Starting resuscitation with extra oxygen did not change what preterm babies needed

When a preterm baby needs help breathing in the first minutes after birth, clinicians must choose how much oxygen to give. Too little risks leaving the newborn short of oxygen; too much exposes immature tissue to the harms of excess. For the most premature infants, guidelines lean toward starting low. For babies born a little earlier than term — 32 to 35 weeks — the right starting concentration has never been settled. AIROPLANE, published in JAMA on 9 September, set out to settle it, and found that the choice made little difference to the outcome it was built around [s1].

What the trial did

AIROPLANE was an unblinded, cluster randomised crossover trial run in 26 Australian maternity hospitals — a deliberately broad mix of tertiary, non-tertiary, metropolitan, regional, public and private sites [s1]. Rather than randomising individual babies, it assigned each hospital to start delivery-room resuscitation with an inspired oxygen fraction of either 0.30 or 0.21 (room air), then had each site cross over to the other setting halfway through its recruitment period [s1]. Newborns at 32 to 35 weeks who began respiratory support within three minutes of birth were enrolled under a waiver of consent, between December 2022 and September 2025, with final follow-up in March 2026 [s1].

A total of 1,818 eligible newborns were included, with a mean gestational age of 34.1 weeks, a mean birth weight of 2,165 g, and 44.7% female [s1]. The primary outcome was whether a baby was still receiving respiratory support when leaving the delivery room [s1].

What it found

On that primary measure, the two strategies were indistinguishable. Ongoing respiratory support on leaving the delivery room occurred in 700 of 964 newborns (72.6%) started on 30% oxygen and in 626 of 854 (73.3%) started on room air — a risk difference of −0.83 percentage points, with a 95% confidence interval from −4.33 to 2.67 [s1]. An interval straddling zero like that is the signature of a null result: the data are consistent with a small benefit either way, and with none.

The cluster crossover design is itself a strength: because each hospital used both settings at different times, differences between sites — staffing, equipment, patient mix — are less able to masquerade as an effect of oxygen. That makes the flat primary result harder to dismiss as noise.

The picture among the 12 secondary outcomes was mixed, and worth reading carefully. Ten showed no significant difference; two favoured the higher starting concentration [s1]. Babies started on 30% oxygen were less likely to escalate to the most intensive delivery-room measures — cardiac compressions or epinephrine were used in 19 of 964 (2.0%) versus 25 of 854 (2.9%), a proportional odds ratio of 0.70 (95% CI 0.52 to 0.95) [s1]. Fewer also needed endotracheal ventilation beyond the delivery room: 61 of 959 (6.4%) versus 79 of 852 (9.3%), a risk ratio of 0.69 (95% CI 0.47 to 0.91) [s1]. Rates of non-invasive positive pressure ventilation were similar, at 606 of 964 (62.9%) and 560 of 854 (65.6%) [s1].

How to read it

The honest summary is the one the trial reached: for babies born at 32 to 35 weeks, starting with 30% oxygen rather than room air did not change the proportion still needing breathing support as they left the delivery room [s1]. That is a genuinely useful null. It tells clinicians that neither starting point is clearly better on the headline measure, which removes a source of practice variation without demanding change.

The secondary signals — fewer escalations to the most aggressive interventions in the 30% group — are the kind of finding that generates the next study, not the kind that settles current practice. Secondary outcomes are not protected against chance the way the primary is, and with 12 of them, two crossing significance is close to what one would expect by coincidence [s1]. The authors present them as they should be presented: as hypotheses.

Two design features shape how far the result travels. The trial was unblinded, so clinicians knew which oxygen setting they were using, which can influence how aggressively support is escalated [s1]. And it was conducted entirely in Australian hospitals under a specific set of resuscitation practices; the balance might differ in settings that intervene earlier or later [s1].

What to watch

AIROPLANE's protocol is registered on the Australian New Zealand Clinical Trials Registry as ACTRN12621001267842 [s2]. The open question the secondary outcomes raise — whether a modestly higher starting oxygen concentration reduces the need for the most intensive interventions without adding harm — is the one a future, adequately powered trial would need to answer directly before practice moves.