DIABETICMedicine DOI: 10.1111/dme.12713
Research: Treatment Continuous glucose monitoring in people with diabetes: the randomized controlled Glucose Level Awareness in Diabetes Study (GLADIS) J. P. New1, R. Ajjan2, A. F. H. Pfeiffer3 and G. Freckmann4 1 Department of Diabetes, Salford Royal NHS Foundation Trust, Salford, 2Division of Cardiovascular and Diabetes Research, The LIGHT Laboratories, University of Leeds and St James University Hospital, Leeds, UK, 3Department of Endocrinology, Diabetes and Nutrition, Charite University Medical School, Berlin and 4Institut €r Diabetes-Technologie Forschungs und Entwicklungsgesellschaft mbH an der Universit€at Ulm, Ulm, Germany fu
Accepted 3 February 2015
Abstract Aims To investigate the best glucose monitoring strategy for maintaining euglycaemia by comparing self-monitoring of blood glucose with continuous glucose monitoring, with or without an alarm function. Methods A 100-day, randomized controlled study was conducted at four European centres, enrolling 160 patients with Type 1 or Type 2 diabetes, on multiple daily insulin injections or continuous subcutaneous insulin infusion. Participants were randomized to continuous glucose monitoring without alarms (n = 48), continuous glucose monitoring with alarms (n = 49) or self-monitoring of blood glucose (n = 48). Results Time spent outside the glucose target during days 80–100 was 9.9 h/day for the continuous glucose monitoring without alarms group, 9.7 h/day for the continuous glucose monitoring with alarms group and 10.6 h/day for the selfmonitoring of blood glucose group (P = 0.18 and 0.08 compared with continuous glucose monitoring without and with alarms, respectively).The continuous glucose monitoring with alarms group spent less time in hypoglycaemia compared with the self-monitoring of blood glucose group (1.0 h/day and 1.6 h/day, respectively; 95% CI 1.2 to 0.1; P = 0.030). Among those treated with continuous subcutaneous insulin infusion, time spent outside the glucose target was significantly different when comparing continuous glucose monitoring without alarms and self-monitoring of blood glucose ( 1.9 h/day; 95% CI 3.8 to 0.0; P = 0.0461) and when comparing continuous glucose monitoring with alarms and self-monitoring of blood glucose ( 2.4 h/day; 95% CI 4.1 to 0.5; P = 0.0134). There was no difference in HbA1c reduction from baseline in the three groups; however, the proportion of participants with a reduction of ≥ 6 mmol/mol (≥ 0.5%) was higher in the continuous glucose monitoring without alarms (27%) and continuous glucose monitoring with alarms groups (25%) than in the self-monitoring of blood glucose group (10.6%). Conclusions This study shows that the use of continuous glucose monitoring reduces time spent outside glucose targets compared with self-monitoring of blood glucose, especially among users of insulin pumps.
Diabet. Med. 00, 000–000 (2015)
Introduction The Diabetes Control and Complications Trial and the UK Prospective Diabetes Study follow-up showed that early, tight glucose control in people with Type 1 and Type 2 diabetes reduced the risk of development or progression of long-term diabetes complications [1,2]. For people using either multiple daily insulin injections (MDI) or continuous subcutaneous insulin infusion (CSII), continuous glucose monitoring (CGM) helps to improve Correspondence to: John New. E-mail: [email protected]
ª 2015 The Authors. Diabetic Medicine ª 2015 Diabetes UK
glycaemic control without increasing the risk of severe hypoglycaemia [3,4]. Indeed, several randomized controlled trials have shown that the use of CGM lowers HbA1c concentrations in people with Type 1 diabetes with no increased risk, and even a reduction in frequency, of hypoglycaemia compared with conventional self-monitoring of blood glucose (SMBG) [4–7]. At a basic level, CGM can alert the user of hypo-/ hyperglycaemia via low and high alarm settings, respectively [8], and CGM device alarms are deemed to be a requirement by both manufacturers and clinicians [9,10]. In a survey among young people with Type 1 diabetes, the continuous
1
DIABETICMedicine
What’s new? • In this study, time spent outside glucose target was not significantly different between continuous glucose monitoring with and without alarm groups. The proportion of participants with an HbA1c concentration reduction ≥ 6 mmol/mol (≥ 0.5%) was similar in the two continuous glucose monitoring groups. • This indicates that, for certain individuals, continuous glucose monitoring without alarms may be as beneficial and safe as continuous glucose monitoring with alarms compared with conventional self-monitoring of blood glucose. • High frequency of alarms, especially false alarms, may lead to alarm fatigue, making users less likely to respond appropriately. data were very popular; however, intermittent users and those who discontinued use described real-time alarms as annoying and as interfering with their lives [11]. High frequency of alarms, especially false alarms, may lead to ‘alarm fatigue’ among users, making them less likely to respond appropriately to hyper- or hypoglycaemia [12]. While CGM technology has been developed and evaluated over the last decade, there has been limited assessment of the utility of alarms in glucose management. Our principal aim in the present study was to investigate the best glucose monitoring strategy for maintaining euglycaemia by comparing SMBG with CGM, with or without an alarm function.
Patients and methods Study cohort
This was a 100-day, prospective, randomized controlled intervention study conducted at four study centres in the UK and Germany. People with Type 1 or Type 2 diabetes on MDI or CSII (> 6 months), aged 18–65 years, with HbA1c values of 53–97 mmol/mol (7–11%), who performed SMBG an average of 2–7 times per day were eligible for inclusion. People with a concomitant disease or a condition influencing metabolic control, participating in another glucose monitoring device study or using drugs that could affect glucose management, who had used CGM in the last 6 months or who were pregnant or planning to be within the planned study duration, were excluded. All participants who used the FreeStyle Navigator Continuous Glucose Monitoring System device (Abbott Diabetes Care, Maidenhead, UK) did so in accordance with product labelling. Patients were enrolled between February 2011 and January 2012; the last participant exited the study in May 2012.
2
Continuous glucose monitoring: the GLADIS study J. P. New et al.
After consent, screening and enrolment, participants completed a 20-day baseline phase during which they selfmanaged blood glucose using the FreeStyle meter built into the masked FreeStyle Navigator, which collected continuous glucose data. Participants also logged their insulin, food and state of health during the 20-day baseline phase.
Randomization
After the masked CGM 20-day baseline phase, participants were randomized by permuted block randomization in blocks of three or six individuals, with stratification by site and type of diabetes, into one of three study arms for the next 80 days, providing they had CGM data for 50% of the baseline period (or at least 1400 individual CGM readings). The CGM with alarms group wore an unmasked FreeStyle Navigator with enabled alarms for the remainder of the study. The CGM without alarms group wore an unmasked FreeStyle Navigator with the low, high and projected alarms switched off (data loss and calibration alarms were still active) and were instructed to leave the alarms disabled for the duration of the study. The control group managed their blood glucose with standard SMBG using a masked FreeStyle Navigator for a further two 20-day periods (study days 40–60 and 80–100). These data were not used by the participant or study staff as part of the participant’s monitoring or management regimens. There is currently no consensus on the management of CGM glucose data in standard care; therefore, advice for using the glucose data to support therapy changes was not offered to any of the participants in the study.
Clinic visits
The first clinic visit was scheduled for 30 days before study start (screening), followed by training (study day 1), sensor removal (day 6) and unmasking, training and clinical oversight (day 21). There were two scheduled interim calls, an interim clinic visit and a final visit at end-of-trial (day 100). The CGM groups had no additional direction or algorithm on the use of CGM data for their therapy. Once participants had completed the study, they were given an option to extend their sensor-wearing by up to 4 weeks with an unmasked FreeStyle Navigator. These data were not included in the study analysis.
Outcome measures Primary outcome
To assess the utility of CGM alarms, the primary outcome of the study was the difference in time spent outside a glucose target of 3.9–10.0 mmol/l (70–180 mg/dl) in the CGM without alarms group vs. the SMBG control group during days 80–100.
ª 2015 The Authors. Diabetic Medicine ª 2015 Diabetes UK
Research article
Secondary outcomes
The secondary outcome measures included: (1) difference in time spent outside the glucose target 3.9–10.0 mmol/l (70–180 mg/dl) in the CGM with alarms vs. CGM without alarms groups during days 80–100; (2) difference in time spent outside of glucose target in the CGM with alarms vs. SMBG groups during days 80–100; (3) HbA1c difference between arms; (4) proportion of participants with a reduction in HbA1c concentration of ≥ 6 mmol/mol (≥ 0.5%) in the three arms; and (5) quality-of-life measures. Safety endpoints included adverse events and insertion-site symptoms. Differences in time spent outside the glucose target are shown for all participants and are grouped by the following subgroups: type of diabetes and insulin administration method (MDI or CSII). An excursion event, for which time spent outside glucose target was recorded, was defined as all consecutive recordings outside the predefined acceptable glucose value boundaries and at least 10 min in duration. The duration of an excursion was defined as the elapsed time from first excursion to the first reading indicating return inside the excursion boundary. The HbA1c values were recorded at baseline and day 100. CGM readings (collected with a masked CGM in the control group) rather than SMBG values were used to assess mean glucose and glycaemic variability. All participants completed the Short-Form-8 Health Survey [13] and the Diabetes Distress Scale questionnaire [14] before any other study activities at both baseline and on day 100. The Short-Form-8 Health Survey assesses eight separate health-related quality of life items, with the results scored on a scale from 0 to 100 (a lower score indicates poorer health-related quality of life). The Diabetes Distress Scale measures diabetes-specific emotional distress, where a mean item score of ≥ 3 (moderate distress) is considered as a level of distress worthy of clinical attention. Sensor-wearing in the two CGM groups was expressed as percentage of total number of days between study visits, included sensors worn from randomization to end of study and excluded withdrawn participants.
Statistical analysis
The primary endpoint for the study was time spent outside of glucose target. To detect a clinically significant difference (1.3 h/day) with hypo-/hyperglycaemic excursions, at the 5% level and a power of 80%, a sample size of 45 participants/ arm would be needed. To allow for 10% dropout, sample size was increased to 50/arm. The primary endpoint was tested for superiority and analyses were performed on the per-protocol population (participants who completed the study as defined in the protocol). Although analyses included all data collected for participants who discontinued prematurely, it cannot be
ª 2015 The Authors. Diabetic Medicine ª 2015 Diabetes UK
DIABETICMedicine
interpreted as an intent-to-treat population because their primary endpoint data were collected when the CGM device was not used. The last observation carried forward method was used for participants in whom an end-of-study HbA1c value was not available; however, missing data were not estimated in the statistical analysis for glucose values recorded with a CGM. Measures of glycaemic variability, except HbA1c, were considered by ANCOVA on baseline measurements, and the difference in between-group HbA1c at 100 days was considered by ANCOVA on baseline HbA1c, both allowing for the study site. Differences between study arms are presented as the difference in least-squares mean from control. All statistical tests and associated P values were two-sided. Analyses were conducted using the GLM procedures in SAS version 9.2 (SAS Institute Inc., Cary, NC, USA).
Results Study population
A total of 160 people provided consent and were enrolled in the study and 145 were randomized to CGM without alarms (n = 48), CGM with alarms (n = 49) or SMBG (n = 48). Participant disposition is shown in Fig. 1. Baseline characteristics were similar among the three arms, except for a slight among-group discrepancy regarding gender: 60, 43 and 58% men in the CGM without alarms, CGM with alarms and SMBG groups, respectively (Table 1).
Time spent outside glucose target
The adjusted mean values on days 80–100 for time spent outside glucose target were 9.9 and 10.6 h/day in the CGM without alarms and SMBG groups, respectively. There was no statistically significant difference between the CGM with or without alarms vs. the SMBG group (Table 2). Significantly less time was spent in hypoglycaemia (< 3.9 mmol/l [< 70 mg/dl]) by participants in the CGM with alarms group than by those in the SMBG group (1.0 vs. 1.6 h/day; 95% CI for difference 1.2 to 0.1; P = 0.030). For participants in the CGM without alarms group, the shorter time spent below 3.9 mmol/l (70 mg/dl) was not statistically significant vs. the SMBG group (1.3 vs. 1.6 h/day; 95% CI for difference 0.8 to 0.3; P = 0.349). When comparing time spent outside glucose target in participants with Type 1 diabetes, the adjusted mean for days 80–100 was 9.9 h/day, 9.6 h/day and 11.0 h/day in the CGM without alarms, CGM with alarms and SMBG groups, respectively. The difference was 1.0 h/day (95% CI 2.2 to 0.0; P = 0.0591) when comparing the CGM without alarms group with the SMBG group, and 1.4 h/day (95% CI 2.5 to 0.3; P = 0.0149) when comparing the CGM with alarms group with the SMBG group. There was no significant
3
DIABETICMedicine
Continuous glucose monitoring: the GLADIS study J. P. New et al.
FIGURE 1 Participant disposition. CGM, continuous glucose monitoring; SMBG, self-monitoring of blood glucose.
difference between the two CGM groups with regard to participants with Type 1 and Type 2 diabetes. For participants using CSII, time spent outside glucose target was significantly different for the CGM without alarms vs. the SMBG group ( 1.9 h/day; 95% CI 3.8 to 0.0; P = 0.0461) and for the CGM with alarms vs. the SMBG group ( 2.4 h/day; 95% CI 4.1 to 0.5; P = 0.0134; Fig. 2a). For participants using MDI, the amount of time spent outside target in the CGM without alarms group was shown to be on average no more than 1.1 h/day greater than in the SMBG group, as determined by the 95% upper confidence limit (Table 2).
and 63 9 mmol/mol (8.0 0.8%), 65 13 mmol/mol (8.1 1.2%) and 64 10 mmol/mol (8.0 1.0%), respectively, for days 80–100. There was no significant difference in the change from baseline in either betweengroup comparison. There tended to be a difference in the proportion of participants who achieved a reduction in HbA1c concentration of ≥ 6 mmol/mol (≥ 0.5%) between the CGM without alarms vs. the SMBG group (P = 0.0652). The proportion of participants who achieved an HbA1c reduction ≥ 6 mmol/mol (≥ 0.5%) in the overall population and by insulin administration is shown in Fig. 2b.
Quality-of-life assessments HbA1c
In the intent-to-treat analysis of all participants, mean SD HbA1c concentration in the CGM without alarms, CGM with alarms and SMBG groups was 65 9 mmol/mol (8.1 0.8%), 66 14 mmol/mol (8.2 1.3%) and 63 11 mmol/mol (8.0 1.0%), respectively, at baseline
4
There was no significant difference in change in mental component summary scores between the three groups. There was a significantly different improvement in physical component summary score from baseline to day 100 in the CGM with alarms group compared with the SMBG group (3.6; P = 0.0245; Table 3). There were no significant differences ª 2015 The Authors. Diabetic Medicine ª 2015 Diabetes UK
Research article
DIABETICMedicine
Table 1 Baseline characteristics of randomized participants CGM without alarms (n = 48) Type of diabetes, n (%) Type 1 42 (87.5) Type 2 6 (12.5) Gender, n (%) Female 19 (39.6) Male 29 (60.4) Median (range) age, 47 (19–65) years BMI, kg/m2 28.5 5.5* Insulin administration, n (%) CSII 13 (27.1) MDI 35 (72.9) Previous use of CGM, n (%) No 39 (81.3) Yes 9 (18.7) HbA1c, mmol/mol 65.7 9.7* HbA1c, % 8.2 0.9* No. of self-reported 5.0 1.6 blood glucose tests/day
CGM with alarms (n = 49)
SMBG (n = 48)
42 (85.7) 7 (14.3)
42 (87.5) 6 (12.5)
28 (57.1) 21 (42.9) 47 (20–65)
20 (41.7) 28 (58.3) 42 (18–65)
27.1 5.8
All participants (n = 145)
126 (86.9) 19 (13.1) 67 (46.2) 78 (53.8) 47 (18–65) 27.2 5.5†
25.9 4.9
16 (32.7) 33 (67.3)
16 (33.3) 32 (66.7)
40 (81.6) 9 (18.4) 68.7 14.4 8.4 1.3 4.9 2.0
36 (75.0) 12 (25.0) 65.9 11.2 8.2 1.0 5.3 1.8
45 (31.0) 100 (69.0) 115 (79.3) 30 (20.7) 66.8 11.9† 8.2 1.1† 5.1 1.8
All values are mean SD unless otherwise specified. CGM, continuous glucose monitoring; CSII, continuous subcutaneous insulin infusion; MDI, multiple daily injections; SMBG, self-monitoring of blood glucose *n = 47; †n = 144.
Table 2 Time spent outside glucose target CGM without alarms
CGM with alarms
SMBG
Time spent outside target (h/day) All participants Baseline Day 80–100 Day 80–100 (adjusted mean) Type 1 diabetes Baseline Day 80–100 Day 80–100 (adjusted mean) Type 2 diabetes Baseline Day 80–100 Day 80–100 (adjusted mean) Insulin delivery with CSII Baseline Day 80–100 Day 80–100 (adjusted mean) Insulin delivery with MDI Baseline Day 80–100 Day 80–100 (adjusted mean)
CGM without alarms vs. CGM with alarms
CGM without alarms vs. SMBG
CGM with alarms vs. SMBG
95% CI
95% CI
95% CI
P
P
n = 45 10.0 3.5 9.6 4.1 9.9
n = 44 10.3 3.3 9.6 3.7 9.7
n = 39 10.5 3.2 10.8 3.7 10.6
0.85; 1.30
0.6762
1.86; 0.35
0.1785
n = 39 10.3 3.5 9.7 4.3 9.9
n = 38 10.5 3.1 9.6 3.5 9.6
n = 33 10.6 3.3 11.1 3.8 11.0
0.75; 1.41
0.5460
2.19; 0.04
0.0591
n=6 8.0 2.5 9.0 2.8 9.3
n=6 9.0 4.6 10.0 5.0 9.4
n=6 9.9 2.8 9.3 3.5 8.6
5.10; 5.04
0.9892
4.19; 5.59
0.7588
n = 12
n = 15
n = 13
10.4 2.7 9.3 3.6 9.1
10.5 3.0 9.1 3.2 8.8
9.5 3.1 10.3 4.1 11.0
1.45;2.15
0.6939
n = 33
n = 29
n = 26
9.9 3.8 9.7 4.3 10.2
10.2 3.5 9.9 3.9 10.1
11.0 3.2 11.1 3.6 10.5
1.27; 1.46
0.8905
3.84;
0.04
0.0461
1.79; 1.07
0.6168
P
2.10; 0.13
2.52;
0.0824
0.28
0.0149
4.42; 5.89
0.7689
4.07;
0.51
0.0134
1.92; 1.00
0.5365
All values are mean SD unless otherwise specified. Adjusted means from ANCOVA allowing for covariates baseline and study site. CGM, continuous glucose monitoring; CSII, continuous subcutaneous insulin infusion; MDI, multiple daily injections; SMBG, selfmonitoring of blood glucose.
ª 2015 The Authors. Diabetic Medicine ª 2015 Diabetes UK
5
DIABETICMedicine
(a)
Continuous glucose monitoring: the GLADIS study J. P. New et al.
alarms. Both CGM groups wore sensors for > 80% of the time.
Blood glucose measurements and checks
(b)
The frequency of capillary SMBG measurements, including calibration measurements, in the per-protocol group was similar in the three groups at baseline with 3.9, 4.0 and 4.0 measurements/day in the CGM without alarms, CGM with alarms and SMBG groups, respectively; however, on days 80–100, the frequency of capillary SMBG measurement was reduced in the CGM without alarms (1.4/day) and CGM with alarms (1.7/day) groups compared with the SMBG group (3.6/day; P < 0.0001 for both). When looking at number of times/day that participants in the three different groups checked their glucose by either a real-time continuous glucose value or capillary SMBG, the frequency of glucose checks in the two CGM groups was higher at any given time after randomization vs. SMBG (Fig. 3).
Discussion
FIGURE 2 (a) Time spent outside glucose target of 3.9–10.0 mmol/l (70–180 mg/dl) in hyperglycaemia and hypoglycaemia in participants treated with CSII and MDI. (b) Percentage of participants with HbA1c reduction ≥ 6.0 mmol/mol (≥ 0.5%). ns, non-significant; CGM, continuous glucose monitoring; SMBG, self-monitoring of blood glucose; CSII, continuous subcutaneous insulin infusion; MDI, multiple daily insulin injections.
in Diabetes Distress Scale score on day 100 between any of the three groups. None of the mean values for the total score or any of the four separate questionnaire items was > 3 at baseline or on day 100.
Safety analysis
Of the 160 participants who provided consent, 157 had a sensor inserted. The 13 adverse events (in 12 participants) considered related or possibly related to the study device were not serious; all symptoms were connected to sensor-site issues. Of the 157 participants with sensor insertions, the most commonly reported insertion-site symptoms were erythema (26%), bleeding (21%) and itching (15%).
Sensor-wearing and alarm settings
Mean sensor-wearing in the main study phase was 83% in the CGM without alarms group and 90% in the CGM with
6
The present study was designed to evaluate the effect of CGM, with and without alarms enabled, compared with conventional SMBG in time spent outside a predefined glucose target (hyper-/hypoglycaemia). The study builds on a trial by Bode et al. [15] that investigated the benefit of alarms in CGM. In the present study, the effectiveness of CGM with and without alarms in diabetes management was assessed and the main findings can be summarized as follows: (1) time spent outside glucose target, especially hypoglycaemia, was reduced with CGM compared with SMBG, and was related to method of insulin administration; and (2) there was no significant difference in HbA1c concentration between groups; however, the proportion of participants with HbA1c reductions ≥ 6.0 mmol/mol (≥ 0.5%) was higher in the CGM without alarms and in the CGM with alarms groups than with SMBG. The frequency of capillary glucose measurements, comparing baseline with end of study, significantly decreased in the CGM groups vs. the SMBG group. Although the present study showed a reduction in hypoglycaemia, there was no significant change in HbA1c concentration from baseline, indicating that the reduced hypoglycaemia was not caused by a worsening of glucose control. This is in accordance with other clinical trials in people with Type 1 diabetes that have shown that the use of CGM compared with SMBG can improve HbA1c without increasing the risk of hypoglycaemia [4–7]. Although the difference in h/day spent outside glucose targets for the three groups was not statistically significant, time spent outside glucose targets was significantly reduced for participants receiving CSII using either CGM without or with alarms compared with the SMBG group, whereas
ª 2015 The Authors. Diabetic Medicine ª 2015 Diabetes UK
Research article
DIABETICMedicine
Table 3 Mental component summary score and physical component summary score from the Short-Form-8 Health Survey at baseline and visit 6 CGM without alarms
CGM with alarms
SMBG
Time spent outside target (h/day) Mental component summary score Baseline Day 100 Day 100 (adjusted mean) Physical component summary score Baseline Day 100 Day 100 (adjusted mean)
n = 44
n = 43
n = 39
49.1 9.4 50.9 9.4 50.4
47.6 11.2 48.9 11.4 49.2
49.0 10.4 49.3 10.7 48.9
n = 44
n = 43
n = 39
48.6 9.7 49.0 9.8 48.9
46.7 8.8 49.4 9.6 50.7
49.1 7.9 47.5 8.5 47.1
CGM without alarms vs. CGM with alarms
CGM without alarms vs. SMBG
CGM with alarms vs. SMBG
95% CI
95% CI
95% CI
P
P
P
2.41; 4.80
0.5128
2.23; 5.15
0.4350
3.45; 3.98
0.8887
4.85; 1.21
0.2369
1.31; 4.88
0.2556
0.47; 6.73
0.0245
All values are mean SD unless otherwise specified. CGM, continuous glucose monitoring; CSII, continuous subcutaneous insulin infusion; MDI, multiple daily injections; SMBG, selfmonitoring of blood glucose.
FIGURE 3 Frequency of blood glucose checks per day. CGM, continuous glucose monitoring; SMBG, self-monitoring of blood glucose.
results for participants using MDI were similar to those found in the overall participant analysis. When compared with the SMBG group, the odds ratios of participants achieving a reduction in HbA1c ≥ 0.5% were similar for MDI and CSII (3.1 and 2.7, respectively), as the SMBG group percentage was higher in participants using MDI. This difference is likely to be because the baseline HbA1c was lower for CSII users. The availability of accurate correction doses to CSII users could account for the difference between the two insulin administration groups. The nature of the delivery method may mean that CSII users’ exposure to diabetes technology helps them become familiar with CGM more quickly than MDI users; however, early adopters of technology may not be representative of the group [10], and CSII users remain a minority in the adult diabetes population. ª 2015 The Authors. Diabetic Medicine ª 2015 Diabetes UK
Other studies have shown that sensor use time is inversely associated with HbA1c concentration [16], and sensor use of 60–70% is associated with a reduction in HbA1c concentration [3,5–7,11,17,18], In the present study, sensor compliance in both CGM groups was > 80%; however, no statistical reduction in HbA1c was seen. One plausible reason for the failure to observe change in HbA1c concentrations is the limited follow-up period. Compared with isolated glucose values, glucose patterns over time offer an effective strategy to optimize diabetes management [19]. Glycaemic variability may be a risk factor for the development of long-term complications associated with diabetes, and reduced glucose variability may result in short- and long-term improvements in HbA1c control [20]. In the present study, no structured education was provided on how to use CGM for self-management. In addition, a separate study showed no difference in time outside target glucose, but greater change in HbA1c concentration in people with diabetes who received an algorithm to guide CGM responses, compared with those who were not educated on CGM use [16]. Up to 30% of CGM alarms may be false [9]. If the false alarm rate is perceived as high, the corresponding response rate is low or none [12,21,22]. Alarm coping strategies may be necessary to improve sensor use [23]. In the present study, sensor-wearing was numerically higher in the CGM with alarms group vs. CGM without alarms, and there was no significant difference between the two CGM groups with regard to quality-of-life measurements. The frequency of SMBG has been linked with improved glycaemic control [24–26] and is a predictor for concordance with CGM use [18]. The self-reported number of glucoselevel checks per day for the SMBG group unsurprisingly remained the same as baseline; however, the frequency of real-time glucose-data checks for the CGM group was much
7
DIABETICMedicine
higher and increased throughout the course of the study, suggesting increased attention to glucose levels in both CGM groups vs. SMBG. Increased use of CGM data over the study period may indicate that some familiarization is required for patients to use CGM data; however, a CGM system without alarms was demonstrated as a safe replacement for SMBG for everyday self-management. A limitation of the present study was the small number of participants who were included, resulting in multiple hypothesis tests being performed on small subgroups. A practical proportion of participants with Type 2 diabetes was not achieved for the study because of recruitment issues at the study sites, limiting any meaningful subgroup analysis. Multiple tests were performed on time spent outside glucose target for subgroups without formal adjustment of the significance level and these results should be considered exploratory. Despite fewer than the assumed number of participants performing SMBG completing the study, the study still has 77% power to detect a difference of 1.3 h/day for the primary endpoint. In conclusion, this study shows that using CGM with or without alarms reduces patients’ time spent outside glucose targets compared with conventional SMBG. This was especially evident for patients who are on CSII therapy. Although no difference in HbA1c reduction was seen among the three groups because of the limited follow-up, clinically meaningful HbA1c reduction was more frequent for individuals in the CGM groups. In addition, participants in the CMG groups spent less time in hypoglycaemia and experienced less variability in glucose levels than those in the SMBG group.
Funding sources
This work was funded by Abbott Diabetes Care. The clinical investigation was sponsored and designed by Abbott Diabetes Care with input from the Chief Investigator. Abbott Diabetes Care also provided assistance with statistical analysis.
Competing interests
J.P.N., R.A. and G.F. have received research funding and consulting fees from Abbott Diabetes Care. G.F. is General Manager of the Institut f€ ur Diabetes-Technologie, which carries out studies evaluating SMBG systems and medical devices for diabetes therapy on behalf of various companies. A.F.H.P. has no potential conflicts of interest associated with this submission to declare.
Acknowledgements
All authors take full responsibility for the content of this manuscript, but gratefully acknowledge the assistance of the study personnel at the four clinical sites for execution of this
8
Continuous glucose monitoring: the GLADIS study J. P. New et al.
study (Appendix S1). We also thank Zoe Welsh of Abbott Diabetes Care for assistance with statistical analysis, and Daniella Pfeifer and Lisa Sullivan of Watermeadow Medical for assistance with preparation and submission of the manuscript (supported by Abbott Diabetes Care).
References 1 The Diabetes Control and Complications Trial Research Group. The effect of intensive treatment of diabetes on the development and progression of long-term complications in insulin-dependent diabetes mellitus. N Engl J Med 1993; 329: 977–986. 2 Holman RR, Paul SK, Bethel MA, Matthews DR, Neil HA. 10-year follow-up of intensive glucose control in type 2 diabetes. N Engl J Med 2008; 359: 1577–1589. 3 Hirsch IB, Abelseth J, Bode BW, Fischer JS, Kaufman FR, Mastrototaro J et al. Sensor-augmented insulin pump therapy: results of the first randomized treat-to-target study. Diabetes Technol Ther 2008; 10: 377–383. 4 Tamborlane WV, Beck RW, Bode BW, Buckingham B, Chase HP, Clemons R et al. (Juvenile Diabetes Research Foundation Continuous Glucose Monitoring Study Group). Continuous glucose monitoring and intensive treatment of type 1 diabetes. N Engl J Med 2008; 359: 1464–1476. 5 Deiss D, Bolinder J, Riveline JP, Battelino T, Bosi E, Tubiana-Rufi N et al. Improved glycemic control in poorly controlled patients with type 1 diabetes using real-time continuous glucose monitoring. Diabetes Care 2006; 29: 2730–2732. 6 Raccah D, Sulmont V, Reznik Y, Guerci B, Renard E, Hanaire H et al. Incremental value of continuous glucose monitoring when starting pump therapy in patients with poorly controlled type 1 diabetes: the Real Trend study. Diabetes Care 2009; 32: 2245– 2250. 7 O’Connell MA, Donath S, O’Neal DN, Colman PG, Ambler GR, Jones TW et al. Glycaemic impact of patient-led use of sensorguided pump therapy in type 1 diabetes: a randomised controlled trial. Diabetologia 2009; 52: 1250–1257. 8 De Salvo D, Buckingham B. Continuous glucose monitoring: Current use and future directions. Curr Diab Rep 2013; 13: 657– 662. 9 Kamath A, Mahalingam A, Brauker J. Methods of evaluating the utility of continuous glucose monitor alerts. J Diabetes Sci Technol 2010; 4: 57–66. 10 Gandhi GY, Kovalaske M, Kudva Y, Walsh K, Elamin MB, Beers M et al. Efficacy of continuous glucose monitoring in improving glycemic control and reducing hypoglycemia: a systematic review and meta-analysis of randomized trials. J Diabetes Sci Technol 2011; 5: 952–965. 11 Ramchandani N, Arya S, Ten S, Bhandari S. Real-life utilization of real-time continuous glucose monitoring: the complete picture. J Diabetes Sci Technol 2011; 5: 860–870. 12 Shivers JP, Mackowiak L, Anhalt H, Zisser H. “Turn it off!”: diabetes device alarm fatigue considerations for the present and the future. J Diabetes Sci Technol 2013; 7: 789–794. 13 Ware JE, Kosinski M, Dewey JE, Gandek B. How to Score and Interpret Single-Item Health Status Measures: A Manual For Users of the SF-8TM Health Survey. Lincoln: QualityMetric, 2001. 14 Polonsky WH, Anderson BJ, Lohrer PA, Welch G, Jacobson AM, Aponte JE et al. Assessment of diabetes-related distress. Diabetes Care 1995; 18: 754–760. 15 Bode B, Gross K, Rikalo N, Schwartz S, Wahl T, Page C et al. Alarms based on real-time sensor glucose values alert patients to hypo- and hyperglycemia: the guardian continuous monitoring system. Diabetes Technol Ther 2004; 6: 105–113.
ª 2015 The Authors. Diabetic Medicine ª 2015 Diabetes UK
Research article
16 Jenkins AJ, Krishnamurthy B, Best JD, Cameron FJ, Colman PG, Farish S et al. Evaluation of an algorithm to guide patients with type 1 diabetes treated with continuous subcutaneous insulin infusion on how to respond to real-time continuous glucose levels: a randomized controlled trial. Diabetes Care 2010; 33: 1242–1248. 17 Bailey TS, Zisser HC, Garg SK. Reduction in hemoglobin A1C with real-time continuous glucose monitoring: results from a 12-week observational study. Diabetes Technol Ther 2007; 9: 203–210. 18 Juvenile Diabetes Research Foundation Continuous Glucose Monitoring Study Group, Fiallo-Scharer R, Cheng J, Beck RW, Buckingham BA, Chase HP et al. Factors predictive of severe hypoglycemia in type 1 diabetes: analysis from the Juvenile Diabetes Research Foundation continuous glucose monitoring randomized control trial dataset. Diabetes Care 2011; 34: 586–590. 19 Jovanovic L. Using meal-based self-monitoring blood glucose (SMBG) data to guide dietary recommendations in patients with diabetes. Diabetes Educ 2009; 35: 1023–1030. 20 Bode B, Silver M, Weiss R, Martin K. Evaluation of a continuous glucose monitoring system for home-use conditions. Manag Care 2008; 17: 40–45. 21 Bliss JP, Gilson RD, Deaton JE. Human probability matching behaviour in response to alarms of varying reliability. Ergonomics 1995; 38: 2300–2312.
ª 2015 The Authors. Diabetic Medicine ª 2015 Diabetes UK
DIABETICMedicine
22 Bliss JP, Dunn MC. Behavioural implications of alarm mistrust as a function of task workload. Ergonomics 2000; 43: 1283–1300. 23 Schmidt S, Duun-Henriksen AK, Nørgaard K. Psychosocial factors and adherence to continuous glucose monitoring in type 1 diabetes. J Diabetes Sci Technol 2012; 6: 986–987. 24 Karter AJ, Ackerson LM, Darbinian JA, D’Agostino RB Jr, Ferrara A, Liu J et al. Self-monitoring of blood glucose levels and glycemic control: the Northern California Kaiser Permanente Diabetes registry. Am J Med 2001; 111: 1–9. 25 Karter AJ, Parker MM, Moffet HH, Spence MM, Chan J, Ettner SL et al. Longitudinal study of new and prevalent use of selfmonitoring of blood glucose. Diabetes Care 2006; 29: 1757–1763. 26 Blonde L, Karter AJ. Current evidence regarding the value of selfmonitored blood glucose testing. Am J Med 2005; 118(Suppl. 9A): 20S–26S.
Supporting Information Additional Supporting Information may be found in the online version of this article: Appendix S1. List of contributors to collection of data at the GLADIS study sites.
9
Research: Treatment Continuous glucose monitoring in people with diabetes: the randomized controlled Glucose Level Awareness in Diabetes Study (GLADIS) J. P. New1, R. Ajjan2, A. F. H. Pfeiffer3 and G. Freckmann4 1 Department of Diabetes, Salford Royal NHS Foundation Trust, Salford, 2Division of Cardiovascular and Diabetes Research, The LIGHT Laboratories, University of Leeds and St James University Hospital, Leeds, UK, 3Department of Endocrinology, Diabetes and Nutrition, Charite University Medical School, Berlin and 4Institut €r Diabetes-Technologie Forschungs und Entwicklungsgesellschaft mbH an der Universit€at Ulm, Ulm, Germany fu
Accepted 3 February 2015
Abstract Aims To investigate the best glucose monitoring strategy for maintaining euglycaemia by comparing self-monitoring of blood glucose with continuous glucose monitoring, with or without an alarm function. Methods A 100-day, randomized controlled study was conducted at four European centres, enrolling 160 patients with Type 1 or Type 2 diabetes, on multiple daily insulin injections or continuous subcutaneous insulin infusion. Participants were randomized to continuous glucose monitoring without alarms (n = 48), continuous glucose monitoring with alarms (n = 49) or self-monitoring of blood glucose (n = 48). Results Time spent outside the glucose target during days 80–100 was 9.9 h/day for the continuous glucose monitoring without alarms group, 9.7 h/day for the continuous glucose monitoring with alarms group and 10.6 h/day for the selfmonitoring of blood glucose group (P = 0.18 and 0.08 compared with continuous glucose monitoring without and with alarms, respectively).The continuous glucose monitoring with alarms group spent less time in hypoglycaemia compared with the self-monitoring of blood glucose group (1.0 h/day and 1.6 h/day, respectively; 95% CI 1.2 to 0.1; P = 0.030). Among those treated with continuous subcutaneous insulin infusion, time spent outside the glucose target was significantly different when comparing continuous glucose monitoring without alarms and self-monitoring of blood glucose ( 1.9 h/day; 95% CI 3.8 to 0.0; P = 0.0461) and when comparing continuous glucose monitoring with alarms and self-monitoring of blood glucose ( 2.4 h/day; 95% CI 4.1 to 0.5; P = 0.0134). There was no difference in HbA1c reduction from baseline in the three groups; however, the proportion of participants with a reduction of ≥ 6 mmol/mol (≥ 0.5%) was higher in the continuous glucose monitoring without alarms (27%) and continuous glucose monitoring with alarms groups (25%) than in the self-monitoring of blood glucose group (10.6%). Conclusions This study shows that the use of continuous glucose monitoring reduces time spent outside glucose targets compared with self-monitoring of blood glucose, especially among users of insulin pumps.
Diabet. Med. 00, 000–000 (2015)
Introduction The Diabetes Control and Complications Trial and the UK Prospective Diabetes Study follow-up showed that early, tight glucose control in people with Type 1 and Type 2 diabetes reduced the risk of development or progression of long-term diabetes complications [1,2]. For people using either multiple daily insulin injections (MDI) or continuous subcutaneous insulin infusion (CSII), continuous glucose monitoring (CGM) helps to improve Correspondence to: John New. E-mail: [email protected]
ª 2015 The Authors. Diabetic Medicine ª 2015 Diabetes UK
glycaemic control without increasing the risk of severe hypoglycaemia [3,4]. Indeed, several randomized controlled trials have shown that the use of CGM lowers HbA1c concentrations in people with Type 1 diabetes with no increased risk, and even a reduction in frequency, of hypoglycaemia compared with conventional self-monitoring of blood glucose (SMBG) [4–7]. At a basic level, CGM can alert the user of hypo-/ hyperglycaemia via low and high alarm settings, respectively [8], and CGM device alarms are deemed to be a requirement by both manufacturers and clinicians [9,10]. In a survey among young people with Type 1 diabetes, the continuous
1
DIABETICMedicine
What’s new? • In this study, time spent outside glucose target was not significantly different between continuous glucose monitoring with and without alarm groups. The proportion of participants with an HbA1c concentration reduction ≥ 6 mmol/mol (≥ 0.5%) was similar in the two continuous glucose monitoring groups. • This indicates that, for certain individuals, continuous glucose monitoring without alarms may be as beneficial and safe as continuous glucose monitoring with alarms compared with conventional self-monitoring of blood glucose. • High frequency of alarms, especially false alarms, may lead to alarm fatigue, making users less likely to respond appropriately. data were very popular; however, intermittent users and those who discontinued use described real-time alarms as annoying and as interfering with their lives [11]. High frequency of alarms, especially false alarms, may lead to ‘alarm fatigue’ among users, making them less likely to respond appropriately to hyper- or hypoglycaemia [12]. While CGM technology has been developed and evaluated over the last decade, there has been limited assessment of the utility of alarms in glucose management. Our principal aim in the present study was to investigate the best glucose monitoring strategy for maintaining euglycaemia by comparing SMBG with CGM, with or without an alarm function.
Patients and methods Study cohort
This was a 100-day, prospective, randomized controlled intervention study conducted at four study centres in the UK and Germany. People with Type 1 or Type 2 diabetes on MDI or CSII (> 6 months), aged 18–65 years, with HbA1c values of 53–97 mmol/mol (7–11%), who performed SMBG an average of 2–7 times per day were eligible for inclusion. People with a concomitant disease or a condition influencing metabolic control, participating in another glucose monitoring device study or using drugs that could affect glucose management, who had used CGM in the last 6 months or who were pregnant or planning to be within the planned study duration, were excluded. All participants who used the FreeStyle Navigator Continuous Glucose Monitoring System device (Abbott Diabetes Care, Maidenhead, UK) did so in accordance with product labelling. Patients were enrolled between February 2011 and January 2012; the last participant exited the study in May 2012.
2
Continuous glucose monitoring: the GLADIS study J. P. New et al.
After consent, screening and enrolment, participants completed a 20-day baseline phase during which they selfmanaged blood glucose using the FreeStyle meter built into the masked FreeStyle Navigator, which collected continuous glucose data. Participants also logged their insulin, food and state of health during the 20-day baseline phase.
Randomization
After the masked CGM 20-day baseline phase, participants were randomized by permuted block randomization in blocks of three or six individuals, with stratification by site and type of diabetes, into one of three study arms for the next 80 days, providing they had CGM data for 50% of the baseline period (or at least 1400 individual CGM readings). The CGM with alarms group wore an unmasked FreeStyle Navigator with enabled alarms for the remainder of the study. The CGM without alarms group wore an unmasked FreeStyle Navigator with the low, high and projected alarms switched off (data loss and calibration alarms were still active) and were instructed to leave the alarms disabled for the duration of the study. The control group managed their blood glucose with standard SMBG using a masked FreeStyle Navigator for a further two 20-day periods (study days 40–60 and 80–100). These data were not used by the participant or study staff as part of the participant’s monitoring or management regimens. There is currently no consensus on the management of CGM glucose data in standard care; therefore, advice for using the glucose data to support therapy changes was not offered to any of the participants in the study.
Clinic visits
The first clinic visit was scheduled for 30 days before study start (screening), followed by training (study day 1), sensor removal (day 6) and unmasking, training and clinical oversight (day 21). There were two scheduled interim calls, an interim clinic visit and a final visit at end-of-trial (day 100). The CGM groups had no additional direction or algorithm on the use of CGM data for their therapy. Once participants had completed the study, they were given an option to extend their sensor-wearing by up to 4 weeks with an unmasked FreeStyle Navigator. These data were not included in the study analysis.
Outcome measures Primary outcome
To assess the utility of CGM alarms, the primary outcome of the study was the difference in time spent outside a glucose target of 3.9–10.0 mmol/l (70–180 mg/dl) in the CGM without alarms group vs. the SMBG control group during days 80–100.
ª 2015 The Authors. Diabetic Medicine ª 2015 Diabetes UK
Research article
Secondary outcomes
The secondary outcome measures included: (1) difference in time spent outside the glucose target 3.9–10.0 mmol/l (70–180 mg/dl) in the CGM with alarms vs. CGM without alarms groups during days 80–100; (2) difference in time spent outside of glucose target in the CGM with alarms vs. SMBG groups during days 80–100; (3) HbA1c difference between arms; (4) proportion of participants with a reduction in HbA1c concentration of ≥ 6 mmol/mol (≥ 0.5%) in the three arms; and (5) quality-of-life measures. Safety endpoints included adverse events and insertion-site symptoms. Differences in time spent outside the glucose target are shown for all participants and are grouped by the following subgroups: type of diabetes and insulin administration method (MDI or CSII). An excursion event, for which time spent outside glucose target was recorded, was defined as all consecutive recordings outside the predefined acceptable glucose value boundaries and at least 10 min in duration. The duration of an excursion was defined as the elapsed time from first excursion to the first reading indicating return inside the excursion boundary. The HbA1c values were recorded at baseline and day 100. CGM readings (collected with a masked CGM in the control group) rather than SMBG values were used to assess mean glucose and glycaemic variability. All participants completed the Short-Form-8 Health Survey [13] and the Diabetes Distress Scale questionnaire [14] before any other study activities at both baseline and on day 100. The Short-Form-8 Health Survey assesses eight separate health-related quality of life items, with the results scored on a scale from 0 to 100 (a lower score indicates poorer health-related quality of life). The Diabetes Distress Scale measures diabetes-specific emotional distress, where a mean item score of ≥ 3 (moderate distress) is considered as a level of distress worthy of clinical attention. Sensor-wearing in the two CGM groups was expressed as percentage of total number of days between study visits, included sensors worn from randomization to end of study and excluded withdrawn participants.
Statistical analysis
The primary endpoint for the study was time spent outside of glucose target. To detect a clinically significant difference (1.3 h/day) with hypo-/hyperglycaemic excursions, at the 5% level and a power of 80%, a sample size of 45 participants/ arm would be needed. To allow for 10% dropout, sample size was increased to 50/arm. The primary endpoint was tested for superiority and analyses were performed on the per-protocol population (participants who completed the study as defined in the protocol). Although analyses included all data collected for participants who discontinued prematurely, it cannot be
ª 2015 The Authors. Diabetic Medicine ª 2015 Diabetes UK
DIABETICMedicine
interpreted as an intent-to-treat population because their primary endpoint data were collected when the CGM device was not used. The last observation carried forward method was used for participants in whom an end-of-study HbA1c value was not available; however, missing data were not estimated in the statistical analysis for glucose values recorded with a CGM. Measures of glycaemic variability, except HbA1c, were considered by ANCOVA on baseline measurements, and the difference in between-group HbA1c at 100 days was considered by ANCOVA on baseline HbA1c, both allowing for the study site. Differences between study arms are presented as the difference in least-squares mean from control. All statistical tests and associated P values were two-sided. Analyses were conducted using the GLM procedures in SAS version 9.2 (SAS Institute Inc., Cary, NC, USA).
Results Study population
A total of 160 people provided consent and were enrolled in the study and 145 were randomized to CGM without alarms (n = 48), CGM with alarms (n = 49) or SMBG (n = 48). Participant disposition is shown in Fig. 1. Baseline characteristics were similar among the three arms, except for a slight among-group discrepancy regarding gender: 60, 43 and 58% men in the CGM without alarms, CGM with alarms and SMBG groups, respectively (Table 1).
Time spent outside glucose target
The adjusted mean values on days 80–100 for time spent outside glucose target were 9.9 and 10.6 h/day in the CGM without alarms and SMBG groups, respectively. There was no statistically significant difference between the CGM with or without alarms vs. the SMBG group (Table 2). Significantly less time was spent in hypoglycaemia (< 3.9 mmol/l [< 70 mg/dl]) by participants in the CGM with alarms group than by those in the SMBG group (1.0 vs. 1.6 h/day; 95% CI for difference 1.2 to 0.1; P = 0.030). For participants in the CGM without alarms group, the shorter time spent below 3.9 mmol/l (70 mg/dl) was not statistically significant vs. the SMBG group (1.3 vs. 1.6 h/day; 95% CI for difference 0.8 to 0.3; P = 0.349). When comparing time spent outside glucose target in participants with Type 1 diabetes, the adjusted mean for days 80–100 was 9.9 h/day, 9.6 h/day and 11.0 h/day in the CGM without alarms, CGM with alarms and SMBG groups, respectively. The difference was 1.0 h/day (95% CI 2.2 to 0.0; P = 0.0591) when comparing the CGM without alarms group with the SMBG group, and 1.4 h/day (95% CI 2.5 to 0.3; P = 0.0149) when comparing the CGM with alarms group with the SMBG group. There was no significant
3
DIABETICMedicine
Continuous glucose monitoring: the GLADIS study J. P. New et al.
FIGURE 1 Participant disposition. CGM, continuous glucose monitoring; SMBG, self-monitoring of blood glucose.
difference between the two CGM groups with regard to participants with Type 1 and Type 2 diabetes. For participants using CSII, time spent outside glucose target was significantly different for the CGM without alarms vs. the SMBG group ( 1.9 h/day; 95% CI 3.8 to 0.0; P = 0.0461) and for the CGM with alarms vs. the SMBG group ( 2.4 h/day; 95% CI 4.1 to 0.5; P = 0.0134; Fig. 2a). For participants using MDI, the amount of time spent outside target in the CGM without alarms group was shown to be on average no more than 1.1 h/day greater than in the SMBG group, as determined by the 95% upper confidence limit (Table 2).
and 63 9 mmol/mol (8.0 0.8%), 65 13 mmol/mol (8.1 1.2%) and 64 10 mmol/mol (8.0 1.0%), respectively, for days 80–100. There was no significant difference in the change from baseline in either betweengroup comparison. There tended to be a difference in the proportion of participants who achieved a reduction in HbA1c concentration of ≥ 6 mmol/mol (≥ 0.5%) between the CGM without alarms vs. the SMBG group (P = 0.0652). The proportion of participants who achieved an HbA1c reduction ≥ 6 mmol/mol (≥ 0.5%) in the overall population and by insulin administration is shown in Fig. 2b.
Quality-of-life assessments HbA1c
In the intent-to-treat analysis of all participants, mean SD HbA1c concentration in the CGM without alarms, CGM with alarms and SMBG groups was 65 9 mmol/mol (8.1 0.8%), 66 14 mmol/mol (8.2 1.3%) and 63 11 mmol/mol (8.0 1.0%), respectively, at baseline
4
There was no significant difference in change in mental component summary scores between the three groups. There was a significantly different improvement in physical component summary score from baseline to day 100 in the CGM with alarms group compared with the SMBG group (3.6; P = 0.0245; Table 3). There were no significant differences ª 2015 The Authors. Diabetic Medicine ª 2015 Diabetes UK
Research article
DIABETICMedicine
Table 1 Baseline characteristics of randomized participants CGM without alarms (n = 48) Type of diabetes, n (%) Type 1 42 (87.5) Type 2 6 (12.5) Gender, n (%) Female 19 (39.6) Male 29 (60.4) Median (range) age, 47 (19–65) years BMI, kg/m2 28.5 5.5* Insulin administration, n (%) CSII 13 (27.1) MDI 35 (72.9) Previous use of CGM, n (%) No 39 (81.3) Yes 9 (18.7) HbA1c, mmol/mol 65.7 9.7* HbA1c, % 8.2 0.9* No. of self-reported 5.0 1.6 blood glucose tests/day
CGM with alarms (n = 49)
SMBG (n = 48)
42 (85.7) 7 (14.3)
42 (87.5) 6 (12.5)
28 (57.1) 21 (42.9) 47 (20–65)
20 (41.7) 28 (58.3) 42 (18–65)
27.1 5.8
All participants (n = 145)
126 (86.9) 19 (13.1) 67 (46.2) 78 (53.8) 47 (18–65) 27.2 5.5†
25.9 4.9
16 (32.7) 33 (67.3)
16 (33.3) 32 (66.7)
40 (81.6) 9 (18.4) 68.7 14.4 8.4 1.3 4.9 2.0
36 (75.0) 12 (25.0) 65.9 11.2 8.2 1.0 5.3 1.8
45 (31.0) 100 (69.0) 115 (79.3) 30 (20.7) 66.8 11.9† 8.2 1.1† 5.1 1.8
All values are mean SD unless otherwise specified. CGM, continuous glucose monitoring; CSII, continuous subcutaneous insulin infusion; MDI, multiple daily injections; SMBG, self-monitoring of blood glucose *n = 47; †n = 144.
Table 2 Time spent outside glucose target CGM without alarms
CGM with alarms
SMBG
Time spent outside target (h/day) All participants Baseline Day 80–100 Day 80–100 (adjusted mean) Type 1 diabetes Baseline Day 80–100 Day 80–100 (adjusted mean) Type 2 diabetes Baseline Day 80–100 Day 80–100 (adjusted mean) Insulin delivery with CSII Baseline Day 80–100 Day 80–100 (adjusted mean) Insulin delivery with MDI Baseline Day 80–100 Day 80–100 (adjusted mean)
CGM without alarms vs. CGM with alarms
CGM without alarms vs. SMBG
CGM with alarms vs. SMBG
95% CI
95% CI
95% CI
P
P
n = 45 10.0 3.5 9.6 4.1 9.9
n = 44 10.3 3.3 9.6 3.7 9.7
n = 39 10.5 3.2 10.8 3.7 10.6
0.85; 1.30
0.6762
1.86; 0.35
0.1785
n = 39 10.3 3.5 9.7 4.3 9.9
n = 38 10.5 3.1 9.6 3.5 9.6
n = 33 10.6 3.3 11.1 3.8 11.0
0.75; 1.41
0.5460
2.19; 0.04
0.0591
n=6 8.0 2.5 9.0 2.8 9.3
n=6 9.0 4.6 10.0 5.0 9.4
n=6 9.9 2.8 9.3 3.5 8.6
5.10; 5.04
0.9892
4.19; 5.59
0.7588
n = 12
n = 15
n = 13
10.4 2.7 9.3 3.6 9.1
10.5 3.0 9.1 3.2 8.8
9.5 3.1 10.3 4.1 11.0
1.45;2.15
0.6939
n = 33
n = 29
n = 26
9.9 3.8 9.7 4.3 10.2
10.2 3.5 9.9 3.9 10.1
11.0 3.2 11.1 3.6 10.5
1.27; 1.46
0.8905
3.84;
0.04
0.0461
1.79; 1.07
0.6168
P
2.10; 0.13
2.52;
0.0824
0.28
0.0149
4.42; 5.89
0.7689
4.07;
0.51
0.0134
1.92; 1.00
0.5365
All values are mean SD unless otherwise specified. Adjusted means from ANCOVA allowing for covariates baseline and study site. CGM, continuous glucose monitoring; CSII, continuous subcutaneous insulin infusion; MDI, multiple daily injections; SMBG, selfmonitoring of blood glucose.
ª 2015 The Authors. Diabetic Medicine ª 2015 Diabetes UK
5
DIABETICMedicine
(a)
Continuous glucose monitoring: the GLADIS study J. P. New et al.
alarms. Both CGM groups wore sensors for > 80% of the time.
Blood glucose measurements and checks
(b)
The frequency of capillary SMBG measurements, including calibration measurements, in the per-protocol group was similar in the three groups at baseline with 3.9, 4.0 and 4.0 measurements/day in the CGM without alarms, CGM with alarms and SMBG groups, respectively; however, on days 80–100, the frequency of capillary SMBG measurement was reduced in the CGM without alarms (1.4/day) and CGM with alarms (1.7/day) groups compared with the SMBG group (3.6/day; P < 0.0001 for both). When looking at number of times/day that participants in the three different groups checked their glucose by either a real-time continuous glucose value or capillary SMBG, the frequency of glucose checks in the two CGM groups was higher at any given time after randomization vs. SMBG (Fig. 3).
Discussion
FIGURE 2 (a) Time spent outside glucose target of 3.9–10.0 mmol/l (70–180 mg/dl) in hyperglycaemia and hypoglycaemia in participants treated with CSII and MDI. (b) Percentage of participants with HbA1c reduction ≥ 6.0 mmol/mol (≥ 0.5%). ns, non-significant; CGM, continuous glucose monitoring; SMBG, self-monitoring of blood glucose; CSII, continuous subcutaneous insulin infusion; MDI, multiple daily insulin injections.
in Diabetes Distress Scale score on day 100 between any of the three groups. None of the mean values for the total score or any of the four separate questionnaire items was > 3 at baseline or on day 100.
Safety analysis
Of the 160 participants who provided consent, 157 had a sensor inserted. The 13 adverse events (in 12 participants) considered related or possibly related to the study device were not serious; all symptoms were connected to sensor-site issues. Of the 157 participants with sensor insertions, the most commonly reported insertion-site symptoms were erythema (26%), bleeding (21%) and itching (15%).
Sensor-wearing and alarm settings
Mean sensor-wearing in the main study phase was 83% in the CGM without alarms group and 90% in the CGM with
6
The present study was designed to evaluate the effect of CGM, with and without alarms enabled, compared with conventional SMBG in time spent outside a predefined glucose target (hyper-/hypoglycaemia). The study builds on a trial by Bode et al. [15] that investigated the benefit of alarms in CGM. In the present study, the effectiveness of CGM with and without alarms in diabetes management was assessed and the main findings can be summarized as follows: (1) time spent outside glucose target, especially hypoglycaemia, was reduced with CGM compared with SMBG, and was related to method of insulin administration; and (2) there was no significant difference in HbA1c concentration between groups; however, the proportion of participants with HbA1c reductions ≥ 6.0 mmol/mol (≥ 0.5%) was higher in the CGM without alarms and in the CGM with alarms groups than with SMBG. The frequency of capillary glucose measurements, comparing baseline with end of study, significantly decreased in the CGM groups vs. the SMBG group. Although the present study showed a reduction in hypoglycaemia, there was no significant change in HbA1c concentration from baseline, indicating that the reduced hypoglycaemia was not caused by a worsening of glucose control. This is in accordance with other clinical trials in people with Type 1 diabetes that have shown that the use of CGM compared with SMBG can improve HbA1c without increasing the risk of hypoglycaemia [4–7]. Although the difference in h/day spent outside glucose targets for the three groups was not statistically significant, time spent outside glucose targets was significantly reduced for participants receiving CSII using either CGM without or with alarms compared with the SMBG group, whereas
ª 2015 The Authors. Diabetic Medicine ª 2015 Diabetes UK
Research article
DIABETICMedicine
Table 3 Mental component summary score and physical component summary score from the Short-Form-8 Health Survey at baseline and visit 6 CGM without alarms
CGM with alarms
SMBG
Time spent outside target (h/day) Mental component summary score Baseline Day 100 Day 100 (adjusted mean) Physical component summary score Baseline Day 100 Day 100 (adjusted mean)
n = 44
n = 43
n = 39
49.1 9.4 50.9 9.4 50.4
47.6 11.2 48.9 11.4 49.2
49.0 10.4 49.3 10.7 48.9
n = 44
n = 43
n = 39
48.6 9.7 49.0 9.8 48.9
46.7 8.8 49.4 9.6 50.7
49.1 7.9 47.5 8.5 47.1
CGM without alarms vs. CGM with alarms
CGM without alarms vs. SMBG
CGM with alarms vs. SMBG
95% CI
95% CI
95% CI
P
P
P
2.41; 4.80
0.5128
2.23; 5.15
0.4350
3.45; 3.98
0.8887
4.85; 1.21
0.2369
1.31; 4.88
0.2556
0.47; 6.73
0.0245
All values are mean SD unless otherwise specified. CGM, continuous glucose monitoring; CSII, continuous subcutaneous insulin infusion; MDI, multiple daily injections; SMBG, selfmonitoring of blood glucose.
FIGURE 3 Frequency of blood glucose checks per day. CGM, continuous glucose monitoring; SMBG, self-monitoring of blood glucose.
results for participants using MDI were similar to those found in the overall participant analysis. When compared with the SMBG group, the odds ratios of participants achieving a reduction in HbA1c ≥ 0.5% were similar for MDI and CSII (3.1 and 2.7, respectively), as the SMBG group percentage was higher in participants using MDI. This difference is likely to be because the baseline HbA1c was lower for CSII users. The availability of accurate correction doses to CSII users could account for the difference between the two insulin administration groups. The nature of the delivery method may mean that CSII users’ exposure to diabetes technology helps them become familiar with CGM more quickly than MDI users; however, early adopters of technology may not be representative of the group [10], and CSII users remain a minority in the adult diabetes population. ª 2015 The Authors. Diabetic Medicine ª 2015 Diabetes UK
Other studies have shown that sensor use time is inversely associated with HbA1c concentration [16], and sensor use of 60–70% is associated with a reduction in HbA1c concentration [3,5–7,11,17,18], In the present study, sensor compliance in both CGM groups was > 80%; however, no statistical reduction in HbA1c was seen. One plausible reason for the failure to observe change in HbA1c concentrations is the limited follow-up period. Compared with isolated glucose values, glucose patterns over time offer an effective strategy to optimize diabetes management [19]. Glycaemic variability may be a risk factor for the development of long-term complications associated with diabetes, and reduced glucose variability may result in short- and long-term improvements in HbA1c control [20]. In the present study, no structured education was provided on how to use CGM for self-management. In addition, a separate study showed no difference in time outside target glucose, but greater change in HbA1c concentration in people with diabetes who received an algorithm to guide CGM responses, compared with those who were not educated on CGM use [16]. Up to 30% of CGM alarms may be false [9]. If the false alarm rate is perceived as high, the corresponding response rate is low or none [12,21,22]. Alarm coping strategies may be necessary to improve sensor use [23]. In the present study, sensor-wearing was numerically higher in the CGM with alarms group vs. CGM without alarms, and there was no significant difference between the two CGM groups with regard to quality-of-life measurements. The frequency of SMBG has been linked with improved glycaemic control [24–26] and is a predictor for concordance with CGM use [18]. The self-reported number of glucoselevel checks per day for the SMBG group unsurprisingly remained the same as baseline; however, the frequency of real-time glucose-data checks for the CGM group was much
7
DIABETICMedicine
higher and increased throughout the course of the study, suggesting increased attention to glucose levels in both CGM groups vs. SMBG. Increased use of CGM data over the study period may indicate that some familiarization is required for patients to use CGM data; however, a CGM system without alarms was demonstrated as a safe replacement for SMBG for everyday self-management. A limitation of the present study was the small number of participants who were included, resulting in multiple hypothesis tests being performed on small subgroups. A practical proportion of participants with Type 2 diabetes was not achieved for the study because of recruitment issues at the study sites, limiting any meaningful subgroup analysis. Multiple tests were performed on time spent outside glucose target for subgroups without formal adjustment of the significance level and these results should be considered exploratory. Despite fewer than the assumed number of participants performing SMBG completing the study, the study still has 77% power to detect a difference of 1.3 h/day for the primary endpoint. In conclusion, this study shows that using CGM with or without alarms reduces patients’ time spent outside glucose targets compared with conventional SMBG. This was especially evident for patients who are on CSII therapy. Although no difference in HbA1c reduction was seen among the three groups because of the limited follow-up, clinically meaningful HbA1c reduction was more frequent for individuals in the CGM groups. In addition, participants in the CMG groups spent less time in hypoglycaemia and experienced less variability in glucose levels than those in the SMBG group.
Funding sources
This work was funded by Abbott Diabetes Care. The clinical investigation was sponsored and designed by Abbott Diabetes Care with input from the Chief Investigator. Abbott Diabetes Care also provided assistance with statistical analysis.
Competing interests
J.P.N., R.A. and G.F. have received research funding and consulting fees from Abbott Diabetes Care. G.F. is General Manager of the Institut f€ ur Diabetes-Technologie, which carries out studies evaluating SMBG systems and medical devices for diabetes therapy on behalf of various companies. A.F.H.P. has no potential conflicts of interest associated with this submission to declare.
Acknowledgements
All authors take full responsibility for the content of this manuscript, but gratefully acknowledge the assistance of the study personnel at the four clinical sites for execution of this
8
Continuous glucose monitoring: the GLADIS study J. P. New et al.
study (Appendix S1). We also thank Zoe Welsh of Abbott Diabetes Care for assistance with statistical analysis, and Daniella Pfeifer and Lisa Sullivan of Watermeadow Medical for assistance with preparation and submission of the manuscript (supported by Abbott Diabetes Care).
References 1 The Diabetes Control and Complications Trial Research Group. The effect of intensive treatment of diabetes on the development and progression of long-term complications in insulin-dependent diabetes mellitus. N Engl J Med 1993; 329: 977–986. 2 Holman RR, Paul SK, Bethel MA, Matthews DR, Neil HA. 10-year follow-up of intensive glucose control in type 2 diabetes. N Engl J Med 2008; 359: 1577–1589. 3 Hirsch IB, Abelseth J, Bode BW, Fischer JS, Kaufman FR, Mastrototaro J et al. Sensor-augmented insulin pump therapy: results of the first randomized treat-to-target study. Diabetes Technol Ther 2008; 10: 377–383. 4 Tamborlane WV, Beck RW, Bode BW, Buckingham B, Chase HP, Clemons R et al. (Juvenile Diabetes Research Foundation Continuous Glucose Monitoring Study Group). Continuous glucose monitoring and intensive treatment of type 1 diabetes. N Engl J Med 2008; 359: 1464–1476. 5 Deiss D, Bolinder J, Riveline JP, Battelino T, Bosi E, Tubiana-Rufi N et al. Improved glycemic control in poorly controlled patients with type 1 diabetes using real-time continuous glucose monitoring. Diabetes Care 2006; 29: 2730–2732. 6 Raccah D, Sulmont V, Reznik Y, Guerci B, Renard E, Hanaire H et al. Incremental value of continuous glucose monitoring when starting pump therapy in patients with poorly controlled type 1 diabetes: the Real Trend study. Diabetes Care 2009; 32: 2245– 2250. 7 O’Connell MA, Donath S, O’Neal DN, Colman PG, Ambler GR, Jones TW et al. Glycaemic impact of patient-led use of sensorguided pump therapy in type 1 diabetes: a randomised controlled trial. Diabetologia 2009; 52: 1250–1257. 8 De Salvo D, Buckingham B. Continuous glucose monitoring: Current use and future directions. Curr Diab Rep 2013; 13: 657– 662. 9 Kamath A, Mahalingam A, Brauker J. Methods of evaluating the utility of continuous glucose monitor alerts. J Diabetes Sci Technol 2010; 4: 57–66. 10 Gandhi GY, Kovalaske M, Kudva Y, Walsh K, Elamin MB, Beers M et al. Efficacy of continuous glucose monitoring in improving glycemic control and reducing hypoglycemia: a systematic review and meta-analysis of randomized trials. J Diabetes Sci Technol 2011; 5: 952–965. 11 Ramchandani N, Arya S, Ten S, Bhandari S. Real-life utilization of real-time continuous glucose monitoring: the complete picture. J Diabetes Sci Technol 2011; 5: 860–870. 12 Shivers JP, Mackowiak L, Anhalt H, Zisser H. “Turn it off!”: diabetes device alarm fatigue considerations for the present and the future. J Diabetes Sci Technol 2013; 7: 789–794. 13 Ware JE, Kosinski M, Dewey JE, Gandek B. How to Score and Interpret Single-Item Health Status Measures: A Manual For Users of the SF-8TM Health Survey. Lincoln: QualityMetric, 2001. 14 Polonsky WH, Anderson BJ, Lohrer PA, Welch G, Jacobson AM, Aponte JE et al. Assessment of diabetes-related distress. Diabetes Care 1995; 18: 754–760. 15 Bode B, Gross K, Rikalo N, Schwartz S, Wahl T, Page C et al. Alarms based on real-time sensor glucose values alert patients to hypo- and hyperglycemia: the guardian continuous monitoring system. Diabetes Technol Ther 2004; 6: 105–113.
ª 2015 The Authors. Diabetic Medicine ª 2015 Diabetes UK
Research article
16 Jenkins AJ, Krishnamurthy B, Best JD, Cameron FJ, Colman PG, Farish S et al. Evaluation of an algorithm to guide patients with type 1 diabetes treated with continuous subcutaneous insulin infusion on how to respond to real-time continuous glucose levels: a randomized controlled trial. Diabetes Care 2010; 33: 1242–1248. 17 Bailey TS, Zisser HC, Garg SK. Reduction in hemoglobin A1C with real-time continuous glucose monitoring: results from a 12-week observational study. Diabetes Technol Ther 2007; 9: 203–210. 18 Juvenile Diabetes Research Foundation Continuous Glucose Monitoring Study Group, Fiallo-Scharer R, Cheng J, Beck RW, Buckingham BA, Chase HP et al. Factors predictive of severe hypoglycemia in type 1 diabetes: analysis from the Juvenile Diabetes Research Foundation continuous glucose monitoring randomized control trial dataset. Diabetes Care 2011; 34: 586–590. 19 Jovanovic L. Using meal-based self-monitoring blood glucose (SMBG) data to guide dietary recommendations in patients with diabetes. Diabetes Educ 2009; 35: 1023–1030. 20 Bode B, Silver M, Weiss R, Martin K. Evaluation of a continuous glucose monitoring system for home-use conditions. Manag Care 2008; 17: 40–45. 21 Bliss JP, Gilson RD, Deaton JE. Human probability matching behaviour in response to alarms of varying reliability. Ergonomics 1995; 38: 2300–2312.
ª 2015 The Authors. Diabetic Medicine ª 2015 Diabetes UK
DIABETICMedicine
22 Bliss JP, Dunn MC. Behavioural implications of alarm mistrust as a function of task workload. Ergonomics 2000; 43: 1283–1300. 23 Schmidt S, Duun-Henriksen AK, Nørgaard K. Psychosocial factors and adherence to continuous glucose monitoring in type 1 diabetes. J Diabetes Sci Technol 2012; 6: 986–987. 24 Karter AJ, Ackerson LM, Darbinian JA, D’Agostino RB Jr, Ferrara A, Liu J et al. Self-monitoring of blood glucose levels and glycemic control: the Northern California Kaiser Permanente Diabetes registry. Am J Med 2001; 111: 1–9. 25 Karter AJ, Parker MM, Moffet HH, Spence MM, Chan J, Ettner SL et al. Longitudinal study of new and prevalent use of selfmonitoring of blood glucose. Diabetes Care 2006; 29: 1757–1763. 26 Blonde L, Karter AJ. Current evidence regarding the value of selfmonitored blood glucose testing. Am J Med 2005; 118(Suppl. 9A): 20S–26S.
Supporting Information Additional Supporting Information may be found in the online version of this article: Appendix S1. List of contributors to collection of data at the GLADIS study sites.
9
