Objective To systematically review outcomes of infants ≤28+6 weeks gestation randomised to resuscitation with low (≤0.3) vs high (≥0.6) fraction of inspired oxygen (FiO₂) at delivery.

Design Systematic review of randomised controlled trials of low (≤0.3) vs high (≥0.6) FiO₂ resuscitation. Information was obtained from databases (Medline/Pub Med, EMBASE, ClinicalTrials.gov, Cochrane) and meeting abstracts between 1990 to 2015. Search index terms: preterm/ resuscitation/oxygen. Data for infants ≤28+6 weeks gestation were independently extracted and pooled using a random effects model. Analyses were performed with Revman V.5.

Main outcome measures Death in hospital, bronchopulmonary dysplasia (BPD), retinopathy of prematurity >grade 2 (ROP), intraventricular haemorrhage >grade 2 (IVH), patent ductus arteriosus (PDA) and necrotising enterocolitis (NEC).

Results A total of 251 and 253 infants were enrolled in 8 studies (6 masked, 2 unmasked) in the lower and higher oxygen groups, respectively, (mean gestation 26 weeks) between 2005 and 2014. There were no differences in BPD (relative risk, 95% CIs 0.88 (0.68 to 1.14)), IVH (0.81 (0.52 to 1.27)), ROP (0.82 (0.46 to 1.46)), PDA (0.95 (0.80 to 1.14)) and NEC (1.61 (0.67 to 3.36)) and overall mortality (0.99 (0.52 to 1.91)). Mortality was lower in low oxygen arms of masked studies (0.46 (0.23 to 0.92), p=0.03) and higher in low oxygen arms of unmasked studies (1.94 (1.02 to 3.68), p=0.04).

Conclusions There is no difference in the overall risk of death or other common preterm morbidities after resuscitation is initiated at delivery with lower (≤0.30) or higher (≥0.6) FiO₂ in infants ≤28+6 weeks gestation. The opposing results for masked and unmasked trials may represent a Type I error, emphasising the need for larger, well designed studies.