Continuous glucose monitoring in type 2 diabetes: Evidence for glycaemic metrics and patient‐reported outcomes

Abstract

Aims

In adults with type 2 diabetes (T2D), continuous glucose monitoring (CGM) provides insights beyond HbA1c. A comprehensive understanding of clinical benefits and patient‐reported outcomes (PROs) associated with CGM is needed to inform its integration into routine care.

Methods

A systematic literature search was conducted across PubMed, Scopus, Cochrane Library and CINAHL from inception to June 2025 to identify randomised controlled trials and non‐randomised studies including quasi‐experimental and observational designs. Outcomes of interest included glycaemic measures (HbA1c, time in range, hypoglycaemia, glycaemic variability) and PROs. Given the heterogeneity in study design and outcome reporting, findings were synthesised narratively.

Results

Thirty studies (N ≈ 3275) were included. CGM use was associated with improvement in HbA1c, particularly among insulin‐treated individuals and those with higher baseline HbA1c. CGM use also increased time in range by approximately 9–15 percentage points (1.5–3.6 h/day) and was generally associated with reduced hypoglycaemia risk and glycaemic variability, although effects on hypoglycaemia were not uniform. In contrast, diabetes‐specific PROs, including treatment satisfaction and diabetes‐related distress, showed more consistent improvement, whereas changes in overall quality‐of‐life measures were less pronounced.

Conclusions

CGM use was associated with clinically meaningful improvements in glycaemic metrics and PROs in adults with T2D. Although evidence remains limited for long‐term outcomes and special populations, current findings support the clinical value of CGM in diabetes management.

What's new?

CGM improves glycaemic control in type 1 diabetes, but its clinical impact and influence on patient‐reported outcomes in T2D are less robustly synthesised.

CGM use facilitates a reduction in HbA1c and an increase in time in range, though improvements are contingent upon continuous use, device generation and modality. Furthermore, CGM consistently enhances diabetes‐specific PROs, including treatment satisfaction and reduced distress, across diverse T2D populations.

Findings support the integration of CGM into routine T2D care to enhance patient‐centred management, though long‐term cost and sustainability require further evaluation.

1. INTRODUCTION

Type 2 diabetes (T2D) accounts for approximately 90–95% of diabetes cases worldwide, with diagnoses increasingly reported among younger populations. ¹ Identifying effective technologies for diabetes management is therefore crucial to support optimal glycaemic control and reduce the risk of diabetes‐related microvascular and macrovascular complications. ² , ³ Self monitoring of blood glucose (SMBG) using intermittent finger‐prick blood tests has long been the standard approach to guide therapeutic decisions. ⁴ However, SMBG provides only limited intermittent snapshots of glucose levels and cannot adequately capture glycaemic fluctuations such as nocturnal hypoglycaemia and postprandial hyperglycaemia. ⁵

In contrast, continuous glucose monitoring (CGM), including real‐time (rtCGM) and intermittently scanned CGM (isCGM) or flash CGM, provides continuous insight into glucose trends, supports timely therapeutic adjustments and enhances participant engagement. ⁶ , ⁷ While CGM has been widely adopted in type 1 diabetes management, its use in T2D, particularly among individuals not receiving intensive insulin regimens, has gained increasing interest. ⁴ Clinical studies have suggested that both rtCGM and isCGM can reduce glycated haemoglobin (HbA1c) in adults with T2D, including those not treated with insulin, ⁸ in both short‐ and longer‐term studies. ⁹ In addition to HbA1c reduction, CGM use has been associated with improvements in time in range (TIR) and time above range (TAR), ¹⁰ , ¹¹ and may support dietary and lifestyle modifications by increasing awareness of glucose responses to daily behaviours. ⁶ , ⁷

Importantly, CGM use is also associated with improvements in patient‐reported outcomes (PROs), including treatment satisfaction, diabetes‐related distress and confidence in self management, partly through reducing the burden of finger‐stick testing. ¹² , ¹³ Individuals with T2D face not only physical complications but also psychological distress, which can reduce quality of life. ¹⁴ , ¹⁵ Consistent improvements in treatment satisfaction, lower distress and fear of hypoglycaemia, and higher confidence have been reported among both insulin‐treated and non‐insulin‐treated T2D individuals using CGM. ¹⁵ , ¹⁶ Reflecting these evolving findings, ADA Standards of Care now recommend consideration of the use of CGM at diabetes onset and anytime thereafter for children, adolescents and adults with diabetes receiving insulin therapy, therapies that may cause hypoglycaemia or any treatment where CGM can assist in glycaemic management. ¹⁷

Although the evidence is well documented in T1D, it is less robust in T2D, especially in real‐world settings. Furthermore, most systematic reviews of CGM focus primarily on HbA1c as the indicator of effectiveness, ¹³ , ¹⁸ with limited attention to PROs. Evidence on the emotional, behavioural and quality of life domains critical for adherence to diabetes technologies and long‐term self management remains mixed. ¹⁹ , ²⁰ This systematic review aims to examine the effects of CGM on HbA1c, TIR and glycaemic variability, alongside PROs such as treatment satisfaction. By including observational studies alongside clinical trials, we aim to provide a comprehensive and clinically meaningful assessment of CGM use in adults with T2D, supporting a more balanced interpretation of its role beyond HbA1c alone.

2. METHODS

This systematic review followed the Preferred Reporting Items for Systematic Reviews and Meta‐analyses (PRISMA) 2020. ²¹ This review was not prospectively registered on a public database. However, a predefined protocol outlining the search strategy and eligibility criteria was established prior to study selection and data extraction to ensure methodological consistency. Additionally, a narrative synthesis was conducted due to variation across outcomes assessed, study designs and findings.

2.1. Databases search and search strategy

Two reviewers conducted a comprehensive literature search in PubMed, Scopus, Cochrane Library and CINAHL from inception to June 2025. In each database, controlled vocabulary was combined (e.g. MeSH, CINAHL headings) with free‐text keywords to capture four main concepts: type 2 diabetes, continuous glucose monitoring, glycaemic control and PROs. Boolean operators (AND, OR) and database‐specific syntax were used to link terms. The full search strategies for each database are provided in Table S1. Grey literature sources like conference proceedings, government reports and dissertations were also screened using the same key terms to ensure comprehensive coverage.

2.2. Inclusion and exclusion of studies

All randomised controlled trials (RCTs) and non‐randomised studies, including quasi‐experimental or observational studies that assessed the effects of CGM, rtCGM or isCGM in the adult population with T2D were included. While this review explicitly targets the T2D population, several included studies utilised mixed cohorts comprising individuals with T1D. Where subgroup data were available, outcomes specific to participants with T2D were extracted and used in the synthesis. In studies where subgroup data were not separately reported, findings were interpreted cautiously in relation to the T2D population. Eligible studies reported at least one of the following clinical outcomes: HbA1c, hypoglycaemia, hyperglycaemia, time in range or PROs (e.g. satisfaction, quality of life). Only studies published in English were included. We excluded studies that reported CGM effects on children and adolescents, as well as case reports, editorials and reviews.

2.3. Screening and data extraction

Two reviewers independently screened titles, abstracts and full texts using Rayyan software and extracted data to minimise bias and errors. Data were extracted using a predefined form covering study title, authors, study design, region/setting, participants, intervention, outcomes, key findings and limitations. Inter‐rater reliability between the two reviewers (AWP and WLC) was calculated using Cohen's kappa coefficient. There was near‐perfect agreement between the two independent reviewers (Cohen's kappa = 0.92). Any disagreements were resolved through discussion or consultation with a third reviewer (HZH).

2.4. Quality assessment

Methodological quality was independently assessed by two reviewers using study design‐specific tools, with discrepancies resolved through consensus. RCTs were evaluated with the Cochrane Risk of Bias 2.0 (RoB 2) tool ²² appraising five domains: randomisation process, deviations from intended interventions, missing outcome data, outcome measurement and selection of reported results. The methodological quality of non‐randomised studies was assessed using design‐specific Joanna Briggs Institute (JBI) Critical Appraisal Checklists to ensure robust internal validity. Quasi‐experimental studies ²³ were evaluated across nine domains, prioritising causal clarity, participant similarity, the comparability of control groups and the reliability of multiple outcome measurements. Diagnostic test accuracy (DTA) ²⁴ studies were appraised using a 10‐item tool that evaluated the representativeness of the participants' spectrum, the validity of reference standards, blinding protocols for both index and reference tests and the clear definition of cut‐offs. Finally, the cohort study ²⁵ was assessed against 11 criteria, focusing on recruitment strategies, valid exposure measurement, the identification and management of confounding factors and the completeness of follow‐up.

3. RESULTS

The systematic search identified 1038 records. After removing 405 duplicates, 629 unique records underwent initial title and abstract screening. Of these, 508 records were excluded at the screening phase because they clearly did not meet the inclusion criteria (e.g. unrelated to diabetes/CGM). The remaining 121 reports were sought for full‐text retrieval, resulting in 117 reports assessed for eligibility. Specific reasons for the exclusion of 87 full‐text reports are detailed in Figure 1.

FIGURE 1.

PRISMA flow diagram of selection of studies.

3.1. Study characteristics

This systematic review synthesised evidence from 30 studies presented in Table 1, 18 RCTs and 12 non‐randomised studies assessing CGM efficacy in T2D or mixed individuals with diabetes including open‐label and pragmatic designs and compared real‐time or isCGM with traditional SMBG. Non‐randomised studies included pre‐ and post‐interventions, DTA and a cohort study. These studies were conducted in 12 countries, including the United States, Saudi Arabia, Australia, Japan, Singapore, Denmark, Canada, China, France, Israel, the United Kingdom and Germany, with a total of 3275 participants aged 50–65 years with poorly controlled T2D. RCTs' sample sizes ranged from 40 to 404 participants, with follow‐up from 6 weeks to 18 months. Quasi/observational studies enrolled 10–174 participants with 2 weeks to 6 months of follow‐up. Across all studies, the participants' age group and gender did not affect CGM use outcomes. Note that we only analysed outcomes related to the T2D population. Most studies included participants with a baseline 53–108 mmol/mol (7.0–12.0%). Interventions comprised rtCGM (n = 14) or isCGM (n = 16) compared with SMBG or usual care. Variations included episodic use, ²⁶ integrated pump systems, ²⁷ and telehealth support. ²⁸ Study populations included individuals on basal insulin, ⁷ basal‐bolus, ²⁹ insulin pump, ³⁰ non‐insulin agents ²⁸ and special subgroups post‐myocardial infarction, ³¹ dialysis ³² and gastroparesis. ³³ Among all studies 17 focused exclusively on T2D, 13 studies utilised mixed T1D/T2D cohorts (Figure S1). Among mixed cohort studies, 10 out of 13 reported isolated results for the T2D cohort. ¹⁵ , ³² , ³³ , ³⁴ , ³⁵ , ³⁶ , ³⁷ , ³⁸ , ³⁹ , ⁴⁰ Whereas three studies ⁴¹ , ⁴² , ⁴³ focused on specialised settings like pregnancy and acute COVID‐19 wards and reported pooled data for both cohorts.

TABLE 1.

Characteristics of the included studies.

Study title	Author (year)	Study design	Region/setting	Participants	Interventions	Outcomes	Limitations
Effect of Continuous Glucose Monitoring on Glycaemic Control in Patients With Type 2 Diabetes Treated With Basal Insulin (MOBILE Study)	Martens et al. (2021)	Multicentre randomised clinical trial (open‐label), 8 months	15 primary care centres, United States (2018–2020)	175 adults (mean age 57 years) with poorly controlled T2D on basal insulin‐only HbA1c 76 mmol/mol (9.1%)	CGM (Dexcom G6) vs. traditional self monitoring with blood‐glucose meter (2:1 randomisation)	Primary: HbA1c at 8 months; Secondary: time in range 70–180 mg/dL, time >250 mg/dL, mean glucose, severe hypoglycaemia	Open label; 8‐month duration; US primary‐care sample; few severe hypoglycaemia events; results may not generalise to all populations
The Impact of Flash Glucose Monitoring on Markers of Glycaemic Control and Patient Satisfaction in Type 2 Diabetes	Al Hayek et al. (2021)	Single‐arm prospective observational (12 weeks)	Prince Sultan Military Medical City, Riyadh, Saudi Arabia	54 adults with T2D on multiple daily insulin injections, HbA1c ≥ 53 mmol/mol (≥7.0%)	Initiation of isCGM (FreeStyle Libre)	HbA1c, hypoglycaemia episodes/month, scan frequency, DTSQ & GMSS satisfaction scores	No control group; single centre; small sample; short 12‐week follow‐up
The role of continuous glucose monitoring in clinical decision‐making in diabetes in pregnancy	McLachlan et al. (2007)	Prospective observational evaluation of 72‐h CGMS traces	Werribee Mercy Hospital, Victoria, Australia	55 pregnant women (37 GDM, 10 T2D, 8 T1D)	72‐h Medtronic CGMS during routine antenatal care	Proportion of traces altering management; patient tolerability; device accuracy vs. finger‐prick glucose	Observational, no control group; small single‐site sample; lacked long‐term maternal/fetal outcomes
Effect of the FreeStyle Libre flash glucose monitoring system on glycaemic control in individuals with type 2 diabetes treated with basal–bolus insulin therapy	Ogawa et al. (2021)	Prospective open‐label multicentre single‐arm trial (90 days)	Five diabetes centres, Japan	94 adults with T2D on basal–bolus insulin therapy	14‐day masked baseline then 11‐week open use of FreeStyle Libre	Primary: time <70 mg/dL; Secondary: time in range, hyperglycaemia, est. A1c, finger‐stick frequency, DTSQ score	Single‐arm; short duration; low baseline hypoglycaemia; open‐label self management; no control comparator
Continuous Glucose Monitoring Feedback in the Subsequent Development of Gestational Diabetes (I‐PROFILE)	Quah et al. (2024)	Prospective single‐centre pilot RCT	KK Women's & Children's Hospital, Singapore	206 first‐trimester pregnant women without pre‐existing diabetes	Unblinded isCGM vs. masked CGM worn at 3 gestational periods	Primary: GDM incidence via 75 g OGTT; Secondary: OGTT glucose levels, glycaemic variability	Pilot, single centre, lack of personalised education, limited power
The Influence of Real‐Time CGM on Psychosocial Outcomes in Insulin‐Using Type 2 Diabetes	Soriano & Polonsky (2023)	6‐month prospective before‐after study	Scripps Whittier Diabetes Institute, USA	174 adults with insulin‐treated T2D	Initiation of Dexcom G5 rtCGM (no control arm)	Diabetes distress, hypoglycaemia fear & confidence, emotional burden, trust; HbA1c	No control group, older CGM requiring calibration, mainly non‐Hispanic White sample
Comparing CGM and Blood Glucose Monitoring in Adults With Inadequately Controlled, Insulin‐Treated T2D (Steno2tech)	Lind et al. (2024)	12‐month single‐centre open‐label RCT	Steno Diabetes Center Copenhagen, Denmark	76 adults with insulin‐treated T2D, HbA1c ≥58 mmol/mol (≥7.5%)	Dexcom G6 CGM vs. conventional BGM	Primary: Time‐in‐Range 3.9–10 mmol/L; Secondary: HbA1c, insulin dose, weight, PROs	Single‐centre, open‐label, small sample, COVID‐related recruitment delays
Effectiveness of CGM with Remote Telemonitoring‐Enabled Virtual Educator Visits in Non‐Insulin T2D	Lau et al. (2024)	Open‐label randomised trial (6‐week intervention, 12‐week follow‐up)	University of Alberta, Canada	105 adults with T2D (not on insulin) HbA1c >53 mmol/mol (>7.0%)	FreeStyle Libre 2 CGM + telemonitoring & two virtual education visits vs. enhanced usual care	Change in HbA1c at 12 weeks; weight; treatment satisfaction	Short duration, single‐centre, open‐label, small sample
GP‐OSMOTIC: Quarterly Professional‐Mode Flash CGM in Adults With T2D in General Practice	Speight et al. (2021)	Pragmatic open‐label 12‐month RCT (secondary psychological analysis)	25 general practices, Victoria, Australia	299 adults with T2D	Professional‐mode isCGM (FreeStyle Libre Pro) every 3 months vs. usual care	WHO‐5 well‐being, diabetes‐specific quality of life, monitoring satisfaction, self care, PICS at 12 months	Secondary outcomes; open label; possible intervention fidelity issues
Personal CGM With Professional Support in Insulin‐Treated Diabetes, Hong Kong	Wong (2023)	Single‐centre retrospective cohort with propensity‐matched controls	Ruttonjee Hospital outpatient clinic, Hong Kong	90 CGM users vs. 90 usual‐care controls (insulin‐treated T1D/T2D)	10–14‐day rtCGM or isCGM plus physician/nurse feedback	Change in HbA1c (12–20 weeks); clinical interventions logged	Retrospective, small, single‐centre, insulin users only, potential selection bias
Feasibility of Real‐Time Continuous Glucose Monitoring Telemetry System in an Inpatient Diabetes Unit	Dillmann et al. (2022)	Prospective observational single‐centre pilot study	University Hospital of Strasbourg, France	53 insulin‐requiring inpatients (28 T1D, 25 T2D)	Guardian Connect rtCGM with remote telemetry during admission	TIR (70–180 mg/dL), TAR, TBR from admission start to discharge	Small sample, single‐centre, short hospital stay, no control arm
Flash Glucose Monitoring Combined With Insulin Pump in Type 2 Diabetes	Wang et al. (2021)	Randomised controlled trial (n = 80)	Second Affiliated Hospital, Nanjing Medical University, China	40 intervention vs. 40 control, T2D aged ~72 years on insulin pumps	Insulin pump + is CGM (FreeStyle Libre) vs. pump + finger‐stick SMBG	Glycaemic indices, insulin dose, hypoglycaemia, comfort, mental health, quality of life	Single‐centre, elderly cohort, short follow‐up, open label
CGM Versus Usual Care in Type 2 Diabetes on Multiple Daily Injections (DIAMOND)	Beck et al. (2017)	24‐week multicentre open‐label RCT	25 endocrinology practices, USA & Canada	158 adults (mean age 60 years) with T2D, HbA1c 58–85 mmol/mol (7.5–9.9%)	Dexcom G4 rtCGM vs. SMBG	Primary: HbA1c change; secondary: hypoglycaemia, quality of life	Six‐month horizon, unblinded, technology dated
Effect of Flash Glucose Monitoring Technology on Glycaemic Control and Treatment Satisfaction in Patients With Type 2 Diabetes	Yaron et al. (2019)	Open‐label randomised controlled trial, 10 weeks	Two diabetes centres, Israel	101 adults with T2D on MDI ≥1 year	isCGM (FreeStyle Libre) monitoring vs. SMBG with insulin‐dose support	DTSQ (primary), HbA1c, hypoglycaemia, quality of life	Short duration, open‐label, single device; moderate sample size
Self‐Monitoring Using CGM with Real‐Time Feedback Improves Exercise Adherence	Bailey et al. (2016)	8‐week pilot RCT with 1‐month follow‐up	Community exercise programme, British Columbia, Canada	13 adults with prediabetes/T2D (6 CGM‐SM, 7 control)	Group‐mediated intervention teaching self monitoring via rtCGM vs. standard exercise education	Self monitoring behaviour, goal‐setting, attendance, re‐enrolment	Very small, single‐site, quasi‐experimental allocation, self‐reported PA
Effect of Flash Glucose Monitoring on Mental Well‐being & Treatment Satisfaction in Japanese People with Diabetes	Mitsuishi et al. (2017)	Pre–post single‐arm study (14 days)	Five centres, Japan	80 insulin‐treated adults (57 T1D, 23 T2D)	FreeStyle Libre worn for 14 days	WHO‐5 well‐being index, DTSQ scores	No control group, very short exposure, subjective outcomes
Glucose Sensor‐Augmented CSII in Diabetic Gastroparesis (GLUMIT‐DG)	Calles‐Escandón et al. (2018)	24‐week open‐label prospective multicentre pilot, non rct	7 NIDDK Gastroparesis Consortium sites, USA	45 adults (T1D/T2D) with HbA1c >64 mmol/mol (>8.0%) and scintigraphy‐confirmed gastroparesis	MiniMed Paradigm CSII + real‐time CGM with algorithm‐based titration	Safety (hypoglycaemia), CGM metrics, HbA1c, GCSI, PAGI‐QOL, meal tolerance	Non‐randomised, no control, complex protocol, specialised population
CGM + Problem‐Solving Counselling to Change Physical Activity in Women with T2D	Allen et al. (2011)	12‐week pilot RCT	Massachusetts, USA (health‐system recruitment)	29 women with T2D (14 CGM + PS, 15 CGM + Education)	Single CGM feedback session plus 90‐min problem‐solving vs. CGM plus general education	PA, diet adherence, problem‐solving skills, HbA1c, weight	Small, under‐powered, short, both arms received CGM, so metabolic differences are minimal
MITRE Trial: Minimally Invasive Glucose Monitoring vs. Conventional Care in Insulin‐Treated Diabetes	Newman et al. (2009)	Four‐arm pragmatic RCT (18 months)	Eight UK hospitals	404 adults with poorly controlled insulin‐treated T1D/T2D	GlucoWatch Biographer or MiniMed CGMS vs. attention control vs. usual care	Primary: HbA1c at 3, 6, 12, 18 months; acceptability; health economics	Older first‐generation devices, usability issues, high attrition, modest nurse feedback impact
Effectiveness of Continuous Glucose Monitoring in Dialysis Patients with Diabetes: The DIALYDIAB Pilot Study	Joubert et al. (2015)	Before–after monocentric pilot (12 weeks: 6 w SMBG → 6 w iterative CGM)	University Hospital of Caen dialysis unit, France	15 adults on chronic haemodialysis with diabetes (mean 61 years; 20% diet only, 80% insulin)	3–6 SMBG/day for 6 w followed by 5‐day blinded CGM every 2 w with remote physician counselling	Mean CGM glucose, AUC >10 mmol/L, AUC <3.3 mmol/L, treatment‐change frequency	Small, single‐centre, no control, short duration; the remote‐management model may limit generalisability
Flash Glucose Monitoring Helps Achieve Better Glycaemic Control Than SMBG in Non‐Insulin‐Treated Type 2 Diabetes	Wada et al. (2020)	Multicenter open‐label RCT, 24 weeks (12 w monitoring +12 w follow‐up)	Five hospitals in Japan	100 adults with non‐insulin T2D, baseline HbA1c 58–69 mmol/mol (7.5–8.5%)	FreeStyle Libre for 12 w vs. SMBG	Δ HbA1c (primary), DTSQ, mean glucose, GV indices, hyperglycaemia time	Open label; modest sample; monitoring limited to first 12 w
Episodic Real‐Time CGM Use in Adults with Type 2 Diabetes: Pilot Randomised Trial	Price et al. (2021)	Multicenter pilot RCT, 12 w intervention +9 m follow‐up	8 sites, North America	70 adults with T2D on ≥2 non‐insulin agents (baseline HbA1c 62–91 mmol/mol) (7.8–10.5%)	Three 10‐day unblinded rtCGM sessions vs. SMBG	HbA1c, time‐in‐range (TIR), % achieving HbA1c <7.5%	Pilot size; lifestyle‐only; not powered for HbA1c; benefits not durable
LIBERATES: isCGM vs. SMBG After Myocardial Infarction in Type 2 Diabetes	Ajjan et al. (2023)	Multicenter open‐label RCT, 3 months	8 UK hospitals	141 insulin‐/SU‐treated T2D with recent MI (median age 63 years)	isCGM (FreeStyle Libre) vs. SMBG	Primary TIR days 76–90; hypoglycaemia, HbA1c, quality of life, cost‐effectiveness	Marginal TIR effect; open label; 3‐month horizon; unclear long‐term benefits
Short‐ and Long‐Term Effects of Real‐Time CGM in Type 2 Diabetes (52‐Week RCT)	Vigersky et al. (2012)	Randomised trial: 12 weeks intermittent RT‐CGM then 40 weeks follow‐up	Walter Reed Health Care System, USA	100 adults with T2D not on prandial insulin (HbA1c 53–108 mmol/mol) (7–12%)	2‐w‐on/1‐w‐off RT‐CGM cycles vs. SMBG	A1C at 12,24,38,52 weeks; medication intensification	Single health system military cohort; older sensor tech; open label
A randomised, controlled cross‐over study of short‐term CGM using the Dexcom G6 to support self‐management behaviour in complex type 2 diabetes (DISCO GM)	Parsons et al. (2024)	Single‐centre 36‐week randomised crossover (12 weeks usual care ↔ 12 weeks Dexcom G6) with 12 weeks follow‐up	Specialist diabetes clinic, Swansea, UK	51 adults with T2D ≥1 year, HbA1c ≥75 mmol/mol(≥9.0%); 37 (73%) completed	Dexcom G6 real‐time CGM plus education vs. usual specialist care	Diabetes Self‐Management Questionnaire (DSMQ) total & sub‐scores	Single site; behaviour self report; 27% attrition; no biochemical endpoints
Continuous glucose monitoring in people with diabetes: the randomised controlled Glucose Level Awareness in Diabetes Study (GLADIS)	New et al. (2015)	100‐day multicentre RCT: CGM ± alarms vs. SMBG	4 centres in the UK & Germany	145 T1/T2 on MDI or CSII	FreeStyle Navigator with alarms, without alarms or SMBG only	Primary: time outside 3.9–10 mmol/L days 80–100; HbA1c; hypoglycaemia	Alarm fatigue; modest T2D subgroup; no long‐term follow‐up
The effect of real‐time continuous glucose monitoring in pregnant women with diabetes	Secher et al. (2013)	Randomised controlled trial, intermittent rtCGM vs. routine care	Rigshospitalet, Copenhagen, Denmark	123 T1D & 31 T2D pregnancies (<14 weeks gestation)	6‐day rtCGM at 8, 12, 21, 27, 33 weeks gestation plus SMBG vs. SMBG alone	Primary: large‐for‐gestational‐age (LGA) infants; HbA1c; severe hypoglycaemia	Only 64% used CGM per protocol; intermittent use; single centre
Feasibility and acceptability of Libre Flash™ monitoring in insulin‐requiring unstable diabetes with complex needs	Bevelander et al. (2024)	6‐week observational feasibility study	St Vincent's Hospital, Melbourne, Australia	38 adults (87% T2D) with unstable diabetes & psychosocial complexity	isCGM (Libre Flash CGM) with weekly DNE support	Device scans/day, time‐in‐range (TIR), HbA1c, user questionnaire	No baseline TIR; small, uncontrolled; intensive staff input required
Evaluation of the performance and usability of a novel SiJoy GS1 continuous glucose monitoring system	Yan et al. (2023)	Multicentre performance trial, 14‐day wear	3 tertiary hospitals, China	70 adults (56% T1D, 35% T2D)	SiJoy GS1 CGM sensors (two per participant)	20/20% accuracy, Clarke/Consensus A + B, MARD, usability, safety	Short duration; clinical not real‐world; no comparator CGM
Practical implementation of remote real‐time CGM in hospitalised patients with diabetes (COVID‐19 cohort)	Hellman et al. (2022)	Retrospective observational evaluation of accuracy	North Kansas City Hospital, USA (non‐ICU COVID‐19 ward)	10 adult inpatients (9 T2D, 1 T1D) on insulin	Dexcom G6 rtCGM streaming to central station; POC BG for therapy	MARD vs. POC; Clarke/Surveillance error grids; hypothetical insulin‐dose discordance	Small pilot; single community hospital; limited to COVID‐19 cohort

3.2. Risk of bias

3.2.1. RCTS

In total, 18 RCT trials were assessed using the RoB 2 tool presented in Figure 2. Three RCTs ⁷ , ³⁴ , ³⁵ were ‘low‐risk’, while seven RCTs ²⁶ , ³¹ , ³⁶ , ³⁷ , ³⁸ , ³⁹ , ⁴⁰ had ‘some concerns’ and eight RCTs ²⁸ , ³⁰ , ⁴¹ , ⁴² , ⁴³ , ⁴⁴ , ⁴⁵ , ⁴⁶ were ‘high risk’. Most studies were low risk for randomisation, while nearly all showed ‘some concerns’ for deviations from intended intervention, due to limited blinding and protocol departures. Missing‐data handling and outcome measurement were acceptable in most cases (13 and 12 studies low risk each), though five trials were high risk from attrition. Selective reporting was least problematic (13 low risk). Overall, high bias arose from protocol deviations and attrition, whereas ‘some concerns’ usually reflected minor single‐domain issues.

FIGURE 2.

RoB 2 quality appraisal of RCTs.

3.2.2. Non‐randomised studies

Due to heterogeneity in study designs, the risk of bias (presented in Table 2) was evaluated across three distinct study designs: quasi‐experimental, DTA and cohort studies using appropriate JBI tools. ²³ , ²⁴ , ²⁵ Within the quasi‐experimental category, all studies were classified as moderate risk ¹⁵ , ²⁹ , ³² , ³³ , ⁴⁷ , ⁴⁸ , ⁴⁹ , ⁵⁰ primarily due to the absence of a control group, though some also lacked multiple measurements. The DTA designs ⁵¹ , ⁵² were considered low risk, whereas the study ⁵³ was considered a moderate risk. Even though these studies primarily focused on CGM accuracy against SMBG or venous blood monitoring, their secondary outcomes evaluated the clinical value of CGM regarding therapeutic decision‐making and participant safety. However, their designs relied on validating device performance against a reference standard. Therefore, to ensure robust internal validity, we applied the JBI DTA checklist. Additionally, a single cohort study ⁵⁴ was classified as low risk, satisfying nine criteria, including the use of a control group.

TABLE 2.

Risk of Bias of non‐randomised studies.

Study	Q1	Q2	Q3	Q4	Q5	Q6	Q7	Q8	Q9	Q10	Q11	Overall
Quasi‐experimental
Ogawa et al. (2021)	✓	✗	✓	✓	✓	✓	✓	✓	✓	—	—	Moderate
Mitsuishi et al. (2017)	✓	✗	✓	✓	✓	✓	✓	✓	✓	—	—	Moderate
Calles‐Escandon (2018)	✓	✗	✓	✓	✓	✓	✓	✓	✓	—	—	Moderate
Joubert et al. (2015)	✓	✗	✓	✓	✓	✓	✓	✓	✓	—	—	Moderate
Al Hayek (2021)	✓	✗	✓	✓	✓	✓	✓	✓	✓	—	—	Moderate
Soriano (2023)	✓	✗	✓	✓	✓	✓	✓	✓	✓	—	—	Moderate
Dillmann (2022)	✓	✗	✓	✗	✓	✓	✓	✓	✓	—	—	Moderate
Bevelander (2024)	✓	✗	✓	✗	✓	✓	✗	✓	✓	—	—	Moderate
Diagnostic accuracy
McLachlan et al. (2007)	✗	✓	?	✗	✓	✗	✓	✓	✓	✗	—	Low
Baker et al. (2022)	✗	✓	?	✓	✓	✗	✓	✓	✓	✓	—	Low
Yan et al. (2023)	✗	✓	✓	✓	✓	✓	✓	✓	✓	✓	—	Moderate
Cohort studies
Wong (2023)	✓	✓	✓	✓	✓	✓	✓	✓	✓	✗	✓	Low

Note: ✓ = Low Risk (Yes); ✗ = High Risk (No); ? = Unclear; — = Not Applicable.

3.3. Clinical outcomes

A range of glycaemic metrics is measured as the clinical outcomes of CGM. These outcomes can be categorised into four groups: (1) HbA1c change, TIR, hypoglycaemia and glycaemic variability.

3.3.1. HbA1c change

Across both RCTs and non‐randomised studies (n = 18), CGM was generally associated with reductions in HbA1c compared with SMBG, though results were not uniform. The impact of CGM is associated with device generation, CGM modality and population characteristics.

Several large RCTs utilising modern rtCGM (Dexcom G6, Dexcom G4) such as the MOBILE, Steno2tech, ⁷ , ³⁵ and DIAMOND ³⁴ studies demonstrated significant reductions in HbA1c, particularly in individuals treated with insulin and those with higher baseline HbA1c levels. For instance, the MOBILE trial reported a 12 mmol/mol (1.1%) reduction with rtCGM compared with 7 mmol/mol (0.6%) with SMBG. Similarly, the Steno2tech and DIAMOND trials reported 10 mmol/mol (0.9%) and 8 mmol/mol (0.7%) reductions respectively with rtCGM use. While isCGM use demonstrated generally lesser reductions in HbA1c 3–5 mmol/mol (0.3–0.5%) across studies, ⁴⁰ , ⁴⁵ , ⁵⁴ some studies found no significant between‐group differences in overall HbA1c reductions with isCGM use. ³¹ , ³⁸ , ⁵⁵ Moreover, early generation systems, such as the GlucoWatch Biographer and early MiniMed devices evaluated in the MITRE trial, ³⁶ suffered from high discontinuation rates and failed to demonstrate long‐term HbA1c advantages over standard care. These devices are no longer in use. In pregnancy, however, HbA1c levels did not differ significantly between rtCGM and SMBG users. ³⁷

3.3.2. Time in range

CGM use was generally associated with clinically meaningful improvements in TIR across both randomised and non‐randomised studies (n = 16), with rtCGM trials reporting TIR increases of up to 15.2%, equating to approximately 3.6 to 3.8 additional hours per day in the target range. Specifically, the MOBILE trial ⁷ demonstrated a 15% adjusted difference in TIR (an increase to 59% from a 40% baseline) for rtCGM users compared to standard blood glucose meter users, while the Steno2tech trial ³⁵ reported a 15.2% greater increase in TIR, translating to exactly 3 h and 39 min of additional time in range.

Moreover, the use also resulted in substantial benefits, with one study ⁴⁰ demonstrating that users gained over 2 h a day in TIR (increasing from 15.00 to 17.37 h) compared to standard self monitoring. Observational studies support these findings with moderate TIR gains, such as a significant increase from 75.7% to 82.2% observed in the general ward in individuals with type 2 diabetes. ⁴⁹ However, some interventions report no significant TIR change; for example, a feasibility study evaluating Libre Flash ⁴⁸ in populations with complex needs found no statistically significant differences in TIR over time despite participants achieving a significant 1.1% reduction in HbA1c (Table 3).

TABLE 3.

Cross‐study comparison of HbA1c and TIR.

Outcome	rtCGM	isCGM	Interpretation
HbA1c	~8–12 mmol/mol (~0.7–1.1%)	~3–5 mmol/mol (~0.3–0.5%)	rtCGM shows greater reductions
TIR	~15% or 3.6–3.8 h/day	~8% or ~2 h/day	rtCGM shows greater improvement
Evidence base	Consistent across major RCTs	Variable, some reported non‐significant findings	More robust for rtCGM

3.3.3. Hypoglycaemia

Most studies reported either no increase in time below range (TBR) or a reduction in hypoglycaemia. isCGM without alarms did not change time spent <3.9 mmol/L, whereas isCGM with alarms reduced hypoglycaemia duration from 1.6 h/day to 1.0 h/day. ⁴⁶ While one study reported zero episodes of hypoglycaemia in the flash CGM group, four episodes were observed in the SMBG group. ³⁰ Inpatients undergoing therapeutic education saw TBR fall from 3.4% to 0%. ⁴⁹

3.3.4. Glycaemic variability

Multiple studies reported reduced glycaemic variability ²⁹ , ³⁴ , ³⁵ reflected by decreases in glucose standard deviation and mean amplitude of glycaemic excursions (MAGE). CGM was associated with more consistent reductions in variability compared with SMBG. ⁴⁰ While mean glucose and time spent in hyperglycaemic ranges (>10, >13.3 and >16.7 mmol/L) decreased, coefficients of variation were often unchanged. ²⁹

3.3.5. Special populations

For the pregnant population, CGM altered clinical management in up to 62% of cases, ⁵² highlighting its potential role in supporting real‐time decision‐making in a physiologically insulin‐resistant state. In a cohort of pregnant women without pre‐existing diabetes, CGM‐derived metrics identified differences in early glycaemic variability and glucose patterns; however, CGM use did not result in consistent improvements in gestational diabetes outcomes. ⁴⁴ While this population does not directly reflect individuals with established type 2 diabetes, it provides contextual insight into CGM performance in metabolically dynamic conditions. These findings remain indirect and were not used to inform the primary conclusions of this review. Among individuals with end‐stage renal disease on dialysis, CGM facilitated more frequent treatment adjustments, resulting in improved mean glucose levels and reduced hyperglycaemia. ³² In individuals with diabetic gastroparesis, sensor‐augmented pump CGM improved TIR, reduced hypoglycaemia and improved gastroparesis symptoms and quality of life. ³³ In individuals with T2D following myocardial infarction, isCGM modestly increased TIR and reduced hypoglycaemia compared with SMBG. ³¹

3.4. Patient‐reported outcomes

A total of (n = 26) studies reported PROs alongside clinical benefits of CGM, including treatment satisfaction or usability (n = 12), psychosocial or quality of life (n = 10) and behavioural or adherence (n = 4) (Table 4). Among these studies, more than half (n = 14) reported statistically significant improvements in at least one domain. In contrast, (n = 6) studies reported no significant changes, and an additional (n = 6) provided descriptive findings only.

TABLE 4.

Patient‐reported outcomes across study design.

Domain	Study	Author (year)	Design	Intervention	PRO outcomes
Treatment satisfaction/usability	The Impact of Flash Glucose Monitoring on Markers of Glycaemic Control and Patient Satisfaction	Al Hayek et al. (2021)	Non‐randomised	DTSQ‐s, DTSQ‐c and GMSS	Significant: 2‐fold ↑ satisfaction (p < 0.001); ↓ burden & ↑ openness (all p < 0.001).
	Effect of the FreeStyle Libre flash glucose monitoring system on glycaemic control in individuals with type 2 diabetes treated with basal–bolus insulin therapy	Ogawa et al. (2021)	Non‐randomised	DTSQ‐c (Japanese Version)	Significant: ↑ satisfaction (p < 0.0001); ↑ convenience, flexibility, and recommendation (p < 0.001); ↑ hypo/hyper perception (p < 0.05).
	Personal CGM With Professional Support in Insulin‐Treated Diabetes, Hong Kong	Wong (2023)	Non‐randomised	GMSS	Descriptive: 71–90% reported ↑ satisfaction and understanding; no comparative p‐values.
	Feasibility of Real‐Time Continuous Glucose Monitoring Telemetry System in an Inpatient Diabetes Unit	Dillmann et al. (2022)	Non‐randomised	Satisfaction surveys	Descriptive: 94% found the system useful; 87% desired future use; no comparative p‐values.
	Feasibility and acceptability of Libre Flash™ monitoring in insulin‐requiring unstable diabetes with complex needs	Bevelander et al. (2024)	Non‐randomised	Acceptability survey	Descriptive: 80% rated easy; 97% would continue; 27% cited cost concerns; no p‐values.
	Evaluation of the performance and usability of a novel SiJoy GS1 continuous glucose monitoring system	Zhao et al. (2023)	Non‐randomised	18‐item usability questionnaire	Descriptive: High usability reported (mean score 86.6/90); no p‐values provided.
	Effectiveness of CGM with Remote Telemonitoring‐Enabled Virtual Educator Visits in Non‐Insulin T2D	Lau et al. (2024)	RCT	DTSQ‐s and DTSQ‐c	Significant: ↑ satisfaction vs. usual care (DTSQc diff +7.95, p < 0.001; DTSQs p = 0.01).
	Effect of Continuous Glucose Monitoring on Glycaemic Control in Patients With Type 2 Diabetes Treated With Basal Insulin (MOBILE Study)	Martens et al. (2021)	RCT	CGM satisfaction scale	Descriptive: High satisfaction reported (mean 4.1/5); no comparative p‐values.
	Effect of Flash Glucose Monitoring Technology on Glycaemic Control and Treatment Satisfaction in Patients With Type 2 Diabetes	Yaron et al. (2019)	RCT	DTSQ‐s and DTSQ‐c (Hebrew Version), ADDQoL	Mixed: ↑ flexibility (p = 0.019) and recommendation (p = 0.023); overall DTSQc change p = 0.053; no ADDQoL change.
	Flash Glucose Monitoring Helps Achieve Better Glycaemic Control Than SMBG in Non‐Insulin‐Treated Type 2 Diabetes	Wada et al. (2020)	RCT	DTSQ (unspecified)	Significant: ↑ satisfaction, convenience, and flexibility (p < 0.001); ↑ hyper perception (p = 0.047).
	Continuous Glucose Monitoring Feedback in the Subsequent Development of Gestational Diabetes (I‐PROFILE)	Quah et al. (2024)	RCT	Semistructured satisfaction survey	Significant: ↑ satisfaction (4.4 vs. 4.1, p = 0.002); ↑ relevance (p = 0.005) and motivation (p = 0.006).
	The role of continuous glucose monitoring in clinical decision‐making in diabetes in pregnancy	McLachlan et al. (2007)	Non‐randomised	Self report questionnaire	Descriptive: 77% felt benefits > inconvenience; 90% improved understanding; no p‐values.
Psychological/Quality of Life	The Influence of Real‐Time CGM on Psychosocial Outcomes in Insulin‐Using Type 2 Diabetes	Soriano & Polonsky (2023)	Non‐randomised	Distress, hypoglycaemia, fear & confidence scales	Significant: ↓ distress (p < 0.001) and fear (p = 0.031); ↑ hypo confidence (p < 0.001); ↓ burden (p < 0.001).
	Effect of Flash Glucose Monitoring on Mental Well‐being & Treatment Satisfaction in Japanese People with Diabetes	Mitsuishi et al. (2017)	Non‐randomised	WHO‐5, DTSQ (unspecified)	Non‐Significant: No change in WHO‐5 (p = 0.218) or DTSQ satisfaction (p = 0.422) for the T2D subset.
	Glucose Sensor‐Augmented CSII in Diabetic Gastroparesis (GLUMIT‐DG)	Calles‐Escandón et al. (2018)	Non‐randomised	PAGIQOL	Significant: ↑ PAGI‐QOL (2.4 to 3.1, p < 0.0001); ↑ fullness/bloating symptoms (p < 0.001).
	Comparing CGM and Blood Glucose Monitoring in Adults With Inadequately Controlled, Insulin‐Treated T2D (Steno2tech)	Lind et al. (2024)	RCT	WHO‐5, Distress scale, satisfaction scales, DTSQ‐s and DTSQ‐c	Significant: ↑ WHO‐5 (p = 0.041), DTSQ (p < 0.005), and GMSS (p < 0.0001); diet/activity non‐significant.
	Short‐ and Long‐Term Effects of Real‐Time Continuous Glucose Monitoring in Patients With Type 2 Diabetes	Vigersky et al. (2012)	RCT	Problem Areas in Diabetes (PAID)	Non‐Significant: No difference in emotional distress (Wilcoxon p = 0.85; Friedman p = 0.18).
	Continuous Glucose Monitoring Versus Usual Care in Patients With Type 2 Diabetes Receiving Multiple Daily Insulin Injections	Beck et al. (2017)	RCT	EQ‐5D‐5L, WHO‐5, DDS, Fear‐of‐ hypoglycaemia, hypoglycaemia confidence scale	Non‐Significant: No statistically significant between‐group differences in general or diabetes‐specific QoL at 24 weeks.
	Impact of quarterly professional‐mode flash glucose monitoring in adults with type 2 diabetes in general practice (GP‐OSMOTIC): Secondary psychological and self‐care outcomes of a pragmatic, open‐label, 12‐month, randomised controlled trial	Speight et al. (2021)	RCT	WHO‐5, DAWN Impact of Diabetes Profile (DIDP), Glucose Monitoring Experiences Questionnaire (GME‐Q) for satisfaction, Perceived Involvement in Clinical Care Scale (PICS)	Non‐Significant: No differences in WHO‐5 (p = 0.65), DIDP (p = 0.58), GME‐Q (p = 0.06) or PICS facilitation (p = 0.80).
	Continuous glucose monitoring in people with diabetes: the randomised controlled Glucose Level Awareness in Diabetes Study (GLADIS)	New et al. (2015)	RCT	SF‐8 (physical/mental), Distress scale	Mixed: ↑ physical SF‐8 (p = 0.0245); no significant change in mental component or overall distress.
	Role of Flash Glucose Monitoring System Combined With Insulin Pump in Blood Glucose Treatment of Patients with Type 2 Diabetes Mellitus	Wang et al. (2021)	RCT	GCQ, SAS, SDS, PSQI, WHOQOL‐BREF	Significant: ↑ comfort, sleep, and QoL (all p < 0.05); ↓ anxiety and depression (p < 0.05).
	Multicenter Randomized Trial of Intermittently Scanned Continuous Glucose Monitoring Versus Self‐Monitoring of Blood Glucose in Individuals With Type 2 Diabetes and Recent‐Onset Acute Myocardial Infarction: LIBERATES Trial	Ajjan et al. (2023)	RCT	EQ5D‐5L, DTSQ (unspecified), ADDQoL	Non‐Significant: No differences in EQ5D‐5L utility (p > 0.05) or DTSQ satisfaction (diff 1.1, p > 0.05).
Behavioural/adherence	Self‐Monitoring Using CGM with Real‐Time Feedback Improves Exercise Adherence	Bailey et al. (2016)	Non‐randomised	Self monitoring behaviours, goal‐setting, HRQL	Significant: ↑ mental HRQL (p = 0.046) and self monitoring (p = 0.03); physical HRQL was non‐significant.
	CGM + Problem‐Solving Counselling to Change Physical Activity in Women with T2D	Allen et al. (2011)	Non‐randomised	Satisfaction survey & problem‐solving skills	Significant: ↑ problem‐solving (p = 0.02) and dietary adherence (p = 0.01); depression was non‐significant.
	A randomised, controlled cross‐over study of short‐term CGM (DISCO GM)	Parsons et al. (2024)	RCT	Diabetes Self‐Management Questionnaire (DSMQ)	Significant: ↑ DSMQ Total Score (p = 0.001), Monitoring (p = 0.014), and Activity (p = 0.011).
	MITRE Trial: Minimally Invasive Glucose Monitoring vs. Conventional Care in Insulin‐Treated Diabetes	Newman et al. (2009)	RCT	ADDQoL, SDSCA, fear‐of‐ hypoglycaemia scale, DTSQ‐s and DTSQ‐c	Non‐Significant: No advantages for ADDQoL (p = 0.089), HFS (p = 0.171) or PMD effectiveness (p = 0.559).

Note: DTSQ‐c, change version; DTSQ‐s, status version; DTSQ (unspecified), version not reported in the original study.

3.4.1. Treatment satisfaction and usability

Studies conducted in Canada, Japan, Israel and Saudi Arabia showed that CGM use was associated with higher treatment satisfaction than SMBG or usual care across both insulin‐treated and non‐insulin‐treated populations. ²⁸ , ⁴⁰ , ⁴⁵ , ⁴⁷ Improvements were reported in overall Diabetes Treatment Satisfaction Questionnaire (DTSQs/DTSQc) scores (p < 0.001 to p < 0.0001), ²⁸ , ⁴⁰ , ⁴⁵ , ⁴⁷ largely driven by perceived treatment flexibility (p < 0.001 to p = 0.019), awareness of hyperglycaemia and hypoglycaemia (p < 0.05) and willingness to recommend (p < 0.001 to p = 0.023). Complementing these scores, GMSS‐based assessments indicated reductions in behavioural burden and emotional burden alongside increases in perceived usefulness and worthwhileness of glucose monitoring, ⁴⁷ and immediate satisfaction gains were also noted in short‐term studies. ⁵⁴ Non‐RCT studies also reported high satisfaction and usability rates, with most participants describing CGM as easy to use and reporting improved understanding of the effects of diet and physical activity on glucose level. ⁴³ , ⁵⁰ , ⁵² However, acceptability was found to be device‐dependent. In the MITRE trial, earlier CGM systems such as the GlucoWatch were reported to interfere with daily activities, contributing to lower acceptability. ³⁶

3.4.2. Psychological impact and QoL

Evidence of the impact of CGM on general quality of life is mixed. While one study using the WHOQOL‐BREF reported higher quality of life scores across multiple domains (physiological, psychological, environmental and social) among CGM users, ³⁰ most studies employing WHO‐5, EuroQol‐5D or ADDQoL across short‐(14‐day), medium‐(24‐week) and long‐term (18‐month) follow‐up periods reported no significant differences between CGM users and control groups. ³⁴ , ⁵⁰ A notable exception was the Steno2tech trial, which observed a significant improvement in WHO‐5 well‐being scores (p = 0.041). ³⁵

In contrast, findings related to diabetes‐specific distress were more consistent. A study using the Diabetes Distress Scale (DDS) found a significant overall reduction in distress from baseline, particularly in the emotional burden and regimen distress domains. ¹⁵ CGM use also influenced hypoglycaemia‐related psychological outcomes; in the DIAMOND trial, 94% of participants reported increased perceived safety about hypoglycaemia. ³⁴ Among individuals with impaired awareness of hypoglycaemia (IAH), early reductions in fear and distress were not sustained over longer follow‐up. When CGM use was combined with an exercise program, the mental component (SF‐36) of participants improved significantly over time. ⁴² Similarly, when CGM was featured with alarms, participants' perception of their physical health improved compared to SMBG. ⁴⁶ Additionally, in a pilot study involving individuals with diabetic gastroparesis, combining CSII and CGM increased the PAGIQOL score by 0.7. ³³

3.4.3. Behavioural/adherence

Few RCTs directly assessed behaviour change or adherence. Real‐world and pilot studies reported that CGM use may enhance engagement and adherence. When CGM was combined with problem‐solving or lifestyle interventions, such as an 8‐week exercise adherence study in prediabetes/T2D adults, participants showed greater self monitoring, goal‐setting and adherence. These effects are often attributed to real‐time feedback, although this mechanism has been rarely examined in trials. ⁴² Regarding device tolerability, mild insertion or wear‐related symptoms (e.g. bleeding, erythema, itching) were commonly reported with newer CGM systems and generally resolved without treatment. However, earlier devices like GlucoWatch were associated with high burden of adverse skin reactions, and high discontinuation rates, highlighting the influence of the device on participant experience and sustained use. ³⁶

4. DISCUSSION

This systematic review demonstrates that CGM provides significant clinical and psychosocial benefits for adults with T2D compared to SMBG. While T2D remains difficult to manage, with individuals requiring interventions ranging from lifestyle modification and oral agents to intensive insulin injections, ⁵⁶ treatment responses may vary due to differences in clinical characteristics and genetic factors influencing antidiabetic therapies. ⁵⁷ Compared with traditional SMBG, CGM provides a detailed assessment of glycaemic patterns and has been associated with an initial robust reduction in HbA1c. ⁴⁰ Furthermore, by capturing glucose trends throughout the day and night, CGM empowers individuals to understand the glycaemic impact of diet and activity while supporting clinicians in optimising therapy. ⁵⁸

Importantly, the magnitude of this glycaemic improvement is influenced by the specific CGM modality used. While both rtCGM and isCGM improve glycaemic outcomes, rtCGM generally demonstrates greater reductions in HbA1c and more consistent improvements in TIR. This may be explained by the real‐time feedback and alert functions available in rtCGM, which enable more immediate behavioural and therapeutic adjustments, whereas isCGM requires user‐initiated scanning and may result in less timely intervention. ¹⁷ Although the reduction in HbA1c may appear modest in some individuals, these reductions become clinically relevant when considered alongside improvements in TIR. For example, a 3 mmol/mol (0.3%) drop in HbA1c combined with a 10% increase in TIR reduces microvascular risk. ⁵⁹

To ensure appropriate interpretation of evidence, this review distinguished between high‐quality RCTs and non‐randomised studies, thereby reducing the risk of counting. As the search strategy included studies from database inception to June 2025, with the earliest study from 2007, technological heterogeneity across CGM generations was anticipated. To address this, greater emphasis was placed on contemporary RCTs when interpreting clinical outcomes, while earlier studies were considered primarily to provide historical context regarding patient experience and the evolution of CGM hardware.

Consequently, our review favours evidence from the three high‐quality RCTs on CGM, ⁷ , ³⁴ , ³⁵ wherein the HbA1c reduction was between 5 and 10 mmol/mol (0.5–0.9%) between groups. However, care should be taken when interpreting higher reductions, as they are population‐dependent. The three major RCTs, ⁷ , ³⁴ , ³⁵ had different baseline HbA1c levels, with the greatest reductions observed in subsets with the highest baselines. Notwithstanding this effect, CGM use favoured HbA1c improvement across various study designs.

Moreover, even when HbA1c reductions are modest, robust improvements in TIR are consistently observed, highlighting a distinct physiological mechanism that CGM aids primarily in reducing variability and postprandial spikes without drastically shifting the absolute glycaemic mean. TIR serves as a vital, actionable metric for daily management, ⁷ , ³⁵ aligning with current ADA guidelines recommending both HbA1c and TIR to assess glycaemic control. ¹⁷ In practical terms, a 6–9% increase in TIR translates to an extra 1.5–2 h per day in the euglycaemic range. Across multiple studies, ²⁹ , ⁴⁰ reductions in SD and MAGE indicate a smoother glucose pattern. Achieving this glycaemic stability is important, as excessive glucose fluctuations contribute to microvascular and macrovascular complications through oxidative stress, inflammation and endothelial dysfunction, ⁶⁰ exacerbating the metabolic abnormalities often associated with T2D. ⁶¹

To realise these physiological benefits, however, the effectiveness of the device is contingent on the user's ability to interpret their data. Diabetes self management education and counselling synergistically improve the clinical outcomes of CGM by helping patients improve glycaemic control and reduce hypoglycaemia. ²⁶ , ⁶² However, the durability of these behavioural responses and the primary clinical value of the device itself differ significantly based on the patient's underlying treatment regimen.

For insulin‐treated individuals, diabetes management requires balancing glycaemic targets with the risk of hypoglycaemia. ³ , ⁴ In this context, CGM provides continuous glucose data that support treatment adjustment. ⁴ Major trials such as MOBILE, DIAMOND and Steno2tech, ⁷ , ³⁴ , ³⁵ demonstrate that CGM yields HbA1c reductions and significant increases in TIR compared to traditional SMBG. Furthermore, in high‐risk populations like those on multiple daily injections or with a recent myocardial infarction, CGM significantly reduces dangerous hypoglycaemic exposure. ³¹ , ⁵⁴

Conversely, for non‐insulin‐treated patients, baseline hypoglycaemia is minimal, shifting the utility of CGM toward learning and behavioural modification. ⁴¹ , ⁶³ Real‐time glycaemic feedback presents the immediate consequences of dietary choices. ⁶³ This empowers patients to reach HbA1c targets without escalating pharmacotherapy, as demonstrated in the COMMITED pilot trial, ²⁶ particularly when supported by virtual education. ²⁸ Ultimately, these findings suggest the role of CGM may differ according to treatment context, with greater emphasis on safety and treatment optimisation in insulin‐treated patients, and on behavioural insight and self management in non‐insulin‐treated individuals. ⁴ , ⁵⁶

While CGM demonstrates clear utility in the general T2D population, its application in special populations requires specific consideration. For example, in the pregnant population, findings remain inconsistent, not because CGM lacks utility, but due to heterogeneity in the evidence base regarding diabetes subtypes, timing of initiation, device modality, and outcomes assessed. ⁶⁴ , ⁶⁵ Furthermore, intermittent sensor use may attenuate the measurable impact of CGM on longer‐term outcomes like HbA1c. ³⁷ , ⁶⁴ Importantly, improvements in CGM‐derived metrics do not always translate into better maternal or neonatal clinical endpoints, highlighting uncertainty regarding the most appropriate glycaemic targets in pregnancy. ⁶⁵ , ⁶⁶ While device lag in measuring interstitial glucose during rapid physiological changes plays a role, ⁶⁵ current evidence suggests inconsistency is more strongly driven by study heterogeneity and sample sizes rather than device accuracy alone. ⁶⁴ While one included study evaluated an insulin‐resistant state rather than established type 2 diabetes, its findings provide critical mechanistic context by demonstrating CGM's capacity to capture early glycaemic excursions. ⁴⁴

Beyond pregnancy, CGM provides critical utility in other complex conditions where standard metrics fall short. For individuals with a recent myocardial infarction, standard HbA1c often fails to capture the acute hypoglycaemic events that trigger pro‐arrhythmic variability, ⁶⁷ suggesting the benefit of CGM lies in its predictive alerts rather than overall glucose lowering. For dialysis candidates, HbA1c is rendered unreliable by altered erythrocyte lifespan and uremic environments, ⁶⁸ making CGM essential for clinical decisions. Additionally, for individuals with gastroparesis, where uncoupled insulin action and nutrient absorption cause early hypoglycaemia followed by a dramatic spike, CGM is critical for identifying these delayed post‐prandial peaks. ⁶⁹

Despite these diverse physiological applications, the effective use of CGM ultimately depends on user adherence and experience. Managing diabetes is an overwhelming burden, ⁷⁰ making it essential to capture PROs. Validated instruments like the Diabetes Treatment Satisfaction Questionnaire (DTSQ) ⁷¹ and the Glucose Monitoring System Satisfaction Survey (GMSS) ⁷² consistently report higher treatment satisfaction with CGM across RCTs and non‐randomised studies. Importantly, these improvements are observed even in studies reporting modest glycaemic changes, indicating that satisfaction is driven heavily by user experience and emotional response. The elimination of finger‐prick testing provides ease and convenience, ¹⁰ leaving users feeling more informed and confident. However, real‐world implementation barriers remain. While skin irritation was primarily an issue with older devices, ³⁷ the ongoing financial burden of modern sensors continues to limit widespread adoption, particularly in under‐resourced settings where out‐of‐pocket costs prohibit continuous access. ³¹ , ⁴⁸

Beyond treatment satisfaction, CGM use significantly reduces the fear of hypoglycaemia and diabetes‐related distress. This alleviation of emotional burden is closely linked to improved clinical outcomes, as users gain the confidence to titrate insulin and enjoy greater dietary flexibility. ³³ Interestingly, generic quality‐of‐life measures, such as EQ‐5D and WHO‐5, often reveal little to no significant change. ⁷³ This discrepancy likely reflects differences in measurement sensitivity: generic instruments primarily evaluate broad functional domains (e.g. mobility) prone to ceiling effects, whereas CGM‐centric questionnaires specifically capture the psychosocial relief provided by the device.

Overall, these findings indicate that access to real‐time glucose trends supports day‐to‐day decision‐making and offers benefits extending far beyond glycaemic metrics. ⁶³ However, translating these perceived benefits into durable clinical changes requires mitigating the novelty effect. Initial improvements in PROs and behaviours often drop off if CGM use is episodic, ²⁶ , ⁴³ whereas long‐term access as demonstrated in Steno2tech, ³⁵ can sustain positive outcomes. Therefore, transforming the transient novelty of a new device into durable behavioural change likely requires continuous access rather than intermittent use.

5. STRENGTHS AND LIMITATIONS

This systematic review provides a comprehensive and up‐to‐date synthesis of CGM in T2D, combining evidence from RCT and observational studies to capture controlled efficacy and real‐world effectiveness. A key strength is the inclusion of both RCT and non‐randomised studies, allowing evaluation of CGM effects under controlled conditions as well as in real‐world clinical settings. In addition, this review examined both clinical outcomes and PROs, enabling a more complete assessment of clinical and user‐centred effects of CGM use. While taking into consideration the quality of evidence for clinical and PROs. This review provides a thorough and nuanced understanding of CGM use, its sensitivity in various populations subsets and its effects on user experiences. However, several limitations must be considered. The overall quality of included studies was moderate, with many RCTs being open‐label and single‐centre, and limited by small sample sizes, increasing the risk of performance and detection bias. Considerable heterogeneity in interventions duration, study populations, and outcome definitions prevented meta‐analysis and limits the precision of pooled estimates. The evidence base for non‐insulin‐treated and pregnant populations remains limited. Moreover, only studies published in English were included, which may have excluded relevant research conducted in non‐English‐speaking regions where the T2D prevalence is high.

Despite these limitations, the convergence of evidence across diverse study designs provides reasonable confidence that CGM benefited adults with T2D clinically and psychosocially when used consistently and supported by CGM user education.

6. CONCLUSION

CGM use is associated with improvement in glycaemic management and user experience in adults with T2D. CGM provides access to real‐time glucose data and glycaemic metrics beyond SMBG, including TIR and glycaemic variability. Improvements in HbA1c and time in range are clinically meaningful and were not associated with an increased risk of hypoglycaemia across studies. PROs, particularly treatment satisfaction and diabetes‐related distress, showed more consistent improvement than generic QoL measures. Taken together, these findings highlight the clinical value of CGM in T2D, with its effects best interpreted by considering both glycaemic metrics and CGM user‐reported experiences.

Future research on CGM must address existing methodological limitations, long‐term effects, and real‐world implementation barriers. To enhance evidence quality, subsequent trials should employ rigorous designs such as sham‐sensor controlled protocols to minimise open‐label biases and directly compare real‐time versus blinded professional‐mode systems. Longitudinal studies extending beyond 6 months are important to find whether glycaemic and behavioural improvements are sustained or a short‐term novelty effect. Furthermore, study populations must be diversified to include underrepresented ethnic minorities and complex clinical cohorts, such as those with diabetic gastroparesis, recent myocardial infarction or severe hypoglycaemia unawareness.

FUNDING INFORMATION

The authors received no financial support for the research.

CONFLICTS OF INTEREST STATEMENT

The authors declare no conflicts of interest.

Supporting information

Figure S1. Mixed Cohort Studies.

Table S1. Search Strategy for searching databases.