High-Flow Nasal Cannula: Impact on Oxygenation and
Ventilation in an Acute Lung Injury Model
Meg Frizzola, DO,1* Thomas L. Miller, PhD,2,3Maria Elena Rodriguez, MD,1Yan Zhu, MD,1
Jorge Rojas, MD,4Anne Hesek, AS, RLT,1Angela Stump, BS, RT,1Thomas H. Shaffer, MSE, PhD,1,2,5
and Kevin Dysart, MD1,2
Summary. Introduction: High-flow nasal cannula therapy (HFNC) has been shown to be more
effective than continuous positive airway pressure (CPAP) in reducing intubations and
ventilator days. HFNC likely provides mechanisms to support respiratory efficiency beyond
application of distending pressure. We reason that HFNC washout of nasopharyngeal dead space
impacts CO2removal along with oxygenation. The aim of this study was to demonstrate the flow
IV oleic acid and supported with HFNC at 2 through 8L/min. High and low leak around the nasal
physiologic equilibration. Tracheal pressures were recorded by transmural catheters. Results: With
HFNC, CO2trended downward in a flow-dependent manner independent of leak. Oxygenation and
tracheal pressures increased in a flow-dependent manner with the greatest effect during double
exchange in a flow-dependent manner; double prong had greater impact on O2;single prong had
greater impact on CO2elimination. Pediatr Pulmonol. ? 2010 Wiley-Liss, Inc.
Key words: high-flow nasal cannula; animal model; gas exchange; dead space; lung
Funding source: NIH COBRE, Number P20 RR 020173-06; Nemours Foundation;
Continuous positive airway pressure (CPAP) is often a
preferred intervention over mechanical ventilation with
neonatal respiratory distress syndrome in that oxygen-
ation is improved and lung volume is recruited while
supporting spontaneous ventilatory efforts.1,2
essential clinical criteria to remain on CPAP are effective
spontaneous respiratory effort and CO2 elimination.
Hypercapnia, or apnea that may be secondary to hyper-
capnia, arecited as some of the morecommonreasons for
Therefore, it is thought that if CO2retention during CPAP
can be reduced or eliminated, many infants can be spared
potential lung injury and subsequent chronic lung disease
induced by mechanical ventilation.
High-flow nasal cannula (HFNC; >2Liters per minute
(lpm)) has been proposed by some physicians as an
alternative to CPAP in maintaining infants without the
need for intubation and mechanical ventilation.3The use
of HFNC in place of conventional methods has been
controversial and as yet not well-documented in the
literature as being more effective than CPAP. Given
the paucity of translational research investigating the
1Nemours Children’s Clinic of Wilmington, The Nemours Foundation,
Alfred I. duPont Children’s Hospital, Primary Research Institution,
2Department Pediatrics, Jefferson Medical College, Philadelphia, Pennsyl-
3Vapotherm, Inc., Stevensville, Maryland.
4Pediatrix, Nashville, Tennessee.
5Departments of Physiology and Pediatrics, Temple University School of
Medicine, Philadelphia, Pennsylvania.
Data from this manuscript was presented at the 2008 Annual Meeting of the
Pediatric Academic Societies.
*Correspondence to: Meg Frizzola, DO, Department of Critical Care,
Nemours Children’s Clinic—Wilmington of The Nemours Foundation,
Alfred I. duPont Hospital for Children, 1600 Rockland Road, Wilmington,
DE 19803. E-mail: firstname.lastname@example.org
Received 30 December 2009; Revised 14 June 2010; Accepted 15 June
Published online in Wiley Online Library
? 2010 Wiley-Liss, Inc.
mechanism of action of HFNC, we sought to investigate
this mechanism further. We hypothesized that the
mechanism of gas exchange with HFNC is associated
with the washout of nasopharyngeal anatomical dead
In a recent study, we used tracheal gas insufflation
(TGI) with CPAP in lung-injured piglets using a specially
designed endotracheal tube to allow for insufflation of
fresh gas flow (No. 6501.30; Vygon, Ecouen, France).4
The results showed a significant reduction in CO2
retention associated with prosthetic dead space washout
increase the number of infants that can be sustained
without intubation or mechanical ventilation and these
infants may be weaned sooner.
Given our TGI results, we reasoned that the beneficial
clinical responses to HFNC may be related to the high-
flow rates effectively flushing the nasopharyngeal cavity
of expiratory gas. Thus, we proposed that the HFNC may
provide a practical means to wash out nasopharyngeal
for intubation. The present study was designed to
demonstrate this washout effect and the flow dependence
of CO2elimination and oxygenation during HFNC. In
addition, the study was conducted to demonstrate the
independent effects of airway pressure under different
would affect CO2retention in a flow-dependent manner
without markedly increasing intra-tracheal pressure and
improve respiratory parameters compared to a control
MATERIALS AND METHODS
mask CPAP of 5cmH2O and injured by intravenous
administration of oleic acid. All 13 piglets were treated
with CPAP (minimal leak), single prong (SP) HFNC with
ahighdegree ofleak around the nasal prongs(HILEAK),
and double prong (DP) HFNC with a low degree of leak
around the nasal prongs (LOW-LEAK). The three treat-
ment conditions were administered in a randomized,
repeated measure, crossover design described below.
Following experimentation, animals were sacrificed with
pentobarbital (50mg/kg) and saturated potassium chlor-
ide (2mEq). At baseline, injury and at each treatment
measurements were made of arterial blood chemistry,
pulmonary function, and intra-tracheal pressure. All
procedures for the animal protocol in this study were
Committee at Nemours Children’s Clinic—Wilmington
of The Nemours Foundation.
Instrumentation and Injury Protocol
The piglets were anesthetized with an anesthetic
solution (ketamine: 23mg/kg; azepromazine: 0.58mg/
kg; and xylazine: 0.8mg/kg [KAX]) given as two 1ml/kg
intramuscular injections separated by 10min. The skin
and soft tissues were locally infiltrated with 0.5%
lidocaine HCl (4mg/kg) and instrumented with 5.0F
catheters placed in the jugular vein and carotid artery.
Additionally, a 5.0F fluid filled catheter was placed
directly into the mid trachea approximately 2cm below
the cricoid ring at a perpendicular orientation for
measurements of dynamic tracheal pressure. The tracheal
catheter insertion site was sealed with surgical glue to
eliminate leak around the catheter. Subsequent anesthesia
was maintained with an intravenous infusion of KAX at
0.4ml/kg/hr, along with diazepam every 2hr and addi-
at a rate of 6ml/kg/hr. Arterial blood pressure was
monitored by attaching the arterial catheter to a standard
pressure transducer via bedside patient monitor (Model
M1175A; Hewlett Packard, Palo Alto, CA). ECG electro-
des were placed for monitoring. Throughout the protocol
the animal’s rectal temperature was monitored and
maintained at 37–388C on a radiant warmer bed
(Resuscitaire1; Hill-Rom Air-Shields, Hatboro, PA).
Following instrumentation, baseline measurements
were made and then lung injury was initiated. Oleic acid
(0.08ml/kg; Sigma, St. Louis, MO) was emulsified in
four equal doses with ample time to allow stabilization
between doses. The injury model has been previously
described4and the oleic acid dose was refined in a pilot
study to achieve a 50% reduction in PaO2and respiratory
compliance while preserving spontaneous respiration.
Following the injury procedure and a 30min stabiliza-
tion period, all 13 animals were exposed to HI-LEAK,
LOW-LEAK, and CPAP (minimal leak) in randomly
assigned order. For each treatment scenario, flow rate
then incrementally back to the least; the two sets of data
for each increment were then averaged to filter out the
residual effects of the previous level of treatment. To be
consistent with current neonatal clinical practice, HFNC
flow rates ranged from 2 to 8lpm and CPAP pressures
used ranged from 2 to 6cmH2O flow was held constant at
8L/min. Inspired oxygen fraction was always held at 1.0
for consistency and so that PaO2reflects the a/A ratio.
2 Frizzola et al.
HFNC therapy was administered using a Vapotherm
2000i (Vapotherm, Inc., Stevensville, MD) with a
pediatric nasal cannula. Given the small diameter of the
piglet nares relative to the human, the pediatric cannula
fully occluded both nares of the piglet, thus representing
the LOW-LEAK condition. To achieve the HIGH-LEAK
condition where only half the area of the nares was
obstructed (as per manufacturer’s recommendation), one
of the prongs of a pediatric cannula was removed and the
hole sealed, such that one nare of the piglet was occluded
with the cannula prong and the other nare was open to
CPAP was administered by a VIP Bird infant ventilator
(Bird Products Corp., Palm Springs, CA) set to the CPAP
mode. The ventilator circuit was attached to the piglet
using a semi-cone shaped face mask lined with silicone
that conformed to the piglet’s head shape to minimize the
addition of prosthetic dead space. Furthermore, the base
of the face mask (around the piglets head) allowed
for minimal gas leak to allow washout of any remaining
theventilator was able togenerate the prescribed levels of
CPAP as indicated by ventilator circuit and pig tracheal
Pulmonary Function Assessment
Arterial blood gases and chemistrywere measured bya
standard patient blood analyzer (Stat Profile1; Nova
Biomedical, Waltham, MA). Blood gas parameters
measured included pH, PaCO2, PaO2, hemoglobin,
hematocrit, HCO3, and base excess values. The blood
in respiratory support to ensure adequate physiologic
equilibration given that the respiratory system compen-
sates for changes in acid–base status within this time
period.5Respiratory parameters were assessed by induc-
tive plethysmography (SomnoStar PT; SensorMedics,
Yorba Linda, CA), including relative tidal volume,
respiratory rate, and thoracoabdominal synchrony. In
all animals, relative tidal volumes, respiratory rates,
and minute ventilations were determined prior to lung
injury duringmanual bagging (via mask CPAP) with low-
(2–4ml/kg) and high-(6–8ml/kg) tidalvolumestrategies
in order to establish relative calibration standards for
Data were analyzed using linear regression or analysis
of variance (ANOVA) to assess differences associated
with type and/or magnitude of therapy. Post-hoc analyses
were done, where appropriate, using Bonferroni compar-
isons. Significance was accepted with P-values<0.05.
Thirteen piglets were utilized for the study and
demonstrated a stable injury; age 13?8 days and weight
6.0?0.2kg. The injury was associated with a substantial
increase in PaCO2(37.3mmHg?6.6 vs. 49.1mmHg?
6.1;P<0.01)despite thetransitionfromroom air toFiO2
of 1.0 and an increase in A-a gradient (23.3?12.1 vs.
254.3?88.9; P<0.01). These findings are demonstrated
in Figure 1.
As shown in Figure 2, it was possible to record direct
temporal changes in tracheal pressure at each specific
therapeutic intervention. A data acquisition recorder
(DASH 8Xe) was used for the tracheal measurements.
As such, these traces provided evaluation of pressure
response to increments in CPAP or flow during HFNC
perturbations. Based on summarized data shown in
Figure 3A,B, there was a direct linear relationship
between each prescribed CPAP setting and tracheal
pressure (slope¼0.6; r2¼0.99; P¼0.02). Similarly, a
direct linear relationship existed between flow rate and
Fig. 1. Baseline and injury: gas exchange alterations following oleic acid injury. Left panel:
change in arterial PCO2, baseline versus injury. Right panel: change in A-a gradient baseline
versus injury (N¼13; P<0.01).
High-Flow Nasal Cannula3
tracheal pressure under the LOW-LEAK condition
(slope¼0.21; r2¼0.82; P¼0.04). Under the HIGH-
resultant tracheal pressure showed a significant rise in
tracheal pressure between 2 and 8lpm of flow (r2¼0.81;
P¼0.04). Overall, mean tracheal pressures were higher
duringthe low-leak condition compared to high-leakflow
Pulmonary and Physiologic Parameters
tidal volume across therapies or orders of magnitude (i.e.,
relative minute ventilation values determined through
integration of rates and relative tidal volumes (inductive
plethysmography) also showed no differences as demon-
strated in Table 1.
Physiologic parameters were measured for all piglets
under every condition. Measurements included rectal
temperature, systolic blood pressure, diastolic blood
pressure, mean arterial blood pressure, heart rate,
deviations for each parameter under all conditions are
depicted in Table 2. There was no significance found for
any of these values.
Figure 4 shows the effects of incremental increases in
CPAP and HFNC therapies on PaCO2. With initiation of
CPAP (2cmH2O), arterial carbon dioxide tension
(P<0.05); however, no further decease in PaCO2was
seen with the incremental increases in CPAP pressure.
With HFNC, in the LOW-LEAK condition a significant
relationship was found between flow rate and partial
pressure of carbon dioxide (slope¼?0.73; r2¼0.92;
P<0.01) where PaCO2 decreased with incremental
increases in flow rate in a sigmoidal fashion. In the
HIGH-LEAK condition, PaCO2was restored to baseline
levels immediately (2lpm of flow) and therefore did not
decrease further with incremental increases in flow rate.
the basalinjury level
Oxygenation versus CPAP/flow rate data is shown in
Figure 5. With CPAP administration, arterial oxygen
tension (PaO2) rose to over 400mmHg butdid not change
with incremental increases in CPAP pressure. However,
PaO2showed a flow-dependent increase for both LOW-
LEAK and HIGH-LEAK conditions. In the LOW-LEAK
condition, PaO2responded in a significant linear relation-
ship with throughout the range of flow rates 2–8lpm
(slope¼30.6; r2¼0.85; P¼0.03), and there was a
400mmHg plateau from 8 to 10lpm. Under the HIGH-
LEAK condition, PaO2rapidly increased from 2 to 6lpm
before reaching a 400mmHg plateau above 6lpm.
This study demonstrated two important findings. First,
with HFNC tracheal pressures were comparable to CPAP
pressures at the same flow range. Second, washout of
nasopharyngeal dead space is associated with improved
gas exchange using HFNC and may allow patients to
Fig. 2. Real-time tracheal pressures: typical tracing of real-time
tracheal pressure as recorded using a water-filled catheter at SP
Fig. 3. Tracheal pressures: summarized data for tracheal pressures during CPAP; r2¼0.99;
P¼0.02 (left panel). Tracheal pressures for high- (r2¼0.81; P¼0.04) and low-leak conditions
(r2¼0.82; P¼0.04; right panel).
4 Frizzola et al.
breathe more comfortably as represented clinically.6As
ourresults demonstrate, tracheal pressuresin the neonatal
piglets were no greater than with traditional CPAP
ventilation and pressures. This is consistent with HFNC
clinical studies that report lower esophageal pressures
compared to CPAP at 6cmH2O.7–11
With respect to gas exchange, the impact of increasing
of tracheal pressure generation alone. With increasing
CPAP pressure in this injured lung model, neither PaCO2
nor PaO2demonstrated progressive change. However,
with HFNC under both leak conditions PaCO2and PaO2
by saturation curves (PaCO2decreased with increasing
flow until saturation and PaO2increased with increasing
flow until saturation). These saturation relationships are
as demonstrated in the literature from TGI.
catheter inserted into an endotracheal tube, or a specially
designed endotracheal tube for gas insufflation.6,12–15
the catheter at low flows (typically 0.5–1.0lpm). This
method supplies fresh gas during inspiration, and also
tip and the oral/nasal openings during expiration and the
end expiratory pause. By flushing prosthetic dead space
during the expiratory phase of the respiratory cycle, the
less residual end-expiratory gas. Therefore, alveolar gas
fractions are moved toward fresh gas values; that is, less
carbon dioxide and more oxygen, thereby improving
is that it requires intubation.4
of dead space to be flushed is determined by the time
Therefore, greater gas flows will flush more dead space
anatomical/prosthetic dead space in the allotted time;
beyond this liter flow, increased flow rates accomplish no
thus arterial gas composition.
of 8lpm, while both ventilation and oxygenation
responses reached saturation near 6–8lpm flow. There-
fore, we determined that nasopharyngeal dead space
washout was a more predominant factor in determining
LOW-LEAK, the partial pressure of carbon dioxide was
lower for flow rates (<6lpm) in the high-leak condition.
We attribute this phenomenon to a better washout of the
TABLE 1—Minute Ventilation Values (AU) as a Function Flow
Minute ventilation data is presented in arbitrary units (AU) as determined by respiratory inductive plethysmography (RIP) measurements. Data represent mean?SD.
High-Flow Nasal Cannula5
be achieved at a lower tracheal pressure.
of tracheal pressure, positiveend distending pressure, and
CPAP generated during the use of high-frequency nasal
cannula therapy, these studies are confounded by several
measurement limitations. The results to date have been
collected by means of nasal pharyngeal pressure monitor-
ing, oral cavity pressure monitoring, or esophageal
pressure monitoring. Although these methods are
all minimally invasive; they provide only an estimation
of actual airway pressure. The major complication with
these forms of measurement are the following: inaccurate
pressure assessment due to catheter placement or catheter
Fig. 4. Partial pressure carbon dioxide: effects of incremental increases in CPAP on PCO2
r2¼0.5; P¼0.5 (left panel). Effects of incremental increases in HFNC flow rate on PCO2(low leak
r2¼0.92; P<0.01; low leak r2¼0.7; P¼0.07; right panel).
Fig. 5. Partial pressure oxygen: effects of incremental increases in CPAP on PO2(left panel).
top right panel). Effects of incremental increases in HFNC flow rate on O2in high-leak condition
(bottom right panel).
6 Frizzola et al.
design, catheter orientation with respect to the gas flow,
precise, accurate, and responsive dynamic tracheal
pressure monitoring/recording via a fluid-filled catheter
surgically placed mid trachea with a perpendicular
orientation (with respect to gas flow). Using this
technique, as described, we did not find significantly
elevated tracheal pressures over the flow ranges tested.
This is the first time this method has been used to directly
Taking intoaccount the possibilityof otherphysiologic
adjustments in minute ventilation impacting alveolar and
relative minute ventilation using inductive plethysmog-
raphy. No changes in this relative minuteventilation were
noted, indicating that changes in blood gas values were
directly associated with the change in therapy paradigm.
In addition, as therapeutic interventions were randomized
alveolar dead space, which could have accounted for
changes in gas exchange. Taken together, these data
support the conclusion that HFNC may augment carbon
dioxide elimination by means of nasopharyngeal dead
is the translation of data to clinical application. The
anatomy of a piglet, especially the nasopharyngeal
anatomy is very different from a human. In this regard,
we believe that the cavernous human nasopharynx, as
opposed to the elongated pig nasopharynx would flush
more easily without as much pressure development.
Nonetheless, we contend that the anatomy is similar
model for treatment of lung injury has been well-
The current data suggest that HFNC therapy may be
used as a first line option over traditional CPAP support.
This study shows that HFNC provides improved levels of
arterial CO2, whereas CPAP does not, independent of
changes in minute ventilation. Therefore, HFNC can be
viewed as a truly non-invasive means of ventilatory
support. In addition, by way of nasopharyngeal dead
space elimination, HFNC can serve as a means of
oxygenation support independent of, or in addition to,
supplemental oxygen administration. Clinical studies are
now warranted to refine flow needs in the human scenario
to optimize these effects demonstrated in this preclinical
The authors gratefully acknowledge The Nemours
Foundation, National Institutes of Health COBRE grant
TABLE 2—Physiologic Parameters as a Function of CPAP, SP, and DP Across Experimental Conditions
Single prong (SP)
Double prong (DP)
165.1?32.3 132.8?15.1 138.8?32.3 138.4?25.1
134.2?29.8 128.2?21.1 135.7?33.2 142.5?32.1 136.5?29.9
Respiratory rate 48.8?17.1
All data are presented as mean?SD and no parameters were found to be significant as a function of condition or flow.
High-Flow Nasal Cannula7
and Vapotherm, Inc. (Stevensville, MD) in part for Download full-text
financial supportof the studyand provision of equipment.
1. Gregory GA, Kitterman JA, Phibbs RH, Tooley WH, Hamilton
WK. Treatment of the idiopathic respiratory-distress syndrome
with continuous positive airway pressure. N Engl J Med
2. Saunders RA, Milner AD, Hopkin IE. The effects of continuous
positive airway pressure on lung mechanics and lung volumes in
the neonate. Biol Neonate 1976;29:178–186.
3. Shoemaker MT, Pierce MR, Yoder BA, DiGeronimo RJ. High
flow nasal cannula versus nasal CPAP for neonatal respiratory
disease: a retrospective study. J Perinatol 2007;27:85–91.
4. Miller TL, Blackson TJ, Shaffer TH, Touch SM. Tracheal gas
insufflation-augmented continuous positive airway pressure in a
spontaneously breathing model of neonatal respiratory distress.
Pediatr Pulmonol 2004;38:386–395.
5. Hall J. Regulation of acid base balance. In: Guyton AC, Hall JE,
editors. Guyton and Hall Textbook of Medical Physiology. 11th
edition. Philadelphia: Elsevier Saunders; 2006. pp. 383–401.
6. Danan C, Dassieu G, Janaud JC, Brochard L. Efficacy of dead-
space washout in mechanically ventilated premature newborns.
Am J Respir Crit Care Med 1996;153:1571–1576.
7. Kubicka ZJ, Limauro J, Darnall RA. Heated, humidifiedhigh flow
nasal cannula therapy: yet another way to deliver continuous
positive airway pressure? Pediatrics 2008;121:82–88.
8. Spence KL, Murphy D, Kilian C, McGonigle R, Kilani RA. High
flow nasal cannula as a device to provide continuous positive
airway pressure in infants. J Perinatol 2007;27:772–775.
9. Wilkinson DJ, Andersen CC, Smith K, Holberton J. Pharyngeal
pressure with high-flow nasal cannulae in premature infants.
J Perinatol 2008;28:42–47.
10. Saslow JG, Aghai ZH, Nakhla TA, Hart JJ, Lawrysh R, Stahl GE,
Pyon KH. Work of breathing using high-flow nasal cannula in
preterm infants. J Perinatol 2006;26:476–480.
11. Kahn DJ, Courtney SE, Steele AM, Habib RH. Unpredictability
of delivered bubble nasal continuous positive airway pressure:
role of bias flow magnitude and nares-prong air leaks. Pediatric
12. Dassieu G, Brochard L, Agudze E, Patkaı ¨ J, Janaud JC, Danan C.
Continuous tracheal gas insufflation enables a volume reduction
strategy in hyaline membrane disease: technical aspects and
clinical results. Intensive Care Med 1998;24:1076–1082.
13. Claure N, D’Ugard C, Bancalari E. Elimination of ventilator dead
space during synchronized ventilation in premature infants.
J Pediatr 2003;143:315–320.
14. Burke WC, Nahum A, Ravenscraft SA, Nakos G, Adams AB,
Marcy TW, Marini JJ. Modes of tracheal gas insufflation.
Comparison of continuous and phase-specific gas injection in
normal dogs. Am Rev Respir Dis 1993;148:562–568.
15. Bernath MA, Henning R. Tracheal gas insufflation reduces
requirements for mechanical ventilation in a rabbit model of
respiratory distress syndrome. Anaesth Intensive Care 1997;25:
16. Nahum A. Animal and lung model studies of tracheal gas
insufflation. Respir Care 2001;46:149–157.
17. Dassieu G, Brochard L, Benani M, Avenel S, Danan C.
Continuous tracheal gas insufflation in preterm infants with
hyaline membrane disease. A prospective randomized trial. Am J
Respir Crit Care Med 2000;162:826–831.
18. Nakos G, Lachana A, Prekates A, Pneumatikos J, Guillaume M,
Pappas K, Tsagaris H. Respiratory effects of tracheal gas
insufflation in spontaneously breathing COPD patients. Intensive
Care Med 1995;21:904–912.
19. Parke R, McGuinness S, Eccleston M. Nasal high-flow therapy
delivers low level positive airway pressure. Br J Anaesth
20. Lampland A, Plumm B, Meyers P, Worwa C, Mammel MC.
Observational study of humidified high-flow nasal cannula
compared with nasal continuous positive airway pressure.
J Pediatr 2009;154:177–182.
21. Locke RG, Wolfson MR, Shaffer TH, Rubenstein SD, Greenspan
JS. Pediatrics 1993;91:135–138.
22. Miller TL, Singhaus CJ, Sherman TI, Greenspan JS, Shaffer TH.
Physiologic implications of helium as a carrier gas for inhaled
nitric oxide in a neonatal model of Bethanecol-induced broncho-
constriction. Pediatr Crit Care Med 2006;7:159–164.
8 Frizzola et al.