Introduction

Intensive insulin therapy is important in avoiding diabetes-related microvascular and macrovascular complications [1], but may also have a role in preserving residual beta cell function [2]. Levels of C-peptide are now being increasingly recognised as a primary endpoint in intervention trials in patients with newly diagnosed type 1 diabetes [3]. High number of blood glucose tests per day was shown to be the most important factor related to an improved HbA1c level in adolescents [4]. In short-term studies, the use of continuous glucose monitoring (CGM) has been shown to improve glycaemic control and variability, both in children and in adults with long-standing type 1 diabetes [5, 6], whereas in a long-term study lasting 26 weeks, a beneficial effect of CGM was seen only in adults [7]. Despite problems that may be related to the inconvenience of wearing CGM devices, low compliance and consequently poor glycaemic results, particularly in young patients, may be associated with difficulties in implementing the information from the system and adjusting the therapy accordingly. Integrating CGM into the diabetes treatment from the onset of clinical disease could lead to a better understanding of the individual interplay between exogenous insulin requirements, prandial glycaemic excursions and sport activities, and so, may imprint the patient’s treatment approach in the long term.

Therefore, we performed a multicentre randomised controlled trial to assess the acceptance, efficacy and safety of the use of CGM in combination with insulin pump therapy from the diagnosis of type 1 diabetes in children and adolescents. Particularly, we set out to determine whether the use of sensor-augmented insulin pump therapy leads to better glycaemic control, lower daily insulin requirements, higher residual beta cell function, lower incidence of severe hypoglycaemia and better quality of life after 1 year of treatment compared with the use of a conventional insulin pump combined with conventional self-monitoring of blood glucose.

Research design and methods

Patients

Children and adolescents between 1 and 16 years of age with a diagnosis of type 1 diabetes within 4 weeks before study entry were eligible for inclusion. Patients were treated in five European paediatric centres experienced in insulin pump treatment and CGM use (Hannover and Leipzig, Germany; Vienna, Austria; Warsaw, Poland; Angers, France). A minimum of ten patients per centre was required. The study was approved by the Ethical Committee of each participating centre. Written informed consent was obtained from patients and their parents or guardians.

Patients were assigned by a central randomisation procedure to receive either sensor-augmented insulin pumps (Paradigm REAL-Time Insulin Pump and Continuous Glucose Monitoring System, Medtronic MiniMed, Northridge, CA, USA) or MiniMed Paradigm 515/715 insulin pumps (Medtronic MiniMed) and blood glucose meters.

Patients were instructed to use the CGM device on a daily basis, replace the sensors every 3 days and to confirm glucose measurements by self-monitoring of blood glucose (SMBG) before making management decisions, as required from the labelling on the regulatory device. Patients in the control group were given blood glucose meters and asked to perform SMBG at least four times a day. Glycaemic targets were identical in all study centres and corresponded to those suggested by the International Society of Paediatric and Adolescent Diabetes [8] with optimal target pre-meal glucose values between 5.0 and 8.0 mmol/l, peak postprandial values below 10.0 mmol/l, bedtime values between 6.7 and 10.0 mmol/l and overnight values between 4.5 and 9.0 mmol/l.

Follow-up

Study visits were conducted at baseline and at 6, 26 and 52 weeks (±1 week) thereafter. Patients of both arms received the same local standard care as given to the off-study patients of each centre including phone contact in case of acute complications and routine diabetes management visits at the diabetes outpatient clinic, occurring regularly every 6–10 weeks. Before the study visits at 6 and 52 weeks, the control group used a blinded continuous glucose monitor (Guardian REAL-Time Clinical device, Medtronic MiniMed) for 6 days. At each visit, the insulin pump data were downloaded in a blinded fashion using the web-based application CareLink Clinical (Medtronic MiniMed). Variables (dose of insulin per day, percentages of basal and bolus insulin dose as well as average number of daily boluses) and sensor data (glucose average and variability, i.e. glucose SD and mean amplitude of glycaemic excursions [MAGE]) [9] were calculated on a 24 h basis. At baseline and at 52 weeks, C-peptide levels were determined in fasting conditions. At the start of the study and at 24 and 52 weeks, children aged 8–18 years and the primary caregiver were asked to complete the DISABKIDS [10] and KIDSCREEN-27 [11, 12] questionnaires to evaluate patient’s health-related quality of life and caregiver’s impression of patient’s quality of life and own wellbeing assessed by the WHO-5 questionnaire [13].

Laboratory assessment

HbA1c was measured centrally at baseline as well as at every study visit using HPLC (Bio-Rad Laboratories, Munich, Germany; LKF-Laboratorium für Klinische Forschung, Kiel, Germany). Diabetes-associated autoantibodies (islet cell antibodies [ICA]; insulin autoantibodies [IAA]; glutamic acid decarboxylase antibodies [GADA]; tyrosine phosphatase-related islet antigen-2 antibodies [IA-2A]; zinc transporter 8 antibodies [ZnT8A]) at diagnosis of diabetes were analysed centrally using in-house methods by the Scientific Laboratory, Hospital for Children and Adolescents, University of Helsinki, Finland. In the IAA assay, mono-125I (TyrA14) human recombinant insulin (GE Healthcare, Chalfont St Giles, UK; activity 7,400 × 1010 Bq/mmol) was used as the labelled antigen and in the GADA, IA-2A and ZnT8A assays antigens produced in house through in vitro transcription and translation were used. Fasting C-peptide and plasma glucose concentrations were analysed in the LKF-laboratory using the Immuno-Analyzer Centaur Advia System and the hexokinase method, respectively. All laboratory results were blinded to the investigators.

Statistical analysis

Sample size estimate was based on the primary outcome variable (HbA1c after 12 months) and a previous analysis of paediatric patients with an estimated average of HbA1c of 8.0% (SD 1.0%) after 1 year of type 1 diabetes (O. Kordonouri and T. Danne, unpublished data). The hypothesis of the current study was to detect an HbA1c difference of 0.5% comparing the two treatment groups at the end of the treatment phase using a two-sample t test with two-sided alpha of 0.05. A total sample size of 160 randomised patients with equal numbers in each group was planned to achieve a statistical power of 80% and reach an evaluable dataset of 64 patients per group, allowing for a 20% dropout rate. The statistical evaluation was performed by intention-to-treat analysis according to treatment groups defined as randomised using the statistical software package SPSS 12.0. Standard procedures were used such as Fisher’s χ2 test, t test and nonparametric Mann–Whitney U test or Wilcoxon test when appropriate. The association between rank-ordered variables was assessed by Spearman correlation coefficient (ρ). The Kaplan–Meier method with the logrank test was used to analyse and compare HbA1c between the study groups in relation to sensor use. A statistically significant difference was assumed with a two-sided level of p < 0.05 for each test.

Results

Patients

Between February 2007 and October 2008, 357 eligible children and adolescents were diagnosed with type 1 diabetes in the five participating centres (Table 1). Study participation was offered to 295 patients. A total of 160 patients agreed to take part in the study and were randomised on average 9.6 ± 6.0 days (mean ± SD) after diagnosis, with 80 patients assigned to each of the two randomisation groups (acceptance rate 54.2%). Only 154 patients completed all visits and were evaluable for statistical analyses (sensor-augmented pump vs pump alone: 76 vs 78, dropout rate 3.8%). Clinical and biochemical characteristics at baseline did not differ between the treatment groups (Table 2).

Table 1 Study cohort flowchart
Full size table
Table 2 Patient characteristics at baseline
Full size table

Glycaemic control and variability

HbA1c of both treatment groups was not significantly different throughout the total period. At each follow-up visit, HbA1c levels in patients with sensor-augmented pump therapy were persistently below those of patients treated with insulin pump and SMBG (Table 3). HbA1c at 12 months was 7.5 ± 1.1% in the total cohort. No significant differences between treatment groups were seen within the age groups. In total, 30 out of 76 patients (39.5%) with sensor-augmented pump had HbA1c levels below 7.0% at 12 months compared with 26 of 77 patients (33.8%) with insulin pump alone (p = 0.464). At 12 months, the 24 h glucose average was comparable between the groups (p = 0.966), but glycaemic variability was lower in the sensor group and reached statistical significance for MAGE (Table 4).

Table 3 Glycaemic outcomes during the study, according to age
Full size table
Table 4 Glycaemic variability, insulin requirements and C-peptide at 52 weeks
Full size table

Documented sensor use data over the whole study period were available for 55 patients. Patients with sensor-augmented insulin pump had an average of 1.4 ± 0.7 sensor uses per week. Sensor usage was significantly higher at 6 weeks (2.1 ± 0.9 sensors per week) and declined significantly at 26 weeks (1.4 ± 1.0 sensors per week) and 52 weeks (1.1 ± 0.7 sensors per week), respectively (p < 0.001). There were no significant differences between the age groups in the frequency of sensor use (p = 0.497).

Patients with regular sensor use (at least one sensor per week during the first year) had significantly lower HbA1c (mean 7.1%, 95% CI 6.8–7.4%) compared with the combined group of patients with no or low sensor usage, i.e. less than 1 sensor per week (7.6%, 95% CI 7.3–7.9%; p = 0.032) (Fig. 1). Nineteen of 37 (51.4%) patients with regular sensor use had HbA1c below 7.0% at 12 months. There was a significant inverse correlation between HbA1c and frequency of sensor usage at the 6 week visit (ρ = −0.268, p = 0.026), but no such relationship was seen at the 26 week (ρ = 0.009, p = 0.942) or 52 week visit (ρ = −0.066, p = 0.615). Patients with sensor-augmented insulin pumps performed fewer self-monitoring blood glucose finger sticks per day (5.2 ± 2.0) than those with insulin pump alone (6.5 ± 2.1, p < 0.001).

Fig. 1
figure 1

HbA1c in patients with regular sensor use (≥1 sensor/week; n = 37; upper curve) and those in the combined group with no or low sensor use (<1 sensor/week; n = 95; lower curve) compared by Kaplan–Meier plot and logrank test (p = 0.032)

Full size image

Insulin requirements and endogenous insulin secretion

Insulin requirements did not differ significantly between the groups during the study. At 12 months, the total daily insulin dose was 0.59 ± 0.22 U/kg body weight in patients with sensor-augmented insulin pump and 0.64 ± 0.23 U/kg body weight in those with insulin pump only (p = 0.248). The proportion of basal insulin was 29.7 ± 10.4% in the control group, which was significantly lower than in the sensor-augmented pump group (34.0 ± 11.8%, p = 0.021). In terms of the average number of daily boluses, patients in the sensor-augmented pump group injected 7.9 ± 3.6 boluses per day compared with 7.0 ± 2.7 in the control group (p = 0.097).

Fasting C-peptide concentration at baseline was not associated with HbA1c (ρ = −0.099, r = 0.225) and did not significantly differ between the groups (Table 2). In the total cohort, we observed a slight decrease in basal C-peptide secretion over the first year (Table 3). The proportion of patients with an increase in fasting C-peptide concentration from baseline to 12 months was 39.2% (29 of 74) in the sensor-augmented pump group and 34.2% (26 of 76) in the control group (p = 0.528). Among patients with regular sensor use, C-peptide increase was seen in 17 of 37 patients (45.9%). Significantly higher C-peptide concentrations were observed at 12 months in the sensor group (Table 4). There was a close to significant correlation between the frequency of sensor usage during the study and C-peptide concentrations at 12 months (ρ = 0.250, p = 0.057). In the sensor-augmented pump group, more frequent sensor use during the first year was associated with a reduced decrease in fasting C-peptide at 12 months. Interestingly, more frequent sensor use at 26 weeks was also associated with lower C-peptide loss at 12 months as well (ρ = 0.323, p = 0.008).

Severe hypoglycaemia and quality of life

No episode of severe hypoglycaemia was reported in patients with a sensor-augmented insulin pump compared with four episodes in patients with insulin pump alone (p = 0.046). As assessed with the WHO-5 questionnaire, the dominant caregiver fulfilled on average the clinical diagnosis of depression (score below 48) at baseline (Table 5). The scores reached normal values at 6 months and remained normal after 1 year. The children’s health-related quality of life showed significantly lower scores compared with European norm data (t values standardised: mean 50 ± 10) for physical, psychological, social support, and school at baseline, normalising after 6 months and remaining normal after 12 months with no difference between the intervention and control groups.

Table 5 Psychosocial adaptation and quality of life
Full size table

Discussion

This is the first study to examine the efficacy of implementing both CGM and continuous subcutaneous insulin infusion (CSII) in children and adolescents at the time of diagnosis of type 1 diabetes. The main results of this study can be summarised as follows: the primary outcome measure (HbA1c at 12 months) did not differ between the two treatment groups, fasting C-peptide and change in C-peptide in the total cohort were not significantly different and quality of life measures did not differ between the groups. Although we found that patients who were able to use CGM regularly had lower HbA1c levels than those who did not, there were not enough of these patients to affect the overall results, indicating that the sensor systems are not yet effective enough or user-friendly enough for children and parents to have a clinically important impact on metabolic control during the first 12 months of the disease.

In our randomised, controlled trial, we observed beneficial effects of insulin pump therapy from the diagnosis of diabetes particularly in children with regular continuous sensor use. Although the 0.2% difference in HbA1c values lacked statistical significance, the use of continuous glucose monitoring was associated with a better preservation of endogenous insulin secretion after 1 year of diabetes. Similar results with increasing C-peptide levels from baseline up to 2 years were also reported in a small cohort of patients treated with insulin pump and intensive subcutaneous insulin therapy from the IMDIAB Group [14]. It is well known that the loss of beta cell function tends to be slower in teenagers than in young children [15]. Therefore, one may speculate whether implementation of CGM from the onset of clinical disease may have an especially positive effect in the adolescent patient cohort independent of glycaemic control as seen in the present study. Moreover, imperfect adherence with long-term diabetes management has been recognised as a limitation to achieve good metabolic control in adolescents and young adults [15–17]. Taken together, the rapid introduction of continuous glucose monitoring into diabetes treatment may be helpful and effective from both metabolic and educational aspects. A stimulated C-peptide concentration may have been even more sensitive in detecting such effects of beta cell protection, but we refrained from using this measure because of the young age range included in this study. Nevertheless, the effects of sensor-augmented pump treatment starting from the clinical manifestation of type 1 diabetes have to be considered when immunological interventions aimed at the preservation of endogenous insulin secretion (such as anti-CD3-antibody treatment [18] or GAD vaccination [19]) are evaluated.

Compared with historical data, such as those recently reported by the Hvidøre Study Group in 275 children with multiple daily injections (mean HbA1c 7.9 ± 1.5% after 12 months of diabetes), we observed beneficial effects of insulin pump therapy during the first year of diabetes [15]. The overall HbA1c was 7.5 ± 1.1% and additional use of continuous glucose monitoring has no statistically significant impact on the overall glycaemic control, except in children with regular sensor use. In the latter group, average HbA1c was 0.8% lower than in the Hvidøre report with more than half of these patients achieving HbA1c below 7.0% without any severe hypoglycaemic event. Moreover, episodes of severe hypoglycaemia were not reported at all in patients with sensor-augmented pump treatment, but occurred in four patients in the control group. This difference has to be interpreted with caution because of the low frequency of severe hypoglycaemia in the whole study, particularly as previous studies have shown limited effectiveness of CGM as a preventive tool for hypoglycaemia [20, 21].

Interestingly, in our study, patients with sensor-augmented insulin pump treatment had significantly lower glycaemic variability at 12 months of overt diabetes. Such reduction in variability has been observed previously in short-term interventions in children and adults [6]. Patients with CGM performed fewer finger pricks per day than those without CGM but it should be noted that CGM systems cannot eliminate the need for self monitoring blood glucose using an exogenous meter. Whereas insulin requirements were similar in both treatment groups, patients in the control group had a lower proportion of basal insulin accompanied by a lower number of daily boluses. This may be a consequence of the additional historical, current and predictive information from CGM allowing greater accuracy and application of treatment decisions. One may speculate that the use of CGM from the diagnosis of the disease allows patients to gain insight into their glycaemic homeostasis that may have a long-lasting educational effect and, therefore, may influence their long-term therapeutic handling of diabetes.

Although no differences were observed in the tests for quality of life and disease adjustment between the groups at any time points, the scores compared favourably with those from other studies [22]. Of particular note, a recent German study of parents’ wellbeing assessed with the WHO-5 questionnaire from the onset up to 1 year of type 1 diabetes in 81 children (8.1 ± 2.9 years) with multiple injection therapy (K. Lange and T. Danne, unpublished data) showed a comparably low score at diagnosis (48 ± 28) with a significantly slower recovery after 6 months (52 ± 21, p < 0.03) and still tending to be lower at 12 months (58 ± 20) than in either of the two CSII-based therapies in the present study. It may be that the greater flexibility of CSII alone may have allowed the rapid normalisation of psychosocial variables, which the questionnaires may not have been sensitive enough to detect between pumps and sensor-augmented pumps. However, there is no indication that the technical requirements of these modern therapies lead to a deterioration of health-related quality of life, as has been implicated repeatedly.

So far previous studies have shown that CGM can improve HbA1c levels and enhance the management of type 1 diabetes in patients, particularly adults, motivated to use this technology. In a very recently published report of the STAR 3 Study Group, beneficial effects of sensor-augmented pump therapy compared with injection therapy have been clearly confirmed in a large cohort of adults with inadequately controlled type 1 diabetes as well as in children over 6 years of age [23]. As type 1 diabetes predominantly occurs in young patients with an increasing rate in the last decades [24], it is essential to achieve the best metabolic results in this relevant group of patients, taking into account that continuous use of CGM would increase insulin treatment costs by approximately 19% compared with CSII use alone. However, initial glycaemic achievements are closely related to long-term glycaemic control [25], whereas poor control early on may have consequences through a ‘metabolic memory’ despite improvements during the subsequent course of the disease [26]. It is important to note that the use of sensors declined progressively from the initial frequency of 2.1 sensors per week to 1.1 at 52 weeks in our study, indicating that CGM acceptance decreased over time and that repeated diabetes education for patients and parents is mandatory. Few physicians who treat children with type 1 diabetes would argue that the technology described in this paper makes it easier to achieve and maintain good glycaemic control in most patients. Any tool that improves the parent–patient knowledge is also valuable, but only if used effectively. The reason this study showed no difference between the groups may reflect the quality of education and care delivered to all of our patients and certainly the results in both groups are excellent, although data on preservation of beta cell function is less convincing.

The results of the present study indicate that children with type 1 diabetes can benefit from using a sensor-augmented insulin pump from the diagnosis of their disease, both in terms of achieving good glycaemic control and decreased glycaemic variability. According to more frequent sensor use and improved glycaemic control at week 52, the younger age groups seem to profit mostly from sensor-augmented pump treatment in our study. However, the most important question is whether sensor-augmented pump therapy is universally better and whether it is cost effective for benefit gained. Another question is whether or not it is better to apply this therapy from diagnosis rather than moving through a slower, staged education and therapy process, especially given the C-peptide results. Further work is needed to assess whether a decrease of glycaemic variability during the first year of clinical diabetes may have protective effects on beta cell function that are not reflected by the measurement of HbA1c.