SOHO — High-Flow vs Standard Oxygen in Acute Hypoxemic Respiratory Failure

1. Publication

• Title: High-Flow or Standard Oxygen in Acute Hypoxemic Respiratory Failure
• Acronym: SOHO
• Year & Journal: New England Journal of Medicine, epublished March 17, 2026 (print: 2026;394(21):2095-2106)
• Citation: Frat JP, Quenot JP, Guitton C, et al; SOHO Trial Group and the REVA Network. High-Flow or Standard Oxygen in Acute Hypoxemic Respiratory Failure. N Engl J Med. 2026;394(21):2095-2106. doi:10.1056/NEJMoa2516087

2. Context & Rationale

Background: Standard oxygen remained the default first-line interface for de novo acute hypoxemic respiratory failure (AHRF) despite delivering an imprecise FiO2 and doing little to unload inspiratory effort. High-flow nasal oxygen (HFNO) delivers more stable FiO2, generates modest PEEP, washes out dead space, and is better tolerated. Prior evidence was directionally favorable but inconsistent: the original FLORALI trial suggested a survival signal in more severe hypoxemia; the HIGH trial in immunocompromised patients was neutral. By 2024, guidelines already favored HFNO mainly on presumed intubation reduction rather than proven mortality benefit.

Research Question/Hypothesis: In adults with de novo AHRF, does first-line HFNO reduce 28-day mortality compared with standard oxygen?

Why This Matters: SOHO was designed as the largest direct HFNO-vs-standard-oxygen superiority trial in this syndrome, explicitly mortality-focused (not just intubation-focused), testing whether HFNO's expanding clinical adoption had outrun its evidence base.

3. Design & Methods

• Study Type: Investigator-initiated, multicenter, open-label, randomized, parallel-group superiority trial
• Setting & Centers: 42 ICUs in France
• Population:
• Inclusion: Adults with AHRF, respiratory rate >25/min, pulmonary infiltrates on imaging, PaO2:FiO2 ≤200 while receiving ≥10 L/min oxygen via non-rebreather mask; randomized within 24h of ICU admission
• Exclusions: PaCO2 >45 mmHg; severe COPD/chronic lung disease requiring long-term O2/ventilation; cardiogenic pulmonary edema; hemodynamic instability (vasopressors >0.3 μg/kg/min or hypoperfusion); GCS ≤12; post-extubation respiratory failure; postoperative abdominal/cardiothoracic respiratory failure within 7 days; need for emergency intubation; do-not-intubate order
• Intervention: Continuous HFNO via heated humidification, flow ≥50 L/min, FiO2 titrated to SpO2 92–96% for ≥48h, then step-down criteria
• Comparator: Continuous standard oxygen via facemask/non-rebreather, flow ≥10 L/min, same SpO2 target and step-down logic; rescue NIV discouraged in both arms
• Blinding: Open-label bedside care; centralized, concealed randomization pre-allocation
• Statistical Power & Follow-Up: 1110 patients required to detect a 6% absolute mortality reduction (18%→12%) at 80% power, two-sided α=0.05; no interim analyses/stopping rules. ITT analysis. Primary endpoint at day 28; additional outcomes to day 90.

4. Key Results

Not stopped early. 1116 randomized; 1110 (556 HFNO, 554 standard) analyzed.

Subgroups: No convincing heterogeneity; intubation benefit numerically larger in immunocompromised patients and PaO2:FiO2 ≤100, but all subgroup CIs crossed zero.

5. Internal Validity Assessment

• Randomization & Allocation: Central, web-based, permuted-block, stratified by immunosuppression status — strong concealment.
• Protocol Adherence & Separation: Good physiological separation (initial flow 51±9 L/min, FiO2 0.67±0.17 vs O2 flow 12±3 L/min). Crossover: 30 HFNO→standard, 14 standard→HFNO (device intolerance). Rescue NIV rare (15 vs 22).
• Blinding & Detection Bias: Open-label; mortality is a hard endpoint, but intubation and dyspnea are more performance-bias-susceptible.
• Missing Data: Minimal attrition — 6/1116 excluded (5 consent withdrawals, 1 investigator withdrawal).
• Overall Internal Validity Conclusion: Moderate-to-strong — early treatment delivery, low attrition, and a hard primary endpoint support the null result; power was reduced for smaller effects since observed control mortality (14.6%) was lower than the assumed 18%.

6. External Validity Assessment

• Population Representativeness: Mean PaO2:FiO2 ~132, pneumonia in 88.1%, immunocompromise in 22.3%, bilateral infiltrates 87% — representative of de novo AHRF in high-income ICUs; nearly half the cohort had COVID-19-related illness.
• Practice Context: Requires humidification systems, high-flow capability, and trained staff — may limit adoption in lower-resource settings. Findings do not apply to hypercapnic failure, severe COPD, cardiogenic edema, hemodynamic instability, depressed consciousness, post-extubation or postoperative respiratory failure (all excluded).
• Overall External Validity Conclusion: Good for adults in well-resourced ICUs with de novo moderate-to-severe AHRF, especially pneumonia-dominant populations; limited outside ICU or in excluded syndromes.

7. Strengths & Limitations

Strengths:

• Largest direct HFNO-vs-standard-oxygen mortality-focused trial in this syndrome
• Early randomization, credible intervention dose, hard primary endpoint
• Prespecified intubation criteria; low crossover, minimal rescue NIV use

Limitations:

• Open-label design; no data and safety monitoring board
• Lower-than-expected control mortality reduced power for smaller effects
• Secondary outcomes not multiplicity-adjusted
• High proportion of viral/COVID-19 pneumonia limits generalizability to that specific case mix

8. Interpretation & Practice Impact

• Clinical Implications: HFNO should not be presented as a proven mortality-reducing therapy in unselected AHRF; its case rests on probable intubation avoidance and comfort, not survival.
• Mechanistic Coherence: Similar time-to-intubation between groups argues against HFNO merely delaying an inevitable, harmful intubation.
• Systems-Level Takeaway: Supports HFNO as foundational first-line respiratory support for intubation avoidance and physiological unloading, not for a demonstrated survival gain.

9. Controversies & Subsequent Evidence

• Editorial Commentary: An accompanying NEJM editorial (Serpa Neto) argued 28-day mortality may be an unrealistically distal endpoint for an oxygen-delivery interface, and that intubation, comfort, and treatment burden may be more mechanistically coherent outcomes.
• Guideline Integration: Consistent with (rather than overturning) recent guideline thinking (2024 SRLF-SFMU guideline) and the RENOVATE trial (HFNO vs NIV), which together support HFNO as a major first-line respiratory-support component even without a demonstrated mortality edge over standard oxygen.

10. Summary & Executive Takeaway

Summary: SOHO randomized 1110 ICU patients with de novo AHRF to HFNO or standard oxygen. 28-day mortality was identical (14.6% both arms; P=0.98). HFNO probably reduced intubation (42.4% vs 48.4%) and improved early respiratory physiology and comfort, without evidence of delayed intubation.

Overall Takeaway: SOHO recalibrates the case for HFNO in AHRF — valuable for intubation avoidance, symptom relief, and physiological unloading, but not a mortality-reducing intervention in unselected patients.

11. Bibliography

• Frat JP, Thille AW, Mercat A, et al. High-flow oxygen through nasal cannula in acute hypoxemic respiratory failure (FLORALI). N Engl J Med. 2015;372(23):2185-2196.
• Azoulay E, Lemiale V, Mokart D, et al. HIGH trial. JAMA. 2018;320(20):2099-2107.
• RENOVATE Investigators and the BRICNet Authors. JAMA. 2025;333(10):875-890.
• Serpa Neto A. Rethinking high-flow oxygen in acute hypoxemic respiratory failure [editorial]. N Engl J Med. 2026.