Eur J Anaesthesiol 2019; 36:135–143

ORIGINAL ARTICLE

Monitoring haemodynamic response to fluid-challenge in

ICU: comparison of pressure recording analytical method
and oesophageal Doppler
A prospective observational study
Romain Barthe  le
my, Arthur Neuschwander, Fatou Drame
, Maximilien Redoute
, David Ditchi,
Jules Stern, Alexandre Mebazaa, Romain Pirracchio and Benjamin G. Chousterman
Downloaded from http://journals.lww.com/ejanaesthesiology by BhDMf5ePHKbH4TTImqenVDiEotxSR1UH0aQm8ky4rg4fiVBT7tTnbCHWOZxMC3qQ9OJahBxy6uU= on 02/24/2019

BACKGROUND The ability of the pressure recording ana-
lytical method (PRAM) in tracking change in cardiac output
(DCO) after a fluid challenge in ICU needs to be evaluated
with the most contemporary comparison methods recom-
mended by experts.
OBJECTIVE Our objective was to report the trending ability
of PRAM in tracking DCO after a fluid challenge in ICU and to
compare this with oesophageal Doppler monitoring (ODM).
DESIGN Prospective, observational study.
SETTING Hôpital Lariboisière and Hôpital Europ
een George
Pompidou, Paris, France, from April 2016 to December 2017.
PATIENTS Critically ill patients admitted to ICU with
monitoring of CO monitored by ODM and invasive arterial
pressure.
INTERVENTION DCO after fluid challenge was simulta-
neously registered with ODM and PRAM connected to
the arterial line.
and clinical concordance evaluation (error grid and clinical
concordance rate). Predictors of bias were determined.
RESULTS Sixty-eight fluid challenge were administered in
49 patients. At the time of fluid challenge, almost all were
mechanically ventilated (99%), with 85% receiving norepi-
nephrine. Admission diagnosis was septic shock in 70% of
patients. Patients had a Sequential Organ Failure Assess-
ment score of 10 [7 to 12] and a median Simplified Acute
Physiology Score II of 61 [49 to 69]. Relative DCO bias was
7.88 (6.38) with radial limits of agreement of 41.78, polar
concordance rate 80% and clinical concordance rate 74%.
DCO bias was associated with baseline bias (P ¼ 0.007).
Baseline bias was associated with radial location of the
arterial line (P ¼ 0.03).
CONCLUSION When compared with ODM, PRAM has
insufficient performance to track DCO induced by fluid
challenge in ICU patients. Baseline bias is an independent
predictor of trending bias.

MAIN OUTCOME MEASURE Polar statistics (mean angular TRIAL REGISTRATION IRB 00010254-2016-033.
bias, radial limits of agreement and polar concordance rate) Published online 11 December 2018

Introduction
Fluid resuscitation of critically ill patients remains a
major concern for clinicians.1 Fluid challenge, to correct
hypotension or signs of hypoperfusion, is the most fre-
quent intervention in ICUs.2,3 Measurement of cardiac
output (CO) is recommended to evaluate the response to
fluid in patients with shock, and less invasive devices are
recommended if they have been validated in this con-
text.4 A correct evaluation of a new device needs static
comparison with a reference method, but also comparison
of trending ability.5

From the Department of Anaesthesia, Burn, and Critical Care, Saint-Louis Lariboisière University Hospital (RB, FD, MR, JS, AM, BGC), Department of Anaesthesia and
Critical Care, European Hospital Georges-Pompidou, APHP (AN, DD, RP), Paris Diderot University, Sorbonne Paris Cit e (AM, BGC), Inserm U942 (AM), Paris Descartes
University, Sorbonne Paris Cit
e (RP), Inserm U1153 (RP) and Inserm U1160, Paris, France (BGC)
Correspondence to Romain Barth el
emy, MD, Department of Anaesthesia, Burn, and Critical Care, Saint-Louis Lariboisière University Hospital, AP-HP, 2 Rue Ambroise
Par
e, 75475 Paris Cedex 10, France
E-mail: romain.barthelemy@aphp.fr

0265-0215 Copyright ß 2018 European Society of Anaesthesiology. All rights reserved. DOI:10.1097/EJA.0000000000000924

Copyright © European Society of Anaesthesiology. Unauthorized reproduction of this article is prohibited.

 136 Barth
el
emy et al.
The pressure recording analytical method (PRAM) is a
pulse-contour-based CO monitor that was first described
in 2002.6 This device allows continuous and beat-to-beat
CO measurement and is minimally invasive as it can be
implemented on any arterial catheter by connecting it to
the transducer. Few studies have addressed the ability of
PRAM to track changes in CO (DCO) related to fluid
challenge in ICU and their results are conflicting.7,8
Therefore, there is still an unmet need for the evaluation
of COPRAM accuracy in tracking DCO related to fluid
challenge in the most severely ill ICU patients using the
most contemporary comparison methods recommended
by experts.5,9
The aim of this study was to investigate the trending
ability of PRAM to track DCO after fluid challenge in a
mixed cohort of ICU patients in comparison with oeso-
phageal Doppler monitoring (ODM).
Methods
The current study is a prospective, two-centre, observa-
tional study aiming to evaluate the trending ability of
PRAM in tracking DCO after fluid challenge in ICU
patients compared with ODM. Ethical approval (IRB
00010254-2016-033) was provided by an institutional
ethics committee (Comite d’ethique de la Societe Française
d’Anesthesie-Reanimation, Paris, France, chairman Prof J.-
E. Bazin) on 14 April 2016. Requirement of a written
informed consent was in accordance with French law.
Patients
The study was conducted in two French surgical ICUs
between April 2016 and December 2017. Site A (surgical
ICU, Hôpital Lariboisière, Paris, France) had a 5 year
experience of PRAM monitor use and participated in a
multicentre validation study.10 Site B (surgical ICU,
Hôpital Europ
een George Pompidou, Paris) was new
to the technique. Both sites have used ODM as their
primary CO monitor for more than 10 years.
Inclusion criteria were: presence of an intra-arterial cath-
eter and an ODM, a decision by the physician in charge to
perform a fluid challenge for haemodynamic instability,
availability of the PRAM device, availability of an inves-
tigator trained in the use of both devices. Exclusion
criteria were age less than 18 years old, pregnancy, cardiac
dysrhythmia11–13 and poor signal quality of either of the
methods (see above).
Measurements and data collection
Patient characteristics, diagnosis at admission, haemody-
namic status, current vasopressor treatment, mechanical
ventilation, localisation of the arterial line, Simplified
Acute Physiology Score (SAPS II), Sequential Organ
Failure Assessment (SOFA), Charlson comorbidity
index, dead or alive status at day 28 were retrieved.
Eur J Anaesthesiol 2019; 36:135–143
Copyright © European Society of Anaesthesiology. Unauthorized reproduction of this article is prohibited.
PRAM (MostCareUp; Vytech, Padova, Italy) is a beat-to-
beat CO monitor connected to an arterial line and does
not require any specific catheter. Stroke volume (SV) is
computed from the area under the systolic portion of the
arterial pressure curve and the dynamic impedance of the
cardiovascular system. Impedance is derived from an
analysis of the arterial waveform signal sampled at
1000 Hz and requires neither internal nor external cali-
bration.6,13,14
ODM (CardioQ; Deltex, Chichester, Sussex, UK) is a
beat-to-beat CO monitor measuring descending aortic
blood flow velocity with a 4 MHz continuous ultrasound
probe assuming a fixed angle of 458 with aorta. The
relationship between left ventricular SV and descending
aortic blood flow velocity is internally computed using
patient’s age, height and weight.15
Before inclusion, the quality of the ODM signal was
assessed by a trained investigator: the probe was posi-
tioned to capture an optimal wave pattern with correct
delineation of the signal and absence of diastolic flow.
Patient characteristics entered in the device were
checked by the physician and the cycle averaging for
CO measurement was set to 15 cycles. The quality of the
arterial pressure signal was assessed with a fast-flush test
ensuring no abnormal signal damping. Once PRAM is
connected with a Y-connector to the arterial line, zeroing
at the phlebostatic level was performed and accuracy of
algorithm detection of the dicrotic notch was checked
and corrected if necessary. The internal filter was left at
the default setting and the averaging for CO measure-
ment was set to 15 cycles.
All fluid challenges were performed with crystalloid
solutions, infused in less than 10 min (free-flow adminis-
tration) as widely performed in ICU.3 The volume (either
250 or 500 ml) of fluid challenge was at the discretion of
the attending physician. To ensure consistency in timing
a unique simultaneous photograph containing both moni-
tors was taken immediately after the 15 averaging cycles
of the ODM, at baseline and 1 min after fluid challenge.
ODM probe positioning and PRAM dicrotic notch accu-
racy were checked again after fluid challenge and
adjusted if necessary. Doses of medications and mechan-
ical ventilation settings were not modified between the
two time points.
Statistical analysis
The results are expressed as mean (SD) or as median
[interquartile range] depending on the normal distribu-
tion of the data.
Agreement between cardiac output measured with
oesophageal Doppler monitoring and cardiac output
measured with pressure recording analytical method
Bland and Altman analysis for repeated measurements
was used to assess agreement between pooled data of

COODM and COPRAM, with calculation of bias and its 95%
confidence interval (CI95), upper and lower limits of
agreement (LOA) and their CI95,16,17 and percentage
error (PE).18 The updated method for repeated measure-
ments according to Olofsen et al. was also performed.19
To take into account the fact that one patient may receive
more than one fluid challenge, the effect of fluid chal-
lenge on bias was tested in linear mixed models with bias
as a dependent variable, time of CO measurement
(before or after fluid challenge) as a fixed effect and
patient, nested by centre, as a random effect.
Directional concordance of change in cardiac output
Directional concordance was graphically evaluated using
the standard 4-quadrant plot initially used by Perrino et al.
with a central zone of 10% for relative DCO and
0.5 l min1 for absolute DCO.20 Global clinical concor-
dance was defined as the proportion of paired measures in
the same quadrant or in the central zone over all
paired measures.
Clinical concordance of change in cardiac output
Clinical concordance was determined using the 4-quad-
rant plot with an error grid delimiting 4 zones as recently
proposed by Montenij et al.5 The error grid was plotted
with cut-off values of 10% delimiting no significant
DCO as it is the lower reported definition for fluid
challenge responsiveness,21 and 30% delimiting large
DCO. For absolute DCO, cut-off values of 0.5 l min1
were chosen for no significant change and 1.5 l min1
for large DCO. Zone 1 corresponds to the clinical concor-
dance categories in which COPRAM and COODM change
in the same direction and to the same extent, zone 2 in
which COPRAM and COODM change in the same direction
but not to the same extent, zone 3 in which COPRAM
changes while COODM is constant or vice versa, and zone
4 represents opposite changes in COPRAM and COODM.
Montenij et al. did not propose any cut-off value to
validate a device as ‘clinically acceptable’. We propose
a concordance rate (zone 1 þ zone 2) above 90% as a
clinically acceptable trending ability.
Polar plot analysis of change in cardiac output
Polar plot analysis was performed using the method
described by Critchley et al. Results are expressed as
mean angular bias, radial LOA (rLOA) and polar concor-
dance rate (PCR, percentage of paired DCO that lies
between the 308 limits).22,23 Polar statistics were com-
puted for repeated measures, after exclusion of a central
zone of 10% for relative DCO and 0.5 l min1 for
absolute DCO. Criteria used to define acceptable polar
concordance: were: radial bias less than 58, rLOA less
than 308 and PCR more than 90%.
Subgroup analysis
Similar analyses were performed in subgroups of patients
according to SAPS II, Charlson comorbidity index, age,
Eur J Anaesthesiol 2019; 36:135–143
Copyright © European Society of Anaesthesiology. Unauthorized reproduction of this article is prohibited.
Monitoring haemodynamic response to fluid-challenge in ICU 137
history of hypertension, BMI, location of arterial line,
septic shock, dose of norepinephrine, mean arterial pres-
sure, heart rate, investigation site, baseline bias and
volume of fluid challenge. To take into account the fact
that a patient may receive more than one fluid challenge,
the effect of the subgroup variables on CO bias and DCO
bias were tested in linear mixed models with bias as
dependant variables, subgroup variables as fixed effects
and patient nested by centre as random effects. Variables
with a P value less than 0.20 in univariate analysis were
tested in multivariate analysis. Classification according to
PCR and zones of error grid were tested in generalised
linear mixed models with patients nested by centre as
random effects.
Sensitivity analysis
A sensitivity analysis based on the first fluid challenge
was performed for directional concordance, clinical con-
cordance and polar plot analyses.
Statistical analyses were performed using R statistical soft-
ware (The ‘R’ Foundation for Statistical Computing,
Vienna, Austria). Updated Bland–Altman analysis was per-
formed with the web-based program available at https://
sec.lumc.nl/method_agreement_analysis/. A P value less
than 0.05 was considered statistically significant.
Sample size
We assumed that at least 30 significant DCO after fluid
challenge were needed to expect a normal distribution of
DCO differences to compute parametric statistics for
repeated measures. Half of patients usually improve their
CO after fluid challenge21 leading to 60 fluid challenge
needed. We aimed to include 66 fluid challenge, allowing
a safety a margin of 10%.
Results
A total of 68 fluid challenge administered to 49 subjects
was included in the study. A diagram depicting the flow of
patients is provided as supplementary data (see Figure,
Supplemental Digital Content 1, http://links.lww.com/
EJA/A182). Patients’ characteristics are summarised in
Table 1 and haemodynamic variables are presented in
Table 2. The main admission diagnosis was septic shock
(70% of patients). At the time of fluid challenge, almost
all patients (99%) were receiving mechanical ventilation
and the majority (85%) were treated with norepineph-
rine: median dose of 0.39 mg kg1 min1 [0.23 to 0.98].
The severity of illness was high with SAPS II of 61 [49 to
69], SOFA of 10 [7 to 12] and a mortality rate at day 28 of
41%.
Agreement between cardiac output measured with
oesophageal Doppler monitoring and cardiac output
measured with pressure recording analytical method
Bland and Altman analyses of static paired CO compari-
son are provided as supplementary data (see Text File,
 138 Barth
el
emy et al.

Table 1 Main characteristics of patients Supplemental Digital Content 2, http://links.lww.com/

n¼49 EJA/A182).
Age (years) 65 [55 to 77]
Female 18 (36%)
Directional concordance of change in cardiac output
BMI (kg m2) 25 [22 to 30]
Admission category Relative change in cardiac output
Surgical 39 (80%) The 4-quadrant plot with central exclusion zone of 10%
Medical 10 (20%) is presented in supplementary data (see Figure, Supple-
Admission diagnosis
Septic shock 34 (70%)
mental Digital Content 3, http://links.lww.com/EJA/
Pneumonia 21 A182). Both methods agreed in identifying DCO less
Peritonitis 9 than 10% in 16 paired measurements. Concordant direc-
Soft-tissues infection 3
tional changes (i.e. paired CO outside the central zone are
Arthritis 1
Brain injury 6 (12%) in upper-right or lower left quadrant) were found in 48 of
Haemorrhagic shock 3 (6%) the remaining 52 paired measurements (92%). The global
Cardiac arrest 3 (6%) concordance rate was then 94%.
Polytraumatism 2 (4%)
Major abdominal surgery 1 (2%)
Charlson comorbidity index 4 [2 to 6] Absolute change in cardiac output
Comorbidities The 4-quadrant plot with central exclusion zone of
History of hypertension 27 (55%) 0.5 l min1 is presented in supplementary data (see
Diabetes mellitus 11 (22%)
Coronary artery disease 5 (10%) Figure, Supplemental Digital Content 3, http://
Chronic heart failure 4 (8%) links.lww.com/EJA/A182). Both methods agreed in iden-
SAPS II 61 [49 to 69] tifying DCO less than 0.5 l min1 in 17 paired measure-
SOFA score 10 [7 to 12]
Mortality at day 28 20 (41%)
ments. Concordant directional change was found in 47 of
Number of fluid challenge by patient 1 [1 to 2] the remaining 51 paired measurements (92%). The global
Number of patients receiving 1/2/3/4 fluid challenge 37/6/5/1 concordance rate was then 94%.
Values are median [interquartile range] and number (percentage). SAPS, simpli-
fied acute physiology score; SOFA, sequential organ failure assessment. Clinical concordance of change in cardiac output
Relative change in cardiac output
The 4-quadrant plot with four zones error grid is pre-
sented in Fig. 1a. Good clinical concordance (zone 1) was
found in 53% of fluid challenges. Acceptable clinical
Table 2 Characteristics of fluid challenges concordance (both methods showed significant DCO
n ¼ 68 but magnitude differed, zone 2) was found in 21% of
Time from ICU admission (days) 1 [1 to 3] fluid challenges. No clinical concordance (one method
Mechanical ventilation 67 (99%) showed a significant DCO and the other did not, zone 3)
Tidal volume (ml kg1) 6.4 [6.1 to 6.7]
End expiratory pressure (cmH2O) 5 [5 to 6]
represented 26% of fluid challenges. No fluid challenges
Mode of ventilation showed opposite directional change (zone 4). The total of
Controlled 64 (96%) good and acceptable clinical concordance (zone 1 þ 2)
Partially assisted 3 (4%) was 74%.
Norepinephrine 58 (85%)
Dose (mg kg1 min1) 0.39 [0.23 to 0.98]
Arterial line location Absolute change in cardiac output
Femoral 33 (49%) The 4-quadrant plot with four zones error grid is pre-
Radial 35 (51%)
sented in Fig. 1b. Zone 1, zone 2 and zone 3, respectively,
Volume of FC 250 [250 to 500]
250 ml 50 (74%) represented 56, 16 and 28% of fluid challenges. No
500 ml 18 (26%) opposite directional change was found (zone 4). The
ml kg1 3.9 [3.0 to 5.2] total of zones 1 þ 2 was 72%.
Baseline haemodynamic
COODM (l min1) 4.0 [3.1 to 5.0]
COPRAM (l min1) 3.7 [3.3 to 4.1] Polar plot analysis of change in cardiac output
SVODM (ml) 42 [35 to 51] Relative change in cardiac output
SVPRAM (ml) 40 [33 to 56]
A total of 44 positive fluid challenge administered to 33
Heart rate (bpm) 88 [79 to 106]
MAP (mmHg) 74 [68 to 83] patients was included in the analysis. The polar plot is
SAP (mmHg) 110 [95 to 128] shown in Fig. 2a. The central zone (mean DCO < 10%)
DAP (mmHg) 58 [51 to 63] contained 24 paired DCO. Mean bias was 7.88 (6.38) and
CVP (mmHg) 6 [4 to 9]
rLOA was 41.78 with rLOACI of 10.98. PCR was 80%.
Values are median [interquartile range] and number (percentage). CO, cardiac
output; CVP, central venous pressure; DAP, diastolic arterial pressure; FC, fluid Absolute change in cardiac output
challenge; MAP, mean arterial pressure; ODM, oesophageal Doppler monitoring;
PRAM, pressure recording analytical method; SAP, systolic arterial pressure; SV, A total of 40 positive fluid challenges administered to 31
stroke volume. patients was included in the analysis. The polar plot is

Eur J Anaesthesiol 2019; 36:135–143

Copyright © European Society of Anaesthesiology. Unauthorized reproduction of this article is prohibited.
 Monitoring haemodynamic response to fluid-challenge in ICU 139

Fig. 1

(a) (b)
Zone 1: 36 (53%) Zone 1: 38 (56%)
50 (74%) 49 (72%)
Zone 2: 14 (21%) Zone 2: 11 (16%)
Zone 3: 18 (26%) Zone 3: 19 (28%)
Zone 4: 0 Zone 4: 0

3
100
2

COODM (l min 1 )
50
1
COODM (%)

100 50 50 100 3 2 1 1 2 3
1
50
2
100
3

COPRAM (%) COPRAM (l min 1 )

Error grid plots of change in cardiac output after fluid challenge measured with oesophageal Doppler monitoring and pressure recording analytical
method. Subfigure (a) shows relative change in cardiac output expressed as percentage of baseline value and subfigure (b) shows absolute change
in cardiac output in l min1. Zone 1 corresponds to the clinical concordance categories in which cardiac output after fluid challenge measured with
pressure recording analytical method and cardiac output after fluid challenge measured with oesophageal Doppler monitoring change in the same
direction and to the same extent, zone 2 in which cardiac output after fluid challenge measured with pressure recording analytical method and
cardiac output after fluid challenge measured with oesophageal Doppler monitoring change in the same direction but not to the same extent, zone 3
in which cardiac output after fluid challenge measured with pressure recording analytical method changes while cardiac output after fluid challenge
measured with oesophageal Doppler monitoring is constant or vice versa, and zone 4 represents opposite changes in cardiac output after fluid
challenge measured with pressure recording analytical method and cardiac output after fluid challenge measured with oesophageal Doppler
monitoring. The number of data pairs falling within each zone is depicted above the figures.

shown in Fig. 2b. The central zone (mean
DCO < 0.5 l min1) contained 28 paired DCO. Mean bias
was 9.18 (7.58) and rLOA was 47.18 with rLOACI of
12.98. PCR was 78%.
Subgroup analysis
Subgroup analysis of baseline CO measurement are
shown in supplementary data (see Table, Supplemental
Digital Content 4, http://links.lww.com/EJA/A182). In
multivariable analysis, only radial location of the arterial
line was independently associated with CO bias
(P ¼ 0.03).
Subgroup analysis of DCO concordance is shown in Table
3. In multivariable analysis, only baseline bias was inde-
pendently associated with DCO bias (P ¼ 0.007). A his-
tory of hypertension is associated with lower PCR
(P ¼ 0.03) and lower clinical concordance rate (P ¼ 0.02).
Sensitivity analysis based on first fluid challenge
Relative change in cardiac output
Both methods agreed in identifying DCO less than 10%
in 12 paired measurements. Concordant directional
changes (i.e. paired CO outside the central zone are in
upper-right or lower left quadrant) were found in 34 of
the remaining 37 paired measurements (92%). The global
concordance rate was then 94%.
Good clinical concordance (zone 1) was found in 57% of
fluid challenges. Acceptable clinical concordance was
found in 20% of fluid challenges. No clinical concordance
represented 22% of fluid challenges. No fluid challenges
showed opposite directional change (zone 4). The total of
good and acceptable clinical concordance (zone 1 þ 2)
was 78%.
A total of 32 positive fluid challenges were included in the
polar plot analysis. The central zone (mean DCO < 10%)
contained 17 paired DCO. Mean bias was 4.18 (7.88) and
rLOA was 44.28 with rLOACI of 13.58. PCR was 81%.
Absolute change in cardiac output
Both methods agreed in identifying DCO less than
0.5 l min1 in 12 paired measurements. Concordant direc-
tional change was found in 34 of the remaining 37 paired
measurements (92%). The global concordance rate was
then 94%.

Eur J Anaesthesiol 2019; 36:135–143

Copyright © European Society of Anaesthesiology. Unauthorized reproduction of this article is prohibited.
 140 Barth
el
emy et al.

Fig. 2

(a) 90 (b) 90
120 60 120 60

CO (l min 1 )
CO (%)
150 30 150 30

180 0 180 0
20 0.5
1
40
150 30 150 1.5 30
80
2
10 0 2.5
120 60 120 60
90 90

Polar plots of change in cardiac output after fluid challenge measured with oesophageal Doppler monitoring and pressure recording analytical
method. Subfigure (a) shows relative change in cardiac output expressed as percentage of baseline value after exclusion of paired measurements of
mean change in cardiac output less than 10% and subfigure (b) shows absolute change in cardiac output in l min1 after exclusion of paired
measurement of mean change in cardiac output less than 0.5 l min1. Solid lines delimit the 308 zone of polar concordance rate, dashed lines are
for mean radial bias and dotted lines for radial limits of agreement.

Zone 1, zone 2 and zone 3, respectively, represented 57, 18
and 24% of fluid challenges. No opposite directional change
was found (zone 4). The total of zones 1 þ 2 was 76%.
A total of 29 positive fluid challenge was included in
the polar plot analysis. The central zone (mean
DCO < 0.5 l min1) contained 20 paired DCO. Mean bias
was 5.18 (8.08) and rLOA was 43.18 with rLOACI of
13.98. PCR was 86%.
Figures related to sensitivity analysis can be found in
Supplementary Digital Content 5, http://links.lww.com/
EJA/A182.
Discussion
In this study, we report poor performance of PRAM
compared with ODM when tracking DCO in response
to a fluid challenge in ICU patients. Indeed, PRAM
measures had higher bias, larger rLOA and lower PCR
than usually considered to be acceptable, and more than
25% clinical discordance compared with ODM. Perfor-
mances of PRAM seem to be only dependent of the
baseline bias with no other identified risk factors.
We chose ODM in our study as the reference method. It
is not considered the traditional ‘gold standard’ for CO
measurement but has practically no bias and with a
limited clinical agreement when compared with pulmo-
nary thermodilution.15 Variations of aortic diameter may
lead to an underestimation of DCO24 but it is well
demonstrated that ODM has a good trending ability
when measuring relative change in CO.15 Moreover,
ODM has several advantages regarding the design of this
study. It is a beat-to-beat continuous CO monitor with
immediate response to haemodynamic changes. This
allowed us to test the ability of PRAM to detect DCO
in the same timeframe. This is particularly important as
the haemodynamic effect of fluid challenge is not stable
over time.25,26 We set up both devices to average the CO
measurement over 15 beats and ensured a perfect syn-
chronisation. ODM also has the advantage of good
repeatability with the coefficient of variation reported
between 3.827 and 8%.28
In our study, the main objective was to evaluate the
trending ability of PRAM when tracking DCO compared
with ODM. Only three previous studies evaluated the
ability of PRAM to detect DCO in ICU patients, with
conflicting results. Two of these studies reported good
trending ability after interventions in ICU patients such
as fluid challenge or variation of vasopressor doses,8,29 and
one found poor trending ability when performing a fluid
challenge.7 In our study, we also report a poor agreement
between static CO values with the Bland–Altman meth-
odology because of large LOA despite a low mean bias.
Our results are however in conflict with a recent multi-
centre study conducted by Scolletta et al.10 including 400
ICU patients with 400 paired measurement of CO com-
paring PRAM and Doppler echocardiography. The latter

Eur J Anaesthesiol 2019; 36:135–143

Copyright © European Society of Anaesthesiology. Unauthorized reproduction of this article is prohibited.
 Monitoring haemodynamic response to fluid-challenge in ICU 141

Table 3 Subgroup analysis of DCO concordance in fluid challenge responders

n Bias (95% CI) rLOA (95% CI) Pa Pb PCR Pc Z1R2 Pd

1
Baseline bias (L min )
1 23 2.0 (5.5 to 9.5) 36.1 (23 to 49.2) 0.005 0.01 87 0.07 83 0.83
>1 21 14.1 (3.7 to 24.5) 47.9 (29.8 to 66) 71 64
Arterial line location
Radial 21 13.5 (5.4 to 21.6) 38.7 (24.7 to 52.7) 0.06 0.06 74 0.34 69 0.51
Femoral 23 1.5 (8.5 to 11.5) 45.7 (28.4 to 63) 86 79
Sepsis
No 15 5.1 (7.3 to 17.5) 47.9 (26.5 to 69.3) 0.57 67 0.24 70 0.74
Yes 29 9.2 (1.5 to 16.9) 41.4 (28.1 to 54.7) 86 76
Norepinephrine (mg kg1 min1)
0.33 19 7.4 (4.3 to 19.1) 50.8 (30.6 to 71) 0.52 68 0.21 62 0.15
>0.33 25 8 (0.4 to 16.6) 37.8 (24.7 to 50.9) 88 85
MAP (mmHg)
74 23 7.8 (1.7 to 17.3) 45.4 (29 to 61.8) 0.64 74 0.63 74 0.85
>74 21 7.8 (1.4 to 17) 42 (26 to 58) 86 73
Heart rate (bpm)
88 22 8.3 (1.2 to 17.8) 44.7 (28.2 to 72.9) 0.76 77 0.93 77 0.42
>88 22 7.2 (2 to 16.4) 42.9 (27 to 58.8) 82 70
BMI (kg m2)
<25 20 7.4 (2.6 to 17.4) 44.7 (27.4 to 62) 0.74 80 0.54 71 0.29
25 24 8.1 (0.7 to 16.9) 43.1 (27.9 to 58.3) 79 76
Age (years)
65 20 9.5 (0.7 to 19.7) 45.6 (28 to 63.2) 0.29 70 0.11 65 0.08
>65 24 6.3 (2.3 to 14.9) 42.2 (27.3 to 57.1) 88 82
SAPS II
61 22 8.7 (0.7 to 18.1) 44.2 (27.9 to 60.5) 0.60 82 0.57 74 0.79
>61 22 6.8 (2.5 to 16.1) 43.4 (27.4 to 59.4) 77 74
Charlson comorbidity index
4 23 10.4 (1.4 to 19.4) 43.3 (27.7 to 71) 0.94 78 0.55 70 0.51
>4 21 4.9 (4.6 to 14.4) 43.7 (27.2 to 60.2) 81 79
History of hypertension
No 25 3 (4.1 to 10.1) 35.3 (23.2 to 47.3) 0.07 0.10 92 0.03 84 0.02
Yes 19 14 (2.4 to 25.6) 50.4 (30.4 to 70.4) 63 61
Site
A 36 9.8 (2.9 to 16.7) 41.2 (29.3 to 53.1) 0.21 83 0.20 76 0.17
B 8 1.3 (19.4 to 16.8) 51.1 (19.8 to 82.4) 63 64
Volume
250 mL 33 11.1 (4 to 18.2) 40.8 (28.5 to 53.1) 0.08 0.06 82 0.52 76 0.23
500 ml 11 2.3 (16.3 to 11.7) 46.6 (22.3 to 68.9) 73 67

Data are Continuous variables are dichotomised according to median values and polar plot statistics are performed on each subgroup after exclusion of the central zone of
mean DCO<10%. rLOA, radial limits of agreement; PCR, polar concordance rate. PCR is the percentage of paired DCO within the 308 radial range. Z1þ2 is the
percentage of paired DCO within zone 1 and zone 2 of error grid. pa is for univariate analysis: bias is tested for each variable in a linear mixed model with patient as random
effect. pb value is for multivariable analysis: any variable with p < 0.20 in univariate analysis was included in the linear mixed model. pc is for PCR between group and pd is for
Z1þ2 between group tested in a generalised linear mixed model. CI, Confidence Interval.

study showed a quasi-null bias with narrow LOA
(1.5 l min1), but DCO was not investigated.
The observed discrepancies with previous studies could
be related to different causes. First, we used ODM as the
reference method while other studies used pulmonary
thermodilution or Doppler echocardiography. Although
ODM is not the gold standard for CO measurement, this
limitation applies only to ‘‘absolute’’ values, for trending
values ODM is known to be robust. Second, our patients
were more severely ill than those previously studied. The
mean SAPS II score in our study was 61 compared with 35
in Scolletta et al. study. In our cohort, almost all patients
were mechanically ventilated, the majority had septic
shock and required high doses of norepinephrine for
haemodynamic instability compared with 24.5% in Scol-
letta et al. study. Inclusion criteria specified that patient
had to be already monitored with ODM and invasive
arterial pressure possibly selecting patient with the most
complex haemodynamic states.
We identified initial bias as an independent predictor of
trending bias during fluid challenge. Therefore, a high
baseline bias is a possible cause of poor agreement
between PRAM and ODM. We investigated risk factors
of baseline bias and trending bias that could explain the
observed discordances. We identified radial location of
arterial line as an independent risk factor of baseline bias.
This can be explained by the femoral location being
closer to a ‘central’ measurement of arterial pressure
and less influenced by peripheral reflection waves.30,31
This could impair the reliability of arterial signals and
impact on PRAM performance. However, this explana-
tion is discordant with the findings in the study by

Eur J Anaesthesiol 2019; 36:135–143

Copyright © European Society of Anaesthesiology. Unauthorized reproduction of this article is prohibited.
 142 Barth
el
emy et al.
Scolletta et al.10 where a higher bias was observed when
the arterial catheter was in the femoral artery. We are also
unable to exclude possible biases related to patient
characteristics possibly influencing both the choice of
the site of arterial line insertion and the accuracy of
the device. Larger LOA are observed with higher comor-
bidities, higher disease severity, higher norepinephrine
doses and in older patients. All of these conditions might
also impair reliability of the arterial signal and could
impact on PRAM performance. The arterial site was
not an independent predictor of trending bias.
In our study, we used up to date analysis methods for
trending ability: the quantitative analysis with polar plot
and polar statistics as proposed by Critchley et al.23 and
the recent semi-quantitative evolution of the traditional
4-quadrant analysis as proposed by Montenij et al.5
termed the ‘error grid’. The error grid is a step forward
in the analysis of trending ability because it focuses on
the clinical relevance of the results. Here, our results
reveal a clinical discordance in more than 25% of
fluid challenge.
Our study has some limitations. It is a two-centre study
with a limited number of patients. Inclusion rate was low.
This could have induced a selection bias. We excluded
patients with cardiac dysrhythmia because it has been
suggested to be a source of error when using PRAM.11,12
Therefore, our results cannot be extrapolated to all ICU
patients. Fluid challenges were repeated in some
patients. We therefore performed an additional sensitiv-
ity analysis based on the first fluid challenge and found
similar results. Another point is that the volume of fluid
challenge is limited to 250 ml in a majority of patients in
our study. However, we did not find any interaction
between PRAM performance and the volume of fluid
challenge. Most of the fluid challenge were made in
patients receiving controlled mechanical ventilation, this
prevent generalisation to patients that are not under
mechanical ventilation or have spontaneous breaths.
The clinical implication of our study is that the PRAM
cannot be used in critically ill patients to track DCO after
a fluid challenge except if high bias is excluded by
comparing CO obtained with PRAM with another non-
invasive device such as Doppler echocardiography. Fur-
ther studies assessing PRAM performances when a low
initial bias has been confirmed are awaited.
In conclusion, when compared with ODM, PRAM has
insufficient performance to track DCO induced by fluid
challenge in ICU patients. Baseline bias is an indepen-
dent predictor of trending bias. Baseline bias is higher
when the arterial line is in the radial artery.
Acknowledgements relating to this article
Assistance with the study: none.
Financial support and sponsorship: none.
Eur J Anaesthesiol 2019; 36:135–143
Copyright © European Society of Anaesthesiology. Unauthorized reproduction of this article is prohibited.
Conflicts of interest: RB received travel expenses and unrestricted
grants for meeting and congresses from Vygon France.
Presentation: preliminary data of this study were presented as a
poster presentation at the annual scientific meeting of the French
Society of Anaesthesiology and Critical Care Medicine (SFAR Le
Congrès), 21 to 23 September 2017, Paris.
References
1 Perner A, Hjortrup PB, Pettilä V. Focus on fluid therapy. Intensive Care Med
2017; 43:1907–1909.
2 Cecconi M, Hofer C, Teboul J-L, et al. Fluid challenges in intensive care: the
FENICE study: a global inception cohort study. Intensive Care Med 2015;
41:1529–1537.
3 Boulain T, Boisrame-Helms J, Ehrmann S, et al. Volume expansion in the
first 4 days of shock: a prospective multicentre study in 19 French intensive
care units. Intensive Care Med 2015; 41:248–256.
4 Cecconi M, De Backer D, Antonelli M, et al. Consensus on circulatory
shock and hemodynamic monitoring. Task force of the European Society of
Intensive Care Medicine. Intensive Care Med 2014; 40:1795–1815.
5 Montenij LJ, Buhre WF, Jansen JR, et al. Methodology of method
comparison studies evaluating the validity of cardiac output monitors: a
stepwise approach and checklist. Br J Anaesth 2016; 116:750–758.
6 Romano SM, Pistolesi M. Assessment of cardiac output from systemic
arterial pressure in humans. Crit Care Med 2002; 30:1834–1841.
7 Biais M, Cottenceau V, Stecken L, et al. Evaluation of stroke volume
variations obtained with the pressure recording analytic method. Crit Care
Med 2012; 40:1186–1191.
8 Franchi F, Falciani E, Donadello K, et al. Echocardiography and pulse
contour analysis to assess cardiac output in trauma patients. Minerva
Anestesiol 2013; 79:137–146.
9 Saugel B, Grothe O, Wagner JY. Tracking changes in cardiac output:
statistical considerations on the 4-quadrant plot and the polar plot
methodology. Anesth Analg 2015; 121:514–524.
10 Scolletta S, Franchi F, Romagnoli S, et al. Comparison between Doppler-
echocardiography and uncalibrated pulse contour method for cardiac
output measurement: a multicenter observational study. Crit Care Med
2016; 44:1370–1379.
11 Maj G, Monaco F, Landoni G, et al. Cardiac index assessment by the
pressure recording analytic method in unstable patients with atrial
fibrillation. J Cardiothorac Vasc Anesth 2011; 25:476–480.
12 Barile L, Landoni G, Pieri M, et al. Cardiac index assessment by the
pressure recording analytic method in critically ill unstable patients after
cardiac surgery. J Cardiothorac Vasc Anesth 2013; 27:1108–1113.
13 Romagnoli S, Franchi F, Ricci Z, et al. The pressure recording analytical
method (PRAM): technical concepts and literature review. J Cardiothorac
Vasc Anesth 2017; 31:1460–1470.
14 Romano S, Conti A, Giglioli C, et al. Blood flow assessment by arterial
pressure wave without external calibration. Comput Cardiol 2006;
33:293–296.
15 Dark PM, Singer M. The validity of trans-esophageal Doppler
ultrasonography as a measure of cardiac output in critically ill adults.
Intensive Care Med 2004; 30:2060–2066.
16 Bland JM, Altman DG. Measuring agreement in method comparison
studies. Stat Methods Med Res 1999; 8:135–160.
17 Bland JM, Altman DG. Agreement between methods of measurement
with multiple observations per individual. J Biopharm Stat 2007; 17:
571–582.
18 Critchley LA, Critchley JA. A meta-analysis of studies using bias and
precision statistics to compare cardiac output measurement techniques.
J Clin Monit Comput 1999; 15:85–91.
19 Olofsen E, Dahan A, Borsboom G, et al. Improvements in the application
and reporting of advanced Bland–Altman methods of comparison. J Clin
Monit Comput 2015; 29:127–139.
20 Perrino AC, O’Connor T, Luther M. Transtracheal Doppler cardiac output
monitoring: comparison to thermodilution during noncardiac surgery.
Anesth Analg 1994; 78:1060–1066.
21 Messina A, Longhini F, Coppo C, et al. Use of the fluid challenge in critically
ill adult patients: a systematic review. Anesth Analg 2017; 125:1532–
1543.
22 Critchley LA, Lee A, Ho AM-H. A critical review of the ability of continuous
cardiac output monitors to measure trends in cardiac output. Anesth Analg
2010; 111:1180–1192.
23 Critchley LA, Yang XX, Lee A. Assessment of trending ability of cardiac
output monitors by polar plot methodology. J Cardiothorac Vasc Anesth
2011; 25:536–546.
 Monitoring haemodynamic response to fluid-challenge in ICU 143

24 Monnet X, Chemla D, Osman D, et al. Measuring aortic diameter improves
accuracy of esophageal Doppler in assessing fluid responsiveness. Crit
Care Med 2007; 35:477–482.
25 Aya HD, Ster IC, Fletcher N, et al. Pharmacodynamic analysis of a fluid
challenge. Crit Care Med 2016; 44:880–891.
26 Nunes TSO, Ladeira RT, Bafi AT, et al. Duration of hemodynamic effects of
crystalloids in patients with circulatory shock after initial resuscitation. Ann
Intensive Care 2014; 4:25.
27 Singer M, Clarke J, Bennett ED. Continuous hemodynamic
monitoring by esophageal Doppler. Crit Care Med 1989; 17:447 –
452.
28 Valtier B, Cholley BP, Belot JP, et al. Noninvasive monitoring of cardiac
output in critically ill patients using transesophageal Doppler. Am J Respir
Crit Care Med 1998; 158:77–83.
29 Franchi F, Silvestri R, Cubattoli L, et al. Comparison between an uncalibrated
pulse contour method and thermodilution technique for cardiac output
estimation in septic patients. Br J Anaesth 2011; 107:202–208.
30 Nichols WW, Edwards DG. Arterial elastance and wave reflection
augmentation of systolic blood pressure: deleterious effects and
implications for therapy. J Cardiovasc Pharmacol Ther 2001; 6:5–21.
31 Pauca AL, Wallenhaupt SL, Kon ND, et al. Does radial artery pressure
accurately reflect aortic pressure? Chest 1992; 102:1193–1198.

Eur J Anaesthesiol 2019; 36:135–143

Copyright © European Society of Anaesthesiology. Unauthorized reproduction of this article is prohibited.