Fluid loading is a common therapeutic step after cardiac surgery, and predictors and monitors of fluid responses of the heart are continuously explored. Transpulmonary thermodilution (PiCCO)-derived blood volumes such as the global end-diastolic blood volume index (GEDVI) may better predict and monitor preload-dependent fluid responses of cardiac output than central venous pressure (CVP) or pulmonary artery occlusion pressure (PAOP) obtained by a pulmonary artery catheter (PAC) [1–13]. Indeed, transmitted airway pressure in mechanically ventilated patients may confound filling pressures as indicators of fluid responses. Volumes may be less dependent on mechanical ventilation than pressures (and dynamic indicators) [7,9,12], but superiority of the former can be doubted. Indeed, observations on the superior predictive and monitoring value of transpulmonary volumes for fluid responses may be confounded by a shared measurement error and mathematical rather than physiologic coupling, when both volumes and cardiac output are derived from the same transpulmonary thermodilution curve . For instance, echocardiographic dimensions may less well predict transpulmonary cardiac output than transpulmonary volumes . Even though cardiac output may vary independently from volumes by infusion of inotropes or ß-blockers suggesting independent variance in spite of mathematical coupling [14,15], the effect of the latter on assessment of fluid responses, when volumes and cardiac output change in similar directions, has been addressed only once, unsupported by data .
Mathematical coupling can be evaluated by using an independent measurement [3,16–18]. For comparison with cardiac output measured by transpulmonary thermodilution, which agrees with that obtained by pulmonary artery thermodilution [1,4,10,19–25], we used the continuous thermodilution cardiac output measurement technique via a PAC. The latter does not require repeated injections and is an acceptable alternative to the intermittent bolus technique, at least, perhaps, in the absence of tricuspid regurgitation and rapid rewarming, and provided that the relative slowness of the system is taken into account [26–34]. Continuous methods are attractive in monitoring fluid responses and the continuous pulmonary and intermittent transpulmonary thermodilution cardiac output responses proved almost interchangeable in that respect, prior to coronary artery surgery, in the single available comparison . However, methods may also require comparison after surgery, when haemodynamic stability may be less [2,23,24,32–34].
The goal of the current study was to compare intermittent transpulmonary (PiCCO) and continuous pulmonary (PAC) thermodilution methods and associated static cardiac filling indicators in predicting and monitoring fluid responses, in mechanically ventilated patients after coronary artery surgery. This was done to test the hypothesis that mathematical coupling with cardiac output accounts, at least in part, for the alleged superiority of GEDVI over pressures in predicting and monitoring fluid responses.
Patients and methods
Eleven consecutive patients undergoing elective coronary artery bypass grafting below 85 years of age were included and studied within 2 h after surgery in the intensive care unit. The Local Ethics Committee approved the study protocol and written informed consent was obtained from all patients before surgery. Inclusion criteria were life expectancy more than 24 h and presumed hypovolaemia and tolerance to intravenous (i.v.) fluids, arbitrarily defined by PAOP and CVP less than 13 mmHg and a continuous cardiac index measured in the pulmonary artery (CCIp) less than 4.0 l min−1 m−2. Exclusion criteria were surgical bleeding greater than 100 ml h−1, clinical evidence for pulmonary oedema, known aneurysms of thoracic or abdominal aorta and presence of atrial fibrillation. All patients underwent cardiopulmonary bypass. There were no visible V waves in the CVP curve, thereby disfavouring haemodynamically important tricuspid regurgitation. Doses of drugs, including vasoactive agents, were given according to local guidelines. Patients were sedated with propofol (1 mg kg−1 h−1) or midazolam (0.1 mg kg−1 h−1) and sufentanil (0.5 μg kg−1 h−1). Patients were on volume-controlled ventilation with tidal volumes of 10 ml kg−1 and a positive end-expiratory pressure (PEEP) of 5 cmH2O. Ventilation frequency was adjusted between 10 and 13 breaths min−1 to maintain the arterial PCO2 in the normal range (4.5–6.0 kPa).
Pulmonary artery catheter
A thermal filament-rapped pulmonary artery thermodilution catheter (8.5-Fr, CCO/VIP, Edwards Life Sciences, Santa Ana, CA, USA) for continuous cardiac output monitoring had been preoperatively inserted via the internal jugular vein. The catheter was connected to a computer (Vigilance, Edwards Life Sciences, Irvine, CA, USA), displaying continuous cardiac output data at 0.5–1 min intervals. After pooling of data, measured at 1–2 min intervals and normalized for individual means (n = 59), the reproducibility of CCIp, expressed as the coefficient of variation around the mean, was 4.2% in our patients. The PAC was also used to determine PAOP and CVP at the end of expiration, after calibration and zeroing to atmospheric pressure, at the mid-chest level, after, for the PAOP, proper wedging, and with patients in the supine position.
The transpulmonary thermodilution technique (PiCCO, Pulsion Medical, Munich, Germany) was used to determine GEDVI, representing the sum of volumes of the cardiac chambers at end diastole [4,6,11,15,35]. The GEDVI is derived from the transpulmonary cardiac index (CItp) multiplied by the difference in mean transit time of the thermal indicator and its down-slope time. We found in the 132 recorded individual of triplicate transpulmonary measurements in this study, which were pooled after normalization for individual mean values, a reproducibility, expressed as the coefficient of variation around the mean, of 6.7% for CItp and of 11.3% for GEDVI, in accordance with the literature [1,4,11,19,21]. Body surface area calculated from weight and height was used to normalize measured variables (SVI equals stroke volume index).
Within 2 h after arrival in the intensive care unit, a 5F thermodilution catheter (Pulsiocath PV 2015; Pulsion Medical, Munich, Germany) was inserted via the femoral artery into the aorta and connected to the PiCCO computer. Demographic data, including the EuroSCORE and body surface area, calculated from height and weight were collected. At baseline (T = 0), heart rate (HR), mean arterial pressure (MAP), PAOP, CVP, GEDVI, CItp and CCIp were measured. To this end, three central venous injections of 20 ml of cold (<8°C) saline were done, regardless of the respiratory cycle, and averaged values were taken. Five fluid-loading steps of 250 ml each (1250 ml in total) were performed at 30 min intervals (from T = 0 to T = 5), in eight patients (n = 40 steps). One patient received 250 ml of fluids (one step), one patient 750 ml (three steps), and one patient 1000 ml (four steps). Fluid loading was discontinued in the three latter patients, in case of sudden changes of MAP, increasing more than 10 mmHg, PAOP, increasing more than 6 mmHg or CCIp, increasing more than 2.1 l min−1 m−2 over values in preceding step. After each 15 min loading step, 15 min were allowed for reaching a new steady state before haemodynamic variables were measured , and the CCIp was taken before CItp measurements. If coagulation was impaired, 250 ml of fresh frozen plasma (n = 11 fluid steps) was given. In all other cases (n = 37 fluid steps), a starch (MW 130) solution (Voluven, Fresenius, Germany) was used. Heart rate was recorded with one of the standard leads of the electrocardiogram. Two patients were sequential atrial–ventricular pacemaker-dependent during the study protocol; the pacemaker frequency was set at 80 beats min−1. Doses of continuously administered vasoactive and sedative drugs and ventilator settings were unchanged during the study protocol.
We expected about 50% rises more than 5% or at least 10% (with no-rises ≤5 and <10%, respectively) in cardiac output among fluid-loading step. The cut-offs were chosen to account for the volumes administered and the literature on fluid responsiveness [9,10,12,13]. Generalized estimating equations, taking repeated measurements in the same patients into account, were used to evaluate changes over time, to study the concordance of fluid responses among cardiac output techniques and to study the value of baseline values to predict and changes in variables to monitor fluid responses by either technique. The type of fluid was entered as a covariate. The CItp and CCIp were then compared by means of Bland and Altman plots of differences versus means, yielding bias, precision (SD) and limits of agreement (bias ± 2SD), and were correlated with help of linear regression . This allowed identifying the patients with the greatest potential errors in CItp and it was reasoned that their exclusion would decrease mathematical coupling to GEDVI and overestimation of correlations, if any. We also calculated bias and limits of agreement with help of a standard deviation corrected for that of differences between techniques within each patient, to account for repeated measurements . To judge the effect on assessing fluid responses by mathematical coupling between GEDVI and CItp versus CCIp, we compared relations between pressure, volume and cardiac output (changes), prior to and after fluid loading, prior to and after exclusion of CItp outliers (versus CCIp), and prior to and after averaging of triplicate measurements. Partial linear correlation coefficients, with patient numbers as a covariate to account for repeated measurements, were calculated for the relations and compared after z transformation. Exact two-sided P values greater than 0.001 are given and considered statistically significant if less than 0.05. Data were summarized as mean ± standard deviation (SD), since they were normally distributed (Kolmogorov–Smirnov test).
Patient characteristics are described in Table 1. Table 2 summarizes results of haemodynamic measurements, prior to and after each fluid-loading step.
In Table 3, fluid steps have been pooled, according to rises and no-rises for both cardiac output techniques, defined by two cut-off values. Fluid responses were similar, when judged from changes in CItp or CCIp, and there was a high concordance (69%) between techniques for responses more than 5% and at least 10% (P = 0.001). The table shows that filling pressures (and MAP) were of poor value in predicting and monitoring fluid responses. In contrast, changes in GEDVI paralleled fluid responses, but only when judged by CItp, regardless of type of fluid.
Comparing CItp to CCIp and relations with pressures and volumes
Figure 1 shows the relationship between CItp and CCIp for pooled data (r = 0.70, P < 0.001; n = 59). Pooled changes in time correlated also (r = 0.44, P = 0.002; n = 48). The bias between techniques was 0.40 l min−1 m−2 and limits of agreement were −0.68 and 1.48 l min−1 m−2, since precision was 0.54 l min−1 m−2. After correction for repeated measurements, bias was 0.33 l min−1 m−2 and limits of agreement (bias ± 2SD) were −1.04 and 1.70 l min−1 m−2, since precision (SD) was 0.70 l min−1 m−2. Hence, there was greater between-patient than within-patient variability.
After excluding two patients with greatest within-patient variability between techniques (Fig. 1), CItp and CCIp correlated at r = 0.74 (P < 0.001; n = 49) and changes in time at r = 0.63 (P < 0.001; n = 40). Then, the bias was 0.38 l min−1 m−2 and limits of agreement (bias ± 2SD) were −0.55 and 1.29 l min−1 m−2, since precision (SD) was 0.46 l min−1 m−2. After correction for repeated measurements, bias was 0.21 l min−1 m−2 and limits of agreement (bias ± 2SD) were −1.14 and 1.56 l min−1 m−2, since precision (SD) was 0.69 l min−1 m−2, indicating almost unaltered between-patient variability.
For pooled data (n = 59), CVP and GEDVI similarly correlated to CItp (r = 0.33, P = 0.012; r = 0.52, P < 0.001, respectively), whereas only CVP correlated to CCIp (r = 0.32, P = 0.013) and GEDVI did not (r = 0.24, P = 0.064). After excluding two patients with outliers of CItp versus CCIp (Fig. 1), the GEDVI still did not relate to CCIp (r = 0.18, P = 0.231; n = 49), whereas the CVP did (r = 0.32, P = 0.028), and CVP and GEDVI both related to CItp (r = 0.37 or higher, P = 0.010 or lower). The PAOP did not relate to any of the variables except for CVP (r = 0.38, P = 0.004). Also, changes in GEDVI correlated to changes in CItp (r = 0.66, P < 0.001; n = 48) but not to changes in CCIp (r = 0.12, P = 0.435; P = 0.002 versus r for CItp) (Fig. 2), whereas changes in CVP did not relate to changes in either CCIp or CItp. In contrast, changes in CVP and GEDVI similarly related to those of CItp (r = 0.38 or higher, P = 0.019 or lower; n = 40), after excluding two patients with CItp outliers (Fig. 1) and the rs did not differ from each other. Changes in PAOP only related to those in CVP (r = 0.51, P < 0.001). Table 4 shows that, for individual of triplicate measurements, changes in GEDVI best related to those in CItp (and not in CCIp) when derived from the same thermodilution curve.
We compared the intermittent transpulmonary and the continuous pulmonary thermodilution techniques and associated measurements in predicting and monitoring fluid responses after coronary artery surgery. This allowed us to assess the dependency of the alleged superiority of filling volumes over pressures on cardiac output measurement techniques and their suitability to monitor fluid responses. We found fair overlap of cardiac output responses and relatively poor predictive and monitoring values of filling pressures. Mathematical coupling nevertheless overestimated the value, that is correlation with cardiac output (changes) of volumes over pressures, when derived from the same transpulmonary thermodilution curve as cardiac output.
The two cardiac output measurement techniques are supposedly less dependent on modulations of cardiac output by the mechanical respiratory cycle than bolus thermodilution measurements in the pulmonary artery. We found a fair correlation between CItp and CCIp and changes therein, in the absence of overt tricuspid regurgitation and rapid rewarming that may confound measurements. The limits of agreement were around or even above the 30% of the mean cardiac output, the criterion by Critchley and Critchley  for a clinically useful alternative to a standard technique, and lower after excluding two CItp outliers, with systematic overestimation (bias) by the transpulmonary technique as described before [4,19–23,25]. We cannot formally judge whether the relatively high limits of agreement, particularly after adjusting for within-patients variability, can be attributed to one technique or the other (or both), but our results suggest that the two patients with highest within-patient variability indeed had the greatest errors in CItp (versus CCIp), so that their exclusion hardly affected between-patients variability. Moreover, use of a Bland Altman analysis taking repeated measurements into account, thereby elevating limits of agreement in case of higher between-patient than within-patient variability, is not beyond criticism and the Critchley criterion may not apply . Otherwise, the overestimation may partly relate to thermal loss , whereas the continuous thermodilution measurement does not carry such a problem. In spite of incomplete agreement of absolute values, the cardiac output response to fluid loading was similar among techniques, in accordance with the single prior study on this issue , and, together with the high reproducibility of CCIp, this suggests that CCIp had reached a steady state after fluid-loading steps. Apparently, the 15 min observation period after each fluid loading step was sufficient to reach a steady state CCIp response, as noted before [26–28].
Fluid responses were not predicted or monitored well by filling pressures, in accordance with the literature [1,2,4–10,12,13], in spite of the low PEEP used in this study. Apparent superiority of CVP over PAOP can be explained if left ventricular function was well maintained after surgery. The poor value was independent of the manner to obtain and define fluid responses. In contrast, transpulmonary volumes have been claimed to better predict and monitor fluid responses than filling pressures, since, among others, fluid-induced changes better correlated with changes in transpulmonary stroke volume and cardiac output than pressures after cardiac surgery [1,2,4–10,12]. However, this has rarely been compared with an independent cardiac output measurement technique .
Our study suggests that the correlation between volumes and CItp changes which is greater than that between volumes and CCIp changes can be accounted for, at least in part, by mathematical coupling of shared measurement error between CItp and volumes, when derived from the same thermodilution curve. The greater value of transpulmonary volumes than of filling pressures in monitoring (transpulmonary) cardiac output (changes) was abolished after excluding two patients with CItp outliers as compared to CCIp. Hence, the correlation between volumes and CItp was overestimated and spuriously increased over that with pressures. Indeed, a shared measurement error may spuriously increase a correlation coefficient when the range of observations is relatively small and/or the measurement error is relatively large [3,16–18]. However, animal and clinical studies [14,15], using the transpulmonary dilution technique, showed that cardiac output may change independently from volumes following infusion of inotropes or ß-blockers, but this does not exclude spurious covariance by mathematical coupling during fluid loading, when volumes and cardiac output change in the same direction. In comparing transpulmonary with continuous pulmonary thermodilution cardiac output measurements after coronary artery surgery, Hofer et al. did not find overestimation of transpulmonary volumes over pressures as monitors of fluid responses based on mathematical coupling, in the absence of numerical data, perhaps because their fluid-induced increases in cardiac output were greater than in our study and the range of observations relative to measurement error was larger.
We did not include dynamic indicators to predict fluid responses in this study, because the goal was to compare static ones, and because dynamic indicators like stroke volume variations may also depend on ventilatory conditions [7,9,12,21,22,24,31]. The number of patients studied was relatively small, even after presuming fluid responses in 50% of steps, but apparently sufficient to demonstrate the effect of mathematical coupling. A formal power analysis was not considered feasible for comparison of techniques to evaluate the role of mathematical coupling.
In conclusion, our data suggest that, in mechanically ventilated patients after coronary artery surgery, fluid responses can be assessed by either intermittent transpulmonary or continuous pulmonary thermodilution measurements of cardiac output. Whereas filling pressures are poor predictors and monitors of fluid responses, the superiority of transpulmonary volumes over filling pressures can be overestimated by mathematical coupling of a shared measurement error when derived from the same transpulmonary thermodilution curve as cardiac output.
1 Gödje O, Peyerl M, Seebauer T, et al
. Central venous pressure, pulmonary capillary wedge pressure and intrathoracic blood volumes as preload indicators in cardiac surgery patients. Eur J Cardiothorac Surg 1998; 13:533–539.
2 Hinder F, Poelaert JI, Schmidt C, et al
. Assessment of cardiovascular volume status by transoesophageal echocardiography and dye dilution during cardiac surgery. Eur J Anaesthesiol 1998; 15:633–640.
3 Bindels AJGH, van der Hoeven JG, Graafland AD, et al
. Relationships between volume and pressure measurements and stroke volume in critically ill patients. Crit Care 2000; 4:193–199.
4 Wiesenack C, Prasser C, Keyl C, Rödig G. Assessment of intrathoracic blood volume as an indicator of cardiac preload: single transpulmonary thermodilution technique versus assessment of pressure preload parameters derived from a pulmonary artery catheter. J Cardiothorac Vasc Anesth 2001; 15:584–588.
5 Brock H, Gabriel C, Bibl D, Necek S. Monitoring intravascular volumes for postoperative volume therapy. Eur J Anaesthesiol 2002; 19:288–294.
6 Reuter DA, Felbinger TW, Moerstedt K, et al
. Intrathoracic blood volume index measured by thermodilution for preload monitoring after cardiac surgery. J Cardiothorac Vasc Anesth 2002; 16:191–195.
7 Reuter DA, Felbinger TW, Schmidt C, et al
. Stroke volume variations for assessment of cardiac responsiveness to volume loading in mechanically ventilated patients after cardiac surgery. Intensive Care Med 2002; 28:392–398.
8 Rex S, Brose S, Metzelder S, et al
. Prediction of fluid responsiveness in patients during cardiac surgery. Br J Anaesth 2004; 93:782–788.
9 Hofer CK, Müller SM, Furrer L, et al
. Stroke volume and pulse pressure variation for prediction of fluid responsiveness in patients undergoing off-pump coronary artery bypass grafting. Chest 2005; 128:848–854.
10 Hofer CK, Furrer L, Matter-Ensner S, et al
. Volumetric preload measurement by thermodilution: a comparison with transoesophageal echocardiography. Br J Anaesth 2005; 94:748–755.
11 Nirmalan M, Willard TM, Edwards DJ, et al
. Estimation of errors in determining intrathoracic blood volume using the single transpulmonary thermal dilution technique in hypovolemic shock. Anesthesiology 2005; 103:805–812.
12 Preisman S, Kogan S, Berkenstadt H, Perel A. Predicting fluid responsiveness in patients undergoing cardiac surgery: functional haemodynamic parameters including the respiratory systolic variation test and static preload indicators. Br J Anaesth 2005; 95:746–775.
13 Marik PE, Baram M, Vahid B. Does central venous pressure predict fluid responsiveness? A systematic review of the literature and the tale of seven mares. Chest 2008; 134:172–178.
14 McLuckie A, Bihari D. Investigating the relationship between intrathoracic blood volume index and cardiac index. Intensive Care Med 2000; 26:1376–1378.
15 Buhre W, Kazmaier S, Sonntag H, Weyland A. Changes in cardiac output and intrathoracic blood volume: mathematical coupling
of data? Acta Anaesthesiol Scand 2001; 45:863–867.
16 Moreno LF, Stratton HH, Newell JC, Feustel PJ. Mathematical coupling
of data: correction of a common error for linear calculations. J Appl Physiol 1986; 60:335–343.
17 Stratton HH, Feustel PJ, Newell JC. Regression of calculated variables in the presence of shared measurement error. J Appl Physiol 1987; 62:2083–2093.
18 Squara P. Mathematic coupling of data: a frequently misused concept. Intensive Care Med 2008; 34:1916–1921.
19 Gödje O, Peyerl M, Seebauer T, et al
. Reproducibility of double indicator dilution measurements of intrathoracic blood volume compartments, extravascular lung water, and liver function. Chest 1998; 113:1070–1077.
20 Gust R, Gottschalk A, Bauer H, et al
. Cardiac output measurement by transpulmonary versus conventional thermodilution technique in intensive care patients after coronary artery bypass grafting. J Cardiothorac Vasc Anesth 1998; 12:519–522.
21 Gödje O, Hoeke K, Lichtwarck-Aschoff M, et al
. Continuous cardiac output by femoral arterial thermodilution calibrated pulse contour analysis: comparison with pulmonary arterial thermodilution. Crit Care Med 1999; 27:2407–2412.
22 Zöllner C, Haller M, Weis M, et al
. Beat-to-beat measurement of cardiac output by intravascular pulse contour analysis: a prospective criterion standard study in patients after cardiac surgery. J Cardiothorac Vasc Anesth 2000; 14:125–129.
23 Mielck F, Buhre W, Hanekop G, et al
. Comparison of continuous cardiac output measurements in patients after cardiac surgery. J Cardiothorac Vasc Anesth 2003; 17:211–212.
24 Sander M, von Heymann C, Foer A, et al
. Pulse contour analysis after normothermic cardiopulmonary bypass in cardiac surgery patients. Crit Care 2005; 9:R729–R734.
25 Hüter L, Schwarzkopf KR, Preussler NP, et al
. The level of cardiac output affects the relationship and agreement between pulmonary artery and transpulmonary aortic thermodilution measurements in an animal model. J Cardiothorac Vasc Anesth 2007; 21:659–663.
26 Haller M, Zöllner C, Briegel J, Forst H. Evaluation of a new continuous thermodilution cardiac output monitor in critically ill patients: a prospective criterion standard study. Crit Care Med 1995; 23:860–866.
27 Siegel LC, Hennesy MM, Pearl RG. Delayed time response of the continuous cardiac output pulmonary artery catheter. Anesth Analg 1996; 83:1173–1177.
28 Lazor MA, Pierce ET, Stanley GD, et al
. Evaluation of the accuracy and response time of STAT-mode continuous cardiac output. J Cardiothor Vasc Anesth 1997; 11:432–436.
29 Critchley LAH, Critchley JAJH. A meta-analysis of studies using bias and precision statistics to compare cardiac output measurement techniques. J Clin Monit Comp 1999; 15:85–91.
30 Medin DL, Brown DT, Wesley R, et al
. Validation of continuous thermodilution cardiac output in critically ill patients with analysis of systemic errors. J Crit Care 1999; 82:525–530.
31 Rödig G, Prasser C, Keyl C, et al
. Continuous cardiac output measurement: pulse contour analysis vs. thermodilution technique in cardiac surgical patients. Br J Anaesth 1999; 82:525–530.
32 Rauch H, Müller M, Fleischer F, et al
. Pulse contour analysis versus thermodilution in cardiac surgery patients. Acta Anaesthesiol Scand 2002; 46:424–429.
33 Bendjelid K, Schütz N, Suter PM, Romand JA. Continuous cardiac output monitoring after cardiopulmonary bypass: a comparison with bolus thermodilution. Intensive Care Med 2006; 32:919–922.
34 Button D, Weibel L, Reuthebuch O, et al
. Clinical evaluation of the FloTrac/VigileoTM system and two established continuous cardiac output monitoring devices in patients undergoing cardiac surgery. Br J Anaesth 2007; 99:329–336.
35 Sakka SG, Ruhl CC, Pfeiffer UJ, et al
. Assessment of cardiac preload and extravascular lung water by single transpulmonary thermodilution. Intensive Care Med 2000; 26:180–187.
36 Bland JM, Altman DG. Measuring agreement in method comparison studies. Stat Meth Med Res 1999; 8:135–160.
37 Myles PS, Cui J, Editorial I. Using the Bland-Altman method to measure agreement with repeated measures. Br J Anaesthesia 2007; 99:309–311.