Phelps 2018.

Methods	Randomised clinical trial included infants enrolled from 18 neonatal intensive care centres throughout the USA from 17 April 2014 to 4 September 2015; final date of follow‐up was 12 February 2016.
Participants	638 infants < 28 weeks’ PMA, surviving for at least 12 hours, and admitted to 1 of the 18 Neonatal Research Network centres before 72 hours’ postnatal age.
Interventions	A 40 mg/kg dose of myo‐inositol was given every 12 hours (initially intravenously, then enterally when feeding; (N = 317) or for up to 10 weeks. The active drug was an isotonic, sterile, pyrogen‐ and preservative‐free aqueous solution of 5% myo‐inositol (50 mg/mL) at neutral pH and was provided by Abbott Laboratories. A dose of 40 mg/kg every 12 hours was selected to achieve serum concentrations similar to those previously reported. A therapeutic duration of up to 10 weeks was chosen to sustain serum myo‐inositol levels similar to those found in utero throughout the period of normal retinal vascular development and because of the reported benefits in the treatment of ROP and survival. Placebo (N = 321) (5% glucose for IV use from pharmacy stock).
Outcomes	The unfavourable primary outcome was type 1 ROP, which was defined as meeting the criteria for ophthalmological intervention to prevent retinal detachment, a more severe ROP type than ROP type 1 (e.g. aggressive posterior ROP or Rush disease), or death before the ROP outcome could be determined.
Notes	The planned enrolment of 1760 participants would permit detection of an absolute reduction in death or type 1 ROP of 7% with 90% power. The trial was terminated early due to a statistically significantly higher mortality rate in the myo‐inositol group. The favourable primary outcome was survival with only milder ROP or no ROP. Infants were followed up as outpatients to determine the primary outcome up to a maximum of 55 weeks’ PMA.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Computer‐generated and centrally administered randomisation.
Allocation concealment (selection bias)	Low risk	With the exception of pharmacists, who prepared the daily unit doses of myo‐inositol or placebo according to randomisation assignment, all other clinical and research personnel and families were blind to group assignment.
Blinding (performance bias and detection bias) All outcomes	Low risk	See above for allocation concealment.
Incomplete outcome data (attrition bias) All outcomes	Low risk	Outcome group: of the 317 infants randomised to receive myo‐inositol 313 received the intervention as randomised; 3 died prior to receiving the intervention and 1 was withdrawn prior to intervention. 313 included in primary analysis. Of the 321 infants randomised to placebo 315 received the intervention as randomised; 2 died prior to receiving intervention and 4 received < 2 doses of myo‐inositol. 320 included in the primary analysis of type 1 retinopathy of prematurity or death. 1 excluded from primary analysis.
Selective reporting (reporting bias)	Low risk	The protocol for the study was available to us and we did not identify any deviations except that the study was terminated prior to reaching the full sample size because of safety concerns. The authors state: “No changes to the protocol occurred during the trial”. Registered as: NCT01954082.
Other bias	Low risk	Appears free of other bias.

BW = birth weight

PMA = postmenstrual age

PK = pharmacokinetics