical practice.

They are summarized in Table 1. A narrative synthesis is presented. Figure 2 presents the search process flow chart.

The details of the included SRs (n = 16) and SRs plus MAs (n = 16) are presented in Table 2. They were published in the period 2008–2019 and the number of analyzed studies (randomized clinical trials, cohort, crossover or parallel or prospective observational studies, etc.) in each separate SR or MA varied between 2 and 44. Diabetes monitoring systems observed were: Continuous Glucose Monitoring (CGMS), Sensor Augmented Pump Therapy (SAPT), Self-monitoring Blood Glucose (SMBG), Continuous subcutaneous insulin infusion (CSII), Flash Glucose Monitoring (FGM) or Intermittent-scanned continuous glucose monitoring (isCGM), Closed-loop systems, and telemedicine. All studies analyzed and reported the changes in glycated hemoglobin (HbA1c) levels as a primary clinical outcome. Some of them reported results about well-being, hypoglycemic fear and episodes, hyperglycemia incidents, overall diabetes distress, quality of life, patients’ satisfaction. Standard care (for example a weekly venipuncture protocol), conventional treatment (non-meter through urine tests and blood-glucose levels measured at the fortnightly clinic visits), or no comparator was used mainly as comparators for all CGMS. CGMSs were also compared with SMBG with or without CSII.

The technologies, that were assessed in the analyzed systematic reviews and meta-analyses, and the relevant number of studies identified are presented on Figures 3 and 4. The most reviewed technologies were the continuous glucose monitoring systems with 11 systematic reviews and 15 meta-analyses. One analysis focused on self-monitoring approaches. It should be noted that CGMS studies also evaluated continuous subcutaneous injections as a method of insulin delivery and had many analyzed subsections. Most analyses focused on Type 1 diabetes (26) with only a small amount analyzing separately intensive insulin control in Type 2 patients, or in a mixed sample of Type 1 and Type 2 diabetes patients. The number of studies investigating gestational diabetes patients are limited (6) as they mostly compare SMBG vs. CGM and SMBG vs. standard care.

Blood glucose levels are an important clinical measure for estimating effectiveness, and overall analysis of all included studies showed that there is a reduction in HbA1c levels. These can be seen in both Type 1 and Type 2 diabetes, for all interventions.The reductions are indeed almost negligble. Since the aim of control, even self-control of diabetes mellitus (DM), is to reduce HbA1c levels, the purpose of the new diabetes technologies, apart from blood glucose control is to reduce glycemic excursions, hypo- or hyperglycemia episodes, increase time spent within desired ranges, and improve compliance, especially in children. The reduction in glycated hemoglobin (HbA1c) levels varies from 0·17 to 0·70% during use of CGM. It is most significant for CGM identified in meta-analysis—from 0·23 to 0·37% for patients with Type 1 diabetes and from 0·20 to 0·48% for Type 2 diabetes. In systematic reviews a wider range and greater reduction of HbA1c were observed: from 0·20 to 0·70% in Type 1 and from 0·17 to 0·50% in Type 2. Telemonitoring would support the reduction of the HbA1c-levels in patients with Type 1 diabetes from 0·53 to 1·10%. Data for gestational diabetes was identified only for CGM where the reduction is 0·26–0·34% (23). Change in QALY (ΔQALY) was observed only in one study (19) and it is higher in Type 2 diabetes (0·272) than in Type 1 diabetes (0·046) when comparing real-time CGM vs. SMBG.

When using CGM the time spent in hypoglycemia is expected to be lower vs. SMBG—0·4–0·1 h/day vs. 0·65–0·6 h/day (31). Hill et al. (20), reported significant reduction in time spent in hyperglycemia—the mean difference was −68.56 min/day favoring RT-CGM (95% CI, −101·17 to −35·96) vs. SMBG. CGM ensured the rate for hypoglycemia of 3·50 (95% CI, 1·07 to 11·44) and for hyperglycemia 1·42 (95% CI, 0·26 to 7·82) when comparing SMBG and CGM. Significant decrease in time in hypo- and hyperglycemia for closed-loop systems in comparison with CSII was observed—mean difference of 0·67 and 3·01%, respectively (16).

We took a much broader approach of including systematic reviews of randomized clinical trials (RCTs) and observational studies because regulatory authorities prefer to have a real-world evidence for the decision making when they decide to reimburse a particular device. The most notable conclusions, with the level of evidence used are summarized in Table 2. Since we predominantly focused on SRs and MAs of randomized trials, the strength of evidence for most studies is high, such as in DMS, where 16 of the 26 studies were of high quality and 9 of moderate. We used the GRADE system to assess study quality, and it should be noted, that GRADE includes some subjectivity, since it is implemented manually and not mechanically. Although the GRADE system is transparent, the decision ultimately falls on the reviewer whether to downgrade a randomized trial based on the GRADE system’s “factors which may influence the quality level of a body of evidence.” The closed-loop systems significantly reduce complications such as hypo- or hyperglycemia, as well as the CGMs, which although moderately effective, increased the time spent in range regardless if the monitoring system was with self-injection or continuous injection delivery methods. Patients with high-compliance benefit less than patients with low compliance. Based on the collected evidence the current systematic review could highlight that the CGM is an effective and suitable method for monitoring of blood glucose levels. It could ensure reductions in HbA1c as they vary in a wide range—between 0.20–0.70% in Type 1 and between 0.17 and 0.50% in Type 2 diabetes patients. The accuracy and benefits of CGM utilization are deeply examined and confirmed. The evidences reveal that real-time CGM in Type 1 diabetes improves clinical parameters, whereas a smaller number of studies consider the results of patients with Type 2 diabetes. The recommendations mainly concern improvement of unforeseen hypoglycemia risk and glucose variability in Type 2 diabetes patients (49, 51). In our study we also found that larger number of analysis are focused on Type 1 diabetes, while those analyzing the results in Type 2 or combined studies are a smaller number.

The latest years precision of CGM systems has improved and their accuracy within glucose levels >80–200 mg/dl is similar (46). CGM could be used for self-adjustment of dosage, interpretation of hypoglycemia results, and measurement of response to therapy as it covers a wide range of glucose values. Its accuracy depends also on glucose levels variability (39). CGM is likely to improve treatment results, improve glycemic control and quality of life, as well as lower micro- and macro-vascular outcomes despite the existing barriers and educational needs for physicians and patients (55, 65, 66). The utilization of CGM with remote monitoring in children with Type 1 diabetes leads to better quality of life, parental sleep, and decreases family stress (56), while utilization in youth with Type 1 diabetes resulted in improved adherence, glycemic control, as well as a low psychosocial distress (57). Moreover, CGM is able to pick up asymptomatic hypoglycemic episodes in older patients with diabetes Type 1 or 2 and to ensure a reduction in severe hypoglycemic episodes (47). Argento et al. reported that the severe hypoglycemic episodes dropped from 52 (5 years before CGM initiation) to 12 after starting CGM (50). The proportion of patients with Type 1 diabetes with any severe hypoglycemia felt from 79 to 31% after initiation of CGM. However, Lagarde et al. concluded that no difference exists between the number of minor hypoglycemic episodes between the CGMS and the control group of children with Type 1 diabetes (mean difference 0·53, 95% CI, −0·68 to 1·74; p = 0·39) (47). Similar results are presented by Langendam et al. —no significant difference is revealed in risk of severe hypoglycemia or ketoacidosis between CGM and SMBG adults with Type 1 diabetes (29). However, due to the small number of participants and limitations, findings should be interpreted with significant caution. Hill et al. did not find any significant difference in severe hypoglycemia events between the rt-CGM and SMBG groups, but found a significant reduction in time spent in the hyperglycemic range (31). Decreased time in hypoglycemia (13 out of 15 studies) in Type 1 diabetes as well as increased time in range (TIR) as a result of CGM usage were also observed in a systematic review by De Ridder et al. (33). Other study reported that adult patients with Type 1 diabetes who use CGM perceive improvements in their quality of life, especially related to hypoglycemia fear (22). Only a few studies have found positive outcomes regarding hypoglycemia when using CGM (reduction in nocturnal hypoglycemia episodes by 54% with is CGM vs. SMBG (−0·29 ± 0·08 h per 7 h; P = 0·0001) (45, 63) but no decrease in time spent in hypoglycemia was observed. The other important outcome, Time in Glycemic Range (70–180 mg/dl), is reported to increase with isCGM (intermittently scanned) among well-controlled patients with Type 1 diabetes (33). Whereas, the results are controversial among adult patients with uncontrolled T2DM using insulin (62).

Jones et al., reported that CGM is able to reduce neonatal hypoglycemia (RR 0·66, 95% CI 0·48 to 0·93; 3 studies, 428 infants) (21). Latest updated evidence from 2019 by Yu et al. suggests that CGM is superior to SMBG among pregnant women with gestational diabetes mellitus as it is able to detect hypoglycemic and hyperglycemic episodes (36). Therefore, regular monitoring of glucose levels in pregnant women with diabetes through specific glucose monitoring devices could ensure limitation of hypoglycemic episodes and then influence the outcomes for both mother and child. No significant improvement in the frequency of neonatal hypoglycemia or any other primary outcomes were detected among pregnant women with pre-existing diabetes when using glucose monitoring technique.

Our study reveals a small number of MAs and SRs exploring results in CGM utilization in children. The findings show that CSII was associated with improved quality of life compared with MDI and similar results on HbA1c levels and severe hypoglycemia. Regarding closed-loop systems, meta-analyses by Karageorgiou et al., and by Weisman et al., showed that these systems lead to significantly higher percentage of time spent in the target glycemic range and to lower percentages of time in hyperglycemia and hypoglycemia for non-adult Type 1 diabetes patients (16, 17). Free Style Libre Flash Glucose Self-Monitoring System also showed promising results for reduction in time and number of events with glucose levels <3·9 in 24 h in comparison with SMBG. Evidence regarding isCGM (Intermittent-scanned continuous glucose monitoring) impact on improving time in glycemic range, glycemic variability, and hypoglycemia are variable and further clinical trials should investigate these devices (12).

Because of the variability and lack of enough strong evidence, no general conclusion or recommendation about the patients target groups who might be most suitable for particular DMS could be highlighted. Basing on the available evidence, it could be mentioned that patients’ satisfaction, preferences, lifestyle habits, age, therapy applied, and severity of the condition (type of diabetes, duration, concomitant diseases, etc.) are some of the main criteria for choosing a method for monitoring and control.

The current systematic review has significant strengths as it gathers evidence for the effectiveness of a variety of diabetes monitoring devices both from controlled interventional studies and observational studies from the everyday clinical practices thus providing the opportunity to assess the effects of DMS from the perspective of different study desings. Moreover, a wide range of diabetes patients were included in the analysis—diagnosed with Type 1, Type 2, or gestational diabetes. Multiple databases were searched to identify relevant studies which answer to the research question. This systematic review of reviews provides evidence to inform both clinical practice and future research.

The main limitation is that the number of evidences for some diabetes monitoring systems such as telemonitoring, closed-loop systems, and SAPT, is too narrow, not sufficiently enough, and lack of statistical significance to make general conclusions. Moreover, different outcomes are measured and compared in the different studies which is a strong complication for a more comprehensive synthesis and analysis. Because of the limited data for assessing the effectiveness of monitoring technologies only the reduction in HbA1c levels for some DMS was analyzed. There are other criteria for assessing the quality of glycemic control such as number, duration of hypo- and hyperglycemic episodes, and time in glycemic range. Due to insufficient and controversial evidence for all valuable parameters, we assessed and presented only the variability in HbA1c for different patients’ groups and by type of study analysed (SR or SR+MA). Moreover, due to the heterogeneity of the methodologies, patient populations, and gathered data, we were not able to perform formal meta-analysis. So, a narrative synthesis is presented which could also be highlighted as a strong limitation of the study. Some DMS are innovative and not commonly applied so there is not enough relevant evidence. No studies comparing FGM and rt-CGM were included and analyzed in the current review which could be highlighted as another limitation of the review.

We made a distinction between the studies only on the criteria of whether they are only systematic reviews or include and meta-analysis of the analysed studies. No other selection criteria were applied for the primary studies included in each one of the observed reviews. We recognize that it might be a limitation of our analysis but trusted the authors performing the systematic reviews in their proper selection of the comparable RCTs or observational studies.

Very few studies performing a systematic review of already published SR and SR+MA of continuous glucose monitoring systems were identified. Published studies focus mainly on specific group of patients with Type 1 or Type 2 diabetes and on specific intervention (58).

However, a similar approach of systematizing published MA assessing supported self-management for people with Type 2 diabetes was identified. The authors focus on the role of self-management mechanism as one of main factors affecting the treatment outcomes and quality of life of patients. Some of the studies in the systematic review of meta-analysis show that self-monitoring systems and tele-health may provide some advantages in the process of self-management. Authors’ conclusions focus mainly on informative type of these findings which could be in favor to policy makers and health care professionals (59).

Other systematic review of reviews evaluates technology-enabled diabetes self-management. The study shows that mobile technologies for self-management of diabetes improve patient-generated health data and communication between patients and health care professionals. The results show that technology-enabled diabetes self-management solutions significantly improve HbA1c (67).

Similar to our results, a conducted narrative review showed that use of CGM in Type 2 diabetes patients leads to greater reductions in HbA1c in comparison with traditional self-monitoring as higher compliance to CGM was also reported. Logically, addition of other methods to CGM such as lifestyle counseling could lead to further improvements (52).

Evidences of key publications associating CGMS reported improvement in clinical outcomes, reducing of hypoglycemia and impact on physical, emotional, and relational aspects of everyday life (48). The positive effects of CGMS utilization are discussed from American Association of Clinical Endocrinologists and the American College of Endocrinology. The conclusion they reach is that increasing utilization of CGMS will probably improve the health outcomes, decrease health care resource costs for acute and chronic complications (53).

A review on studies observing glucose management reported that according to current literature evidence utilization of CGMS is mainly recommended in T1DM patients with a poor control of HbA1c levels after SMBG and risk of hypoglycemia, which confirms our findings. Utilization of CGMS is favored for patients with Type 2 diabetes who reported severe hypoglycemia or suspected hypoglycemia, particularly nocturnal (64).

We found very few studies reporting the M-value (durable nyctohemeral measurement of glycemic behavior) (16, 17) and SD of 24-h glucose (24, 32, 33) and did not find a study reporting the Mean amplitude of glycemic excursion (MAGE). We recognize that those measures of continuous diabetes control are currently introduced and might be more informative for the endocrinologist. Further studies need to be done to explore the previous research/systematic reviews investigating the effect of DMS on indices of glycemic variability such as MAGE, M-value, and SD of average 24-h glucose concentrations in patients with diabetes.

Only a few studies, however, assess the role of meta-analyses and systematic reviews on diabetes monitoring systems and their efficacy in terms of the HTA perspective. Insufficient cost-effectiveness studies and randomized clinical trials in specific patient populations used in HTA resulted in different criteria and rate of reimbursement among countries (54). In 2018 EUnetHTA published an HTA core model for rapid relative effectiveness assessment of continuous glucose monitoring (CGM real time, rtCGM) and flash glucose monitoring (FGM) as personal, standalone system in patients with diabetes mellitus treated with insulin (60). This report shows that in the light of the increasing number of different systems for rtCGM available on the market, systematic reviews and meta-analyses assessing their relative clinical effectiveness could be of a great importance in the assessment of their cost-effectiveness in terms of decision making. These studies could provide a summary of the best scientific available evidence which could facilitate the appraisal process and decision making and could favor the national/regional/local HTA. Health Technology Wales evidence appraisal report also included systematic review on the clinical and cost-effectiveness when providing decision for FreeStyle Libre flash glucose monitoring for the management of Type 1 or Type 2 diabetes (61).