Skip to main content

Trial Summaries & Critiques

AIROPLANE

image: shutterstock / BelezaPoy

Premature infant in the NICU

Publication

  • Title: Oxygen vs Air at Birth for Moderate- to Late-Preterm Infants: The AIROPLANE Cluster Randomized Crossover Trial.
  • Acronym: AIROPLANE — Air or Oxygen for Preterm Newborns; An Embedded Trial.
  • Year: 2026; published online 9 September.
  • Journal published in: JAMA.
  • Citation: Peart SR, Manley BJ, Cheong JLY, Oei JL, Grobler AC, Moore CL, et al; AIROPLANE Trial Investigators and IMPACT Clinical Trials Network for Mothers’ and Babies’ Health. Oxygen vs air at birth for moderate- to late-preterm infants: the AIROPLANE cluster randomized crossover trial. JAMA. Published online 9 September 2026.

Context & Rationale

  • Background
    • Successful transition at birth requires lung aeration, increased pulmonary blood flow and adequate respiratory drive. Supplemental oxygen may facilitate this transition, but excessive exposure can generate oxidative injury.
    • Evidence concerning initial oxygen concentration has largely concerned term or very preterm infants. Direct randomised evidence for infants born at 32–35 completed weeks was lacking.
    • The 2025 ILCOR consensus identified insufficient evidence to recommend an initial oxygen concentration specifically at 32–34+6 weeks; its recommendations for infants below 32 weeks and those at or above 35 weeks differed. This is a more precise description of the evidence gap than treating all infants at 32–35 weeks as one guideline category.1
    • A prospective cohort of 76 moderate- to late-preterm infants requiring resuscitation found substantial difficulty attaining early oxygen saturation targets. Its observational design supported the need for a trial but could not establish the superiority of supplemental oxygen over air.2
  • Research Question/Hypothesis
    • Would starting respiratory support with a fraction of inspired oxygen (FiO₂) of 0.30 rather than 0.21 improve early transition sufficiently to reduce ongoing respiratory support when infants left the delivery room?
    • The intervention tested a brief difference in initial oxygen concentration followed by clinically guided titration in both groups.
  • Why This Matters
    • A small change to the starting oxygen concentration is inexpensive and readily implementable wherever blended oxygen is available.
    • Demonstrating reduced respiratory support could influence neonatal admission, invasive ventilation and transfer requirements across both specialist and non-specialist maternity hospitals.
    • AIROPLANE addressed a common clinical decision in a population that had been poorly represented in resuscitation trials.

Design & Methods

  • Research Question:
    • In infants born at 32+0–35+6 weeks who commenced respiratory support within three minutes of birth, did initial FiO₂ 0.30 versus 0.21 reduce ongoing respiratory support at delivery room departure?
    • The primary outcome included continuous positive airway pressure (CPAP), nasal high-flow therapy or intermittent positive pressure ventilation through any interface. Supplemental oxygen alone did not qualify.
    • Infants who died in the delivery room were counted as having the primary outcome.
  • Study Type:
    • Investigator-initiated, pragmatic, multicentre, open-label, cluster-randomised crossover superiority trial in Australian maternity hospitals.
    • Twenty-six hospitals recruited: eight tertiary and 18 non-tertiary centres, including metropolitan and regional, public and private hospitals in Victoria and New South Wales.
    • Hospitals were allocated 1:1 to the sequence 30% then 21%, or 21% then 30%, using simple block randomisation; 13 recruiting hospitals followed each sequence.
    • Each hospital used both treatments for equal periods, with one crossover and a one-week washout during which births were excluded from analysis. Recruitment periods differed between hospitals.
    • Treatment was delivered under a waiver of informed consent. One hospital used an opt-out approach for data collection; consumer representatives contributed to trial development.3
  • Population:
    • Inclusion: gestation 32+0–35+6 weeks, active neonatal care, and respiratory support commencing within three minutes of birth.
    • Exclusion: known major cardiorespiratory or craniofacial anomalies likely to affect transition or oxygen requirements, planned palliative care, or failure to meet the gestational-age and early-support criteria.
    • Of 5,168 births within the gestational-age range, 1,828 infants received study treatment and 1,818 were analysed: 964 in the 30% group and 854 in the 21% group.
    • Mean gestation was 34.1 weeks, mean birth weight 2,165 g, and 44.7% were female.
  • Intervention:
    • Commence respiratory support with FiO₂ 0.30, delivered with the respiratory support interface and mode selected by the treating clinician.
    • While respiratory support remained necessary, maintain the allocated concentration until three minutes of age or until at least one minute of respiratory support had been delivered, whichever was later.
    • Apply a preductal pulse oximeter to the right hand or wrist.
    • Earlier oxygen escalation was permitted for a heart rate below 60 beats/min or the need for endotracheal intubation; effective ventilation and emergency resuscitation took priority.
  • Comparison:
    • Commence respiratory support with FiO₂ 0.21 under the same timing, monitoring and emergency-escalation rules.
    • After the initial treatment window, both groups underwent oxygen titration against peripheral oxygen saturation according to local resuscitation guidance and clinical judgement.
    • CPAP, positive pressure ventilation and subsequent neonatal treatment otherwise followed local practice. All respiratory support began after umbilical cord clamping; deferred cord clamping was permitted.3
  • Blinding:
    • Clinicians knew their hospital’s current allocation and recorded the delivery room outcomes.
    • The trial therefore evaluated an openly implemented oxygen policy, including the clinical decisions made under that policy.
  • Statistics:
    • A minimum of 1,200 infants across at least 20 hospitals was required to detect an eight-percentage-point reduction in the primary outcome, from 51% to 43%, with 80% power (β=0.20) and a two-sided α of 0.05, assuming an intracluster correlation coefficient of 0.02, an average of 30 infants per hospital per period and a cluster-size coefficient of variation of four.
    • The principal analysis followed hospital-period allocation among eligible infants whose data were retained, regardless of treatment actually received.
    • Mixed-effects logistic regression included fixed effects for treatment and period, with random effects for hospital and multiple birth; risk differences and risk ratios were derived from the models.
    • Prespecified sensitivity adjustment included antenatal corticosteroids, gestational age and delivery mode; subgroup analyses examined gestational age and hospital capability.
    • The highest delivery room support category was analysed using a proportional odds model. Twelve secondary outcomes were assessed without adjustment for multiple comparisons.
    • The statistical analysis plan was dated 15 July 2025 and made public before analysis; later journal publication of the protocol and analysis plan should not be mistaken for their initial preparation dates.4
  • Follow-Up Period:
    • Recruitment occurred from December 2022 to September 2025; follow-up extended to death or first discharge home, with final follow-up stated as March 2026.
    • One infant remained hospitalised at data lock; hospital duration was measured to that point.
    • Long-term neurodevelopmental outcomes were not available in this report. The protocol planned additional developmental follow-up through linkage for a subset of participants.3

Key Results

This trial was not stopped early. Recruitment followed the planned time-based approach. An independent safety committee reviewed mortality after 600 infants and recommended continuation; no interim efficacy analysis was performed.

Outcome Initial FiO₂ 0.30 Initial FiO₂ 0.21 Effect p value / 95% CI Notes
Ongoing respiratory support at delivery room departure 700/964 (72.6%) 626/854 (73.3%) RR 0.99
Risk difference −0.83 percentage points		 RR: 95% CI 0.94 to 1.03
Risk difference: 95% CI −4.33 to 2.67
P=0.64		 Primary outcome; oxygen alone excluded, delivery room death included.
Apgar scores at one and five minutes Median 6 and 8 Median 6 and 8 Median difference 0 at both times One minute: 95% CI −0.14 to 0.14
Five minutes: 95% CI −0.27 to 0.27		 No detected difference at either time.
Highest level of delivery room support CPAP/high flow: 319/964 (33.1%)
Noninvasive positive pressure ventilation: 606/964 (62.9%)
Endotracheal/supraglottic ventilation: 20/964 (2.1%)
Compressions/adrenaline: 19/964 (2.0%)		 CPAP/high flow: 245/854 (28.7%)
Noninvasive positive pressure ventilation: 560/854 (65.6%)
Endotracheal/supraglottic ventilation: 24/854 (2.8%)
Compressions/adrenaline: 25/854 (2.9%)		 Common proportional OR 0.70 95% CI 0.52 to 0.95 One ordinal comparison across mutually exclusive highest-support categories; not separate treatment effects for each component.
Maximum delivery room FiO₂ Mean 0.53 Mean 0.55 Modelled mean difference −0.01 95% CI −0.04 to 0.01 Peak oxygen concentration; does not measure cumulative exposure.
Endotracheal ventilation after delivery room departure 61/959 (6.4%) 79/852 (9.3%) RR 0.69
Risk difference −2.79 percentage points		 RR: 95% CI 0.47 to 0.91
Risk difference: 95% CI −5.30 to −0.27		 Exploratory secondary finding; no multiplicity adjustment.
Noninvasive respiratory support after delivery room departure 673/959 (70.2%) 608/852 (71.4%) RR 0.98 95% CI 0.93 to 1.04 No detected reduction.
Supplemental oxygen after delivery room departure 414/959 (43.2%) 386/852 (45.3%) RR 0.96 95% CI 0.86 to 1.06 No detected reduction.
Any respiratory support or supplemental oxygen after delivery room departure 697/959 (72.7%) 640/852 (75.1%) RR 0.97
Risk difference −2.32 percentage points		 RR: 95% CI 0.92 to 1.02
Risk difference: 95% CI −6.07 to 1.43		 No detected reduction in the broader post-delivery-room outcome.
Exogenous surfactant 73/959 (7.6%) 84/852 (9.9%) RR 0.77
Risk difference −2.46 percentage points		 RR: 95% CI 0.54 to 0.99
Risk difference: 95% CI −5.26 to 0.34		 Published intervals give discordant conclusions about exclusion of the null; see critique.
Hospital admission duration Mean 23.2 days (n=959) Mean 23.0 days (n=852) Published mean difference −0.13 days Published 95% CI −1.65 to 1.91 Values reproduced as published; the estimate and interval require clarification.
Interhospital transfer for escalating care 102/964 (10.6%) 73/854 (8.5%) RR 1.19
Risk difference 2.39 percentage points		 RR: 95% CI 0.88 to 1.49
Risk difference: 95% CI −1.28 to 6.07		 No detected difference; the point estimate favoured air.
Death before hospital discharge 2/959 (0.2%) 7/852 (0.8%) Not estimated P value and comparative CI not reported Descriptive analysis was prespecified; nine deaths cannot establish a mortality effect.
  • Primary finding: Starting with 30% oxygen did not demonstrate superiority for ongoing respiratory support at departure. The confidence interval excludes the planned eight-percentage-point benefit, but remains compatible with a smaller benefit or harm; equivalence was not tested.
  • Secondary outcomes and harms:
    • The ordinal support result and lower subsequent endotracheal ventilation rate favour 30% oxygen, but are exploratory findings among multiple correlated secondary comparisons.
    • Among infants receiving endotracheal ventilation, median age at the final day of ventilation was 2.5 days (95% CI 1.7 to 3.3) versus 3.5 days (95% CI 1.9 to 5.1). These are descriptive, treatment-conditioned summaries, not an intention-to-treat estimate of a reduction in ventilation duration.5
    • The small mortality counts and absence of comprehensive long-term outcomes preclude a reliable conclusion about comparative safety.
  • Prespecified subgroups:
    • At 32–33+6 weeks, the primary outcome occurred in 378/448 infants (84.4%) versus 297/354 (83.9%): RR 1.01; 95% CI 0.95 to 1.06.
    • At 34–35+6 weeks, it occurred in 322/516 (62.4%) versus 329/500 (65.8%): RR 0.96; 95% CI 0.89 to 1.03. The gestational-age interaction was not significant (P=0.39).
    • In non-tertiary hospitals, rates were 288/393 (73.3%) versus 239/334 (71.6%): RR 1.01; 95% CI 0.93 to 1.10.
    • In tertiary hospitals, rates were 412/571 (72.2%) versus 387/520 (74.4%): RR 0.97; 95% CI 0.91 to 1.02. The hospital-capability interaction was not significant (P=0.25).5

Internal Validity

  • Randomisation and allocation:
    • Random allocation of treatment sequence and exposure of each hospital to both strategies reduced confounding by stable institutional characteristics.
    • Allocation was known during each recruitment period. Clinicians could therefore know the assigned treatment before deciding whether an infant required respiratory support and met the clinical eligibility criterion.
    • This creates a potential for differential recruitment after cluster allocation, although the available data do not demonstrate that such selection occurred.
  • Withdrawals, exclusions and missing data:
    • Three of 29 randomised hospitals withdrew before recruiting because of changes in capability or operational and governance requirements.
    • Ten treated infants were excluded: five after parental requests to remove data and five because known major anomalies made them ineligible.
    • The analysis therefore retained allocation among eligible infants with usable consent arrangements, rather than including every treated infant.
    • Primary outcome data were complete for all 1,818 analysed infants; most hospital outcomes were available for 959 versus 852 infants. Missingness was below 1%, and the prespecified threshold for multiple imputation was not reached.
  • Baseline characteristics:
    • Mean gestation was 34.0 versus 34.2 weeks, mean birth weight 2,147 versus 2,185 g, and caesarean delivery occurred in 79.8% versus 79.5%.
    • At least two antenatal corticosteroid doses within seven days were recorded in 35.3% versus 29.7%; cord clamping at or beyond 60 seconds occurred in 41.4% versus 36.7%.
    • These imbalances are relevant to respiratory transition, but adjustment for prespecified prognostic factors did not materially change the primary conclusion: adjusted risk difference −1.59 percentage points; 95% CI −4.90 to 1.73; P=0.35.
    • The population had substantial respiratory morbidity: approximately three-quarters still required support at departure. It was not a cohort in which the primary event was too uncommon to detect benefit.
  • Protocol adherence and treatment received:
    • The exact allocated starting concentration was delivered to 882/964 infants (91.5%) versus 815/854 (95.4%). Including permitted emergency exceptions, commencement was classified as compliant in 91.9% versus 96.3%.
    • The assigned concentration was maintained throughout the specified initial window in 78.7% versus 75.2%, as tabulated in the supplement.
    • Major starting-concentration deviations occurred in 78 infants (8.1%) versus 32 (3.7%); 69 allocated to 30% actually started in air, while 24 allocated to air started at 30%.
    • Early changes classified as minor deviations occurred in 9.9% versus 16.2%; permitted changes for clinical indications occurred in 2.9% versus 3.9%.5
  • Timing, dose and separation:
    • Support commenced within the intended early transition period, but the duration of exposure depended on when support started and whether it remained necessary.
    • The assigned contrast was FiO₂ 0.30 versus 0.21, followed by titration in both groups.
    • Oxygen was subsequently increased above the allocated starting concentration in 612 infants (63.5%) versus 705 (82.7%); mean maximum FiO₂ converged to 0.53 versus 0.55.
    • Oxygen saturation trajectories, time spent below or above target, cumulative oxygen exposure and detailed timing of concentration changes were not reported. Separation in prescribed starting concentration therefore cannot establish separation in physiological oxygenation.
  • Crossover and co-interventions:
    • The planned crossover occurred at hospital level and involved different infants in each period; individual biological carryover was not relevant.
    • The one-week washout addressed staff practice and implementation carryover.
    • Later oxygen escalation was an intended component of both treatment strategies, rather than a conventional switch between fixed treatment arms.
    • Respiratory support decisions and subsequent therapies remained clinician-directed, supporting pragmatic relevance while permitting variation in co-interventions.
  • Outcome assessment and blinding:
    • Whether a support device was in use was readily observable, but the decision to continue it was clinician-dependent.
    • Delivery room departure was not a fixed physiological assessment time.
    • CPAP or nasal high flow accounted for most primary events: 672/964 infants (69.7%) versus 594/854 (69.6%) left receiving these modalities.
    • Unblinded decisions could influence both the primary outcome and the secondary outcomes concerning escalation or intubation; the direction and magnitude of any bias cannot be quantified.
  • Clustering, heterogeneity and statistical robustness:
    • The analysis accounted for hospital, period and multiple birth, matching the principal dependencies in the design.
    • The observed intracluster correlation coefficient was 0.30, compared with the assumed 0.02; the control event rate was 73.3%, compared with the assumed 51%.
    • A post hoc model allowing treatment effects to vary between hospitals produced a risk difference of −0.90 percentage points; 95% CI −5.87 to 4.06; P=0.72.
    • The treatment-by-period interaction was not significant (P=0.12), but this does not prove the absence of temporal or implementation effects.
    • The SAP allowed logistic modelling if the planned binomial approach failed to converge. The main analysis used logistic models; median comparisons used quantile regression with hospital bootstrap resampling, rather than the originally described quantile mixed model.4
    • The primary conclusion was consistent across the principal and sensitivity analyses. Secondary confidence intervals were not adjusted for multiplicity.

Conclusion on Internal Validity: Internal validity is moderate overall. Randomised treatment sequences, almost complete outcome capture and consistent sensitivity analyses support the primary estimate, while unblinded clinical decisions, incomplete adherence and limited physiological measurements constrain interpretation of the secondary signals.

External Validity

  • Population representativeness:
    • Broad recruitment under waived consent included infants requiring urgent support who might be missed by antenatal consent procedures.
    • The findings directly concern infants born at 32+0–35+6 weeks who begin respiratory support within three minutes.
    • They do not directly address infants born before 32 weeks, infants at 36 weeks, term infants, those needing support only later, or those with major known anomalies.
    • The high caesarean-delivery rate and inclusion of infants receiving relatively low-intensity support should be considered when applying the overall event rates elsewhere.
  • Clinical settings:
    • Participation by 18 non-tertiary hospitals and both metropolitan and regional centres improves applicability beyond specialist neonatal intensive care units.
    • The trial remained confined to two Australian states with established neonatal support and transfer systems.
    • Implementation requires reliable oxygen blending, preductal oximetry, respiratory support equipment and personnel able to titrate oxygen promptly.
    • Applicability is less certain where these resources are unavailable or where thresholds for neonatal admission, CPAP continuation and transfer differ substantially.
  • Care pathways:
    • Both groups received subsequent titration and rescue treatment; the findings apply to these complete care strategies.
    • All support began after cord clamping. The findings cannot establish the optimal starting oxygen concentration during intact-cord stabilisation.
    • No treatment interaction was demonstrated by gestational-age band or hospital capability, but the subgroup analyses do not establish identical effects across all infants or centres.

Conclusion on External Validity: External validity is good for similarly resourced maternity services caring for infants born at 32–35+6 weeks who need immediate respiratory support. Extrapolation to other gestational ages, intact-cord resuscitation or settings without dependable oxygen titration remains limited.

Strengths & Limitations

  • Strengths:
    • A large randomised evaluation directed specifically at an important and previously understudied gestational-age group.
    • A pragmatic intervention embedded across a broad range of maternity hospitals, with each hospital contributing to both treatment conditions.
    • Waived consent, consumer involvement, complete primary outcome data and very little missing hospital outcome information.
    • Publicly available protocol and SAP, appropriate recognition of clustering and multiple births, and sensitivity analyses supporting the primary conclusion.
    • Academic sponsorship and public or institutional funding; funding bodies had no role in trial conduct, analysis or publication decisions.
  • Limitations:
    • A brief and modest oxygen contrast, incomplete adherence, and frequent subsequent escalation in both groups.
    • Unblinded eligibility and treatment decisions, with a primary endpoint strongly influenced by local support and departure practices.
    • Only 26 recruiting randomised clusters, substantial intracluster correlation and variation in treatment effects and recruitment periods.
    • Multiple secondary comparisons, rare deaths and insufficient information to assess long-term benefit or harm.
    • No direct measurement of the physiological exposure that might explain benefit, harm or the absence of a primary effect.

Interpretation & Why It Matters

  • Clinical effect
    • Replacing an air-first policy with a 30%-oxygen-first policy did not materially change the observed proportion leaving the delivery room on respiratory support.
    • The primary confidence interval makes the anticipated eight-percentage-point benefit unlikely under the tested conditions, while leaving smaller effects unresolved.
  • Potential benefit
    • Avoiding endotracheal ventilation would be clinically meaningful, and the 6.4% versus 9.3% finding warrants further evaluation.
    • That finding does not establish a general reduction in respiratory morbidity: any post-delivery-room support or oxygen, hospital stay and transfer were not demonstrably improved.
  • Practice implication
    • AIROPLANE provides direct evidence to inform the choice between two commonly used starting concentrations, but does not establish 30% as the uniquely preferred concentration.
    • Starting concentration remains one component of stabilisation; effective ventilation, early monitoring and responsive oxygen titration remain central to the care strategy actually tested.
    • The trial does not justify maintaining either concentration despite an inadequate clinical response.

Controversies & Other Evidence

  • What does the primary endpoint establish?
    • The trial estimates the effect of assigning an initial oxygen policy within routine care, including non-adherence and subsequent rescue. It does not isolate the biological effect of sustained exposure to 30% versus 21% oxygen.
    • Respiratory support at departure is operationally important but is not a direct measure of lung injury, oxygenation or recovery. Its predominance by CPAP/high flow may make it less responsive to an intervention that chiefly alters the intensity of early resuscitation.
    • The accompanying editorial emphasises the clinician-dependent nature of these outcomes and the influence of time spent in the delivery room, supporting the need for physiological endpoints in future trials.6
  • How much weight should the favourable secondary findings carry?
    • Both the trial discussion and editorial regard starting at 30% as reasonable, drawing on lower support intensity and fewer subsequent intubations.6
    • That is a defensible clinical option, but a claim of established superiority would exceed the evidence: the primary hypothesis was not supported and the secondary family had no multiplicity control.
    • The two favourable outcomes are correlated clinical decisions, rather than independent replications of the treatment effect.
    • The ordinal OR of 0.70 describes a common shift across support thresholds. It must not be presented as a demonstrated 30% reduction in intubation, compressions or adrenaline individually.
    • Antenatal corticosteroid and cord-clamping imbalances could contribute to secondary differences. The robust adjusted primary analysis does not automatically resolve this uncertainty for each secondary outcome.
  • Does frequent oxygen escalation show that starting in air is inadequate?
    • Escalation above the starting concentration in 82.7% of the air group shows that air was often insufficient later in stabilisation under the treatment algorithms used.
    • It does not establish that commencing with air caused harmful hypoxaemia or that every infant would benefit from higher oxygen immediately after birth.
    • The absence of oxygen saturation and heart-rate trajectories prevents assessment of whether 30% shortened hypoxaemia, increased hyperoxia, or changed the time needed to achieve adequate ventilation.
    • The editorial identifies this missing physiological information and the narrow concentration contrast as major priorities for further research.6
  • Precision, clustering and safety:
    • Recruitment exceeded the minimum sample size, but the event rate and intracluster correlation differed substantially from the planning assumptions. Precision should therefore be judged from the reported confidence intervals rather than from the original power statement alone.
    • The analysis allowing treatment effects to vary between hospitals remained compatible with no primary benefit and still excluded the planned eight-percentage-point reduction.
    • Nine deaths, several associated with major antenatal or congenital pathology, provide very limited evidence about an oxygen-mediated mortality effect.
    • The protocol did not require conventional comprehensive adverse-event reporting for this comparison of existing practices. Mortality oversight cannot substitute for systematic assessment of uncommon morbidity or neurodevelopment.35
  • Reporting points requiring clarification:
    • For surfactant, the published risk-ratio interval excludes one, whereas the risk-difference interval includes zero; the text identifies only two statistically significant secondary outcomes. These scale-dependent inferences should be reconciled before surfactant is promoted as a third benefit.
    • The hospital-duration mean difference is printed as −0.13 days with a 95% CI of −1.65 to 1.91, which is not centred on that estimate despite the stated linear-model analysis. The values should be clarified, not silently corrected.
    • Statistical analysis is dated 29 October 2025, while final follow-up is stated as March 2026. The sequence of initial analysis, subsequent outcome updates and final data lock is not clear from these dates.
    • These issues warrant clarification but do not, by themselves, overturn the consistently null primary comparison.
  • Related observational evidence:
    • In the 2024 prospective cohort, 58 of 76 infants began support in air; among 43 with five-minute saturation measurements, 18 (42%) remained below 80%. This documents a physiological concern, but incomplete measurements and clinician-selected oxygen treatment prevent a causal comparison of starting concentrations.2
  • Meta-analysis in more immature infants:
    • The 2024 individual participant data network meta-analysis included 1,055 infants from 12 trials below 32 weeks’ gestation. High initial oxygen, at least 90%, was associated with lower mortality than low oxygen, at most 30%: OR 0.45; 95% credible interval 0.23 to 0.86; low-certainty evidence.7
    • This is indirect evidence for AIROPLANE: the gestational ages differ, and both AIROPLANE concentrations fall within that review’s low-oxygen category. It cannot establish that 30% is superior to air at 32–35 weeks.
  • TORPIDO 30/60:
    • TORPIDO 30/60 compared initial FiO₂ 0.60 with 0.30 in infants born at 23–28 weeks. Death or brain injury occurred in 330/703 infants (46.9%) versus 344/720 (47.8%): RR 0.98; 95% CI 0.89 to 1.09.8
    • Its combination of improved early resuscitation measures without improvement in the primary clinical endpoint reinforces the distinction between physiological transition and downstream benefit.
    • Its markedly more immature population and higher oxygen comparison prevent direct extrapolation to AIROPLANE.
  • Guidelines and the gestational-age boundary:
    • The 2025 AHA/AAP guidelines allow initial oxygen of 21–30% for infants at 32–34+6 weeks and recommend starting at 21% from 35 weeks; oxygen is subsequently titrated to saturation targets.9
    • The 2025 ERC guidelines recommend starting at 21% from 32 weeks and at least 30% below 32 weeks.10
    • These guidelines predate AIROPLANE; no guideline incorporating its results was identified by 10 September 2026.
    • AIROPLANE contributes directly to the uncertainty underlying these differing recommendations, but its 34–35+6-week subgroup does not isolate infants at 35 weeks. It therefore cannot establish a separate treatment effect at that guideline threshold.

Summary

  • AIROPLANE compared initial oxygen concentrations of 30% and 21% in 1,818 infants born at 32–35+6 weeks across 26 Australian hospitals using a cluster-randomised crossover design.
  • Ongoing respiratory support at delivery room departure was similar: 72.6% versus 73.3%; risk difference −0.83 percentage points; 95% CI −4.33 to 2.67; P=0.64.
  • Lower delivery room support intensity and fewer subsequent endotracheal ventilations favoured 30% oxygen, but remain exploratory findings without multiplicity adjustment.
  • Randomisation, broad recruitment and complete primary outcome capture support the main result; unblinded decisions, incomplete adherence and absent physiological exposure data limit mechanistic interpretation.
  • The trial supports informed choice between starting strategies with prompt titration, while leaving superiority, the optimal concentration and long-term comparative safety unresolved.

Overall Takeaway

AIROPLANE provides substantial direct randomised evidence that starting respiratory support with 30% rather than 21% oxygen does not achieve the anticipated reduction in ongoing support at delivery room departure for infants born at 32–35+6 weeks. Its favourable secondary findings make 30% a reasonable option for further evaluation, but do not establish superiority or long-term safety; clinical interpretation must retain the trial’s subsequent monitoring, titration and rescue treatment.

Overall Summary

  • Primary outcome: No demonstrated reduction in respiratory support at delivery room departure with initial 30% oxygen.
  • Potential benefit: Less intensive early support and fewer later intubations, requiring confirmation.
  • Clinical implication: Starting concentration should remain part of a monitored, responsive resuscitation strategy; AIROPLANE does not identify a universally superior initial concentration.

Bibliography

Added September 10th, 2026