JORNAL CLUB

1. Copyright © 2017 International Anesthesia Research Society. Unauthorized reproduction of this article is prohibited.
XXX XXX • Volume XXX • Number XXX www.anesthesia-analgesia.org 1
DOI: 10.1213/ANE.0000000000002341
A
critical component of the safe delivery of emergency
anesthesia is the avoidance of hypoxemia during
the apneic phase of a rapid sequence intubation.1,2
Preoxygenation describes the process of having the patient
breathe 100% oxygen for a brief period, washing out nitrogen
(denitrogenation) and creating an oxygen reservoir in the
functional residual capacity of the lungs to extend the dura-
tion of apnea before desaturation.3,4
Optimal preoxygen-
ation, whether inside or outside the operating theatre may
not always be possible due to facial trauma, edentulous, or
uncooperative patients where a mask leak is unavoidable,
or environmentally determined factors such as the unavail-
ability of specialized anesthetic circuits, limited personnel,
or an austere prehospital working environment.5–8
Recently, investigators have studied whether supple-
mental oxygen supplied via nasal cannula (NC) can
enhance the completeness of preoxygenation achieved with
equipment commonly used for critical care patients in and
out of the operating room such as anesthetic circuits, non-
rebreather masks and bag valve masks (BVM).5,9,10
These
studies suggested that supplemental NC oxygen increased
end-tidal oxygen (ETo2) values for BVM, anesthetic circuit,
and nonrebreather masks when the NC flow rate of 10 liters
per minute (lpm)9
or 5 lpm5
and the mask seal was imper-
fect. When the BVM mask seal was optimal however, 10
lpm did not improve ETo2.9
One possible explanation for
these results is that the NC, when added to the BVM, pro-
duces a leak allowing entrainment of room air that is only
overcome with NC flow rates of 5 to 10 lpm.
Because NC oxygen is also advocated for apneic oxygen-
ation during laryngoscopy and thus likely to be placed on
the patient before induction,2
the combination of NC oxy-
gen and mask oxygenation is common. Clarifying whether
the presence of NC impairs preoxygenation when used with
a BVM and what NC oxygen flow is needed to compensate
would facilitate clinical care. In principle, a wide range of
NC oxygen flow are possible as rates up to 15 lpm are rea-
sonably tolerated.11
Identifying the optimum NC oxygen
flow rate that allows optimal preoxygenation may assist in
preoxygenation in uncooperative patients.
To characterize the effect of supplemental NC oxygen flow
on preoxygenation using a BVM, we performed a random-
ized crossover study on healthy volunteers. Our primary out-
come was to compare the ETo2 after 3-minute preoxygenation
between a well-sealed BVM and BVM + NC of different flow
rates. Our secondary outcomes were the comparison of ETo2
for different flow rates at 1- and 2-minute preoxygenation.
BACKGROUND: A critical safety component of emergency anesthesia is the avoidance of hypox-
emia during the apneic phase of a rapid sequence intubation. Preoxygenation with a bag valve
mask (BVM) or anesthetic circuit may be improved with supplemental oxygen by nasal cannula
(NC) if there is a mask leak. In addition, NC is recommended for apneic oxygenation after induc-
tion and may be placed before preoxygenation. However, the optimum NC flow rate for preoxy-
genation or whether the presence of NC alone creates a mask leak remains unclear.
METHODS: We performed a randomized crossover study on healthy volunteers comparing BVM
alone and BVM with NC flow rates of 0 (NC-0), 5 (NC-5), 10 (NC-10), and 15 (NC-15) liters per
minute (lpm). Our primary outcome was end-tidal oxygen (ETo2) after 3-minute preoxygenation.
RESULTS: There was no difference in ETo2 between NC-15, NC-10, or BVM-only at 3 minutes.
NC-0 and NC-5 recorded significantly lower ETo2 at all times compared with NC-15, NC-10, or
BVM-only (least difference NC-5, −7% [95% confidence interval {CI}, −4% to −10%), NC-0, 16%
[95% CI, 13%–19%]). There was a difference in ETo2 between NC-15 and BVM-only at 1 minute
(7%; 95% CI, 5%–9%), but not at 2 or 3 minutes. There was no difference in ETo2 between NC-10
and NC-15.
CONCLUSIONS: Our study found that NC at 0 and 5 lpm with a BVM is deleterious to preoxy-
genation and should be avoided. In addition, a lack of difference between NC-10 and BVM-only
demonstrates that NC at flows of at least 10 lpm should not impair the preoxygenation process.
While NC-15 may offer a benefit by reaching maximal ETo2 at 1 minute, this would need to be
balanced against patient comfort. (Anesth Analg 2017;XXX:00–00)
Addition of Nasal Cannula Can Either Impair or
Enhance Preoxygenation With a Bag Valve Mask:
A Randomized Crossover Design Study Comparing
Oxygen Flow Rates
David McQuade, MBChB,* Matthew R. Miller, MBChB,† and Clare Hayes-Bradley, MBBS‡
From the *Wellington Hospital, C&CDHB, Wellington, New Zealand;
†Department of Anesthesia, Toronto Western Hospital, Toronto, Ontario,
Canada; and ‡Greater Sydney Area HEMS, NSW Ambulance Aeromedical
Service, Rozelle, New South Wales, Australia.
Accepted for publication June 8, 2017.
Funding: None.
The authors declare no conflicts of interest.
IRB approval: Sydney Local Health District Ethics Review Committee
(HREC/15/RPAH/587).
Address correspondence to Matthew R. Miller, MBChB, 3402-318 Richmond
St W, Toronto, ON M5V 0B4, Canada. Address e-mail to mattdotmiller@
gmail.com.
Copyright © 2017 International Anesthesia Research Society

2. Copyright © 2017 International Anesthesia Research Society. Unauthorized reproduction of this article is prohibited.
2 www.anesthesia-analgesia.org ANESTHESIA ANALGESIA
Nasal Cannula Flow Rate and Preoxygenation
METHODS
Study Design
We performed a randomized crossover study using healthy
volunteers. Participants were recruited from all persons
aged over 18 years attending clinical governance days at our
institution. Exclusion criteria were pregnancy, known lung
disease, suspected coronary or cerebrovascular disease, pre-
vious exposure to bleomycin or amiodarone, and beards,
moustaches, dentures, or facial abnormalities that might
affect mask seal. The study received ethics approval from
the Sydney Local Health District Ethics Review Committee
(HREC/15/RPAH/587) and has been registered as a clini-
cal trial (ACTRN12617000337370). Written informed con-
sent was obtained from all participants. This manuscript
adheres to the EQUATOR guidelines.
All participants completed the preoxygenation trials
lying supine with a pillow under the head for comfort. Each
preoxygenation period consisted of 3 minutes of tidal vol-
ume breathing with the mask held in place by 1 of 2 inves-
tigators, Attending Physicians in Anesthesia (C.H.B.) and
Emergency Medicine (D.M.; Figure 1). During these 3 min-
utes, we recorded ETo2 via a single expiratory breath into
an in-circuit measuring device at 1, 2, and 3 minutes. A rest
period of 2 to 3 minutes between trials allowed ETo2 values
to return to baseline before commencing the next trial and
was confirmed by measuring the ETo2 so it was within 2%
of each participant’s baseline. The sequence of the trials was
randomized in a balanced 5 treatment 5 period crossover
design, so that another participant completing the trials in
an opposite order would balance a participant completing
the trials in 1 order. This sequence minimized any potential
residual effect of the order of trials. To achieve this goal, a
randomized table of 20 trial sequences was generated by 1
investigator (M.R.M.) using the statistical software R (ver-
sion 3.1.2; R Foundation for Statistical Computing, Vienna,
Austria)12
and these sequences were then reversed for the
remaining 20 trials. The order in which these trials were
performed was again randomized using R software. The
investigators that conducted the trials were not blinded to
the trial sequences. The 5 preoxygenation trials were BVM-
only, BVM and NC with no oxygen flow (NC-0), BVM and
NC at 5 lpm (NC-5), BVM and NC at 10 lpm (NC-10), and
BVM and NC at 15 lpm (NC-15; Figure 2).
The BVM used was a disposable self-inflating resusci-
tator with a 2-L reservoir bag (Mayo Healthcare, Mascot,
NSW, Australia)13
and expiratory cap connected to a heat
moisture exchange filter and catheter mount at a flow rate
of 15 L/min (Figure 2). Single use adult straight-prong NC
(Mayo Healthcare, Mascot, NSW, Australia) were used for
all participants with the flow rate varied for each trial.
The ETo2 was measured with a Datex Capnomac Ultima
side-stream gas analyzer (Datex-Engstrom, Helsinki, Finland).
This device was serviced before commencing the first trial
and had an accuracy of ±1%. The measurements were per-
formed by having the participants exhale a controlled forced
expiratory breath into the tight fitting mask over 4 seconds to
achieve an alveolar plateau on the monitor.
Statistical Analysis
Analysis was performed using the statistical software R
(version 3.1.2, R Foundation for Statistical Computing).12
ETo2 was analyzed using type-III 2-way repeated measures
analysis of variance (ANOVA), with preoxygenation trial
and time of ETo2 sampling as the within-subjects variables
and ETo2 as the dependent variable. Both the main effects of
each variable and interaction between trial and time were
tested. Sphericity was examined with Mauchly’s test, and if
these were violated, Greenhouse-Geisser correction to P val-
ues was used. Our primary and secondary outcomes were
whether a main effect of NC flow rate existed, and also an
interaction effect between NC flow rate and time. Planned
post hoc tests were pairwise comparisons of NC flow rate
versus time if a significant effect was found for the interac-
tion, or fixed effects for NC flow rate if only a main effect
Figure 1. Photo of the bag valve mask
circuit with end-tidal oxygen sampling
line and potential dead spaces as
indicated.

3. Copyright © 2017 International Anesthesia Research Society. Unauthorized reproduction of this article is prohibited.
XXX XXX • Volume XXX • Number XXX www.anesthesia-analgesia.org 3
was found. This analysis was performed using the R pack-
age post hoc interaction analysis,14
with Bonferroni correc-
tion to P values to account for multiple comparisons (these
are applied within the statistical test). Confidence intervals
(CIs) for the ANOVA effect sizes were then calculated with
the method described by Hollands and Jarmasz15
which
takes into account the degrees of freedom and number of
levels of variables to account for multiple comparisons. A
2-sided P value .05 was considered significant, and was
used for all tests.
A sample size calculation was performed using G*Power
316
(Version 3.1.9.2, University of Dusseldorf) for repeated
measures ANOVA before beginning the study. The effect
size we chose was initially 0.35 (converted from Cohen’s
d from a mean difference of 5% and a standard deviation
of 7%); however, we settled on using a more conservative
medium effect size (f = 0.25)17
across 5 trials and 3 measure-
ments and this determined 30 participants were required.
We chose a 5% difference in ETo2 as clinically relevant as it
would equate to 30 seconds of additional safe apnea time
([5% × 2400 mL]/250 mL/min oxygen consumption) in an
80 kg male and is consistent with previous research.3,9
We
aimed to recruit 40 participants to allow for dropout.
RESULTS
Overall 40 participants completed the trial. One ETo2 mea-
surement was missing leaving 39 with complete data for
analysis (29 males and 10 females). The mean age of par-
ticipants was 41 (standard deviation = 9), with a mean body
mass index of 26 (standard deviation = 3).
We found that both the nasal prong flow rate (P .001),
and time primarily affected the efficacy of preoxygenation
(P .001). In addition, we found an interaction between
nasal prong flow rate and time (P .001). Table 1 and
Figure 3 present ETo2 results for all NC flow rates tested
at each time period. For our primary outcome, pairwise
comparisons revealed no difference in ETo2 between NC-15,
NC-10, or BVM-only groups at 3 minutes (Figure 3, column
3). The ETo2 for NC-0 at 3 minutes was lower than all other
flow rates, while NC-5 was greater than NC-0 (9%, 95% CI,
8%–12%; P = .001) but lower than BVM-only (−11%, 95% CI,
−8% to −14%; P .001), NC-10 (−9%, 95% CI, −8% to −12%;
P .001) and NC-15 (−12%, 95% CI, −9% to −15%; P .001).
For the secondary outcomes (difference in ETo2 at 1 and
2 minutes), pairwise comparisons showed NC-0 and NC-5
recorded the lowest ETo2 at all times (Figure 3 and Table 2).
There was no difference in ETo2 between NC-10 and NC-15
at all measurements (Table 2). There was a difference in
ETo2 between NC-15 and BVM-only at 1 minute (7%; 95%
CI, 5%–9%; P = .009).
DISCUSSION
In this volunteer study, we found that increasing rates of
NC oxygen flow during preoxygenation leads to sequen-
tially higher ETo2, but that even 15 lpm NC oxygen flow did
not improve ETo2 compared to a tightly fitted BVM face-
mask at 3 minutes.
Our results are consistent with existing literature. When
compared to preoxygenation with BVM-only, NC-0 and
NC-5 were less effective, a finding similar to Groombridge et
al.10
One possible explanation for this observation is that the
concurrent presence of NC introduces a leak to the intended
tight fit of a BVM facemask. This leak allows entrainment
of room air, thereby diluting the fraction of inspired oxygen
(Fio2). Taken together, current evidence thus suggests that
concurrent NC with no flow or a flow rate of 5 lpm may
impair the efficacy of preoxygenation with a BVM.
The ETo2 achieved with BVM and NC-10 is equivalent
to BVM without NC-10. This finding is consistent with our
previous study,9
was observed across all time points in our
study, and suggests that 10 lpm may be sufficient to over-
come the mask leak introduced by NC but not enough to
increase Fio2 overall. In addition, while NC-15 did increase
ETo2 after 1 minute of preoxygenation, the magnitude of
this difference decreased at the 2- and 3-minute time points,
and NC-15, NC-10, and BVM-only were statistically and
clinically equivalent at the 3-minute time point.
Assessed for eligibilty
(n =40)
preoxygenation trials
(order randomised)
BVM-only
BVM+NC at 0 lpm
BVM+NC at 5 lpm
BVM+NC at 10 lpm
BVM+NC at 15 lpm
1 participant missing
ETO2 data
Analysis
(n=39)
Figure 2. Flow chart outlining study design. BVM indicates bag valve
mask; ETo2, end-tidal oxygen; LPM, liters per minute; NC, nasal
cannula.
Table 1. ETo2 (%) at Each Minute for Each
Preoxygenation Condition (Mean [SD])
Condition
Time (Min)
1 2 3
BVM-only 74 (9) 82 (10) 84 (9)
NC-0 59 (10) 64 (9) 66 (10)
NC-5 68 (9) 73 (8) 74 (8)
NC-10 77 (8) 81 (8) 83 (7)
NC-15 82 (6) 85 (6) 86 (6)
Abbreviations: BVM, bag valve mask; ETo2, end-tidal oxygen; lpm, liters per
minute; NC-0, BVM with nasal cannulae at 0 lpm; NC-5, BVM with nasal
cannulae at 5 lpm; NC-10, BVM with nasal cannulae at 10 lpm; NC-15, BVM
with nasal cannulae at 15 lpm; SD, standard deviation.

4. Copyright © 2017 International Anesthesia Research Society. Unauthorized reproduction of this article is prohibited.
4 www.anesthesia-analgesia.org ANESTHESIA ANALGESIA
Nasal Cannula Flow Rate and Preoxygenation
Our findings and those of others have clinical relevance. If
NC is used concurrently with BVM, either for preoxygenation
or apneic diffusion oxygenation, setting the oxygen flow rate
through those cannula at 10 lpm may result in slower and
less complete denitrogenation. In addition, our data suggest
that if an inadvertent leak were introduced, then the NC equal
or greater to 10 lpm may offset the diminished Fio2 that occurs
with the entrainment of room air.9
Also, if NC are to be used for
apneic oxygenation, by placing the NC on the patient before
the preoxygenation phase, they need not be removed during
the time critical apneic period. Our data also raise the possibil-
ity that NC at 15 lpm may be able to achieve the same level
of preoxygenation in 1 minute as BVM-only can at 3 minutes.
Our study has limitations. We used healthy vol-
unteers, which may limit the generalizability of our
findings. For example, the effects we report may not
be reproducible in obese patients or those with respi-
ratory failure. In such populations, our study is thus
only hypothesis generating. In addition, the equipment
tested was the standard used by our service and results
achieved with different equipment may not be the same.
In particular, the gooseneck introduces a deadspace of
up to 40 mL. Our in-line ETo2 sampling method differs
from the single-exhalation technique used in previous
studies9,10
and may theoretically lead to higher abso-
lute ETo2 readings due to continuous oxygen flow con-
taminating the sampling line. To mitigate these effects,
we sought to record the ETo2 at the end of the alveo-
lar plateau, so that a typical tidal volume breath of 500
mL should washout the deadspace of the circuit during
measurement. In addition, this technique is consistent
with previous anesthetic literature18–20
and allowed us
to make serial measurements each minute. The ETo2 lev-
els measured in our study is similar to previous stud-
ies9,10
suggesting this method is reliable. Finally, our
study looks at the period of preoxygenation created by
the participants’ spontaneous respiratory efforts before
the onset of apnea. Oxygenation after apnea but before
laryngoscopy by way of ventilating the patient either
manually21
or mechanically22
remains an option in all
patients but this remains controversial and not part of a
classical rapid sequence induction.
In conclusion, we found that NC at 0 and 5 lpm pre-
vent optimal preoxygenation with a BVM. In addition,
we found no decrement in preoxygenation efficacy
when NC-10 was used concurrently with BVM. Our
findings indicate that oxygen flow 10 lpm should not
impair the preoxygenation process. Further study is
needed to better understand the clinical consequences
of our ­findings. E
50
60
70
80
90
1 minute 2 minutes 3 minutes
Duration of preoxygenation
ETO2(%)
NC flow rate
10LPM
15LPM
5LPM
BVM-only
oLPM
Figure 3. Line plot of ETo2 for each nasal cannula
flow rate at each minute of preoxygenation. Error
bars represent the 95% CI for the mean. BVM
indicates bag valve mask; CI, confidence interval;
ETo2, end-tidal oxygen; LPM, liters per minute; NC,
nasal cannula.
Table 2. ETo2 Difference (%) and 95% CI Between
Preoxygenation Conditions at 1, 2, and 3 Minutes
(Some CI May Be Asymmetric Due to Rounding)
Comparison
Measurement Time
1 Min 2 Min 3 Min
Difference in
Means (95% CI)
Difference in
Means (95% CI)
Difference in
Means (95% CI)
NC-15 to
BVM-only 7 (4–10) 3 (0–6) 2 (−1 to 5)
NC-0 22 (19–25) 20 (17–23) 20 (17–23)
NC-5 14 (11–17) 12 (9–15) 12 (9–15)
NC-10 4 (0–7) 3 (0–6) 3 (0–6)
NC-10 to
BVM-only 2 (−1 to 5) 1 (−2 to 4) 1 (−2 to 4)
NC-0 18 (15–21) 17 (14–20) 17 (14–20)
NC-5
9 (6–12) 8 (5–11) 9 (8–12)
NC-5 to
BVM-only −7 (−4 to –10) −9 (−6 to –12) −11 (−8 to –14)
NC-0 9 (6–12) 9 (6–12) 7 (3–11)
BVM-only to
NC-0 16 (13–19) 17 (14–20) 18 (15–21)
Abbreviations: BVM, bag valve mask; CI, confidence interval; ETo2, end-tidal
oxygen; lpm, liters per minute; NC-0, BVM with nasal cannulae at 0 lpm; NC-5,
BVM with nasal cannulae at 5 lpm; NC-10, BVM with nasal cannulae at 10
lpm; NC-15, BVM with nasal cannulae at 15 lpm.

5. Copyright © 2017 International Anesthesia Research Society. Unauthorized reproduction of this article is prohibited.
XXX XXX • Volume XXX • Number XXX www.anesthesia-analgesia.org 5
DISCLOSURES
Name: David McQuade, MBChB.
Contribution: This author helped design the study, write the proto-
col, liaise with the ethics committee, recruit and perform the mea-
surements on the participants, and contribute to the introduction
and discussion and review the manuscript as a whole.
Name: Matthew R. Miller, MBChB.
Contribution: This author helped design the study, write the pro-
tocol, perform the statistical analysis, write the methods and results
section, and contribute to the introduction and discussion and
review the manuscript as a whole.
Name: Clare Hayes-Bradley, MBBS.
Contribution: This author helped design the study, liaise with the
ethics committee, recruit and perform the measurements on the
participants, and contribute to the introduction and discussion and
review the manuscript as a whole.
This manuscript was handled by: Avery Tung, MD, FCCM.
ACKNOWLEDGMENTS
The authors would like to thank the staff of the Greater
Sydney Area Helicopter Emergency Medical Service as
well as attendees of the clinical governance days for vol-
unteering their time. Assistance was graciously given by
Dr Karel Habig, Dr Brian Burns, and Sandra Ware of NSW
Ambulance Aeromedical Service.
REFERENCES
1. Weingart SD, Levitan RM. Preoxygenation and prevention of
desaturation during emergency airway management. Ann
Emerg Med. 2012;59:165–75.e1.
2. Nimmagadda U, Salem MR, Crystal GJ. Preoxygenation:
physiologic basis, benefits, and potential risks. Anesth Analg.
2017;124:507–517.
3. Tanoubi I, Drolet P, Donati F. Optimizing preoxygenation in
adults. Can J Anaesth. 2009;56:449–466.
4. Heller ML, Watson TR Jr. Polarographic study of arterial oxy-
genation during apnea in man. N Engl J Med. 1961;264:326–330.
5. Russell T, Ng L, Nathan E, Debenham E. Supplementation of
standard pre-oxygenation with nasal prong oxygen or machine
oxygen flush during a simulated leak scenario. Anaesthesia.
2014;69:1133–1137.
6. Kundra P, Stephen S, Vinayagam S. Techniques of preoxy-
genation in patients with ineffective face mask seal. Indian J
Anaesth. 2013;57:175–179.
7. McGowan P, Skinner A. Preoxygenation—the importance of a
good face mask seal. Br J Anaesth. 1995;75:777–778.
8. Bowles TM, Freshwater-Turner DA, Janssen DJ, Peden CJ; RTIC
Severn Group. Out-of-theatre tracheal intubation: prospective
multicentre study of clinical practice and adverse events. Br J
Anaesth. 2011;107:687–692.
9. Hayes-Bradley C, Lewis A, Burns B, Miller M. Efficacy of nasal
cannula oxygen as a preoxygenation adjunct in emergency air-
way management. Ann Emerg Med. 2016;68:174–180.
10. GroombridgeC,ChinCW,HanrahanB,HoldgateA.Assessment
of common preoxygenation strategies outside of the operating
room environment. Acad Emerg Med. 2016;23:342–346.
11. Brainard A, Chuang D, Zeng I, Larkin GL. A randomized trial
on subject tolerance and the adverse effects associated with
higher- versus lower-flow oxygen through a standard nasal
cannula. Ann Emerg Med. 2015;65:356–361.
12. R Core Team. R:Alanguage and environment for statistical com-
puting.RFoundationforStatisticalComputing,Vienna,Austria.
2016. Available at: https://www.R-project.org/. Accessed
April 1, 2016.
13. Mayo Healthcare. Airway management: resuscitation devices.
Reusable manual resuscitator. Available at: http://www.mayo-
healthcare.com.au/products/Resp_airManage_resuscitation.
htm#disp_resus. Accessed May 5, 2015.
14. De Rosario-Martinez H. Phia: Post-Hoc Interaction Analysis. R
package version 0.2–1; 2015. Available at: http://CRAN.R-
project.org/package=phia. Accessed April 1, 2016.
15. Hollands JG, Jarmasz J. Revisiting confidence inter-
vals for repeated measures designs. Psychon Bull Rev.
2010;17:135–138.
16. Faul F, Erdfelder E, Lang AG, Buchner A. G*Power 3: a flexible
statistical power analysis program for the social, behavioral,
and biomedical sciences. Behav Res Methods. 2007;39:175–191.
17. Cohen J. Statistical Power Analysis for the Behavioral Sciences.
New York, NY: Routledge; 1988.
18. Baillard C, Depret F, Levy V, Boubaya M, Beloucif S. Incidence
and prediction of inadequate preoxygenation before induction
of anaesthesia. Ann Fr Anesth Reanim. 2014;33:e55–e58.
19. Carmichael FJ, Cruise CJ, Crago RR, Paluck S. Preoxygenation:
a study of denitrogenation. Anesth Analg. 1989;68:406–409.
20. Nimmagadda U, Salem MR, Joseph NJ, Miko I. Efficacy of pre-
oxygenation using tidal volume and deep breathing techniques
with and without prior maximal exhalation. Can J Anesth.
2007;54:448–452.
21. Davis DP, Lemieux J, Serra J, Koenig W, Aguilar SA.
Preoxygenation reduces desaturation events and improves
intubation success. Air Med J. 2015;34:82–85.
22. Baillard C, Fosse JP, Sebbane M, et al. Noninvasive ventila-
tion improves preoxygenation before intubation of hypoxic
patients. Am J Respir Crit Care Med. 2006;174:171–177.