HIGH

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

Azoulay E et al. JAMA October 2018; DOI:10.1001/jama.2018.14282

Clinical Question

  • In immunocompromised patients with acute hypoxaemic respiratory failure (AHRF), does the use of high-flow oxygen compared with standard oxygen therapy, reduce 28-day mortality?

Background

  • Early small randomised trials (1, 2) suggested that the use of non-invasive ventilation (NIV) may reduce the need for endotracheal intubation and improve outcomes in immunocompromised patients with AHRF
  • In a larger more recent trial, NIV did not improve outcomes vs standard oxygen. However, the trial was underpowered for its primary outcome given the lower than expected mortality in the control arm
  • The FLORALI trial found improved ICU and 90-day mortality with the use of high-flow nasal oxygen compared to other oxygen delivery methods in a broadly defined group of patients with AHRF
  • Whether high-flow nasal oxygen improves outcomes specifically for immunocompromised patients with AHRF remains unclear

Design

  • Multicentre randomised parallel-group trial
  • Non-blinded given the nature of the intervention
  • 1:1 assignment
  • Block randomisation
  • Allocation concealment maintained through the use of an electronic system
  • Stratification according to study centre, oxygen flow rate at randomisation, need for vasopressors, and time since ICU admission
  • Power calculation
    • Based on an expected 28-day mortality of 30% in the standard oxygen arm and 20% in the high-flow oxygen arm
    • 779 patients targeted to provide the trial with a 90% power to detect a decrease in 28-day mortality (alpha set at 5%)
  • Intention to treat analysis
  • The protocol was initially designed as a non-inferiority trial with different secondary endpoints. At the request of a grant review jury (in part in reaction to the results of the FLORALI study), the protocol was modified to a superiority trial. Both the initial and the revised protocol are provided in the published data supplement.

Setting

  • 32 French hospitals (24 university-affiliated and 8 non-university-affiliated)
  • Data collected: May 2016 to December 2017

Population

  • Inclusion criteria: Immunocompromised adult patients with AHRF
    • 18 years of age or older
    • AHRF defined as any of the following: PaO2 < 60 mmHg, SpO2 < 90% on room air, respiratory rate (RR) > 30 breaths/minute, labored breathing, respiratory distress
    • Need for oxygen flow ≥ 6 L/min
    • Known immunosuppression defined as any of the following: use of ≥ 3 months of > 0.5 mg/kg/day steroids, use of other immunosuppressant drugs, solid organ transplantation, solid tumour requiring chemotherapy in the last 5 years, haematologic malignancy regardless of the time of diagnosis, primary immunodeficiency
  • Exclusion criteria:
    • AIDS
    • Imminent death
    • Anatomical factors precluding the use of nasal cannula
    • Hypercapnia (PaCO2 ≥ 50 mmHg)
    • Isolated cardiogenic pulmonary edema
    • Pregnancy or breastfeeding
    • Surgery within the last 6 days
    • Absence of coverage by the French statutory health care insurance system
  • 778 patients randomised, 2 patients withdrew consent
  • Baseline characteristics were well matched
    • Median age 64 years
    • 33.3% women
    • 78.9% with underlying cancer (56.8% of which were a haematologic malignancy)
    • 57.9% receiving chemotherapy at the time of ICU admission
    • 14.9% with poor performance status
    • 63.8% enrolled on the day of ICU admission; 16% enrolled >=2 days after ICU admission
    • Median PaO2/FiO2 < 150 in both groups immediately prior to randomisation
    • Median SOFA score at randomisation 6
    • 11% received high-flow nasal oxygen prior to randomisation
    • Cause of AHRF
      • 41% bacterial pneumonia
      • 12% invasive fungal infection
      • 10% lung involvement from underlying disease

Intervention

  • Continuous high-flow nasal oxygen (n=389)
    • initiated at 50 L/min and 1.0 FiO2. If SpO2 < 95%, flow increased to 60 L/min. FiO2 subsequently titrated to achieve SpO2 > 95%. The minimal flow rate during the first 3 days was 50 L/min
    • High-flow nasal oxygen was weaned if the patient had improved clinical signs of respiratory distress, PaO2/FiO2 > 300, and could maintain SpO2 > 95% with < 6 L/min of standard O2
      • Patients returned to high-flow nasal oxygen for recurrent hypoxia
    • 376 patients received intervention as randomised
    • 1 patient withdrew consent and 12 patients did not receive treatment with high-flow due to discomfort

Control

  • Normal oxygen delivery (n=389)
    • Oxygen delivered via nasal prong or mask; with or without a reservoir and with or without a Venturi system
      • Flow set to achieve SpO2 > 95%
      • High-flow nasal oxygen allowed only for Do Not Intubate orders in patients for whom standard O2 had failed
    • 358 patients received intervention as randomised
    • 1 patient withdrew consent and 30 patients received high-flow

Management common to both groups

  • NIV only allowed for hypercapnia or pulmonary oedema
  • Standardised criteria for invasive mechanical ventilation
  • Ventilator management followed consensus guidelines
  • All other management decisions were left to the treating clinician

Outcome

  • Primary outcome: 28-day all-cause mortality – no significant difference
    • 35.6% (intervention) vs. 36.1% (control)
    • HR, 0.98 [95% C.I., 0.77 to 1.24], p = 0.94
    • No significant interaction between the intervention effect and any pre-defined subgroup
  • Secondary outcomes (comparing intervention vs. control)
    • No significant difference in:
      • Intubation rate
        • 38.7% vs 43.8%
        • cause-specific HR, 0.85 [95% C.I., 0.68 to 1.06], p = 0.17
      • ICU acquired infections
        • 10.0% vs. 10.6%
        • absolute risk difference -0.6 [95% C.I. -4.6% to 4.1%]
      • ICU length of stay
        • 8 vs. 6 days
        • mean difference 0.6 days [95% C.I. -1.0 to 2.2]
      •  ICU mortality
        • 31.7% vs. 31.4%
        • mean difference 0.3 [95% C.I. -6.3 to 6.8]
      • Hospital length of stay
        • 24 vs. 27 days
        • mean difference -2 days [95% C.I. -7.3 to 3.3]
      • Hospital mortality
        • 41.2% vs. 41.7%
        • mean difference -0.5 [95% C.I. -7.5 to 6.4]
      • Cumulative incidence of mechanical ventilation, patient comfort (assessed using a visual analogic scale) or patient dyspnea (assessed using visual analogic scale)
    • Significantly reduced in the intervention group
      • Respiratory rate after 6 hours
        • 25 breaths/min vs. 26 breaths/min
        • mean difference, -1.8 [95% CI, -3.2 to -0.3]
    • Significantly increased in the intervention group
      • PaO2/FiO2 until day 4
        • 150 vs. 119
        • mean difference, 19.5 [95% CI, 4.4 to 13.6]
  • Post-hoc outcomes
    • No significant difference between groups in need for vasopressors and renal replacement therapy, median time from randomisation to intubation, mortality in intubated patients, decisions to limit treatment, 28-day mortality in patients with and without cancer as the cause of immunosuppression, and 90-day mortality

Authors’ Conclusions

  • High-flow nasal oxygen does not decrease 28-day mortality compared to standard oxygen in immunocompromised patients with AHRF.

Strengths

  • Addresses a clinically relevant question
  • The largest trial to date on the subject
  • Trial conducted at both academic and community hospitals, increasing the generalisability of the results
  • Baseline characteristics were well balanced between the two study arms
  • The number of patients in the high-flow oxygen arm who crossed over to the standard oxygen arm was low (3%)
  • No patients were lost to follow-up during the study period

Weaknesses

  • While multicentre, the trial was performed only in France which may limit generalisability to other practice settings
  • Lack of blinding may have led to differential treatment between groups
  • The minimum allowable SpO2 in either arm was 95%. As the benefits of a conservative oxygenation threshold in the ICU become clear, the aggressive SpO2 target used in this trial seems increasingly out of line with best practice (although the harms of this approach – if any – should have been evenly balanced by randomisation)
  • The study hypothesis and protocol lumps a heterogeneous patient population under the common label of “immunocompromised” and presumes this group shares a common physiologic response to different oxygen delivery devices. This conflation may mask benefits or harm to high-flow oxygen within specific patient sub-groups not expressly highlighted in the trial
  • The study protocol including many of the secondary endpoints was adjusted after initial submission (which the authors acknowledge in the manuscript)
  • 11% of patients received high-flow prior to randomisation and 8% of patients randomised to control group received high flow oxygen

The Bottom Line

  • In a broadly defined group of immunocompromised patients with AHRF, the use of high-flow nasal oxygen did not reduce 28-day mortality compared to standard oxygen
  • This trial should allow clinicians to move away from prescriptive rules for oxygen delivery (e.g., all immunocompromised patients with AHRF must receive high-flow nasal oxygen) and towards individualised care guided by the clinical situation, available resources, and the patient’s response to the chosen therapy
  • Based on the strong physiologic rationale for the use of high-flow nasal oxygen in AHRF, the results of the FLORALI study, and several trials (1, 2) supporting the use of high-flow nasal oxygen to prevent post-extubation respiratory failure, I will continue to view high-flow nasal oxygen as a reasonable front-line therapy for immunocompromised patients with AHRF

External Links

Metadata

Summary author:  James M. Walter @macwalterMD
Summary date: 15th November 2018
Peer-review editor: @davidslessor

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.