To investigate whether intermittently scanned continuous glucose monitoring (isCGM) significantly improves glycemic control compared with capillary self-monitored blood glucose (SMBG) in youth with type 1 diabetes and high-risk glycemic control.
This multicenter 6-month randomized, controlled, parallel-arm trial included 64 participants aged 13–20 years with established type 1 diabetes and glycated hemoglobin (HbA1c) ≥9% (≥75 mmol/mol). Participants were allocated to 6-month intervention (isCGM; FreeStyle Libre; Abbott Diabetes Care, Witney, U.K.) (n = 33) or control (SMBG; n = 31) using minimization. The primary outcome was the difference in change in HbA1c from baseline to 6 months.
There was no evidence of a difference between groups for changes in HbA1c at 6 months (adjusted mean 0.2% greater improvement for isCGM [95% CI −0.9 to 0.5] [−2.1 mmol/mol (95% CI −9.6 to 5.4)]; P = 0.576). However, glucose-monitoring frequency was 2.83 (95% CI 1.72–4.65; P < 0.001) times higher in the isCGM group compared with that in the SMBG group at 6 months. The change in the Diabetes Treatment Satisfaction Questionnaire mean item score also favored isCGM at 6 months (P = 0.048), with no significant differences between groups for fear of hypoglycemia and quality of life (both general and diabetes specific) (all P > 0.1).
For youth with high-risk glycemic control, isCGM led to improvements in glucose testing frequency and diabetes treatment satisfaction. However, these did not translate to greater improvement in glycemic control over usual care with SMBG at 6 months.
Introduction
Glucose monitoring is a fundamental component of type 1 diabetes care (1). Frequency of self-monitoring of blood glucose (SMBG) via intermittent finger-prick capillary glucose measurements is related to glycemic control, with improvements in glycated hemoglobin (HbA1c) of up to 0.5% (5 mmol/mol) for each additional glucose check, up to five per day (2). However, this degree of monitoring represents a considerable management burden and can lead to reduced adherence (3).
Disease burden and reduced adherence to SMBG are particularly common during adolescence and young adulthood. Youth may struggle to adhere to self-monitoring recommendations for many reasons, including physical discomfort from finger pricks, social pressure to “fit in” (4), fear of adverse reactions to glucose readings by others (3), and adjusting to less parental involvement in their diabetes management (5). The impact of these challenges is exemplified in the most recent T1D Exchange Clinic Registry data, in which 2016–2018 glycemic control for youth was the least healthy of any age group and, worryingly, also deteriorated compared with the 2010–2012 time period (6,7). Only 17% of those aged <18 years achieved 2018 American Diabetes Association targets of <7.5% (<58 mmol/mol) (6), and even fewer achieved the recent International Society for Pediatric and Adolescent Diabetes target of <7.0% (<53 mmol/mol) (1). Clearly, new management approaches for this population are needed.
Newer diabetes technologies may enable people with diabetes to achieve healthier glycemic control. While youth potentially have the most to gain in this regard, to date, results for technology interventions in this population have been mixed. Real-time continuous glucose monitoring (CGM) technology has shown considerable promise as a tool to help people with type 1 diabetes achieve better glycemic control (8), but appears less efficacious and less well tolerated in youth, in part due to the burden of use, including some systems requiring frequent calibration, and alarm fatigue (9,10).
Flash glucose monitoring, also known as intermittently scanned CGM (isCGM), provides many of the “on-demand” benefits of real-time CGM. However, in the first-generation system, there are no glucose alerts for hypoglycemia or hyperglycemia, raising questions of reduced efficacy when compared with real-time CGM (11). While there is growing evidence that isCGM may improve glycemic control and reduce hypoglycemia (12), there is little direct evidence supporting long-term use in youth, especially those with high-risk glycemic control, defined as an HbA1c ≥9% (≥75 mmol/mol) (13).
Given the propensity for youth to engage with new technology (14), isCGM may offer particular advantages over real-time CGM and may provide a valuable opportunity to engage youth in their diabetes care and help them increase adherence to glucose monitoring and subsequent self-management behaviors. However, there is a dearth of research on isCGM use in adolescents, particularly those with high-risk glycemic control. We, therefore, conducted a randomized controlled trial investigating the effectiveness of isCGM for improving glycemic control in youth with high-risk glycemic control, together with psychosocial impacts associated with its use.
Research Design and Methods
This randomized, two-arm, parallel, controlled, open-label study was conducted across three academic diabetes centers (Canterbury, Capital & Coast, and Southern District Health Boards) in New Zealand from April 2018 to May 2019 (see Supplementary Information 1 for a link to the previously published full protocol and description of a single change to eligibility criteria after recruitment commenced [15]).
Eligibility criteria were: aged 13–20 years (inclusive), type 1 diabetes duration ≥12 months, mean HbA1c ≥9% (≥75 mmol/mol) in the 6 months before enrollment, and planning to continue with routine clinical care during the 6-month trial period. Exclusion criteria included: any severe diabetes-related complication; other uncontrolled medical or psychiatric comorbidity; current or continuous use of real-time CGM or isCGM device within the previous 4 months (intermittent hospital or clinic-based use was not an exclusion); current participation in another device or drug study that could affect glucose measurements; and pregnancy. Any potentially eligible subject from the participating diabetes centers was invited to participate during routine clinic visits.
Following consent, eligible participants were allocated by an offsite statistician to either the intervention group (isCGM plus usual care) or the control group (capillary SMBG plus usual care) using a 1:1 ratio through minimization based on sex (male or female) and HbA1c (9.0% to <11.3% [75 to <100 mmol/mol] and ≥11.3% [≥100 mmol/mol]) with a small random component (20%).
Basic participant and diabetes demographics were obtained during the screening visit, and outcome measures were assessed at baseline and 3 and 6 months. To maximize participation, study visits were rescheduled up to three times. HbA1c was measured by a calibrated point-of-care device (DCA Vantage Analyzer; Siemens Healthcare Diagnostics, Dublin, Ireland). Glucose monitoring was recorded from device downloads for the previous 14 days. Self-reported quality of life (PedsQL generic core scale [16]), diabetes-specific health-related quality of life (PedsQL diabetes module [17]), fear of hypoglycemia (Hypoglycemia Fear Survey-II [18] and Hypoglycemia Fear Survey for Children [19]), diabetes treatment satisfaction (Diabetes Treatment Satisfaction Questionnaire status [DTSQs]) at baseline and 6 months, and DTSQ change at 6 months only (20) were also assessed using validated instruments prior to being informed of study group allocation. A nonvalidated self-report questionnaire was used from previously published research investigating the same glucose-monitoring system in a pediatric population (21) to evaluate isCGM acceptability. Data on adverse events, including cutaneous problems, severe hypoglycemia, and diabetic ketoacidosis, were self-reported in electronic safety questionnaires administered via a link sent by e-mail and text message every 2 weeks over the 6 months of the trial.
The treatment group allocation was revealed at the baseline visit. Participants allocated to the intervention group were provided with isCGM (FreeStyle Libre system; Abbott Diabetes Care), education on using the system (per the manufacturer’s user manual), and a 3-month supply of sensors. Intervention group participants also attended one additional 14-day visit, which included further brief education regarding sensor insertion (including observation by study staff of participants technique) and using advanced features such as trend arrows for making treatment decisions. Participants in the control group self-monitored blood glucose concentrations using their usual glucometer. All participants continued standard diabetes care, with all diabetes management advice provided by their usual diabetes care provider.
The study protocol was approved by the Southern Health and Disability Ethics Committee (17/STH/240; Wellington, New Zealand) (15). The trial was registered with the Australian New Zealand Clinical Trials Registry (ACTRN12618000320257; ANZCTR.org.au) and was issued a Universal Trial Number (U1111-1205-5784) by the World Health Organization International Clinical Trials Registry. Neither the isCGM manufacturer nor the distributor were involved with planning, funding, or conducting the study.
Statistical Analysis
We based the sample size of 64 participants on a power calculation with a two-sided α of 0.05 and power of 0.80 to detect a 1% (10.93 mmol/mol) between-group difference in mean HbA1c change at 6 months (this comparison is the primary outcome for the study) based on an estimated SD of 14.4 mmol/mol for changes, allowing for a small amount (∼9%) of missing data, and without making additional assumptions about the final statistical model (further details are in the study protocol [15]). HbA1c change of 1% was based on the following assumptions: 1) that glucose testing frequency would increase from a hypothesized baseline of 1–2/day to 3–4/day, which, based on previous studies, was feasible to translate into a 1% improvement in HbA1c (2,22), and 2) that baseline HbA1c would be high risk, ensured by the study design restricting entry to a minimum HbA1c of 9% (75 mmol/mol).
The primary outcome and all secondary outcome analyses followed a modified intention-to-treat principle with all participants analyzed in the group to which they were allocated. For HbA1c, additional per-protocol analyses were investigated by including only those from the isCGM group who scanned at least 12 out of 14 days in the 2 weeks prior to the 6-month study visit.
Mixed tobit regression models were used to test for difference in changes in HbA1c at 6 months between groups due to right censoring of values >130 mmol/mol, using data from all time points (baseline, 3 months, and 6 months) with a random participant effect used to accommodate the repeated measures and a group-time interaction to model differences in longitudinal changes.
Secondary outcomes included differences in changes at 3 months in HbA1c and differences in changes at 3 and 6 months in the number of glucose measurements performed (two aspects of glucose monitoring were examined: 1) combined capillary plus interstitial [intervention] vs. capillary [control]; and 2) interstitial only [intervention] vs. capillary [control]), along with participant-reported outcome measures (i.e., quality of life, diabetes-specific health-related quality of life, fear of hypoglycemia, and diabetes treatment satisfaction). As recommended by the developer, diabetes treatment satisfaction, DTSQs mean item score, was calculated due to differences in the number of items in the DTSQs for teens (13–17 years) and young adults (18–20 years). Poisson regression models (negative binomial regression for which there was evidence of overdispersion) tested for differences in count outcomes, and linear mixed models tested for differences in continuous outcomes with all collected data included.
All models included minimization variables (sex and baseline HbA1c, 9.0% to <11.3% [75 to <100 mmol/mol] and ≥11.3% [≥100 mmol/mol]) and incorporated baseline values as a repeated measure. Model residuals were examined, along with the distribution of random effects when relevant, to check that model assumptions were sufficiently well satisfied. All mixed models used restricted maximum likelihood estimation for the variance components in which this was supported (i.e., for linear models), and maximum likelihood estimation was used otherwise.
Statistical analyses were performed by a biostatistician blinded to group allocation until primary analyses were completed using Stata 15.1 software, with a two-sided P value <0.05 considered significant.
Results
Figure 1 presents the participant flow from recruitment through the 6-month follow-up. In total, 64 adolescents and young adults were enrolled in the study. Thirty-three participants were allocated to receive isCGM and 31 to continue SMBG. Overall, baseline characteristics were similar for both groups (Table 1).
Consolidated Standards of Reporting Trials (CONSORT) participant flow diagram.
Baseline demographic and clinical characteristics
. | All participants (n = 64) . | Intervention (n = 33) . | Control (n = 31) . |
---|---|---|---|
Age (years), mean ± SD | 16.6 ± 2.1 | 16.5 ± 1.9 | 16.7 ± 2.2 |
Sex, n (%) | |||
Female | 31 (48) | 16 (48) | 15 (48) |
Male | 33 (52) | 17 (52) | 16 (52) |
Prioritized ethnicity, n (%) | |||
New Zealand European | 37 (58) | 18 (55) | 19 (61) |
Māori* | 16 (25) | 9 (27) | 7 (23) |
Pacific Islander | 10 (16) | 5 (15) | 5 (16) |
Asian | 1 (2) | 1 (3) | 0 (0) |
NZDep13, n (%) | |||
Low deprivation (1–3) | 19 (30) | 10 (30) | 9 (29) |
Medium deprivation (4–7) | 26 (41) | 12 (36) | 14 (45) |
High deprivation (8–10) | 19 (30) | 11 (33) | 8 (26) |
Education/employment, n (%) | |||
In education (secondary) | 42 (66) | 21 (64) | 21 (68) |
In education (tertiary) | 11 (17) | 5 (15) | 6 (19) |
In employment | 10 (16) | 6 (18) | 4 (13) |
NEET | 1 (2) | 1 (3) | 0 (0) |
BMI (z score), mean ± SD | 0.70 ± 1.00 | 0.67 ± 1.05 | 0.73 ± 0.96 |
Duration of diabetes (years), mean ± SD | 7.5 ± 3.8 | 7.0 ± 3.5 | 8.0 ± 4.0 |
Insulin therapy, n (%) | |||
MDI | 55 (86) | 29 (88) | 26 (84) |
CSII | 9 (14) | 4 (12) | 5 (16) |
Insulin estimated total daily dose (units), median (IQR) | 72 (31) | 73 (33) | 70 (24) |
HbA1c (%), mean ± SD | 10.9 ± 1.7 | 10.8 ± 1.7 | 11.2 ± 1.6 |
HbA1c (mmol/mol), mean ± SD | 96.0 ± 18.0 | 94.0 ± 18.0 | 98.8 ± 17.8 |
SMBG checks per day, mean ± SD | 1.9 ± 2.7 | 1.8 ± 1.6 | 1.9 ± 3.6 |
. | All participants (n = 64) . | Intervention (n = 33) . | Control (n = 31) . |
---|---|---|---|
Age (years), mean ± SD | 16.6 ± 2.1 | 16.5 ± 1.9 | 16.7 ± 2.2 |
Sex, n (%) | |||
Female | 31 (48) | 16 (48) | 15 (48) |
Male | 33 (52) | 17 (52) | 16 (52) |
Prioritized ethnicity, n (%) | |||
New Zealand European | 37 (58) | 18 (55) | 19 (61) |
Māori* | 16 (25) | 9 (27) | 7 (23) |
Pacific Islander | 10 (16) | 5 (15) | 5 (16) |
Asian | 1 (2) | 1 (3) | 0 (0) |
NZDep13, n (%) | |||
Low deprivation (1–3) | 19 (30) | 10 (30) | 9 (29) |
Medium deprivation (4–7) | 26 (41) | 12 (36) | 14 (45) |
High deprivation (8–10) | 19 (30) | 11 (33) | 8 (26) |
Education/employment, n (%) | |||
In education (secondary) | 42 (66) | 21 (64) | 21 (68) |
In education (tertiary) | 11 (17) | 5 (15) | 6 (19) |
In employment | 10 (16) | 6 (18) | 4 (13) |
NEET | 1 (2) | 1 (3) | 0 (0) |
BMI (z score), mean ± SD | 0.70 ± 1.00 | 0.67 ± 1.05 | 0.73 ± 0.96 |
Duration of diabetes (years), mean ± SD | 7.5 ± 3.8 | 7.0 ± 3.5 | 8.0 ± 4.0 |
Insulin therapy, n (%) | |||
MDI | 55 (86) | 29 (88) | 26 (84) |
CSII | 9 (14) | 4 (12) | 5 (16) |
Insulin estimated total daily dose (units), median (IQR) | 72 (31) | 73 (33) | 70 (24) |
HbA1c (%), mean ± SD | 10.9 ± 1.7 | 10.8 ± 1.7 | 11.2 ± 1.6 |
HbA1c (mmol/mol), mean ± SD | 96.0 ± 18.0 | 94.0 ± 18.0 | 98.8 ± 17.8 |
SMBG checks per day, mean ± SD | 1.9 ± 2.7 | 1.8 ± 1.6 | 1.9 ± 3.6 |
CSII, continuous subcutaneous insulin infusion; IQR, interquartile range; MDI, multiple daily injections; NEET, not in employment, education, or training; NZDep13, New Zealand index of socioeconomic deprivation 2013 (in which 1 represents the least socioeconomic deprivation and 10 represents the most deprivation).
Māori are the indigenous people of New Zealand.
Glycemic Control
HbA1c data were available for 61 out of 64 participants at 3 months and all participants at 6 months. At baseline, the mean ± SD HbA1c was 10.8 ± 1.7% (94.0 ± 18.0 mmol/mol) in the isCGM group and 11.2 ± 1.6% (98.8 ± 17.8 mmol/mol) in the SMBG group. The mean HbA1c decreased in the isCGM group (10.1 ± 1.6% [87.3 ± 16.8 mmol/mol]) and was unchanged (11.2 ± 1.7% [98.7 ± 18.2 mmol/mol]) in the SMBG group at 3 months. Mean HbA1c decreased further in the isCGM group (10.0 ± 1.5% [86.1 ± 16.5 mmol/mol]) and decreased (10.7 ± 1.5% [93.2 ± 15.9 mmol/mol]) in the SMBG group at 6 months. At the end of the study, there was a nonsignificant between-group difference for change in favor of isCGM (adjusted mean −0.2% [95% CI −0.9 to 0.5] [−2.1 mmol/mol (95% CI −9.6 to 5.4)]; P = 0.576). Removing 17 participants from the isCGM group who did not have at least 12 days of use in the final 14 days prior to the 6-month visit did not change this result (0.2% in favor of isCGM [95% CI −0.7 to 1.1] [2.4 mmol/mol (95% CI −7.1 to 12.0)]; P = 0.618). There was also no evidence of a statistically significant difference at 3 months with all participants included (−0.6% [95% CI −1.3 to 0.1] [−6.3 mmol/mol (95% CI −14.0 to 1.3)]; P = 0.105) or when 17 participants without at least 12 days of isCGM use in the 14 days prior to the 3-month visit were removed (0.5% [95% CI −0.4 to 1.4] [5.5 mmol/mol (95% CI −4.1 to 15.1)]; P = 0.263). Full results of analyses are presented in Table 2. The mean decreases in HbA1c within both the isCGM and SMBG group were clinically significant, defined as a 5 mmol/mol (0.5%) or greater change in HbA1c (Supplementary Fig. 2). No participants achieved the American Diabetes Association target of <7.5% (<58 mmol/mol) at 3 or 6 months.
HbA1c and glucose-monitoring frequency outcomes at 3 and 6 months
. | Baseline . | 3 months . | 6 months . | Difference in adjusted changes at 6 months (95% CI) . | P value . | |||
---|---|---|---|---|---|---|---|---|
Intervention (n = 33) . | Control (n = 31) . | Intervention (n = 31) . | Control (n = 30) . | Intervention (n = 33) . | Control (n = 31) . | |||
Glycemic control* | 0.576 | |||||||
HbA1c (%) | 10.8 ± 1.7 | 11.2 ± 1.6 | 10.1 ± 1.6 | 11.2 ± 1.7 | 10.0 ± 1.5 | 10.7 ± 1.5 | −0.2 (−0.9 to 0.5) | |
HbA1c (mmol/mol) | 94.0 ± 18.0 | 98.8 ± 17.8 | 87.3 ± 16.8 | 98.7 ± 18.2 | 86.1 ± 16.5 | 93.2 ± 15.9 | −2.1 (−9.6 to 5.4) | |
Glucose monitoring† | ||||||||
Interstitial (intervention) vs. capillary (control) | 1.8 ± 1.6 | 1.9 ± 3.6 | 4.1 ± 4.1 | 1.5 ± 3.5 | 3.5 ± 3.2 | 1.4 ± 3.0 | 2.6 (1.56–4.26) | <0.001 |
Interstitial + capillary (intervention) vs. capillary (control) | 1.8 ± 1.6 | 1.9 ± 3.6 | 4.5 ± 4.3 | 1.5 ± 3.5 | 3.8 ± 3.1 | 1.4 ± 3.0 | 2.8 (1.72–4.65) | <0.001 |
. | Baseline . | 3 months . | 6 months . | Difference in adjusted changes at 6 months (95% CI) . | P value . | |||
---|---|---|---|---|---|---|---|---|
Intervention (n = 33) . | Control (n = 31) . | Intervention (n = 31) . | Control (n = 30) . | Intervention (n = 33) . | Control (n = 31) . | |||
Glycemic control* | 0.576 | |||||||
HbA1c (%) | 10.8 ± 1.7 | 11.2 ± 1.6 | 10.1 ± 1.6 | 11.2 ± 1.7 | 10.0 ± 1.5 | 10.7 ± 1.5 | −0.2 (−0.9 to 0.5) | |
HbA1c (mmol/mol) | 94.0 ± 18.0 | 98.8 ± 17.8 | 87.3 ± 16.8 | 98.7 ± 18.2 | 86.1 ± 16.5 | 93.2 ± 15.9 | −2.1 (−9.6 to 5.4) | |
Glucose monitoring† | ||||||||
Interstitial (intervention) vs. capillary (control) | 1.8 ± 1.6 | 1.9 ± 3.6 | 4.1 ± 4.1 | 1.5 ± 3.5 | 3.5 ± 3.2 | 1.4 ± 3.0 | 2.6 (1.56–4.26) | <0.001 |
Interstitial + capillary (intervention) vs. capillary (control) | 1.8 ± 1.6 | 1.9 ± 3.6 | 4.5 ± 4.3 | 1.5 ± 3.5 | 3.8 ± 3.1 | 1.4 ± 3.0 | 2.8 (1.72–4.65) | <0.001 |
Data are mean ± SD.
Glycemic control P value based on HbA1c measurements recorded in millimoles per mole.
Baseline and 3- and 6-month capillary glucose checks from blood glucose meter 14-day summary (intervention group and control group). The 3- and 6-month interstitial glucose checks were from the FreeStyle Libre reader 14-day summary (intervention group). Difference in changes at 3 and 6 months were adjusted for sex and baseline HbA1c (9.0% to <11.3% [75 to <100 mmol/mol] and ≥11.3% [≥100 mmol/mol]) and incorporated baseline values as repeated measures.
Glucose Monitoring
In the isCGM group, there was an increase in mean ± SD glucose checks performed per day from 1.8 ± 1.6 capillary checks at baseline to 4.5 ± 4.3 combined interstitial and capillary checks per day at 3 months and a small decrease to 3.8 ± 3.1 interstitial and capillary checks per day at 6 months (Table 2). Of the 33 isCGM group participants, scanning frequency data were available for 28 participants at 2 weeks, 30 participants at 3 months, and 31 participants at 6 months. The mean daily scanning frequency was 7.2 ± 3.1 at 2 weeks, 4.5 ± 4.1 at 3 months, and 3.7 ± 3.2 at 6 months. In the SMBG group, the mean glucose checks performed per day decreased from 1.9 ± 3.6/day at baseline to 1.5 ± 3.5 checks/day at 3 months and decreased further to 1.4 ± 3.0 checks/day at 6 months. The between-group differences for change in glucose checks per day was significant at 3 months (intervention group rate of glucose checking 3.2 times greater than the SMBG group [95% CI 1.97–5.23]; P < 0.001) and at 6 months (intervention group rate of glucose checking 2.8 times the SMBG group [95% CI 1.72–4.65]; P < 0.001). Results were similar if the isCGM group’s SMBGs during the intervention phase were excluded (3-month intervention group rate of glucose checking 2.9 times the SMBG group [95% CI 1.75–4.69], P < 0.001; 6-month intervention group rate of glucose checking 2.6 times the SMBG group [95% CI 1.56–4.26], P < 0.001).
Quality of Life Assessments
At 6 months, there were no significant differences between study groups on the PedsQL Generic Total score, PedsQL Diabetes Module Total score, and Fear of Hypoglycemia (Table 3), except for an improvement in the intervention group for the DTSQs treatment satisfaction (0.47 [95% CI 0.00–0.93]; P = 0.048) and a relative improvement in the control group for the diabetes subscale of the PedsQL (9.2 [95% CI 3.3–15.2]; P = 0.002) (one out of five subscales).
Comparison of psychosocial outcomes between study groups at 6 months
Questionnaire . | Baseline . | 6 months . | Difference in change at 6 months (95% CI) . | P value . | ||
---|---|---|---|---|---|---|
Intervention . | Control . | Intervention . | Control . | |||
PedsQL Generic | (n = 33) | (n = 31) | (n = 32) | (n = 31) | ||
Total score | 73.8 ± 14.0 | 75.2 ± 11.9 | 77.5 ± 15.7 | 79.7 ± 11.6 | −1.2 (−6.5 to 4.1) | 0.661 |
PedsQL Diabetes | (n = 33) | (n = 31) | (n = 32) | (n = 31) | ||
Diabetes subscale | 56.9 ± 16.2 | 57.2 ± 18.3 | 56.5 ± 17.7 | 65.7 ± 16.7 | −9.2 (−15.2 to −3.3) | 0.002 |
Treatment I subscale | 57.4 ± 20.2 | 65.9 ± 20.7 | 68.9 ± 21.3 | 68.5 ± 17.9 | 8.1 (−0.1 to 16.4) | 0.053 |
Treatment II subscale | 67.2 ± 18.6 | 68.2 ± 19.5 | 74.4 ± 17.3 | 70.0 ± 21.6 | 4.8 (−2.6 to 12.3) | 0.204 |
Worry subscale | 61.6 ± 22.2 | 67.7 ± 21.5 | 71.6 ± 23.4 | 69.6 ± 21.6 | 7.7 (−2.3 to 17.8) | 0.130 |
Communication subscale | 64.9 ± 28.2 | 70.2 ± 22.3 | 68.2 ± 30.5 | 77.4 ± 24.2 | −5.2 (−16.7 to 6.2) | 0.370 |
Total score | 60.9 ± 14.6 | 63.7 ± 15.6 | 65.7 ± 15.9 | 68.9 ± 15.1 | −1.1 (−6.2 to 4.1) | 0.688 |
HFS | (n = 31) | (n = 31) | (n = 32) | (n = 31) | ||
Behavior subscale | 1.75 ± 0.58 | 1.91 ± 0.63 | 1.70 ± 0.67 | 1.69 ± 0.47 | 0.18 (−0.08 to 0.44) | 0.182 |
Worry subscale | 1.19 ± 0.59 | 1.26 ± 0.73 | 0.94 ± 0.55 | 1.14 ± 0.75 | −0.13 (−0.37 to 0.11) | 0.302 |
DTSQ | (n = 32) | (n = 31) | (n = 32) | (n = 31) | ||
Treatment satisfaction mean item score | 3.96 ± 0.88 | 4.36 ± 0.90 | 4.33 ± 1.12 | 4.28 ± 1.02 | 0.47 (0.00–0.93) | 0.048 |
Questionnaire . | Baseline . | 6 months . | Difference in change at 6 months (95% CI) . | P value . | ||
---|---|---|---|---|---|---|
Intervention . | Control . | Intervention . | Control . | |||
PedsQL Generic | (n = 33) | (n = 31) | (n = 32) | (n = 31) | ||
Total score | 73.8 ± 14.0 | 75.2 ± 11.9 | 77.5 ± 15.7 | 79.7 ± 11.6 | −1.2 (−6.5 to 4.1) | 0.661 |
PedsQL Diabetes | (n = 33) | (n = 31) | (n = 32) | (n = 31) | ||
Diabetes subscale | 56.9 ± 16.2 | 57.2 ± 18.3 | 56.5 ± 17.7 | 65.7 ± 16.7 | −9.2 (−15.2 to −3.3) | 0.002 |
Treatment I subscale | 57.4 ± 20.2 | 65.9 ± 20.7 | 68.9 ± 21.3 | 68.5 ± 17.9 | 8.1 (−0.1 to 16.4) | 0.053 |
Treatment II subscale | 67.2 ± 18.6 | 68.2 ± 19.5 | 74.4 ± 17.3 | 70.0 ± 21.6 | 4.8 (−2.6 to 12.3) | 0.204 |
Worry subscale | 61.6 ± 22.2 | 67.7 ± 21.5 | 71.6 ± 23.4 | 69.6 ± 21.6 | 7.7 (−2.3 to 17.8) | 0.130 |
Communication subscale | 64.9 ± 28.2 | 70.2 ± 22.3 | 68.2 ± 30.5 | 77.4 ± 24.2 | −5.2 (−16.7 to 6.2) | 0.370 |
Total score | 60.9 ± 14.6 | 63.7 ± 15.6 | 65.7 ± 15.9 | 68.9 ± 15.1 | −1.1 (−6.2 to 4.1) | 0.688 |
HFS | (n = 31) | (n = 31) | (n = 32) | (n = 31) | ||
Behavior subscale | 1.75 ± 0.58 | 1.91 ± 0.63 | 1.70 ± 0.67 | 1.69 ± 0.47 | 0.18 (−0.08 to 0.44) | 0.182 |
Worry subscale | 1.19 ± 0.59 | 1.26 ± 0.73 | 0.94 ± 0.55 | 1.14 ± 0.75 | −0.13 (−0.37 to 0.11) | 0.302 |
DTSQ | (n = 32) | (n = 31) | (n = 32) | (n = 31) | ||
Treatment satisfaction mean item score | 3.96 ± 0.88 | 4.36 ± 0.90 | 4.33 ± 1.12 | 4.28 ± 1.02 | 0.47 (0.00–0.93) | 0.048 |
Data are mean ± SD unless otherwise indicated. PedsQL Generic Total scores range from 0 to 100; higher scores indicate better health-related quality of life. PedsQL Diabetes subscale scores range from 0 to 100; higher scores indicate lower problems with diabetes symptoms. PedsQL Treatment I subscale scores range from 0 to 100; higher scores indicate lower problems with diabetes-specific barriers. PedsQL Treatment II subscale scores range from 0 to 100; higher scores indicate lower problems with adherence. PedsQL Worry subscale scores range from 0 to 100; higher scores indicate lower problems with diabetes-specific worry. PedsQL Communication subscale scores range from 0 to 100; higher scores indicate lower problems with diabetes-specific communication. PedsQL Diabetes Total scores range from 0 to 100; higher scores indicate better diabetes-specific quality of life. Fear of hypoglycemia (HFS) Behavior subscale mean item scores range from 0 to 4; higher scores indicate a greater tendency to avoid hypoglycemia. HFS Worry subscale mean item scores range from 0 to 4; higher scores indicate more worry concerning episodes of hypoglycemia and its consequences. DTSQs mean item scores range from 0 to 6; higher scores indicate higher treatment satisfaction.
Adverse Events
Six participants (18%) in the intervention group and five (16%) participants in the control group experienced at least one episode of diabetic ketoacidosis, with no significant differences between groups. There were no severe hypoglycemic events during the trial. A further 11 participants (5 in the intervention group and 6 participants in the control group) were hospitalized for reasons not attributable to participation in the study, including treatment for a viral infection and self-admission due to ongoing unhealthy glycemic control.
Acceptance of isCGM
All 33 participants in the isCGM group would recommend the device to a friend, and 32 planned to continue using the system. Most participants in the intervention group reported isCGM was less painful (97%), quicker (100%), and easier (100%) compared with SMBG (Supplementary Fig. 3).
Conclusions
The Managing Diabetes in a “Flash” study investigated the use of isCGM in a youth population with high-risk glycemic control. The study, a randomized controlled trial, was carried out independently (i.e., was not industry funded) to fill a gap in research investigating isCGM in people with type 1 diabetes who have been largely excluded from previous technology research. Despite a sustained increase in glucose-checking frequency and improved diabetes treatment satisfaction in the isCGM group, there was no evidence of a greater benefit on glycemic control by using isCGM over SMBG at 6 months. It is noteworthy that both groups saw improving glycemic control, with 52% (16 out of 31) of participants in the isCGM group and 20% (6 out of 30) of participants in the SMBG group experiencing a clinically significant improvement in glycemic control at 3 months with further increases at 6 months (61% [20 out of 33] and 44% [14 out of 31], respectively).
Overall, the improvement in HbA1c observed in the isCGM group is similar to previous nonrandomized studies in pediatric and adult populations (12,23–25), with evidence that increasing glucose monitoring is associated with lower HbA1c (26). Importantly, the scanning frequency in the isCGM group at 3 months was an increase from baseline, and at 6 months, the isCGM group performed significantly more glucose-level checks compared with the control group. These findings may suggest that overall, the novelty of isCGM did not diminish over 6 months and that overall, isCGM facilitated more frequent glucose monitoring compared with SMBG. However, our finding that HbA1c in the control group also decreased may highlight the influence of behavior change associated with awareness of being in a research trial (3,27,28). In order to minimize missing data, multiple attempts were made to conduct study visits with some participants (up to three times), clinical contact that may not be possible in routine care in which people with high-risk glycemic control may only have sporadic clinical support. Therefore, it is plausible that control group participants changed their self-management behavior in response (29). The relatively small further decreased HbA1c in the isCGM group may suggest the device is unlikely to be appreciably more effective than SMBG without additional clinical support, something specifically not provided in this study, although the widths of the 95% CIs were large and included clinically important effect sizes. This result warrants caution when interpreting recent observational data suggesting those with less healthy glycemic control may benefit the most from isCGM (25). Lack of effectiveness in this specific population has also been seen in the wider CGM literature, in which only those youth continuously using CGM at least 6 or 7 days each week achieve glycemic gains (10).
Youth are well documented to have the least healthy glycemic control (7), and many experience considerable diabetes- and nondiabetes-related burden, including diabetes distress (30), stigma (31), depression (32), and family conflict (33). Clearly, despite improved glucose-checking frequency and diabetes satisfaction and overwhelmingly positive participant and family experiences with isCGM in this trial (34,35), gains in traditionally measured psychosocial outcomes such as quality of life were not seen at 6 months. For the majority of these individuals, improving psychosocial outcomes may require more intensive input, beyond just reducing invasive glucose monitoring and access to more glucose data. This point was highlighted in the Diabetes Control and Complications Trial, in which weekly nursing input was paired with intensive therapy to achieve the well-reported HbA1c gains and long-term health benefits (36–38). This is important as support and the art of diabetes care are likely needed for intensive management to be most successful (39). Alternatively, our sample may have experienced challenges outside of diabetes that took precedence over self-management (40).
To our knowledge, this multicenter randomized trial is the first such trial to assess the effectiveness of isCGM in adolescents and young adults with type 1 diabetes and high-risk glycemic control. A key strength of this study is the investigation of the effect of isCGM alone, rather than the potential impact of added support and attention, and therefore, the findings reflect the impact of the technology alone. Other strengths of this study include the multisite design, which enhances external validity, study retention (100% at 6 months), and a sample that was highly representative of ethnic minority groups, which is lacking in previous research. The importance of properly designed randomized controlled trials is highlighted by the improved glycemic control from baseline seen in both the usual care and isCGM groups. This Hawthorne effect has been well described (3) and may explain why routinely measuring HbA1c and reporting the outcome to participants in this population changed self-management practices and impacted glycemic control as well as diabetes symptoms in the control group. Caution is therefore needed in interpreting the findings of observational studies for isCGM and other new diabetes interventions. This study intentionally included people with a high-risk prestudy HbA1c (≥75 mmol/mol [9%]), which limits the generalizability of these findings outside of this complex population. The age inclusion criteria is both a strength and a weakness given the many developmental changes that occur across the range spanning from 13 to 20 years. While change in HbA1c was the primary outcome of interest, research is needed to investigate if isCGM is more effective than SMBG for improving other important glycemic outcomes in young people with high-risk glycemic control, such as time in range. Having available these other glycemic variables, such as time in range, time in hypoglycemia, and time in hyperglycemia between the two groups would have strengthened the study. A further limitation was using point-of-care HbA1c measures at clinical sites instead of a more stringent central laboratory (a pragmatic compromise for this age group and multisite study). A final limitation is the study sample size, which, despite being powered to detect clinically important differences, resulted in CIs, as shown in Supplementary Fig. 2, that were sufficiently wide so that clinically important effects cannot be ruled out. Future randomized studies investigating isCGM in larger samples of youth with high-risk glycemic control are needed.
In summary, this randomized controlled trial in adolescents and young adults with high-risk glycemic control reveals that isCGM does engage this complex population, as evidenced by increased glucose-monitoring behavior and diabetes satisfaction. However, in this study, this did not translate to a statistically significant greater improvement in glycemic control over usual care with SMBG at 6 months. Ongoing efforts to find likely combined technology and psychosocial strategies to help this population are required.
Clinical trial reg. no. ACTRN12618000320257, www.anzctr.org.au
This article contains supplementary material online at https://doi.org/10.2337/figshare.12689951.
Article Information
Funding. Funding was provided by Cure Kids grant 3582; the Department of Women’s and Children’s Health Research Committee, University of Otago; and the Dunedin School of Medicine, University of Otago.
Neither the CGM manufacturer nor the distributor was involved with planning, funding, or conducting the study.
Duality of Interest. No potential conflicts of interest relevant to this article were reported.
Author Contributions. S.E.B., A.R.G., E.J.W., M.I.d.B., B.C.G., P.A.T., J.A.R., KE.M., H.C., and B.J.W. were involved with the design of the study protocol. S.E.B., E.J.W., M.I.d.B., P.A.T., J.A.R., K.E.M., H.C., S.R., and B.J.W. were involved with recruitment and data collection. A.R.G. performed the statistical analyses. All authors worked collaboratively to review and prepare the final manuscript. B.J.W. is the guarantor of this work and, as such, had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.