Fluid administration in critically ill patients is typically performed to increase cardiac preload, followed by a raise in cardiac output. However, studies conducted during the past few years have shown that about 50% of critically ill patients do not exhibit the desired effect (they are not fluid responsive) 1. Thus, we require an accurate and reliable technique to guide fluid management. Pressure preload variables (central venous pressure and pulmonary capillary wedge pressure), which continue to be used, often fail to provide reliable information regarding cardiac preload 2 and are incapable of predicting cardiac response to fluid therapy 3. On the other hand, the volumetric preload variables that are assessed by transpulmonary thermodilution may better reflect left ventricular preload 4, but they do not allow assessment of fluid responsiveness 35.

As an alternative to these static variables, a dynamic approach may be used in the form of preload monitoring to guide fluid therapy. With passive leg rising in spontaneously breathing patients, the heart's reaction (increased venous return) can be assessed without any fluid administration 6, and in mechanically ventilated patients the interaction between heart and lung can be used to predict fluid responsiveness 7.

Different, less invasive haemodynamic monitoring systems based on arterial pulse contour analysis allow stroke volume variation (SVV) to be tracked continuously. SVV assessed using the PiCCO_plus™ system (Pulsion Medical Systems, Munich, Germany; SVV_PiCCO) has repeatedly been shown to predict fluid responsiveness well in various clinical settings3891011, whereas only sparse data are available for SVV determined using the recently introduced FloTrac™/Vigileo™ system (Edwards Lifesciences, Irvine, CA, USA; SVV_FloTrac). In a study conducted by de Waal and coworkers 12, SVV_FloTrac failed to predict fluid responsiveness. This finding may be attributable to the initial version of the device's software, adapting vascular compliance every 10 minutes. Limited accuracy in assessment of cardiac output (and thus of stroke volume) could be achieved using this early software version 13. Modification to the software (reduction in the time window for vascular adjustment to 1 minute) resulted in improved accuracy in measuring cardiac output 14.

The aim of the present study was to compare SVV_FloTrac with SVV_PiCCO during a volume shift manoeuvre (by changing body positioning) in patients who had undergone elective off-pump coronary artery bypass grafting.

40 patients (american society of anesthesiologists risk classification III, female/male ratio 1/4, age 66.5 ± 9.2 years, left ventricular ejection fraction 56.1 ± 10.0%) were enrolled in this study. During the study period, no significant changes in ventilatory tidal volumes or peak inspiratory pressure were identified.

Change in body position from 30° head-up to 30° head-down resulted in significantly increased SV, GEDV and CVP, whereas SVV_FloTrac, SVV_PiCCO and PPV significantly decreased (Table 1 and Figure 1). The mean increase in SV was 24 ± 14%. Twenty-three patients (58%) had an increase in SV of greater than 25% (mean change 34 ± 9%). A SV increase of under 10% was observed in eight patients (20%; mean change = 5 ± 3%).

For patients with an increase in SV of greater than 25%, baseline SVV_FloTrac and SVV_PiCCO were 16 ± 4% and 19 ± 5%, respectively. In patients with an increase in SV of under 10%, baseline SVV_FloTrac and SVV_PiCCO were 9 ± 2% and 11 ± 3%, respectively. Results of ROC curve and linear regression analyses for the prediction of SV changes induced by altered body positioning are summarized in Figure 2 and Table 2. There was no significant difference between AUCs with respect to identifying a SV increase of more than 25% for SVV_FloTrac and SVV_PiCCO (Table 3). Optimal threshold values given by the ROC curves were 9.6% for SVV_FloTrac (sensitivity 91% and specificity 83%) and 12.1% for SVV_PiCCO (sensitivity 87% and specificity 76%). Based on these threshold values, positive and negative predictive values were 80% and 92% for SVV_FloTrac, respectively; corresponding values for SVV_PiCCO were 77% and 90%. There was no significant difference between identification of patients with a SV increase of more than 25% using SVV_FloTrac and SVV_PiCCO (P = 0.523). Correlations between SVV_FloTrac and SVV_PiCCO and changes of SV (ΔSV) were similar (Tables 2 and 4, and Figure 3).

The AUCs for PPV, GEDV and CVP were significantly lower than the AUCs for SVV_FloTrac and SVV_PiCCO (Table 3). The optimal threshold value was 15.4% for PPV, yielding a sensitivity of 60% and a specificity of 62%, and negative and positive predictive values being 68% and 60%, respectively. Moreover, the correlation between PPV with ΔSV was significant, whereas no significant correlations of GEDV and CVP with ΔSV could be established (Tables 2 and 4).

Bland-Altman analysis (SVV_FloTrac - SVV_PiCCO) revealed a mean bias ± 2 standard deviations (limits of agreement) of -2.5 ± 6.1% for head-up measurements and -1.5 ± 3.6% for head-down measurements (Figure 4). Correlation coefficients for SVV_FloTrac versus SVV_PiCCO (r²) were 0.612 and 0.757 for head-up and head-down measurements, respectively.

SVVs as assessed using the modified FloTrac™/Vigileo™ and the PiCCOplus™ systems were comparable predictors of SV changes related to a fluid shift manoeuvre in patients who had undergone off-pump coronary artery bypass grafting. There was clinically acceptable agreement and a strong correlation between SVV assessed using the two devices. However, in comparison with SVV_PiCCO, it must be noted that the SVV_FloTrac device has a lower threshold value.

Differences between the devices in terms of absolute values, and thus their threshold values, may primarily be explained by different signal detection sites and specific differences in the measurement techniques, despite the fact that both systems use a similar SVV algorithm. Both the FloTrac/Vigileo™ and the PiCCOplus™ system are based on analysis of pulse pressure which is the result of the interaction between SV and the systemic vascular system. Therefore, vascular resistance and compliance at the site of signal detection must be considered in order to determine SV and SVV reliably. The FloTrac/Vigileo™ system assesses SV typically using signal detection via a peripheral radial artery. It analyzes the impact of vascular tone on pressure during a period of 20 seconds, and adjusts for actual vascular tone at intervals of 1 minute based on wave form analysis and patient characteristics. In contrast, with the PiCCOplus™ system aortic pressure wave forms are recorded usually via a specific thermistor-tipped arterial catheter in the femoral artery. SV is then calculated using an algorithm that assesses the area under the systolic part of the pressure wave form beat-to-beat after calibration by transpulmonary thermodilution 16.

Only limited data are available regarding SVV as assessed using the FloTrac/Vigileo™ system. Recently, de Waal and coworkers 12 demonstrated that SVV_FloTrac was unable to predict fluid responsiveness in cardiac surgery patients; their findings contradict those of the present study. However, de Waal and coworkers used the FloTrac/Vigileo™ system with an early version of the software, which employed a time window for adjustment to vascular tone of 10 minutes. Based on the limited accuracy of cardiac output assessment observed with this early software version 13, the algorithm has been modified and – a major change – the time window has been reduced to 1 minute. These modifications resulted in improved measurement of cardiac output 14, and in the present study, in which we used a new software version, we demonstrated that SVV_FloTrac is useful for predicting fluid responsiveness.

The SVV_PiCCO findings in the present study are in agreement with those of a variety of previous investigations. SVV_PiCCO could predict fluid responsiveness in patients with severe sepsis 8, in those undergoing neurosurgery 19 and in cardiac surgery patients with preserved 3910 as well as reduced left ventricular function 9. Interestingly, Wiesenack and coworkers 20 identified no correlation between SVV_PiCCO and SV changes during a fluid trial using an older version of the PiCCO algorithm. However, favourable results however were later reported by the same study group 11 using a PiCCO algorithm that has been modified to address better aortic compliance in the individual patient.

Despite the clinical importance of SVV threshold values, little information is available in the literature. A SVV threshold value represents the 'trigger' value at which initiation of fluid replacement therapy is expected to result in a positive cardiac response. ROC analyses performed thus far have revealed that SVV_PiCCO at the threshold value of 9.5% results in a stepwise SV index increase of 5% or more in studies involving incremental volume administration 919. In a previous study 3 we observed that a SVV_PiCCO threshold value of 12.5% resulted in an increase in SV index of 25% or more after fluid loading. In the present study, by inducing a fluid shift by altering body positioning, a comparable SVV_PiCCO threshold value (12.1%) for SV increase greater than 25% was identified, whereas the SVV threshold value for the FloTrac/Vigileo™ system was 9.6%.

Variations in SV are followed by concomitant changes in arterial pressure, which can be assessed by measuring PPV. In addition to SVV, the PiCCO_plus system allows automated assessment of PPV. Reuter and coworkers 21 found a close correlation between PPV and SVV. Moreover, in a previous study 3, we demonstrated that both SVV and PPV were similarly able to predict fluid responsiveness. However in the present study the predictive performance of PPV was inferior to those of SVV_Flotrac and SVV_PiCCO; it is likely that this finding is mainly accounted for by the study design. In this study preload changes were induced by changes in body positioning, whereas in the previous study 3 a fluid trial using hetastarch solution was performed. Body positioning might have influenced vascular tone, and PPV is known to be more susceptible to changes in vasomotor tone than SVV 22.

In contrast to SVV and PPV, conventional static preload parameters assessed in this study failed to predict fluid responsiveness, confirming the findings of previous work related to this issue 323. CVP, the standard pressure preload variable, did not adequately reflect preload status 24 and is therefore unsuitable for predicting ventricular response to fluid loading. GEDV, a volumetric preload variable, more adequately reflected changes in cardiac preload 424, but this parameter also failed to predict the reaction of the heart to fluid loading.

When interpreting of the data presented in this study, some methodological aspects and limitations must be considered. First, we induced a fluid shift by changing body positioning and not by subjecting patients to a 'real' fluid challenge. Changes in body positioning might have resulted in alterations to systemic vascular resistance based on sympathetic activity and not on changes in preload in isolation. Thus, a fluid shift due to changed body positioning and a true episode of hypovolaemia followed by intravenous fluid administration are not necessarily comparable. However, the sympathetic effect was minimized by using adequate sedation and analgesia in all patients during the study. Furthermore, the significant increase in GEDV observed after changed body positioning indicates an increase in preload. Moreover, body positioning is a typical manoeuvre for assessing fluid responsiveness 16. In addition, a similar approach has previously been used to determine fluid responsiveness of SVV_PiCCO 10, and we previously observed a similar haemodynamic pattern and comparable threshold value for SVV_PiCCO in a true fluid challenge 3. Second, mechanical ventilation and a regular heart rhythm are prerequisites for reliable SVV assessment. In order to satisfy these requirements, patients were fully sedated, neuromuscular blockade was used and fixed cardiac pacing was applied in all patients. However, pacing resulted in a lack of heart rate response to the fluid shift. Moreover, it may not be possible to achieve these optimal conditions in clinical practice.

Thus, our findings may not be directly applicable to other situations or to patients other than the population studied (patients with preserved left ventricular function after elective cardiac surgery). Therefore, further evaluation of SVV_FloTrac using intravenous fluid administration under different conditions and populations of critically ill patients is needed.

In conclusion, SVV assessed using the FloTrac™/Vigileo™ and the PiCCOplus™ system exhibited comparable performance in terms of predicting fluid responsiveness. However, in comparison with SVV_PiCCO, SVV_FloTrac has a lower threshold value.

• Fluid responsiveness can reliably be assessed using SVV determined using the modified FloTrac™/Vigileo™ system.

• SVV assessed using the modified FloTrac™/Vigileo™ and the PiCCOplus™ systems exhibited similar performance.

• The threshold value was lower for SVV_FloTrac (9.6%) than for SVV_PiCCO (12.1%).

• Compared with SVV, PPV was inferior in terms of predicting fluid responsiveness; static preload variables assessed in the study failed to provide valuable information on fluid responsiveness.

AUC = area under the receiver operating characteristic curve; CVP = central venous pressure; GEDV = global end-diastolic volume; PPV = pulse pressure variation; ROC = receiver operating characteristic; SV = stroke volume; SVV = stroke volume variation; SVV_FloTrac = SVV assessed using FloTrac™/Vigileo™; SVV_PiCCO = SVV assessed using PiCCO_plus™.

This study was supported in part by a research grant from Edwards Lifesciences (Irvine, CA, USA). In the past, the Institute of Anaesthesiology and Intensive Care Medicine, Triemli City Hospital have held research grants from Pulsion Medical Systems (Munich, Germany) and Edwards Lifesciences. CKH has received lecture fees from both Pulsion Medical Systems and Edwards Lifesciences in the past.

CKH was responsible for study design and protocol, patient recruitment, measurements, data collection, statistical analysis and manuscript writing. AS was responsible for patient recruitment, measurements, data collection and manuscript writing. LW was responsible for patient recruitment, data collection and technical support. AZ was responsible for study design and protocol.