June 15, 2021


Beardsall K, Thomson L, Guy C, Iglesias-Platas I, van Weissenbruch MM, Bond S, Allison A, Kim S, Petrou S, Pantaleo B, Hovorka R, Dunger D, and the REACT collaborative. Real-time continuous glucose monitoring in preterm infants (REACT): an international, open-label, randomised controlled trial. Lancet Child Adolesc Health 2021; PMID: 33577770.


Daniela Dinu
Assistant Professor
Baylor College of Medicine, Houston, TX, United States

Paul Rozance
University of Colorado, Aurora, CO, United States




In preterm infants <34 weeks’ gestation and <1200g does real-time continuous glucose monitoring (CGM), compared with standard of care intermittent monitoring, increase the time spent within glucose target range (2.6-10 mmol) during the first week after birth?


  • Design: open label, randomized controlled trial
  • Allocation: babies were randomized 1:1 using a central web randomization system, stratified by recruiting center and gestational age (<26 weeks or ≥26 weeks)
  • Blinding: not feasible
  • Follow-up period: 7 days of life for primary and secondary outcomes. Exploratory outcomes were reported at 1 week (nutritional intake), 2 weeks (use and dose of insulin), 36 weeks (mortality, BPD, growth), and at hospital discharge (ROP, intracerebral pathology, infection, NEC, PDA).
  • Setting: 13 NICUs in the UK, Spain and the Netherlands
  • Patients: 180 randomly assigned
    • Inclusion Criteria:
      • <24 hours of age
      • Parental written informed consent
      • BW ≤1200 g
      • GA <34 weeks
    • Exclusion criteria:
      • Any lethal congenital malformation
      • Any congenital metabolic disorder
  • Intervention: All infants had an Enlite glucose sensor (Medtronic, Northridge, CA, USA) inserted subcutaneously into the thigh connected to a transmitter. For the intervention group real-time CGM values were available and sensor glucose was recorded every one hour. The clinical staff made adjustments in the rate of dextrose infusion or dose of insulin based on a specifically designed protocol using the CGM data. In the standard care group glucose was monitored according to the local standard clinical practice using intermittent sampling and all the clinical care was at the discretion of the local staff. The real-time CGM data was still collected in the standard care group, but values were masked.
  • Outcomes:
    • Primary efficacy outcome: proportion of time that sensor glucose concentrations were in the target range (2.6-10 mmol/L) up to 7 days of life
    • Secondary efficacy outcomes:
      • Proportion of time sensor glucose concentrations were in the target range of 4–8 mmol/L
      • Cverall mean sensor glucose concentration
      • Sensor glucose concentration variability (assessed by within-patient standard deviation)
      • Proportion of time that sensor glucose concentrations were in the severe hyperglycemic range (>15 mmol/L)
    • Safety outcomes:
      • Incidence of hypoglycemia: any recorded blood glucose concentration of 2.2–2.6 mmol/L or any continuous episode of sensor glucose concentration of <2.6 mmol/L for >1 h
      • Incidence of severe hypoglycemia: any recorded blood glucose ≤2.2 mmol/L
    • Clinical outcomes:
      • Mortality before 36 weeks corrected gestational age
      • Retinopathy of prematurity (maximum grade across all examinations)
      • Bronchopulmonary dysplasia (need for supplemental oxygen or respiratory support at 36 weeks corrected gestational age)
      • Infection (confirmed microbiologically or clinically suspected late onset invasive infection) until hospital discharge
      • Necrotizing enterocolitis requiring surgical intervention or causing death
      • Patent ductus arteriosus requiring medical or surgical treatment
      • Intracerebral pathology before discharge
      • Growth at the end of 1 week and at 36 weeks corrected gestational age
      • Nutritional intake during the first week of life
      • Use of insulin during the first two weeks of life
  • Analysis and Sample Size:
    • 182 infants were enrolled, 180 were randomized
    • 85 assigned to CGM, 70 had CGM data and included in the primary and the secondary analyses
    • 95 assigned to standard care, 85 had CGM data and included in the primary and the secondary analyses
  • Patient follow-up: 11 patients withdrew, there were 8 deaths (2 in the CGM group and 6 in the standard care group), 1 patient was excluded from the final analysis due to ineligibility (older than 24 hours of age). 155 patients were included in the final analysis.


Randomized infants had similar baseline characteristics (Table 1).

Table 1. Main characteristics of infants
Intervention group (CGM)

n= 84

Standard of care group


Mean gestational age ± SD (range) 27.7 ± 2.1 (24.0-33.7) 27.4 ± 1.9 (23.3-31.3)
Birth weight g ± SD 910 ± 160 880 ± 180
Birth weight standard SD score ± SD -0.82 ± 0.98 -0.80 ± 0.92
Male % (n) 54% (45) 47% (45)
Received steroids >24h prior to delivery % (n) 82% (69) 73% (69)
Maternal diabetes % (n) 6% (5) 8% (8)

Infants in the intervention arm spent more time having glucose within target range (94%) compared with the standard of care group (84%), equivalent to 13 hours, and spent less time in the hyperglycemic range (0.3% versus 1.3%). While infants in the intervention arm had more episodes of hypoglycemia (sensor glucose = 2.2-2.6 mmol/L) than the standard of care group: 15% compared to 12% and severe hypoglycemia (blood glucose <2.2 mmol/L) 13% compared to 7%, they had less continuous episodes of sensor glucose <2.6 mmol/L lasting for more than one hour and spent less time in hours having sensor glucose <2.6 mmol/L (Table 2).

Table 2 Main results
Intervention group (CGM) n=70 Standard of care group n=85 P value
Proportion of time sensor glucose was 2.6-10 mmol/L
Mean 94% 84% 0.005
Median 99% 97%
Sensor glucose concentration mmol/L
Mean 6.5 ± 1.2 7 ± 2.1 N/A
Median 6.4 (5.8-9.9) 6.5 (5.2-8.4) N/A
Proportion of time in hyperglycemic range (sensor glucose >15 mmol/L)
Mean 0.3% ± 2.1 1.3% ± 4.2 N/A
Median 0.0% (0.0-0.0) 0.0% (0.0-0.0) N/A
Proportion of time in hypoglycemia range (sensor glucose <2. 6 mmol/L)
Mean 1.0 ± 5.3 1.1 ± 3.2 N/A
Median 0.0 (0.0-0.2) 0.0 (0.0-0.1) N/A
Number of infants with episodes of blood glucose 2.2-2.6 mmol/L 14.7% (11/75) 11.8% (11/93) 0.7
Number of infants with episodes of severe hypoglycemia (blood glucose <2.2 mmol/L) 13.3 (10/75) 6.5 (6/93) 0.2
Continuous episode of sensor glucose <2.6 mmol/L) for more than 1h 5.7% (4/70) 15.3% (13/85) 0.1
Length of time with sensor glucose <2.6 mmol/L in hours mean ± SD 0.5  ± 1.7 1  ± 3.2

Although underpowered to detect differences in mortality and morbidities, the exploratory analyses showed a tendency towards decreased mortality and, interestingly, lower rates of NEC in the intervention arm, potentially related to less time spent in the hyperglycemic range for this particular group (Table 3).

Table 3. Clinical outcomes/Exploratory analyses
CGM/Intervention group Standard group P value
Mortality 2% (2/84) 6%  (6/95) 0.13
Infection 58%  (50/74) 62% (53/85) 0.99
NEC 13% (10/75) 28% (24/85) 0.014
PDA 26 % (19/74) 31% (26/85) 0.16
Intracerebral pathology 33% (25/75) 32% (27/84) 0.95
Maximum ROP 1.1 (3.8) 1.5 (3.6) 0.64
Weight standard deviation score day 7 -1.26 (0.79) -1.3 (0.75) 0.69
Total insulin week 2 (units/kg) 1.1 (3.0) 1.1 (2.6) 0.82


The authors conclude that real-time CGM can reduce exposure to prolonged or severe hyperglycemia and hypoglycemia.


Continuous interstitial glucose monitoring (CGM) is used to assess glycemic status in some adults and older children with diabetes (1, 2). Additionally, CGM has been used in term, preterm, and low birth weight newborns to evaluate glucose concentrations during the first days of life (3), to evaluate efficacy of dextrose gel in preventing hypoglycemia (4), or to detect episodes of low glucose which would go unrecognized with standard, intermittent blood sampling (3, 4). CGM could potentially decrease the number of invasive capillary samplings (5).

In the current study, the efficacy of CGM in maintaining glucose within a specified range was investigated. Overall, the intervention arm spent more time within the target range (94%) compared to the standard group (84%) or a total of 13 more hours. Although not statistically significant, more infants in the intervention group (15% versus 12%) experienced at least one episode of hypoglycemia (glucose = 2.2-2.6 mmol/L) with 13% in the intervention group versus 7% in the standard group having at least one episode of severe hypoglycemia (glucose <2.2 mmol/L), consistent with previous studies showing that hypoglycemic episodes might go unnoticed if relying solely on intermittent sampling.  Alternatively, the different incidence of hypoglycemia could be explained by higher rate of insulin use in the intervention arm (61%) compared to the standard group (37%) and a post-hoc analysis of infants receiving insulin might have clarified this possibility. The intervention group had less continuous episodes longer than one hour with sensor glucose <2.6 mmol/L and less time in the hyperglycemic range. Use of CGM was safe regarding episodes of infection or discomfort.

The protocol’s (6) recommendation to start or increase the insulin dose when glucose levels reach 8 mmol/L might explain why more infants in the intervention arm received insulin in the first week of life compared with the standard group. The total dose of insulin was similar between groups, raising the possibility that infants in the intervention arm might have received insulin for transient elevated glucose levels. Different glucose thresholds and different durations of elevated levels before starting insulin might change the impact of CGM on insulin exposure.

One interesting exploratory finding was a higher incidence of NEC in the standard group, possibly due to more time spent in the hyperglycemic range. This deserves further investigation.

The authors did not mention the proportion of growth restricted (IUGR) or small for gestational age infants. It is worth considering whether IUGR or other subgroups of infants with increased glycemic variability in the first week of life (infants of diabetic mothers, congenital infections, HIE) would have demonstrated a different efficacy of CGM.

This study adds to the mounting evidence that CGM in preterms used to achieve glycemic control is safe and efficacious. It extends the results from previous smaller studies to a larger, multicenter one. The optimal glucose concentration target range for preterms was not addressed. However, the results show that CGM can be used to improve future studies investigating different target glucose ranges in relationship to short- and long-term outcomes.


Dr. Rozance has received a StatStrip Glucose Meter from Nova Biomedical for use in his research laboratory.


  1. Danne T, Nimri R, Batellino T, Bergenstal RM, Close KL, DeVries HJ et al. L. International Consensus on Use of Continuous Glucose Monitoring. Diabetes Care 2017; 40:1631-1640.
  2. Klonoff DC, Buckingham B, Christiansen JS, Montori VM, Tamborlane WV, Robert A. Vigersky RA, et al. Continuous Glucose Monitoring: An Endocrine Society Clinical Practice Guideline. J Clin Endocrinol Metab 2011; 96(10):2968–2979.
  3. Harris DL, Weston PJ, Gamble GD, Harding, JE. Glucose Profiles in Healthy Term Infants in the First 5 Days: The Glucose in Well Babies (GLOW) Study. J Pediatr 2020; 223:34-41.e4.
  4. Harris DL, Weston PJ, Signal M, Chase GJ, Harding, JE. Dextrose gel for neonatal hypoglycaemia (the Sugar Babies Study): a randomized, double-blind, placebo-controlled trial. Lancet 2013; 382:2077-83.
  5. Uettwiller,F,, Chemin A, Bonnemaison E, FavraiesG, Saliba E, Labarthe F. Real-Time Continuous Glucose Monitoring Reduces the Duration of Hypoglycemia Episodes: A Randomized Trial in Very Low Birth Weight Neonates. PLoS ONE 10(1):e0116255sed.
  6. Beardsall K, ThormsonL, Guy C, van Weissenbruch MM, Iglesias I, Muthukumar P et al. Protocol of a randomised controlled trial of real time continuous glucose monitoring in neonatal intensive care ‘REACT’ BMJ Open 2018:8e020816.