(Last Updated On: August 6, 2018)

Preoxygenation With Flush Rate Oxygen: Comparing the Nonrebreather Mask With the Bag-Valve Mask

Driver, Brian E et al. Annals of Emergency Medicine, March 2018. [paper]

Why I chose this study

While preoxygenating during rapid sequence intubation for spontaneously breathing patients, I’ve typically used bag valve mask (BVM) and nonrebreather mask (NRB) interchangeably. This is mainly based on the path of least resistance with proximity of supplies and respiratory staff experience. Anecdotally, I have used NRB more than BVM. I’ve been interested in how these two devices compare with varying flow rates and practices.


Both BVM and NRB are utilized for preoxygenation during rapid sequence intubation to denitrogenate and create an oxygen reservoir, subsequently reducing the risk of hypoxemia. At 15 L/min flow, NRB can’t generate a reservoir as well as BVM can. If the NRB is at flush rate (>40 L/min) it achieves similar fractional expired oxygen (FeO2) as BVM at 15 L/min (at least in healthy volunteers). This is of course assuming a the ideal circumstance where a perfect seal is obtained and there is synchrony with the patient’s spontaneous respirations. The caveat is that in a real world situation, hopefully one can allocate cognitive and physical resources (i.e. free hands, personnel) to achieve this ideal set up.

Barriers to BVM are manifold: in situ high flow nasal cannula or other devices on the face that we forget to move out of the way, poor seal due to facial characteristics, lack of extra hands for two-person BVM ventilation, high stress/emergent circumstances, etc. On a given intubation event, I can route anywhere from 15-99% of my brain to simply not panicking — this may sometimes leave little else to the remainder of the resuscitation! Additionally, many of our patients weigh 2 or more Chicago Units (CU; 1 CU ≅ 100 kg), have substantial beard growth potential, or may have facial abnormalities from trauma, adding to the challenge of a good seal with BVM. So, is it worth it to use BVM, all this considered? Or am I not much worse off using NRB at flush rate, patient permitting?

Research Question

Is NRB mask with flush rate oxygen flow noninferior to flush rate BVM preoxygenation (assisted, unassisted, with or w/o air leak)?

Study Design

Two parallel cross-over studies at Hennepin County Medical Center Emergency Department (ED).

Inclusion criteria

“Healthy” ED volunteer personnel >17 years of age.

Exclusion criteria

Facial hair more than stubble present, pregnancy, significant smoking history, or symptomatic respiratory disease at time of study.


  • Subjects underwent multiple trials of preoxygenation at random sequence (order effects mitigated using a balanced Latin Square design).
  • All preoxygenation done at flush rate (O2 turned all the way up, ends up delivering 40-60 L/min).
  • BVMs had one-way exhalation valves to prevent inhalation of room air.
  • No formal screening for medical comorbidities.
  • Head of bed at 30 degrees.
  • Reservoir bags of both NRB and BVM filled x 15 seconds prior to administration.
  • 3 minute preoxygenation periods.
  • 2 minute washout periods with verification of return to subject’s baseline FeO2 prior to next trial.

Study 1 composed of 3 trials: (both subject and investigator could maintain BVM seal, no assisted BVM breaths delivered)

  • NRB
  • BVM
  • BVM with a simulated leak (16 French nasogastric tubes x2 taped above lip)

Study 2 composed of 2 trials: (only investigator maintaining BVM seal while assisting)

  • NRB
  • BVM with inspiratory assistance


Primary outcome: FeO2 at end of each preoxygenation trial (quantifies denitrogenation), as measured by complete exhalation into gas analyzer.

Study 1, primary comparison: whether NRB with flush rate oxygen was noninferior to BVM at flush rate.

Study 2, primary comparison: whether NRB with flush rate oxygen was noninferior to BVM at flush rate with assistance.


Noninferiority margin was set at 10%, with rationale being a 10% absolute difference in FeO2 would provide ~1 additional minute of safe apnea time in a normal adult with average lung volumes, and this would be deemed a clinically significant difference.

Investigators estimated needing 26 subjects per study to have 80% power for a test of noninferiority with an absolute difference of 10% at a significance of .05.

Values expressed as FeO2 with 95% CI.


Study 1: 30 subjects enrolled, Mean age 23, SD 4 years, 12 men, Mean BMI 18 kg/m², SD 3 kg/m².

  • At a difference of 5%, FeO2 after NRB was noninferior to BVM (95% CI -1%-10%).
  • At a difference of 51%, FeO2 after NRB was higher than BVM with a leak (95% CI 46%-55%).

Study 2: 27 subjects enrolled, Mean age 23, SD 5 years, 13 men, Mean BMI 24 kg/m², SD 4 kg/m².

  • At a difference of 6%, FeO2 after NRB was noninferior to BVM with bag squeezed in synchrony with inhalation (95% CI 3%-10%).

Copyright: Tasas Cerniasukas


In these two cross-over studies of “healthy” young volunteers, preoxygenation with flush rate NRB was noninferior to flush rate BVM assisted and unassisted. Additionally, FeO2 was higher with flush rate NRB as compared to flush rate BVM with air leak. The authors conclude that flush rate NRB is a reasonable default preoxygenation method in spontaneously breathing patients not requiring positive pressure.

Some limitations to the studies:

  • Study conditions included cooperative, “healthy” participants therefore it is difficult to generalize to critically ill patients in the ED.
  • In severe dyspnea, the max inspiratory flow rate can exceed the O2 flow rate, which decreases final FiO2, and it’s not clear what effect this might have on study outcomes.
  • Participants and/or investigators in study 1 were allowed to fix any perceived air leaks.
  • No screening for medical comorbidities.
  • BVMs outfitted with one-way exhalation valves — BVMs without this will have increased room air on inspiration.
  • This is a noninferiority trial, not an equivalence or superiority trial.

NRB isn’t better than BVM, but it’s not worse by the authors’ preset standards. This paper makes me feel better about initially reaching for NRB at flush rates for preoxygenation. The finding of a markedly decreased FeO2 with a simulated leak gives me pause. It isn’t a primary endpoint, and remember this trial is not powered for superiority, but if I assume there will be a significant leak with BVM I am going to insist on the NRB. I’m guessing that’s why it was included in the paper — it bolsters the case for the overall utility of NRB in addition to their noninferiority findings.

Consequently, the threshold in my mind in order to use BVM over NRB is now a bit higher after reading this paper. If a patient is amenable to NRB (no shunt physiology, hypoventilation, no upper airway obstruction, need for positive pressure), I’ll favor it for its simplicity, low maintenance, and cognitive offloading ability. There’s also something to be said for not having to worry about high peak airway pressures using NRB and theoretically less risk of gastric insufflation or aspiration. Taking the noninferiority margin to its boundary, a loss of 10% FeO2 or ~1 minute before desaturation should hopefully be balanced or even outweighed with apneic oxygenation by using a nasal cannula.

What are your thoughts?

How much does this affect your practice? Any die-hard BVM or NRB fans out there? What flow rates can we achieve within the settings we work (ICU, ED sites, etc)? Do our BVMs come with one-way exhalation ports?

With some higher mean FeO2s in the NRB groups of this study, might there be utility in devising a trial of equivalence, or even superiority of NRB over BVM? At least in the same demographic of healthy, cooperative volunteers…


Sal Calo is a EM/IM Resident, Class of 2020




EM/IM Sessions are reviewed in journal club style by the current attendings and residents, as well as alumni of the UIC IM/EM program prior to publication. This post was specifically reviewed by Wes Eilbert, MD Associate Professor in Emergency Medicine. Elspeth Pearce, MD Editor.