Intensive Care Med (2004) 30:17341739 DOI 10.

1007/s00134-004-2361-y

ORIGINAL

Antoine Vieillard-Baron Karim Chergui Anne Rabiller Olivier Peyrouset Bernard Page Alain Beauchet Franois Jardin

Superior vena caval collapsibility as a gauge of volume status in ventilated septic patients

Received: 14 October 2003 Accepted: 1 June 2004 Published online: 26 June 2004  Springer-Verlag 2004 An editorial regarding this article can be found at http://dx.doi.org/10.1007/s00134004-2362-x Electronic Supplementary Material Supplementary material is available for this article if you access the article at http:// dx.doi.org/10.1007/s00134-004-2361-y. A link in the frame on the left on that page takes you directly to the supplementary material. A. Vieillard-Baron K. Chergui A. Rabiller O. Peyrouset B. Page F. Jardin ()) Medical Intensive Care Unit, University Hospital Ambroise Par, Assistance Publique Hpitaux de Paris, 9 avenue Charles de Gaulle, 92104 Boulogne Cedex, France e-mail: francois.jardin@apr.ap-hop-paris.fr Tel.: +33-1-49095604 Fax: +1-49095892 A. Beauchet Department of Biostatistics, University Hospital Ambroise Par, Assistance Publique Hpitaux de Paris, 9 avenue Charles de Gaulle, 92104 Boulogne Cedex, France

Abstract Objective: In mechanically ventilated patients inspiratory increase in pleural pressure during lung inflation may produce complete or partial collapse of the superior vena cava. Occurrence of this collapse suggests that at this time external pressure exerted by the thoracic cavity on the superior vena cava is greater than the venous pressure required to maintain the vessel fully open. We tested the hypothesis that measurement of superior vena caval collapsibility would reveal the need for volume expansion in a given septic patient. Design and setting: Prospective data collection for 66 successive patients in septic shock admitted in a medical intensive care unit and mechanically ventilated for an associated acute lung injury. Measurements and results: We simultaneously measured superior vena caval collapsibility by echocardiography and cardiac index by the Doppler technique at baseline and after a 10 ml/kg volume expansion by 6% hydroxyethyl starch in 30 min. The threshold superior vena caval collapsibility of 36%, calculated as

(maximum diameter on expirationminimum diameter on inspiration)/maximum diameter on expiration, allowed discrimination between responders (defined by an increase in cardiac index of at least 11% induced by volume expansion) and nonresponders, with a sensitivity of 90% and a specificity of 100%. Conclusions: Superior vena cava measurement should be systematically performed during routine echocardiography in septic shock as it gives an accurate index of fluid responsiveness. Keywords Septic shock Superior vena caval collapsibility Fluid responsiveness

Introduction
Cardiovascular dysfunction in sepsis is complex, usually associated with an inadequate cardiac preload [1], depressed left ventricular (LV) systolic function [2, 3], and abnormal arterial dilatation leading to hyperdynamic

shock syndrome [4]. In this setting first-line therapy is blood volume expansion (VE), and a large amount of fluid may be required during the first days of management [5]. However, fluid responsiveness is not constant in septic patients [6], and excessive volume loading may seriously worsen the commonly associated respiratory

1735

failure [7]. A predictive index of fluid responsiveness has long been sought by clinicians dealing with this crucial therapeutic problem [8]. Baseline values of central venous pressure, pulmonary capillary wedge pressure, and echocardiographic LV diastolic dimensions have successively been found to be inaccurate [8]. To date only respiratory changes in arterial pulse have been convincingly demonstrated as accurately predicting fluid responsiveness in septic patients under mechanical ventilation, provided they do not exhibit spontaneous breathing [9]. In our unit transesophageal echocardiography is a routine procedure used to assess hemodynamic status in mechanically ventilated patients. This procedure permits evaluation of left and right ventricular function. We have also recently observed in several patients a partial or complete collapse of the superior vena cava (SVC) during tidal ventilation [10]. Occurrence of this collapse suggests that at this time external pressure exerted by the thoracic cavity on SVC is greater than the venous pressure required to maintain the vessel fully open, and this was illustrated in our previous report in which this collapse was corrected by VE [10]. The present prospective open study was performed to evaluate whether the amplitude of SVC dimensional changes during mechanical ventilation predicts the hemodynamic effect of blood VE in a given patient and to compare this index with the previously validated changes in arterial pulse pressure. Preliminary results concerning the first 48 patients of this study have been previously summarized in a clinical commentary [11].

urinary origin, and 1 septic arthritis). Thirty-five patients received vasopressor support (norepinephrine alone in 23, combined with dobutamine in 12, because of a LV ejection fraction by transesophageal echocardiography lower than 40%), and the 31 remaining patients had severe hypotension (systolic blood pressure=7911 mmHg). All patients studied were free of previous cardiopulmonary disease. In 48 patients there was acute lung injury and in 18 acute respiratory distress syndrome. The mean Average Simplified Acute Physiology Score II in the overall series was 6124. Forty patients survived. Pressure measurements Heart rate and systemic arterial pressure from an indwelling radial artery catheter were recorded throughout the study, together with airway pressure obtained from a side port of the tracheal tube. Central venous pressure was measured at end-expiration through an internal jugular catheter previously inserted for fluid administration, with the midaxillary line as a zero reference level. All patients were continuously monitored by a nail-bed oxymeter. From arterial pressure recording we measured arterial pulse as systolic pressure minus diastolic pressure of the preceding beat and calculated the expiratory decrease in arterial pulse (dP), as proposed by Michard et al. [9]. Echo Doppler measurements Echo-Doppler studies were performed with a Corevision model SSA-350A (Toshiba France, Puteaux, France) equipped with a multiplane 5-MHz transesophageal echocardiography transducer. Using the signal from the respirator, airway pressure was displayed on the screen of the echo Doppler device permitting accurate timing of cardiac events during the respiratory cycle. In particular, it allowed us to differentiate inspiration from expiration. The SVC was examined from a long-axis view, using the twodimensional view to direct the M-mode beam across the maximum diameter. From this view we measured SVC diameter during the respiratory cycle. The collapsibility index of SVC, or SVC collapsibility, that is, the inspiratory decrease in SVC diameter, was determined as (maximum diameter on expirationminimum diameter on inspiration)/maximum diameter on expiration, and was expressed as a percentage [10]. LV stroke volume was measured by the Doppler technique. Doppler aortic flow by pulsed wave was obtained at the level of aortic annulus by a transgastric, long-axis approach. By tracing the envelope of aortic flow the velocity-time integral (AoVTI) was automatically processed and was averaged during the whole respiratory cycle. A long-axis view of LV outflow tract was used to measure aortic diameter at the level of the aortic annulus. Stroke volume was obtained by multiplying AoVTI by aortic area, which was calculated from the systolic aortic diameter. Cardiac output was calculated as stroke volume times heart rate. All measurements were indexed to body surface area, and expressed as stroke index and cardiac index (CI). Study protocol Echocardiographic and pressure measurements were first performed at baseline. Then a rapid blood VE was performed by infusion of 10 ml/kg 6% hydroxyethylstarch in 30 min. Measurements were then repeated.

Materials and methods

Patients We studied 66 mechanically ventilated patients diagnosed with acute circulatory failure related to sepsis (48 men, 18 women; mean age 6115 years, range 1984). Inclusion criteria were as follows: (a) sepsis defined by the criteria of the American College of Chest Physicians/Society of Critical Care Medicine Consensus Conference [12], (b) acute circulatory failure defined by a systolic blood pressure lower than 90 mmHg or the need for norepinephrine infusion, (c) the need for mechanical ventilation for associated acute lung injury or acute respiratory distress syndrome as defined by the North American/European Consensus Conference [13], and (d) a perfect adaptation to the respirator without any spontaneous breathing efforts as visualized by continuous monitoring of airway pressure and flow on the respirator screen. All patients were studied during the first day of hospitalization in our intensive care unit. Mechanical ventilation was used in the controlled mode, with a limited plateau pressure (<30 cmH2O), an average tidal volume of 82 ml/kg, and a respiratory rate of 15 breaths/min. External positive end-expiratory pressure (PEEP) was between 5 and 7 cmH2O, and no patient exhibited intrinsic PEEP with this external PEEP. During the period of the study six patients meeting the inclusion criteria were excluded because of an inadequate Doppler aortic signal by a transgastric approach in four cases and an irregular cardiac rhythm in two. The 66 patients studied had clear evidence of sepsis (39 bacterial pneumonia; 20 abdominal sepsis; 5 meningitis; 1 sepsis of

1736

Statistical analysis The effects of VE on hemodynamic and echocardiographic parameters were assessed using a nonparametric Wilcoxon rank sum test. Patients were divided into two groups according to the percentage increase in CI in response to VE. From a previous report of Doppler measurements of cardiac output [14] we assumed that an 11% change in CI was needed for clinical significance. Therefore patients with a VE-induced increase in CI greater than 11% and lower than 11% were classified as responders and nonresponders, respectively. Between groups comparisons were performed using a nonparametric Mann-Whitney U test. Results are expressed as mean values SD. Receiver operating characteristic (ROC) curves were generated for SVC collapsibility, dP, and maximal (expiratory) SVC diameter, varying the discriminating threshold of each parameter. The area under the ROC curve (SEM) was calculated for each parameter and compared. Least square linear regression analysis was also used when appropriate.

Table 3 Main hemodynamic data at baseline Responders 1 (n=20) Heart rate (beats/min) Systolic arterial pressure (mmHg) Central venous pressure (mmHg) Cardiac index (l min1 m2) Stroke index (ml m2) dP (%) Stroke index diametera (mm) SVC collapsibility (%)
a

Nonresponders (n=46) 9919* 11128* 125 2.71 278* 96* 175* 1411*

11021 9321 94 2.40.7 227 2712 154 6417

Maximum diameter measured at expiration * p<0.05 responders vs. nonresponders

Results
Intra observer and interobserver variability in SVC diameter measurements is presented in Table 1. The same information concerning cardiac output measurements by the Doppler technique have been previously published [11]. In all patients the maximum value of SVC diameter was observed at expiration and the minimum value during inspiration. Conversely, maximum pulse and systolic pressure were exhibited during inspiration and minimum pulse and systolic pressure during expiration. Hemodynamic parameters before and after VE are presented in Table 2.

Fig. 1 This plot of proportional superior venal caval (SVC) collapsibility and relative frequency (number of patients exhibiting this collapsibility index) in the overall group illustrates its bimodal distribution

Table 1 Intraobserver and interobserver variability (percentages) in SVC diameter measurement established in our laboratory, from M-mode echocardiography guided by two-dimensional imaging. Values were obtained in a group of ten patients
Intraobserver (observer A) SVC maximum diameter SVC minimum diameter 23 56 Intraobserver (observer B) 33 44 Interobserver 77 77

Table 2 Main hemodynamic data at baseline and after volume expansion in all 66 patients Baseline Heart rate (beats/min) Systolic arterial pressure (mmHg) Central venous pressure (mmHg) Cardiac index (l min1 m 2) Stroke index (ml m2) dP (%) SVC diametera (mm) SVC collapsibility index (%) 10320 10627 105 2.60.9 268 1412 175 3026 VE 10119 11829* 124* 2.91 298* 64* 184 148*

* p<0.05 vs. baseline a Maximum diameter measured at expiration

VE induced a significant increase in CI (from 2.40.7 before VE to 3.31 l min1 m2 after VE) in 20 responder patients (group 1). VE did not change significantly CI (2.71 before VE to 2.71 l min1 m2 after VE) in 46 nonresponder patients (group 2). Comparisons between baseline hemodymic values in both groups are presented in Table 3. Before VE SVC collapsibility ranged from 0% to 100%, was weakly correlated with SVC maximum diameter (r=0.3, p=0.2), and was not correlated with central venous pressure. Baseline SVC collapsibility differed significantly between group 1 and group 2 patients (64 17%, vs. 1411%, respectively). However, distribution of SVC collapsibility was essentially bimodal: among the 17 of the 20 group 1 patients (85%) had SVC collapsibility greater than 50%, whereas and 41 of the 46 group 2 patients (89%) SVC collapsibility less than 30%. Thus only 8 patients (12% of the overall group) had an intermediate SVC collapsibility between 30% and 50% (Fig. 1). The high collapsibility index (>50%) observed in the majority of group 1 patients suggests partial or complete collapse of SVC during inspiration in these patients. After VE the mean SVC collapsibility was significantly reduced in group 1 patients (177%) but remained unchanged in

1737

Fig. 2 Two examples of SVC examination by combining twodimensional imaging and Mmode study. A A complete collapse of SVC was observed at baseline, relieved by volume expansion (VE). At the same time cardiac index (CI) significantly increased. B SVC diameter was not influenced by lung inflation at baseline, and remained unchanged after VE. In this patient cardiac index was not modified by VE. TP Tracheal pressure

Fig. 3 ROC curves comparing the ability of SVC collapsibility and dP to discriminate responders (CI increase !11%) and nonresponders to VE. The area under the ROC curve for SVC collapsibility (0.9930.013) did not differ significantly from the area under the ROC curve for dP (0.9400.038; P=0.19) Threshold values for these parameters (arrows) are 36% and 12%, respectively

group 2 patients (139%). Examples of patients in each group are shown in Fig. 2. Before VE dP ranged from 0 to 52% and was significantly different between group 1 and group 2 patients (2712%, vs. 96%, respectively). After VE dP was significantly reduced in group 1 patients (65%) and unchanged in group 2 patients (65%). The areas under the ROC curves for SVC collapsibility and dP did not differ significantly (Fig. 3). The threshold SVC collapsibility of 36% allowed discrimination between responders and nonresponders, with a sensitivity of 90% and a specificity of 100%. Individual changes in CI produced by VE are presented in Fig. 4. The threshold dP of 12% allowed discrimination between responders and nonresponders with a sensitivity of 90% and a specificity of 87%. Finally, discordant predictions concerning fluid responsiveness using SVC collapsibility and dP were ob-

served in ten cases, an erroneous prediction by dP being obtained in eight cases (12%) and by SVC collapsibility in two (3%). Three examples of discordant cases are shown in Figs. S1, S2, and S3 (Electronic Supplementary Material). Use of both evaluations simultaneously permitted a 100% specificity and a 100% sensitivity in predicting fluid responsiveness. SVC maximum diameter (154 mm vs. 175 mm, p<0.02) was smaller in responders than in nonresponders. A threshold value of 16 mm permitted discrimination of responders, with a sensitivity of 90% and a specificity of 56%. The area under the ROC curve (0.700.066) was significantly lower than the area for SVC collapsibility (p<0.001). Central venous pressure (94 mmHg vs. 125 mmHg) did not differ significantly between responders and nonresponders.

1738

Fig. 4 Individual changes in cardiac index (CI) produced by volume expansion (VE). Patients were divided into two groups, according to the prediction of fluid responsiveness by SVC collapsibility index, a value >36% predicting a positive response (left) and a value 36% predicting a negative response (right)

Discussion
By permitting an accurate assessment of both left [15] and right ventricular function [16] bedside transesophageal echocardiography has gained broad acceptance as a noninvasive procedure for hemodynamic evaluation in hemodynamically unstable patients [17]. The present study indicates that this evaluation should be systematically completed by SVC examination. In this group of 66 mechanically ventilated patients with septic shock the SVC collapsibility index was found to be an excellent indicator of fluid responsiveness. The presence of SVC collapsibility greater than 36% before VE indicated fluid responsiveness, as defined by an increase in cardiac output of at least 11% after VE, with a sensitivity of 90% and a specificity of 100%. In the present study the quite bimodal distribution of SVC collapsibility suggests that this vessel obeys the all-or-nothing law. The vessel stays open when external pressure remains lower than opening pressure and suddenly collapses when external pressure becomes higher than opening pressure. Seventeen patients had a collapse of SVC (complete or partial) produced by lung inflation, with more than 50% decrease in inspiratory diameter, and all these patients, constituting 85% of the responder group, significantly increased their cardiac index after VE. On the other hand, in nonresponders the SVC was only minimally or not at all influenced by lung inflation. Recently an expiratory drop in arterial pulse pressure greater than 13% was also evidenced as a powerful index of fluid responsiveness in mechanically ventilated septic patients [9]. Our present data also confirm this report, with a very similar threshold of 12%. However, six falsepositives and two false-negatives were observed with dP variations. Obviously the fact that a patient who did not exhibit marked respiratory changes in arterial pulse pressure could show increased cardiac output after VE (false-negative finding) is not surprising: VE in a patient without cardiac failure is thought to increase cardiac output, and there was also one false-negative case in the group of 40 patients described by Michard et al. [9]. More

interesting was the finding that among the six false-positive cases when using dP three had a marked inspiratory increase in arterial pulse, and three others exhibited severe acute cor pulmonale, as previously defined by our group [16]. The latter finding suggests that marked inspiratory increase or severe right ventricular dysfunction would limit the use of pulse pressure variations in predicting fluid responsiveness in a mechanically ventilated septic patient. Whereas a large inspiratory increase in pulse is easy to detect, as illustrated in Fig. S3, echocardiographic examination is required to detect severe right ventricular dysfunction. We have previously illustrated this limitation [11], which might be relevant in a context in which right ventricular dysfunction is often observed [6, 18, 19, 20]. Both experimental [21] and clinical [22, 23] studies have demonstrated that the expiratory drop in arterial pulse is produced by a relative decrease in LV stroke volume during lung deflation, when compared with a relative increase during lung inflation. Because the LV is filled by the blood present in the pulmonary capillary bed, the amount of this filling reserve plays a key role in modulating cyclic changes in arterial pulse during respiratory support [23]. During lung inflation the pulmonary capillary bed is suddenly emptied by the increase in transpulmonary pressure, which boosts blood toward the left cavities [23, 24], and this transient improvement in LV preload also transiently improves LV ejection [22, 23]. Additionally, the sudden rise in trans-pulmonary pressure increases right ventricle outflow impedance [23, 25], precluding any immediate refilling of the pulmonary capillary bed. Thus a patient with a diminished capillary blood volume at baseline may exhibit a marked preload defect in the LV at the onset of expiration, producing a major drop in LV stroke output and in arterial pulse pressure. This expiratory drop can be corrected by VE, provided this volume reaches the pulmonary capillary bed. This is obviously not possible when severe right ventricular dysfunction is present, as illustrated by three of our false-positive patients when using dP.

1739

In conclusion, our study describes a new index of fluid responsiveness obtainable during routine bedside echocardiography in mechanically ventilated septic patients.

As illustrated by the study, SVC collapsibility may be used in clinical practice as an echocardiographic gauge of volume status.

References
1. Jardin F, Fourme T, Page B, Loubires Y, Vieillard-Baron A, Beauchet A, Bourdarias JP (1999) Persistent preload defect in severe sepsis despite fluid loading: A longitudinal echocardiographic study in patients with septic shock. Chest 116:13541359 2. Parker M Shelhamer J, Bacharach S, Gren M, Natanson C, Frederick T, Damske B, Parrillo J (1984) Profound but reversible myocardial depression in patients with septic shock. Ann Intern Med 100:483490 3. Jardin F, Brun-Ney D, Auvert B, Beauchet A, Bourdarias JP (1990) Sepsis-related cardiogenic shock. Crit Care Med 18:10551060 4. Parrillo J (1993) Pathogenic mechanism of septic shock. N Engl J Med 328:14711477 5. Task Force of the American College of Critical Care Medicine, Society of Critical Care Medicine (1999) Practice parameters for hemodynamic support of sepsis in adult patients in sepsis. Crit Care Med 27:639660 6. Schneider A, Teule G, Groeneveld A, Nauta J, Heidendal G, Thijs L (1988) Biventricular performance during volume loading in patients with early septic shock, with emphasis on the right ventricle: a combined hemodynamic and radionuclide study. Am Heart J 116:103112 7. Krausz M, Perel A, Eimerl D, Cotev S (1977) Cardiopulmonary effects of volume loading in patients in septic shock. Ann Surg 185:429434 8. Bendjelid K, Romand J (2003) Fluid responsiveness in mechanically ventilated patients: a review of indices used in intensive care. Intensive Care Med 29:352360 9. Michard F, Boussat S, Chemla D Anguel N, Mercat A, Lecarpentier Y, Richard C, Pinsky M, Teboul JL (2000) Relation between respiratory changes in arterial pulse pressure and fluid responsiveness in septic patients with acute circulatory failure. Am J Respir Crit Care Med 162:134138 10. Vieillard-Baron A, Augarde R, Prin S, Page B, Beauchet A, Jardin F (2001) Influence of superior vena caval zone conditions on cyclic changes in right ventricular outflow during respiratory support. Anesthesiology 95:10831088 11. Vieillard-Baron A, Prin S, Chergui K, Dubourg O, Jardin F (2003) Hemodynamic instability in sepsis: bedside assessment by Doppler echocardiography. Am J Respir Crit Care Med 168:1270 1276 12. American College of Chest Physicians/ Society of Critical Care Medicine Consensus Conference (1992) Definition for sepsis and organ failure and guidelines for the use of innovative therapies in sepsis. Crit Care Med 20:864874 13. American-European Consensus Conference on ARDS (1994) Definition, mechanisms, relevant outcomes, and clinical trial coordination. Am J Respir Crit Care Med 149:818824 14. Darmon PL, Hillel Z, Mogtader A, Mindich B, Thys D (1994) Cardiac output by transesophageal echocardiography using continuous-wave Doppler across the aortic valve. Anesthesiology 80:796805 15. Vieillard-Baron A, Schmitt JM, Beauchet A, Augarde R, Prin S, Page B, Jardin F (2001) Early preload adaptation in septic shock? A transesophageal echocardiographic study. Anesthesiology 94:400406 16. Vieillard-Baron A, Prin S, Chergui K, Dubourg O, Jardin F (2002) EchoDoppler demonstration of acute cor pulmonale at the bedside in the medical intensive care unit. Am J Respir Crit Care Med 166:13101319 17. Fellahi JL, Jardin F (1999) Doppler echocardiography in intensive care. In: Webb A, Shapiro M, Singer M, Suter P (eds) Oxford textbook of critical care. Oxford University Press, Oxford, pp 11451153 18. Kimchi A, Ellrodt G, Berman D, Riedinger M, Swan H, Murata G (1984) Right ventricular performance in septic shock: a combined radionuclide and hemodynamic study. J Am Coll Cardiol 5:945951 19. Vincent JL, Reuse N, Frank N, Contempr B, Kahn R (1989) Right ventricular dysfunction in septic shock: assessment by measurements of right ventricular ejection fraction using thermodilution technique. Acta Anaesthesiol Scand 33:3438 20. Parker M, McCarthy K, Ognibene F, Parrillo J (1990) Right ventricular dysfunction and dilatation, similar to left ventricular changes, characterize the cardiac depression of septic shock in humans. Chest 97:126131 21. Morgan B, Martin W, Hornbein T, Crawford E, Guntheroth W (1966) Hemodynamic effect of intermittent positive pressure respiration. Anesthesiology 27:584590 22. Jardin F, Farcot JC, Gueret P, Prost JF, Ozier Y, Bourdarias JP (1983) Cyclic changes in arterial pulse during respiratory support. Circulation 68:266274 23. Vieillard-Baron A, Chergui K, Augarde R, Prin S, Page B, Beauchet A, Jardin F (2003) Cyclic changes in arterial pulse during respiratory support revisited by Doppler Echocardiography. Am J Respir Crit Care Med 168:671676 24. Versprille A (1990) The pulmonary circulation during mechanical ventilation. Acta Anaesthesiol Scand 34 [Suppl 94]:5162 25. Vieillard-Baron A, Loubires Y, Schmitt JM, Page B, Dubourg O, Jardin F (1999) Cyclic changes in right ventricular output impedance during mechanical ventilation. J Appl Physiol 87:16441650