CHEST Original Research

CRITICAL CARE

Patient-Ventilator Asynchrony During

Noninvasive Ventilation
A Bench and Clinical Study
Guillaume Carteaux, MD; Aissam Lyazidi, PhD; Ana Cordoba-Izquierdo, MD;
Laurence Vignaux; Philippe Jolliet, MD; Arnaud W. Thille, MD, PhD;
Jean-Christophe M. Richard, MD, PhD; and Laurent Brochard, MD

Background: Different kinds of ventilators are available to perform noninvasive ventilation (NIV)
in ICUs. Which type allows the best patient-ventilator synchrony is unknown. The objective was
to compare patient-ventilator synchrony during NIV between ICU, transport—both with and
without the NIV algorithm engaged—and dedicated NIV ventilators.
Methods: First, a bench model simulating spontaneous breathing efforts was used to assess
the respective impact of inspiratory and expiratory leaks on cycling and triggering functions
in 19 ventilators. Second, a clinical study evaluated the incidence of patient-ventilator asyn-
chronies in 15 patients during three randomized, consecutive, 20-min periods of NIV using
an ICU ventilator with and without its NIV algorithm engaged and a dedicated NIV ventilator.
Patient-ventilator asynchrony was assessed using flow, airway pressure, and respiratory muscles
surface electromyogram recordings.
Results: On the bench, frequent auto-triggering and delayed cycling occurred in the presence of
leaks using ICU and transport ventilators. NIV algorithms unevenly minimized these asynchronies,
whereas no asynchrony was observed with the dedicated NIV ventilators in all except one. These
results were reproduced during the clinical study: The asynchrony index was significantly lower
with a dedicated NIV ventilator than with ICU ventilators without or with their NIV algorithm
engaged (0.5% [0.4%-1.2%] vs 3.7% [1.4%-10.3%] and 2.0% [1.5%-6.6%], P , .01), especially
because of less auto-triggering.
Conclusions: Dedicated NIV ventilators allow better patient-ventilator synchrony than ICU and
transport ventilators, even with their NIV algorithm. However, the NIV algorithm improves, at least
slightly and with a wide variation among ventilators, triggering and/or cycling off synchronization.
CHEST 2012; 142(2):367–376

Abbreviations: AI 5 asynchrony index; ICUniv2 5 ICU ventilator with the noninvasive ventilation algorithm turned
off; ICUniv1 5 ICU ventilator with the noninvasive ventilation algorithm turned on; NIV 5 noninvasive ventilation;
NIVv 5 dedicated noninvasive ventilation ventilator; PEEP 5 positive end-expiratory pressure; TD 5 triggering delay;
Tiexcess 5 insufflation time in excess; Tisim 5 simulated active inspiration time; Tivent 5 time between the beginning of
a simulated inspiratory effort and the end of the ventilator’s insufflation

Noninvasive ventilation (NIV) has become a stan-

dard of care for the management of many causes
with a high incidence during NIV in critically ill
patients.7
of acute respiratory failure. 1-3 During NIV, the Different ventilators are now used to conduct NIV
unavoidable presence of leaks around the mask4 can in ICU: ICU ventilators,2 dedicated NIV ventilators,8
interfere with the ventilator performance. Expira-
tory leaks can mimic an inspiratory effort for the For editorial comment see page 274
ventilator, leading to auto-triggering5; and inspiratory
leaks can mimic a sustained inspiration, leading to and also transport ventilators when needed.9-11 Most
delayed cycling.6 Not surprisingly, patient-ventilator ICU ventilators were initially built to work without any
asynchronies have, therefore, been reported to occur leak, at least in adults, and are prone to be disrupted

journal.publications.chestnet.org CHEST / 142 / 2 / AUGUST 2012 367

Downloaded From: http://journal.publications.chestnet.org/ on 08/11/2012

 by the presence of leaks during NIV.12 To address we conducted a clinical study in critically ill patients
this issue, manufacturers have implemented NIV algo- to compare the incidence of patient-ventilator asyn-
rithms (so called “NIV modes”) on the latest gen- chrony between ICU ventilators with and without
eration of ICU ventilators to compensate and better their NIV algorithm engaged, and a dedicated NIV
manage the leaks. Both bench12,13 and clinical14 studies ventilator.
assessing the performance of NIV algorithms on ICU
ventilators have shown mixed results, partly due to
large variations among the ventilators, making it dif- Materials and Methods
ficult to draw an overall conclusion. Dedicated NIV
This study involved a bench part and a clinical part. An exten-
ventilators stem from bilevel home ventilator tech- sive description of both the bench and clinical protocols is provided
nology, which has been particularly oriented toward in e-Appendix 1.
leakage management and comfort. Some bench studies
suggested that a dedicated NIV ventilator could pro- Bench Study
duce better performance and synchronization than All 19 ventilators tested are reported in Table 1 and included
ICU ventilators in the presence of leaks.13,15 However, eight ICU ventilators, five transport ventilators, and six dedi-
no bench model concerning ventilator synchroniza- cated NIV ventilators. The test lung, an Active Servo Lung 5000
tion during NIV has been clinically validated, raising (ASL 5000; IngMar Medical, Ltd), was used to simulate a mod-
the question of their clinical relevance in critically ill erate inspiratory effort in the presence of an 80 mL/cm H2O respira-
tory system compliance and 10 cm H2O/L/s resistance to mimic
patients. Consequently, the kind of ventilator that a mild obstructive condition. The simulated respiratory rate was
allows the best synchronization during NIV in the 15 breaths/min and the inspiratory time 0.8 s. Three leak condi-
ICU is still unknown. In some areas, NIV is mainly tions were generated (Fig 1A): absence of leak, continuous leak
delivered with dedicated NIV ventilators,8 whereas (to reveal triggering asynchronies during expiratory leak), and
in other countries ICU ventilators are almost exclu- inspiratory leak (to reveal cycling-off asynchronies). For this last
experiment, the leak started at a pressure corresponding to a
sively preferred,2 and this distribution reflects local water column of 7 cm H2O, as detailed in e-Appendix 1. The
habits rather than an evidence-based approach. inspiratory leak was characterized by a nonlinear pressure-flow
The purpose of this study was to compare patient- relationship with a flow varying from 0 to 22 L/min for a pressure
ventilator synchronization during NIV using ICU from 7 to 15 cm H2O. The continuous (expiratory) leak was char-
and transport ventilators with or without their NIV acterized by a flow of 16 L/min at 5 cm H2O pressure.
Ventilators were set in pressure support ventilation, with a
algorithm, and finally dedicated NIV ventilators. We pressure support level at 15 cm H2O and a positive end-expiratory
designed a bench model to assess ventilator syn- pressure (PEEP) at 5 cm H2O. ICU and transport ventilators were
chronization with a simulated inspiratory effort in tested with and without their NIV algorithm engaged, except
different leak conditions, simulating the different the Elisee 250, whose NIV algorithm cannot be turned off. Data
challenges to be faced by the ventilator. Furthermore, were acquired at 512 Hz from ASL 5000 and stored in a laptop
computer for subsequent analysis (Acqknowledge 3.7.3; BIOPAC
Systems, Inc). Inspiratory triggering synchronization was assessed
Manuscript received September 7, 2011; revision accepted using the triggering delay, the triggering pressure-time product,
January 20, 2012. and the incidence of auto-triggering, expressed as a percentage
Affiliations: From the Réanimation Médicale (Drs Carteaux,
and calculated as follows: auto-triggering incidence (%) 5 (auto-
Lyazidi, Cordoba-Izquierdo, Thille, and Brochard), AP-HP, Groupe
Hospitalier Albert Chenevier-Henri Mondor; INSERM Unité 955 triggered cycles/total ventilator cycles) 3 100. The pressurization
(Equipe 13) (Drs Carteaux, Lyazidi, Cordoba-Izquierdo, Thille, was assessed using the pressure-time product at 300 milliseconds.
and Brochard), Université Paris EST, Créteil, France; the Depart- Cycling synchronization was assessed by determining ventilator
ment of Intensive Care (Ms Vignaux and Dr Brochard), Geneva insufflation time in excess (Tiexcess), expressed as a percentage and
University Hospital and Geneva University, Geneva; the Service calculated as follows: Tiexcess 5 [(Tivent 2 Tisim)/Tisim] 3 100,
de Médecine Intensive Adulte et Centre des brulés (Dr Jolliet), where Tivent is the time between the beginning of the simu-
Centre Hospitalier Universitaire Vaudois, Lausanne, Switzerland; lated inspiratory effort and the end of the ventilator’s insufflation,
the Faculté de biologie et de medicine (Dr Jolliet), Université de and Tisim the simulated active inspiration time. Delayed cycling
Lausanne, Lausanne, Switzerland; and the Réanimation médicale
was defined by a Tivent ⱖ 2 Tisim and premature cycling by a
(Dr Richard), Centre Hospitalier Universitaire Charles Nicolle,
Rouen, France. Tivent ⱕ 2/3 Tisim.
Funding/Support: This study was supported in part by a research
grant from Philips Respironics (€10,000). This study was per- Clinical Study
formed while Dr Carteaux was funded by an institutional grant,
the Année Recherche, from the Ministère de l’éducation nationale, A prospective, randomized, crossover study was conducted in
de l’enseignement supérieur et de la recherche (French Ministry two university hospital ICUs. The protocol was approved by the
for Education and Research). ethics committee CPP-Ile-de-France IX (number: 08-021), and
Correspondence to: Guillaume Carteaux, MD, Service de informed consent was obtained from all patients. We included
Réanimation Médicale, Hôpital Henri Mondor, 51 avenue du 15 patients in the ICU receiving NIV in pressure support ventila-
Maréchal de Lattre de Tassigny 94010 Créteil, France; e-mail:
[email protected] tion mode with PEEP via a standard oronasal mask. The venti-
© 2012 American College of Chest Physicians. Reproduction lator settings chosen by the clinician in charge of the patient were
of this article is prohibited without written permission from the kept identical for the study. Three consecutive NIV sessions were
American College of Chest Physicians. See online for more details. applied in a random order, using the same oronasal mask: (1) use
DOI: 10.1378/chest.11-2279 of an ICU ventilator whose NIV algorithm has been turned off

368 Original Research

Downloaded From: http://journal.publications.chestnet.org/ on 08/11/2012

 Table 1—Bench Study: Characteristics of the ICU, Transport, and NIV Ventilators Tested in the Bench Study

Ventilator Supplier Use Gas Source Circuit NIV Mode ET Range IT Range
Avea CareFusion ICU Pressurized Double Manual 5%-45% 0.1-20 L/min
Engstrom GE Healthcare ICU Pressurized Double Manual 5%-50% 1-9 L/min; 21 to 210 cm H2O
Evita XL Dräger ICU Pressurized Double Automatic Automatic 0.3-15 L/min
G5 Hamilton Co ICU Pressurized Double Manual 5%-70% 0.5-15 L/min
PB840a Covidien ICU Pressurized Double Manual 1%-80% 0.2-20 L/min
Servo-i MAQUET GmbH ICU Pressurized Double Manual 1%-40% 0%-100%; 220 to 0 cm H2O
& Co KG
V500 Dräger ICU Pressurized Double Automatic/manual Automatic; 5%-70% Automatic; 0.2-15 L/min
Vela CareFusion ICU Turbine Double Manual 5%-40% 1-8 L/min
Elisee 250 ResMed Transport Turbine Double Automatic/manual Automatic; 1%-6% Automatic
Medumat Weinmann Medical Transport Pneumatic Single Automatic 5%-50% 1-15 L/min
Technology
Oxylog 3000 Dräger Transport Pneumatic Single Automatic Automatic Automatic
Supportair Covidien Transport Turbine Single Manual 5%-95% 01-05
T1 Hamilton Co Transport Turbine Double Manual 5%-80% 1-20 L/min
BiPAP Vision Philips Respironics NIV Turbine Single Automatic Automatic Automatic
Carina Dräger NIV Turbine Single Automatic Automatic Analogical (sensible/normal)
Trilogy 100 Philips Respironics NIV Turbine Single Automatic Automatic Automatic
V60 Philips Respironics NIV Turbine Single Automatic Automatic Automatic
Vivo 40 Breas NIV Turbine Single Automatic Automatic Automatic
VPAP 4 ResMed NIV Turbine Single Automatic Automatic Automatic
ET 5 expiratory trigger, expressed as a percentage of peak inspiratory flow; IT 5 inspiratory trigger; NIV 5 noninvasive ventilation.
Version comprising both an NIV mode and leak compensation.
a

(ICUniv2), (2) use of an ICU ventilator whose NIV algorithm has leaks exhibited a total triggering delay (TD) of 117 mil-
been turned on (ICUniv1), and (3) use of a dedicated NIV venti- liseconds (99-131 milliseconds) and 143 millisec-
lator (NIVv). Each session was 20 min long. ICU ventilators used onds (114-174 milliseconds), respectively (P 5 .37)
in the clinical study were: Evita XL or EVITA 4 (Dräger) (n 5 12)
and Engstrom Carestation (GE Healthcare) (n 5 3). The dedicated
(Fig 2).The addition of inspiratory leaks did not sig-
NIVv was the BiPAP Vision (Philips Respironics). We selected nificantly modify these values except for the Engstrom,
this ventilator because it is widely used in ICUs using NIV ven- G5, and T1, which had an increased TD, and the
tilators and also because it has been used in many clinical and Medumat, which showed a reduced TD. Turning
physiologic studies concerning NIV. Flow, airway pressure, and on the NIV algorithm while maintaining inspiratory
diaphragmatic and inspiratory neck muscles surface electromyo-
grams were continuously recorded throughout the three NIV ses-
leaks led to different behaviors among ICU and trans-
sions and stored in a laptop for subsequent analysis, as described port ventilators: TD significantly increased for five
in e-Appendix 1. All tracings were analyzed by one investigator ventilators (Medumat, Evita XL, Servo-i, V500, Sup-
(G. C.). The methodology used was previously described with- portair), decreased for three (Engstrom, PB840, T1),
out noticing any interobserver difference,7,14 and allowed the and was not modified for the others. In this last
quantification of major asynchrony events (ineffective triggering,
double-triggering, auto-triggering, premature cycling, and delayed
condition, the TD of ICU, transport, and dedicated
cycling) (Fig 1B). A global asynchrony index (AI), expressed as a NIV ventilators were 107 (83-120), 126 (112-190),
percentage, was computed as follows16: AI (%) 5 (number of asyn- and 125 (102-145) milliseconds, respectively (P . .05
chronies/[ineffective breaths 1 ventilator cycles]) 3 100. for every intergroup comparison). When NIV algo-
rithms were used in the presence of inspiratory leaks,
Statistics six ICU ventilators (Avea, Engstrom, PB840, Servo-i,
Statistical analyses were performed with Statistical Package V500, Vela), two transport ventilators (Elisee 250,
for the Social Sciences (version 16.0, SPSS). Continuous data Supportair), and two NIV ventilators (BiPAP Vision, V60)
are expressed as the median (25th-75th percentile). In both exhibited a TD , 117 milliseconds (ie, the median
the bench and clinical study, the variables did not display
a normal distribution, so only nonparametric tests, detailed in
TD of ICU ventilators with the NIV algorithm turned
e-Appendix 1, were used. A P value of , 0.05 was considered sta- off in absence of leaks). The additional assessment of
tistically significant. the triggering pressure-time product is reported in
e-Appendix 1 and e-Figure 1.

Results Auto-Triggering: Occurrence of auto-triggering

was assessed during the presence of continuous leaks
Bench Study
(Fig 3). Expiratory leaks induced an incidence of auto-
Triggering Delay: The ICU and transport ventilators triggering between 0% and 100% among ICU and
with their NIV algorithm turned off in the absence of transport ventilators when their NIV algorithm was

journal.publications.chestnet.org CHEST / 142 / 2 / AUGUST 2012 369

Downloaded From: http://journal.publications.chestnet.org/ on 08/11/2012

 tilators whose NIV algorithm can be turned off. The
NIV algorithms generally minimized the insufflation
time, which remained significantly higher than without
leaks for only two ICU ventilators (Avea, G5) and three
transport ventilators (Oxylog 3000, Supportair, T1).
With NIV algorithm and inspiratory leaks, ICU,
transport, and dedicated NIV ventilators exhibited a
Tiexcess of 34% (29%-43%), 37% (25%-43%), and
37% (18%-49%), respectively. In this condition, the
Tiexcess was , 32% for four ICU ventilators (Engstrom,
Evita XL, Servo-i, V500), two transport ventilators
(Medumat, Supportair), and three dedicated NIV
ventilators (BiPAP Vision, Trilogy 100, V60).
During inspiratory leaks when NIV algorithms were
turned off, delayed cycling occurred with four ICU
ventilators (Avea, G5, PB840, Vela) and three trans-
port ventilators (Medumat, Oxylog 3000, T1). The
activation of the NIV algorithm eliminated delayed
cycling for all of these ventilators but one (G5). How-
ever, the NIV algorithm of the Servo-i overcorrected
the Tiexcess (24%). Concerning dedicated NIV ven-
tilators subjected to inspiratory leaks, one of them
(VIVO 40) exhibited delayed cycling.
We also assessed the ability of the ventilators to pres-
surize the airway in the first 300 milliseconds with or
without leaks. For the sake of simplicity, these data
are only shown in e-Appendix 1 and e-Figure 2.
Figure 1. Experimental protocols. A, Bench study experimental Clinical Study
design. To experimentally reproduce noninvasive ventilation (NIV)
conditions with calibrated leaks, we placed a T-piece between the Fifteen patients of median age 68 years old
ASL5000 (lung simulator) and the ventilator circuit. Three situa-
tions were generated: no leak, in which the free extremity of the (61-76 years) were included, 13 men and two women,
T-piece was closed; inspiratory leak, in which the free extremity with a median BMI of 24 kg/m2 (20-27 kg/m2). At
of the T-piece was connected to a tube immersed in a 7 cm H2O inclusion, Simplified Acute Physiology Score II was
column, allowing leaks to occur during insufflation only when the
pressure in the circuit was higher than the height of the water 47 (32-62) and arterial blood gas levels were as fol-
column; and continuous leak using the same experimental assembly lows: pH 5 7.36 (7.29-7.42), Pa co 2 5 48 mm Hg
without water in the receptacle, allowing leaks to occur during the
whole respiratory cycle. B, Clinical study representative record of
an auto-triggered cycle. EMGd 5 diaphragmatic electromyogram;
EMGn 5 neck muscles electromyogram; Paw 5 airway pressure.

turned off. The activation of the NIV algorithm led to

a heterogeneous response among these ventilators:
the incidence of auto-triggering fell to or remained at
0% for three ICU ventilators (PB840, Servo-i, V500)
and three transport ventilators (Elisee 250, Suppor-
tair, T1), was not modified for one ICU ventilator
(Avea), and decreased slightly for the other ICU and
transport ventilators. By contrast, no auto-triggering
occurred with any NIV ventilator.

Cycling and Insufflation Time: ICU and transport Figure 2. Bench study triggering delay. Representation of the
ventilators without their NIV algorithm in the absence triggering delay for ICU and transport ventilators with their NIV
algorithm turned off in the absence of any leak (NIV2/Leaks2,
of leaks exhibited a Tiexcess of 32% (30%-34%) and white bars), then in the presence of inspiratory leaks (NIV2/Leaks1,
49% (24%-75%), respectively (P 5 .93) (Fig 4). Inspi- gray bars); and for ICU and transport ventilators with their
ratory leaks led to a significant increase in insuffla- NIV algorithm turned on as well as for NIV ventilators in the
presence of inspiratory leaks (NIV1/Leaks1, black bars). # P , .05
tion time for six ICU ventilators (Avea, Engstrom, vs NIV2/Leaks2. $ P , .05 vs NIV2/Leaks1. See Figure 1 legend
G5, PB840, Servo-i, Vela) and all four transport ven- for expansion of abbreviation.

370 Original Research

Downloaded From: http://journal.publications.chestnet.org/ on 08/11/2012

 Figure 3. Bench study incidence of auto-triggering during continuous leaks. Incidence of auto-triggering
is represented as a percentage of the total ventilator cycles ([Auto-triggered cycles]/[total ventilator
cycles] 3 100) during continuous leaks with ICU and transport ventilators without NIV algorithm
(NIV2, white bar) and with the same ventilators with the NIV algorithm turned on, and with NIV
ventilators (NIV1, black bar). The activation of the NIV algorithm on ICU and transport ventilators
unequally led to an improvement in inspiratory triggering synchronization, whereas no auto-triggering
occurred with any NIV ventilator. See Figure 1 legend for expansion of abbreviation.

(41-63 mm Hg), Pao2/Fio2 5 206 mm Hg (183- for both comparisons) ( Fig 5 ). The incidence of
252 mm Hg). Patients had spent one median day each asynchrony during the three NIV sessions is
(0.3-1.0 days) under NIV before inclusion. Indica- represented in Figure 6. Auto-triggering had the
tions for NIV were the following: to avert respira- highest incidence. The incidence of auto-triggering,
tory failure after extubation (n 5 5), exacerbation of however, was significantly lower with NIVv than
COPD (n 5 4), cardiogenic pulmonary edema (n 5 3), with ICUniv2 and ICUniv1, 0.1/min (0.1-0.1/min)
community-acquired pneumonia (n 5 2), and post vs 0.5/min (0.1-1.1/min) and 0.3/min (0.1-1.2/min),
thoracic surgery (n 5 1). Eight patients (53%) had P , .001, and the proportion of patients who exhib-
COPD. Ventilator settings were pressure support ited a high incidence of auto-triggering (. 1/min) was
level 5 10 cm H2O (8-11 cm H2O), PEEP 5 4 cm H2O significantly lower with NIVv than with ICUniv2 and
(4-5 cm H2O), inspiratory trigger 5 1 L/min (1-2 L/min), ICUniv 1 ( Table 3 ). Four patients (27%) had an
pressurization slope 5 100 milliseconds (100-100 mil- AI . 10% with ICUniv2, two (13%) with ICUniv1,
liseconds), and Fio2 5 40% (30%-50%). There was and none with NIVv (P 5 .091). The level of leaks
no significant difference between the three NIV ses- throughout the clinical study was noticeably high
sions regarding ventilator settings, respiratory param- in these two last patients (14 and 16 L/min, respec-
eters, and the measured level of leaks (Table 2). ICU tively). The proportion of patients who exhibited at
ventilators used in the clinical study had a similar least one asynchrony with a high incidence (. 1/min)
response to leaks as during the bench study in terms was significantly higher with ICUniv2 and ICUniv1
of asynchrony: a propensity to auto-triggering with than with NIVv (Table 3).
expiratory leaks, partially corrected by the NIV algo-
rithm, but no delayed cycling with the NIV algorithm
and inspiratory leaks (Figs 3, 4). Discussion
Patient-Ventilator Synchrony: The asynchrony index To our knowledge, this study is the first to com-
(AI) did not significantly differ when using ICU venti- pare patient-ventilator synchronization during NIV
lators without (ICUniv2) or with (ICUniv1) their between ICU, transport, and dedicated NIV ventilators,
NIV algorithm engaged, 3.7% (1.4%-10.3%) vs 2.0% with both a bench and a clinical evaluation. The obser-
(1.5%-6.6%), respectively, P 5 .118. By contrast, AI vations made with these two approaches were con-
was significantly lower with NIVv (0.5% [0.4%-1.2%]) sistent, offering a strong validation of the bench
than with both ICUniv2 and ICUniv1 ( P 5 .001 model, a logical explanation for the clinical data, and

journal.publications.chestnet.org CHEST / 142 / 2 / AUGUST 2012 371

Downloaded From: http://journal.publications.chestnet.org/ on 08/11/2012

 Figure 4. Bench study on the effect of inspiratory leaks on insufflation time. Representation of the
insufflation time for ICU and transport ventilators without their NIV algorithm in the absence of leak
(NIV2/Leaks2, white bars), then in the presence of inspiratory leaks (NIV2/Leaks1, gray bars); and
for ICU and transport ventilators with their NIV algorithm turned on as well as for dedicated NIV ven-
tilators in the presence of inspiratory leaks (NIV1/Leaks1, black bars). The simulated inspiratory time
was 0.8 s (solid line). When insufflation time reached 1.6 s (dotted line) it corresponded to delayed
cycling. For ICU and transport ventilators, the introduction of inspiratory leaks led to an increase in
insufflation time when the NIV algorithm was turned off. This prolongation of insufflation due to
leaks was partly and unequally minimized by the NIV algorithm. #P , .05 vs NIV2/Leaks2. $P , .05
vs NIV2/Leaks1. See Figure 1 legend for expansion of abbreviation.

lending strength to the main results of this study, of leaks equivalent to that of the ICU ventilators
which are: in absence of leaks.
• Synchronization performance in the presence of
• In NIV conditions, most dedicated NIV ventila- leaks remains heterogeneous among ICU as well
tors allowed better patient-ventilator synchroni- as transport ventilators, and each machine should
zation than ICU and transport ventilators, even be considered individually.
when the NIV algorithm was engaged, especially • The NIV algorithm usually improved, at least
regarding the risk of auto-triggering. slightly, the triggering and/or cycling synchroni-
• Most of the dedicated NIV ventilators exhibited zation of ICU and transport ventilators in the
a synchronization performance in the presence presence of leaks.

Table 2—Clinical Study: Main Respiratory Parameters

Respiratory Parameters ICUniv2 ICUniv1 NIVv P Value

RRp, per min 29 (22-31) 27 (22-31) 26 (24-30) .982
Tip, ms 780 (599-914) 674 (558-957) 749 (629-923) .057
Tiexcess, % 14 (4-24) 12 (6-23) 13 (11-21) .344
Vte, mL 467 (269-633) 465 (322-548) 487 (278-539) .931
Vte, mL/kg 6.5 (4.3-9.4) 6.9 (4.6-8.3) 7.0 (4.6-9.0) .797
e, L/min 11.5 (8.7-15.5) 10.3 (9.2-16.7) 10.6 (8.6-14.0) .683
Leaks, L/min 6.3 (4.3-10.8) 6.2 (2.6-12.1) 7.3 (3.0-11.7) .947
Leaks, % e 55 (39-101) 47 (26-113) 81 (16-121) .612
Main respiratory parameters recorded throughout the three NIV sessions during the clinical study. ICUniv2 5 NIV session using an ICU
ventilator whose NIV algorithm has been turned off; ICUniv1 5 NIV session using an ICU ventilator whose NIV algorithm has been turned on;
NIVv 5 NIV session using a dedicated NIV ventilator; RRp 5 patient’s respiratory rate measured with the use of the electromyogram signal;
Tiexcess 5 percentage of insufflation time that exceeds the neural inspiratory time; Tip 5 patient’s neural inspiratory time; e 5 minute ventilation;
Vte 5 expired tidal volume. See Table 1 legend for expansion of other abbreviation.

372 Original Research

Downloaded From: http://journal.publications.chestnet.org/ on 08/11/2012

 dence of patient-ventilator asynchronies during NIV
in a clinical study involving 65 patients and five
ICU ventilators. Without the NIV algorithm engaged,
46% of the patients had an AI . 10%. The NIV algo-
rithm permitted a decrease in the incidence of asyn-
chronies due to leaks but without a decrease in the
overall incidence of patient-ventilator asynchronies
(38% vs 46%, P 5 .69), due to a high incidence of
asynchronies not directly related to leaks. We report
a lower proportion of patients exhibiting an AI . 10%
due to a lower incidence of some major asynchronies.
Several reasons explain this discrepancy. First, the
level of assistance in our study was lower than the one
observed in the study by Vignaux et al,14 leading to a
lower tidal volume, which might explain our low inci-
dence of ineffective efforts.20 Second, we have modi-
fied the definition of premature cycling, considering
that the previous one was too sensitive in terms of
clinical relevance and what can be considered as a
Figure 5. Clinical study asynchrony index during the three “major” patient-ventilator asynchrony. This definition
NIV sessions. The box plots represent the asynchrony index modification has automatically led to less recorded
(thick horizontal bar: median; extremities of the boxes: 25th
and 75th percentiles; thin horizontal bars: fifth and 95th per- premature cycling, so to a lower AI. Third, the ICU
centiles) during each 20-min NIV session: ICUNIV2, ICUNIV1, ventilators used in our clinical assessment had the
and NIVV. The asynchrony index was significantly lower with same behavior during our bench evaluation: a propen-
NIVV than with ICUNIV2 and ICUNIV1. ICUNIV2 5 ICU ventilator
with NIV algorithm turned off; ICUNIV1 5 ICU ventilator with sity to auto-triggering with expiratory leaks, but no
NIV algorithm turned on; NIVV 5 dedicated NIV ventilator. See delayed cycling in the presence of inspiratory leaks.
Figure 1 legend for expansion of abbreviation. Although the strength of our bench model was to
assess separately the impact of expiratory and inspira-
tory leaks on triggering and cycling synchronizations,
Patient-Ventilator Interactions During NIV
respectively, the originality of our clinical study was
Patient-ventilator asynchrony is frequent during to use ICU ventilators that had the same behavior
both invasive16,17 and noninvasive7,14 mechanical ven- during their bench evaluation. This led to intelligible
tilation. However, the respective proportion of each results and gave a mutual validation to the two assess-
type of major asynchrony markedly differs between ments. In the meantime, as a part of this behavior
these two techniques. During invasive mechanical ven- was to avoid delayed cycling, this logically led to a
tilation, ineffective effort represents the most preva- decrease in the overall AI during the clinical study
lent asynchrony.16,18 Its occurrence is largely favored as compared with previous studies conducted with
by overassistance and can frequently be avoided by other ventilators. Finally, an AI . 10% in our clinical
reducing the amount of support both in terms of tidal study was mainly related to a high incidence of auto-
volume and inspiratory time.19,20 By contrast, during triggering, which reflects the ventilator’s ability to
NIV, additional asynchronies, especially auto-triggering manage leaks rather than the relevance of the settings
and delayed cycling, are induced by the presence of chosen by the clinician.
leaks around the mask4,7 and reflect more the ventila- As with ICU ventilators, our bench evaluation also
tor’s ability to manage leaks than the settings chosen showed very uneven performances of transport ven-
by the clinician. Our bench study showed a wide var- tilators and their NIV algorithms in the presence of
iation in this ability among ICU ventilators and their leaks. Such heterogeneity has also been previously
NIV algorithms, which is consistent with previous reported with transport ventilators assessed in inva-
bench studies.12,13 More interestingly, our bench results sive conditions.21,22
were also well reproduced during our clinical study. On the whole, our results suggest that rather than
In fact, auto-triggering represented the most frequent being considered as belonging to a group of venti-
asynchrony with ICU ventilators used in the clinical lators, each ICU and transport ventilator should be
study, as predicted during their bench evaluation. Fur- examined individually regarding its ability to manage
thermore, there was a trend toward less asynchrony NIV conditions. By contrast, dedicated NIV venti-
with the NIV algorithm, which usually minimized lators exhibited more homogeneous behavior during
asynchronies during the bench study. Vignaux et al14 our bench evaluation, with an ability to avoid auto-
assessed the impact of the NIV algorithm on the inci- triggering or delayed cycling while keeping a short

journal.publications.chestnet.org CHEST / 142 / 2 / AUGUST 2012 373

Downloaded From: http://journal.publications.chestnet.org/ on 08/11/2012

 Figure 6. Clinical study incidence of each patient-ventilator asynchrony during the three NIV sessions. Each patient-ventilator asyn-
chrony is represented as box plots (thick horizontal bar: median; extremities of the boxes: 25th and 75th percentiles; thin horizontal bars:
fifth and 95th percentiles) for each 20-min NIV session: ICUNIV2, ICUNIV1, and NIVV. *P , .05 vs NIVV, †P , .05 vs auto-triggering. See
Figure 1 and 5 legends for expansion of abbreviations.

triggering delay despite the presence of leaks. This indication for NIV in ICU.23 Second, only one level of
is consistent with two previous bench studies that both inspiratory and expiratory leaks was designed.
showed a better synchronization ability of a dedicated These experimental conditions may not reproduce
NIV ventilator as compared with several ICU ventila- what happens in clinical conditions. However, our
tors without15 or with13 their NIV algorithm engaged. clinical study showed that our bench model succeeded
Our clinical study is the first to our knowledge to con- in capturing the kind of asynchronies that may occur
firm that the use of a NIV ventilator to perform NIV in the presence of leaks with each ventilator in the
in critically ill patients led to a significant decrease in clinical setting.
the incidence of patient-ventilator asynchrony.
Clinical Relevance
Limitations
It is currently unknown if patient-ventilator asyn-
Several limitations of this study should be under- chronies, especially those due to leaks, can affect
lined. First, during the bench study, only mild the clinical outcome of NIV and therefore influence
obstructive respiratory mechanics were simulated, as ventilator choice by clinicians. However, several argu-
respiratory mechanics are known to affect the cycling ments favor the best possible synchronization dur-
delay. Our aim was to uncover delayed cycling in the ing NIV. First, it seems reasonable to assume that
presence of inspiratory leaks, which could be mini- auto-triggering and delayed cycling will reduce the
mized in the case of restrictive respiratory mechanics.12 tolerance of the procedure, an important key to NIV
In addition, COPD represents the most recognized success.24,25 Second, the occurrence of delayed cycling

Table 3—Clinical Study Patients Presenting Each Type of Asynchrony With a High Incidence ( . 1/min) or an
Asynchrony Index . 10%

Type of Asynchrony ICUniv2 ICUniv1 NIVv P Value

Auto-triggering 5 (33) 5 (33) 0 .016
Double-triggering 0 1 (7) 0 …
Ineffective effort 0 0 0 …
Delayed cycling 0 0 0 …
Premature cycling 3 (20) 1 (7) 0 .097
At least one asynchrony 6 (40) 5 (33) 0 .012
Asynchrony index . 10% 4 (27) 2 (13) 0 .091
Data are presented as No. (%). See Table 1 and 2 legends for expansion of abbreviations.

374 Original Research

Downloaded From: http://journal.publications.chestnet.org/ on 08/11/2012

 can lead to dynamic hyperinflation and contribute Fisher-Paykel. Drs Lyazidi, Cordoba-Izquierdo, Thille, and Richard
and Ms Vignaux have reported that no potential conflicts of inter-
to the development of ineffective efforts,6,19 which est exist with any companies/organizations whose products or ser-
are associated with a prolongation of the ventilation vices may be discussed in this article.
during invasive mechanical ventilation.26 Given the Role of sponsors: The sponsor had no role in the study design,
the collection, analysis, interpretation of the data, the writing of
benefits of NIV when avoiding intubation,23,25,27,28 each the manuscript, or the decision to submit the paper for publication.
factor potentially involved in its success should logi- Other contributions: This work was performed at Réanimation
cally be promoted. However, if no patient exhibited Médicale, AP-HP, Centre Hospitalier Albert Chenevier-Henri
Mondor, Créteil, France and the Department of Intensive Care,
a high incidence of asynchrony with the NIV venti- Geneva University Hospital, Geneva, Switzerland.
lator in our study, just a few had an AI . 10% with Additional information: The e-Appendix and e-Figures can be
ICU ventilators. We cannot know to what extent found in the “Supplemental Materials” area of the online article.
this difference may be clinically relevant and fur-
ther clinical studies addressing the impact of dif- References
ferent devices on the outcome of different groups
1. Brochard L. Noninvasive ventilation for acute respiratory fail-
of patients under NIV are needed to formulate some ure. JAMA. 2002;288(8):932-935.
recommendations. 2. Demoule A, Girou E, Richard JC, Taillé S, Brochard L.
Increased use of noninvasive ventilation in French intensive
care units. Intensive Care Med. 2006;32(11):1747-1755.
Conclusion 3. Keenan SP, Mehta S. Noninvasive ventilation for patients
presenting with acute respiratory failure: the randomized
In conclusion, our study shows that dedicated NIV controlled trials. Respir Care. 2009;54(1):116-126.
4. Schettino GP, Tucci MR, Sousa R, Valente Barbas CS,
ventilators allow a better patient-ventilator synchrony Passos Amato MB, Carvalho CR. Mask mechanics and leak
in the presence of leaks than ICU and transport ven- dynamics during noninvasive pressure support ventilation:
tilators, even if their NIV algorithm is engaged, espe- a bench study. Intensive Care Med. 2001;27(12):1887-1891.
cially for what concerns auto-triggering. When using 5. Bernstein G, Knodel E, Heldt GP. Airway leak size in neo-
an ICU or transport ventilator to perform NIV, the nates and autocycling of three flow-triggered ventilators. Crit
Care Med. 1995;23(10):1739-1744.
NIV algorithm usually improves, at least slightly and 6. Calderini E, Confalonieri M, Puccio PG, Francavilla N, Stella L,
with variations among ventilators, triggering and/or Gregoretti C. Patient-ventilator asynchrony during noninva-
cycling synchronization. sive ventilation: the role of expiratory trigger. Intensive Care
Med. 1999;25(7):662-667.
7. Vignaux L, Vargas F, Roeseler J, et al. Patient-ventilator asyn-
Acknowledgments chrony during non-invasive ventilation for acute respiratory
Author contributions: Dr Carteaux is the guarantor of the paper. failure: a multicenter study. Intensive Care Med. 2009;35(5):
Dr Carteaux: contributed to the study design, patient enrollment, 840-846.
data collection, data analysis, data interpretation, and manuscript 8. Schettino G, Altobelli N, Kacmarek RM. Noninvasive positive-
preparation, and read and approved the final manuscript. pressure ventilation in acute respiratory failure outside clin-
Dr Lyazidi: contributed to the study design, patient enrollment, ical trials: experience at the Massachusetts General Hospital.
data collection, data analysis, data interpretation, and manuscript Crit Care Med. 2008;36(2):441-447.
preparation, and read and approved the final manuscript. 9. Kluge S, Baumann HJ, Kreymann G. Intrahospital transport
Dr Cordoba-Izquierdo: contributed to the study design, patient of a patient with acute exacerbation of chronic obstructive
enrollment, data collection, data analysis, data interpretation,
and manuscript preparation, and read and approved the final manu- pulmonary disease under noninvasive ventilation. Intensive
script. Care Med. 2005;31(6):886.
Ms Vignaux: contributed to the study design, patient enrollment, 10. Scala R, Naldi M. Ventilators for noninvasive ventilation
and data collection, and read and approved the final manuscript. to treat acute respiratory failure. Respir Care. 2008;53(8):
Dr Jolliet: contributed to the study design and patient enrollment, 1054-1080.
and read and approved the final manuscript. 11. Taylor DM, Bernard SA, Masci K, MacBean CE, Kennedy MP,
Dr Thille: contributed to the study design, data analysis, data inter- Zalstein S. Prehospital noninvasive ventilation: a viable treat-
pretation, and manuscript preparation, and read and approved ment option in the urban setting. Prehosp Emerg Care. 2008;
the final manuscript. 12(1):42-45.
Dr Richard: contributed to reading and approving the final manu-
script. 12. Vignaux L, Tassaux D, Jolliet P. Performance of noninvasive
Dr Brochard: contributed to the study design, data analysis, data ventilation modes on ICU ventilators during pressure sup-
interpretation, and manuscript preparation, and read and approved port: a bench model study. Intensive Care Med. 2007;33(8):
the final manuscript. 1444-1451.
Financial/nonfinancial disclosures: The authors have reported 13. Ferreira JC, Chipman DW, Hill NS, Kacmarek RM. Bilevel
to CHEST the following conflicts of interest: Dr Carteaux vs ICU ventilators providing noninvasive ventilation: effect of
received reimbursement from Covidien for expenses relative to system leaks: a COPD lung model comparison. Chest. 2009;
travel, accommodation, and registration for the 2010 American 136(2):448-456.
Thoracic Society annual meeting for a specific presentation that 14. Vignaux L, Tassaux D, Carteaux G, et al. Performance of
did not concern this study. Dr Jolliet received a research grant
from MAQUET GmbH & Co KG and runs a laboratory that noninvasive ventilation algorithms on ICU ventilators during
received research grants for specific research projects from pressure support: a clinical study. Intensive Care Med. 2010;
Dräger and ResMed. Dr Brochard runs a laboratory that received 36(12):2053-2059.
research grants for specific research projects from Dräger, General 15. Miyoshi E, Fujino Y, Uchiyama A, Mashimo T, Nishimura M.
Electric Company, MAQUET GmbH & Co KG, Covidien, and Effects of gas leak on triggering function, humidification, and

journal.publications.chestnet.org CHEST / 142 / 2 / AUGUST 2012 375

Downloaded From: http://journal.publications.chestnet.org/ on 08/11/2012

 inspiratory oxygen fraction during noninvasive positive airway 23. Brochard L, Mancebo J, Wysocki M, et al. Noninvasive ven-
pressure ventilation. Chest. 2005;128(5):3691-3698. tilation for acute exacerbations of chronic obstructive pul-
16. Thille AW, Rodriguez P, Cabello B, Lellouche F, Brochard L. monary disease. N Engl J Med. 1995;333(13):817-822.
Patient-ventilator asynchrony during assisted mechanical venti- 24. Antonelli M, Conti G, Moro ML, et al. Predictors of failure
lation. Intensive Care Med. 2006;32(10):1515-1522. of noninvasive positive pressure ventilation in patients with
17. Chao DC, Scheinhorn DJ, Stearn-Hassenpflug M. Patient- acute hypoxemic respiratory failure: a multi-center study.
ventilator trigger asynchrony in prolonged mechanical venti- Intensive Care Med. 2001;27(11):1718-1728.
lation. Chest. 1997;112(6):1592-1599. 25. Carlucci A, Richard JC, Wysocki M, Lepage E, Brochard L;
18. Tobin MJ, Jubran A, Laghi F. Patient-ventilator interaction. SRLF Collaborative Group on Mechanical Ventilation. Non-
Am J Respir Crit Care Med. 2001;163(5):1059-1063. invasive versus conventional mechanical ventilation. An
19. Tassaux D, Gainnier M, Battisti A, Jolliet P. Impact of expira- epidemiologic survey. Am J Respir Crit Care Med. 2001;
tory trigger setting on delayed cycling and inspiratory muscle 163(4):874-880.
workload. Am J Respir Crit Care Med. 2005;172(10):1283-1289. 26. de Wit M, Miller KB, Green DA, Ostman HE, Gennings C,
20. Thille AW, Cabello B, Galia F, Lyazidi A, Brochard L. Epstein SK. Ineffective triggering predicts increased dura-
Reduction of patient-ventilator asynchrony by reducing tidal tion of mechanical ventilation. Crit Care Med. 2009;37(10):
volume during pressure-support ventilation. Intensive Care 2740-2745.
Med. 2008;34(8):1477-1486. 27. Girou E, Schortgen F, Delclaux C, et al. Association of non-
21. Chipman DW, Caramez MP, Miyoshi E, Kratohvil JP, invasive ventilation with nosocomial infections and survival
Kacmarek RM. Performance comparison of 15 transport in critically ill patients. JAMA. 2000;284(18):2361-2367.
ventilators. Respir Care. 2007;52(6):740-751. 28. Nourdine K, Combes P, Carton MJ, Beuret P, Cannamela A,
22. Zanetta G, Robert D, Guérin C. Evaluation of ventilators used Ducreux JC. Does noninvasive ventilation reduce the ICU
during transport of ICU patients — a bench study. Intensive nosocomial infection risk? A prospective clinical survey.
Care Med. 2002;28(4):443-451. Intensive Care Med. 1999;25(6):567-573.

376 Original Research

DownloadedFrom:http://journal.publications.chestnet.org/on08/11/2012