Hegarty 2016a.

Study characteristics
Methods	Two centre randomised, double‐blind, placebo‐controlled trial
Participants	416 infants (one infant randomised in error) Inclusion criteria: infants of diabetic mothers or late preterm infants (35 to 36 weeks' gestation) or small‐for‐gestational age (< 2.5 kg or < 10th percentile) or large‐for‐gestational age (> 4.5 kg or > 90th percentile) infants Exclusion criteria: major congenital abnormality, previously fed by formula or received intravenous fluid, previous diagnosis of hypoglycaemia, admission to NICU or imminent admission to NICU Setting: 2 hospitals providing maternity and neonatal services (Auckland City Hospital and Waitakere Hospital) in Auckland, New Zealand Timing: Recruitment: August 2013 to November 2014 Follow‐up: August 2015 to February 2017
Interventions	40% dextrose gel massaged into the buccal mucosa as a single dose (0.5 mL/kg or 1 mL/kg at 1 hour) or multiple doses (additional 0.5 mL/kg 3 times pre‐feed in first 12 hours) (n = 277) vs Placebo gel massaged into the buccal mucosa using same protocol and volume as the intervention (n = 138)
Outcomes	Primary outcome Hypoglycaemia, defined as any blood glucose concentration < 2.6 mmol/L in the first 48 hours after birth ‐ lower incidence in any dextrose group Secondary outcomes Admission to NICU (defined as admission for > 4 hours) ‐ no difference between any dextrose and placebo groups Admission to NICU for hypoglycaemia ‐ reduced admission in any dextrose group Hyperglycaemia (blood glucose concentration > 10 mmol/L) ‐ no cases Breastfeeding at discharge from hospital (full or exclusive) ‐ no difference between any dextrose and placebo groups Receipt of any formula before discharge from hospital ‐ no difference between any dextrose and placebo groups Formula feeding at 6 weeks of age ‐ no difference between any dextrose and placebo groups Cost of care until discharge home (to be reported separately) Maternal satisfaction at 6 weeks ‐ no difference between any dextrose and placebo groups Neurosensory disability at two years' corrected age (defined as any of: legal blindness; sensorineural deafness requiring hearing aids; cerebral palsy; Bayley‐III cognitive, language or motor score > two standard deviations below the mean) ‐ these data were received from the study authors because the publication used a cut‐off on the Bayley‐III of more than one standard deviation below the mean. Developmental delay (defined as Bayley‐III cognitive, language or motor composite score > one standard deviation below the mean) measured at two years' corrected age ‐ additional data received from the study authors showed no difference between any dextrose and placebo groups Executive function z score below ‐1.5 within the cohort measured at two years' corrected age ‐ unadjusted model showed a reduced risk of low executive function in any dextrose group (RR 0.48, 95% 0.23 to 0.99), but no difference in the adjusted model (aRR 0.50, 95% CI 0.24 to 1.06). Cerebral palsy measured at two years' corrected age ‐ no cases Deafness defined as requiring hearing aids measured at two years' corrected age ‐ not estimable as only one case Blindness defined as visual acuity < 3/60 or > 1.3 logMAR measured at two years' corrected age ‐ no cases
Notes	This trial was funded by the A+ Trust (www.adhb.govt.nz; A+5696); Auckland Medical Research Foundation (www.medicalresearch.org.nz; 1113012); Cure Kids (www.curekids.org.nz; 3537); and Lottery Health Research (http://www.communitymatters.govt.nz; 326844), and by philanthropic donations to the University of Auckland Foundation (www.auckland.ac.nz). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Australian New Zealand Clinical Trials Registry ACTRN12613000322730
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Quote: "We used computer‐generated blocked randomisation with variable block sizes"
Allocation concealment (selection bias)	Low risk	Centralised allocation. Quote: "Research staff entered demographic and entry criteria data into an online randomisation website that provided a number corresponding to a numbered trial pack"
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Clinicians, families, and all study investigators were masked to treatment group allocation throughout the study and remain so for planned follow‐up
Blinding of outcome assessment (detection bias) All outcomes	Low risk	At 24 months’ corrected age, children underwent a comprehensive assessment of neurodevelopment, growth and general health by doctors trained in all assessments who were unaware of the child’s randomisation group
Incomplete outcome data (attrition bias) All outcomes	Low risk	Except for the one infant randomised in error, all infants had primary outcome data available and were included in the intention‐to‐treat analysis. The follow‐up rate at two years' corrected was high (87% of those randomised and 90% of those eligible for follow‐up) and maternal and child characteristics were similar between those who were and were not assessed at two years of age.
Selective reporting (reporting bias)	Low risk	All prespecified outcomes are reported. Trial registration was viewed
Other bias	Low risk	Demographic and baseline characteristics were similar for all randomisation groups. Primary risk factors for hypoglycaemia were similar across all treatment groups. The authors concluded that the study is not at risk of other bias.

NICU: neonatal intensive care unit; SCBU: special care baby unit; vs: versus; aRR: adjusted risk ratio