Skip to main content

Foundational Trials

RENOVATE

Image: Shutterstock / ChaNaWiT

Publication

  • Title: High-Flow Nasal Oxygen vs Noninvasive Ventilation in Patients With Acute Respiratory Failure: The RENOVATE Randomized Clinical Trial
  • Acronym: RENOVATE
  • Year: 2025
  • Journal published in: JAMA
  • Citation: Maia IS, Kawano-Dourado L, Tramujas L, et al; 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):875-889.

Context & Rationale

  • Background
    NIV is established therapy for selected acute respiratory failure syndromes (notably COPD exacerbation with acidosis and acute cardiogenic pulmonary oedema), but it can be poorly tolerated and is often delivered intermittently; HFNO is widely used for acute hypoxaemic respiratory failure (including COVID-19), offers humidified high-flow oxygen with a modest PEEP effect, and is generally better tolerated, but comparative effectiveness vs NIV across syndromes has been uncertain.
  • Research Question/Hypothesis
    In adults with acute respiratory failure requiring noninvasive respiratory support, HFNO is non-inferior to NIV for preventing endotracheal intubation or death within 7 days, evaluated within prespecified clinical syndrome groups.
  • Why This Matters
    If HFNO is reliably non-inferior across syndromes, it could simplify first-line support selection and improve comfort/tolerance; if effects differ by syndrome or immunocompromised status, substituting HFNO for NIV could expose some subgroups to avoidable harm.

Design & Methods

  • Research Question: Whether HFNO is non-inferior to NIV for preventing endotracheal intubation or death within 7 days, assessed within prespecified acute respiratory failure syndrome groups.
  • Study Type: Multicentre, pragmatic, Bayesian adaptive, open-label, randomised clinical trial conducted across 33 hospitals in Brazil; participants enrolled into prespecified clinical groups (including a COVID-19 hypoxaemia group defined by confirmed SARS-CoV-2 infection during admission).
  • Population:
    • Adults with acute respiratory failure requiring noninvasive respiratory support, categorised into one of five clinical groups: hypoxaemic nonimmunocompromised; hypoxaemic immunocompromised; acute cardiogenic pulmonary oedema; COPD exacerbation with respiratory acidosis; hypoxaemic COVID-19.
    • Key exclusions: immediate indication for intubation; do-not-intubate at enrolment; contraindications to NIV/HFNO (eg inability to protect airway).
    • Enrolled: 1800 randomised; primary analysis population: 1766 (randomised with informed consent obtained).
  • Intervention:
    • HFNO delivered via heated humidified high-flow system; flow and FiO2 titrated to oxygenation and work-of-breathing targets.
    • Early delivered settings: at hour 1, typical HFNO flows were 50–60 L/min (FiO2 varied by syndrome; higher in COVID-19 and immunocompromised hypoxaemia).
  • Comparison:
    • NIV via mask interface; pressure settings titrated per syndrome/physiology (including higher pressure support in COPD), with FiO2 adjusted similarly.
    • Among patients still receiving noninvasive support on day 1, support interruption occurred in 53.3% with NIV vs 14.2% with HFNO.
  • Blinding: Open-label; outcomes were objective (intubation/death) but intubation is clinician-mediated; protocolised intubation criteria were used to reduce decision bias.
  • Statistics: Non-inferiority framework using Bayesian hierarchical modelling (dynamic borrowing) with non-inferiority margin OR < 1.55 and posterior probability of non-inferiority threshold >0.992 (group-specific); power calculation (effect size, alpha, beta/power, required sample size): Not reported; analysis population: modified intention-to-treat (randomised with consent obtained).
  • Follow-Up Period: Primary endpoint at 7 days; additional outcomes followed to 28 and 90 days.

Key Results

This trial was not stopped early. Recruitment within individual clinical groups stopped according to prespecified Bayesian rules (non-inferiority stopping in some groups and futility stopping in the immunocompromised hypoxaemia group), while the platform continued.

Outcome HFNO NIV Effect p value / 95% CI Notes
Primary (overall): Intubation or death within 7 days 344/883 (39.0%) 336/883 (38.1%) OR 1.15 95% CrI 0.83 to 1.60; posterior probability of non-inferiority 0.989 Did not meet the prespecified overall non-inferiority probability threshold; primary inference was group-specific
Primary: Nonimmunocompromised hypoxaemia 81/249 (32.5%) 78/236 (33.1%) OR 1.02 95% CrI 0.81 to 1.26; posterior probability of non-inferiority 0.999 Non-inferiority met
Primary: Immunocompromised hypoxaemia 16/28 (57.1%) 8/22 (36.4%) OR 1.07 95% CrI 0.81 to 1.39; posterior probability of non-inferiority 0.989 Non-inferiority not met; group stopped for futility
Primary: Acute cardiogenic pulmonary oedema 14/136 (10.3%) 29/136 (21.3%) OR 0.97 95% CrI 0.73 to 1.23; posterior probability of non-inferiority 0.997 Non-inferiority met; raw event rates favoured HFNO but Bayesian estimate was near-null
Primary: COPD exacerbation with respiratory acidosis 10/35 (28.6%) 11/42 (26.2%) OR 1.05 95% CrI 0.79 to 1.36; posterior probability of non-inferiority 0.992 Non-inferiority met at the prespecified decision threshold
Primary: Hypoxaemic COVID-19 223/435 (51.3%) 210/447 (47.0%) OR 1.13 95% CrI 0.94 to 1.38; posterior probability of non-inferiority 0.997 Non-inferiority met for the composite endpoint; high event rates in both arms
Selected secondary (all patients): Discomfort score at 12 hours (0–100) Mean 25.89 (SD 24.42) Mean 32.25 (SD 24.15) Mean difference −6.73 95% CrI −9.25 to −4.24 HFNO was more comfortable
Selected safety (all patients): ≥1 serious adverse event 83/883 (9.4%) 87/883 (9.9%) Not reported Not reported Serious adverse events were similar between groups (no model estimate reported)
  • HFNO met prespecified non-inferiority criteria vs NIV for the day-7 composite endpoint in 4/5 syndrome groups; the immunocompromised hypoxaemia group did not meet non-inferiority.
  • HFNO was delivered more continuously and was better tolerated: day 1 support interruption 14.2% (HFNO) vs 53.3% (NIV); discomfort at 12 hours mean 25.89 vs 32.25 (mean difference −6.73; 95% CrI −9.25 to −4.24).
  • Overall non-inferiority for the composite endpoint was not met (posterior probability 0.989, below the prespecified threshold), emphasising that inference was designed to be group-specific.

Internal Validity

  • Randomisation and Allocation: Central randomisation supports allocation concealment and minimises selection bias.
  • Drop out or exclusions: 34/1800 randomised patients were excluded from the primary analysis population (1766 analysed), because informed consent was not obtained post randomisation; this modified intention-to-treat approach introduces potential post-randomisation exclusion bias.
  • Performance/Detection Bias: Open-label design creates risk of performance bias for clinician-mediated intubation decisions; use of objective endpoints and protocolised intubation criteria mitigates but does not remove this risk.
  • Protocol Adherence: Delivered “dose” differed meaningfully, consistent with intended separation: day 1 support interruption among those still receiving support was 14.2% with HFNO vs 53.3% with NIV; discomfort at 12 hours was lower with HFNO (mean difference −6.73; 95% CrI −9.25 to −4.24).
  • Baseline Characteristics: Groups were broadly comparable at baseline overall (mean age 64 years; 40% women); syndrome heterogeneity was prespecified by design.
  • Heterogeneity: The trial intentionally enrolled five physiologically distinct syndromes; hierarchical modelling with dynamic borrowing assumes partial exchangeability across strata, increasing model dependence in small groups.
  • Timing: Primary outcome at 7 days aligns with early noninvasive support failure; timing of initiation relative to syndrome onset: Not reported in the provided text.
  • Dose: Early HFNO flow was typically 50–60 L/min at hour 1; NIV pressures were titrated per syndrome; detailed per-syndrome dose metrics beyond these high-level descriptors: Not reported in the provided text.
  • Separation of the Variable of Interest: Day 1 support interruption 14.2% (HFNO) vs 53.3% (NIV), and discomfort at 12 hours mean 25.89 vs 32.25 support meaningful separation in tolerability and continuity of delivery.
  • Crossover: Rescue NIV use in the HFNO arm occurred and was highest in the COPD group (22.9%); impact is likely towards dilution of between-group differences in COPD.
  • Outcome Assessment: Intubation/death are clinically meaningful and relatively objective outcomes, but intubation remains clinician-mediated.
  • Statistical Rigor: Bayesian adaptive design with prespecified decision thresholds; results in smaller strata are sensitive to modelling assumptions, particularly dynamic borrowing. 12

Conclusion on Internal Validity: Overall internal validity is moderate-to-strong, supported by central randomisation and objective outcomes, but limited by open-label care, post-randomisation consent-based exclusions (modified intention-to-treat), and model dependence (dynamic borrowing) in small syndrome strata.

External Validity

  • Population Representativeness: Adults requiring noninvasive respiratory support across five common acute respiratory failure syndromes, enhancing pragmatic relevance; syndrome-specific exclusions and thresholds may vary by centre and were not fully detailed in the provided text.
  • Applicability: Findings are most applicable to ICUs/EDs with established capability in both HFNO and NIV and similar intubation thresholds; centres using alternative NIV interfaces (eg helmet) or achieving higher NIV tolerance may observe different comparative effects.
  • Context effects: COVID-19 subgroup reflects pandemic-era case mix and care pathways; applicability to current COVID-19 practice depends on local standards of care and patient characteristics.

Conclusion on External Validity: External validity is generally good for real-world “first device” decisions, but syndrome-specific application should account for local expertise, interface choice, and evolving standards of care (particularly for COVID-19).

Strengths & Limitations

  • Strengths: Pragmatic multicentre randomised design; clinically meaningful early failure endpoint; prespecified adaptive Bayesian framework enabling group-specific inference; detailed reporting of tolerability and delivery continuity (interruptions; discomfort).
  • Limitations: Open-label design with clinician-mediated intubation decisions; modified intention-to-treat analysis excluding 34 randomised patients without consent; small samples in some strata (notably immunocompromised hypoxaemia and COPD) increasing uncertainty and model dependence.

Interpretation & Why It Matters

  • Syndrome-specific device choice
    HFNO met prespecified non-inferiority vs NIV for the day-7 composite endpoint in four groups, supporting HFNO as a reasonable first-line option where NIV tolerance is poor; failure to demonstrate non-inferiority in immunocompromised hypoxaemia supports caution against assuming interchangeability in this subgroup.
  • Tolerance and delivery
    HFNO was more comfortable and more continuously delivered (day 1 interruption 14.2% vs 53.3%; discomfort mean 25.89 vs 32.25), which is clinically meaningful where patient tolerance and staffing constraints influence escalation decisions.

Controversies & Other Evidence

  • Dynamic borrowing and subgroup interpretability: Hierarchical modelling can shrink subgroup effects towards an overall mean and makes smaller strata particularly sensitive to modelling assumptions; editorials highlighted that qualitative interpretation may change under alternative (no-borrowing) analyses, especially in small groups. 12
  • Non-inferiority margin acceptability: A margin of OR 1.55 may permit clinically meaningful harm in high-event settings; acceptability differs by syndrome given differing baseline risk, established efficacy of NIV in some indications, and patient-centred trade-offs (comfort vs avoiding intubation). 2
  • Guideline and prior trial context: NIV remains positioned as first-line therapy in conditions with robust evidence (eg COPD with acidosis; acute cardiogenic pulmonary oedema) and oxygen strategies are typically framed as syndrome-specific; RENOVATE supports HFNO as an alternative pathway but does not, alone, displace NIV in established indications. 34
  • De novo acute hypoxaemic respiratory failure: Earlier RCT evidence supports HFNO as an effective, well-tolerated strategy in acute hypoxaemic respiratory failure, aligning directionally with RENOVATE’s nonimmunocompromised hypoxaemia findings. 5
  • COVID-19 evidence triangulation: COVID-era RCTs comparing noninvasive strategies have shown heterogeneous results across interfaces and protocols; RENOVATE’s COVID subgroup should be interpreted alongside other large COVID noninvasive strategy trials rather than in isolation. 67
  • Further Reading
    • Other trials: FLORALI. 5
    • Systematic reviews & meta-analyses: HFNO vs NIV in acute heart failure-related respiratory failure. 8
    • Guidelines: SRLF-SFMU consensus on oxygen therapy in acute hypoxaemic respiratory failure; ERS/ATS NIV guideline for acute respiratory failure. 34

Summary

  • Pragmatic Brazilian multicentre randomised trial comparing HFNO vs NIV across five prespecified acute respiratory failure syndromes using a Bayesian adaptive non-inferiority framework.
  • HFNO met prespecified non-inferiority criteria for the day-7 composite endpoint (intubation or death) in 4/5 groups; non-inferiority was not demonstrated in immunocompromised hypoxaemia.
  • Overall non-inferiority for the composite endpoint was not met (posterior probability 0.989, below threshold), reinforcing that inference was designed to be syndrome-specific.
  • HFNO was better tolerated and more continuously delivered: day 1 support interruption 14.2% vs 53.3%; discomfort at 12 hours mean 25.89 vs 32.25 (mean difference −6.73; 95% CrI −9.25 to −4.24).
  • Interpretation in small strata is sensitive to modelling assumptions (dynamic borrowing) and to open-label care with clinician-mediated intubation decisions.

Overall Takeaway

RENOVATE supports HFNO as a broadly reasonable first-line noninvasive support option across several acute respiratory failure syndromes, with a clear tolerability advantage over NIV. However, the immunocompromised hypoxaemia group did not demonstrate non-inferiority, and subgroup inferences—especially in smaller strata—remain sensitive to modelling assumptions and the realities of open-label care.

Overall Summary

  • Across five acute respiratory failure syndromes, HFNO met prespecified Bayesian non-inferiority vs NIV for day-7 intubation/death in four groups, but not in immunocompromised hypoxaemia.
  • HFNO was better tolerated and more continuously delivered (day 1 interruption 14.2% vs 53.3%; discomfort mean 25.89 vs 32.25).
  • Inference in smaller strata is model-dependent (dynamic borrowing) and should be applied with caution alongside syndrome-specific evidence and guidelines. 12

Bibliography