BJA Advance Access published March 3, 2015 British Journal of Anaesthesia Page 1 of 8 doi:10.1093/bja/aev001
Comparison of cardiac output measured by oesophageal Doppler ultrasonography or pulse pressure contour wave analysis A. Caillard 1,2,3, E. Gayat1,2,3, A. Tantot1,2,3, G. Dubreuil 1,2, E. M’Bakulu1,2, C. Madadaki 1,2, F. Bart 1,2, D. Bresson 2,4, S. Froelich 2,4, A. Mebazaa1,2,3 and F. Valle´e1,2* 1
Department of Anesthesiology and Critical Care, Hoˆpitaux Universitaires St-Louis-Lariboisie`re-Fernand Widal, Paris, France University Paris Diderot, Paris, France 3 INSERM UMR-942, Paris, France 4 Department of Neurosurgery, Hoˆpitaux Universitaires St-Louis-Lariboisie`re, Paris, France 2
Editor’s key points † Pulse pressure contour wave analysis of cardiac output (PPCO) is a promising monitoring technique. † PPCO analysed using 9 novel algorithms was compared with oesophageal Doppler monitoring of cardiac output (DCO) in 62 neurosurgical subjects. † There were significant differences in bias and concordance between PPCO and DCO depending on algorithm and therapeutic interventions. † Continuous PPCO monitoring with a specific algorithm could offer advantages compared with DCO alone.
Background. Maintaining adequate organ perfusion during high-risk surgery requires continuous monitoring of cardiac output to optimise haemodynamics. Oesophageal Doppler Cardiac Output monitoring (DCO) is commonly used in this context, but has some limitations. Recently, the cardiac output estimated by pulse pressure analysis- (PPCO) was developed. This study evaluated the agreement of cardiac output variations estimated with 9 non-commercial algorithms of PPCO compared with those obtained with DCO. Methods. High-risk patients undergoing neurosurgery were monitored with invasive blood pressure and DCO. For each patient, 9 PPCO algorithms and DCO were recorded before and at the peak effect for every haemodynamic challenge. Results. Sixty-two subjects were enrolled; 284 events were recorded, including 134 volume expansions and 150 vasopressor boluses. Among the 9 algorithms tested, the LiljestrandZander model led to the smallest bias (0.03 litre min21 [21.31, +1.38] (0.21 litre min21 [21.13; 1.54] after volume expansion and 20.13 litre min21 [21.41, 1.15] after vasopressor use). The corresponding percentage of the concordance was 91% (86% after volume expansion and 94% after vasopressor use). The other algorithms, especially those using the Winkessel concept and the area under the pressure wave, were profoundly affected by the vasopressor. Conclusions. Among the 9 PPCO algorithms examined, the Liljestrand-Zander model demonstrated the least bias and best limits of agreement, especially after vasopressor use. Using this particular algorithm in association with DCO calibration could represent a valuable option for continuous cardiac output monitoring of high risk patients. Clinical Trial Registration. Comite´ d’e´thique de la Socie´te´ de Re´animation de Langue Franc¸aise No. 11 –356. Keywords: cardiac output; Doppler; pulse wave analyses Accepted for publication: 17 December 2014
In high risk surgical patients, haemodynamic monitoring is recommended to maintain suitable organ perfusion.1 Cardiac output is a key cardiovascular parameter and a major determinant of tissue oxygenation.2 Continuous measurement is preferable to intermittent measurement, especially in the operating theatre when rapid changes in cardiovascular function are often observed. Alternative minimally invasive methods for continuously monitoring cardiac output have then been developed. Two continuous monitors are more frequently used: 1) the oesophageal Doppler method (DCO), and 2) the pulse pressure contour method (PPCO). DCO is accurate in the
operating theatre;3 however, this technique has limitations, including its inability to monitor a continuous signal during surgical haemostasis as a result of electrical interference and lack of access to the mouth in oral surgery and neurosurgery. PPCO is minimally invasive and not operator dependent. A method of calculating stroke volume from the contour of the arterial pressure curve was first described in 1899.4 PPCO is based on the hypothesis that the waveform of blood pressure is directly related to the variation in the stroke volume. Many algorithms have been proposed: some make a comparison with a closed hydraulic or electric circuit, such as the
& The Author 2015. Published by Oxford University Press on behalf of the British Journal of Anaesthesia. All rights reserved. For Permissions, please email: [email protected]
Downloaded from http://bja.oxfordjournals.org/ at University of California, San Francisco on March 11, 2015
* Corresponding author. E-mail: [email protected]
BJA
Materials and methods This prospective, non-interventional, non-randomized study was conducted in neurosurgical patients under general anaesthesia, between November 2011 and July 2012, in the Department of Anaesthesiology at Lariboisie`re University Hospital (Paris, France). This study was approved by the institutional ethics committee (Comite´ d’e´thique de la Socie´te´ de Re´animation de Langue Franc¸aise No. 11–356). All subjects gave written informed consent. Haemodynamic monitoring, including invasive blood pressure and DCO, represented the usual care of patients undergoing neurosurgery with reduced cerebral compliance and/or potential bleeding in our institution. The exclusion criteria were age ,18 yrs, pregnancy, contraindication to the use of DCO and chronic cardiac arrhythmia.
Study protocol All subjects were orally premedicated with hydroxyzine (1 mg kg21) 1 h before surgery. Anaesthesia was induced with propofol and remifentanil target controlled infusion. Tracheal intubation was facilitated with a atracurium 0.5 mg kg21, and mechanical ventilation used the volume-controlled mode, with tidal volumes of 7 ml kg21 with positive end-expiratory pressure of 5 cm H2O and a respiratory rate of 12– 16 breaths min21 to maintain end tidal CO2 of approximately 4.7 Kpa. After induction of anaesthesia, a radial arterial line and DCO probe were inserted. Both were connected to CombiQw (Deltex Medical Ltd.). DCO values were averaged over 5 cardiac cycles.
Haemodynamic management Mean arterial pressure (MAP), measured the day before the surgical intervention, was considered to be the MAP of reference (MAPref). According to our standard of care, a decrease in MAP.20% of MAPref led to a therapeutic intervention according to the choice of the physician in charge: including a fluid
Page 2 of 8
challenge (250 ml of saline in 10 min) or a bolus of vasopressor. Three different vasopressors were available according to the choice of the physician (bolus of 50 mcg of phenylephrine, 5 mcg of norepinephrine or 9 mg of ephedrine).
Data collection After zeroing the system to atmospheric pressure, the arterial pressure waveform was carefully checked using a fast flush test to ensure optimal harmonics of the arterial pressure measurement system. The arterial pressure signal was transferred from the bedside monitor to the DCO system using a serial cable. Nine nonproprietary PPCO algorithms built into the CardioQ-Combi software provide continuous cardiac output values. The 9 PPCO algorithms are described in Table 1 and in the Supplementary material. Haemodynamic data were automatically recorded by the monitor every 5 s. Initial calibration of the PPCO was performed using the DCO value before any intervention and after stabilizing heart rate and arterial pressure (,5% variation over a 1-min period). Each therapeutic event was automatically recorded on the monitor using two predefined categories: fluid challenge or vasopressor bolus. The peak effect of the therapeutic action was recorded 5 min after the fluid challenge or at the peak MAP after administration of vasopressor.
Statistical analysis The results are expressed as mean (SD) or as median (interquartile range [IQR]), depending on normal distribution of the
Table 1 Characteristics study subjects Characteristics
All subjects (n 562) (%)
Gender (Male/Female)
26 (42)/36 (58)
Age (yrs)
54 [39 –65]
BMI (kg m22)
25 [21 –29]
American Society of Anesthesiologists Physical Status 1
11 (18)
2
36 (73)
3
6 (10)
Comorbidities History of hypertension
21 (34)
Chronic heart failure
3 (5)
Coronary artery disease
1 (2)
Chronic obstructive pulmonary disease
21 (34)
Diabetes mellitus
5(8)
Type of surgery Cerebral tumour
52 (84)
Aneurism
4 (6)
Spine surgery
4 (6)
Other Duration of surgery (min)
2 (3) 335 [248 –440]
Perioperative fluid administration (ml)
3512 [3750 –4500]
Perioperative bleeding (ml)
221 [100 –363]
Perioperative diuresis (ml)
957 [500 –1350]
Downloaded from http://bja.oxfordjournals.org/ at University of California, San Francisco on March 11, 2015
Windkessel concept, and others analyse the area under the systolic portion of the arterial pressure waveform. Algorithm validity has been verified in a variety of patients and circumstances, but performance could be compromised in the presence of haemodynamic instability, cardiac arrhythmias, or other factors that disturb the arterial pressure waveform, particularly during vasopressor use.5 6 To increase the reliability of PPCO, some commercial systems have calibrated their algorithms using thermodilution.5 A prototype for continuously monitoring cardiac output was recently introduced (CombiQw, Deltex Medical, Chichester, Sussex, UK) that combines oesophageal Doppler and PPCO and includes 9 algorithms to estimate cardiac output after calibration using the DCO value according to a study by Sun and colleagues.7 The present study compared, in patients under general anaesthesia, the agreement of cardiac output measurements obtained using 9 non-commercial algorithms of PPCO with measures of cardiac output obtained using DCO alone used as a reference. We also analysed variations in cardiac output during haemodynamic challenges, including volume expansion or administration of a vasopressor bolus.
Caillard et al.
BJA
Continuous cardiac output measure calibrated with oesophageal Doppler
X1
X2
3
1 0 –1 –2
2 1 0 –1 –2 –3
–3 2
3
4
5
6
7
8
3
4
6
7
0 –1 –2
2
6
7
1 0 –1 –2
3
4
5
6
7
–2 –3 6
7
DCO-PPCO/2 (l m–1)
–1 –2
8
3
5
4
6
7
8
DCO-PPCO/2 (l m–1) 3
X8
2 1 0 –1 –2 –3
5
0
2
DCO-PPCO (l m–1)
DCO-PPCO (l m–1)
–1
8
1
8
3
0
7
2
DCO-PPCO/2 (l m–1)
1
6
–3 2
X7
5
X6
2
8
3
4
4
3
DCO-PPCO/2 (l m–1)
3
3
DCO-PPCO/2 (l m–1)
–3
2
–2
8
DCO-PPCO (l m–1)
DCO-PPCO (l m–1)
DCO-PPCO (l m–1)
1
2
–1
X9
2 1 0 –1 –2 –3
2
3
4
5
6
7
DCO-PPCO/2 (l m–1)
8
2
3
4
5
6
7
8
DCO-PPCO/2 (l m–1)
Fig 1 Bland-Altman plots for absolute values of PPCO compared with DCO. Agreement between PPCO and DCO measurements according to Bland-Altman analysis. Only one marker for the subject was represented in the graph. Marker size is relative to the number of observations per subject.
variables. Bland-Altman analysis for repeated measurements was used to assess agreement between DCO and each PPCO algorithm.8 Bias was defined as the mean difference between DCO and PPCO measurements. Limits of agreement (LOA) were calculated as mean bias (1.96) SD. Percentage error (PE) was calculated to compare CO as PE (%)¼LOA/[(mean DCO+ mean PPCO)/2]. A PE≤30% indicates agreement of the technique with the reference method.9 The concordance of the variation of cardiac output measured by DCO and each PPCO algorithm was assessed by Bland-Altman analysis.8 Percentage of the concordance was estimated using the percentage of data in which the direction of change was in agreement in 4-quandrant plots. Acceptable concordance was .90% in
4-quadrant plots. Analyses were repeated in the fluid challenge and vasopressor bolus subgroups. Ability of the PPCO algorithms to detect responders to the fluid challenge was also assessed and compared with DCO. Subjects were defined as a «responder» to volume expansion when fluid challenge led to an increase in stroke volume .10%. P value ,0.05 was considered statistically significant. All statistical analyses were performed using R statistical software (The ‘R’ Foundation for Statistical Computing, Vienna, Austria).
Results A total of 62 subjects were included; subject characteristics are summarized in Table 1.
Page 3 of 8
Downloaded from http://bja.oxfordjournals.org/ at University of California, San Francisco on March 11, 2015
2
5
0
X5
–3
DCO-PPCO (l m–1)
5
3 X4
4
1
DCO-PPCO/2 (l m–1)
3
3
2
–3 2
DCO-PPCO/2 (l m–1)
2
X3
3 DCO-PPCO (l m–1)
2
DCO-PPCO (l m–1)
DCO-PPCO (l m–1)
3
BJA
Caillard et al.
Table 2 Agreement and concordance analysis between DCO and nine PPCO algorithms. PE, percentage of error; LOA, limit of agreement; As, area under the systolic arterial pressure curve; DBP, diastolic blood arterial pressure; HR, heart rate; k, calibration factor; MAP, mean arterial pressure; SBP, p systolic blood pressure; T, duration of cardiac cycle (T¼ HR/60); Tdia, duration of diastole (Tdia¼T – Tsys); Tsys, duration of systole (estimated as 30% of cardiac cycle time) Formulae
Bias [lower LOA; upper LOA]
Percentage of error (PE) %
X1
Mean arterial pressure
K1×MAP
20.28 [22.37; 1.81]
43
X2
Windkessel
K2×(SAP-DAP)×HR
20.20 [22.08; 1.68]
39
X3
Windkessel with RC decay
K3×(MAP×HR/60)×ln(SAP/DAP)
20.21 [22.10; 1.67]
39
X4
Liljestrand-Zander
0.05 [21.18; 1.28]
26
20.24 [22.19; 1.71]
40
20.27 [22.32; 1.77]
42
X5
Pressure root-mean-square
K4×[(SAP-DAP)×HR]/(SAP+DAP) p K5× (SBP-MAP)2×HR
X6
Herd
K6×(MAP-DBP)
Systolic area
K7×HR×As
20.26 [22.20; 1.68]
40
Systolic area with correction
K8×HR×As×(1+Ts/Td)
20.24 [22.13; 1.65]
39
X9
Systolic area with corrected impedance
K9×HR×(163+HR - 0.48×MAP)×As
20.18 [21.88; 1.52]
35
Concordance between DCO and each PPCO algorithm A total of 16 474 measurement pairs of cardiac output were analyzed with a median of 215 [105 –363] pairs per subject. Of the 9 algorithms considered, the X4 (Liljestrand-Zander model) was associated with the most limited bias at 0.05 litre min21, narrowest limits of agreement (upper and lower LOA at 21.18 litre min21 and 1.28 litre min21, respectively), and lowest PE (26%) (Table 2 and Fig. 1). Algorithms based on the Windkessel model (X2 and X3) or on the area under the systolic portion model (X7, X8 and X9) had an average bias of 20.2 litre min21 with PE of 39%.
Table 3 Agreement between the changes in cardiac output measurements using DCO and each PPCO algorithm. The table compiles the values of the bias, the upper and lower LOA and percentage of concordance with 10% exclusion zone of each algorithm for all therapeutic interventions and separation in fluid or vasopressor administration. LOA, limit of agreement Bias [lower LOA; upper LOA]
Percentage of concordance (with 10% exclusion zone)
X1 Mean arterial Pressure Vasopressor
20.96 [23.65; 1.73]
19
0.01 [21.77; 1.79]
42
Vasopressor
20.82 [23.09; 1.45]
47
Fluid
20.01 [21.68; 1.67]
74
Fluid X2 Windkessel
Concordance of changes in cardiac output after therapeutic intervention A total of 284 events were recorded, including 134 fluid challenges and 150 boluses of vasopressor, consisting of 32 bolus administrations of ephedrine, 61 of pheylephrine and 57 of norepinephrine (Table 3, Figs 2 and 3). Variations of DCO and PPCO induced by fluid challenge or vasopressor bolus are related in Table 1 of Supplementary material. Vasopressor induced a decrease in DCO of 58% with a mean value of 20.11 litre min21 [20.54, +0.02] (Supplementary Table 1). With respect to variation of cardiac output measurements after therapeutic intervention, X4 (Liljestrand-Zander Model) again was the most efficient with a bias of 0.03 litre min21 [21.31, +1.38] (0.21 litre min21 [21.13, +1.54] after volume expansion and 21.31 litre min21 [21.41, +1.15] after vasopressor bolus), with PA of 91% (86% after volume expansion and 94% after vasopressor). All other algorithms showed a low agreement after vasopressor. For a Windkessel model with a correction (X3), PA was 76% after volume expansion and 47% after vasopressor. For the area under the curve model with impedance correction (X9), PA was 64% after volume expansion and 43% after vasopressor.
Page 4 of 8
X3 Windkessel with RC decay Vasopressor
20.85 [23.16; 1.45]
47
Fluid
20.02 [21.7; 1.66]
76
20.13 [21.41; 1.15]
94
0.21 [21.13; 1.54]
86
X4 Liljestrand-Zander Vasopressor Fluid
X5 Pressure root-mean-square Vasopressor
20.9 [23.31; 1.5]
43
20.01 [21.68; 1.67]
79
Vasopressor
20.98 [23.55; 1.59]
37
Fluid
20.08 [21.85; 1.68]
75
20.87 [23.26; 1.52]
26
0.02 [21.6; 1.63]
68
Fluid X6 Herd
X7 Systolic area Vasopressor Fluid
X8 Systolic area with correction Vasopressor
20.82 [23.21; 1.56]
29
Fluid
20.01 [21.71; 1.69]
58
X9 Systolic area with correction Vasopressor Fluid
20.63 [22.62; 1.36]
43
0.07 [21.46; 1.59]
64
Downloaded from http://bja.oxfordjournals.org/ at University of California, San Francisco on March 11, 2015
X7 X8
BJA
Continuous cardiac output measure calibrated with oesophageal Doppler
X2
2 0 –2 –4 –2
–1
–0
1
2 0 –2 –4
2
–2 –4 1
0 –2 –4
–2 –4 1
–0
1
2
Average D (I m–1)
0 –2 –4 –1
–0
1
Average D (I m–1)
2
0 –2 –4 –1
–0
1
2
Average D (I m–1)
2
–2
1
X6
–2
X8
4
–0
2
2
Difference D (I m–1)
Difference D (I m–1)
Difference D (I m–1)
0
–0
–1
–2
–1
4
Average D (I m–1)
2
–1
–4
Average D (I m–1)
2
2
X7
–2
–2
–2
X5
4
Average D (I m–1) 4
0
2
Difference D (I m–1)
Difference D (I m–1)
Difference D (I m–1)
0
–0
1
2
2
X9
4 2 0 –2 –4 –2
–1
–0
1
2
Average D (I m–1)
Fig 2 Bland-Altman plots for DCO compared with PPCO during fluid challenge (black points) or bolus of vasopressors (blue points). The figure shows agreement between changes in cardiac output measurements by DCO and each PPCO algorithm according to the Bland-Altman representation during therapeutic challenge.
Differentiation between the responders and non-responders to fluid challenge using the Liljestrand-Zander Model 25% (33/134) of fluid challenges were responsive according to DCO with an increase in stroke volume .10%. When using X4 (Liljestrand-Zander Model) for classification of responders to fluid challenge, only 13% (17/134) were positive. In comparison with DCO, the sensibility for determining fluid challenge response by the Liljestrand–Zander model was 30% and specificity was 93%.
Discussion We analysed the agreement between DCO and 9 noncommercial PPCO algorithms during fluid challenge and vasopressors
administration. The model of Liljestrand-Zander showed the best performance, namely the lowest error rate and greater concordance with DCO value for static value and after therapeutic intervention, especially after use of vasopressors. The other 8 algorithms showed correct bias after fluid challenge compared with the DCO, but there was poor agreement after vasopressor administration. Although the PPCO seems to be a good alternative to monitoring the cardiac output during surgery, many of the commercialized algorithms had difficulties, especially during vasopressor administration. Meng and colleagues showed that an algorithm derived from the Windkessel concept accurately tracks changes in cardiac output after whole-body tilting to increase preload, whereas phenylephrine-induced change in vasomotor tone profoundly affected its ability to track changes in cardiac output. In this study, vasopressor
Page 5 of 8
Downloaded from http://bja.oxfordjournals.org/ at University of California, San Francisco on March 11, 2015
2
–1
–0
4
Average D (I m–1)
X4
–2
–1
–2
Average D (I m–1) 4
X3
4
Difference D (I m–1)
Difference D (I m–1)
Difference D (I m–1)
X1 4
BJA
Caillard et al.
X2
50
50
50
0 –50
0 –50 –100
0
50
–100
100
D by X1 (%)
D by X1 (%)
0 –50
50
100
–100
–50
0
50
0 –50
X7
100
D by X1 (%)
0 –50 –100
0
50
0
50
0 –50
–100
100
100
X8
0 –50
D by DTO (%)
0
50
100
X9
100
50
–100
–50
D by DTO (%)
–100 –50
100
–100 –50
100
50
50
50
D by DTO (%)
D by DTO (%)
0
X6
100
50
–100
100
–50
D by DTO (%)
–100
–100
–100
0
X5
100
50
–100
–50
50 0 –50 –100
–50
0
50
100
D by DTO (%)
–100
–50
0
50
100
D by DTO (%)
Fig 3 Concordance between changes in DCO and PPCO represented in four quadrants plots.
administration could induce opposite changes in cardiac ouput measuring by DCO or PPCO. The software enhancement does not seem to improve the ability of PPCO to adequately measure cardiac output when arterial compliance is modified. Our study confirmed the poor performance of the Windkessel model after a vasopressor bolus, which is similar to findings of previously published studies in the intensive care unit or operating theatre. The concordance between cardiac ouput measured with DCO and algorithms based on pressure wave form such as the Windkessel models (X4.5), Herd model (X6) and area under the curve (X7, X8, and X9), were poor after vasopressor administration. Furthermore, the relation between cardiac output variations and vasoconstrictor administration is complex and, as recently reported; depends on the volume status of the patient.10 11 Our results are in accordance with those studies: we observed a decrease
Page 6 of 8
of DCO induced by vasopressor which was mostly expected as our neurosurgical subjects were fluid non-responders in 75% of cases.10 11 Interestingly, as recently published,12 13 the LiljestrandZander model retained a good percentage of concordance with DCO (94%) during vasopressor use. The Liljestrand-Zander model takes into consideration vessel compliance and modulates it with continuous pressure changes in the system, unlike the Windkessel models.14 The Liljestrand-Zander model seems very effective in tracking cardiac output changes induced by vasopressors and could be useful for monitoring cardiac output during haemodynamic instability in the OR during procedures when the Doppler signal could be unusable because of interference from electric scalpels. Moreover, PPCO could be recalibrated anytime with DCO value, and the combination of these two devices could facilitate
Downloaded from http://bja.oxfordjournals.org/ at University of California, San Francisco on March 11, 2015
X4
100
–50
D by DTO (%)
D by DTO (%)
D by X1 (%)
–50
0
–100
D by X1 (%)
–100
D by X1 (%)
100
–100
D by X1 (%)
X3
100
D by X1 (%)
D by X1 (%)
X1 100
BJA
Continuous cardiac output measure calibrated with oesophageal Doppler
continuous cardiac output monitoring in the operating theatre and increase accuracy of cardiac output monitoring.5 For the response to fluid challenge, a significant discrepancy was observed between results obtained with DCO and the Liljestrand-Zander model. Indeed, 14 of the 33 positive fluid challenges classified with the DCO were negative according to the Liljestrand-Zander model (sensitivity of 30%). Therefore, for the titration of fluid in the operating room, the Liljestrand-Zander model cannot replace DCO for classifying fluid responders and nonresponders. Furthermore, we observed that the Windkessel models and area under the curve models had small bias after fluid challenge, which confirmed the possible utility of those algorithms in commercial monitors for fluid titration.
First, we used DCO as our reference technique, which does not represent the usual gold standard method; of the effect aortic diameter changes is still a matter of debate.15 16 Nevertheless, DCO is a less invasive beat-to-beat monitor allowing detection of rapid and transient changes in cardiac output and is frequently used in studies dealing with cardiac output variations induced by vasoactive drugs, despite these limitations.6 17 Second, our comparison did not include the third-generation of commercial PPCO algorithms. Those algorithms could have better results for vasopressor administration than the noncommercial algorithms that we tested. Third, a small percentage of our subjects were responders to fluid challenge. Therefore, further studies are needed to confirm our results on different populations of patients. Finally we used automated data collection and only two items could be collected directly by the monitor just before the event: ‘fluid infusion’ and ‘vasopressor bolus’. Thus, we could not compare the differential effects of phenylephrine, ephedrine and norepinephrine on cardiac output.
Conclusions Our study showed that the Liljestrand-Zander PPCO model, after initial calibration with DCO, demonstrated the least bias and best limits of agreement, especially after vasopressor use. The combination of these two devices could then facilitate continuous cardiac output monitoring in the operating theatre. Further studies are needed to confirm our results and to better assess the usefulness of the Liljestrand-Zander model of PPCO in clinical practice.
Supplementary material Supplementary material is available at British Journal of Anaesthesia online.
Acknowledgements The work should be attributed to the Department of Anesthesiology, Hoˆpitaux Universitaires St-Louis-Lariboisie`re, and University Paris Diderot, Paris, France, presented at the 2013 annual meeting of the ‘Socie´te´ Franc¸aise d’Anesthe´sie – Re´animation’.
A.C. participated in the collection of the data and drafted the manuscript. E.G. performed the statistical analysis and drafted the manuscript. A.T. participated in the collection of the data. G.D. drafted the manuscript. E.M.D. participated in the collection of the data. C.M. participated in editing the manuscript. F.B. participated in the collection of data in ICU. D.B. performed the surgical treatment. S.F. performed the surgical treatment. A.M. participated in the study design and coordination and helped to draft the manuscript. F.V. participated in the design of the study, the collection of the data and drafted the manuscript. All authors read and approved the final manuscript.
Declaration of interest None declared.
References 1 Hamilton MA, Cecconi M, Rhodes A. A systematic review and meta-analysis on the use of preemptive hemodynamic intervention to improve postoperative outcomes in moderate and high-risk surgical patients. Anesth Analg 2011; 112: 1392– 402 2 National Institute for Health and Clinical Excellence. Medical technologies guidance MTG3: CardioQODM oesophageal doppler monitor, 2011. Available from: http://www.nice.org.uk/MTG3. (accessed 15 September 2013) zotero://attachment/179/ 3 Mowatt G, Houston G, Herna´ndez R, et al. Systematic review of the clinical effectiveness and cost-effectiveness of oesophageal Doppler monitoring in critically ill and high-risk surgical patients. Health Technol Assess Winch Engl 2009; 13: iii –iv, ix – xii, 1– 95 4 Nelson LD. The new pulmonary arterial catheters. Right ventricular ejection fraction and continuous cardiac output. Crit Care Clin 1996; 12: 795–818 5 Monnet X, Anguel N, Naudin B, Jabot J, Richard C, Teboul J-L. Arterial pressure-based cardiac output in septic patients: different accuracy of pulse contour and uncalibrated pressure waveform devices. Crit Care Lond Engl 2010; 14: R109 6 Meng L, Tran NP, Alexander BS, et al. The impact of phenylephrine, ephedrine, and increased preload on third-generation VigileoFloTrac and esophageal doppler cardiac output measurements. Anesth Analg 2011; 113: 751–7 7 Sun JX. Cardiac output estimation using arterial blood pressure waveforms. Masters Thesis. Cambridge: Massachusetts Institute of Technology, 2006 8 Bland JM, Altman DG. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 1986; 1: 307– 10 9 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 10 Cannesson M, Jian Z, Chen G, Vu TQ, Hatib F. Effects of phenylephrine on cardiac output and venous return depend on the position of the heart on the Frank-Starling relationship. J Appl Physiol Bethesda Md 1985 2012; 113: 281– 9 11 Maas JJ, Pinsky MR, de Wilde RB, de Jonge E, Jansen JR. Cardiac output response to norepinephrine in postoperative cardiac surgery patients: interpretation with venous return and cardiac function curves. Crit Care Med 2013; 41: 143–50 12 Sun JX, Reisner AT, Saeed M, Heldt T, Mark RG. The cardiac output from blood pressure algorithms trial. Crit Care Med 2009; 37: 72–80
Page 7 of 8
Downloaded from http://bja.oxfordjournals.org/ at University of California, San Francisco on March 11, 2015
Limitations of the study
Authors’ contributions
BJA 13 Monge Garcı´a MI, Gracia Romero M, Gil Cano A, Rhodes A, Grounds RM, Cecconi M. Impact of arterial load on the agreement between pulse pressure analysis and esophageal Doppler. Crit Care Lond Engl 2013; 17: R113 14 Liljestrand G, Zander E. Vergleichende Bestimmungen des Minutenvolumens des Herzens beim Menschen mittels der Stickoxydulmethode und durch Blutdruckmessung, 1928. Available from: http://www.ncbi .nlm.nih.gov/pubmed/19200414 (accessed 3 February 2014) 15 Singer M, Clarke J, Bennett ED. Continuous hemodynamic monitoring by esophageal Doppler. Crit Care Med 1989; 17: 447– 52
Caillard et al.
16 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: 477–82 17 Monnet X, Robert J-M, Jozwiak M, Richard C, Teboul J-L. Assessment of changes in left ventricular systolic function with oesophageal Doppler. Br J Anaesth 2013; 111: 743–9
Handling editor: H. C. Hemmings
Downloaded from http://bja.oxfordjournals.org/ at University of California, San Francisco on March 11, 2015
Page 8 of 8
Comparison of cardiac output measured by oesophageal Doppler ultrasonography or pulse pressure contour wave analysis A. Caillard 1,2,3, E. Gayat1,2,3, A. Tantot1,2,3, G. Dubreuil 1,2, E. M’Bakulu1,2, C. Madadaki 1,2, F. Bart 1,2, D. Bresson 2,4, S. Froelich 2,4, A. Mebazaa1,2,3 and F. Valle´e1,2* 1
Department of Anesthesiology and Critical Care, Hoˆpitaux Universitaires St-Louis-Lariboisie`re-Fernand Widal, Paris, France University Paris Diderot, Paris, France 3 INSERM UMR-942, Paris, France 4 Department of Neurosurgery, Hoˆpitaux Universitaires St-Louis-Lariboisie`re, Paris, France 2
Editor’s key points † Pulse pressure contour wave analysis of cardiac output (PPCO) is a promising monitoring technique. † PPCO analysed using 9 novel algorithms was compared with oesophageal Doppler monitoring of cardiac output (DCO) in 62 neurosurgical subjects. † There were significant differences in bias and concordance between PPCO and DCO depending on algorithm and therapeutic interventions. † Continuous PPCO monitoring with a specific algorithm could offer advantages compared with DCO alone.
Background. Maintaining adequate organ perfusion during high-risk surgery requires continuous monitoring of cardiac output to optimise haemodynamics. Oesophageal Doppler Cardiac Output monitoring (DCO) is commonly used in this context, but has some limitations. Recently, the cardiac output estimated by pulse pressure analysis- (PPCO) was developed. This study evaluated the agreement of cardiac output variations estimated with 9 non-commercial algorithms of PPCO compared with those obtained with DCO. Methods. High-risk patients undergoing neurosurgery were monitored with invasive blood pressure and DCO. For each patient, 9 PPCO algorithms and DCO were recorded before and at the peak effect for every haemodynamic challenge. Results. Sixty-two subjects were enrolled; 284 events were recorded, including 134 volume expansions and 150 vasopressor boluses. Among the 9 algorithms tested, the LiljestrandZander model led to the smallest bias (0.03 litre min21 [21.31, +1.38] (0.21 litre min21 [21.13; 1.54] after volume expansion and 20.13 litre min21 [21.41, 1.15] after vasopressor use). The corresponding percentage of the concordance was 91% (86% after volume expansion and 94% after vasopressor use). The other algorithms, especially those using the Winkessel concept and the area under the pressure wave, were profoundly affected by the vasopressor. Conclusions. Among the 9 PPCO algorithms examined, the Liljestrand-Zander model demonstrated the least bias and best limits of agreement, especially after vasopressor use. Using this particular algorithm in association with DCO calibration could represent a valuable option for continuous cardiac output monitoring of high risk patients. Clinical Trial Registration. Comite´ d’e´thique de la Socie´te´ de Re´animation de Langue Franc¸aise No. 11 –356. Keywords: cardiac output; Doppler; pulse wave analyses Accepted for publication: 17 December 2014
In high risk surgical patients, haemodynamic monitoring is recommended to maintain suitable organ perfusion.1 Cardiac output is a key cardiovascular parameter and a major determinant of tissue oxygenation.2 Continuous measurement is preferable to intermittent measurement, especially in the operating theatre when rapid changes in cardiovascular function are often observed. Alternative minimally invasive methods for continuously monitoring cardiac output have then been developed. Two continuous monitors are more frequently used: 1) the oesophageal Doppler method (DCO), and 2) the pulse pressure contour method (PPCO). DCO is accurate in the
operating theatre;3 however, this technique has limitations, including its inability to monitor a continuous signal during surgical haemostasis as a result of electrical interference and lack of access to the mouth in oral surgery and neurosurgery. PPCO is minimally invasive and not operator dependent. A method of calculating stroke volume from the contour of the arterial pressure curve was first described in 1899.4 PPCO is based on the hypothesis that the waveform of blood pressure is directly related to the variation in the stroke volume. Many algorithms have been proposed: some make a comparison with a closed hydraulic or electric circuit, such as the
& The Author 2015. Published by Oxford University Press on behalf of the British Journal of Anaesthesia. All rights reserved. For Permissions, please email: [email protected]
Downloaded from http://bja.oxfordjournals.org/ at University of California, San Francisco on March 11, 2015
* Corresponding author. E-mail: [email protected]
BJA
Materials and methods This prospective, non-interventional, non-randomized study was conducted in neurosurgical patients under general anaesthesia, between November 2011 and July 2012, in the Department of Anaesthesiology at Lariboisie`re University Hospital (Paris, France). This study was approved by the institutional ethics committee (Comite´ d’e´thique de la Socie´te´ de Re´animation de Langue Franc¸aise No. 11–356). All subjects gave written informed consent. Haemodynamic monitoring, including invasive blood pressure and DCO, represented the usual care of patients undergoing neurosurgery with reduced cerebral compliance and/or potential bleeding in our institution. The exclusion criteria were age ,18 yrs, pregnancy, contraindication to the use of DCO and chronic cardiac arrhythmia.
Study protocol All subjects were orally premedicated with hydroxyzine (1 mg kg21) 1 h before surgery. Anaesthesia was induced with propofol and remifentanil target controlled infusion. Tracheal intubation was facilitated with a atracurium 0.5 mg kg21, and mechanical ventilation used the volume-controlled mode, with tidal volumes of 7 ml kg21 with positive end-expiratory pressure of 5 cm H2O and a respiratory rate of 12– 16 breaths min21 to maintain end tidal CO2 of approximately 4.7 Kpa. After induction of anaesthesia, a radial arterial line and DCO probe were inserted. Both were connected to CombiQw (Deltex Medical Ltd.). DCO values were averaged over 5 cardiac cycles.
Haemodynamic management Mean arterial pressure (MAP), measured the day before the surgical intervention, was considered to be the MAP of reference (MAPref). According to our standard of care, a decrease in MAP.20% of MAPref led to a therapeutic intervention according to the choice of the physician in charge: including a fluid
Page 2 of 8
challenge (250 ml of saline in 10 min) or a bolus of vasopressor. Three different vasopressors were available according to the choice of the physician (bolus of 50 mcg of phenylephrine, 5 mcg of norepinephrine or 9 mg of ephedrine).
Data collection After zeroing the system to atmospheric pressure, the arterial pressure waveform was carefully checked using a fast flush test to ensure optimal harmonics of the arterial pressure measurement system. The arterial pressure signal was transferred from the bedside monitor to the DCO system using a serial cable. Nine nonproprietary PPCO algorithms built into the CardioQ-Combi software provide continuous cardiac output values. The 9 PPCO algorithms are described in Table 1 and in the Supplementary material. Haemodynamic data were automatically recorded by the monitor every 5 s. Initial calibration of the PPCO was performed using the DCO value before any intervention and after stabilizing heart rate and arterial pressure (,5% variation over a 1-min period). Each therapeutic event was automatically recorded on the monitor using two predefined categories: fluid challenge or vasopressor bolus. The peak effect of the therapeutic action was recorded 5 min after the fluid challenge or at the peak MAP after administration of vasopressor.
Statistical analysis The results are expressed as mean (SD) or as median (interquartile range [IQR]), depending on normal distribution of the
Table 1 Characteristics study subjects Characteristics
All subjects (n 562) (%)
Gender (Male/Female)
26 (42)/36 (58)
Age (yrs)
54 [39 –65]
BMI (kg m22)
25 [21 –29]
American Society of Anesthesiologists Physical Status 1
11 (18)
2
36 (73)
3
6 (10)
Comorbidities History of hypertension
21 (34)
Chronic heart failure
3 (5)
Coronary artery disease
1 (2)
Chronic obstructive pulmonary disease
21 (34)
Diabetes mellitus
5(8)
Type of surgery Cerebral tumour
52 (84)
Aneurism
4 (6)
Spine surgery
4 (6)
Other Duration of surgery (min)
2 (3) 335 [248 –440]
Perioperative fluid administration (ml)
3512 [3750 –4500]
Perioperative bleeding (ml)
221 [100 –363]
Perioperative diuresis (ml)
957 [500 –1350]
Downloaded from http://bja.oxfordjournals.org/ at University of California, San Francisco on March 11, 2015
Windkessel concept, and others analyse the area under the systolic portion of the arterial pressure waveform. Algorithm validity has been verified in a variety of patients and circumstances, but performance could be compromised in the presence of haemodynamic instability, cardiac arrhythmias, or other factors that disturb the arterial pressure waveform, particularly during vasopressor use.5 6 To increase the reliability of PPCO, some commercial systems have calibrated their algorithms using thermodilution.5 A prototype for continuously monitoring cardiac output was recently introduced (CombiQw, Deltex Medical, Chichester, Sussex, UK) that combines oesophageal Doppler and PPCO and includes 9 algorithms to estimate cardiac output after calibration using the DCO value according to a study by Sun and colleagues.7 The present study compared, in patients under general anaesthesia, the agreement of cardiac output measurements obtained using 9 non-commercial algorithms of PPCO with measures of cardiac output obtained using DCO alone used as a reference. We also analysed variations in cardiac output during haemodynamic challenges, including volume expansion or administration of a vasopressor bolus.
Caillard et al.
BJA
Continuous cardiac output measure calibrated with oesophageal Doppler
X1
X2
3
1 0 –1 –2
2 1 0 –1 –2 –3
–3 2
3
4
5
6
7
8
3
4
6
7
0 –1 –2
2
6
7
1 0 –1 –2
3
4
5
6
7
–2 –3 6
7
DCO-PPCO/2 (l m–1)
–1 –2
8
3
5
4
6
7
8
DCO-PPCO/2 (l m–1) 3
X8
2 1 0 –1 –2 –3
5
0
2
DCO-PPCO (l m–1)
DCO-PPCO (l m–1)
–1
8
1
8
3
0
7
2
DCO-PPCO/2 (l m–1)
1
6
–3 2
X7
5
X6
2
8
3
4
4
3
DCO-PPCO/2 (l m–1)
3
3
DCO-PPCO/2 (l m–1)
–3
2
–2
8
DCO-PPCO (l m–1)
DCO-PPCO (l m–1)
DCO-PPCO (l m–1)
1
2
–1
X9
2 1 0 –1 –2 –3
2
3
4
5
6
7
DCO-PPCO/2 (l m–1)
8
2
3
4
5
6
7
8
DCO-PPCO/2 (l m–1)
Fig 1 Bland-Altman plots for absolute values of PPCO compared with DCO. Agreement between PPCO and DCO measurements according to Bland-Altman analysis. Only one marker for the subject was represented in the graph. Marker size is relative to the number of observations per subject.
variables. Bland-Altman analysis for repeated measurements was used to assess agreement between DCO and each PPCO algorithm.8 Bias was defined as the mean difference between DCO and PPCO measurements. Limits of agreement (LOA) were calculated as mean bias (1.96) SD. Percentage error (PE) was calculated to compare CO as PE (%)¼LOA/[(mean DCO+ mean PPCO)/2]. A PE≤30% indicates agreement of the technique with the reference method.9 The concordance of the variation of cardiac output measured by DCO and each PPCO algorithm was assessed by Bland-Altman analysis.8 Percentage of the concordance was estimated using the percentage of data in which the direction of change was in agreement in 4-quandrant plots. Acceptable concordance was .90% in
4-quadrant plots. Analyses were repeated in the fluid challenge and vasopressor bolus subgroups. Ability of the PPCO algorithms to detect responders to the fluid challenge was also assessed and compared with DCO. Subjects were defined as a «responder» to volume expansion when fluid challenge led to an increase in stroke volume .10%. P value ,0.05 was considered statistically significant. All statistical analyses were performed using R statistical software (The ‘R’ Foundation for Statistical Computing, Vienna, Austria).
Results A total of 62 subjects were included; subject characteristics are summarized in Table 1.
Page 3 of 8
Downloaded from http://bja.oxfordjournals.org/ at University of California, San Francisco on March 11, 2015
2
5
0
X5
–3
DCO-PPCO (l m–1)
5
3 X4
4
1
DCO-PPCO/2 (l m–1)
3
3
2
–3 2
DCO-PPCO/2 (l m–1)
2
X3
3 DCO-PPCO (l m–1)
2
DCO-PPCO (l m–1)
DCO-PPCO (l m–1)
3
BJA
Caillard et al.
Table 2 Agreement and concordance analysis between DCO and nine PPCO algorithms. PE, percentage of error; LOA, limit of agreement; As, area under the systolic arterial pressure curve; DBP, diastolic blood arterial pressure; HR, heart rate; k, calibration factor; MAP, mean arterial pressure; SBP, p systolic blood pressure; T, duration of cardiac cycle (T¼ HR/60); Tdia, duration of diastole (Tdia¼T – Tsys); Tsys, duration of systole (estimated as 30% of cardiac cycle time) Formulae
Bias [lower LOA; upper LOA]
Percentage of error (PE) %
X1
Mean arterial pressure
K1×MAP
20.28 [22.37; 1.81]
43
X2
Windkessel
K2×(SAP-DAP)×HR
20.20 [22.08; 1.68]
39
X3
Windkessel with RC decay
K3×(MAP×HR/60)×ln(SAP/DAP)
20.21 [22.10; 1.67]
39
X4
Liljestrand-Zander
0.05 [21.18; 1.28]
26
20.24 [22.19; 1.71]
40
20.27 [22.32; 1.77]
42
X5
Pressure root-mean-square
K4×[(SAP-DAP)×HR]/(SAP+DAP) p K5× (SBP-MAP)2×HR
X6
Herd
K6×(MAP-DBP)
Systolic area
K7×HR×As
20.26 [22.20; 1.68]
40
Systolic area with correction
K8×HR×As×(1+Ts/Td)
20.24 [22.13; 1.65]
39
X9
Systolic area with corrected impedance
K9×HR×(163+HR - 0.48×MAP)×As
20.18 [21.88; 1.52]
35
Concordance between DCO and each PPCO algorithm A total of 16 474 measurement pairs of cardiac output were analyzed with a median of 215 [105 –363] pairs per subject. Of the 9 algorithms considered, the X4 (Liljestrand-Zander model) was associated with the most limited bias at 0.05 litre min21, narrowest limits of agreement (upper and lower LOA at 21.18 litre min21 and 1.28 litre min21, respectively), and lowest PE (26%) (Table 2 and Fig. 1). Algorithms based on the Windkessel model (X2 and X3) or on the area under the systolic portion model (X7, X8 and X9) had an average bias of 20.2 litre min21 with PE of 39%.
Table 3 Agreement between the changes in cardiac output measurements using DCO and each PPCO algorithm. The table compiles the values of the bias, the upper and lower LOA and percentage of concordance with 10% exclusion zone of each algorithm for all therapeutic interventions and separation in fluid or vasopressor administration. LOA, limit of agreement Bias [lower LOA; upper LOA]
Percentage of concordance (with 10% exclusion zone)
X1 Mean arterial Pressure Vasopressor
20.96 [23.65; 1.73]
19
0.01 [21.77; 1.79]
42
Vasopressor
20.82 [23.09; 1.45]
47
Fluid
20.01 [21.68; 1.67]
74
Fluid X2 Windkessel
Concordance of changes in cardiac output after therapeutic intervention A total of 284 events were recorded, including 134 fluid challenges and 150 boluses of vasopressor, consisting of 32 bolus administrations of ephedrine, 61 of pheylephrine and 57 of norepinephrine (Table 3, Figs 2 and 3). Variations of DCO and PPCO induced by fluid challenge or vasopressor bolus are related in Table 1 of Supplementary material. Vasopressor induced a decrease in DCO of 58% with a mean value of 20.11 litre min21 [20.54, +0.02] (Supplementary Table 1). With respect to variation of cardiac output measurements after therapeutic intervention, X4 (Liljestrand-Zander Model) again was the most efficient with a bias of 0.03 litre min21 [21.31, +1.38] (0.21 litre min21 [21.13, +1.54] after volume expansion and 21.31 litre min21 [21.41, +1.15] after vasopressor bolus), with PA of 91% (86% after volume expansion and 94% after vasopressor). All other algorithms showed a low agreement after vasopressor. For a Windkessel model with a correction (X3), PA was 76% after volume expansion and 47% after vasopressor. For the area under the curve model with impedance correction (X9), PA was 64% after volume expansion and 43% after vasopressor.
Page 4 of 8
X3 Windkessel with RC decay Vasopressor
20.85 [23.16; 1.45]
47
Fluid
20.02 [21.7; 1.66]
76
20.13 [21.41; 1.15]
94
0.21 [21.13; 1.54]
86
X4 Liljestrand-Zander Vasopressor Fluid
X5 Pressure root-mean-square Vasopressor
20.9 [23.31; 1.5]
43
20.01 [21.68; 1.67]
79
Vasopressor
20.98 [23.55; 1.59]
37
Fluid
20.08 [21.85; 1.68]
75
20.87 [23.26; 1.52]
26
0.02 [21.6; 1.63]
68
Fluid X6 Herd
X7 Systolic area Vasopressor Fluid
X8 Systolic area with correction Vasopressor
20.82 [23.21; 1.56]
29
Fluid
20.01 [21.71; 1.69]
58
X9 Systolic area with correction Vasopressor Fluid
20.63 [22.62; 1.36]
43
0.07 [21.46; 1.59]
64
Downloaded from http://bja.oxfordjournals.org/ at University of California, San Francisco on March 11, 2015
X7 X8
BJA
Continuous cardiac output measure calibrated with oesophageal Doppler
X2
2 0 –2 –4 –2
–1
–0
1
2 0 –2 –4
2
–2 –4 1
0 –2 –4
–2 –4 1
–0
1
2
Average D (I m–1)
0 –2 –4 –1
–0
1
Average D (I m–1)
2
0 –2 –4 –1
–0
1
2
Average D (I m–1)
2
–2
1
X6
–2
X8
4
–0
2
2
Difference D (I m–1)
Difference D (I m–1)
Difference D (I m–1)
0
–0
–1
–2
–1
4
Average D (I m–1)
2
–1
–4
Average D (I m–1)
2
2
X7
–2
–2
–2
X5
4
Average D (I m–1) 4
0
2
Difference D (I m–1)
Difference D (I m–1)
Difference D (I m–1)
0
–0
1
2
2
X9
4 2 0 –2 –4 –2
–1
–0
1
2
Average D (I m–1)
Fig 2 Bland-Altman plots for DCO compared with PPCO during fluid challenge (black points) or bolus of vasopressors (blue points). The figure shows agreement between changes in cardiac output measurements by DCO and each PPCO algorithm according to the Bland-Altman representation during therapeutic challenge.
Differentiation between the responders and non-responders to fluid challenge using the Liljestrand-Zander Model 25% (33/134) of fluid challenges were responsive according to DCO with an increase in stroke volume .10%. When using X4 (Liljestrand-Zander Model) for classification of responders to fluid challenge, only 13% (17/134) were positive. In comparison with DCO, the sensibility for determining fluid challenge response by the Liljestrand–Zander model was 30% and specificity was 93%.
Discussion We analysed the agreement between DCO and 9 noncommercial PPCO algorithms during fluid challenge and vasopressors
administration. The model of Liljestrand-Zander showed the best performance, namely the lowest error rate and greater concordance with DCO value for static value and after therapeutic intervention, especially after use of vasopressors. The other 8 algorithms showed correct bias after fluid challenge compared with the DCO, but there was poor agreement after vasopressor administration. Although the PPCO seems to be a good alternative to monitoring the cardiac output during surgery, many of the commercialized algorithms had difficulties, especially during vasopressor administration. Meng and colleagues showed that an algorithm derived from the Windkessel concept accurately tracks changes in cardiac output after whole-body tilting to increase preload, whereas phenylephrine-induced change in vasomotor tone profoundly affected its ability to track changes in cardiac output. In this study, vasopressor
Page 5 of 8
Downloaded from http://bja.oxfordjournals.org/ at University of California, San Francisco on March 11, 2015
2
–1
–0
4
Average D (I m–1)
X4
–2
–1
–2
Average D (I m–1) 4
X3
4
Difference D (I m–1)
Difference D (I m–1)
Difference D (I m–1)
X1 4
BJA
Caillard et al.
X2
50
50
50
0 –50
0 –50 –100
0
50
–100
100
D by X1 (%)
D by X1 (%)
0 –50
50
100
–100
–50
0
50
0 –50
X7
100
D by X1 (%)
0 –50 –100
0
50
0
50
0 –50
–100
100
100
X8
0 –50
D by DTO (%)
0
50
100
X9
100
50
–100
–50
D by DTO (%)
–100 –50
100
–100 –50
100
50
50
50
D by DTO (%)
D by DTO (%)
0
X6
100
50
–100
100
–50
D by DTO (%)
–100
–100
–100
0
X5
100
50
–100
–50
50 0 –50 –100
–50
0
50
100
D by DTO (%)
–100
–50
0
50
100
D by DTO (%)
Fig 3 Concordance between changes in DCO and PPCO represented in four quadrants plots.
administration could induce opposite changes in cardiac ouput measuring by DCO or PPCO. The software enhancement does not seem to improve the ability of PPCO to adequately measure cardiac output when arterial compliance is modified. Our study confirmed the poor performance of the Windkessel model after a vasopressor bolus, which is similar to findings of previously published studies in the intensive care unit or operating theatre. The concordance between cardiac ouput measured with DCO and algorithms based on pressure wave form such as the Windkessel models (X4.5), Herd model (X6) and area under the curve (X7, X8, and X9), were poor after vasopressor administration. Furthermore, the relation between cardiac output variations and vasoconstrictor administration is complex and, as recently reported; depends on the volume status of the patient.10 11 Our results are in accordance with those studies: we observed a decrease
Page 6 of 8
of DCO induced by vasopressor which was mostly expected as our neurosurgical subjects were fluid non-responders in 75% of cases.10 11 Interestingly, as recently published,12 13 the LiljestrandZander model retained a good percentage of concordance with DCO (94%) during vasopressor use. The Liljestrand-Zander model takes into consideration vessel compliance and modulates it with continuous pressure changes in the system, unlike the Windkessel models.14 The Liljestrand-Zander model seems very effective in tracking cardiac output changes induced by vasopressors and could be useful for monitoring cardiac output during haemodynamic instability in the OR during procedures when the Doppler signal could be unusable because of interference from electric scalpels. Moreover, PPCO could be recalibrated anytime with DCO value, and the combination of these two devices could facilitate
Downloaded from http://bja.oxfordjournals.org/ at University of California, San Francisco on March 11, 2015
X4
100
–50
D by DTO (%)
D by DTO (%)
D by X1 (%)
–50
0
–100
D by X1 (%)
–100
D by X1 (%)
100
–100
D by X1 (%)
X3
100
D by X1 (%)
D by X1 (%)
X1 100
BJA
Continuous cardiac output measure calibrated with oesophageal Doppler
continuous cardiac output monitoring in the operating theatre and increase accuracy of cardiac output monitoring.5 For the response to fluid challenge, a significant discrepancy was observed between results obtained with DCO and the Liljestrand-Zander model. Indeed, 14 of the 33 positive fluid challenges classified with the DCO were negative according to the Liljestrand-Zander model (sensitivity of 30%). Therefore, for the titration of fluid in the operating room, the Liljestrand-Zander model cannot replace DCO for classifying fluid responders and nonresponders. Furthermore, we observed that the Windkessel models and area under the curve models had small bias after fluid challenge, which confirmed the possible utility of those algorithms in commercial monitors for fluid titration.
First, we used DCO as our reference technique, which does not represent the usual gold standard method; of the effect aortic diameter changes is still a matter of debate.15 16 Nevertheless, DCO is a less invasive beat-to-beat monitor allowing detection of rapid and transient changes in cardiac output and is frequently used in studies dealing with cardiac output variations induced by vasoactive drugs, despite these limitations.6 17 Second, our comparison did not include the third-generation of commercial PPCO algorithms. Those algorithms could have better results for vasopressor administration than the noncommercial algorithms that we tested. Third, a small percentage of our subjects were responders to fluid challenge. Therefore, further studies are needed to confirm our results on different populations of patients. Finally we used automated data collection and only two items could be collected directly by the monitor just before the event: ‘fluid infusion’ and ‘vasopressor bolus’. Thus, we could not compare the differential effects of phenylephrine, ephedrine and norepinephrine on cardiac output.
Conclusions Our study showed that the Liljestrand-Zander PPCO model, after initial calibration with DCO, demonstrated the least bias and best limits of agreement, especially after vasopressor use. The combination of these two devices could then facilitate continuous cardiac output monitoring in the operating theatre. Further studies are needed to confirm our results and to better assess the usefulness of the Liljestrand-Zander model of PPCO in clinical practice.
Supplementary material Supplementary material is available at British Journal of Anaesthesia online.
Acknowledgements The work should be attributed to the Department of Anesthesiology, Hoˆpitaux Universitaires St-Louis-Lariboisie`re, and University Paris Diderot, Paris, France, presented at the 2013 annual meeting of the ‘Socie´te´ Franc¸aise d’Anesthe´sie – Re´animation’.
A.C. participated in the collection of the data and drafted the manuscript. E.G. performed the statistical analysis and drafted the manuscript. A.T. participated in the collection of the data. G.D. drafted the manuscript. E.M.D. participated in the collection of the data. C.M. participated in editing the manuscript. F.B. participated in the collection of data in ICU. D.B. performed the surgical treatment. S.F. performed the surgical treatment. A.M. participated in the study design and coordination and helped to draft the manuscript. F.V. participated in the design of the study, the collection of the data and drafted the manuscript. All authors read and approved the final manuscript.
Declaration of interest None declared.
References 1 Hamilton MA, Cecconi M, Rhodes A. A systematic review and meta-analysis on the use of preemptive hemodynamic intervention to improve postoperative outcomes in moderate and high-risk surgical patients. Anesth Analg 2011; 112: 1392– 402 2 National Institute for Health and Clinical Excellence. Medical technologies guidance MTG3: CardioQODM oesophageal doppler monitor, 2011. Available from: http://www.nice.org.uk/MTG3. (accessed 15 September 2013) zotero://attachment/179/ 3 Mowatt G, Houston G, Herna´ndez R, et al. Systematic review of the clinical effectiveness and cost-effectiveness of oesophageal Doppler monitoring in critically ill and high-risk surgical patients. Health Technol Assess Winch Engl 2009; 13: iii –iv, ix – xii, 1– 95 4 Nelson LD. The new pulmonary arterial catheters. Right ventricular ejection fraction and continuous cardiac output. Crit Care Clin 1996; 12: 795–818 5 Monnet X, Anguel N, Naudin B, Jabot J, Richard C, Teboul J-L. Arterial pressure-based cardiac output in septic patients: different accuracy of pulse contour and uncalibrated pressure waveform devices. Crit Care Lond Engl 2010; 14: R109 6 Meng L, Tran NP, Alexander BS, et al. The impact of phenylephrine, ephedrine, and increased preload on third-generation VigileoFloTrac and esophageal doppler cardiac output measurements. Anesth Analg 2011; 113: 751–7 7 Sun JX. Cardiac output estimation using arterial blood pressure waveforms. Masters Thesis. Cambridge: Massachusetts Institute of Technology, 2006 8 Bland JM, Altman DG. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 1986; 1: 307– 10 9 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 10 Cannesson M, Jian Z, Chen G, Vu TQ, Hatib F. Effects of phenylephrine on cardiac output and venous return depend on the position of the heart on the Frank-Starling relationship. J Appl Physiol Bethesda Md 1985 2012; 113: 281– 9 11 Maas JJ, Pinsky MR, de Wilde RB, de Jonge E, Jansen JR. Cardiac output response to norepinephrine in postoperative cardiac surgery patients: interpretation with venous return and cardiac function curves. Crit Care Med 2013; 41: 143–50 12 Sun JX, Reisner AT, Saeed M, Heldt T, Mark RG. The cardiac output from blood pressure algorithms trial. Crit Care Med 2009; 37: 72–80
Page 7 of 8
Downloaded from http://bja.oxfordjournals.org/ at University of California, San Francisco on March 11, 2015
Limitations of the study
Authors’ contributions
BJA 13 Monge Garcı´a MI, Gracia Romero M, Gil Cano A, Rhodes A, Grounds RM, Cecconi M. Impact of arterial load on the agreement between pulse pressure analysis and esophageal Doppler. Crit Care Lond Engl 2013; 17: R113 14 Liljestrand G, Zander E. Vergleichende Bestimmungen des Minutenvolumens des Herzens beim Menschen mittels der Stickoxydulmethode und durch Blutdruckmessung, 1928. Available from: http://www.ncbi .nlm.nih.gov/pubmed/19200414 (accessed 3 February 2014) 15 Singer M, Clarke J, Bennett ED. Continuous hemodynamic monitoring by esophageal Doppler. Crit Care Med 1989; 17: 447– 52
Caillard et al.
16 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: 477–82 17 Monnet X, Robert J-M, Jozwiak M, Richard C, Teboul J-L. Assessment of changes in left ventricular systolic function with oesophageal Doppler. Br J Anaesth 2013; 111: 743–9
Handling editor: H. C. Hemmings
Downloaded from http://bja.oxfordjournals.org/ at University of California, San Francisco on March 11, 2015
Page 8 of 8
