Publication

• Title: Oxygen saturation thresholds in children with acute respiratory distress (OxyKids): a multicentre, open, parallel-group, randomised clinical trial
• Acronym: OxyKids
• Year: 2026
• Journal published in: The Lancet Respiratory Medicine
• Citation: Louman S, van Stralen KJ, Koppelman GH, Vaessen-Verberne AAPH, Bekhof J, Bosmans JE, et al. Oxygen saturation thresholds in children with acute respiratory distress (OxyKids): a multicentre, open, parallel-group, randomised clinical trial. Lancet Respir Med. 2026 May 12:S2213-2600(26)00087-1.

Context & Rationale

• Background

  • Acute respiratory distress due to bronchiolitis, lower respiratory tract infection and viral-induced wheeze is among the commonest reasons for paediatric hospital admission.
  • Oxygen therapy is central supportive care when hypoxaemia is present, but the threshold for starting and stopping oxygen in children has historically been based more on consensus and local practice than on high-certainty trial evidence.
  • Guideline and hospital thresholds have commonly ranged from SpO₂ 90% to 94%, creating practice variation and a risk that children receive oxygen for too long.
  • The 2023 systematic review by the OxyKids investigators concluded that lower SpO₂ thresholds, including thresholds as low as 88%, appeared potentially safe in otherwise healthy children with acute respiratory distress and might reduce hospitalisation, but that higher-quality evidence was needed in a broader ward population. ¹
  • Before OxyKids, the best relevant evidence came from narrower populations: infants with bronchiolitis in BIDS, children with pneumonia in the COAST setting, and mechanically ventilated critically ill children in Oxy-PICU. ²³⁴
• Research Question/Hypothesis

  • The trial tested whether managing otherwise healthy children admitted to general paediatric wards with an SpO₂ threshold of 88%, rather than 92%, would safely shorten time to meeting discharge criteria.
  • The investigators prespecified a 12-hour reduction in time to discharge readiness as clinically meaningful.
• Why This Matters

  • Oxygen is a treatment, but in ward-level paediatric respiratory illness it is also a major determinant of monitoring intensity, bed occupancy, parental anxiety, nursing workload and discharge timing.
  • A safe lower threshold could reduce unnecessary oxygen administration, shorten hospital stays, lessen child and family disruption, and relieve winter pressure on paediatric beds.
  • The trial is important because it moved beyond single-disease bronchiolitis evidence and tested a pragmatic oxygen strategy across a broad age range and common acute respiratory diagnoses.

Design & Methods

• Research Question: In children aged 6 weeks to 12 years admitted to general paediatric wards with acute respiratory distress requiring oxygen under usual 92% threshold care, does an 88% SpO₂ threshold safely reduce time from admission to meeting discharge criteria compared with a 92% threshold?
• Study Type: Pragmatic, multicentre, open-label, parallel-group, randomised controlled clinical trial conducted in ten general and teaching hospitals in the Netherlands; all sites had general paediatric wards and no paediatric intensive care unit.
• Population:
  • Setting: General paediatric wards; children requiring PICU were transferred to academic hospitals but retained in the intention-to-treat analysis.
  • Age: 6 weeks to 12 years.
  • Clinical syndromes: Bronchiolitis, lower respiratory tract infection, or acute viral-induced wheeze, diagnosed by treating clinicians using usual-care criteria.
  • Oxygen requirement: Required supplementary oxygen according to usual care at the 92% SpO₂ threshold; children already started on oxygen before ward admission could be enrolled if oxygen therapy had begun less than 6 hours before randomisation.
  • Major exclusions: Known pre-existing cardiopulmonary, immunological, neurological or haematological conditions; birth before 32 weeks’ gestation; concurrent interfering trial participation; no stable internet access; insufficient Dutch or English; and children aged 6–12 years with acute asthma.
  • Randomised population: 566 children were randomised; 557 were included in the intention-to-treat analysis.
• Intervention:
  • Children allocated to the intervention arm were managed using an SpO₂ threshold of 88% for starting, restarting and stopping supplementary oxygen.
  • Oxygen therapy was mandated when SpO₂ fell below the allocated threshold.
  • The published protocol specified oxygen initiation if SpO₂ dropped 1–2 percentage points below the threshold for at least 15 minutes continuously, or more than 2 percentage points below the threshold for any duration. ⁵
  • Oxygen could be delivered by low-flow nasal cannula, high-flow nasal cannula, or face mask using equipment and local ward protocols.
  • Clinicians could initiate oxygen above the allocated threshold for clinical symptoms such as tachypnoea, tachycardia or increased work of breathing, reflecting Dutch standard practice.
  • No upper SpO₂ target was specified after oxygen was started.
• Comparison:
  • Children allocated to the control arm were managed using an SpO₂ threshold of 92% for starting, restarting and stopping supplementary oxygen.
  • Oxygen delivery devices, weaning procedures, rescue escalation, bronchodilators, corticosteroids, feeding and other care were managed according to local protocols.
  • Children transferred to PICU received standard institutional PICU care and remained analysed in their original randomised group.
• Blinding: The trial was unblinded; participants, clinicians, nurses and investigators knew allocation. Masking was not feasible because manufacturers were unwilling to alter device algorithms under revised EU Medical Device Regulation and liability constraints.
• Statistics: A total of 560 children were required to detect a 12-hour difference in time to meeting all discharge criteria, assuming a mean length of stay of 72 hours, SD 48 hours, 80% power, a two-sided alpha of 0.05, and 10% loss to follow-up. The primary analysis was intention-to-treat using a log-transformed time outcome in a multilevel linear regression model with a random intercept for hospital; adjusted analyses included age, sex, diagnosis, baseline vitals and whether oxygen was started before randomisation.
• Follow-Up Period: The published clinical safety follow-up was to 28 days after discharge for serious adverse events, unscheduled health-care use, readmissions and parental anxiety; the protocol included 90-day quality-of-life and cost-effectiveness follow-up, with health-economic results to be reported separately. ⁵

Key Results

This trial was not stopped early. One interim analysis at 25% recruitment assessed safety outcomes and length of stay but not the primary outcome, and there were no predefined stopping guidelines.

• The primary outcome favoured the 88% threshold across prespecified age and diagnostic subgroups: age 6 weeks to <1 year GMR 0.53; 95% CI 0.41 to 0.68; P<0.0001; age 1 to <4 years GMR 0.71; 95% CI 0.58 to 0.87; P=0.0011; age 4 to ≤12 years GMR 0.69; 95% CI 0.47 to 1.01; P=0.054.
• Diagnostic subgroup effects also favoured 88%: bronchiolitis GMR 0.62; 95% CI 0.49 to 0.78; P<0.0001; viral wheeze GMR 0.60; 95% CI 0.46 to 0.79; P=0.0002; lower respiratory tract infection GMR 0.73; 95% CI 0.56 to 0.94; P=0.016.
• The main clinical signal was reduced treatment burden and earlier discharge readiness, with no detected difference in serious short-term harms, but a numerical excess of GP visits in the 88% group.

Internal Validity

• Randomisation and Allocation: Randomisation was 1:1, computer-generated, stratified by centre and age group, using permuted blocks of sizes 2 to 8 and opaque sealed envelopes drawn by trained paediatric nursing staff. Allocation concealment was therefore reasonable, although envelope-based systems are inherently less robust than fully centralised randomisation.
• Drop out or exclusions: Of 566 randomised children, 557 were included in the intention-to-treat analysis: 278/282 in the 88% group and 279/284 in the 92% group. Nine children withdrew before the primary endpoint: 4 in the 88% group and 5 in the 92% group. The primary outcome was missing in 13 children, and covariate data were missing in 35; these were handled by prespecified imputation approaches.
• Pre-randomisation selection: 1602 children met inclusion criteria, but 1036 were not randomised: 577 were not approached, 427 declined and 32 had other reasons. This is more important for external validity than internal validity, because randomisation occurred after this selection.
• Performance/Detection Bias: The open-label design created substantial potential for clinician behaviour to differ by group, especially for oxygen initiation, weaning and discharge readiness. The primary outcome was structured and protocolised, but it incorporated the allocated saturation threshold, making shorter time to meeting discharge criteria a direct consequence of the intervention rather than a fully independent biological endpoint.
• Protocol Adherence: Protocol deviations were uncommon: 4 in the 88% group and 8 in the 92% group. Deviations included temporarily accepting a lower saturation, temporary transfer to another hospital with a different threshold and one switch to control on parental request.
• Baseline Characteristics: Groups were well balanced. Male sex was 161/278 (58%) vs 156/279 (56%); median age was 1 year in both groups; bronchiolitis was 105/278 (38%) vs 107/279 (38%); viral wheeze 90/278 (32%) vs 76/279 (27%); lower respiratory tract infection 83/278 (30%) vs 96/279 (34%); SpO₂ at randomisation was mean 89.4% (SD 1.9) vs 89.0% (SD 1.8).
• Heterogeneity: Clinical heterogeneity was deliberate and pragmatic rather than a major internal validity threat. Prespecified subgroup analyses were directionally consistent, although the 4–12 year subgroup was less precise and did not reach conventional statistical significance for the primary outcome in the main analysis.
• Timing: Enrolment was constrained to less than 6 hours after oxygen initiation if oxygen had already begun. However, only 114/278 (41%) in the 88% group and 127/279 (46%) in the 92% group had oxygen started after randomisation, meaning that for many children the trial primarily tested oxygen weaning and restarting thresholds rather than first initiation in the acute presentation phase.
• Dose: The “dose” was the oxygen saturation threshold itself. The 88% threshold achieved clinically meaningful separation in oxygen exposure, but no upper SpO₂ target was imposed once oxygen was started, so the trial tested a lower threshold strategy rather than tightly conservative oxygen titration.
• Separation of the Variable of Interest: Separation was strong for treatment burden: no oxygen use 68/278 (24%) vs 12/279 (4%); oxygen duration 24.1 h vs 35.7 h; oxygen restarts 0.4 vs 0.7 per patient; post-randomisation oxygen initiations 123 vs 296. Physiological separation in measured SpO₂ at key timepoints was modest: final oxygen stop 95% vs 96%; discharge criteria 95% vs 96%.
• Key Delivery Aspects: The intervention was deliverable in ordinary wards using existing devices and local protocols, which strengthens pragmatic validity. However, allowing oxygen for symptoms despite SpO₂ above threshold introduced a clinician-dependent co-intervention within both groups.
• Crossover: True crossover was minimal. One child in the 88% group switched to control on parental request; temporary threshold deviations occurred but were few and unlikely to explain the large effect.
• Adjunctive therapy use: Bronchodilators, corticosteroids, feeding support and other therapies followed local protocols. Detailed intergroup numeric balance for all adjunctive therapies was not reported in the main manuscript, so residual performance variation cannot be fully excluded.
• Outcome Assessment: The primary outcome used predefined discharge criteria rather than actual discharge alone, reducing bias from social or logistical discharge delays. The trade-off is that one discharge criterion—stable SpO₂ above the allocated threshold in room air for at least 4 hours—was intrinsically determined by treatment allocation.
• Statistical Rigor: The analysis followed a prespecified SAP, used intention-to-treat principles, adjusted for important covariates, accounted for hospital clustering and showed concordant sensitivity analyses: complete-case GMR 0.64; 95% CI 0.55 to 0.74; excluding PICU transfers GMR 0.63; 95% CI 0.55 to 0.73; Cox HR 1.44; 95% CI 1.28 to 1.61. Secondary outcomes and subgroups were not multiplicity-adjusted, and the trial was not powered to exclude rare harms or long-term neurocognitive effects.

Conclusion on Internal Validity: Internal validity is strong for the pragmatic question of whether adopting an 88% ward oxygen threshold reduces oxygen use and discharge-readiness time. It is more limited for mechanistic inference, rare safety outcomes and fully blinded assessment, because the primary endpoint was partly threshold-defined and clinicians were unblinded.

External Validity

• Population Representativeness: The trial population is highly relevant to otherwise healthy children admitted to general paediatric wards with bronchiolitis, viral wheeze or lower respiratory tract infection. Mean age was 2.0 years (SD 2.4), and the cohort included children across infancy, preschool and school age.
• Important Exclusions: Findings should not be extrapolated uncritically to children with chronic cardiopulmonary disease, immunological disease, neurological disease, haematological disease, prematurity before 32 weeks’ gestation, older children with acute asthma, or children managed primarily in PICU.
• Health-System Applicability: The results are most applicable to high-income general paediatric ward settings with trained nursing staff, pulse oximetry, access to escalation pathways and capacity for post-discharge review. Applicability to resource-limited systems is plausible but not directly established by OxyKids.
• Skin Tone and Measurement Equity: The enrolled cohort was predominantly Dutch and light-skinned: 437/480 (91%) with ethnicity data were Dutch, and dark Fitzpatrick skin type was reported in only 11/228 (5%) vs 9/238 (4%). This matters because paediatric data show pulse oximetry can overestimate arterial oxygen saturation in Black children compared with White children, which could alter the safety margin of a lower SpO₂ threshold. ⁶
• Applicability to Emergency Department Decisions: The trial enrolled hospitalised children requiring oxygen under usual care and primarily informs ward oxygen management and discharge readiness. It does not directly define emergency department discharge thresholds for children who never require admission.
• Applicability to Asthma: Children aged 6–12 years with acute asthma were excluded, so the findings should not be applied to acute asthma attacks in older children without dedicated trial evidence.

Conclusion on External Validity: External validity is good for otherwise healthy children with common acute respiratory illnesses on general paediatric wards in similar health-care systems. Generalisability is limited for chronic disease, darker skin tones, older acute asthma, low-resource settings and PICU-level illness.

Strengths & Limitations

• Strengths:
  • Pragmatic design embedded in ordinary general paediatric ward practice.
  • Large sample size for the clinical ward question, with 566 randomised and 557 analysed.
  • Multicentre participation across ten Dutch general and teaching hospitals.
  • Broad diagnostic inclusion across bronchiolitis, viral wheeze and lower respiratory tract infection.
  • Prespecified, clinically meaningful 12-hour effect threshold.
  • Structured discharge-readiness endpoint designed to reduce non-medical discharge-delay bias.
  • Low post-randomisation withdrawal and low protocol-deviation rates.
  • Consistent results across sensitivity analyses and most prespecified subgroups.
  • Patient and parent representatives contributed to outcome selection and trial design.
• Limitations:
  • Open-label design with plausible performance and detection bias.
  • The primary endpoint included the allocated SpO₂ threshold, so part of the treatment effect was structurally built into the endpoint.
  • Many children had already started oxygen before randomisation, limiting inference about the earliest initiation decision.
  • Clinicians could start oxygen for respiratory symptoms above the saturation threshold, adding subjectivity and potential group-dependent behaviour.
  • Short-term safety follow-up only; long-term neurocognitive outcomes were not assessed.
  • The trial was not powered to exclude rare adverse outcomes such as unexpected severe deterioration.
  • Post-discharge GP visits were numerically higher in the 88% group, and the complete-case analysis reached statistical significance.
  • Limited representation of children with darker skin tones.
  • Important exclusions constrain application to chronic disease, older acute asthma and PICU populations.
  • Cost-effectiveness results were not included in the index publication.

Interpretation & Why It Matters

• Practice signal

  For otherwise healthy children admitted to general paediatric wards with bronchiolitis, viral wheeze or lower respiratory tract infection, an 88% SpO₂ threshold substantially reduced time to discharge readiness, hospital stay, oxygen exposure and oxygen restarts.
• Deimplementation

  The trial supports a “less oxygen when not hypoxaemic” strategy and challenges the entrenched habit of treating modest desaturation or work of breathing with oxygen when tissue oxygen delivery is unlikely to be impaired.
• Safety interpretation

  Short-term safety outcomes were reassuring: serious adverse events were 19/278 (7%) vs 19/279 (7%), PICU admissions were 5/278 (2%) vs 5/279 (2%), and no deaths occurred. These data do not prove equivalence for rare harms or long-term neurodevelopment.
• What changes

  OxyKids extends lower-threshold oxygen evidence beyond infant bronchiolitis into a broader ward population and provides a practical basis for revising local oxygen protocols in suitable children.
• What does not change

  The trial should not be used as a universal licence to accept lower saturations in chronic cardiopulmonary disease, older acute asthma, darker skin tones without awareness of oximetry bias, or children who are clinically deteriorating despite apparently acceptable SpO₂.

Controversies & Other Evidence

• The accompanying editorial emphasised that the direction of the primary result was expected because the discharge-readiness criteria required room-air SpO₂ above the allocated threshold; this does not invalidate the pragmatic result, but it explains why the effect size is large. ⁷
• The lower-threshold group had more residual physiological abnormality when considered discharge-ready: heart rate above the 90th percentile 13% vs 6%, and mild work of breathing 38% vs 25%. This supports the interpretation that children were being discharged earlier in the natural history of the same illness, rather than that oxygen itself accelerated disease resolution.
• Oxygen initiation for symptoms despite SpO₂ ≥92% occurred in 15% of post-randomisation oxygen starts in the 88% group and 6% in the 92% group. This likely reflects clinician discomfort with lower thresholds and may have diluted the apparent benefit of the intervention.
• The editorial’s most important practice critique is that oxygen for increased work of breathing without hypoxaemia has weak physiological justification; OxyKids preserved this behaviour for pragmatic reasons, but the results strengthen the case for abandoning oxygen as a treatment for breathlessness alone. ⁷
• The broader bronchiolitis oximetry literature shows how saturation displays can drive hospitalisation decisions: in Schuh et al, infants whose oximetry display was artificially increased by 3 percentage points were less likely to be hospitalised, without a significant safety signal. ⁸
• COAST provides important low-resource evidence around even lower thresholds, but it was conducted in African children with pneumonia, used a different setting and oxygen-infrastructure context, and faced recruitment challenges related to strong prior beliefs about oxygen benefit. ³⁹
• Oxy-PICU supports conservative oxygenation in a very different population: critically ill mechanically ventilated children, where an 88–92% target reduced the composite of organ support duration and death compared with SpO₂ ≥94%. ⁴
• The 2026 Surviving Sepsis Campaign paediatric guideline conditionally suggests titrating supplementary oxygen after resuscitation to SpO₂ 88–92% rather than >94% in children with sepsis or septic shock, reflecting the broader move toward conservative oxygenation in paediatrics; this guideline predated OxyKids and therefore did not incorporate its results. ¹⁰
• The key unresolved safety questions are rare deterioration, neurocognitive effects of brief lower saturations, implementation in darker skin tones, and the correct threshold for acute asthma in older children.

Summary

• OxyKids randomised 566 otherwise healthy children aged 6 weeks to 12 years with bronchiolitis, viral wheeze or lower respiratory tract infection to oxygen thresholds of 88% vs 92%.
• The 88% threshold reduced time to meeting discharge criteria from 46.6 h to 27.6 h; adjusted GMR 0.64; 95% CI 0.55 to 0.74; P<0.0001.
• Hospital stay was shorter by an adjusted 17.6 h, oxygen duration by 12.9 h, and many more children avoided oxygen altogether: 24% vs 4%.
• Short-term harms were not increased: serious adverse events were 7% in both groups, PICU admissions were 2% in both groups, and there were no deaths.
• The main caveats are open-label conduct, a threshold-dependent primary endpoint, limited darker-skin representation, exclusion of chronic disease and older acute asthma, and no long-term neurocognitive follow-up.

Overall Takeaway

OxyKids is a practice-shaping paediatric ward trial showing that an 88% SpO₂ threshold can materially reduce oxygen exposure and hospital stay in otherwise healthy children with acute respiratory illness, without detected short-term harm. Its strongest message is pragmatic rather than mechanistic: many children are kept in hospital because clinicians are treating a number on a pulse oximeter rather than clinically important hypoxaemia.

Bibliography

• 1.Louman S, van Stralen KJ, Pijnenburg MWH, Koppelman GH, Boehmer ALM. Oxygen saturation targets for children with respiratory distress: a systematic review. ERJ Open Res. 2023;9:00256-2023.
• 2.Cunningham S, Rodriguez A, Adams T, Boyd KA, Butcher I, Enderby B, et al. Oxygen saturation targets in infants with bronchiolitis (BIDS): a double-blind, randomised, equivalence trial. Lancet. 2015;386:1041-1048.
• 3.Maitland K, Kiguli S, Olupot-Olupot P, Hamaluba M, Thomas K, Alaroker F, et al. Randomised controlled trial of oxygen therapy and high-flow nasal therapy in African children with pneumonia. Intensive Care Med. 2021;47:566-576.
• 4.Peters MJ, Gould DW, Ray S, Thomas K, Chang I, Orzol M, et al. Conservative versus liberal oxygenation targets in critically ill children (Oxy-PICU): a UK multicentre, open, parallel-group, randomised clinical trial. Lancet. 2024;403:355-364.
• 5.Louman S, van Stralen KJ, Koppelman G, Vaessen-Verberne A, Bekhof J, Bosmans J, et al. Safe, effective and cost-effective oxygen saturation targets for children and adolescents with respiratory distress: protocol for a randomised controlled trial (OxyKids study). BMJ Open. 2024;14:e087891.
• 6.Ruppel H, Makeneni S, Faerber JA, Lane-Fall MB, Foglia EE, O’Byrne ML, et al. Evaluating the accuracy of pulse oximetry in children according to race. JAMA Pediatr. 2023;177:540-543.
• 7.Mitting R, Peters MJ. Challenging dogmata of oxygen therapy in acute childhood illness. Lancet Respir Med. 2026 May 12:S2213-2600(26)00123-2.
• 8.Schuh S, Freedman S, Coates A, Allen U, Parkin PC, Stephens D, et al. Effect of oximetry on hospitalization in bronchiolitis: a randomized clinical trial. JAMA. 2014;312:712-718.
• 9.Peters MJ, Macharia W, Molyneux E. A COASTal view: where prior beliefs and uncertainty collide. Intensive Care Med. 2021;47:591-593.
• 10.Weiss SL, Peters MJ, Oczkowski SJW, Belley-Cote E, Buysse C, Choong KLM, et al. Surviving Sepsis Campaign international guidelines for the management of sepsis and septic shock in children 2026. Pediatr Crit Care Med. 2026;27:379-434.