Renal dysfunction is a relatively frequent problem in hospitalized patients with cirrhosis. In particular, acute renal failure (ARF) occurs in approximately 20% of patients . Although different criteria were used for the definition of renal dysfunction  in patients with cirrhosis, the most commonly used is the serum creatinine level (SCr) ≥1.5 mg/dl (133 μmol/L). A percentage increase in SCr of 50% or more (1.5-fold from baseline) to a final value of SCr ≥ 1.5 mg/dl (133 μmol/L) is the most commonly used criterion for the diagnosis of ARF in these patients [8, 9]. Both diagnostic criteria are totally arbitrary as they have never been validated in prospective clinical studies. It is also worth noting that other diagnostic criteria have been recently proposed for the diagnosis of ARF. The ADQI Working Group [10, 11], and the Acute Kidney Injury Network (AKIN)  developed and published a consensus definition of ‘acute kidney injury’ (AKI), a new term for ARF. The AKI is defined as an abrupt (arbitrarily set at 48 h) impairment in renal function defined by an absolute increase in SCr of 0.3 mg/dl or more (≥26.4 μmol/L), or a percentage increase in SCr of 50% or more (1.5-fold from baseline), or a urine output of less than 0.5 ml/kg per hour for more than 6 h [10-12]. Now, if it is immediately obvious that the criterion based on the urine volume cannot be applied to patients with decompensated cirrhosis, it is important to focus on the remaining part of the definition. When compared with the conventional criteria for ARF in patients with cirrhosis, the main innovative aspects introduced by the new criteria are: (a) the absolute increase in SCr provided beyond the percentage increase equal to or higher than 50% (b) the abolition of the final value of SCr > 1.5 mg/dl (133 μmol/L) .
To date, few data on these issues are available. Two studies involving critically ill patients with cirrhosis admitted into an intensive care unit already showed that the new criteria for AKI were a good predictor of hospital survival [13, 14]. Another study in a similar cohort of patients showed that the new criteria were good predictors of both in hospital and 6 months mortality . One can argue that these studies were conducted on critically ill patients with cirrhosis, long way from everyday clinical practice. Nevertheless, it has been shown that AKIN criteria predict in-hospital mortality with a sensitivity similar to the conventional criteria for the definition of renal dysfunction in patients with cirrhosis. . In addition, recent results showed that the definition of AKI on the new criteria has high sensitivity when it comes to identifying patients with advanced cirrhosis at increased risk of 3 month mortality . These results also showed that the addition of a final value of SCr ≥ 1.5 mg/dl (133 μmol/L) to the new criteria provided a slight significant increase in the sensitivity and that a SCr ≥ 1.5 mg/dl (133 μmol/L) per se had no prognostic value (Table. 1). Interestingly, this also seems to agree with the opinion of most ICA members, who were questioned about the usefulness of the cut-off value of SCr ≥ 1.5 mg/dl (133 μmol/L) when defining renal dysfunction. Most of them (83.3 vs 16,7%) admitted to considering it a possible limit in the prompt management of patients with renal dysfunction. Therefore, further and larger prospective studies are needed to assess the ability of new criteria vs. the conventional criteria of renal dysfunction in the prediction of survival in patients with cirrhosis, as suggested by a recent Working Party .
Table 1. Three month survival (%) according to the presence of acute kidney injury (AKN) defined by Acute Kidney Injury Network (AKIN) criteria and/or serum creatinine (sCr) value [data from Fagundes et al. ()]
|AKI defined by AKIN criteria||38%|
|AKI defined by AKIN criteria and sCr ≥1.5 mg/dl||29%|
|AKI defined by AKIN criteria and sCr<1.5 mg/dl||58%|
|AKI defined by sCr ≥1.5 mg/dl and no AKIN criteria||80%|
|No AKI according to the AKIN criteria||87%|
Beyond this, a critical reflection must also be made on SCr and particularly on the use of SCr as a marker of renal function in patients with advanced cirrhosis. The increase in SCr is delayed compared with renal injury and is, therefore, a late marker of decreased renal function, meaning that a significant renal disease can exist with minimal or no changes in SCr because of enhanced tubular creatinine secretion, or other factors . The SCr is also greatly influenced by numerous non-renal factors such as body weight, race, age, gender. Moreover, in patients with cirrhosis, SCr may be an even poorer marker of kidney function because of a reduced muscle mass. Thus, the release of creatinine in these patients may be considerably reduced, and therefore, they may have a normal SCr in the context of a very low glomerular filtration rate (GFR) . Based on these considerations, it cannot be disregarded that SCr is a key component of the MELD score . Although a criticism of the MELD score is beyond the scope of this review, it should be emphasized that the use of SCr in the MELD score can be a potential source of injustice in the allocation of grafts in liver transplantation (LT) [22, 23], and that its specific weight in the calculation of the score is now debated . The simplified ‘Modification of Diet in Renal Disease’ formula (MDRD6) provides an estimate of GFR relative to SCr corrected by age, race and gender in chronic kidney disease (CKD) when SCr is at a steady state, although it is not useful in AKI when SCr is not in equilibrium. In addition, although MDRD6 had greater precision than other equations to assess GFR in patients with cirrhosis, its accuracy was lower than that reported in other populations, as it overestimates GFR in patients with cirrhosis and renal dysfunction [25-28]. Therefore, better methods for estimating the GFR are required for evaluation of renal function in patients with advanced cirrhosis. In this context, cystatin C seems to be a promising biomarker of renal function , even if it requires further validation using ‘gold standard’ measures of GFR, such as iodothalamate or inulin clearance. Therefore, nowadays, despite all its limitations, SCr remains in clinical partice the key biomarker in the diagnosis of ARF in cirrhosis .
The most common phenotype of ARF in hospitalized patients with cirrhosis is pre-renal ARF, accounting for approximately 45% of the cases . It is followed by intra-renal ARF (32%) including acute tubular necrosis (ATN) and glomerulonephritis, hepatorenal syndrome (HRS) (23%) and post-renal ARF (<1%) . It may be difficult to make a differential diagnosis among the main phenotypes of ARF in patients with cirrhosis because precipitating factors, as well as clinical presentation, may almost be the same as for example it is the case for pre-renal ARF and HRS. Nevertheless, an algorithm helping in the differential diagnostic process can be proposed (Fig. 1). This is based mainly on the more recent diagnostic criteria for HRS  (Fig. 2). Differentiation between pre-renal ARF and HRS is based, by definition, on the fact that pre-renal ARF improves with volume expansion. However, an assessment of intravascular filling achieved by albumin administration is not required; therefore the rate of response in terms of reduction of SCr may be highly variable among these patients . A more accurate evaluation of the response to volume challenge is needed and should be covered in the future research agenda. Differentiation between HRS and intra-renal ARF is based on the finding of abnormal urine analysis, significant proteinuria and abnormal ultrasound examination in patients with intra-renal ARF. Nevertheless, this process does not take into account two potential confounding clinical situations: (a) HRS may develop in a patient with a parenchymal CKD and (b) the features of HRS may evolve in time. These two situations can undermine some differential criteria, such as those based on urine analysis, degree of proteinuria or on renal ultrasound examination. In addition, recent evidence suggests that even the definition of HRS as a functional renal dysfunction should be critically revised. The first line of evidence is based on a French study which is based on the results of renal biopsies in patients with cirrhosis who developed an AKI defined by the new diagnostic criteria on a CKD . As the new criteria are not helpful in differentiating HRS from the other phenotypes of AFR, one may wonder what this has to do with the diagnosis of HRS. Nevertheless, it has to be recognized that patients with normal urine analysis and without significant proteinuria, could be considered as having a rapidly progressive HRS, namely type 1 HRS, developing on a previous stable HRS that was a type 2 HRS. Now, the results of kidney biopsies in these patients often evidenced signs of both vascular and tubular-interstitial damage . The second line of evidence is based on some preliminary results from the use of new biomarkers of tubular damage, such as the urinary excretion of neutrophil gelatinase–associated lipocalin (NGAL) in patients with cirrhosis and type 1 HRS. The results of a recent study showed that the mean urinary excretion of NGAL in patients with cirrhosis and type 1 HRS was more than double compared with that of patients with cirrhosis and normal renal function . In addition, several patients with type 1 HRS had values of urinary excretion of NGAL higher than the limit above which nephrologists considered an AKI to be an intra-renal ARF . On the basis of these results the functional nature of HRS is now under debate, with relevant implications for its treatment to be discussed subsequently.
Pathophysiology of hepatorenal syndrome
Renal vasoconstriction, which is thought to be the main factor in the pathophysiology of HRS , develops in a context of a marked reduction of effective circulating volume related to peripheral arterial vasodilation. The involvement of endogenous vasoconstrictor systems induced by the reduction of effective circulating volume is documented both at the clinical and experimental level . More recent advances in the pathogenesis of HRS have focused mainly on a new aspect, namely on cardiac dysfunction. In fact evidence has been provided that in patients with HRS cardiac ouput (CO) becomes relatively insufficient [3, 35, 36] to compensate for the severe reduction of effective circulating volume because of the splanchnic arterial vasodilation in patients with HRS. The reasons why CO may decrease in end stage liver disease still remains unknown. However, in recent years several specific cardiac abnormalities such as reduced systolic and diastolic responses to stress stimuli, electrophysiological repolarization changes and enlargement of cardiac chambers have been described and associated with the so-called ‘cirrhotic cardiomyopathy’ . In addition, the release of endotoxins, inflammatory citokynes, nitric oxide, carbon monoxide and other biologically active substances as a result of an overt bacterial infection  or of bacterial translocation  can further impair cardiovascular function in patients [35, 40], as well as in experimental models, with end stage liver disease .
Despite recent advances in the knowledge of the pathophysiology of HRS, there is still a long way to go. Many factors can impair GFR beyond the control of renal blood flow, such as changes in intrarenal haemodynamics, the degree of intracapsular pressure and/or endothelial dysfunction. To focus on the potential role of endothelial dysfunction, the results obtained in an experimental model of ARF, induced by the administration of lipopolysaccharide (LPS), can be considered. Namely, agents that counteract either the oxidative stress  or the biological effect of tumour necrosis factor α  are able to prevent the fall in GFR induced by LPS without modifying, or modifying slightly, mean arterial pressure (MAP) or renal blood flow.
Treatment of hepatorenal syndrome
For the purpose of this review only the most widely used pharmacological option, namely terlipressin plus albumin will be considered. After the demonstration of its effectiveness in recovering renal function in patients with cirrhosis and type 1 HRS [44, 45], the management of these patients has changed dramatically. Despite its modest effects on overall survival , this therapeutic option has probably increased the number of patients undergoing LT  and improved also its outcome [48, 49]. However, if one looks at the facts it can be stated that: (a) terlipressin and albumin are effective in less than 50% of the patients with type 1 HRS; (b) it improves survival slightly and (c) it cannot be used in all patients with type 1 HRS because of contraindications and d) it have to be discontinued in up to 25% of the patients because of adverse effects. Therefore, the limits of this therapeutic option can be divided into inherent and extrinsic limits.
Inherent limitations derive from our insufficient knowledge on the nature and the pathophysiology of HRS [8, 9]. As previously discussed there is evidence that type 1 HRS is associated with some degree of parenchymal damage  and, in particular, tubular damage. Tubular damage seems to be more severe in patients who do not respond to terlipressin and albumin . Moreover, there are some data showing that the extent of tubular damage in non-responders can be progressively worsen . As a consequence, it may be hypothesized that an underlying tubular damage may hinder the response to terlipressin and albumin and that type 1 HRS can progress to ATN in non-responders with very important clinical implications if they are candidates to LT. Further studies are needed to confirm these issues.
It should be outlined that the rationale of terlipressin administration is to counteract the splanchnic arterial vasodilation to increase marterial pressure and overall the effective circulating volume . However, as previously discussed, the reduction of the effective circulating volume in decompensated cirrhosis is not only related to the splanchnic arterial vasodilation but also to an inadequate CO. This adresses the question of the effect of terlipressin and albumin on CO in patients with cirrhosis and type 1 HRS. This has never be tested, but it can be observed that terlipressin per se does not change stroke volume , meaning that it could even reduce CO as it reduces the cardiac rate and/or reduces left ventricular wall motion by increasing after-load and end-diastolic volume . Meanwhile, it is worth observing that albumin can improve CO and that, according to some data , this effect is not only related to the expansion of blood volume and to the consequent increase in cardiac preload, but also to an improvement in cardiac contractility . Moreover, there are preliminary data showing that bacterial translocation may reduce the effect of terlipressin on systemic vascular resistance . All these factors may contribute to explaining why terlipressin and albumin are effective in less than 50% of the patients with type 1 HRS. In addition, the observation of a blunted haemodynamic effect of terlipressin in the presence of endotoxaemia raises the question as to what would be the effect of the drug in the presence of an ongoing bacterial infection. Indeed, the data on the efficacy of terlipressin and albumin were obtained in patients with type 1 HRS, diagnosed on the basis of the older diagnostic criteria , and therefore excluding patients with an ongoing bacterial infection. On the other hand, the problem is complicated even further and made more interesting considering that: (a) the terlipressin is able to reduce the expression of inducible nitric oxide synthesis during sepsis in the arterial wall of rats with cirrhosis  and (b) albumin is, in itself, effective in inhibiting the release of nitric oxide in patients with cirrhosis and SBP .
As far as extrinsic limitations are concerned, one should consider those related to how terlipressin and albumin are being currently administered in the treatment of type 1 HRS. Among the recently identified factors predicting the lack of response to treatment with terlipressin and albumin in patients with cirrhosis and type 1 HRS, there are among them a high value of SCr and serum bilirubin before the start of treatment and a modest increase in MAP as a result of the treatment [58, 59]. Focusing on the first factor, it can be observed that the probability of response decreases from 50 to 30% and then to 10% for SCr values, increasing from less than 3 mg/dl (266 μmol(l) to 3–5 mg/dl (266–444 μmol/L) and then to more than 5 mg (444 μmol/L) respectively . These findings emphasize the recommandation that the treatment of type 1 HRS should be started as soon as possible. For this to happen, the diagnosis of type 1 HRS should be performed within a very short-time frame. Nevertheless, the current cut-off value for SCr for the diagnosis of type 1 HRS is very close to 3 mg/dl (266 μmol/L). Moreover, in accordance with current guidelines [8, 9, 60], a period of 48 h should elapse between the finding of a creatinine >2.5 mg/dl (222 μmol/L) and the assessment of the lack of response to the discontinuation of diuretics and volume expansion by albumin before a diagnosis of type 1 HRS is made. During this time, SCr can increase further beyond 3 mg/dl (266 μmol/L). This can explain why, the answer to the question ‘Do you think that a rigid cut-off value of serum creatinine of >2.5 mg/dl (222 μmol/L) may limit the prompt management of patients with type 1 HRS?', by the majority of the members of the International Ascites Club was affirmative (87.5 yes vs 12.5% no). To change the current diagnostic criteria of type 1 HRS, randomized controlled clinical studies are needed to prove that by reducing the cut-off for Scr of 2.5 mg/dl (222 μmol/L) it would be possible to increase the rate of response to terlipressin and albumin. Nevertheless, to prevent clinical practice from going beyond what is evidence-based and stated in the guidelines, an important recommendation needs to be translated into clinical practice: the withdrawal of diuretics and the plasma volume expansion with albumin should be made very early without waiting for SCr to reach 2.5 mg/dl (222 μmol/L).
As reported above a limited increase in MAP (less that 5 mm Hg at day 3 of treatment) lowers the probability of response to terlipressin and albumin . The potential fallout in translating this issue into clinical practice is that future studies, shoud be focused on the increase in MAP, rather than the per cent decrease in SCr level as the goal to be achieved by means of treatment, as it was recently proposed . Finally, the currently recommended schedule for the use of terlipressin in the treatment of type 1 HRS, implies the use of iv boluses, which are given at an initial dose of 1 g every 4 or 6 h. However, this schedule does not appear to be the best way to maximize the effect of terlipressin. In a pharmacodynamic study performed in patients with cirrhosis the effect of terlipressin on portal pressure has been shown to last less than 4 h . Consequently, it will be interesting to evaluate the efficacy of terlipressin given by continuous infusion instead of by iv bolus. Studies in experimental animals and patients with septic shock suggest that the beneficial hemodynamic effects of terlipressin are enhanced when used as continuous infusion compared with intermittent administration . A preliminary report of a randomized study suggests that this might also be the case in type 1 HRS [64, 65], but more definitive results are needed.
Most members of the ICA agreed that a possible treatment for non-responders to terlipressin and albumin should be based on a combination of different drugs, or combining drugs and non-pharmacological supports. As far as the second strategy is concerned, the combination between the pharmacological approach and transjugular intrahepatic portosystemic shunt (TIPS), despite it appears to be promising in pathophysiological terms , would likely be hampered in clinical practice because TIPS is contraindicated in a high percentage of patients with type 1 HRS . On the other hand, a recent study showed that the combination of terlipressin and albumin and an extracorporeal liver support system using fractionated plasma separation and adsorption may improve survival in patients with type 1 HRS . However, as this observation was made using a subanalysis of a larger study, more specific studies assessing this strategy in a specific population of patients with type 1 HRS should be performed.