Anaesth Intensive Care 2013; 41: 736-741

Assessment of the plethysmographic variability index as a

predictor of fluid responsiveness in critically ill patients: a
pilot study
A. K. Baker*, R. J. O. Partridge†, E. Litton‡, K. M. Ho§
Department of Intensive Care Medicine, Royal Perth Hospital, Perth, Western Australia, Australia

Summary
Optimising intravascular volume in patients with hypotension requiring vasopressor support is a key challenge
of critical care medicine. The optimal haemodynamic parameter to assess fluid responsiveness in critically ill
patients, particularly those requiring a noradrenaline infusion and mechanical ventilation, remains uncertain.
This pilot study assessed the accuracy of the plethysmographic variability index (PVI), (Radical-7 pulse
co-oximeter, Masimo®, Irvine, CA, USA) in predicting fluid responsiveness in 25 patients who required
noradrenaline infusion to maintain mean arterial pressure over 65 mmHg and were mechanically ventilated
with a ‘lung-protective’ strategy, and whether administering a fluid bolus was associated with a change in PVI
(Δ PVI). In this study, fluid responsiveness was defined as an increase in stroke volume of greater than 15%
after a 500 ml bolus of colloid infusion over 20 minutes. Of the 25 patients included in the study, only 12 (48%)
were considered fluid responders. As static haemodynamic parameters, PVI, central venous pressure and
inferior vena cava distensibility index were all inaccurate at predicting volume responsiveness with PVI being
the least accurate (area under the receiver operating characteristic curve=0.41, 95% confidence interval 0.18
to 0.65). However, fluid responsiveness was associated with a change in PVI, but not a change in heart rate or
central venous pressure. This association between Δ PVI and fluid responsiveness may be a surrogate marker of
improved cardiac output following a fluid bolus and warrants further investigation.
Key Words: fluid therapy, haemodynamic monitoring, preload, pulse variations

Optimising intravascular volume in patients with It is well established that left ventricular stroke
hypotension requiring vasopressor is a key challenge volume will change with changes in intrathoracic
of critical care medicine. Evidence suggests that both pressure, and by measuring variation in the arterial
inadequate and excessive fluid resuscitation may pressure waveform during mechanical ventilation
worsen clinical outcomes1,2. An ideal haemodynamic inadequate preload or fluid responsiveness can be
parameter to guide fluid optimisation should be assessed4. Advances in technology have led to the
accurate, non-invasive and continuous, allowing development of a non-invasive device similar in
treatment to be titrated according to the patient’s appearance to a pulse oximeter that measures the
response. Central venous pressure (CVP) has been plethysmographic wave form amplitude (perfusion
widely used as a guide to reflect the cardiac preload. index [PI]) and the variation of this amplitude over a
However, recent evidence suggests that both as a static given respiratory cycle (plethysmographic variability
and dynamic trend parameter, CVP is not useful to index [PVI]).
predict fluid responsiveness in critically ill patients3.
PVI is calculated as follows:
PVI = [(PImax – PImin) / PImax] × 100%.
* MBBS, MRCP, FRCA, DICM, Senior Registrar.
† MB, BS, BA, MRCP, FRCA, Senior Registrar.
‡ MB, ChB, FCICM, MSc, Staff Specialist and Clinical Senior Lecturer, Recent meta-analyses have suggested that
School of Medicine and Pharmacology, University of Western Australia.
§ MB, BS, MPH, PhD, FRCP, FANZCA, FCICM, Staff Specialist and plethysmographic indices such as PVI are accurate
Clinical Associate Professor, School of Polulation Health, University of in predicting fluid responsiveness, especially during
Western Australia.
positive pressure ventilation5,6. However, the
Address for correspondence: Dr A. K. Baker. Email: andrewkellasbaker@
doctors.org.uk majority of these studies included mainly elective
Accepted for publication on July 14, 2013 surgical patients. The accuracy of PVI to predict
Anaesthesia and Intensive Care, Vol. 41, No. 6, November 2013
 Assessment of plethysmographic variability index as a predictor 737
fluid responsiveness
fluid responsiveness in critically ill patients who are blood pressure, urine output, arterial lactate
mechanically ventilated with a small tidal volume and concentration, noradrenaline dose, positive end
require vasopressor such as noradrenaline remains expiratory pressure, peak airway pressure, tidal volume,
controversial7–11. CVP, PI, PVI (Radical 7 signal extraction pulse co-
Anatomically, the inferior vena cava (IVC) is oximeter, Masimo, Irvine, CA, USA), maximum and
directly connected to the right atrium without any minimum IVC diameter with ventilation, and stroke
venous valves. As such, its dimension is related to volume. The IVC dimensions were measured using
the right atrial pressure and changes in intrathoracic ultarasound via a subcostal view (iE33xMATRIX,
pressure. Although non-invasive measurement of Koninklijke Philips®, Andover, MA, USA) just
IVC diameter to assess preload and predict fluid proximal to the origin of the suprahepatic vein, and
responsiveness makes physiological sense and is the IVC distensibility index was defined as (maximum
supported by several small studies12–14, its accuracy IVC diameter – minimum IVC diameter) / minimum
and general utility in patients who are mechanically IVC diameter)12. In this study, stroke volume was
ventilated with a small tidal volume (6 to 7 ml/ kg) estimated by multiplying the velocity-time integral,
and require noradrenaline to maintain blood pressure averaged over three consecutive heart beats, with the
remains unproven15. cross-sectional area of the left ventricular outflow
We hypothesised that PVI is accurate in predicting tract using apical 5-chamber and parasternal short-axis
fluid responsiveness in patients who are mechanically views, respectively. All echocardiographic data were
ventilated with a small tidal volume and require recorded and reviewed by an investigator who has
noradrenaline to maintain blood pressure. In a Postgraduate Diploma in Echocardiography and
this study, we compared PVI to CVP and IVC
was blinded to the fluid responsive status of the
distensibility index for predicting fluid responsiveness
patients. Disagreements between the investigator who
in such patients. Furthermore, we assessed whether
fluid responsiveness was associated with a change in
Table 1
PVI (Δ PVI), CVP (Δ CVP) or heart rate (Δ HR). Characteristics of the patients receiving a fluid bolus in whom PVI
and IVC distensibility index were compared as means of predicting
fluid responsiveness (n=25)
METHODS
After obtaining Royal Perth Hospital Ethics Median (IQR)
Committee approval (Approval No. EC2012/031), 25 unless otherwise
stated
critically ill adult patients were recruited for this study.
All were mechanically ventilated with a small tidal Male:female, n 17:8
volume (<8 ml/kg) and had hypotension requiring Age, years 62 (47–72)
noradrenaline to maintain a mean arterial blood Weight, kg 73 (67–81)
pressure >65 to 70 mmHg. The sample size of this HR, beats/min 88 (74–99)
pilot study was based on the sample size of similar
Systolic blood pressure, mmHg 107 (98–120)
studies reported in the literature13,14. Patients were
Diastolic blood pressure, mmHg 56 (50–61)
eligible for inclusion if they fulfilled all of the
following criteria: (a) requiring noradrenaline infusion Mean arterial pressure, mmHg 73 (69–79)
for hypotension through an internal jugular or sub- Noradrenaline dose, µg/kg/min 0.09 (0.04–0.27)
clavian central venous catheter; (b) requiring mandatory Central venous pressure, mmHg 9 (8–13)
mode positive pressure mechanical ventilation with Positive end-expiratory pressure, cmH2O 5 (5–8)
a small tidal volume; and (c) for clinical reasons, the
Peak airway pressure, cmH2O 22 (17–27)
attending physician planned to administer a 500 ml
intravenous colloid fluid bolus (either 4% albumin Tidal volume, ml/kg 7 (6–8)
or Gelofusine [B. Braun, Melsungen, Germany]) to Admission category, n
optimise the haemodynamics of the patient. Patients Medical 7
were excluded if they were not in sinus rhythm or had Trauma 6
any spontaneous respiratory effort above the mand- General surgical 5
atory ventilation because PVI can be less accurate in
Cardiothoracic 5
patients with arrhythmias or who are spontaneously
breathing6. Vascular surgery 2

Before administering the fluid bolus, baseline

PVI=plethysmographic variability index, IVC=inferior vena cava,
measurements included body weight, HR, arterial IQR=interquartile range, HR=heart rate.

Anaesthesia and Intensive Care, Vol. 41, No. 6, November 2013

738 A. K. Baker, R. J. O. Partridge et al

Table 2
Baseline and changes in PVI and echocardiographic data after fluid bolus (n=25)

Variables Median (IQR)

Before fluid bolus After fluid bolus P value
CVP, mmHg 9 (8–13) 12 (10–14) 0.001
HR, beats/min 88 (74–99) 87 (77–97) 0.148
PVI 13 (10–17) 11 (9–15) 0.009
PI 0.7 (0.3–1.3) 1.1 (0.7–1.7) 0.103
Minimum IVC diameter, cm 1.8 (1.6–2.1) Not available
Respiratory variation in IVC 11 (6–22) Not available
diameter, %
Stroke volume, ml 51 (42–60) 60 (47–70) 0.001
Cardiac output, litre/min 4.5 (3.2–5.5) 5.0 (3.7–5.9) 0.001

PVI=plethysmographic variability index, IQR=interquartile range, CVP=central venous pressure,

HR=heart rate, PVI=plethysmographic variability index, PI=perfusion index, IVC=inferior vena cava.

was blinded to the fluid responsiveness status and the is associated with an increased likelihood of fluid
investigators who performed the echocardiography responsive (e.g. PVI). Conversely, an area under
were about 10%, mainly in some measurements of the receiver operating characteristic curve that is
stroke volume by the velocity-time intergral. substantially <0.5 will suggest that a lower value of
After administering 500 ml of colloid fluid, PVI, that haemodynamic parameter is associated with an
PI, CVP, HR and stroke volume measurements increased likelihood of fluid responsiveness (e.g. CVP).
were immediately repeated. The ventilatory settings, In order to assess whether the trend or change
sedation and noradrenaline doses were kept of a haemodynamic parameter was associated with
constant throughout the study period, as was the fluid therapy, the difference in haemodynamic para-
patient’s position. In this study, fluid responsiveness meters before and after fluid challenge was assessed by
was defined as an increase in stroke volume of a paired t-test. We used the non-parametric Spearman
>15% after the fluid challenge6. correlation coefficient (rs) with its 95% confidence
The discriminatory ability of each haemodynamic interval (CI), after applying Bonferroni corrections16,
parameter to determine fluid responsiveness was to assess whether ΔCVP, ΔHR and ΔPVI had any
assessed by area under the receiver operating correlations with the changes in stroke volume after
characteristic curve. When the area under the receiver fluid therapy. All statistical analyses were conducted
operating characteristic curve of a haemodynamic using SPSS for Windows (version 19.0, 2012 IBM,
parameter is substantially >0.5, this suggests that Armonk, NY, USA) and a P value <0.05 was taken
a higher value of that haemodynamic parameter as significant.

RESULTS
Table 3 Twenty-five patients were enrolled in the study over
Area under the receiver operating characteristic curves of different static
haemodynamic parameters to differentiate fluid responsiveness, defined a period of nine months (April to December 2012).
by an increment in stroke volume of at least 15% from baseline In one patient, accurate IVC measurements could
not be obtained due to a thick dressing over a high
Static variable Area under the curve (95% CI)
midline laparotomy incision, but in all other patients
CVP 0.66 (0.44–0.089) the required measurements were taken. No significant
HR 0.55 (0.31–0.78) tricuspid regurgitation waveform on the CVP
IVC minimum diameter 0.55 (0.31–0.79) tracing or significant right ventricular dysfunction
IVC distensibility index 0.46 (0.22–0.69) in the echocardiographic apical 5-chamber view was
PVI 0.41 (0.18–0.65)
observed in the study patients.
All patients were mechanically ventilated without
CI=confidence interval, CVP=central venous pressure, HR=heart significant spontaneous breathing effort and were
rate, IVC=inferior vena cava, PVI=plethysmographic variability treated with noradrenaline infusion to maintain a
index.

Anaesthesia and Intensive Care, Vol. 41, No. 6, November 2013

 Assessment of plethysmographic variability index as a predictor 739
fluid responsiveness
Table 4 responsiveness in patients who were mechanically
Area under the receiver operating characteristic curves of different ventilated with a small tidal volume and required
dynamic haemodynamic parameters to differentiate fluid responsive-
ness, defined by an increment in stroke volume of at least 15% from noradrenaline. This confirms the results of many
baseline previous studies that a single measurement of a static
Change in variable Area under the curve haemodynamic parameter including CVP is not useful
before and after fluid (95% CI) in predicting fluid responsiveness or the preload
ΔPVI 0.82 (0.64–0.99) status of critically ill patients3. This is, perhaps, not
ΔCVP 0.63 (0.40–0.85) surprising because substantial individual variations
ΔHR 0.55 (0.31–0.78)
in all static haemodynamic parameters are expected
from complicated interactions between intrathoracic
pressure, ventricular compliance, HR, vasomotor
CI=confidence interval, PVI=plethysmographic variability index,
tone and intravascular volume.
CVP=central venous pressure, HR=heart rate. We have rejected our initial hypothesis that PVI
is accurate at predicting fluid responsiveness in
mean arterial blood pressure above 65 to 70 mmHg. this group of patients. In fact, of the parameters
Three patients were also treated with dobutamine to we assessed, PVI was demonstrated to be the least
improve peripheral perfusion (2.5 to 5.7 µg/kg/minute) accurate means of discriminating fluid responders
from non-responders. This conclusion conflicts
but no other vasoactive drugs such as milrinone or
with the outcome of many studies on surgical
levosimendan were used in any of the study patients
patients which show that a single reading of PVI
during the study period. The baseline characteristics
can be useful at predicting fluid responsiveness6.
of the participants are described in Table 1. Overall,
This discrepancy may be due to the fact that all our
the PVI, CVP, stroke volume and cardiac output, but
patients were treated with a vasopressor that would
not HR or PI, of the whole cohort were significantly
affect the vasomotor tone and peripheral perfusion.
different after a fluid bolus (Table 2). However, of the
Furthermore, the tidal volume used in our critically
25 patients included in the study, only 12 (48%, 95%
CI 30 to 67%) were considered as fluid responders
with an increase in stroke volume of greater than 15%
22.5 Fluid responders
after the fluid bolus. No
As a static or single haemodynamic measurement, Yes
the baseline PVI, CVP, IVC diameter and IVC 20.0
distensibility index were not reliable in discriminating
fluid responders from non-responders (Table 3). A
comparison of these static measurements showed 17.5
that PVI was the least accurate at predicting fluid
responsiveness (area under receiver operating
PVI (95% CI)

characteristic curve=0.41, 95% CI 0.18 to 0.65). 15.0

After the fluid bolus, only ΔPVI (i.e. not ΔCVP or
ΔHR) was associated with changes in cardiac ouput
(responders: mean change in PVI = -7.3, standard 12.5

deviation 6.1 versus non-responders: mean change

in PVI = -0.4, standard deviation 5.2, Table 4 and
10.0
Figure 1). After applying Bonferroni corrections, Significant difference ( P=0.006) in
the changes in PVI between fluid
changes in stroke volume after fluid therapy were responder and fluid non-responders
associated with ΔPVI (rs = -0.58, 95% CI 0.24 to 0.90;
7.5
P=0.006) but not ΔCVP (rs = -0.37, P=0.21) or ΔHR
(rs = -0.40, P=0.15, Figure 2). Before fluid bolus After fluid bolus

Figure 1: Difference in the changes in PVI between patients

DISCUSSION who responded to a 500 ml colloid fluid bolus and those who
Our results suggested that as a static single haemo- did not. Mean difference in change in PVI between responders
dynamic measurement, all standard haemodynamic and non-responders= 6.9, 95% CI 2.3 to 11.6. Fluid responsive-
ness was defined by an increment in stroke volume of at least
parameters including PVI, CVP, IVC distensibility 15% from baseline. PVI=plethysmographic variability index,
index and PI were not useful in predicting fluid CI=confidence interval.

Anaesthesia and Intensive Care, Vol. 41, No. 6, November 2013

 740 A. K. Baker, R. J. O. Partridge et al

may have been a source of measurement error

and hence affected accuracy in predicting fluid
60
responsiveness. However, this is a recognised
Spearman
Spearmancorrelation
correlation limitation of any ultrasound measurements in patients
coefficient==-0.58,
coefficient -0.58,95%
95%CICI who are mechanically ventilated. Second, we did not
0.24-0.90;P=0.006
0.21–0.90; p=0.006
plan to repeat the IVC distensibility index after the
40 fluid bolus in this study. As such, it is possible than
SV,(ml)
ml

the Δdistensibility index could be more useful than

DeltaSV

the ΔPVI in discriminating fluid responders from

Delta

20
non-responders12, and this also merits further investig-
ation. Third, the sample size of this study was small
and hence we cannot exclude a type 2 error in concl-
uding that PVI or IVC distensibility index were not
0 useful predictors as a static haemodynamic parameters.
Furthermore, evaluation of the interactions between
airway pressure, vasopressor doses and different
-20 -10 0 10 haemodynamic parameters with the small sample size
Delta PVI
Delta PVI of this pilot study was not possible. Finally, although
this study specifically looked at non-invasive means
Figure 2: Correlations between changes in stroke volume and
changes in PVI after 500 ml of colloid fluid bolus. Dotted lines of predicting fluid responsiveness, it would also have
delineate 95% CI of the best fitted line by a quadrantic equation. been interesting to have arterial waveform-derived
PVI=plethysmographic variability index, CI=confidence interval, haemodynamic variables and stroke volume variations
SV=stroke volume.
as comparisons17.
In summary, in this group of critically ill patients
ill patients was smaller than those used in the elective
requiring a noradrenaline infusion, PVI, CVP
surgical patients (7 versus 10 ml/kg), reducing the
and IVC distensibility index were all inaccurate at
variability in plethysmographic amplitude and hence
predicting volume responsiveness. However, fluid
the accuracy of PVI. Similarly, previous studies that
responsiveness was associated with a change in PVI,
suggested the IVC distensibility index was useful to
suggesting that ΔPVI may be a surrogate marker
predict fluid responsiveness also used a larger tidal
of improved cardiac output following a fluid bolus.
volume (>8 to 8.5ml/ kg) than in our study and this
This observation warrants further investigation.
may also explain why our results were different from
the positive results reported in the previous studies12,13.
An interesting observation of our study was that ACKNOWLEDGMENTS
in those patients who responded to fluid therapy, the We thank the Medical Research Foundation
improvement in cardiac output was associated with a of Royal Perth Hospital for funding the Masimo
significant change in PVI. While this may be of little Radical-7 haemodynamic monitor used in this study.
clinical use in predicting fluid responsiveness because
it is obtained only after the fluid has been given, it may References
be of value as a haemodynamic trend monitor. Our   1. Rivers E, Nguyen B, Havstad S, Ressler J, Muzzin A, Knoblich
results would suggest that ΔPVI may have some weak B et al. Early goal-directed therapy in the treatment of severe
association with improved stroke volume (Figure sepsis and septic shock. N Engl J Med 2001; 345:1368-1377.
 2. Wiedemann HP, Wheeler AP, Bernard GR, Thompson BT,
2). This association is likely due to the fact that PVI Hayden D, deBoisblanc B et al. Comparison of two fluid-man-
reflects peripheral pulse volume and perfusion, and a agement strategies in acute lung injury. N Engl J Med 2006;
change in stroke volume may increase the amplitude 354:2564-2575.
of the peripheral pulse volume or reduce its variation   3. Marik PE, Baram M, Vahid B. Does central venous pressure
predict fluid responsiveness? A systematic review of the litera-
during respiration. Whether a change in PVI is a ture and the tale of seven mares. Chest 2008; 134:172-178.
reliable surrogate marker for a change in cardiac   4. Marik PE, Cavallazzi R, Vasu T, Hirani A. Dynamic changes
output in response to fluid therapy or haemorrhage in in arterial waveform derived variables and fluid responsiveness
patients who are mechanically ventilated and require in mechanically ventilated patients: a systematic review of the
literature. Crit Care Med 2009; 37:2642-2647.
noradrenaline warrants further investigation.   5. Sandroni C, Cavallaro F, Marano C, Falcone et al. Accuracy
The last consideration is the limitations of this of plethysmographic indices as predictors of fluid responsive-
study. First, in some of our patients the subcostal ness in mechanically ventilated adults: a systematic review and
echocardiographic windows were not optimal, which meta-analysis. Intensive Care Med 2012; 38:1429-1437.

Anaesthesia and Intensive Care, Vol. 41, No. 6, November 2013

 Assessment of plethysmographic variability index as a predictor 741
fluid responsiveness
 6. Yin JY, Ho KM. Use of plethysmographic variability index 12. Feissel M, Michard F, Faller JP, Teboul JL. The respiratory vari-
derived from the Massimo((R)) pulse oximeter to predict ation in inferior vena cava diameter as a guide to fluid therapy.
fluid or preload responsiveness: a systematic review and meta- Intensive Care Med 2004; 30:1834-1837.
analysis. Anaesthesia 2012; 67:777-783. 13. Barbier C, Loubieres Y, Schmit C, Hayon J, Ricome JL, Jardin
 7. Natalini G, Rosano A, Taranto M, Faggian B, Vittorielli F et al. Respiratory changes in inferior vena cava diameter are
E, Bernardini A. Arterial versus plethysmographic dynamic helpful in predicting fluid responsiveness in ventilated septic
indices to test responsiveness for testing fluid administration patients. Intensive Care Med 2004; 30:1740-1746.
in hypotensive patients: a clinical trial. Anesth Analg 2006; 14. Machare-Delgado E, Decaro M, Marik PE. Inferior vena cava
103:1478-1484. variation compared to pulse contour analysis as predictors of
  8. Feissel M, Teboul JL, Merlani P, Badie J, Faller JP, Bendjelid K. fluid responsiveness: a prospective cohort study. J Intensive
Plethysmographic dynamic indices predict fluid responsiveness Care Med 2011; 26:116-124.
in septic ventilated patients. Intensive Care Med 2007; 33:993- 15. Wetterslev M, Haase N, Johansen RR, Perner A. Predicting
999. fluid responsiveness with transthoracic echocardiography is not
 9. Loupec T, Nanadoumgar H, Frasca D, Petitpas F, Laksiri L, yet evidence based. Acta Anaesthesiol Scand 2013; 57:692-697.
Baudouin D et al. Pleth variability index predicts fluid respon- 16. Ho KM. Ten caveats of interpreting correlation coefficient in
siveness in critically ill patients. Crit Care Med 2011; 39:294- anaesthesia and intensive care research. Anaesth Intensive
299. Care 2012; 40:595-597.
10. Biais M, Cottenceau V, Petit L, Masson F, Cochard JF, Sztark 17. Mandeville JC, Colebourn CL. Can transthoracic echocardi-
F. Impact of norepinephrine on the relationship between pleth ography be used to predict fluid responsiveness in the criti-
variability index and pulse pressure variations in ICU adult cally ill patient? A systematic review. Crit Care Res Pract 2012;
patients. Crit Care 2011; 15:R168. 2012:513480.
11. Monnet X, Guerin L, Jozwiak M, Bataille A, Julien F,
Richard C et al. Pleth variability index is a weak predictor of
fluid responsiveness in patients receiving norepinephrine. Br
J Anaesth 2013; 110:207-213.

Anaesthesia and Intensive Care, Vol. 41, No. 6, November 2013