Publication

• Title: Early Restrictive vs Liberal Oxygen for Trauma Patients: The TRAUMOX2 Randomized Clinical Trial
• Acronym: TRAUMOX2
• Year: 2025
• Journal published in: JAMA
• Citation: Arleth T, Baekgaard J, Siersma V, et al. Early restrictive vs liberal oxygen for trauma patients: the TRAUMOX2 randomized clinical trial. JAMA. 2025;333(6):479-489

Context & Rationale

• Background

  Early supplemental oxygen is commonly recommended in severe trauma, yet guidance is typically non-specific on dose, duration, and targets, making hyperoxaemia frequent in early trauma care.
  Observational data in trauma and broader critical illness have linked hyperoxaemia to increased mortality and respiratory complications, while randomised evidence in comparable hyperacute trauma settings has been sparse and mixed.
  Systematic reviews of acutely ill adults suggested potential harm with liberal oxygen strategies, supporting equipoise for a targeted approach in trauma, particularly early after injury.¹
• Research Question/Hypothesis

  Whether an early (8-hour) restrictive oxygen strategy (SpO2 target 94%) initiated prehospital or on trauma-centre arrival reduces death and/or major respiratory complications within 30 days compared with a liberal oxygen strategy.
• Why This Matters

  Oxygen is ubiquitous, low-cost, and rapidly administered in trauma; even small absolute effects could have major population impact.
  Clarifying early oxygen targets could standardise practice, reduce unnecessary hyperoxaemia, and inform prehospital and trauma-bay protocols.

Design & Methods

• Research Question: In adult trauma patients with full trauma team activation and anticipated admission ≥24 hours, does an 8-hour restrictive oxygen strategy (SpO2 target 94%) reduce 30-day death and/or major respiratory complications versus a liberal oxygen strategy?
• Study Type: Investigator-initiated, pragmatic, international, multicentre, open-label, parallel-group, superiority RCT with blinded primary outcome assessment; enrolment prehospital or on trauma-centre admission at 15 prehospital bases and 5 major trauma centres (Denmark, the Netherlands, Switzerland).
• Population:
  • Adults ≥18 years with blunt or penetrating trauma, transported directly to a participating trauma centre, triggering full trauma team activation, and clinician anticipated hospital stay ≥24 hours.
  • Key exclusions: suspicion of carbon monoxide intoxication; cardiac arrest prior to randomisation; and post-randomisation “secondary exclusions” where secondary survey revealed no/minor injuries with anticipated discharge <24 hours (pre-specified; removed from modified ITT).²
• Intervention:
  • Restrictive oxygen strategy for 8 hours: lowest oxygen dose (≥21%) needed to maintain SpO2 94%, using no supplemental oxygen, nasal cannula, non-rebreather mask, or mechanical ventilation (if intubated).²
• Comparison:
  • Liberal oxygen strategy for 8 hours: non-intubated patients received 12–15 L/min oxygen by non-rebreather mask; intubated patients received FiO2 0.6–1.0 (FiO2 1.0 prehospital/trauma bay/intrahospital transport, with down-titration permitted if SpO2 ≥98% in downstream locations per protocol).²
• Blinding: Open-label to clinicians/patients; primary outcome (death and major respiratory complications) assessed by specialists blinded to allocation via redaction of allocation-indicative oxygen data in records.
• Statistics: Power: 1420 evaluable patients (≈710/group) with assumed 3.5% dropout to detect a 33% relative risk reduction in the composite outcome (assumed 15% to 10%), 80% power, α=0.05; primary analysis modified intention-to-treat (excluding pre-specified secondary exclusions), with per-protocol analysis excluding major protocol violations; primary and key secondary outcomes analysed with logistic regression reporting odds ratios adjusted for stratification variables (site and intubation at randomisation).³
• Follow-Up Period: 30 days for primary and key secondary outcomes (longer-term follow-up planned separately).

Key Results

This trial was not stopped early. Two interim analyses were performed by an independent monitoring committee; enrolment continued to completion as planned.

• Despite substantial oxygenation separation (PaO2 85 vs 280 mm Hg at ~1 hour), the restrictive strategy did not reduce the composite of death and/or major respiratory complications (16.1% vs 16.7%; OR 1.01; 95% CI 0.75 to 1.37; P=0.94).
• Component outcomes moved in opposite numeric directions (death 8.6% vs 7.3%; major respiratory complications 8.9% vs 10.8%), but neither differed significantly.
• Atelectasis was less common with restrictive oxygen (27.6% vs 34.7%), while hypoxaemic episodes were more frequent (6.0% vs 3.8%).

Internal Validity

• Randomisation and allocation: 1:1 allocation with variable block sizes and stratification by inclusion site and intubation at randomisation; sealed envelopes enabled prehospital enrolment but creates theoretical risk of allocation subversion (mitigated by external generation and concealment).³
• Post-randomisation exclusions: 1979 randomised; 1508 “completed the trial”; primary outcome data available for 1457 (733 restrictive; 724 liberal). Secondary exclusions for minor/no injuries were substantial (165 restrictive; 174 liberal) plus additional post-randomisation exclusions (59 restrictive; 71 liberal) as per CONSORT diagram; this required a modified ITT approach and may introduce selection bias if exclusions relate to prognosis, even if pre-specified.
• Blinding and detection bias: Open-label oxygen delivery could influence co-interventions; primary outcome assessment was blinded and assessors guessed allocation correctly ~50% (50.6% vs 51.0%), supporting effective blinding for adjudicated outcomes.
• Protocol adherence and separation: Clear physiological separation (PaO2 median 85 vs 280 mm Hg at 1h; 86 vs 230 mm Hg at 6h). Major protocol violations occurred in 6.7% (restrictive) vs 13.7% (liberal), indicating more frequent deviation from intended “liberal” exposure in routine care settings.⁴
• Timing: Median time from injury to trauma-centre arrival ~58 vs 55 minutes; median time from trauma to randomisation 54 vs 52 minutes (where available), but one-quarter had >50 minutes of oxygen exposure before randomisation (potentially diluting early exposure contrast).
• Dose: Intervention duration was 8 hours (chosen to reflect typical early liberal oxygen exposure plateau); whether this window is sufficient to influence 30-day clinical outcomes remains uncertain in the absence of demonstrated effect.
• Missing data / attrition: Differential dropout handled with inverse probability weighting in planned analyses; best/worst-case scenarios did not meaningfully change the primary conclusion (per supplement).

Conclusion on Internal Validity: Overall, internal validity appears moderate: randomisation and blinded adjudication were robust with clear exposure separation, but substantial post-randomisation secondary exclusions and open-label co-intervention risk meaningfully temper confidence in a fully unbiased estimate.

External Validity

• Population representativeness: Broad adult trauma cohort (blunt ~89%, penetrating ~11%), median ISS 14 (IQR 9–22) with ~50% ICU admission, consistent with moderately-to-severely injured patients seen in major trauma systems.
• Setting applicability: Conducted across prehospital physician-staffed services and major European trauma centres; results most generalisable to systems with similar trauma team activation, prehospital critical care capability, and ready access to ABGs/monitoring.
• Key exclusions: Cardiac arrest and suspected CO poisoning excluded; findings should not be extrapolated to these subgroups or to patients in whom oxygen targets are driven by specific toxicology/ROSC physiology.
• Pragmatic implementation: Oxygen titration based on SpO2 (rather than PaO2) aligns with real-world feasibility in prehospital and trauma-bay care, supporting pragmatic generalisability, but pulse oximetry limitations (e.g., skin pigmentation effects) were not directly assessed.

Conclusion on External Validity: Generalisability is good for adult major trauma care pathways in high-resource systems (prehospital + trauma centre) where SpO2-guided oxygen is standard, but applicability is less certain in lower-resource settings or markedly different prehospital models.

Strengths & Limitations

• Strengths: Large pragmatic multicentre international trial; early initiation prehospital/arrival; blinded adjudication of clinically important respiratory outcomes; clear physiological exposure separation; prespecified SAP and protocol.
• Limitations: Open-label treatment with potential co-intervention effects; substantial post-randomisation “secondary exclusions” for minor/no injuries (modified ITT); heterogeneous trauma phenotypes with limited power for injury-specific effects; composite outcome components trended in opposite directions without significance.

Interpretation & Why It Matters

• Practice implication

  A routine liberal oxygen approach (high-flow/FiO2 0.6–1.0) for the first 8 hours after major trauma did not improve 30-day death and/or major respiratory complications compared with a targeted SpO2 94% strategy, despite marked PaO2 hyperoxia in the liberal arm.
  The findings support an SpO2-targeted approach rather than automatic high-dose oxygen for all trauma patients who are not hypoxaemic, while recognising that the trial did not demonstrate superiority of restrictive oxygen and cannot exclude small benefits or harms.

Controversies & Other Evidence

• Effect modification by haemoglobin and competing risk considerations: Correspondence questioned whether haemoglobin (oxygen delivery) and competing risk of death could confound respiratory complication estimates; trialists responded that haemoglobin at ~1 hour was comparable between groups and that their approach (including inverse probability weighting) accounted for differential attrition and competing-risk-related missingness in their analysis framework.⁵⁶
• SpO2 versus PaO2 targeting and biological plausibility: Correspondence argued the trial effectively compared normoxaemia vs moderate–extreme hyperoxia (PaO2 85 vs 280 mm Hg at 1h) and that PaO2-based stratification might better identify harm/benefit; authors replied that PaO2 titration is impractical in prehospital/trauma/ward settings and noted planned subgroup work in thoracic injury patients.⁷⁶
• Relationship to broader oxygen-strategy evidence: Prior large meta-analytic work in acutely ill adults suggested potential harm with liberal oxygen therapy, while ICU trials have shown mixed effects depending on population and target definitions; TRAUMOX2 adds trauma-specific evidence in the hyperacute window and suggests no major clinical benefit from early high-dose oxygen across a broad trauma cohort.¹⁸

Summary

• Pragmatic international RCT in adult trauma patients (prehospital and trauma-centre enrolment) comparing 8 hours of SpO2-targeted restrictive oxygen (94%) versus a liberal high-dose strategy.
• Primary outcome (death and/or major respiratory complications by 30 days) was similar: 16.1% vs 16.7% (OR 1.01; 95% CI 0.75 to 1.37; P=0.94).
• Death (8.6% vs 7.3%) and major respiratory complications (8.9% vs 10.8%) did not differ significantly, despite large PaO2 separation (85 vs 280 mm Hg at ~1 hour).
• Atelectasis occurred less often with restrictive oxygen (27.6% vs 34.7%), while hypoxaemic episodes during the intervention were more frequent (6.0% vs 3.8%).
• Interpretation is tempered by open-label delivery and substantial pre-specified post-randomisation “secondary exclusions” for minor/no injuries (modified ITT population).

Overall Takeaway

TRAUMOX2 found no evidence that routine early high-dose oxygen for 8 hours improves 30-day survival or major respiratory outcomes after major trauma compared with an SpO2-targeted strategy, even though the liberal approach produced marked hyperoxaemia. The trial supports moving away from automatic liberal oxygen for non-hypoxaemic trauma patients, while highlighting remaining uncertainty about injury-specific subgroups and the clinical relevance of short-duration hyperoxia.

Bibliography

• 1Chu DK, Kim LHY, Young PJ, et al. Mortality and morbidity in acutely ill adults treated with liberal versus conservative oxygen therapy (IOTA): a systematic review and meta-analysis. Lancet. 2018;391(10131):1693-1705
• 2Baekgaard JS, Arleth T, Siersma V, et al. Comparing restrictive versus liberal oxygen strategies for trauma patients—the TRAUMOX2 trial: protocol for a randomised clinical trial. BMJ Open. 2022;12(11):e064047
• 3Arleth T, Baekgaard J, Siersma V, et al. Comparing restrictive versus liberal oxygen strategies for trauma patients: the TRAUMOX2 trial statistical analysis plan. Acta Anaesthesiol Scand. 2023;67(6):829-838
• 4Arleth T, Baekgaard J, Siersma V, et al. Early restrictive vs liberal oxygen for trauma patients: the TRAUMOX2 randomized clinical trial. JAMA. 2025;333(6):479-489
• 5Hsu CK, Jean SS, Lai CC. Early restrictive vs liberal oxygen for trauma patients. JAMA. 2025;334(2):180-181
• 6Baekgaard J, Siersma V, Steinmetz J. In reply: Early restrictive vs liberal oxygen for trauma patients. JAMA. 2025;334(2):181
• 7Patel JJ, Rosenthal MD. Early restrictive vs liberal oxygen for trauma patients. JAMA. 2025;334(2):180
• 8Schjørring OL, Klitgaard TL, Perner A, et al. Lower or higher oxygenation targets for acute hypoxemic respiratory failure. N Engl J Med. 2021;384(14):1301-1311