Meta-Analysis of a Decade of Studies Assessing Accuracy of Abbott FreeStyle Libre Continuous Glucose Monitor

Abstract

Background and Aims:

This meta-analysis provides a summary of research publications over a decade since the FreeStyle Libre device was first launched by Abbott in 2014. This includes research publications in which the Libre freestyle glucose measurements have been evaluated for accuracy or compared to other glucose devices.

Methods and Results:

The dichotomy is organized into six sections: (1) Comparing Libre to Dexcom CGM, (2) Assessing Libre accuracy, (3) Efficacy for diabetes, (4) Ease of use, (5) Direction of change trend arrows, (6) Exercise while wearing Libre. A comprehensive meta-analysis of studies comparing Libre and Dexcom is presented. Main results include: Libre and Dexcom give close matching glucose measurements. Libre and Dexcom are both clinically acceptable. Libre efficacy is worse for glycemic variability and does not reduce the risk of recurrent hypoglycemia. Our comprehensive meta-analysis integrates data from these key studies; our results demonstrate that the accuracy of Libre was not significantly different from accuracy of Dexcom when compared with a gold standard comparator.

Conclusions:

Several factors reduce Libre accuracy: hypoglycemia, rapid glucose changes, night time, exercise, the first and last days of the 14-day sensor life, placing on abdomen instead or arm or thigh, and variation between left and right arms. The direction of change arrows need more consideration before integrating with clinical decisions. Libre can also perform well when used by children and pregnant women. This provides a current overview of Libre, highlighting open questions for future research direction.

Introduction

The FreeStyle Libre (FSL) (Abbott Diabetes Care Ltd., Witney, UK) or FSL-Pro is a flash glucose monitoring (FGM) or intermittently scanned continuous glucose monitoring (isCGM) device which measures the interstitial glucose every 15 min. The device also has an option to obtain a manual current glucose reading (capillary blood glucose) by the user scanning the sensor with the Libre reader. The FreeStyle Libre device is widely used, with a particularly high uptake among people with type 1 diabetes (T1D) in Belgium.¹

Common alternative devices include the Dexcom G4 platinum (DG4P, Dexcom, San Diego, CA) or G5 Mobile (DG5M) device, which provides continuous glucose monitoring (CGM) or real-time CGM, which measures the interstitial glucose every 5 min.

Medtronic also produce CGM devices including the Medtronic iPro2 Professional CGM and Medtronic Enlite (ENL).

Some studies have compared accuracy of glucose measurements between Libre and Dexcom, or other devices including iPro2 or fingertip pricking.

Libre and Dexcom devices are the most widely used devices,² increasingly distributed in the United States and European Union, aiming to replace blood glucose monitoring (BGM),³ increasingly prevalent in clinical practice⁴ and increasingly used by people on multiple daily insulin injections or continuous subcutaneous insulin infusion.⁵

This following sections of this review summarize recent studies assessing accuracy of the FSL, by comparison with other CGM sensors, accuracy in various cohorts, the efficacy of improving diabetes conditions, the ease of use, and patient-reported outcomes and considers how Libre rate of change (RoC) trend arrows can be used for making decisions and the benefits of the Libre during exercise.

Literature Review

Extensive literature searches were conducted on a range of most popular research publication databases including arXiv, PubMed, Scopus, Google scholar, ASME Digital Collection, and IEEE Xplore. Additional references were identified from related citations within the articles found, including citations within recent systematic reviews. Primary search keywords included “Freestyle Libre,” “Continuous Glucose Monitor,” “Diabetes,” “Exercise,” “Accuracy.” Boolean operators were also applied to refine results further. After excluding duplicates and reading for relevance, a list of 33 articles were identified and are outlined within the following sections of this review.

Comparison between Libre and Dexcom for glucose sensor accuracy

In multiple studies, the accuracy of these two devices was assessed and compared by mean absolute relative difference (MARD). A Dexcom G5 and Libre were both worn together by 28 subjects for 14 days including sessions at two clinics where glucose excursions were induced⁶ in a study funded and data analysis by Dexcom. Glucose measurements from the Dexcom and Libre were both compared with YSI 2300 STAT Plus, YSI Inc., Yellow Springs, OH, United States; glucose oxidase-based. Both Dexcom and Libre devices displayed clinically accurate readings, Dexcom had 91.5% within ±20%/20 mg/dL (MARD 9.5%), Libre had 82.5% within ±20%/20 mg/dL (MARD 13.6%).

The FGM FSL was compared with the gold-standard CGM DG4P with 8 patients of various ages, HbA1c and diabetes duration. Patients wore both devices for 2 weeks,⁷ and this found no significant difference between the devices in clinical diagnostic parameters, as over 10,020 paired values had good correlation with R² = 0.76.

A clinical trial⁸ included 21 adults with T1D wearing both Dexcom and Libre for 2 weeks. Accuracy of both devices was assessed by comparing both with a gold standard fingerstick test. This found no significant difference in MARD or accuracy of the two devices, even after delayed and increased bolus. With both systems, the MARD increased during hypoglycemia. Libre was more accurate than Dexcom during moderate and rapid glucose changes. A follow-up study⁹ compared accuracy of Libre with the newer DG5M, the most recent sensor produced by Dexcom Inc., which is more accurate than the old DG4P.

A study with 38 children wearing either Dexcom G6 or Libre¹⁰ had lower overall MARD with Dexcom (10.3%) than Libre (13.3%) when both devices were being compared with self-monitoring of blood glucose (SMBG).

A 14-day study with 20 participants³ evaluated Libre and Dexcom G5 by comparing both with BGM, including at a study site, home use, and phases of induced rapid glucose changes. Accuracy of both was lower at low blood glucose, because when the glucose concentration was below 100 mg/dL, 25% of DG5 and Libre results showed differences (over 15 mg/dL) from BGM (therefore, the use of CGM devices could affect one in four therapeutic decisions), whereas over 100 mg/dL, differences reduced to 15%, showing higher accuracy. Using consensus error grid (CEG) analysis, both Dexcom and Libre had over 99.5% readings within clinically acceptable zones A–B.

The overall accuracy of FSL and Dexcom G5 sensor was shown to be the same (MARD 12.8% and 12.5%, respectively; P = 0.57)¹¹ in 14 adults with T1D comparing 1930 pairs of FSL sensor with plasma glucose measurements.

A study showed a comparison of the Dexcom G5, Libre Pro, and Senseonics Eversense CGM during 6 weeks home use by 23 subjects with T1D wearing all three devices simultaneously.¹² Accuracy of the three devices was compared with point-of-care plasma-glucose values obtained twice daily by the subjects. Eversense achieved better result than Dexcom and Libre, with the lowest nominal MARD (14.8%), followed by Dexcom G5 (16.3%) and Libre Pro (18.0%).

A study had 20 participants wearing two Dexcom G5s and two Libre sensors for 14 days. For some subjects, there was a considerable difference measured between two sensors of different systems, and Libre had the largest differences between sensors of the same system.¹³

A study with 16 adults without diabetes during 28 days wore Libre and Dexcom simultaneously,¹⁴ generating 27,489 measurements. The devices appear not precise enough for personalized nutrition due to imprecision and discordance of within-subject meal rankings between simultaneous CGM devices.

Metrics and parameters derived from CGM, including time below range and time above range, differed substantially between Libre and Dexcom G5, but the time in range (TIR) was identical,¹⁵ in 24 subjects wearing Dexcom and Libre simultaneously for 7 days.

Accuracy of Libre

A clinical trial compared the Libre manual scanned data to the automatic Libre 15 min scanned data, finding that 5% of manual readings showed relative differences over ±10% from the automatic 15-min readings.¹⁶

The body position of the Libre sensor affects accuracy.¹ Placing the Libre sensor on the upper thigh and upper arm gives similar accuracy and precision but placing the sensor on the abdomen gives poor and unacceptable data. This was evaluated by comparing the Libre 15-min measurements with the Libre manual scan data (built-in FSL BG meter).

When worn by 75 children, comparison between Libre sensor and glucose meters (3143 measurements) showed good MARD index—18.22%.¹⁷ The Clarke Error Grid indicated that 2309 (75.2%) of results were acceptable errors (in zone A). Teenagers and children find managing glycemia a significant burden and glycemic control at this age has the largest health implications. A future planned trial will look into whether Libre can improve glycemic control in adolescents.¹⁸ Another trial confirmed good accuracy on 78 children.¹⁹

A study with 10 patients compared both Libre Pro and Medtronic iPro2⁴ with manual self-monitoring with 5555 paired values. Parkes error grid analysis of glucose values in areas A and B showed that the FSL-Pro had 92.9% and 7.1%, iPro2 had 96.3% and 2.8%. MARD was 8.1% for FSL and 5.0% for iPro2. Also, 65.3% of all glucose values were lower for FSL-Pro than on iPro2. During hypoglycemia, the difference between the two systems was generally increased.

A study including 38 children with T1D showed lower accuracy on the Libre compared with Medtronic Enlite. The Enlite had MARD 8.5%, and the Libre had 13.3% MARD overall.

In pregnant women, there are usually restrictions on use of Libre, however, studies have shown a strong positive correlation between interstitial glucose levels measured by FSL sensor and capillary SMBG in a pregnant woman affected by T1DM,²⁰ with one woman having 178 paired capillary blood glucose and sensor values over 14 days with mean absolute difference of 2.6 and CEG analysis of 100% within zones A + B.

During exercise, Libre has shown diminished accuracy. With moderate-intensity cycling for 55 min, when compared with reference blood glucose levels²¹ for 10 participants with T1D, with 845 glucose values, MARD was 22 and increased to 36 during hypoglycemia. In a follow-up study with 19 participants,²² without exercise, MARD was 14% rising to 31% in hypoglycemia, and when exercising, MARD increased to 29% rising to 45% during hypoglycemia.

The age of Libre sensor matters, with lowest accuracy of MARD 14% during day 1 and day 14, compared with MARD 7% on days 5–7.¹¹ This was evaluated by comparing Libre to Plasma glucose levels using YSI2300 STAT Plus Analyser.

A study has checked whether plasma glucose can be estimated from a Libre in health children. The Libre did correlate with post- Oral Glucose Tolerance Test (OGTT) glucose but underestimated the plasma glucose in obese children.²³

Different equations may be needed for glucose management indicators to be CGM device-specific and race-specific.²⁴

Study²⁵ compared Libre with venous blood glucose over 14 days with 26 T2D patients; the 208 paired values showed high correlation with 100% within zones A and B of the Clarke and Parkes error grid analysis. Linear interpolation was used to match the sensor times. During rapid glucose changes, Libre performed less well and underestimated glucose.

Libre devices with 15 min intervals should make 96 measurements per day, but often in practice, have between 1–5 missing values per day due to data loss, but the extent and possibility of 1–6 h missing data should be taken into account when analyzing data from Libre device.²⁶

In 2019, the newer Libre-B generation glucose tracking algorithm and calibration has improved Libre accuracy from the previous Libre-A generation.¹⁹

The right and left arms gave statistically significant discordance in glucose readings in 10 adults which may not be explained by left or right arm dominance.²⁷

Libre showed good accuracy for calculating the Mean Amplitude of Glycemic Excursions (MAGE)²⁸ in 68 patients, an important component of glycemic control, performed during 2 continuous days, matching closely to the standard method of calculating MAGE over 48-h.

Efficacy of improving diabetes conditions

Glycemic variability (GV) was significantly worse when using Libre, when compared with Dexcom,²⁹ in clinical trials with 40 participants. Most GV measures, (computed using EasyGV software), improved with Dexcom when compared with Libre,²⁹ which is important to minimize severe hypoglycemia.

Efficacy was not improved by Libre for reducing risk of severe hypoglycemia reoccurrence in at-risk individuals when compared to SMBG. Using Libre had no significant difference on the number of hypoglycemic events. Libre does not appear to improve over SMBG for at-risk individuals to prevent recurrent severe hypoglycemia.³⁰

Users of Libre spent 38% less time in hypoglycemia compared with users with SMBG.⁵

Libre is associated with mean reduction in the number of hypoglycemic episodes per week of 3.20 (52%) and significant improvement and mean reduction in HbA1c of −7.29 ± 10.76 mmol/mol (P < 0.001).³¹

Ease of use and patient-reported outcomes of Libre

A study with 75 children wearing Libre¹⁷ evaluated opinions of the device; 94% reported that comport and ease of installation was good or very good. A total of 92% reported ease of reading blood glucose values; 95% had no side effects; and 62% of sensors remained intact for 14 days.

Patient-reported outcome measures have shown positive experiences of Libre with significant improvement in all aspects of a focused Diabetes Distress Scale.³¹ A significant improvement in quality-of-life scores was noted in all five domains with key themes highlighted including “life-changing,” “positive experience,” and “convenient,” from over 90 participants.

Libre can avoid the pain and the inconvenience of regular finger pricking.²⁰

Individuals at a high risk of T2D engaged with wearing Libre in combination with Fitbit for 6 weeks, which could help to provide behavioral and physiological feedback in response to exercise and glucose.³²

Using Libre RoC trend arrows for decisions

Trend arrows provide information on both the direction and RoC of glucose.

There is inherent difficulty formulating practical recommendations for incorporating Dexcom and Libre trend arrows into routine clinical care² due to adjustments such as Bolus timing and insulin sensitivity.

It was reported that only around 60% of direction of change arrows may match measured glucose changes.³³ In over 10% of cases, particularly close to insulin administration or carbohydrate ingestion, trend arrow accuracy was very poor. Also, most RoC arrows overestimate actual change.²

Discussions are ongoing and more rigorous studies are needed about the use of trend arrows to aid insulin bolusing decisions. There has been a review of the integration of RoC trend arrow information into daily glucose management, including rapid-acting insulin dosing decisions.⁵ A person using CGM needs to take into account exercise plans, tine since bolus, and several other factors when viewing their RoC arrows and making decisions.

Libre systems do not have automatic alarms associate with trend arrows.³⁴ Recent approaches to integrate trend arrows for decision making takes into account pre-exercise planning, insulin dose adjustments, and the individual’s insulin sensitivity.³⁴

Careful interpretation is needed and different methods are needed to integrate RoC arrows from Libre.³⁵ Libre has five arrow directions; other CGM devices have seven arrows. Methods proposed for integration take into account premeal time glucose, meal carbs, insulin bolus, how prone the individual is to hypoglycemia, and other factors.³⁵ These have not had a formal clinical trial evaluation.

Exercise benefits of Libre

Libre was shown useful for seven ultramarathon runners.³⁶ It could be used as a practical method during exercise to guarantee optimal carbohydrate intake for each ultramarathon runner during a 160 km race. Glucose concentration differed throughout the race and glucose concentration positively correlated with running speeds.

Libre was useful during a running race worn by 24 people, half with T1D. The Libre found significant differences in TIR between healthy people and people with T1D and higher GV in T1D.³⁷

Libre and Fitbit-2 were worn together in a study using XGBoot AI machine learning modeling to predict GV in patients with T2DM who fast during Ramadan.³⁸ The Fitbit step count was used and not the raw accelerometer data. This achieves high predictive performance for normal and hyperglycemic excursions, but has limited predictive value for hypoglycemia in patients on multiple therapies who fast during Ramadan.

Meta-Analysis of Studies Comparing Libre and Dexcom

This section presents a meta-analysis using pooled data analysis, to quantify the degree of agreement between Libre and Dexcom CGMs. This includes evaluation of all studies outlined in the previous section on the comparison between Libre and Dexcom for glucose sensor accuracy. The main steps of our meta-analysis were the identification of suitable data from the included studies, and the statistical analysis of the data.

Within each study, the paired data are from the same participants using two devices: either Libre or Dexcom and a gold standard comparator. Therefore, to run the meta-analysis, the R package metagen was used as it properly models paired data with custom Standard Error (SEs). The result is shown in the forest plot (Fig. 1).

FIG. 1.

Forest plot summary of the meta-analysis.

Data from all studies included in our meta-analysis must be available in a consistent format. Studies^3,8,11 and¹² report the results in consistent format as MARD (±standard deviation [SD]). However studies⁹ and¹⁰ report median ARD [interquartile range [IQR]]. Therefore standardization was required to enable a meta-analysis to be completed including all studies. To standardize the result formats, we converted median (IQR) to estimated mean (±SD) using established widely used formulas from³⁹ and⁴⁰ (Eq. 1 and 2). This conversion enabled inclusion of studies⁹ and¹⁰ in our meta-analysis, but introduces some uncertainty. However, study¹¹ reports results using both median (IQR) and mean (SD), so study¹¹ was used as a benchmark to quantify accuracy of these estimations using equations 1 and 2. $Estimated Mean = \frac{Q 1 + Media n + Q 3}{3}$ (1) $Estimated Standard Deviation = \frac{Q 3 - Q 1}{1.35}$ (2)

For our meta-analysis, we are interested in common effect (also known as fixed effect) statistical model, which essentially assumes a single, underlying true effect size is shared by all studies included in the analysis. This assumes that all studies are estimating the same true effect size, and any differences observed between studies are due to random chance or sampling error, such as the variability within each study. We additionally also completed a random effects model analysis, shown on the forest plot (Fig. 1). The meta-analysis aims to compare Libre and Dexcom in terms of the MARD— a continuous outcome.

Our meta-analysis uses the number of sensor pairs (not number of participants) as the sample size, due to the level of independence of the data, although we have included both in Table 1. The MARD reported in these studies is calculated over individual glucose measurement pairs; for example, there were 829 paired sensor and reference glucose readings in study¹² (see Table 1). Each reading is treated as an independent data point, as these articles report mean ± SD over these measurements.

Table 1.

Summary of Meta-Analysis Showing All the MARD Values Reported from a Number of Studies with Various Number of Participants

Study	Sub-population	Number of participants	Number of data pairs	Gold standard comparator	Version of Dexcom device	Abbott Libre MARD % (±SD)	Dexcom MARD % (±SD)
3	Overall	20	2700	SMBG	G5	10.5 (9.9)	10.3 (10.8)
3	Overall	20	2700	Venous Capillary BG	G5	13.0 (11.4)	13.4 (11.6)
8	Delayed and Increased bolus	21	710	SMBG	G4P	14.9 (5.5)	18.1 (8.1)
8	Standard bolus	21	424	SMBG	GP4	10.9 (4.1)	13.1 (4.6)
9	Hypoglycemia	20	78	YSI	G5M	15.4 (11.85)	14.87 (5.78)
9	Euglycemia	20	349	YSI	G5M	21.53 (18)	13.73 (12.89)
9	Hyperglycemia	20	242	YSI	G5M	10.93 (7.93)	8.1 (6.89)
9	Hypoglycemia	20	78	SMBG	G5M	15 (12.22)	14.97 (11.48)
9	Euglycemia	20	349	SMBG	G5M	14.33 (12.96)	10.93 (10.15)
9	Hyperglycemia	20	242	SMBG	G5M	10.5 (9.11)	8.9 (7.11)
10	Overall	38	2079	SMBG	G6	13.87 (11.04)	10.93 (8.07)
11	Sensor age 0–1 days	14	1027	YSI 2300 STAT Plus	G5	14.5 (8.9)	13.0 (5.3)
11	Sensor age 5–7 days	14	453	YSI 2300 STAT Plus	G5	7.8 (2.9)	13.6 (9.0)
11	Sensor age 13–14 days	14	450	YSI 2300 STAT Plus	G5	14.7 (11.2)	10.5 (4.0)
12	Overall	23	829	Point-of-care (POC) plasma-glucose (PG)	G5	18.0 (17.9)	16.3 (15.4)

MARD, mean absolute relative difference; SMBG, self-monitoring of blood glucose.

In study 3, the number of data pairs was not directly specified but can be estimated as 2700. Some studies^3,8,9,11 report multiple sets of MARD results, because their results were divided into distinct scenarios, such as using multiple gold standard comparators, or different bolus statuses. Therefore these are included in Table 1 as multiple separate rows per study. Some other studies which are discussed in this review^7,13–15 were not included in the meta-analysis, because although they did include both Libre and Dexcom, they did not report the MARD, due to different experimental setups, mainly because they were comparing accuracy directly with each other instead of to an independent gold standard comparator.

As shown in the forest plot (Fig. 1), our overall meta-analysis of these 15 datasets identified no significant difference between accuracies of Libre and Dexcom. Most studies in this meta-analysis (11 out of 15) reported a positive mean difference, which favors Dexcom and the overall random effects model mean difference is 0.96, suggesting slight favoring of Dexcom. However the 95% CI includes the no-effect value 0, so this shows no statistically significant difference between the two devices. Similarly, the common effects model shows very minimal difference between the two devices of only 0.04 overall.

Discussion

The main breakthrough findings or headline outcomes reported in the reviewed literature and clinical trials are shown below.

There is close match between glucose readings from Libre and Dexcom.^6–9,11,12

Accuracy of glucose measurements are clinically acceptable on Libre,^{1,3,4,6,7,11,12,16,17} on Dexcom,^3,7–9,12 and on iPro2.⁴

Efficacy of Libre was significantly worse than Dexcom for controlling GV.²⁹ Efficacy of Libre for reducing risk of severe hypoglycemia reoccurrence was not improved with Libre compared with SMBG.³⁰

During hypoglycemia, accuracy of glucose is lower on Libre,^{3,4,8,10,19,21} and on Dexcom^3,8 and iPro2.⁴

During the first 12–24 h of wear, there is relatively low accuracy with FGM or CGM sensors.^2,27,33 Also, toward the last day at the end of 14-day sensor life, Libre sensor accuracy is lower.¹¹

During rapid glucose changes, Libre performs less well.²⁵

During exercise, accuracy of Libre is lower^21,22 and exercise intensity is a factor to take into consideration when making decisions based on RoC arrows.⁵

Night-time accuracy was lower than day-time on Libre, Dexcom, and Medtronic Enlite CGM devices.¹⁰

Direction of change arrows need more research and discussion for adequate integration into care decisions.^2,5,33–35

Libre does show good accuracy when worn by children^{10,17–19,23} and by pregnant women³⁵ despite current restrictions.

Metrics and parameters derived from CGM including time in hypo and time above range differed substantially between Libre and Dexcom G5,¹⁵ but TIR was identical.

Conclusions

This review article brings together a collection of articles and studies within this area over a decade since the FSL device was first launched by Abbott in 2014. The articles include research publications in which the FSL glucose measurement has been evaluated for accuracy or compared with other glucose devices.

There is increasing use of the Abbott Libre CGM worldwide and the CGM device is increasingly recommended in guidelines such as the National Institute for Health and Care Excellence (NICE) guidelines. The main findings of this review are summarized as headline messages outlined within the discussion. Additionally, we completed a meta-analysis using the data from key studies, finding that there was no significant difference in the accuracy of Libre and Dexcom when compared with a gold standard comparator (Fig. 1). This review article provides a state-of-art summary of the accuracy and overall uses of the Abbott Libre CGM device. These findings can guide future direction and will influence researchers when planning to integrate CGM data into future research studies. These findings will also be of influence for people with diabetes considering use of CGM devices worldwide.

Footnotes

Acknowledgments

This project utilized high-performance computing funded by the U.K. Medical Research Council (MRC) Clinical Research Infrastructure Initiative (award number MR/M008924/1). The views expressed are those of the author(s) and not necessarily those of the NIHR or the Department of Health and Social Care. This research was supported by the NIHR Exeter Clinical Research Facility (NIHR Exeter CRF) in partnership with the Royal Devon University Healthcare NHS Foundation Trust within the Research Innovation Learning & Development (RILD) Building. For the purpose of open access, the author has applied a ‘Creative Commons Attribution (CC BY) license’ to any Author Accepted Article version arising.

Authors’ Contributions

N.V. led the Conceptualization, Data curation, Formal analysis, Funding acquisition, Investigation, Methodology, Project administration, Resources, Software, Supervision, Validation, Visualization, Writing—original draft, and Writing—review and editing.

Human Ethics Declaration

No human subjects or participants were involved in this research. This research was designed in accordance with the University of Exeter’s ethics committee and Internal Review Board (IRB) guidelines. Clinical trial number: not applicable. This research was completed in accordance with the Declaration of Helsinki.

Consent to Participate Declaration

No human subjects or participants were involved in this research. This research was completed in accordance with the Declaration of Helsinki.

Author Disclosure Statement

N.V. receives remuneration as co-opted member of the Wellcome Trust’s Data Sciences, Tools and Technology Discovery Advisory Group for the Career Development Awards (CDA) and Discovery Awards. N.V. is a member of the panels for NIHR i4i Product Development Awards (PDA), Transformative and disruptive innovations (TDI) and Translate Healthcare Research through InnoVation and Entrepreneurship (THRIVE). N.V. is a member of the Royal Society committee panel for Research Grants. N.V. is a member of the UKRI Medical Research Council (MRC) Panel for Non-Clinical Training and Career Development (NCTCD). No other known competing financial interests or personal relationships could have appeared to influence the work reported in this article.

Funding Information

The Royal Academy of Engineering (RAEng) contributed and supported this research through a Research Fellowship awarded to Neil Vaughan. This research was supported by the National Institute for Health and Care Research (NIHR) Exeter Biomedical Research Centre (BRC) and the Exeter Centre of Excellence in Diabetes Research (ExCEeD).

Abbreviations Used

References

Charleer

, Mathieu

, Nobels

, et al. Accuracy and precision of flash glucose monitoring sensors inserted into the abdomen and upper thigh compared with the upper arm. Diabetes Obes Metab, 2018; 20(6):1503–1507; doi: 10.1111/dom.13239

Fantasia

, Modzelewski

, Steenkamp

. Predictive glucose trends from continuous glucose monitoring: Friend or Foe in clinical decision making? J Diabetes Sci Technol, 2019; 13(5):963–966; doi: 10.1177/1932296818823538

Freckmann

, Link

, Pleus

, et al. Measurement performance of two continuous tissue glucose monitoring systems intended for replacement of blood glucose monitoring. Diabetes Technol Ther, 2018; 20(8):541–549; doi: 10.1089/dia.2018.0105

Kumagai

, Muramatsu

, Fujii

, et al. Comparison of glucose monitoring between Freestyle Libre Pro and iPro2 in patients with diabetes mellitus. J Diabetes Investig, 2019; 10(3):851–856; doi: 10.1111/jdi.12970

Ajjan

, Cummings

, Jennings

, et al. Optimising use of rate-of-change trend arrows for insulin dosing decisions using the FreeStyle Libre flash glucose monitoring system. Diab Vasc Dis Res, 2019; 16(1):3–12; doi: 10.1177/1479164118795252

Link

, Kamecke

, Waldenmaier

, et al. Comparative accuracy analysis of a real-time and an intermittent-scanning continuous glucose monitoring system. J Diabetes Sci Technol, 2021; 15(2):287–293; doi: 10.1177/1932296819895022

Bonora

, Maran

, Ciciliot

, et al. Head-to-head comparison between flash and continuous glucose monitoring systems in outpatients with type 1 diabetes. J Endocrinol Invest, 2016; 39(12):1391–1399; doi: 10.1007/s40618-016-0495-8

Boscari

, Galasso

, Facchinetti

, et al. FreeStyle Libre and Dexcom G4 Platinum sensors: Accuracy comparisons during two weeks of home use and use during experimentally induced glucose excursions. Nutr Metab Cardiovasc Dis, 2018; 28(2):180–186; doi: 10.1016/j.numecd.2017.10.023

Boscari

, Galasso

, Acciaroli

, et al. Head-to-head comparison of the accuracy of Abbott FreeStyle Libre and Dexcom G5 mobile. Nutr Metab Cardiovasc Dis, 2018; 28(4):425–427; doi: 10.1016/j.numecd.2018.01.003

10.

Nagl

, Berger

, Aberer

, et al. Performance of three different continuous glucose monitoring systems in children with type 1 diabetes during a diabetes summer camp. Pediatr Diabetes, 2021; 22(2):271–278; doi: 10.1111/pedi.13160

11.

Tsoukas

, Rutkowski

, El-Fathi

, et al. Accuracy of FreeStyle Libre in adults with type 1 diabetes: The effect of sensor age. Diabetes Technol Ther, 2020; 22(3):203–207; doi: 10.1089/dia.2019.0262

12.

Jafri

, Balliro

, El-Khatib

, et al. Pro, and senseonics eversense continuous glucose monitoring devices in a home-use study of subjects with type 1 diabetes. Diabetes Technol Ther, 2020; 22(11):846–852; doi: 10.1089/dia.2019.0449

13.

Pleus

, Kamecke

, Waldenmaier

, et al. Time in specific glucose ranges, glucose management indicator, and glycemic variability: Impact of Continuous Glucose Monitoring (CGM) system model and sensor on CGM metrics. J Diabetes Sci Technol, 2021; 15(5):1104–1110; doi: 10.1177/1932296820931825

14.

Howard

, Guo

, Hall

. Imprecision nutrition? Different simultaneous continuous glucose monitors provide discordant meal rankings for incremental postprandial glucose in subjects without diabetes. Am J Clin Nutr, 2020; 112(4):1114–1119; doi: 10.1093/ajcn/nqaa198

15.

Freckmann

, Pleus

, Schauer

, et al. Choice of continuous glucose monitoring systems may affect metrics: Clinically relevant differences in times in ranges. Exp Clin Endocrinol Diabetes, 2022; 130(5):343–350; doi: 10.1055/a-1347-2550

16.

Pleus

, Kamecke

, Link

, et al. Flash glucose monitoring: Differences between intermittently scanned and continuously stored data. J Diabetes Sci Technol, 2018; 12(2):397–400; doi: 10.1177/1932296817733095

17.

Deja

, Kłeczek

, Chumięcki

, et al. The usefulness of the Flash Style Libre system in glycemic control in children with type 1 diabetes during summer camp. Pediatr Endocrinol Diabetes Metab, 2018; 24(1):11–19; doi: 10.18544/PEDM-24.01.0098

18.

Boucher

, Gray

, de Bock

, et al. Effect of 6 months’ flash glucose monitoring in adolescents and young adults with type 1 diabetes and suboptimal glycaemic control: Managing diabetes in a ‘flash’randomised controlled trial protocol. BMC Endocr Disord, 2019; 19(1):50; doi: 10.1186/s12902-019-0378-z

19.

Szadkowska

, Michalak

, Łosiewicz

, et al. Impact of factory‐calibrated Freestyle Libre System with new glucose algorithm measurement accuracy and clinical performance in children with type 1 diabetes during summer camp. Pediatr Diabetes, 2021; 22(2):261–270; doi: 10.1111/pedi.13135

20.

Gatti

. Feasibility of FreeStyle Libre flash glucose monitoring system in pregnant woman affected by type 1 diabetes. Acta Diabetol, 2019; 56(4):481–483; doi: 10.1007/s00592-018-1252-6

21.

Moser

, Eckstein

, Mueller

, et al. Impact of physical exercise on sensor performance of the FreeStyle Libre intermittently viewed continuous glucose monitoring system in people with Type 1 diabetes: A randomized crossover trial. Diabet Med, 2019; 36(5):606–611; doi: 10.1111/dme.13909

22.

Moser

, Eckstein

, McCarthy

, et al. Performance of the Freestyle Libre flash glucose monitoring (flash GM) system in individuals with type 1 diabetes: A secondary outcome analysis of a randomized crossover trial. Diabetes Obes Metab, 2019; 21(11):2505–2512; doi: 10.1111/dom.13835

23.

Ghane

, Broadney

, Davis

, et al. Estimating plasma glucose with the freestyle libre pro continuous glucose monitor during oral glucose tolerance tests in youth without diabetes. Pediatr Diabetes, 2019; 20(8):1072–1079; doi: 10.1111/pedi.12910

24.

Angellotti

, Muppavarapu

, Siegel

, et al. The calculation of the glucose management indicator is influenced by the continuous glucose monitoring system and patient race. Diabetes Technol Ther, 2020; 22(9):651–657; doi: 10.1089/dia.2019.0405

25.

Yan

, Li

, Kong

, et al. The accuracy and precision of the continuously stored data from flash glucose monitoring system in type 2 diabetes patients during standard meal tolerance test. Int J Endocrinol, 2020; 2020:5947680; doi: 10.1155/2020/5947680

26.

Kingsnorth

, Whelan

, Orme

, et al. Resistance to data loss from the Freestyle Libre: Impact on glucose variability indices and recommendations for data analysis. Appl Physiol Nutr Metab, 2021; 46(2):148–154; doi: 10.1139/apnm-2020-0386

27.

Kim

, Pham

, Shek

, et al. Differences in glucose level between right arm and left arm using continuous glucose monitors. Digit Health, 2020; 6:2055207620970342; doi: 10.1177/2055207620970342

28.

Vergès

, Pignol

, Rouland

, et al. Glycemic variability assessment with a 14-day continuous glucose monitoring system: When and how long to measure MAGE (Mean Amplitude of Glucose Excursion) for optimal reliability? J Diabetes Sci Technol, 2022; 16(4):982–987; doi: 10.1177/1932296821992060

29.

Avari

, Moscardo

, Jugnee

, et al. Glycemic variability and hypoglycemic excursions with continuous glucose monitoring compared to intermittently scanned continuous glucose monitoring in adults with highest risk type 1 diabetes. J Diabetes Sci Technol, 2020; 14(3):567–574; doi: 10.1177/1932296819867688

30.

Davis

, Dwyer

, England

, et al. Efficacy of intermittently scanned continuous glucose monitoring in the prevention of recurrent severe hypoglycemia. Diabetes Technol Ther, 2020; 22(5):367–373; doi: 10.1089/dia.2019.0331

31.

Nana

, Moore

, Ang

, et al. Flash glucose monitoring: Impact on markers of glycaemic control and patient-reported outcomes in individuals with type 1 diabetes mellitus in the real-world setting. Diabetes Res Clin Pract, 2019; 157:107893; doi: 10.1016/j.diabres.2019.107893

32.

Whelan

, Orme

, Kingsnorth

, et al. Examining the use of glucose and physical activity self-monitoring technologies in individuals at moderate to high risk of developing type 2 diabetes: Randomized trial. JMIR Mhealth Uhealth, 2019; 7(10):e14195; doi: 10.2196/14195

33.

Freckmann

, Link

, Westhoff

, et al. Prediction quality of glucose trend indicators in two continuous tissue glucose monitoring systems. Diabetes Technol Ther, 2018; 20(8):550–556; doi: 10.1089/dia.2018.0112

34.

Kudva

, Ahmann

, Bergenstal

, et al. Approach to using trend arrows in the FreeStyle Libre flash glucose monitoring systems in adults. J Endocr Soc, 2018; 2(12):1320–1337; doi: 10.1210/js.2018-00294

35.

Bianchi

, Aragona

, Rodia

, et al. Freestyle Libre trend arrows for the management of adults with insulin-treated diabetes: A practical approach. J Diabetes Complications, 2019; 33(1):6–12; doi: 10.1016/j.jdiacomp.2018.10.012

36.

Ishihara

, Uchiyama

, Kizaki

, et al. Application of continuous glucose monitoring for assessment of individual carbohydrate requirement during ultramarathon race. Nutrients, 2020; 12(4):1121; doi: 10.3390/nu12041121

37.

Moser

, Mueller

, Eckstein

, et al. Improved glycaemic variability and basal insulin dose reduction during a running competition in recreationally active adults with type 1 diabetes—A single-centre, prospective, controlled observational study. PLoS One, 2020; 15(9):e0239091; doi: 10.1371/journal.pone.0239091

38.

Elhadd

, Mall

, Bashir

, et al.; for PROFAST-Ramadan Study Group. Artificial Intelligence (AI) based machine learning models predict glucose variability and hypoglycaemia risk in patients with type 2 diabetes on a multiple drug regimen who fast during ramadan (The PROFAST–IT Ramadan study). Diabetes Res Clin Pract, 2020; 169:108388; doi: 10.1016/j.diabres.2020.108388

39.

Wan

, Wang

, Liu

, et al. Estimating the sample mean and standard deviation from the sample size, median, range and/or interquartile range. BMC Med Res Methodol, 2014; 14:135; doi: 10.1186/1471-2288-14-135

40.

Luo

, Wan

, Liu

, et al. Optimally estimating the sample mean from the sample size, median, mid-range, and/or mid-quartile range. Stat Methods Med Res, 2018; 27(6):1785–1805; doi: 10.1177/0962280216669183