Publication

• Title: High-Flow or Standard Oxygen in Acute Hypoxemic Respiratory Failure
• Acronym: SOHO
• Year: 2026
• Journal published in: New England Journal of Medicine
• Citation: Frat JP, Quenot JP, Guitton C, et al. High-flow or standard oxygen in acute hypoxemic respiratory failure. N Engl J Med 2026;epublished March 17th.

Context & Rationale

• Background

  Standard oxygen had remained the default first-line interface for many ICU patients with de novo acute hypoxaemic respiratory failure, but it delivers an imprecise inspired oxygen fraction and does little to unload inspiratory effort.
  High-flow nasal oxygen had become increasingly attractive because it can deliver a more stable FiO₂, generate a modest PEEP effect, wash out upper-airway dead space, reduce work of breathing, and is generally better tolerated than mask-based support.
  Before SOHO, the clinical evidence was directionally favourable but internally inconsistent: FLORALI suggested a survival signal and less intubation in more severe hypoxaemia, whereas the HIGH trial in immunocompromised patients was neutral; by 2024, guidelines were already leaning towards high-flow oxygen mainly because of probable intubation reduction rather than proven mortality benefit. ¹²³
• Research Question/Hypothesis

  In adults admitted to ICU with moderate-to-severe de novo acute hypoxaemic respiratory failure, does first-line high-flow nasal oxygen reduce 28-day mortality, as compared with standard oxygen? The protocol hypothesis was that high-flow nasal oxygen would improve prognosis, including mortality and intubation. ⁴
• Why This Matters

  SOHO was the largest direct high-flow-versus-standard-oxygen superiority trial in this syndrome and was explicitly mortality-focused rather than centred only on intubation or physiological surrogates.
  It therefore tested whether the expanding clinical position of high-flow oxygen had outrun the evidence base, or whether its physiological advantages translated into better patient-important outcomes against a genuine standard-oxygen comparator. ⁴⁵

Design & Methods

• Research Question: Was first-line high-flow nasal oxygen superior to standard oxygen for reducing 28-day mortality in adults with de novo acute hypoxaemic respiratory failure?
• Study Type: Investigator-initiated, multicentre, open-label, randomised, parallel-group, superiority trial conducted in 42 ICUs in France.
• Population:
  • Adults admitted to ICU with acute hypoxaemic respiratory failure, respiratory rate >25 breaths/min, pulmonary infiltrates on chest imaging, and PaO₂:FiO₂ ratio ≤200 while receiving oxygen at ≥10 L/min through a non-rebreather mask.
  • Randomisation occurred within 24 hours after ICU admission; allocated treatment had to start within 3 hours after randomisation.
  • Key exclusions: PaCO₂ >45 mm Hg, severe COPD or other chronic lung disease requiring long-term oxygen or ventilatory support, cardiogenic pulmonary oedema, haemodynamic instability with vasopressors >0.3 μg/kg/min or signs of hypoperfusion, Glasgow Coma Scale ≤12, post-extubation respiratory failure, postoperative abdominal or cardiothoracic respiratory failure within 7 days, need for emergency intubation, or do-not-intubate order.
• Intervention:
  • Continuous high-flow nasal oxygen through heated humidification and large-bore binasal prongs, with gas flow ≥50 L/min.
  • FiO₂ was titrated to maintain SpO₂ 92% to 96% for at least 48 hours.
  • After 48 hours, high-flow oxygen could be stepped down if respiratory rate was ≤25 breaths/min and SpO₂ was ≥92% with FiO₂ ≤40%.
• Comparison:
  • Continuous standard oxygen delivered through facemask or non-rebreather mask, with oxygen flow ≥10 L/min.
  • Oxygen flow was titrated to maintain SpO₂ 92% to 96% for at least 48 hours.
  • After 48 hours, standard oxygen could be stepped down if respiratory rate was ≤25 breaths/min and SpO₂ was ≥92% with oxygen flow <6 L/min.
  • Rescue noninvasive ventilation was discouraged in both groups.
• Blinding: Bedside treatment was unblinded. Allocation was centralised and concealed before randomisation. The primary outcome was objective, but intubation remained partly susceptible to clinician judgement despite prespecified criteria.
• Statistics: A total of 1110 patients were required to detect a 6% absolute reduction in 28-day mortality (from 18% to 12%) with 80% power at a two-sided 5% significance level. Analysis was intention-to-treat after blind review of protocol deviations; no interim analyses or stopping rules were prespecified.
• Follow-Up Period: Primary endpoint at day 28, with additional mortality and clinical outcomes through day 90.

Key Results

This trial was not stopped early. A total of 1116 patients underwent randomisation; 1110 patients (556 assigned to high-flow oxygen and 554 assigned to standard oxygen) were included in the intention-to-treat analysis.

• The primary endpoint was unequivocally neutral: 28-day mortality was identical at 14.6% in both groups, and the confidence interval excludes any large mortality benefit.
• High-flow oxygen probably reduced intubation, but this was a secondary outcome and the reported confidence intervals were not adjusted for multiplicity.
• Prespecified subgroup analyses showed no convincing heterogeneity. For intubation, the absolute difference favoured high-flow oxygen in immunocompromised patients (−9.68 percentage points; 95% CI −21.93 to 2.58) and in those with PaO₂:FiO₂ ≤100 mm Hg (−9.74 percentage points; 95% CI −23.30 to 3.82), but all subgroup confidence intervals crossed zero.

Internal Validity

• Randomisation and Allocation: Allocation concealment appears strong. Randomisation was central, web-based, in permuted blocks, and stratified by immunosuppression status.
• Drop out or exclusions: Attrition was minimal. Six of 1116 randomised patients were excluded from analysis: 5 consent withdrawals and 1 investigator withdrawal after urgent intubation. This is a small but real departure from a pure all-randomised analysis.
• Performance/Detection Bias: Open-label delivery was unavoidable and matters especially for intubation, rescue therapy, and patient-reported dyspnoea. Mortality, however, is a hard endpoint.
• Protocol Adherence: Treatment separation was good. Initial settings were 51±9 L/min with FiO₂ 0.67±0.17 in the high-flow group versus oxygen flow 12±3 L/min in the standard-oxygen group. Device intolerance led to crossover in 30 patients from high-flow to standard oxygen and 14 patients from standard to high-flow. Rescue noninvasive ventilation was uncommon: 15 versus 22 patients.
• Baseline Characteristics: Groups were broadly comparable, although the high-flow group had more men, a higher baseline dyspnoea score, and a slightly lower respiratory rate. The adjusted primary analysis remained neutral.
• Heterogeneity: Clinical heterogeneity was present but not excessive. Pneumonia accounted for 88.1% of cases, immunocompromise for 22.3%, and only 17% had PaO₂:FiO₂ ≤100 mm Hg. Nearly half the cohort had COVID-19-related illness, which adds biological heterogeneity but did not create an obvious subgroup signal.
• Timing: Timing was appropriate and early. Median time from ICU admission to randomisation was 2.5 hours with high-flow oxygen and 2.4 hours with standard oxygen. High-flow therapy was started a median of 18 minutes after randomisation.
• Dose: The intervention dose was credible and clinically relevant. High-flow oxygen was delivered continuously for at least 48 hours with flow ≥50 L/min; among non-intubated patients, median high-flow duration was 4 days.
• Separation of the Variable of Interest: Physiological separation was clear. At 1 hour, respiratory rate was 26±7 versus 29±7 breaths/min and PaCO₂ was 34±5 versus 36±5 mm Hg. Oxygenation metrics are harder to interpret because FiO₂ was directly set in the high-flow group but estimated in the standard-oxygen group.
• Key Delivery Aspects: The control was a legitimate standard-oxygen strategy, not a weakened comparator. Prespecified intubation criteria were applied in both groups, which strengthens the intubation comparison.
• Outcome Assessment: The primary outcome was objective and patient-important. Intubation criteria were prespecified, but adherence to those criteria was not prospectively documented among patients who were not intubated.
• Statistical Rigor: The primary analysis matched the protocol and was appropriately powered for a fairly large absolute mortality effect. The main statistical limitation is that actual control mortality was lower than anticipated (14.6% vs 18%), reducing power to detect smaller benefits. Secondary outcomes were not multiplicity-adjusted.

Conclusion on Internal Validity: Internal validity is moderate to strong. Randomisation, early treatment delivery, low attrition, and a hard primary endpoint support the result, but open-label care, modest baseline imbalance, and reduced power for smaller mortality effects temper certainty about more subtle benefits.

External Validity

• Population Representativeness: The cohort is highly representative of ICU patients with de novo acute hypoxaemic respiratory failure in high-income health systems: mean PaO₂:FiO₂ approximately 132, pneumonia in 88.1%, immunocompromise in 22.3%, and bilateral infiltrates in 87%.
• Important Exclusions: The findings do not directly apply to hypercapnic respiratory failure, severe COPD or chronic ventilatory-support populations, cardiogenic pulmonary oedema, significant haemodynamic instability, depressed consciousness, post-extubation failure, or postoperative respiratory failure.
• Setting: This was a 42-centre French ICU trial with high-flow equipment, arterial blood-gas monitoring, and protocolised escalation criteria. Applicability is strongest to similar ICU environments.
• COVID-era Case Mix: Almost half the cohort had COVID-19-related respiratory failure. That increases relevance to the pandemic era but also means case mix was unusually viral-pneumonia heavy.
• Resource Requirements: High-flow oxygen requires humidification systems, high oxygen flow capability, and trained staff. These resource demands may limit adoption or dilute benefit in lower-resource settings.

Conclusion on External Validity: External validity is good for adults in well-resourced ICUs with de novo moderate-to-severe acute hypoxaemic respiratory failure, especially pneumonia-dominant populations. Generalisability is more limited outside ICU settings and in hypercapnic, cardiogenic, postoperative, or chronic respiratory disease syndromes.

Strengths & Limitations

• Strengths: Largest direct high-flow-versus-standard-oxygen trial in this syndrome; mortality-focused design; multicentre ICU recruitment; early randomisation; credible high-flow dose; hard primary endpoint; prespecified intubation criteria; low crossover and little rescue NIV use.
• Limitations: Open-label design; no data and safety monitoring board; lower-than-expected mortality reducing power for smaller effects; secondary outcomes not multiplicity-adjusted; high proportion of viral and COVID-19 pneumonia; no prospective audit of intubation-criteria adherence in non-intubated patients.

Interpretation & Why It Matters

• Mortality signal

  SOHO is a clearly negative superiority trial for its primary endpoint. High-flow oxygen should not now be presented as a proven mortality-reducing therapy in unselected ICU patients with de novo acute hypoxaemic respiratory failure.
• Intubation signal

  The clinically relevant residual signal is lower intubation. Although this remains statistically more fragile than the primary result, the similar time to intubation between groups argues against the idea that high-flow oxygen merely delayed an inevitable and potentially harmful intubation.
• Outcome selection

  The accompanying editorial makes the important methodological point that 28-day mortality may be an unrealistically distal endpoint for an oxygen-delivery interface. For such interventions, intubation, comfort, respiratory effort, and treatment burden may be more mechanistically coherent measures of benefit. ⁶
• Place in modern practice

  SOHO narrows, rather than abolishes, the role of high-flow oxygen. Together with recent guideline thinking and the RENOVATE comparison against noninvasive ventilation, it supports high-flow oxygen as a foundational first-line support strategy, but chiefly for intubation avoidance, tolerability, and physiological unloading rather than for demonstrated survival gain. ³⁷

Controversies & Other Evidence

• Immediately before SOHO, the field was already split over whether high-flow oxygen had effectively become the “reference” treatment in acute hypoxaemic respiratory failure. The 2025 pro/con debate in Journal of Intensive Medicine crystallised that disagreement: one side emphasised lower intubation and ease of use, while the other stressed inconsistent mortality evidence, delayed-intubation concerns, and resource cost. SOHO resolves part of that debate by ruling out any large 28-day mortality benefit, while leaving the intubation question partially open. ⁹¹⁰
• The trial’s neutral mortality result should be interpreted with two competing methodological truths in mind: first, the confidence interval effectively excludes a large mortality benefit; second, the study was powered for a 6% absolute reduction and observed lower-than-expected control mortality, so it cannot exclude a smaller effect. The accompanying editorial is persuasive in arguing that this limitation should recalibrate expectations about what oxygen-interface trials can realistically deliver on mortality. ⁶
• SOHO sits in partial tension with FLORALI, which generated the original survival enthusiasm, but it is more consistent with later neutral mortality trials such as HIGH in immunocompromised patients and SOHO-COVID in severe COVID-19. The overall trajectory of evidence is therefore towards probable intubation benefit without reproducible mortality benefit. ¹²⁸
• Guidelines and evidence syntheses published before SOHO had already placed high-flow oxygen in a favourable position, largely because of reduced intubation and better tolerance. SOHO makes that narrower framing more defensible: the 2024 SRLF-SFMU guideline and the 2023 network meta-analysis remain broadly aligned with present practice, while RENOVATE reinforces that high-flow oxygen is at least a major component of modern noninvasive respiratory support even when superiority over standard oxygen on mortality is absent. ³⁵⁷

Summary

• SOHO randomised 1110 analysed ICU patients with de novo acute hypoxaemic respiratory failure to high-flow oxygen or standard oxygen.
• The primary outcome was neutral: 28-day mortality was 14.6% in both groups.
• High-flow oxygen reduced intubation numerically and probably clinically: 42.4% versus 48.4%, absolute difference −5.93 percentage points; 95% CI −11.78 to −0.08.
• High-flow oxygen improved short-term physiological and patient-centred outcomes, including respiratory rate, PaCO₂, and perceived dyspnoea, without clear evidence of delayed intubation.
• The trial resets expectations: high-flow oxygen remains useful as first-line support, but not as a proven mortality-reducing intervention in unselected ICU acute hypoxaemic respiratory failure.

Overall Takeaway

SOHO is a landmark recalibration trial. It shows that the modern case for high-flow oxygen in acute hypoxaemic respiratory failure should rest chiefly on probable intubation avoidance, symptom relief, and physiological unloading rather than on a reproducible survival benefit. For clinicians and trialists, it is the clearest demonstration to date that a biologically plausible and widely adopted respiratory-support interface can be clinically valuable without being mortality-reducing.

Bibliography

• 1.Frat JP, Thille AW, Mercat A, et al. High-flow oxygen through nasal cannula in acute hypoxemic respiratory failure. N Engl J Med. 2015;372(23):2185-2196.
• 2.Azoulay E, Lemiale V, Mokart D, et al. Effect of high-flow nasal oxygen vs standard oxygen on 28-day mortality in immunocompromised patients with acute respiratory failure: the HIGH randomized clinical trial. JAMA. 2018;320(20):2099-2107.
• 3.Helms J, Catoire P, Abensur Vuillaume L, et al. Oxygen therapy in acute hypoxemic respiratory failure: guidelines from the SRLF-SFMU consensus conference. Ann Intensive Care. 2024;14(1):140.
• 4.Frat JP, Coudroy R, Quenot JP, et al. Effect of high-flow nasal cannula oxygen versus standard oxygen on mortality in patients with acute hypoxaemic respiratory failure: protocol for a multicentre, randomised controlled trial (SOHO). BMJ Open. 2024;14(10):e083232.
• 5.Pitre T, Zeraatkar D, Kachkovski GV, et al. Noninvasive oxygenation strategies in adult patients with acute hypoxemic respiratory failure: a systematic review and network meta-analysis. Chest. 2023;164(4):913-928.
• 6.Serpa Neto A. Rethinking high-flow oxygen in acute hypoxemic respiratory failure. N Engl J Med. Published online March 17, 2026.
• 7.RENOVATE Investigators and the BRICNet Authors. High-flow nasal oxygen vs noninvasive ventilation in patients with acute respiratory failure: the RENOVATE randomized clinical trial. JAMA. 2025;333(10):875-890.
• 8.Frat JP, Quenot JP, Badie J, et al. Effect of high-flow nasal cannula oxygen vs standard oxygen therapy on mortality in patients with respiratory failure due to COVID-19: the SOHO-COVID randomized clinical trial. JAMA. 2022;328(12):1212-1222.
• 9.Frat JP, Le Pape S. High-flow nasal oxygen is the reference treatment in acute hypoxemic respiratory failure: Pro. J Intensive Med. 2025;5(3):222-229.
• 10.Kemoun G, Demoule A. High-flow nasal oxygen is the reference treatment in acute hypoxemic respiratory failure: Con. J Intensive Med. 2025;5(3):230-236.