FreeStyle Libre for glucose monitoring

Bailey et al. (2015)

Study size, design and location

72 adults (median age 48.5) were included in this prospective, within-subject study in 4 US clinical sites. Patients had type 1 or 2 diabetes needing insulin (by pump and/or injections) and stable for at least the past 6 months.

Intervention and comparator(s)

FreeStyle Libre results were compared with:

Key outcomes

Accuracy:

Compared with SMBG, 85.5% of FreeStyle Libre readings were clinically accurate and 99.0% were clinically acceptable on a Clarke error grid, using linear mixed modelling.

Compared with venous blood sampling, 96.5% of FreeStyle Libre readings were clinically accurate and 98.9% were clinically acceptable.

There was 11.4% MARD between FreeStyle Libre and SMBG readings and 12% MARD between FreeStyle Libre and venous blood measurements.

Accuracy was slightly improved when a consensus error grid method was used.

Patient acceptability:

94% of patients reported favourable ratings on 9 out of 9 subjective statements such as sensor wear and pain compared with finger stick.

Skin issues were seen in 202 site exams: moderate to severe itching (0.5% of the time) and moderate erythema (4% of the time).

1% of users reported unacceptable pain on insertion of sensor.

Strengths and limitations

3 variations of grid analysis were employed: Clarke error grid, consensus error grid and continuous glucose error grid without clear explanation why.

Bolinder et al. (2016a)

Study size, design and location

This was an RCT of 241 adults self-managing stable and well-controlled insulin dependent type 1 diabetes from 23 diabetic centres in 5 European countries (Sweden, Austria, Germany, Spain, Holland). Patients known to be allergic to medical grade adhesives were excluded.

Intervention and comparator(s)

FreeStyle Libre compared with SMBG.

Key outcomes

Effectiveness:

Patients using FreeStyle Libre experienced less time in hypoglycaemia than patients using SMBG, averaging 1.24 hours per day (SE 0.24) or 38% less time (p<0.0001) in hypoglycaemia and 1 hour more per day in euglycaemia (p=0.0006).

The number of hypoglycaemic events per day reduced by mean of 0.45 (by over 25%; p<0.0001).

There were no differences in HbA1c between the 2 groups and no differences in mean glucose levels but glucose variability decreased using a number of other measures (BGRI, CV glucose, LBGI, MAGE, CONGA).

The mean number of SMBG tests per day reduced from 5.5 (SD 2.0) to 0.5 (SD 0.7) in the intervention group.

Patient acceptability:

Satisfaction with FreeStyle Libre was higher by 6.1 points on a scale of 0 to 18 compared with SMBG (p<0.0001).

8% of FreeStyle Libre sensor users had non-serious device-related adverse events (itchiness/rash allergy, erythema and oedema).

Strengths and limitations

The RCT included 6-month follow-up.

Neither patients, staff nor investigators were blinded. The population was limited to self-managing stable and well-controlled adults.

Bolinder et al. (2016b)

Study size, design and location

This study was a sub-analysis from the RCT reported by Bolinder et al. (2016a). Outcomes in 18 to 24 year olds (n=19) were compared to ≥25 year olds (n=222) over 6-month follow-up.

Intervention and comparator(s)

FreeStyle Libre compared with SMBG.

Key outcomes

After 6 months of device use there was no significant difference between age groups in the number of times the sensor was scanned daily.

18 to 24 year olds using FreeStyle Libre remained in euglycaemic range for nearly 3 hours longer than those using SMBG: TIR increased significantly by 2.9 (SE 0.89) hours per day (p=0.0055). Adults aged 25 or over significantly increased TIR by 0.9 (SE 0.31) hours per day (p=0.0073).

There was no difference in length of time the sensor was worn (91.4% by 18 to 24 year olds and 92.9% by those aged ≥25 years), and no significant interaction of age with treatment group for the primary end point of time in hypoglycaemia at 6 months.

Strengths and limitations

There were only 10 people aged under 25 years in the intervention group. The study was only available as a conference abstract.

Edge et al. (2017)

Study size, design and location

This was a single-arm prospective within-patient study of 89 4 to 17 year olds with type 1 diabetes treated with multiple daily injections of insulin or continuous insulin infusion, from 9 diabetic centres in the UK.

Patients with a known or suspected allergy to medical grade adhesives were excluded.

Intervention and comparator(s)

FreeStyle Libre compared with SMBG.

Key outcomes

Accuracy:

Compared with SMBG, 83.8% of FreeStyle Libre sensor results were clinically accurate and 99.4% were clinically acceptable on a consensus error grid. MARD was 13.9%.

Patient acceptability:

There was good user acceptability. There were 5 out of 89 mild or moderate sensor-related adverse events (allergy, blister, pink marks, mild bleeding, mild erythema, itching, oedema).

Strengths and limitations

The study ran for 14 days only. Sensor results were masked to patients.

Haak et al. (2017)

Study size, design and location

12-month data from REPLACE RCT (n 224) with adult type 2 diabetes patients having intensive insulin therapy or continuous subcutaneous insulin infusion in 26 sites in Germany (n=10), UK (n=8) and France (n=8).

Exclusions included patients with any severe hypo or hyperglycaemia in the preceding 6 months and patients with allergy to medical adhesives.

Intervention and comparator(s)

Randomised 2:1 to FreeStyle Libre or SMBG.

Key outcomes

Effectiveness:

At 12 months the time in hypoglycaemia reduced by 50% compared with baseline, and the frequency of these events was reduced by 41%. There was no change in the time in hyperglycaemia. The number of SBMG tests reduced from 3.9 tests per day at baseline to 0.1 tests per day at 12 months. The median number of sensor scans per day was 5.7.

9 people reported 16 device-related adverse events, which were all related to the sensor adhesive or site reactions.

Strengths and limitations

Neither patients nor investigators were blinded to group allocation.

Scott and Kautzy-Willer (2017)

Study size, design and location

This was a single-arm prospective within-patient study of 74 pregnant women, 24 with type 1 diabetes, 11 with type 2 and 39 with gestational diabetes from 9 sites in the UK and 4 in Austria. Average gestation was 26.6 weeks. Only 66.2% were taking insulin.

Intervention and comparator(s)

FreeStyle Libre compared with SMBG.

Key outcomes

Clinical accuracy:

88.1% of FreeStyle Libre readings were clinically accurate and 99.8% were clinically acceptable compared with SMBG by consensus error grid analysis. MARD was 11.8%.

Clinically accurate results were reported to be obtained for each type of diabetes.

Patient acceptability:

User satisfaction questionnaires were reported to indicate high levels of acceptance for sensor wear and ease of use of the device. There were no unanticipated device-related adverse events.

Strengths and limitations

Published as a conference abstract with limited data. No information was given on how long the sensor was worn.

Abbreviations: BGRI, blood glucose risk index; CONGA, Continuous overall net glycaemic action; LBGI, low blood glucose index; MAGE, mean amplitude of glycaemic excursions; MARD, mean absolute relative difference (the lower the MARD, the more accurate a device is considered); SMBG, self-monitored blood glucose; TIR, time in range.