Secondary Logo

Journal Logo

Technology, Computing, and Simulation: Research Report

The Ability of a Novel Algorithm for Automatic Estimation of the Respiratory Variations in Arterial Pulse Pressure to Monitor Fluid Responsiveness in the Operating Room

Cannesson, Maxime, MD*; Slieker, Juliette, MD*; Desebbe, Olivier, MD*; Bauer, Christian, MD*; Chiari, Pascal, MD, PhD*; Hénaine, Roland, MD; Lehot, Jean-Jacques, MD*

Author Information
doi: 10.1213/01.ane.0000297291.01615.5c
  • Free
  • Take the CME Test
  • Chinese Language Editions

Goal-directed intraoperative fluid administration has been shown to be able to reduce length of hospital stay,1–4 critical care admissions,5 and mortality6 after major surgery in various settings. In most of these studies, the hemodynamic end points were cardiac output (CO), stroke volume (SV), central venous pressure (CVP), and/or respiratory variations in left ventricular SV assessed using esophageal Doppler.

Static indicators such as CVP, pulmonary capillary wedge pressure (PCWP), or left ventricular end-diastolic area have been shown to be poor predictors of fluid responsiveness.7–14 Dynamic indicators have consistently been demonstrated to be better predictors of fluid responsiveness in patients during mechanical ventilation. During positive pressure ventilation, the inspiratory right ventricular SV decrease is proportional to the degree of hypovolemia and is transmitted to the left heart after two or three beats (pulmonary transit time).12,13 One of the main limitations of these indices is that, to the best of our knowledge, only pulse contour analysis (PCCO) allows for continuous monitoring of these respiratory variations. However, PCCO requires invasive and specific catheters.

Recently, a method based on automatic detection algorithms, kernel smoothing, and rank-order filters has been proposed for automated and continuous variations in arterial pulse pressure (ΔPP) calculation (ΔPPauto).15 However, this method has not been validated in the clinical setting.

The aim of our study was to test the relationship and agreement between manually calculated ΔPP (ΔPPman) and ΔPPauto, and to test the ability of ΔPPauto to predict fluid responsiveness in mechanically ventilated patients in the operating room.


The protocol was approved by the IRB for human subjects of our institution (Comité Consultatif de Protection des Personnes dans la Recherche Biomédicale Lyon B). All patients gave informed, written consent. We studied 25 consecutive patients undergoing coronary artery bypass grafting. Patients with cardiac arrhythmias and intracardiac shunt were excluded.

This group consisted of 15 males and10 females between 39- and 82-yr-old (mean age 65 ± 12 yr). Twenty patients received β-blockers preoperatively. Induction of anesthesia was performed with propofol (3–5 mg/kg) and sufentanil (0.5–1.0 μg/kg), and orotracheal intubation was facilitated with pancuronium (0.1–0.15 mg/kg). After induction of anesthesia, an 8-cm 5F catheter (Arrow International Inc., Reading, PA) was inserted in the left or right radial artery, a triple-lumen 16-cm 8.5F central venous catheter (Arrow International Inc., Reading, PA) and a pulmonary artery catheter (Swan-Ganz catheter, 7.5F; Baxter Edwards, Lifescience, LLC, Irvine, CA) were inserted via the right internal jugular vein. Pressure transducers (Medex Medical Ltd., Rossendale, Lancashire, UK) were leveled at the midaxillary line and fixed to the operation table to keep the transducer at the atrial level according to the study protocol. All transducers were zeroed to atmospheric pressure. Correct position of the pulmonary artery catheter in West’s zone 3 was assessed using the method of Teboul et al.16 CO was measured by thermodilution, using the average of five successive measurements obtained by injection of 10 mL of dextrose at room temperature randomly during the respiratory cycle. Cardiac index (CI) and SV index were calculated by dividing CO and SV, respectively, by the body surface area. Anesthesia was maintained with continuous infusions of propofol (5–8 mg • kg−1 • h−1) and sufentanil (0.7–1.0 μg • kg−1 • h−1) to keep bispectral index (Aspect 1000, Aspect Medical Systems Inc., Natick, MA) between 40 and 50. All patients’ lungs were ventilated in a volume-controlled mode with a tidal volume of 8–10 mL/kg of body weight at a frequency of 12–15 cycles/min. Positive end-expiratory pressure was set between 0 and 2 cm H2O according to the attending physician.

Data Recording and Analysis

Arterial pressure waveforms were recorded from a bedside monitor (Intellivue MP70, Philips Medical Systems, Suresnes, France) to a personal computer using data acquisition software (TrendfaceSolo 1.1, Ixellence GmbH, Wildau, Germany) and were analyzed by an observer who had no knowledge of ΔPPauto or any other hemodynamic data. All hemodynamic data were recorded after 3 min of hemodynamic stability.

Respiratory Variations in Pulse Pressure Analysis

PP was defined as the difference between systolic and diastolic pressure. Maximal (PPmax) and minimal (PPmin) values were determined over the same respiratory cycle. ΔPPman was then calculated as described by its authors7: ΔPPman = (PPmax − PPmin)/[(PPmax + PPmin)/2]. The measurements were repeated on three consecutive respiratory cycles and averaged for statistical analysis.

Automated Calculation of Respiratory ΔPP

The algorithm used in this study is commercially available and has been previously described by Aboy et al.15 PP variation (PPV) is displayed in real-time by Philips Intelivue MP70 monitors (Intellivue MP70, Philips Medical Systems, Suresnes, France) as a percentage reflecting ΔPP. Briefly, it allows for ΔPP determination from the arterial pressure waveform alone with no need for the airway pressure acquisition. This method, based on automatic detection algorithms, kernel smoothing, and rank-order filters, involves seven consecutive steps (beat minima detection, beat maxima detection, beat mean calculation, pulse amplitude pressure, envelope estimation, pulse amplitude pressure estimation, and PPV estimation).15 ΔPPauto was calculated and averaged over four cycles of eight seconds.

Other Hemodynamic Measurements

The following variables were recorded both before and after intravascular volume expansion: systolic arterial blood pressure, mean arterial blood pressure, diastolic arterial blood pressure, heart rate, end-expiratory CVP, end-expiratory PCWP, stroke volume index, CI, and systemic vascular resistance index.

Experimental Protocol

All patients were studied immediately after induction of anesthesia and after a 5-min period of hemodynamic stability with no changes in anesthetic protocol and no intravascular volume expansion. Baseline hemodynamic measurements were obtained and then followed by an IV intravascular volume expansion consisting of 500 mL of hetastarch 6% given over 10 min. Hemodynamic measurements were performed within 3 min after intravascular volume expansion. Thereafter, ΔPPman and ΔPPauto were determined at the following six different times throughout the surgery: immediately after skin incision, immediately after sternotomy, immediately after pericardectomy, immediately after weaning from cardiopulmonary bypass, immediately after chest closure, and immediately before leaving the operating room. ΔPPauto was determined in real-time during the surgery and ΔPPman was determined post hoc based on recorded waveforms.

Statistical Analysis

All data are presented as mean ± sd. Changes in hemodynamic variables induced by intravascular volume expansion were assessed using a nonparametric Mann-Whitney U-test or Wilcoxon’s ranked sum test when appropriate. Patients were divided into two groups according to the percent increase in CI after intravascular volume expansion. Responders were defined as patients demonstrating an increase in CI ≥15% after intravascular volume expansion7 and nonresponders as patients whose CI changed <15%. Receiver operating characteristic (ROC) curves were generated for CI, CVP, PCWP, ΔPPman, and ΔPPauto, varying the discriminating threshold of each parameters and area under the ROC curves, were calculated and compared17 (MedCalc, MedCalc Software, Mariakerke, Belgium). Considering previously published results,7 power analysis showed that 25 patients were necessary to detect differences of 0.15 between ΔPPman and ΔPPauto areas under the ROC curves (5% type I error rate, 80% power, two tailed test). Bland-Altman analysis was performed to assess agreement between ΔPPman and ΔPPauto.18 A P value <0.05 was considered as statistically significant. All statistic analyses were performed using SPSS 13.0 for Windows, SPSS, Chicago, IL.


Agreement (mean bias ± sd) between ΔPPman and ΔPPauto (Bland-Altman analysis) was 0.7% ± 3.4% (Fig. 1). In the first part of the study (before and after intravascular volume expansion), agreement between ΔPPman and ΔPPauto was −1.0% ± 4.4% over the 50 pairs of data. In the second part of the study, agreement was 1.3% ± 2.8% over the 150 pairs of data.

Figure 1.
Figure 1.:
Relationship and between ΔPPman and ΔPPauto, and Bland Altman analysis for the agreement between ΔPPman and ΔPPauto. ΔPPman: manually calculated respiratory variations in arterial pulse pressure, ΔPPauto: automated and continuous respiratory variations in arterial pulse pressure calculation.

Changes in Hemodynamic Variables After Intravascular Volume Expansion

Hemodynamic data at baseline and after intravascular volume expansion are shown in Table 1. As expected, on average, intravascular volume expansion induced a significant increase in CI (from 1.8 ± 0.6 to 2.2 ± 0.7 L • min−1 • m−2; P < 0.001), mean arterial pressure (from 61 ± 9 to 77 ± 11 mm Hg; P < 0.001), CVP (from 8 ± 6 to 16 ± 4 mm Hg; P < 0.001), and PCWP (from 13 ± 5 to 17 ± 5 mm Hg; P < 0.001). At the same time we observed significant decreases in both ΔPPman (from 14% ± 8% to 8% ± 5%; P < 0.01) and ΔPPauto (from 12% ± 6% to 8% ± 3%; P < 0.01). We observed no statistically significant difference between percent change in ΔPPman and ΔPPauto after intravascular volume expansion (13% ± 139% vs 19% ± 63% respectively; P = 0.85).

Table 1
Table 1:
Hemodynamic Data at Baseline and After Intravascular Volume Expansion

ΔPPman to Predict Fluid Responsiveness

Seventeen (68%) patients were responders and eight patients were nonresponders to intravascular volume expansion. Their hemodynamic data are shown in Table 2. ΔPPman and ΔPPauto were significantly higher in responders than in nonresponders (18% ± 7% vs 6% ± 3% and 15% ± 6% vs 6% ± 2%, respectively; P < 0.01 for both), and CVP was significantly lower in responders than in nonresponders (5 ± 5 vs 10 ± 6 mm Hg; P < 0.01). Neither difference in PCWP (13 ± 3 mm Hg in responders vs 11 ± 7 mm Hg in nonresponders; P = 0.43) nor difference in CI (1.81 ± 0.59 mL • min−1 • m−2 in responders vs 1.67 ± 0.53 mL • min−1 • m−2 in nonresponders; P = 0.60) reached statistical significance between these two groups. The areas under the ROC curves (±se) were as follows: 0.923 ± 0.060 for ΔPPman, 0.919 ± 0.058 for ΔPPauto, 0.750 ± 0.106 for CVP, 0.651 ± 0.135 for PCWP, and 0.563 ± 0.121 for CI (Fig. 2). The areas for ΔPPauto and ΔPPman were significantly higher than the areas for CVP, PCWP, and CI (P < 0.05 for both). The difference in area under the curve between ΔPPauto and ΔPPman did not reach significance (P = 0.86). Setting a threshold ΔPPman value of 12% discriminated between responders and nonresponders with a sensitivity of 88% and a specificity of 100%. The threshold ΔPPauto value of 10% discriminated between responders and nonresponders with a sensitivity of 82% and a specificity of 88%. Interestingly, we found that the threshold CVP value of 3.5 mm Hg allowed discrimination between responders and nonresponders with a sensitivity of 77% and a specificity of 63%.

Table 2
Table 2:
Hemodynamic Data at Baseline in Responders and Nonresponders to Intravascular Volume Expansion
Figure 2.
Figure 2.:
Receiver operating characteristic curves comparing the ability of ΔPPman, ΔPPauto, CVP, PCWP, and CI at baseline to discriminate between responders and nonresponders to volume expansion. ΔPPman: manually calculated respiratory variations in arterial pulse pressure, ΔPPauto: automated and continuous respiratory variations in arterial pulse pressure calculation, CVP: central venous pressure; PCWP: pulmonary capillary wedge pressure, CI: cardiac index.

ΔPPman to Quantify Response to Intravascular Volume Expansion

There was a statistically significant positive linear correlation between ΔPPauto at baseline and percent changes in CI induced by intravascular volume expansion (δCI) (r = 0.43; P < 0.05) as well as between ΔPPman and δCI induced (r = 0.44; P < 0.05) indicating that the higher ΔPPman and ΔPPauto at baseline, the higher δCI. We observed no statistically significant relationship between CVP at baseline and δCI (r = −0.39; P = 0.06) and between PCWP at baseline and δCI (r = −0.05; P = 0.84).


The results of this study show that ΔPPauto can be displayed continuously and can predict fluid responsiveness in mechanically ventilated patients in the operating room. This index allows for ΔPP monitoring from the arterial pressure waveform alone and has potential for goal-directed intraoperative fluid administration in the operating room.

Fluid responsiveness has been studied extensively in the operating room and in the intensive care unit.7–10,12,13,19–22 Numerous dynamic indices have been described as accurate predictors of fluid responsiveness in these settings: respiratory variations in invasive arterial PP7 or Finapress® arterial PP,22 inferior vena cava diameter,19 aortic pulsed wave Doppler,8 left ventricular stroke area,23 SV derived from PCCO,24 pulse oxymeter plethysmographic waveform amplitude,25 and esophageal Doppler.26,27 It is now well accepted that dynamic indicators, relying on the cardiopulmonary interactions in patients under mechanical ventilation, are better predictors of fluid responsiveness than static indicators such as CVP, PCWP, and left ventricular end-diastolic area.12 However, only SV derived from PCCO allows for a continuous monitoring of these indices. The main limitations of this technology are that it is invasive, not widely available, it requires calibration, insertion of a central venous catheter and specific arterial catheter.

In 2004, Aboy et al. described a novel algorithm to estimate the PPV index (ΔPP).15 This algorithm allows ΔPP determination from the arterial pressure waveform alone with no need for the airway pressure acquisition. This method based on automatic detection algorithms, kernel smoothing, and rank-order filters involves seven consecutive steps (beat minima detection, beat maxima detection, beat mean calculation, pulse amplitude pressure, envelope estimation, pulse amplitude pressure estimation, and PPV estimation).15 In their study, Aboy et al. found that ΔPPauto was not dependent on respiratory phase shift, was robust to beat misdetection, and was a good estimate of the true ΔPP. However, this algorithm had never been tested in the clinical setting for the assessment of fluid responsiveness.

Our results support the ability of this automated ΔPP determination to be able to predict fluid responsiveness in mechanically ventilated patients under general anesthesia. Moreover, we found that dynamic indicators of fluid responsiveness are better predictors than static indicators. Interestingly, we found that CVP was lower in responders than in nonresponders and that it had a weak but statistically significant predictive value. However, CVP area under the curve was significantly lower than those for ΔPPman and ΔPPauto and cannot be recommended for fluid responsiveness prediction. Consequently, these results are consistent with previously published data.

Agreement between both methods was relatively weak in our study. However, ΔPPauto and ΔPPman are not expected to be exactly the same. Indeed, ΔPPman is by definition the arterial PPV over a single mechanical breath,7 whereas ΔPPauto is by definition the arterial PPV over a much longer period.

The respiratory variations in the arterial blood pressure have been shown to be accurate predictors of fluid responsiveness.9 However, the software we used in this study does not allow monitoring of this index. Moreover, several studies have now shown that this index was less reliable than ΔPP for fluid responsiveness prediction.7

To show that ΔPP can be calculated continuously and monitored in the operating room may have clinical impact. Strategies to optimize fluid management during surgery have been developed over the past 10 yr and have been shown to be able to improve length of hospital stay,1–4 critical care admissions,5 and mortality6 after major surgery. However, most of these studies focused on esophageal Doppler-derived indices and/or invasive static indicators (such as CVP). The main limitations of these techniques are that they are either operator dependent and not widely available (esophageal Doppler) or a poor predictor of fluid responsiveness (CVP). Indices relying on the cardiopulmonary interactions in patients under mechanical ventilations have consistently been shown to be the best predictors of fluid responsiveness.9,10,12,13,22,24 ΔPP is still considered the best predictor of fluid responsiveness in this setting. However, it was not possible to conveniently monitor this index in the operating room or in the intensive care unit. Thus, ΔPPauto has potential clinical application for fluid management optimization in the operating room. Further studies are required to determine whether ΔPPauto is reliable for all patients and, more importantly, whether decisions based upon this variable ultimately influence outcome.


As other indices relying on the respiratory variations of left ventricular SV or its surrogate, ΔPPauto cannot be calculated in patients with cardiac arrhythmia. Also, interpretation of ΔPPauto in patients with right and/or left ventricular failure must be done with caution. We chose to measure ΔPP during closed and opened chest conditions. The cardiopulmonary interactions may be impacted by this change in intrathoracic pressure. However, Reuter et al. have shown that dynamic indicators can be used during open chest conditions.28 Consequently, we feel that it is important to report data recorded in this setting.

The threshold value of 10% for prediction of fluid responsiveness has to be interpreted with caution. As for ΔPPman and for any other indices of fluid responsiveness, the threshold value may vary among studies and settings. To quote Solus-Biguenet et al.,22 “ΔPP values ranging from 8% to 13% may constitute an inconclusive or ‘gray zone’29; where its predictive value is uncertain.” Another major point is that ΔPPauto value has to be considered after at least a 1-min period of hemodynamic stability. Because this algorithm relies on a mean PPV estimation, it is very important to observe a steady hemodynamic state before accepting the ΔPPauto value.

Finally, ΔPPauto can only be performed in mechanically ventilated patients although most patients who require careful fluid management during anesthesia will likely be mechanically ventilated.


1. Sinclair S, James S, Singer M. Intraoperative intravascular volume optimisation and length of hospital stay after repair of proximal femoral fracture: randomised controlled trial. BMJ 1997;315:909–12
2. Venn R, Steele A, Richardson P, Poloniecki J, Grounds M, Newman P. Randomized controlled trial to investigate influence of the fluid challenge on duration of hospital stay and perioperative morbidity in patients with hip fractures. Br J Anaesth 2002;88:65–71
3. Wakeling HG, McFall MR, Jenkins CS, Woods WG, Miles WF, Barclay GR, Fleming SC. Intraoperative oesophageal Doppler guided fluid management shortens postoperative hospital stay after major bowel surgery. Br J Anaesth 2005;95:634–42
4. Gan TJ, Soppitt A, Maroof M, el-Moalem H, Robertson KM, Moretti E, Dwane P, Glass PS. Goal-directed intraoperative fluid administration reduces length of hospital stay after major surgery. Anesthesiology 2002;97:820–6
5. Conway DH, Mayall R, Abdul-Latif MS, Gilligan S, Tackaberry C. Randomised controlled trial investigating the influence of intravenous fluid titration using oesophageal Doppler monitoring during bowel surgery. Anaesthesia 2002;57:845–9
6. Poeze M, Greve JW, Ramsay G. Meta-analysis of hemodynamic optimization: relationship to methodological quality. Crit Care 2005;9:R771–9
7. Michard F, Boussat S, Chemla D, Anguel N, Mercat A, Lecarpentier Y, Richard C, Pinsky MR, Teboul JL. Relation between respiratory changes in arterial pulse pressure and fluid responsiveness in septic patients with acute circulatory failure. Am J Respir Crit Care Med 2000;162:134–8
8. Feissel M, Michard F, Mangin I, Ruyer O, Faller JP, Teboul JL. Respiratory changes in aortic blood velocity as an indicator of fluid responsiveness in ventilated patients with septick shock. Chest 2001;119:867–73
9. Tavernier B, Makhotine O, Lebuffe G, Dupont J, Scherpereel P. Systolic pressure variation as a guide to fluid therapy in patients with sepsis-induced hypotension. Anesthesiology 1998;89:1313–21
10. Rex S, Brose S, Metzelder S, Huneke R, Schalte G, Autschbach R, Rossaint R, Buhre W. Prediction of fluid responsiveness in patients during cardiac surgery. Br J Anaesth 2004;93:782–8
11. Wiesenack C, Fiegl C, Keyser A, Prasser C, Keyl C. Assessment of fluid responsiveness in mechanically ventilated cardiac surgical patients. Eur J Anaesthesiol 2005;22:658–65
12. Michard F. Changes in arterial pressure during mechanical ventilation. Anesthesiology 103:419–28, 2005; quiz 49–5
13. Bendjelid K, Romand JA. Fluid responsiveness in mechanically ventilated patients: a review of indices used in intensive care. Intensive Care Med 2003;29:352–60
14. Pinsky MR, Teboul JL. Assessment of indices of preload and volume responsiveness. Curr Opin Crit Care 2005;11:235–9
15. Aboy M, McNames J, Thong T, Phillips CR, Ellenby MS, Goldstein B. A novel algorithm to estimate the pulse pressure variation index deltaPP. IEEE Trans Biomed Engl 2004;51:2198–203
16. Teboul JL, Besbes M, Andrivet P, Axler O, Douguet D, Zelter M, Lemaire F, Brun-Buisson C. A bedside index assessing the reliability of pulmonary occlusion pressure measurements during mechanical ventilation with positive end-expiratory pressure. J Crit Care 1992;7:22–9
17. Hanley JA, McNeil BJ. A method of comparing the areas under receiver operating characteristic curves derived from the same cases. Radiology 1983;148:839–43
18. Bland JM, Altman DG. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 1986;1:307–10
19. Feissel M, Michard F, Faller JP, Teboul JL. The respiratory variation in inferior vena cava diameter as a guide to fluid therapy. Intensive Care Med 2004;30:1834–7
20. Michard F, Teboul JL. Predicting fluid responsiveness in ICU patients. A critical analysis of the evidence. Chest 2002;121:2000–8
21. Cannesson M, Desebbe O, Hachemi M, Jacques D, Bastien O, Lehot JJ. Respiratory variations in pulse oxymeter waveform amplitude are influenced by venous return in mechanically ventilated patients under general anaesthesia. Eur J Anaesthesiol 2007;24:245–51
22. Solus-Biguenet H, Fleyfel M, Tavernier B, Kipnis E, Onimus J, Robin E, Lebuffe G, Decoene C, Pruvot FR, Vallet B. Non-invasive prediction of fluid responsiveness during major hepatic surgery. Br J Anaesth 2006;97:808–16
23. Cannesson M, Slieker J, Desebbe O, Farhat F, Bastien O, Lehot J. Prediction of fluid responsiveness using respiratory variations in left ventricular stroke area by transoesophageal echocardiographic automated border detection in mechanically ventilated patients. Crit Care 2006;10:R171
24. Reuter DA, Felbinger TW, Schmidt C, Kilger E, Goedje O, Lamm P, Goetz AE. Stroke volume variation for assessment of cardiac responsiveness to volume loading in mechanically ventilated patients after cardiac surgery. Intensive Care Med 2002;28:392–8
25. Cannesson M, Attof Y, Rosamel P, Desebbe O, Joseph P, Metton O, Bastien O, Lehot JJ. Respiratory variations in pulse oximetry plethysmographic waveform amplitude to predict fluid responsiveness in the operating room. Anesthesiology 2007;106:1105–11
26. Monnet X, Rienzo M, Osman D, Anguel N, Richard C, Pinsky MR, Teboul JL. Esophageal Doppler monitoring predicts fluid responsiveness in critically ill ventilated patients. Intensive Care Med 2005;31:1195–2001
27. Vallee F, Fourcade O, De Soyres O, Angles O, Sanchez-Verlaan P, Pillard F, Smail N, Olivier G, Genestal M, Samii K. Stroke output variations calculated by esophageal Doppler is a reliable predictor of fluid response. Intensive Care Med 2005;31:1388–93
28. Reuter DA, Goepfert MS, Goresch T, Schmoeckel M, Kilger E, Goetz AE. Assessing fluid responsiveness during open chest conditions. Br J Anaesth 2005;94:318–23
29. Feinstein AR. The inadequacy of binary models for the clinical reality of three-zone diagnostic decisions. J Clin Epidemiol 1990;43:109–13
© 2008 International Anesthesia Research Society