Respiratory Physiology & Neurobiology 203 (2014) 82–89

Contents lists available at ScienceDirect

Respiratory Physiology & Neurobiology

journal homepage: www.elsevier.com/locate/resphysiol

Physiologic comparison of neurally adjusted ventilator assist,

proportional assist and pressure support ventilation in critically ill
patients
Evangelia Akoumianaki, George Prinianakis, Eumorfia Kondili, Polychronis Malliotakis,
Dimitris Georgopoulos ∗
Intensive Care Medicine Department, University Hospital of Heraklion, Medical School, University of Crete, Heraklion, Crete, Greece

a r t i c l e i n f o a b s t r a c t

Article history: To compare, in a group of difficult to wean critically ill patients, the short-term effects of neurally
Accepted 19 August 2014 adjusted ventilator assist (NAVA), proportional assist (PAV+) and pressure support (PSV) ventilation on
Available online 26 August 2014 patient–ventilator interaction. Methods: Seventeen patients were studied during NAVA, PAV+ and PSV
with and without artificial increase in ventilator demands (dead space in 10 and chest load in 7 patients).
Keywords: Prior to challenge addition the level of assist in each of the three modes tested was adjusted to get the
Inspiratory effort
same level of patient’s effort. Results: Compared to PSV, proportional modes favored tidal volume vari-
Patient–ventilator asynchrony
ability. Patient effort increase after dead space was comparable among the three modes. After chest load,
Transdiaphragmatic pressure
patient effort increased significantly more with NAVA and PSV compared to PAV+. Triggering delay was
significantly higher with PAV+. The linear correlation between tidal volume and inspiratory integral of
transdiaphragmatic pressure (PTPdi) was weaker with NAVA than with PAV+ and PSV on account of a
weaker inspiratory integral of the electrical activity of the diaphragm ( EAdi)–PTPdi linear correlation

during NAVA [median (interquartile range) of r2 , determination of coefficient, 16.2% (1.4–30.9%)]. Con-
clusion: Compared to PSV, proportional modes favored tidal volume variability. The weak EAdi–PTPdi
linear relationship during NAVA and poor triggering function during PAV+ may limit the effectiveness of
these modes to proportionally assist the inspiratory effort.
© 2014 Elsevier B.V. All rights reserved.

1. Introduction proportional modes, such as neurally adjusted ventilator assist

Assisted mechanical ventilation is a first line ventilatory strategy
in critically ill patients able to undertake part of their respiratory
work (Tobin, 2003). A priority of assisted ventilation is to ensure
adequate blood gas exchange, harmonious patient–ventilator
interaction and sufficient unloading of the respiratory muscles
(Kondili et al., 2003). Accomplishment of these targets is often com-
plex and represents one of the greatest challenges in the treatment
of partially supported patients.
Pressure support ventilation (PSV) is a frequently used mode
of assisted mechanical ventilation. Nevertheless, with this mode
an optimal level of assistance and/or a perfect patient–ventilator
synchrony may not be achieved (Kondili et al., 2003). Moreover,
ventilator assistance remains constant, a drawback in the face of
changing patient’s ventilatory demands. This is not the case with
∗ Corresponding author. Tel.: +30 2810392636; fax: +30 2810392636.
E-mail address: georgop@med.uoc.gr (D. Georgopoulos).
(NAVA) (Sinderby et al., 1999) and proportional assist ventilation
(PAV) (Younes, 1992), which permit the patient to modify the
level of ventilator assistance. NAVA uses the signal of the electri-
cal activity of the diaphragm (EAdi) to trigger the ventilator, guide
pressure delivery and terminate mechanical inspiration (Sinderby
et al., 1999). PAV delivers pressure proportional to the instanta-
neous flow (V ) and volume (V) and, hence, to respiratory muscles’
pressure, provided that respiratory system elastance and resistance
are known (Younes, 1992). The latter are automatically calculated
by a ventilator software (PAV+, PAV with load adjustable gain fac-
tors) (Younes et al., 2001a,b; Kondili et al., 2006).
Several studies demonstrated that, compared to PSV, both PAV
(Alexopoulou et al., 2013; Bosma et al., 2007; Xirouchaki et al.,
2008) and NAVA (Colombo et al., 2008) improve patient–ventilator
synchrony and unload the respiratory muscles without the risk
of over-assistance. Nevertheless, a direct comparison of the three
modes, PAV+, NAVA and PSV, has never been performed. In addition
there are limited data on the performance of proportional modes in
difficult to wean patients. The latter represents a challenging group

http://dx.doi.org/10.1016/j.resp.2014.08.012
1569-9048/© 2014 Elsevier B.V. All rights reserved.
 E. Akoumianaki et al. / Respiratory Physiology & Neurobiology 203 (2014) 82–89
of patients in whom different ventilatory strategies might have sig-
nificant impact on weaning outcome. The aim of this study was to
examine, in a group of difficult to wean critically ill patients, the
breathing pattern and patient effort during PSV, NAVA and PAV+
at resting conditions and after an acute artificial increase in venti-
lator demand. In addition, because the ability of a ventilator mode
to meet patient demand would require that tidal volume, which is
partly determined by airway pressure, will match patient inspira-
tory effort, we examined the correlation between tidal volume with
indices of inspiratory efforts such as transdiaphragmatic pressure
(Pdi) and EAdi.
2. Methods (see also supplemental digital content, SDC)
2.1. Patients
Patients with difficult or prolonged weaning (Boles et al., 2007)
were studied. PSV, NAVA and PAV+ were applied in all patients
before and after an artificial increase in respiratory demand. The
study was conducted in two time periods: In the first, the respira-
tory demand in all eligible patients was increased by the addition
of 150 ml dead space in the ventilator circuit (dead space group)
(Ranieri et al., 1985). Patients enrolled during the second period
were tested before and after application of sandbags to the entire
surface of the chest and abdominal wall (load group) (Kondili et al.,
2006). The study was approved by the hospital ethics committee
and informed consent was obtained from the patients or their fam-
ilies.
2.2. Measurements
V , V and airway (Paw), esophageal (Pes), gastric (Pgas) and
transdiaphragmatic (Pdi = Pgas − Pes) pressures were measured as
described previously (Kondili et al., 2006, 2010). EAdi was recorded
using the NAVA electrodes attached to a nasogastric tube (Maquet,
Solna, Sweden) (Colombo et al., 2008). The proper position of the
eosphageal and gastric balloons (to measure Pes and Pgas, respec-
tively), and nasogastric tube (to record EAdi) was verified using
standard tests and procedures (D‘Angelo et al., 1991; Kondili et al.,
2006). Each signal was sampled at 150 Hz and stored on a computer
disk for later analysis.
2.3. Study protocol
The patients were studied in semi-recumbent position (>30◦ ).
Initially, their nasogastric tubes were replaced by a nasogastric tube
equipped with NAVA electrodes (16 French diameter, 125 cm long;
Maquet Critical Care, Sölna, Sweden) and the gastric content was
drained. Subsequently, thin, latex balloon-tipped catheter systems
were inserted into the middle third of the esophagus and stomach
to measure Pes and Pgas, respectively.
Two ventilators were used for the purpose of the study: (1) a
Servo-i® (Maquet Critical Care, Solna, Sweden) to deliver NAVA
and PSV and (2) a Puritan Bennett 840 (Covidien, USA) to deliver
PAV+. Inspired fraction of oxygen (FiO2 ), flow trigger sensitivity
and external positive end-expiratory pressure (PEEP) selected by
the primary physician prior to patient inclusion on the baseline
mode (PAV+ or PSV), remained unaltered throughout the study. Pdi
swings (Pdiswings ) on baseline mode were used to adjust the ventila-
tor assist level, such as to obtain comparable steady-state Pdiswings
among the three modes. In each ventilator mode the patients
underwent two 20-min sessions, without (no challenge) and with
(challenge) artificial increase in ventilatory demands. Arterial gas
samples were obtained at the end of each experimental condition
(see Fig. S1 in SDC for study protocol).
83
2.4. Data analysis
The last 3 min of each 20-min period were analyzed and
averaged to give the breath variables corresponding to each exper-
imental condition. Breaths with low quality Pdi signal (i.e. cough,
peristaltic waves on Pes signal) were excluded. Pdi were reported as
changes from end-expiratory value (Kondili et al., 2006). Since with
NAVA and PSV EAdi and Pdi were recorded by two different systems
(Maquet and Windaq, respectively), an end-inspiratory pause was
performed prior to NAVA or PSV recordings, to allow the offline syn-
chronization of the two recordings using Paw and V as reference
signals (see Fig. S2). Synchronization of EAdi cycles to Pdi cycles was
not feasible with PAV+ as we could not acquire Paw and V wave-
forms from ‘Maquet system’ when the patients were ventilated on
PAV+ (see SDC).
To evaluate the breathing pattern, inspiratory tidal volume (VT ),
patient respiratory rate (RR), minute ventilation (VE ), Pdi-derived
(TiPdi ) and flow-derived (Timech ) inspiratory times, Pdi-derived
total respiratory cycle duration (Ttot) and the duty cycle (TiPdi /Ttot)
were measured as previously described (see Fig. S3) (Kondili et al.,
2006, 2010). Breathing pattern variability was expressed as the
coefficient of variation of VT and RR.
Patient inspiratory effort per breath was quantified by measur-
ing the pressure-time area under the Pdi signal (PTPdi/b) between
the onset of inspiratory effort and the end of neural inspiration, i.e.
from the beginning of Pdi increase to the point at which Pdi started
to decline rapidly (Kondili et al., 2006; Spahija et al., 2010; Fig. S3).
PTPdi per minute (PTPdi/min) and per liter of minute ventilation
(PTPdi/L) were also calculated. The inspiratory effort dissipated to
trigger the ventilator (PTPditrig ) was estimated as the time integral
of Pdi from the beginning of inspiration to the nadir value of Paw.
The rate of rise of Pdi (dPdi/dt) and the level of intrinsic PEEP (PEEPi)
were also computed (Kondili et al., 2006, 2010).
In addition to Pdi-derived indices, the corresponding EAdi-
derived indices were computed [EAdi inspiratory time (TiEAdi , Fig.
S3), EAdiswings , the integral of EAdi over TiEAdi (ʃEAdi/b), ʃEAdi/min,
ʃEAdi/L and dEAdi/dt].
Respiratory system elastance and resistance were measured
with and without the challenge using the automatic technique of
end-inspiratory airway occlusion during PAV+ mode (Kondili et al.,
2006).
Asynchrony, although not the primary aim of the study, was also
computed during the 3 min recording period (Georgopoulos et al.,
2006) (see SDC).
2.5. Statistical analysis
Data were analyzed using non-parametric tests. Continuous
variables were expressed as medians (25–75th interquartile range,
IQR). Variables were compared using the Friedman test for repeated
measurements followed, when indicated, by a pairwise comparison
with Wilcoxon signed rank using Bonferroni post hoc correc-
tion. Categorical variables were compared using the Fisher’s exact
test.
Linear regression analysis was used and the coefficient of deter-

mination (r2 ) was calculated to examine the relationship
 among
EAdi/b, PTPdi/b and VT . The relationship between EAdi/b and
VT could not be examined during PAV+ since the synchroniza-
tion of EAdi and Pdi cycles was not feasible (see SDC). In the
dead space group and for a given mode the regression analy-
sis was performed using all analyzed breaths (with and without
dead space); each patient had one value of r2 for any exam-
ined correlation at each tested mode. In the load group, since
respiratory mechanics were altered by the load, the correlation
analysis was performed separately for breaths without and sep-
arately for breaths with load application and each patient had two
84 E. Akoumianaki et al. / Respiratory Physiology & Neurobiology 203 (2014) 82–89

Table 1
Patients’ characteristics.

Pt Sex Age (years) Diagnosis APACHE II Days on MV PaO2 /FiO2 Assistance level at each mode tested Flow trig PEEP
(lt/min) (cmH2 O)
NAVA PAV+(%) PSV(cmH2 O)
(␮V/cmH2 O)

1 M 80 Trauma ARDS 21 8 166 2.0 60 12 2 9

2 M 68 Sepsis 21 36 325 1.5 50 12 3 7
3 M 64 AECOPD 22 5 152 0.7 40 8 3 8
4 M 68 ARDS 11 7 154 0.9 50 10 2 5
5 M 83 CNS damage 19 20 271 0.8 50 8 3 6
6 F 66 Pneumonia ARDS 18 13 243 0.6 30 6 3 7
7 M 72 CHF Sepsis 24 5 220 1.3 65 18 2 7
8 M 73 Sepsis ARDS 18 14 196 0.9 45 8 2 7
9 M 72 Sepsis ARDS 10 19 147 2.0 70 15 2 5
10 M 72 Sepsis ARDS 20 14 180 0.6 60 12 2 7
11 M 77 Sepsis ARDS 25 16 156 1.0 60 10 2 5
12 F 70 AECOPD 8 13 230 1.6 35 7 2 7
13 M 25 Trauma ARDS 22 7 277 0.8 20 5 3 8
14 M 74 Trauma ARDS 10 6 146 0.3 25 5 3 8
15 M 61 Trauma ARDS 14 16 200 1.2 55 9 2 7
16 M 30 Trauma ARDS 15 26 304 0.5 40 6 2 7
17 F 65 Sepsis ARDS 23 12 140 0.6 45 15 2 10

Median (IQR) 70 (65–73) 19 (14–22) 13 (17–16) 196 0 50 (40–60) 9 (7–12) 7 (7–8)

(154–243) .9 (0.6–1.3)

PSV assistance refers to cmH2 O levels above PEEP M, male; F, female; ARDS, acute respiratory distress syndrome; AECOPD, acute exacerbation of chronic obstructive
pulmonary disease; CNS, central nervous system; CHF, chronic heart failure; APACHE II, simplified acute physiology score II; MV, mechanical ventilation; PSV, pressure
support ventilation; PAV+, proportional assist ventilation with load adjustable gain factors; PEEP, positive end-expiratory pressure; IQR, interquartile range.

r2 for a given correlation and a given mode (one with and one
without challenge) (see SDC for further details). Statistical tests
were two-sided and a p < 0.05 was considered statistically signif-
icant.
3. Results
3.1. Patient characteristics
Twenty-two patients were enrolled; 13 during the ‘dead space
period’ and 9 during the ‘load period’. In 5 patients, 3 in the dead
space group and 2 in the load group, the study was interrupted and
these patients were not analyzed. The reason for study interruption
was respiratory distress in all 3 patients in the dead space group. In
two of them respiratory distress developed before dead space addi-
tion and persisted in all three modes. In a female COPD patient dead
space addition during PSV ventilation resulted in respiratory dis-
tress along with respiratory acidosis. One patient in the load group
developed excessive coughing immediately after catheter insertion
while in another one load application led to tachypnea, diaphore-
sis and contraction of accessory respiratory muscles during NAVA.
Change to PSV or PAV+ did not ameliorate symptoms and the study
was interrupted.
The main characteristics and baseline ventilator settings of the
17 patients who completed the study are summarized in Table 1.
3.3. Breathing pattern and gas exchange
Table 2 illustrates the results on breathing pattern and gas
exchange before and after challenge. Apart from a lower baseline VT
and a longer Timech during PAV+ no other differences were observed
among modes before and after challenge. Since TiPdi did not differ
among the three modes the longer Timech reflected a longer pres-
sure delivery time with PAV+. The coefficient of variation of VT was
significantly higher with the proportional modes compared to PSV.
No effect of mode on RR variability was observed.
3.4. Inspiratory effort
Without the challenge there was no statistically significant dif-
ference in the indices of inspiratory effort among modes, indicating
satisfactory titration of baseline ventilator settings (Table 3). Dead
space and chest load addition significantly increased patient effort.
The magnitude of this effect was not affected by the mode of sup-
port; neither the absolute values of muscle effort indices nor their
absolute change compared to baseline values, differed significantly
among modes (Table 3). However, when the changes in PTPdi/b and
PTPdi/min are expressed as a % difference from their baseline value,
the % increase in the load group was significantly lower during PAV+
than during NAVA and during NAVA and PSV respectively (Fig. 1).
Separate analysis of the two groups with regard to breathing
pattern and diaphragmatic inspiratory effort is provided in the
online supplement (SDC).

3.2. Response to type of challenge
The type of challenge had different influence on the two groups
tested. Dead space (but not chest load) addition significantly
increased VT and VE (Table S1, SDC). Elastic load application sig-
nificantly decreased respiratory system compliance [median (IQR)
decrease of 22% (17–32%)] and significantly increased respiratory
system resistance [median (IQR) increase of 13% (5–21%)]. Dead
space had no significant effect on respiratory mechanics (Table S2,
SDC).
3.5. Correlation between VT and inspiratory effort

During NAVA, EAdi/b was always significantly positively
correlated with VT . This was not the case during PSV: 22% of correla-

tions tested were not significantly linear. During NAVA the median
(IQR) determination coefficient (r2 ) and slope of EAdi/b–VT rela-
tionship were significantly higher than the corresponding values
with PSV (Table 4).
PTPdi/b was significantly positively correlated with VT in
87.5% of PAV+ recordings and in 83.3% of PSV recordings. The
 E. Akoumianaki et al. / Respiratory Physiology & Neurobiology 203 (2014) 82–89 85

Table 2
Breathing pattern indices and arterial blood gases in the 17 patients tested.

Challenge No Yes

Mode NAVA PAV+ PSV NAVA PAV+ PSV

VT (ml) 503 (446–566) 471* , (422–527) 501 (475–539) 537 (469–622) 519** (466–589) 495 (447–615)
RR (br/min) 26 (20–29) 28 (20–30) 27 (19–30) 28 (22–32) 28 (22–30) 29 (22–32)
Mean paw (cmH2 O) 12.25 (10.63–15.56) 11.98 (8.54–14.74) 12.93 (9.75–16.25) 13.3** (11.8–17.16) 13.37** (9.75–16.17) 13.04 (9.98–16.19)
VE (lt/min) 11.7 (9.7–14.5) 10.5 (8.7–15.2) 11.2 (9.5–14.9) 14.4** (11.6–16.3) 11.7** (10.5–17.2) 14.2 (10.5–16.9)
TiPdi (ms) 609 (534–795) 664 (566–822) 606 (529–836) 615 (537–754) 648 (543–809) 586 (525–776)
Timech (ms) 723 (667–950) 798* , (688–1002) 690 (611–938) 667 (634–882) 751 (638–905) 659 (580–840)
TiEAdi (msec) 673 (577–857) 692 (576–830) 652 (572–793) 619 (564–820) 652 (584–823) 627 (578–745)
TiPdi /Ttot (%) 25.2 (22.8–30.8) 26.9 (24.5–30.6) 25.9 (22.2–30.1) 28.4 (23.9–31) 28 (24.3–32.3) 28.9 (23.6–34.2)
CV VT (%) 11.3 (9.7–14.5) 8.1 (6.5–13.5) 6.8† (3.6–9.4) 11.9 (8.4–19) 11.0 (6.2–17.3) 6.6† , (3.6–11.3)
CV RR (%) 10.6 (6.0–16.8) 10.0 (7.1–12.6) 10.0 (5.7–17.9) 9.5 (6.8–12.1) 13.3 (7.6–16) 9.0 (6.3–14)
PaO2 (mmHg) 90 (78–110) 90 (76–113) 86 (77–114) 87 (73–118) 84 (76–104) 89 (76–111)
PaCO2 (mmHg) 40 (38–52) 44 (41–53) 41 (35–55) 41 (36–55) 44 (39–55) 39 (35–56)
Ph 7.38 (7.36–7.42) 7.39 (7.33–7.40) 7.38 (7.36–7.42) 7.39 (7.36–7.41) 7.38 (7.35–7.41) 7.38 (7.34–7.42)

Data are median (interquartile range). VT , tidal volume; RR, patient respiratory rate; VE , minute ventilation; TiPdi , Pdi based inspiratory time; Timech , mechanical inspiratory
time; TiEAdi , time from EAdi increase to its peak value; CV; coefficient of variation.
*
p < 0.05, PAV+ vs. NAVA (at equivalent challenge).
§
p < 0.05, PAV+ vs. PSV (at equivalent challenge).
†
p < 0.05, PSV vs. NAVA (at equivalent challenge).
**
p < 0.05, challenge no vs. challenge yes for the same mode.

Table 3
Patient effort indices in the 17 patients tested.

Challenge No Yes

Mode NAVA PAV+ PSV NAVA PAV+ PSV

Pdiswings (cmH2 O) 8.5 (5.8–14.3) 8.8 (7.8–16.4) 8 (6.4–13.6) 11 (7–18.8) 11.8 (8.9–21.9) 12.1 (8.0–19.8)
PTPdi/b (cmH2 O s) 3.2 (2.3–4.9) 4.6 (3–5.9) 3.5 (2.5–4.8) 5 (2.6–6.5) 5.4 (3.2–6.7) 5.5 (1.8–7.1)
PTPdi/min (cmH2 O.s/min) 68.5 (56.4–115) 101.2 (74.4–137.8) 79.8 (59–141.4) 111.7 (75.5–167.7) 129.9 (81.4–178.6) 118.8 (68.5–210.7)
PTPdi/L (cmH2 O s/L) 6.39 (5.05–9.34) 9.48 (7.72–11.95) 7.95 (5.0–10.06) 8.75 (5.04–10.99) 11.60 (7.56–12.91) 10.94 (7.40–13.18)
 swings (␮V)
EAdi 17 (8–26.6) 17.3 (9–31.8) 15.6 (7.7–24.6) 18.2 (10.7–34.8) 19.6 (10.8–30.8) 18.1 (9.6–37.8)
EAdi/br (␮V s) 5.3 (3.2–8.4) 4.9 (3.1–10.2) 5.5 (2.8–8.2) 5.5 (4.5–12.5) 7.1 (3.9–10.2) 6.6 (4.3–12.5)
dPdi/dt (cmH2 O/s) 15.8 (10.3–32.8) 19.6 (13.1–34.2) 20.1 (10.5–38) 26.2 (11.8–45.1) 21.8 (17.1–49) 19 (13.8–44)
dEAdi/dt (␮V/s) 31.4 (14.4–39.4) 29.9 (18.9–45) 32.8 (13.2–43.4) 32.9 (21.5–55.2) 32.6 (21.9–49.8) 34.8 (19.3–60.3)
PEEPi (cmH2 O) 0.1 (0.0–1.0) 0.5 (0.0–1.3) 0.4 (0.0–1.4) 0.7 (0.0–1.6) 1.4 (0.0–2.5) 0.8 (0.4–2.2)
PTPditrig (cmH2 O s) 0.13 (0.04–0.29) 0.30* , (0.14–0.42) 0.09 (0.06–0.16) 0.30 (0.08–0.44) 0.36 (0.20–0.76) 0.13 (0.07–0.36)
% PTPditrig 4.4% (2.4–6.3%) 8.5% (3.7–12.8%) 3.3% (2.1–4.8%) 5.4% (4.0–12.5%) 9.3% (5.6–15.6%) 4% (2.5–6.9%)

EAdi, electrical activity of the diaphragm; Pdiswings , positive inflections of Pdi per breath; PTPdi/b, Pdi time integral per breath; PTPdi/min, PTPdi/b multiplied by the RR;
PTPdi/L, PTPdi/b multiplied by minute ventilation; EAdiswings ; positive inflections of EAdi per breath; ʃEAdi/b; time integral of EAdi per breath; dPdi/dt; rate of rise in Pdi;
dEAdi/dt; rate of rise in EAdi, PEEPi; intrinsic positive end expiratory pressure, PTPditrig ; transdiaphragmatic pressure time product for triggering, %PTPditrig ; percentage of
PTPdi/b lost for triggering.
*
p < 0.05, PAV+ vs. NAVA (at equivalent challenge).
§
p < 0.05, PAV+ vs. PSV (at equivalent challenge).
†
p < 0.05, PSV vs. NAVA (at equivalent challenge).

Table 4  
Determination coefficient (r2 ) and slope of VT –PTPdi/b, VT – Eadi/b and PTPdi/b– Eadi/b correlation during NAVA, PAV+ and PSV.

Mode NAVA PAV+ PSV

2 2
Relationship r Slope r Slope r2 Slope


VT /PTPdi/b (l/cmH2 O s) 0.26 (0.03–0.60) 0.03 (0.01–0.07) 0.58* (0.43–0.74) 0.05 (0.03–0.08) 0.58† (0.30–0.77) 0.03 (0.02–0.06)
0.32† (0.06–0.58) 0.01† (0.006–0.02)
VT /

Eadi/b (l/␮V s) 0.68 (0.49–0.75) 0.03 (0.01–0.05) NA NA NA NA
PTPdi/b/ Eadi/b [(cmH2 O.s)/(␮V s)] 0.16 (0.02–0.31) 0.12 (0.01–0.35) NA NA NA NA 0.29 (0.04–0.51) 0.18† (0.08–0.45)

VT , tidal volume; PTPdi/b, transdiaphragmatic pressure time integral per breath; ʃEAdi/b, time integral of diaphragmatic electrical activity per breath; NA, non applicable.
*
p < 0.05, PAV+ vs. NAVA.
†
p < 0.05, PSV vs. NAVA.

percentage of NAVA recordings exhibiting significant positive lin-
ear PTPdi/b–VT correlation was lower (54.2%, P < 0.02 NAVA vs.
PAV+, Fisher exact test). This finding was in consistence with
the absence of a significantly positive linear correlation between
EAdi/b and PTPdi/b in 54.2% of cases during NAVA. Since with

NAVA, EAdi drives the ventilator, the relationship between EAdi/b
and PTPdi/b influenced the relationship between PTPdi/b and VT ; r2
of PTPdi/b–VT relationship with NAVA was significantly lower than
with PAV+. Fig. 2 illustrates a patient in whom a negative linear
correlation between ʃEAdi/br and PTPdi/br was observed during
NAVA.
3.6. Secondary outcomes
Triggering delay (median, IQR) with PAV+ both before (160 ms,
120–200 ms) and after (170 ms, 130–210 ms) challenge was sig-
nificantly higher than the corresponding values during NAVA
[110 ms without challenge, (70–140 ms) 120 ms with challenge
86 E. Akoumianaki et al. / Respiratory Physiology & Neurobiology 203 (2014) 82–89

Fig. 1. % Changes in transdiaphragmatic pressure swings (A, Pdiswings ), transdiaphragmatic pressure time product per breath (B, PTPdi/b) and transdiaphragmatic pressure
time product per minute (C, PTPdi/min) in the group of patients with chest load challenge, dead space challenge and in the whole population studied. *p < 0.05, PAV+ vs.
NAVA. §p < 0.05, PAV+ vs. PSV.

(70–150 ms)] and PSV [110 ms without challenge, (80–140 ms)
110 ms with challenge (90–170 ms)]. Accordingly, the PTPditrig was
significantly higher during PAV+ (around 9% of PTPPdi/b lost for
triggering) independent of the presence of challenge and (Table 3).
4. Discussion
The main findings of our study were: (1) Compared to PSV,
proportional modes favored a more variable VT . (2) Load addition
resulted in an increase in inspiratory effort which, when expressed
as a percentage, was lower with PAV+ than that with NAVA and PSV,
while dead space application increased respiratory effort to a simi-
lar extent among modes. (3) The linear correlation between VT and
PTPdi/b was weaker with NAVA than with theother two modes on
account of a non significantly linear PTPdi/b– EAdi/b relationship
in several patients.
Two methodological points are worth mentioning. In com-
parative studies of ventilator modes, it is crucial to ensure that
equivalent levels of assistance are applied. In most studies compar-
ing PAV or NAVA to PSV, assistance between modes was matched
either arbitrarily (Delaere et al., 2003; Ranieri et al., 1996) or accord-
ing to peak Paw (Colombo et al., 2008; Piquilloud et al., 2011;
Spahija et al., 2010), mean Paw (Costa et al., 2011; Varelmann
et al., 2005; Wrigge et al., 1999) or minute ventilation (Coisel et al.,
2010). However, ventilator assistance based on the above variables
may differ considerably among modes making the interpretation
of the results complicated. Notwithstanding that the ideal titration
between different ventilator modes remains unclear, we argued
that patient effort, as expressed with Pdiswings , would more ade-
quately reflect comparable levels of support. Secondly, one cannot
preclude that the presence of an esophageal balloon next to the
NAVA catheter may influence the acquisition of EAdi and Pes sig-
nals. However, no difficulties were encountered in patients enrolled
in the study (including the patients not completing the protocol)
with respect to the acquisition of the EAdi signal: an adequate EAdi
signal, fulfilling all the quality criteria suggested in the literature,
was obtained within 10 min after study inclusion in all patients.
4.1. Breathing variability and gas exchange
In critically ill patients breathing pattern variability is associated
with improvement in gas exchange and respiratory mechanics and
weaning success (Mutch et al., 2000; Wysocki et al., 2006). Although
our study showed that proportional modes increase VT variabil-
ity, arterial blood gas exchange remained unaffected. In agreement
with the literature, we believe that the short-term nature of our
study (20 min) may account for this observation (Colombo et al.,
 E. Akoumianaki et al. / Respiratory Physiology & Neurobiology 203 (2014) 82–89 87

Fig. 2. Representative example of a patient ventilated with NAVA, who exhibited dissociation between the electrical activity of the diaphragm (EAdi) as captured by the
NAVA catheter and transdiaphragmatic pressure (Pdi). (A) Pdi, EAdi, airway pressure (Paw) and flow (V ) as a function of time. Notice that there is negative correlation
between changes in EAdi and Pdi (shaded gray area). The first cycle (first series of arrows) in the gray shaded area has an EAdi swing of 22.4 ␮V, a Pdi swing of 22.7 cmH2 O
and a delivered tidal volume (VT ) of 494 ml (not shown). In the last cycle of this area (second series of arrows) EAdi swing increased to 52.6 ␮V resulting in a VT of 672 ml
while Pdi swing was 17 cmH2 O, considerably lower than the first cycle. B: Relationship between: (1) VT and EAdi integral per breath (ʃEAdi/br), (2) Pdi integral (PTPdi/br)
and ʃEAdi/br and (3) VT and PTPdi/br. Notice that although VT closely follows ʃEAdi/br, there is a negative correlation between ʃEAdi/br and PTPdi/br influencing the VT and
PTPdi/br correlation. r2 = coefficient of determination.

2008; Kondili et al., 2006; Spahija et al., 2010; Xirouchaki et al.,
2008).
4.2. Indices of inspiratory effort and ventilator efficiency
Dead space application resulted in comparable effects in patient
effort in all modes. On the other hand, PAV+ responded better to
elastic load, resulting in less % increase in PTPdi/min than PSV
and NAVA. This finding is not unexpected taken into account the
operational principles of this mode. We observed a significant dete-
rioration in respiratory system mechanics in the load group while
dead space, increased significantly VT and VE but did not influence
respiratory system mechanics. Even if patient’s effort remains sta-
ble, an increase in respiratory system’s elastance and/or resistance
results in higher pressure assistance with PAV+ (since the % of assis-
tance is referred to higher values of elastance and/or resistance).
On the other hand, PSV provides the same pressure irrespective
of patient demands or mechanics while NAVA responds solely to
changes in patient effort. Hence, in NAVA, respiratory mechanics
deterioration, if not accompanied by an increase in neural effort,
will not lead to higher assistance (Kondili et al., 2006). This opera-
tional difference could explain why a better response to challenge
with PAV+ became apparent only in the ‘load group’.
The theoretical superiority of proportional ventilator modes lies
on the fact that, both in PAV+ and in NAVA, the ventilator assis-
tance instantaneously follows patient inspiratory demand. What
the patient wants, during an inspiratory effort, is flow and, hence,
volume insertion into its respiratory system. Flow and volume
are the combined result of ventilator’s output (airway pressure)
and patient’s inspiratory effort. As in these modes airway pres-
sure is proportional to inspiratory demand, VT should be strongly
related to Pdi and EAdi. We indeed found a strong positive lin-
ear correlation between PTPdi/b–VT and EAdi–VT during PAV+ and
NAVA respectively. However, we observed that, in many patients,
VT was less linearly related to PTPdi/b with NAVA than  with
the other two modes as a result of a weaker PTPdi/b– EAdi/b
linear correlation. Actually, in approximately half of the occa-
sions during NAVA, there  was no significant linear relationship
between PTPdi/b and EAdi/b. Although studies have shown a
tight relationship between the electrical and muscular activity of
the diaphragm (Beck et al., 1998, 2001), there are limited data
on how well NAVA-recorded EAdi correlates with Pdi. Recently,
strong linear correlation between NAVA-derived EAdi and respi-
ratory muscles pressure (Pmus) has been reported by Bellani and
coworkers (Bellani et al., 2013). Significant methodological dif-
ferences between the two studies might be responsible for this
discrepancy. Firstly, Bellani and coworkers did not directly mea-
sure Pdi, rather they estimated Pmus from the Pes and chest wall
elastance. Notwithstanding that this method of Pmus estimation
uses several assumptions (Georgopoulos et al., 1997), Bellani and
co-workers did not take into account chest wall resistance, which
in critically ill patients may vary considerably (Bellani et al., 2013).
Secondly, the investigators examined several assistance levels and
correlated the average Pmus of all respiratory cycles at each assis-
tance level with the average EAdi of all cycles at this level. This
may not adequately evaluate the ability of EAdi to track on a
88 E. Akoumianaki et al. / Respiratory Physiology & Neurobiology 203 (2014) 82–89
breath-by-breath basis patient effort variations. However, even
with this type of analysis Bland–Altman diagrams showed sig-
nificant difference between Pmus and EAdi derived indices of
inspiratory effort in several patients (Kondili et al., 2001). Finally,
differences in patient behavior may have significantly influenced
our results. It has been demonstrated that Pdi–EAdi linearity may
be disturbed by the appearance of intense diaphragmatic contrac-
tions and/or recruitment of accessory muscles (Fauroux et al., 2003;
Finucane and Singh, 2009). Both of these could have been apparent
and may have led to poor Pdi–EAdi correlation in some patients in
the current study. Further studies are needed to explore the factors
influencing this relationship.
Contrary to the theoretical prediction that VT and diaphrag-
matic pressure positive correlation would be stronger with PAV+
compared to PSV, no significant difference was found between the
two modes. It is known that several factors may disturb the rela-
tionship between ventilator output and diaphragmatic contraction
during PAV+ such as accuracy of lung mechanics estimation, use
of accessory muscles, non-linearity in the pressure–volume rela-
tionship and triggering function (Kacmarek, 2011; Kondili et al.,
2006). The latter may have played a significant role in the present
study, since approximately 9% of the transdiaphragmatic pressure
time product was lost for triggering. As previously demonstrated,
this means that not only the initial 9% but a considerably higher
percentage of inspiratory muscle pressure remained unsupported
(Akoumianaki et al., 2012; Georgopoulos et al., 2007). This may
have caused the relationship between effort and ventilator pres-
sure to lose its tightness, accounting, at least partly, for the fact
that in some patients during PAV+ the PTPdi/b–VT linear correlation
was weaker. Additionally, if, during PSV, inspiratory muscle con-
traction is sufficiently strong and pressure delivery is in time with
neural inspiration, VT becomes sensitive to patient effort (Younes,
1992). Conclusively, both the triggering delay during PAV+ and the
presence of strong inspiratory muscle contraction may have con-
tributed to the absence of difference between PAV+ and PSV with
regard to PTPdi/b–VT relationship.
Although in a recent bench study, Thille et al. (2009) have shown
that the triggering function during PSV did not differ between
PB840 and Servo-I, we have reasons to assume that the different
ventilators (PB840 and Servo-I) used in this study and not the mode
of support may account for the longer triggering delay associated
with PAV+. In a direct comparison of PSV and PAV+ using PB840
ventilator, we (Kondili et al., 2006) and others (Costa et al., 2011) did
not find any difference regarding the triggering function between
the two modes.
4.3. Limitations
Notably, since we studied awake patients, substantial breathing
variability made difficult the acquisition of Pdiswings that remained
within the targeted value over the long study period (>120 min).
Therefore, despite similar Pdiswings among the modes tested dur-
ing the assistance titration period, its value changed over time and
differences in Pdiswings between NAVA, PAV+ and PSV before chal-
lenge were observed in some patients. These differences, although
not statistically significant, could have masked changes in patient
effort after challenge. To minimize this limitation we expressed the
increase in effort as a percentage of its pre-challenge value.
In addition, the number of studied patients was small, in accor-
dance with most physiological studies requiring interventions such
as catheters insertion and respiratory effort loading in critically
ill intubated patients (Appendini et al., 1999; Costa et al., 2011;
Grasso et al., 2000; Kondili et al., 2006; Leung et al., 1997; Ranieri
et al., 1996, 1997). This, however, is certainly a limitation, because
the small number of studied patients may mask differences among
the modes tested. Furthermore, the population was heterogeneous.
Several advantages of NAVA might be more apparent in specific
patient groups, i.e. in patients with PEEPi and patients with major
asynchronies during PSV. Furthermore, despite the 20 min adap-
tation period at each session, wakefulness may have elicited acute
behavioral responses to respiratory load (Puddy and Younes, 1991).
Finally, study period was short and thus, few assumptions can be
made regarding the long-term effects of each mode.
5. Conclusions
In a small group of difficult to wean critically ill patients, propor-
tional modes of ventilation favored a more variable VT compared to
PSV, which on a long term basis might be clinically relevant. Impor-
tantly, in the face of changing respiratory system mechanics, PAV+
might be superior in unloading the patient’s work of breathing.
However, drawbacks in both NAVA and PAV+ function may limit
the effectiveness of these modes. NAVA does not adapt to respi-
ratory mechanic’s worsening in the absence of change in patient
effort. With PAV+ the poor triggering function which is further
worsened by any presence of PEEPi considerably affects the ability
of this mode to proportionally and effectively assist the inspira-
tory effort. Moreoever, in some patients, the EAdi is not linearly
correlated with Pdi during NAVA. Future studies in larger groups
of patients are required to determine the reasons for this and its
clinical significance as well as to confirm the results of the present
study. Meanwhile, in the daily practice and as a result of the afore-
mentioned drawbacks in proportional modes function, in critically
ill difficult to wean patients a carefully titrated pressure support
may exhibit comparable performance to these modes, provided
that patient’s respiratory system mechanics do not significantly
deteriorate.
Appendix A. Supplementary data
Supplementary data associated with this article can be found, in
the online version, at http://dx.doi.org/10.1016/j.resp.2014.08.012.
References
Akoumianaki, E., Kondili, E., Georgopoulos, D., 2012. Proportional assist ventilation.
Eur. Respir. Mon. 55, 97–115.
Alexopoulou, C., Kondili, E., Plataki, M., Georgopoulos, D., 2013. Patient–ventilator
synchrony and sleep quality with proportional assist and pressure support ven-
tilation. Intensive Care Med. 39, 1040–1047.
Appendini, L., Purro, A., Gudjonsdottir, M., Baderna, P., Patessio, A., Zanaboni, S., Don-
ner, C.F., Rossi, A., 1999. Physiologic response of ventilator-dependent patients
with chronic obstructive pulmonary disease to proportional assist ventilation
and continuous positive airway pressure. Am. J. Respir. Crit. Care Med. 159,
1510–1517.
Beck, J., Gottfried, S.B., Navalesi, P., Skrobik, Y., Contois, N., Rossini, M., Sinderby, C.,
2001. Electrical activity of the diaphragm during pressure support ventilation
in acute respiratory failure. Am. J. Respir. Crit. Care Med. 164, 419–424.
Beck, J., Sinderby, C., Lindstrom, L., Grassino, A., 1998. Effects of lung volume on
diaphragm EMG signal strength during voluntary contractions. J. Appl. Physiol.
(1985) 85, 1123–1134.
Bellani, G., Mauri, T., Coppadoro, A., Grasselli, G., Patroniti, N., Spadaro, S., Sala, V.,
Foti, G., Pesenti, A., 2013. Estimation of patient’s inspiratory effort from the
electrical activity of the diaphragm. Crit. Care Med. 41, 1483–1491.
Boles, J.M., Bion, J., Connors, A., Herridhe, M., Marsh, B., Melot, C., Pearl, R., Silverman,
H., Stanchina, M., Vieillard-Baron, A., Welte, T., 2007. Weaning from mechanical
ventilation. Eur. Respir. J. 29, 1033–1056.
Bosma, K., Ferreyra, G., Ambrogio, C., Pasero, D., Mirabella, L., Braghiroli, A., Appen-
dini, L., Mascia, L., Ranieri, V.M., 2007. Patient–ventilator interaction and sleep
in mechanically ventilated patients: pressure support versus proportional assist
ventilation. Crit. Care Med. 35, 1048–1054.
Coisel, Y., Chanques, G., Jung, B., Constantin, J.M., Capdevila, X., Matecki, S., Grasso, S.,
Jaber, S., 2010. Neurally adjusted ventilatory assist in critically ill postoperative
patients: a crossover randomized study. Anesthesiology 113, 925–935.
Colombo, D., Cammarota, G., Bergamaschi, V., De Lucia, M., Corte, F.D., Navalesi, P.,
2008. Physiologic response to varying levels of pressure support and neurally
adjusted ventilatory assist in patients with acute respiratory failure. Intensive
Care Med. 34, 2010–2018.
Costa, R., Spinazzola, G., Cipriani, F., Ferrone, G., Festa, O., Arcangeli, A., Antonelli,
M., Proietti, R., Conti, G., 2011. A physiologic comparison of proportional assist
 E. Akoumianaki et al. / Respiratory Physiology & Neurobiology 203 (2014) 82–89
ventilation with load-adjustable gain factors (PAV+) versus pressure support
ventilation (PSV). Intensive Care Med. 37, 1494–1500.
D‘Angelo, E., Robatto, F.M., Calderini, E., Tavola, M., Bono, D., Torri, G., Milic-Emili, J.,
1991. Pulmonary and chest wall mechanics in anesthetized paralyzed humans.
J. Appl. Physiol. (1985) 70, 2602–2610.
Delaere, S., Roeseler, J., D‘Hoore, W., Matte, P., Reynaert, M., Jolliet, P., Sottiaux,
T., Liistro, G., 2003. Respiratory muscle workload in intubated, spontaneously
breathing patients without COPD: pressure support vs proportional assist ven-
tilation. Intensive Care Med. 29, 949–954.
Fauroux, B., Hart, N., Luo, Y.M., MacNeill, S., Moxham, J., Lofaso, F., Polkey, M.I.,
2003. Measurement of diaphragm loading during pressure support ventilation.
Intensive Care Med. 29, 1960–1966.
Finucane, K.E., Singh, B., 2009. Human diaphragm efficiency estimated as power out-
put relative to activation increases with hypercapnic hyperpnea. J. Appl. Physiol.
(1985) 107, 1397–1405.
Georgopoulos, D., Mitrouska, I., Webster, K., Bshouty, Z., Younes, M., 1997. Effects
of inspiratory muscle unloading on the response of respiratory motor output to
CO2 . Am. J. Respir. Crit. Care Med. 155, 2000–2009.
Georgopoulos, D., Prinianakis, G., Kondili, E., 2006. Bedside waveforms interpreta-
tion as a tool to identify patient-ventilator asynchronies. Intensive Care Med.
32, 34–47.
Georgopoulos, D., Plataki, M., Prinianakis, G., Kondili, E., 2007. Current sta-
tus of proportional assist ventilation. Int. J. Intensive Care Med. 14,
74–80.
Grasso, S., Puntillo, F., Mascia, L., Ancona, G., Fiore, T., Bruno, F., Slutsky, A.S., Ranieri,
V.M., 2000. Compensation for increase in respiratory workload during mechan-
ical ventilation. Pressure-support versus proportional-assist ventilation. Am. J.
Respir. Crit. Care Med. 161, 819–826.
Kacmarek, R.M., 2011. Proportional assist ventilation and neurally adjusted ventila-
tory assist. Respir. Care 56, 140–148, discussion 149–152.
Kondili, E., Alexopoulou, C., Xirouchaki, N., Vaporidi, K., Georgopoulos, D., 2010. Esti-
mation of inspiratory muscle pressure in critically ill patients. Intensive Care
Med. 36, 648–655.
Kondili, E., Prinianakis, G., Alexopoulou, C., Vakouti, E., Klimathianaki, M.,
Georgopoulos, D., 2006. Respiratory load compensation during mechanical
ventilation–proportional assist ventilation with load-adjustable gain factors
versus pressure support. Intensive Care Med. 32, 692–699.
Kondili, E., Prinianakis, G., Anastasaki, M., Georgopoulos, D., 2001. Acute effects
of ventilator settings on respiratory motor output in patients with acute lung
injury. Intensive Care Med. 27, 1147–1157.
Kondili, E., Prinianakis, G., Georgopoulos, D., 2003. Patient–ventilator interaction.
Br. J. Anaesth. 91, 106–119.
Leung, P., Jubran, A., Tobin, M.J., 1997. Comparison of assisted ventilator modes
on triggering, patient effort, and dyspnea. Am. J. Respir. Crit. Care Med. 155,
1940–1948.
Mutch, W.A., Harms, S., Ruth Graham, M., Kowalski, S.E., Girling, L.G., Lefevre, G.R.,
2000. Biologically variable or naturally noisy mechanical ventilation recruits
atelectatic lung. Am. J. Respir. Crit. Care Med. 162, 319–323.
89
Piquilloud, L., Vignaux, L., Bialais, E., Roeseler, J., Sottiaux, T., Laterre, P.F., Jolliet,
P., Tassaux, D., 2011. Neurally adjusted ventilatory assist improves patient-
ventilator interaction. Intensive Care Med. 37, 263–271.
Puddy, A., Younes, M., 1991. Effect of slowly increasing elastic load on breathing in
conscious humans. J. Appl. Physiol. (1985) 70, 1277–1283.
Ranieri, V.M., Giuliani, R., Mascia, L., Grasso, S., Petruzzelli, V., Puntillo, N., Perchiazzi,
G., Fiore, T., Brienza, A., 1996. Patient–ventilator interaction during acute hyper-
capnia: pressure-support vs. proportional-assist ventilation. J. Appl. Physiol.
(1985) 81, 426–436.
Ranieri, V.M., Grasso, S., Mascia, L., Martino, S., Fiore, T., Brienza, A., Giuliani, R., 1997.
Effects of proportional assist ventilation on inspiratory muscle effort in patients
with chronic obstructive pulmonary disease and acute respiratory failure. Anes-
thesiology 86, 79–91.
Sinderby, C., Navalesi, P., Beck, J., Skrobik, Y., Comtois, N., Friberg, S., Gottfried, S.B.,
Lindstrom, L., 1999. Neural control of mechanical ventilation in respiratory fail-
ure. Nat. Med. 5, 1433–1436.
Spahija, J., de Marchie, M., Albert, M., Bellemare, P., Delisle, S., Beck, J., Sinderby,
C., 2010. Patient–ventilator interaction during pressure support ventilation and
neurally adjusted ventilatory assist. Crit. Care Med. 38, 518–526.
Thille, A.W., Lyazidi, A., Richard, J.C., Galia, F., Brochard, L., 2009. A bench study of
intensive-care-unit ventilators: new versus old and turbine-based versus com-
pressed gas-based ventilators. Intensive Care Med. 35, 1368–1376.
Tobin, M.J., 2003. Advances in mechanical ventilation. N. Engl. J. Med. 344,
1986–1996.
Varelmann, D., Wrigge, H., Zinserling, J., Muders, T., Hering, R., Putensen, C., 2005.
Proportional assist versus pressure support ventilation in patients with acute
respiratory failure: cardiorespiratory responses to artificially increased ventila-
tory demand. Crit. Care Med. 33, 1968–1975.
Wrigge, H., Golisch, W., Zinserling, J., Sydow, M., Almeling, G., Burchardi, H., 1999.
Proportional assist versus pressure support ventilation: effects on breathing
pattern and respiratory work of patients with chronic obstructive pulmonary
disease. Intensive Care Med. 25, 790–798.
Wysocki, M., Cracco, C., Teixeira, A., Mercat, A., Diehl, J.L., Lefort, Y., Derenne, J.P.,
Similowski, T., 2006. Reduced breathing variability as a predictor of unsuccessful
patient separation from mechanical ventilation. Crit. Care Med. 34, 2076–2083.
Xirouchaki, N., Kondili, E., Vaporidi, K., Xirouchakis, G., Klimathianaki, M., Gavri-
ilidis, G., Alexandropoulou, E., Plataki, M., Alexopoulou, C., Georgopoulos, D.,
2008. Proportional assist ventilation with load-adjustable gain factors in crit-
ically ill patients: comparison with pressure support. Intensive Care Med. 34,
2026–2034.
Younes, M., 1992. Proportional assist ventilation, a new approach to ventilatory
support. Theory. Am. Rev. Respir. Dis. 145, 114–120.
Younes, M., Kun, J., Masiowski, B., Webster, K., Roberts, D., 2001a. A method for
noninvasive determination of inspiratory resistance during proportional assist
ventilation. Am. J. Respir. Crit. Care Med. 163, 829–839.
Younes, M., Webster, K., Kun, J., Roberts, D., Masiowski, B., 2001b. A method for
measuring passive elastance during proportional assist ventilation. Am. J. Respir.
Crit. Care Med. 164, 50–60.