The maintenance of a stable circulation is therefore important in cirrhotic patients. Plasma expansion with albumin has been found to protect against renal failure in spontaneous bacterial peritonitis [7]. It has also become a mainstay in the treatment of HRS and a prerequisite of its diagnosis, particularly when, according to a current consensus statement, "true hypovolaemia" has to be excluded by administration of a substantial amount (up to 100 g daily) of albumin over two days [8]. During acute conditions such as infection or haemorrhage, as well as for the treatment of functional renal failure, parameter guided fluid therapy may be useful for avoiding or improving relative hypovolemia, but also for avoiding potential complications of fluid overload. Data on the haemodynamic effects of an albumin infusion in cirrhotic patients, however, are scarce and it is not clear which parameters should be employed to guide fluid therapy in cirrhotic patients. Central venous pressure (CVP) has been shown to be of little value in the assessment of fluid responsiveness in critically ill patients [9-11]. In addition, elevated intra-abdominal pressure, as seen in ascitic patients, may influence CVP [12]. It therefore is of no surprise that in a recent study the traditional target values for CVP apparently failed to exclude hypovolaemia in patients with HRS[13]. Pulmonary artery catheters have traditionally been used for guiding haemodynamic interventions. Their use however remains controversial as they may be associated with complications and their predictive value in the assessment of fluid responsiveness is low [9,14-16]. Based on earlier work on indicator dilution techniques for the measurement of cardiac output (CO) and intra-thoracic blood volumes [17,18], trans-pulmonary thermodilution combined with continuous measurement of CO by pulse contour analysis has in recent years been evaluated [10,11,19]. Trans-pulmonary thermodilution has shown some promise in guiding fluid therapy by providing preload-associated parameters such as the global end-diastolic volume (GEDV). A treatment protocol targeting GEDV assessed by trans-pulmonary thermodilution resulted in reduced need for catecholamines and less time on mechanical ventilation in cardiac surgery patients [20].

It is, however, questionable if results from studies in non-cirrhotic patients are to be extrapolated to cirrhotic patients. Earlier studies found that central blood volume in cirrhotic patients cannot be significantly expanded, thus casting doubt on possible correlations between intra-thoracic blood-volumes, as a measure of preload, and CO [21,22]. To date, trans-pulmonary thermodilution has only been evaluated in ventilated cirrhotic patients during orthotopic transplantation of the liver [23].

The improvement in CI after volume therapy supports the notion that relative central hypovolaemia contributes to circulatory dysfunction in cirrhotic patients. After plasma expansion with albumin we found an increase in central blood volume. This is in contrast to the results of other studies who failed to detect relevant changes in central blood volume after fluid loading in patients with advanced cirrhosis [22,27]. We believe that in these studies the possible effects were missed due to the small number of patients included. Indeed, both studies showed increases in central blood volume after volume loading, however, this failed to reach statistical significance.

In our study, baseline GEDVI correlated with baseline CI and SVI, and GEDVI was lower in patients with a positive response to volume loading than in those with a negative response. Furthermore, increases in GEDVI correlated to increases in CI and SVI, highlighting that GEDVI, evaluated by trans-pulmonary thermodilution, behaves as an indicator of preload in patients with cirrhosis. CVP was lower in patients who responded to volume loading and increased significantly after infusion. The volume loading-induced changes, however, were not proportional to the changes in CI and SVI, and baseline CVP did not correlate with CI or SVI. This confirms previous reports and underlines the limited value of CVP as a marker of cardiac preload. Both parameters performed poorly as predictors of volume responsiveness as has been previously documented in various clinical settings [16,28].

Cardiac preload is defined as myocardial wall tension at end diastole, and, according to Laplace's law, is determined by ventricular geometry and intra-ventricular pressures. Myocardial contractility depends on end-diastolic tension of the myocardial sarcomers and the connection between increasing preload and contractility is given in the sigmoidal Frank-Starling curve. Without knowing the individual myocardial properties at the moment of interest, we cannot determine the position on the Frank-Starling curve of any given preload condition. This explains why good intra-individual correlations between preload markers and CI in paired measurements may be accompanied by a low predictive value of single measurements of preload associated markers for fluid responsiveness.

Following volume challenges we observed substantial decreases in SVRI in our patients. SVRI is dependant on (MAP-CVP) and CI by a linear relationship. Therefore, with constant CVP, any changes in CI must be accompanied by proportional changes in SVRI, MAP, or both. In patients with septic shock, opposite changes of a similar relative size of both MAP and SVRI have been observed after volume loading [11]. In contrast, we found large decreases in SVRI with only minuscule increases in MAP. This contrasts to a previous study on plasma expansion in patients with spontaneous bacterial peritonitis (SBP) [29]. Here the authors described an increase in peripheral vascular resistance after treatment of SBP with antibiotics and albumin. They hypothesized that this may be due to the pharmacological action of albumin as a scavenger of nitric oxide, thus reducing the vasodilatory properties of plasma. However, haemodynamic measurements in this study were days apart and the increased vasotonus, may have been due to reduced septic vasodilation. In our study cohort, care was taken to select patients without infection or haemorrhage, so that any related confounding factors were avoided.

Whereas MAP was not different between the patients who responded to volume loading and those who did not, SVRI was significantly and by a large proportion higher in responders when compared to non-responders. However, CI was significantly lower in responders than in non-responders. Pre-infusion values of SVRI (and CI) were predictive of volume responsiveness in our patients. As suggested previously [30], this may indicate that in a proportion of patients with cirrhotic circulatory dysfunction, relative central hypovolaemia, resulting in further activation of endogenous vasopressor systems to maintain MAP at the cost of high peripheral resistance, may contribute to impaired cardiac output, despite what is essentially a hyperdynamic circulation. Volume therapy may thus decrease vasopressor activation, and may lead to decreased levels of endogenous vasopressors such as norepinephrine, renin and angiotensin, as has been described previously [22]. Renal dysfunction in cirrhosis deteriorates along a continuum starting with an impaired capacity to excrete sodium and free water leading to an oedematous state with increased plasma volume and ascites, to pre-renal failure and, finally, irreversible tubular damage. According to current understanding, an important etiologic factor is elevated levels of vasoconstrictors affecting the renal microcirculation, narrowing the kidneys' capacity to cope with additional haemodynamic insults. Volume management may be relevant to the prevention and treatment of functional renal failure in cirrhosis. Recent studies on vasopressor therapy in HRS highlight the importance of adequate volume status. Whereas it had previously been shown that albumin was necessary for the beneficial effect of terlipressin [31], a recent study by Alessandria et al. showed that a substantial number of patients included in a study on treatment of HRS responded to plasma expansion alone when it was tailored according to CVP instead of using the usual fixed-dose regimen [13]. In this study the aim was a CVP of 10 – 15 mmHg. In our study, 33% of patients with a CVP greater than 10 mmHg and 24% of patients a CVP of over > 15 mmHg still responded to albumin infusion with a further increase in CI. Consequently, neither CVP nor GEDVI should be recommended as parameters to direct fluid resuscitation in cirrhotic patients with pre-renal kidney failure.

In ventilated patients, dynamic parameters such as pulse pressure variation or stroke volume variation have shown much better predictive power for assessing fluid responsiveness [28]. However, the majority of cirrhotic patients at risk of renal failure are breathing spontaneously and these circumstances, dynamic parameters are not applicable. A time honoured method for the assessment of fluid responsiveness, "passive leg raising" (PLR) [32], has recently gained renewed interest in the intensive care setting. PLR generates a transient increase in venous return. The immediate haemodynamic response of mean blood flow to this manoeuvre, assessed by methods such as oesophageal Doppler [33] or trans-thoracic echocardiography [34], has been used to estimate fluid responsiveness. The recently published method of PLR is difficult to apply in the intensive care setting, because a fixed angle of the hips is required throughout the procedure and the whole bed must be tilted instantaneously by 45°. Cirrhotic patients may react differently to tilting than other patients or normal controls [35], and the elevated intra-abdominal pressure of ascitic patients may also affect PLR-induced blood transfer [36]. Therefore, PLR may give different results in cirrhotic patients. This has not to our knowledge been evaluated.

Without static parameters predictive of fluid responsiveness, but a variety of monitoring tools capable of providing data on CI and MAP, iterative protocols of fluid challenges may offer the possibility of increasing cardiac output in patients with reduced effective intravascular volume [14]. Whether this translates to improved kidney function in cirrhotic patients with renal failure should be evaluated in future studies.