Study design

Babies were recruited from the neonatal intensive care unit at University of Cambridge Addenbrooke’s Hospital NHS Trust. All parents of eligible babies admitted to the unit within 48 hours of birth were approached (if a member of the research team was available to do so). The study applied a randomised, open-label, parallel design, with babies randomised to CGM alone supported by a paper algorithm or to CGM with an additional intervention period of closed-loop CGM. Ethics Committee approval, and informed written parental consent were obtained prior to study procedures. Eligibility criteria included birth weight <1200 g and age <48 hours. Babies were excluded if they had a major congenital malformation, an underlying metabolic disorder or the mother’s pregnancy had been complicated by diabetes. Criteria were chosen based on previous data that identified these infants as at most risk from glucose dysregulation, while avoiding potential bias.22

All babies had real-time (CGM) inserted which remained in situ for up to 7 days. The paper guideline advised on the use of insulin or additional dextrose support. For a prespecified period of 24 hours, between 48 and 72 hours postbirth, a closed-loop system controlled glucose in babies during the closed-loop intervention, whereas babies in the control group continued to use CGM alongside the paper guideline. This window for closed-loop intervention was preselected as to be a standard time from birth in all babies, to allow time for informed parental consent to be obtained and as previous data (from masked CGM) had shown this to be the time when hyperglycaemia was most prevalent.22 Randomisation applied the minimisation methods using the Minim randomisation software,23 with stratification according to gestational age and birth weight.

Common study procedures

Apart from glucose control over the prespecified period, all other aspects of care were identical between treatment groups. BG monitoring was undertaken on the point-of-care BG metre StatStrip  metre (Nova Biomedical, Waltham, Massachusetts, USA). Actrapid insulin (Novo Nordisk, Bagsværd, Denmark) in concentration of 25 U/kg in 50 mL of 0.9% saline was used in both treatment groups. For CGM, an Enlite sensor (Medtronic, Watford, UK) was inserted by hand into the lateral thigh of each baby, and linked to the Paradigm Veo (Medtronic) for display of the sensor glucose (SG) concentration. Outside the intervention period (48–72 hours), CGM data were used in combination with the paper algorithm, by the clinical team to guide glucose control in all babies. Nurses calibrated the CGM at least once every 12 hours with a BG measurement taken on the point-of-care BG metre.

Paper algorithm

The paper algorithm for insulin delivery has previously been described (see online supplementary appendix 1).14 It provided guidance based on the absolute SG value as well as glucose trends. If SG levels were outside of the target range it advised to review the clinical care, dextrose infusion rate and to consider modifying insulin delivery, which could be instigated by the nurse at the cot side. The insulin and dextrose were delivered by Alaris pumps (CareFusion, San Diego, California, USA).

Closed-loop glucose control between 48 and 72 hours

The closed-loop system comprised (i) Enlite sensor, (ii) a laptop computer running a model predictive control algorithm and (iii) two Alaris syringe pumps. We used a control algorithm based on the model predictive control approach,20 optimised and tuned in silico using a computer simulation environment validated for glucose control in the critically ill,24 and aiming to keep SG between 4.0 and 8.0 mmol/L. The algorithm calculated insulin requirements or, at low glucose values, 20% dextrose infusion rates based on real-time SG values. A study nurse entered SG values into the laptop and modified insulin and dextrose pumps as directed by the control algorithm every 15 min. During the closed-loop intervention, actrapid insulin in concentration of 5 U/kg in 50 mL of 0.9% saline was used to allow for finer dose titration in accordance with the algorithm. The algorithm calculations used a compartment model of glucose kinetics25 describing the effect of insulin on SG excursions. The algorithm was initialised using a baby’s weight and adapted itself to a particular baby by updating two model parameters—a rapidly changing glucose flux correcting for errors in model-based predictions and a slowly changing estimate of an insulin rate to maintain normoglycaemia. The individualised model forecasted plasma glucose excursions over a 1.5 hour prediction horizon when calculating the insulin rate and a 30–40 min horizon when calculating the dextrose rate. Information about enteral or parenteral nutrition was not provided to the algorithm. A reference BG value was used every 6 hours for calibration of glucose sensor. If sensor readings were not available due to sensor failure or loss of data capture, then hourly BG levels were used to inform the algorithm for up to 4 hours. At this time, the algorithm continued to provide advice every 15 min.

Assessments and data collection

Demographic and clinical characteristics were collected with antenatal variables defined as: antenatal steroids as having received at least one dose prior to delivery, prolonged rupture of membranes as rupture >24 hours prior to delivery, hypertension and chorioamnionitis were based on clinical diagnoses recorded in the maternal medical file. All BG measurements, insulin administration, type and volume of enteral and parenteral nutrition and additional intravenous glucose administration were recorded from the time of randomisation to the end of CGM.

Statistical analysis

The outcome measures and the statistical analysis plan were agreed in advance. Primary outcome being comparison of time SG in target glucose range 4.0–8.0 mmol/L between study arms. Secondary outcomes were time spent with SG levels between 2.6 and 10.0 mmol/L, prevalence of hyperglycaemia (per cent time SG >10.0 mmol/L), mean and SD of SG. Safety end points included frequency of hypoglycaemia (any BG <2.6 mmol/L) and other adverse events. Sample size was selected to reflect this study as a pilot, to ensure we had comparable control data, while minimising patient exposure. Formal power calculations were not performed, and all analyses are based on intention to treat. Unpaired t-test was used to compare normally distributed variables. Non-normally distributed variables were compared using Mann-Whitney U test. Calculations were carried out using SPSS V.23 (IBM Software, Hampshire, UK). Values are given as mean, SD or median (IQR). P value <0.05 was considered to be statistically significant.

Glucose control, insulin and dextrose administration between 48 and 72 hours

There was no difference in the baseline mean SG at 48 hours between study groups (table 2). During the intervention period, the median (IQR) time spent in the target range (SG 4.0–8.0 mmol/L) was significantly higher in babies in the closed-loop group 91% (78, 99) compared with controls 26% (6-64); p<0.001—equivalent to 22 (19-24) hours in the ‘closed loop’ versus 6 (1-15) hours in the control group. Similarly, the time spent in the wider target range 2.6–10.0 mmol/L was higher in the closed-loop group: median 100% compared with control group, median 84%. This was due to the smaller per cent of time with SG values >10 mmol/L with median 16% in the control group compared with median 0% in the closed-loop group. There was no difference in the time spent with SG levels <2.6 mmol/L. Lower SG was observed in the closed-loop group median (IQR) 6.2 (6.1-7.1) mmol/L compared with the control group 8.6 (7.4-11.1) mmol/L (p=0.002). Glucose variability as measured by the SD of SG was similar (p=0.604).

The summative glucose profiles and insulin infused are provided in figure 1. Nine out of the 10 babies in the closed-loop group had received insulin prior to the 24 hours intervention with the one remaining baby being started on insulin during the 24 hours closed loop. This compared with four babies in the control arm having received insulin prior to, and eight babies receiving insulin during the intervention period. Four babies in the closed-loop study arm received additional 20% dextrose for short periods during the intervention period (up to 3.5 hours). The mean infusion rate in these babies ranged from 0.13 to 0.53 mL/kg/hour. The highest rate being infused in a baby who was hypoglycaemia prior to the start of the closed-loop intervention. There was no statistical difference in the total amount of insulin infused or nutritional intake between study groups during the 24 hours intervention period.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Glucose control and insulin delivery median (IQR) of sensor glucose and insulin infused in babies randomised to closed-loop management or continuous glucose monitoring with paper algorithm (control). The closed-loop intervention period is denoted by the vertical lines, and the target glucose range 4.0–8.0 mmol/L is denoted by horizontal lines.

Glucose control, insulin and dextrose administration between 72 and 160 hours

In the postintervention period, after 72 hours, there was a trend of increased time in both glucose target ranges (4.0–8.0 and 2.6–10.0 mmol/L) in the closed-loop group compared with the control group (table 3), but these differences did not reach statistical significance.

Nutrition and clinical care

All babies received parenteral and enteral nutrition according to the standard local neonatal unit protocol. During the closed-loop intervention period between 48 and 72 hours, four babies in each study group were receiving minimal amounts of trophic feeds. In the postintervention period, after 72 hours, there was no difference in the volumes of milk received between the two study groups (table 3). None of the babies received hydrocortisone, but seven babies in the control arm and two in the intervention arm received inotropes during the first week. During the intervention period, the median (IQR) number of BG values taken per day was 5.50 (4.75-6.56) in the control group compared with 6.25 (5.13-6.5) in the intervention group and over the whole study period 5.14 (4.29-5.43) in the control compared with 5.29 (4.64-5.80) in the intervention group.

Safety

There were no reported concerns about the sensor site in terms of inflammation or infection either during the study or after removal. In the closed-loop study group, there were two babies who had documented episodes of hypoglycaemia with BG <2.6 mmol/L. One episode occurred when checking the baseline BG prior to the onset of closed loop, at this time maintenance fluids were being changed, but no insulin was being infused. The second episode was on day 6, again associated with a change of maintenance fluids. There were a further two babies who had periods (after the 72 hours closed loop) when SG fell to below 2.6 mmol/L but the BG at this time was documented above 2.6 mmol/L. In the control study group, no babies had a documented BG value <2.6 mmol/L. One baby in the control group had an episode lasting 205 min when the SG fell to <2.6 mmol/L (BG was not checked at this time). None of the babies were on insulin, and there was no clinical evidence of hypoglycaemia in these babies during any of these episodes.