Without evidence of benefit, an intervention should not be presumed to be beneficial or safe.

- Rogue Medic

Advanced Airway vs. BVM During CPR – Which is Worse?


The authors wanted to find out what method is worst best for ventilating patients during out-of-hospital treatment of cardiac arrest.

Endotracheal tube?

Supraglottic airway (laryngeal mask airway, laryngeal tube, and esophageal-tracheal twin-lumen airway device)?

BVM (Bag Valve Mask)?

This assumes that ventilations provide some sort of benefit to the patient. There is no evidence to support this myth.

Their endpoint was neurological outcome at one month for all out-of-hospital cardiac arrest patients treated January 1, 2005, to December 31, 2010.

In addition, we postulated that both advanced airway techniques (endotracheal intubation or use of supraglottic airways) would be similarly associated with favorable neurological outcome after OHCA.[1]


Intubation training required more intubations in the operating room than is required in the US (zero to ? – 5 are typically given as the minimum requirement, but there is no standard accepted by all agencies), or in the UK.

Beginning in 2004, endotracheal intubation could be performed by specially trained emergency lifesaving technicians who had completed an additional 62 hours of training sessions and performed 30 supervised successful intubations in operating rooms.24 [1]


What were the results?

Click on images to make them larger.

Everything underlined in red is worse than BVM and statistically significant. The endotracheal tube numbers underlined in blue are only statistically significant until adjusted for confounding variables and only for ROSC (Return Of Spontaneous Circulation).

ROSC is important – if we do not mind causing long term harm in order to get this short term benefit. This is not a trade off that helps our patients.

A set of potential confounders was chosen a priori based on biological plausibility and a priori knowledge. These selected variables included age, sex, cause of cardiac arrest, first documented rhythm, witnessed status, type of bystander CPR, use of a public access automated external defibrillator, epinephrine administration, and time intervals from receipt of call to CPR by EMS and from receipt of call to hospital arrival.[1]


Not adjusting for confounders would be wrong.

There is one big confounder that I will get to at the end.

The supraglottic airways did even worse than the endotracheal tubes. There has been research in pigs showing that the inflation of the cuffs of supraglottic airways cause more interruption of carotid circulation than the cuffs of endotracheal tubes and that the cuffs of endotracheal tubes cause more interruption of carotid circulation than BVMs. BVMs have no cuffs to inflate to limit carotid circulation, which is the main source of blood flow to the brain.[2]

What else does not have a cuff?

A nonrebreather mask doesn’t have a cuff.

A nasal cannula doesn’t have a cuff.


Did failed tubes cause the study to be biased against intubation?

However, we defined advanced airway management as successful endotracheal intubation or supraglottic airway placement only. Thus, in our study, failed advanced airway management cases reverted to and were classified as bag-valve-mask ventilation cases. This would have biased our conclusions toward the null.[1]


The authors also calculated the results if all of the missing BVM patient data were negative and all of the missing supraglottic airway data were positive. The supraglottic airway data were associated with so much harm, that even this did not make the results look much less harmful for use of supraglottic airways.

Likewise, the authors calculated the results if all of the missing BVM patient data were negative and all of the missing endotracheal tube data were positive. The endotracheal tube data were also associated with so much harm, that even this did not make the results look much less harmful for endotracheal intubation.

Only 6.5% of the patients had endotracheal intubation, so this seems as if we could be dealing with small numbers producing statistical flukes. However, 6.5% of this huge sample is still 41,972. Therefore, this is not a case of statistical manipulation of small numbers with no clinical significance. There are more intubated patients in this study than there are total patients in most studies of airway management.

Assuming the validity of our study, a more secure airway, regardless of its technique, would be detrimental.[1]


Should we assume the validity of this study?

In addition, multiple studies arrived at similar conclusions despite differing populations, disease groups, and designs.7-10,12,13 [1]


There are two problems.


There is no group with no ventilations. I realize that this was not within the control of the authors, but it would answer a more important question.

We are asking –

What is the best method of ventilation during cardiac arrest?

We have not yet answered –

Should we be ventilating during cardiac arrest?

We have not determined that any ventilation is beneficial during cardiac arrest, but we are spending our time choosing colors and arguing over which brand name is best.


14 minutes to ROSC vs. 6 minutes to ROSC?

This is only mentioned tangentially and I am not satisfied with the explanation.

This presents a lot of different data. They all are similar – right up until the very last one.


This deserves some very specific description.

Unfortunately, the data probably do not include enough information to learn more about these differences.

The difference in time to ROSC is huge.

The IQRs (InterQuartile Ranges) show much more diversity than we should expect when everything else is so similar.

Could this explain the difference in outcomes?

The explanation of the authors (that their subgroup analyses account for this) seems to be unreasonably optimistic.

I want the authors to show that patients who have ROSC in 6 minutes should not be expected to have an outcome that is three times better than patients who take 2 13 times as long to achieve ROSC.

These numbers are large huge. 281,522 in one group and 367,837 in the other. How can such a heterogeneous variable be given so little attention?

If I tell someone that two groups of patients were resuscitated, but it took over twice as long to resuscitate one group as it took for the patients in another group, should I expect anyone to be surprised that the group resuscitated faster has outcomes that are three times better?

In cardiac arrest, as time increases bad outcomes can be expected to increase at an even greater rate.

There is not even much overlap in the IQRs of the times to ROSC. 8 – 20 minutes vs. 3 – 12. A 12 minute IQR with only 4 minutes of overlap.

Although everything else appears to be well matched, I do not see the subgroup analyses reconciling the differences in ROSC times.

I think that this could adequately explain the differences in outcomes. We do need prospective studies, but we should find out if there is any reason to be providing ventilations in cardiac arrest (except for those that are pediatric and/or respiratory in origin).

See also Advanced Airway Loses to BVM and read the comments.


[1] Association of prehospital advanced airway management with neurologic outcome and survival in patients with out-of-hospital cardiac arrest.
Hasegawa K, Hiraide A, Chang Y, Brown DF.
JAMA. 2013 Jan 16;309(3):257-66. doi: 10.1001/jama.2012.187612.
PMID: 23321764 [PubMed – indexed for MEDLINE]

[2] Impairment of carotid artery blood flow by supraglottic airway use in a swine model of cardiac arrest.
Segal N, Yannopoulos D, Mahoney BD, Frascone RJ, Matsuura T, Cowles CG, McKnite SH, Chase DG.
Resuscitation. 2012 Aug;83(8):1025-30. doi: 10.1016/j.resuscitation.2012.03.025. Epub 2012 Mar 28.
PMID: 22465807 [PubMed – in process]

The use of 3 different SGDs (Supraglottic airway devices) during CPR significantly decreased CBF (carotid blood flow ) in a porcine model of cardiac arrest. While the current study is limited to pigs, the findings suggest that further research on the effects of SGD use in humans and the effects on carotid artery blood flow is warranted.

Hasegawa K, Hiraide A, Chang Y, & Brown DF (2013). Association of prehospital advanced airway management with neurologic outcome and survival in patients with out-of-hospital cardiac arrest. JAMA : the journal of the American Medical Association, 309 (3), 257-66 PMID: 23321764

Segal, N., Yannopoulos, D., Mahoney, B., Frascone, R., Matsuura, T., Cowles, C., McKnite, S., & Chase, D. (2012). Impairment of carotid artery blood flow by supraglottic airway use in a swine model of cardiac arrest Resuscitation, 83 (8), 1025-1030 DOI: 10.1016/j.resuscitation.2012.03.025



  1. THANK YOU! I’m working on a podcast diving into the minutia of this paper right now, but your final point is the big issue that folks seem to be paying lip-service to but still somehow glossing over completely.

    It’s utterly crazy how people are conflating correlation with causation when they look at this study. I want ETI-during-CPR to die as much as anyone else, and happen to have a personal bias towards the possibility of someday just using airway adjuncts +/- a NRB with no ventilation, but this study is not an ETI killer. It makes perfect sense that codes that last longer are more likely to receive intubation, and we know that mere seconds matter when it comes to duration of CPR and outcome, so why are we surprised that ETI is being associated with worse outcomes since there’s always going to be that confounding variable of “CPR duration” in a retrospective study of this design.

    I can’t believe this is being referred to as “landmark research” by some. Everyone loves to cite an evidence basis for their practice, but we have to be scientific about this stuff and can’t use nonsense findings just because they support our biases. It is clear that this paper neither supports nor refutes the use of ETI during cardiac arrest, and as much as I would love to do more with a giant data set like this, we have to leave it at that.

    Also, did you notice that Table 1 and Table 4 address list all the same baseline characteristics so that the unadjusted and propensity-matched cohorts can be compared, except one finding was left off of Table 4… “time from CPR by emergency responder to ROSC.” How could they leave out the one variable I really care about when describing their propensity-matched data? I’ve been trying to understand why they might have left it out on purpose and can’t really find a reason. Weird.

    • “It makes perfect sense that codes that last longer are more likely to receive intubation”

      Let’s turn that around, shall we?

      ‘It makes perfect sense that codes that receive intubation are more likely to last longer’

      And finish with this.

      “It’s utterly crazy how people are conflating correlation with causation when they look at this study.”

      • Which is exactly my point… For one of those two reasons there’s always going to be an association with harm when you perform a retrospective, non-interventional, observational study like this. Everyone seems to be assuming that the ETI is the cause, so I’m merely arguing that maybe it’s just the result – the key is that we don’t know either way. All I want to do is balance the ship and force people to approach this data analytically.

        I happen to believe that ETI is entirely useless in cardiac arrest and probably causes harm because it interrupts CPR and leaves you with the unimpeded ability to over-ventilate patients. I also believe that if you’re going to use science to back up your statements, you damn well better use it correctly. While I side with everyone who wants to eliminate ETI from CPR from a practical standpoint, if you want to use evidence to support that position, this is not a valid study on which to stand.

  2. We should not mix apples and oranges. In this case, they compare a device used for one thing to a device used for a different thing, A BVM is a ventilatory device and ETT’s and SGA’s are airways not ventilation devices. A BVM is not even a good ventilatory device; it is a squeeze toy evolved from a fireplace bellows. Unconvinced this is real science.


  1. […] Be sure to also go read Tim’s blog post that we discussed during the show over at RogueMedic […]

  2. […] We also discuss what I wrote in Advanced Airway vs. BVM During CPR – Which is Worse? […]