Continuous glucose monitoring in adults with type 2 diabetes: a systematic review and meta-analysis

Aims/hypothesis Continuous glucose monitoring (CGM) is increasingly used in the treatment of type 2 diabetes, but the effects on glycaemic control are unclear. The aim of this systematic review and meta-analysis is to provide a comprehensive overview of the effect of CGM on glycaemic control in adults with type 2 diabetes. Methods We performed a systematic review using Embase, MEDLINE, Web of Science, Scopus and ClinicalTrials.gov from inception until 2 May 2023. We included RCTs investigating real-time CGM (rtCGM) or intermittently scanned CGM (isCGM) compared with self-monitoring of blood glucose (SMBG) in adults with type 2 diabetes. Studies with an intervention duration <6 weeks or investigating professional CGM, a combination of CGM and additional glucose-lowering treatment strategies or GlucoWatch were not eligible. Change in HbA1c and the CGM metrics time in range (TIR), time below range (TBR), time above range (TAR) and glycaemic variability were extracted. We evaluated the risk of bias using the Cochrane risk-of-bias tool version 2. Data were synthesised by performing a meta-analysis. We also explored the effects of CGM on severe hypoglycaemia and micro- and macrovascular complications. Results We found 12 RCTs comprising 1248 participants, with eight investigating rtCGM and four isCGM. Compared with SMBG, CGM use (rtCGM or isCGM) led to a mean difference (MD) in HbA1c of −3.43 mmol/mol (−0.31%; 95% CI −4.75, −2.11, p<0.00001, I2=15%; moderate certainty). This effect was comparable in studies that included individuals using insulin with or without oral agents (MD −3.27 mmol/mol [−0.30%]; 95% CI −6.22, −0.31, p=0.03, I2=55%), and individuals using oral agents only (MD −3.22 mmol/mol [−0.29%]; 95% CI −5.39, −1.05, p=0.004, I2=0%). Use of rtCGM showed a trend towards a larger effect (MD −3.95 mmol/mol [−0.36%]; 95% CI −5.46 to −2.44, p<0.00001, I2=0%) than use of isCGM (MD −1.79 mmol/mol [−0.16%]; 95% CI −5.28, 1.69, p=0.31, I2=64%). CGM was also associated with an increase in TIR (+6.36%; 95% CI +2.48, +10.24, p=0.001, I2=9%) and a decrease in TBR (−0.66%; 95% CI −1.21, −0.12, p=0.02, I2=45%), TAR (−5.86%; 95% CI −10.88, −0.84, p=0.02, I2=37%) and glycaemic variability (−1.47%; 95% CI −2.94, −0.01, p=0.05, I2=0%). Three studies reported one or more events of severe hypoglycaemia and macrovascular complications. In comparison with SMBG, CGM use led to a non-statistically significant difference in the incidence of severe hypoglycaemia (RR 0.66, 95% CI 0.15, 3.00, p=0.57, I2=0%) and macrovascular complications (RR 1.54, 95% CI 0.42, 5.72, p=0.52, I2=29%). No trials reported data on microvascular complications. Conclusions/interpretation CGM use compared with SMBG is associated with improvements in glycaemic control in adults with type 2 diabetes. However, all studies were open label. In addition, outcome data on incident severe hypoglycaemia and incident microvascular and macrovascular complications were scarce. Registration This systematic review was registered on PROSPERO (ID CRD42023418005). Graphical Abstract Supplementary Information The online version of this article (10.1007/s00125-024-06107-6) contains peer-reviewed but unedited supplementary material.


Introduction
Optimising glycaemic control is a keystone in the management of type 2 diabetes [1].Fingerstick-based self-monitoring of blood glucose (SMBG) has been the most used method for measuring daily glucose levels [2].However, this method does not provide continuous data about glucose levels, and, thus, may miss asymptomatic hypo-or hyperglycaemia and does not provide information about the direction of change in glucose levels.Furthermore, SMBG can be painful and increases disease burden [3].
The development of continuous glucose monitoring (CGM), either intermittently scanned CGM (isCGM) systems or real-time CGM (rtCGM) systems, has enabled the monitoring of glucose levels without fingersticks.CGM consists of a subcutaneous sensor that monitors interstitial glucose levels, which approximates blood glucose levels [3].CGM thereby allows for direct observation of glycaemic excursions and daily glucose profiles that can inform therapy decisions and possibly adjust behaviours [4,5].Recent guidelines have recommended CGM use in individuals with type 2 diabetes treated with insulin [1].However, the extent to which CGM improves glycaemic control in type 2 diabetes is unclear.Furthermore, it is unknown whether any such beneficial effect is different among individuals treated with or without insulin.
Glycaemic control is most often quantified by measurement of HbA 1c levels [4], which reflects average glucose over the last 2-3 months.CGM provides additional parameters of glycaemic control, including time in range (TIR), time below range (TBR) and time above range (TAR).These parameters provide information about glucose control on a daily basis, and are increasingly used in clinical research and daily care [1,4].
To date, there have been seven systematic reviews investigating the effect of CGM on glycaemic control in type 2 diabetes [6][7][8][9][10][11][12].However, these reviews only included a limited number of RCTs, i.e. six studies or fewer, or included studies with a mixed population of both individuals with type 1 diabetes and individuals with type 2 diabetes.Furthermore, previous reviews could not conclude on whether the effect of CGM was different in individuals treated with or without insulin, and most reviews did not investigate the effect of CGM use on the sensor-derived glycaemic parameters TIR, TBR and TAR [8].Also, no review evaluated the effect on the occurrence of severe hypoglycaemia or development of diabetes-related complications [6].
Therefore, the primary aim of this systematic review and meta-analysis was to give an up-to-date comprehensive overview of the effect of CGM (rtCGM or isCGM) compared with SMBG on glycaemic control, as quantified by HbA 1c , in adults with type 2 diabetes treated with or without insulin.Secondary aims were to evaluate the effect of CGM use compared with SMBG on TIR, TAR, TBR, glycaemic variability, incident severe hypoglycaemia and incident diabetesrelated micro-and macrovascular complications.

Methods
This systematic review was registered on PROSPERO (ID CRD42023418005) and is reported in accordance with the Preferred Reporting Items for Systematic Reviews and Metaanalysis (PRISMA) statement [13].The PRISMA checklist is provided in electronic supplementary material (ESM) Table 1.

Data sources and searches
We searched Embase, MEDLINE (via PubMed), Web of Science, Scopus and ClinicalTrials.gov for relevant articles using a combination of the terms 'diabetes mellitus type 2', 'continuous glucose monitoring' and 'HbA 1c ' from inception until 2 May 2023.Details of the search are provided in ESM Table 2. Additionally, we did a manual search by reviewing the reference lists of all relevant articles identified and prior reviews and meta-analyses to identify any remaining articles.Two reviewers (MJ and TACMV) independently performed screening of titles/ abstracts using Rayyan [14], and assessed the full texts for eligibility.Any discrepancies were discussed and resolved by a third reviewer (TTvS).

Study selection
Studies were eligible if they compared CGM (rtCGM or isCGM) to SMBG (or isCGM when rtCGM was the main intervention) and reported HbA 1c as an outcome measure.We included RCTs with a minimum intervention period of 6 weeks of consecutive or intermittent use of CGM among adults with type 2 diabetes (irrespective of diabetes treatment) in an outpatient setting.We excluded studies with pregnant women or individuals with type 1 diabetes, studies that investigated GlucoWatch [15] or a professional CGM (pCGM) device (e.g.Abbott Freestyle Libre Pro IQ or Dexcom G6 Pro) or an intervention that consisted of CGM combined with an additional glucose-lowering treatment strategy.
Data extraction and quality assessment Two reviewers (MJ and TACMV) independently extracted data from the included full-text articles using a standardised form.Disagreements were resolved by consensus or a third reviewer (TTvS).We extracted data on the study authors, year of publication, study design and follow-up duration, attrition rate, intervention type and duration, comparator type, inclusion and exclusion criteria, baseline characteristics (age, sex, diabetes duration and ethnicity), baseline insulin use and use of oral glucose-lowering drugs.In addition, we retrieved information on HbA 1c , TIR, TBR, TAR and glycaemic variability (defined as coefficient of variation [CV]) at baseline and at the endpoint.Finally, data on the incidence of severe hypoglycaemia (as defined in the original publication) and the incidence of microvascular complications (retinopathy, nephropathy and neuropathy) or macrovascular complications (myocardial infarction, cardiovascular death, cerebrovascular disease or peripheral artery disease) at the endpoint.Authors were contacted in case of any missing information.
We used the Cochrane risk-of-bias tool version 2 (RoB 2) to assess risk of bias of the included trials [16].This tool includes five domains: randomisation process, deviations from intended interventions, missing outcome data, measurement of the outcome and selection of the reported results.The quality of each RCT was assessed independently by two reviewers (MJ and TACMV).Any disagreements were resolved by consensus with a third reviewer (TTvS).Studies were rated as having high, moderate or low risk of bias.We labelled trials as low risk of bias if all five domains were scored as low risk of bias.

Data synthesis and analysis
The primary outcome was mean difference in HbA 1c (% and mmol/mol) from baseline to study end and corresponding 95% CI.Secondary outcomes were TIR (percentage of time glucose was between 3.9-10 mmol/l [70-180 mg/dl]), TBR (percentage of time glucose <3.9 mmol/l [<70 mg/dl]), TAR (percentage of time glucose was >10 mmol/l [>180 mg/dl]), glycaemic variability (CV [%]), incident severe hypoglycaemia (RR and 95% CI), incident microvascular (retinopathy, nephropathy and neuropathy) and macrovascular complications (myocardial infarction, cardiovascular death, cerebrovascular disease and peripheral artery disease).When TIR, TBR or TAR were described as hours and minutes, values were converted to percentage of time.For glycaemic outcomes we extracted the mean change between groups from baseline to endpoint and the SD.Incident severe hypoglycaemia and complications outcomes were analysed as RR and corresponding 95% CI.
A meta-analysis with a pooled estimates and random effects model was performed in Review Manager version 5.4 [17,18].A p value of <0.05 was considered statistically significant.Heterogeneity was assessed using I 2 and χ 2 .I 2 values of 0% to 40%, 30% to 60%, 50% to 90% and 75% to 100% were interpreted as low, moderate, substantial and considerable heterogeneity, respectively [18].Funnel plots were visually inspected and the Egger test was used to assess publication bias [19].Grading of Recommendations, Assessment, Development, and Evaluations (GRADE) was used to estimate the certainty of evidence [20].
For the primary outcome, we performed a predefined subgroup analysis considering CGM type (comparing rtCGM to isCGM, isCGM to SMBG, and rtCGM to SMBG).If possible, the main analysis was repeated stratified according to insulin use, median baseline HbA 1c , median age, median diabetes duration, median intervention duration, the presence of micro-or macrovascular complications at baseline, sex (male vs female) and background glucose-lowering therapy (insulin users, oral agents users only, and mix of insulin and oral agent users).

Risk of bias
Four trials [21,22,24,28] had an overall low risk of bias, whereas eight trials [23,[25][26][27][29][30][31][32] had some concerns (ESM Fig. 2).As for the domain of 'randomisation process', three trials [25,29,30] were graded with some concerns due to lack of information about the randomisation and allocation concealment process.All trials  were graded with low risk of bias for the domains 'deviation from the intended interventions', 'missing outcome data' and 'measurement of the outcome'.As for the domain of 'selection of the reported results', seven trials [23, 25-27, 29, 31, 32] were graded with some concerns due to missing trial protocols and statistical analysis plan.
Fig. 2 Forest plot of pooled analysis of change in HbA 1c (mmol/mol) in individuals with type 2 diabetes using rtCGM or isCGM compared with SMBG, stratified according to type of glucose-lowering therapy (insulin users, no insulin users or mixed population of insulin users and no insulin users) Nine studies [21-25, 28, 29, 31, 32] reported on severe hypoglycaemia, yet only three studies [21,23,28] reported one or more events in one of the groups and contributed to the meta-analysis.There was no statistically significant difference in the incidence of severe hypoglycaemia (RR 0.66, 95% CI 0.15, 3.00, p=0.57,I 2 =0%).In addition, based on three trials [21,23,28], there was no statistically significant difference in the incidence of macrovascular complications (RR 1.54, 95% CI 0.42, 5.72, p=0.52,I 2 =29%) in the CGM group compared with SMBG (ESM Fig. [9][10].No trials reported outcome data on microvascular complications. Using GRADE criteria, the outcomes TIR and TAR were graded with moderate certainty.The outcomes TBR, glycaemic variability and severe hypoglycaemia were graded with low certainty, whereas outcome macrovascular complications was graded with very low certainty.Outcomes were downgraded partly because of inconsistency and the low number of events resulting in imprecision (Table 2).

Discussion
This systematic review and meta-analysis on the effect of CGM use (rtCGM or isCGM) on glycaemic control in adults with type 2 diabetes showed a modest reduction of −3.43 mmol/mol (−0.31%) in HbA 1c .This effect was comparable among users of insulin and other oral agents.Furthermore, CGM was associated with a +6.36% increase in TIR and a decrease of −0.66% in TBR, −5.86% in TAR and −1.47% in glycaemic variability.
Our results are in accordance with previous systematic reviews and meta-analyses, which found a significant reduction in HbA 1c ranging from −7.65 mmol/mol (−0.70%) to −2.73 mmol/mol (−0.25%) in individuals with type 2 diabetes [6][7][8][9][10]12].Moreover, our result is comparable with a systematic review showing a reduction of −2.46 mmol/mol (0.23%) in HbA 1c with CGM use compared with SMBG in individuals with type 1 diabetes [33].Our review extends previous reviews [6][7][8][9][10][11] because we include an additional six RCTs including 589 participants, which allowed us to calculate a more precise effect size with higher statistical power.This also allowed us to evaluate the effect of CGM use in users of insulin and oral agents, according to CGM type (rtCGM and isCGM) and in relevant subgroups.
We found a reduction in HbA 1c that was comparable between studies including both users of insulin and other oral agents.Current guidelines suggest CGM as a therapy strategy only in individuals with type 2 diabetes who use insulin [1].Our findings, however, might support the efficacy of CGM in individuals with type 2 diabetes, irrespective of glucose-lowering therapy.CGM use may improve dosing of any glucose-lowering therapy (insulin and other oral agents) and/or stimulate a healthy lifestyle, and this may explain its beneficial effects on glycaemic control compared with SMBG [1].
In our analyses, we found a trend towards a larger reduction in HbA 1c in studies investigating rtCGM rather than those investigating isCGM.This is in accordance with a previous systematic review that reported, in a subgroup analysis, a non-significant change in HbA 1c in both type 1 diabetes or type 2 diabetes [11].This finding might suggest that realtime techniques might provide additional benefit compared with techniques requiring intermittent scanning.However, no study directly compared rtCGM to isCGM in type 2 diabetes and this issue, therefore, requires further study.
The mean reduction in HbA 1c of −3.43 mmol/mol (−0.31%) was relatively modest, but the effect was consistent across all included studies, indicating the robustness of the study findings.Furthermore, the effect was consistent across studies with younger and older individuals, short and long diabetes duration and higher or lower HbA 1c at baseline.Also, consistent beneficial effects of CGM were found on other markers of glycaemic control, i.e.TIR, TBR, TAR and glycaemic variability [1].The beneficial effect on TIR (+6.36%) was more than the current consensus of 5% minimal clinical relevant difference in TIR [34].
We found a non-significant decrease in the incidence of severe hypoglycaemia.This was, however, based on only three studies with a small number of events (eight events in total).Most studies that assessed severe hypoglycaemia reported no events in both groups.Therefore, we likely had insufficient power to detect a difference in incidence of severe hypoglycaemia.Furthermore, only three trials assessed macrovascular complications, with most events detected in one study [21].This study was performed in people who had experienced a recent myocardial infarction.Thus, our aggregated results for macrovascular complications have limited generalisability and should be interpreted with caution.A reduction in HbA 1c would likely translate to lower rates of diabetes-related complications in the long term, but this cannot be substantiated in this data.
Strengths of our analysis include the comprehensive overview of the effect of both rtCGM and isCGM on glycaemic control in adults with type 2 diabetes with different glucoselowering therapies, and the analysis of multiple relevant glycaemic outcome parameters.In addition, our study findings are consistent with, and extend the results of, a recent metaanalysis [12].We did a more recent search and identified one additional RCT [21].Furthermore, we were able to do a large range of prespecified subgroup analyses.Limitations include the fact that we did not include quality of life as outcome.However, previous trials have demonstrated that CGM use in type 2 diabetes is associated with beneficial effects on quality of life compared with SMBG [25,32].In contrast to studies done in individuals with type 1 diabetes [33] we did not find different effects of glucose sensor use according to baseline HbA 1c levels.However, we may not have had sufficient power to detect any differences related to baseline HbA 1c , because all studies except one [24] had a baseline HbA 1c value of 64 mmol/mol (8%) or higher.In addition, all RCTs were open label, the study duration of included RCTs was relatively short (maximum 52 weeks) and no or only limited data were available on incident severe hypoglycaemia and incident microvascular and macrovascular complications.
In conclusion, this systematic review and meta-analysis showed an improvement in HbA 1c and other parameters of glycaemic control related to CGM use (rtCGM or isCGM) in adults with type 2 diabetes.Future studies are needed to compare the effect on glycaemic control of rtCGM to isCGM and assess the effect of CGM use on incident microand macrovascular complications.

Fig. 1
Fig.1 Forest plot of pooled analysis of change in HbA 1c (mmol/mol) in individuals with type 2 diabetes using rtCGM or isCGM compared with SMBG

Fig. 3
Fig. 3 Forest plot of pooled analysis of change in TIR (a) TBR (b) and TAR (c) in individuals with type 2 diabetes using rtCGM or isCGM compared with SMBG
Characteristics of included trialsThe 12 RCTs were published between 2008 and 2023 and included a total of 1248 participants.The sample size ranged from 25 to 224 participants (Table

Table 2
Determining the certainty of evidence of outcomes using GRADE criteria