Publication

• Title: Non-invasive ventilation versus high-flow nasal cannula oxygen therapy with apnoeic oxygenation for preoxygenation before intubation of patients with acute hypoxaemic respiratory failure: a randomised, multicentre, open-label trial
• Acronym: FLORALI-2
• Year: 2019
• Journal published in: The Lancet Respiratory Medicine
• Citation: Frat JP, Ricard JD, Quenot JP, et al; FLORALI-2 study group and REVA network. Non-invasive ventilation versus high-flow nasal cannula oxygen therapy with apnoeic oxygenation for preoxygenation before intubation of patients with acute hypoxaemic respiratory failure: a randomised, multicentre, open-label trial. Lancet Respir Med. 2019;7(4):303-312.

Context & Rationale

• Background

  • Tracheal intubation in acute hypoxaemic respiratory failure (AHRF) carries a high risk of peri-procedural desaturation and cardiovascular collapse, with physiology often dominated by shunt and low oxygen reserve.
  • Non-invasive ventilation (NIV) preoxygenation can deliver high FiO₂, positive end-expiratory pressure (PEEP), and active ventilation; prior randomised evidence showed less severe desaturation versus standard oxygen during ICU intubation preoxygenation.¹
  • High-flow nasal cannula (HFNC) delivers high FiO₂ at high flows, modest distending pressure, and—critically—can remain in place during laryngoscopy to provide apnoeic oxygenation; HFNC had already become widely used in AHRF after earlier trials of non-invasive respiratory support in hypoxaemia.²
  • Specific comparative evidence for NIV versus HFNC (with apnoeic oxygenation) as the preoxygenation strategy immediately before ICU intubation in AHRF remained uncertain; a prior randomised trial of HFNC during intubation in hypoxaemic patients informed feasibility and equipoise for HFNC-based preoxygenation.³
• Research Question/Hypothesis

  • In ICU adults with AHRF requiring intubation, NIV (during preoxygenation and between induction and laryngoscopy) would reduce the incidence of severe hypoxaemia during the intubation procedure compared with HFNC delivered at high flow with apnoeic oxygenation continued throughout laryngoscopy.
• Why This Matters

  • Preoxygenation is a modifiable, immediately actionable component of the “intubation bundle” aimed at preventing catastrophic hypoxaemia and downstream complications.
  • HFNC and NIV each have plausible, competing mechanistic advantages (apnoeic oxygenation continuity vs ventilatory/PEEP support), and the optimal approach likely depends on baseline severity of hypoxaemia.
  • Clear evidence could standardise practice across ICUs, reduce practice variation, and inform guideline recommendations for physiologically difficult airway management in AHRF.

Design & Methods

• Research Question: In ICU adults requiring intubation for acute hypoxaemic respiratory failure, does NIV-based preoxygenation reduce severe hypoxaemia during the intubation procedure compared with HFNC preoxygenation with apnoeic oxygenation continued during laryngoscopy?
• Study Type: Randomised, multicentre, parallel-group, open-label trial in 28 ICUs in France; stratified randomisation by centre and PaO₂/FiO₂ category (≤200 vs >200 mm Hg).
• Population:
  • Adults (>18 years) in ICU requiring intubation for acute hypoxaemic respiratory failure.
  • AHRF definition: respiratory rate >25/min and/or signs of respiratory distress, with PaO₂/FiO₂ ≤300 mm Hg.
  • Key exclusions: intubation for cardiac arrest; Glasgow Coma Score <8; contraindication to NIV (eg, recent laryngeal/oesophageal/gastric surgery, facial fractures); inability to obtain pulse oximetry; pregnancy/breastfeeding; refusal.
• Intervention:
  • NIV via face mask connected to an ICU ventilator for 3–5 min in 30° semi-recumbent position.
  • FiO₂ 1.0; PEEP 5 cm H₂O; pressure support adjusted to achieve expired tidal volume 6–8 mL/kg predicted body weight.
  • NIV used for preoxygenation and continued between induction and laryngoscopy; mask removed for laryngoscopy.
  • Standardised intubation bundle: two operators; rapid sequence induction (eg, etomidate 0.2–0.3 mg/kg or ketamine 1.5–3 mg/kg plus rocuronium 0.6–1 mg/kg or succinylcholine 1 mg/kg); predefined difficult airway algorithm.
• Comparison:
  • HFNC preoxygenation for 3–5 min in 30° semi-recumbent position.
  • FiO₂ 1.0; gas flow 60 L/min via heated humidifier and binasal prongs.
  • HFNC maintained during laryngoscopy until endotracheal tube placement (apnoeic oxygenation).
  • Same intubation bundle and rescue/difficult airway algorithm as intervention group.
• Blinding: Open-label to treating clinicians; primary outcome and complications assessed by an independent adjudication committee blinded to group assignment; investigators blinded to adjudicated outcomes until database lock.
• Statistics: A total of 320 patients were required to detect a 15% absolute reduction in severe hypoxaemia (assumed 25% in the HFNC group, to 10% with NIV) with 95% power at the 5% (two-sided) significance level; primary analysis planned as intention-to-treat (with adjudicated outcome availability).
• Follow-Up Period: In-procedure outcomes from induction to 5 min after intubation; additional outcomes during ICU stay; mortality to day 28.

Key Results

This trial was not stopped early. Recruitment ran from Feb 8, 2016, to May 6, 2017; 322 were randomised and 313 analysed (NIV n=142; HFNC n=171).

• Overall, severe hypoxaemia occurred in 23% (NIV) vs 27% (HFNC) with no statistically significant difference (RD -4.2%; 95% CI -13.7 to 5.5; P=0.39).
• NIV produced numerically higher end-preoxygenation SpO₂ (97% vs 96%) and higher lowest SpO₂ during the procedure (87% vs 84%), without statistical significance in the overall cohort.
• In the pre-specified PaO₂/FiO₂ ≤200 subgroup, severe hypoxaemia was lower with NIV (24% vs 35%), with borderline unadjusted significance (P=0.0553) and statistically significant adjusted analysis (P=0.0459).

Internal Validity

• Randomisation and Allocation: Concealed, web-based randomisation; block size 4 (unknown to investigators); stratified by centre and PaO₂/FiO₂ category (≤200 vs >200 mm Hg), supporting protection against selection bias.
• Drop out or exclusions: 322 randomised; 9 excluded from analysis (NIV: 5; HFNC: 4) due to not intubated (n=1), legal protection (n=1), withdrawal of consent (n=1), and no recorded data (n=5); analysed population 313.
• Performance/Detection Bias: Open-label delivery could influence co-interventions; primary endpoint (SpO₂ threshold) is objective and was adjudicated by a blinded committee, mitigating detection bias.
• Protocol Adherence: Assigned technique duration was similar (both groups: mean 5 min; SD 2 [NIV] vs SD 4 [HFNC]; P=0.45); NIV settings achieved mean pressure support 9 cm H₂O (SD 4), PEEP 5 cm H₂O (SD 0.5), FiO₂ 0.99 (SD 0.06), expired tidal volume 8.3 mL/kg (SD 2.6); HFNC mean flow 58 L/min (SD 9), FiO₂ 0.99 (SD 0.08).
• Baseline Characteristics: Groups were broadly comparable (eg, PaO₂/FiO₂ 142 [SD 65] vs 148 [SD 70]; SAPS II 52 [20] vs 51 [19]); stratification imbalance remained (PaO₂/FiO₂ >200: 18% vs 27%; P=0.06), which is methodologically important given effect heterogeneity by severity.
• Heterogeneity: Multicentre ICU design improves robustness; clinical heterogeneity in AHRF aetiology and operator experience is inherent and clinically relevant; stratification by oxygenation severity and pre-specified subgroup analyses partially address heterogeneity.
• Timing: Median time from ICU admission to randomisation was 1 day (IQR 0–2), indicating enrolment early in ICU course; intervention delivered immediately before intubation as intended.
• Dose: NIV PEEP was protocolised at 5 cm H₂O, which may be submaximal for some patients with severe shunt physiology; HFNC delivered near-protocol flow (mean 58 L/min).
• Separation of the Variable of Interest: Distinct and quantifiable separation was achieved (NIV: pressure support 9/PEEP 5 with tidal volumes 8.3 mL/kg; HFNC: flow 58 L/min; FiO₂ ~0.99 in both), with the intended mechanistic differences (ventilation/PEEP vs apnoeic oxygenation continuity) plausibly preserved.
• Key Delivery Aspects: Use of a standardised intubation bundle (two operators, rapid sequence induction, rescue algorithm) supports internal validity by reducing practice variation around the airway procedure.
• Crossover: Minimal; one HFNC-assigned patient did not receive allocated treatment (per trial profile); protocol deviations were not prominent in reported delivery metrics.
• Outcome Assessment: Primary endpoint is objective but dependent on pulse oximetry (including potential signal lag during rapidly evolving events); the trial used a prespecified severe desaturation threshold (SpO₂ <80%).
• Statistical Rigor: Sample size target met (322 randomised vs 320 planned); primary analysis used absolute risk difference with 95% CI; no multiplicity adjustment was reported for multiple secondary outcomes and subgroup analyses.

Conclusion on Internal Validity: Overall, internal validity appears moderate-to-strong: allocation concealment and protocolised delivery were robust with blinded adjudication of an objective primary endpoint, but open-label management, reliance on pulse oximetry, and post-randomisation exclusions modestly increase bias risk.

External Validity

• Population Representativeness: Adult ICU patients with AHRF (PaO₂/FiO₂ ≤300 mm Hg) requiring intubation; exclusions (GCS <8, cardiac arrest, NIV contraindications, inability to obtain oximetry) limit applicability to some high-risk airway phenotypes.
• Setting and Systems: 28 ICUs in France with access to ICU ventilators for NIV and dedicated HFNC devices; results most applicable to similarly resourced ICUs with staff familiar with an intubation bundle.
• Applicability Across Care Environments: Generalisability to emergency department, prehospital, and resource-limited settings is uncertain due to differences in staffing, monitoring, device availability, and procedural standardisation.
• Subpopulation Applicability: Findings likely most relevant to patients with moderate-to-severe hypoxaemia (PaO₂/FiO₂ ≤200) where physiological reserve is lowest; the mild hypoxaemia subgroup was small (NIV n=25; HFNC n=46).

Conclusion on External Validity: External validity is moderate for adult ICU AHRF populations in similar health systems, and limited for settings without advanced monitoring/equipment or for excluded airway phenotypes (eg, coma, arrest, NIV contraindication).

Strengths & Limitations

• Strengths: Multicentre randomised design; concealed, stratified randomisation; pragmatic ICU population; protocolised intubation bundle; blinded adjudication of an objective primary endpoint; pre-specified oxygenation-severity strata (PaO₂/FiO₂ ≤200 vs >200).
• Limitations: Open-label delivery with potential co-intervention differences; primary endpoint uses a severe desaturation threshold (SpO₂ <80%) that may not capture clinically important lesser hypoxaemia; pulse oximetry lag and artefact are intrinsic measurement constraints; fixed NIV PEEP (5 cm H₂O) may underdose PEEP for some severe AHRF; post-randomisation exclusions and unequal analysed group sizes; not powered for mortality or rare safety outcomes.

Interpretation & Why It Matters

• Clinical signal in overall cohort

  NIV did not significantly reduce severe hypoxaemia versus HFNC with apnoeic oxygenation (23% vs 27%), supporting HFNC as a reasonable default strategy when NIV is impractical or poorly tolerated.
• Severity-dependent effect

  The pre-specified PaO₂/FiO₂ ≤200 subgroup showed a clinically meaningful reduction in severe hypoxaemia (24% vs 35%) with NIV, aligning with the concept that ventilatory support and PEEP matter most when oxygen reserve is critically low.
• Practical bedside relevance

  FLORALI-2 reframes preoxygenation as a physiology-targeted choice rather than a single universal approach: NIV may be preferred in more severe AHRF, whereas HFNC offers procedural simplicity and uninterrupted apnoeic oxygenation during laryngoscopy.

Controversies & Subsequent Evidence

• Endpoint selection: The primary endpoint (SpO₂ <80% for ≥5 s) focuses on profound desaturation; alternative thresholds (eg, <90%) might yield different sensitivity to smaller but still clinically meaningful oxygenation differences.
• Physiology and “what was really tested”: NIV necessarily requires mask removal for laryngoscopy; HFNC remains in place throughout laryngoscopy, so the comparison is not simply “NIV vs HFNC” but “ventilatory/PEEP support before laryngoscopy vs uninterrupted apnoeic oxygenation during laryngoscopy”.
• Subgroup interpretation: Oxygenation-severity strata were pre-specified; however, the borderline unadjusted subgroup P value (P=0.0553) and the reliance on an adjusted analysis (P=0.0459) should be interpreted alongside multiplicity and potential residual confounding despite randomisation.
• Correspondence: Post-publication debate focused on whether HFNC is adequate (or potentially advantageous) in less severe hypoxaemia and how far to generalise subgroup findings beyond the trial’s ICU population.⁴⁵
• Direction of later evidence: Subsequent randomised trials and meta-analyses (including comparisons against standard oxygen) have continued to evaluate NIV, HFNC, and hybrid approaches (eg, NIV for preoxygenation plus apnoeic oxygenation) as strategies to minimise peri-intubation hypoxaemia; guideline recommendations increasingly reflect oxygenation severity and physiological difficulty rather than device dogma.

Summary

• In ICU adults with AHRF requiring intubation, NIV preoxygenation did not significantly reduce severe hypoxaemia versus HFNC with apnoeic oxygenation (23% vs 27%; P=0.39).
• NIV produced numerically higher oxygen saturations at the end of preoxygenation and at the lowest point during the procedure (97% vs 96%; 87% vs 84%), without statistically significant overall differences.
• In the pre-specified PaO₂/FiO₂ ≤200 subgroup, severe hypoxaemia was lower with NIV (24% vs 35%), with borderline unadjusted and significant adjusted analyses.
• Major immediate complications (eg, hypotension) were common and similar; the trial was not powered for rare events or mortality differences.
• The trial supports a physiology-targeted approach to preoxygenation choice, balancing ventilatory/PEEP benefits (NIV) against procedural simplicity and continuous apnoeic oxygenation (HFNC).

Overall Takeaway

FLORALI-2 is a landmark ICU airway trial because it directly tested two widely used, mechanistically distinct preoxygenation strategies in acute hypoxaemic respiratory failure using a pragmatic, bundle-based approach. It found no overall reduction in profound desaturation with NIV compared with HFNC plus apnoeic oxygenation, while providing a coherent severity-stratified signal that NIV may offer greater protection in more severe hypoxaemia—reinforcing physiology-first decision-making at the bedside.

Bibliography

• 1Baillard C, Prat G, Jung B, et al. Effect of noninvasive ventilation for preoxygenation before intubation on oxygenation and on occurrence of severe hypoxemia in critically ill patients: a randomized and controlled study. Am J Respir Crit Care Med. 2006;174:171-177.
• 2Frat JP, Thille AW, Mercat A, et al. High-flow oxygen through nasal cannula in acute hypoxemic respiratory failure. N Engl J Med. 2015;372:2185-2196.
• 3Vourc'h M, Asfar P, Volteau C, et al. High-flow nasal cannula oxygen during endotracheal intubation in hypoxemic patients: a randomized controlled clinical trial. Intensive Care Med. 2015;41:1538-1548.
• 4Coudroy R, Frat JP. Preoxygenation before intubation in severe hypoxaemic respiratory failure—a step too far for high-flow nasal cannula? Lancet Respir Med. 2019;7(6):e17-e18.
• 5Frat JP, Ricard JD, Quenot JP, et al. Preoxygenation before intubation in severe hypoxaemic respiratory failure—a step too far for high-flow nasal cannula? Authors’ reply. Lancet Respir Med. 2019;7(6):e18.