Management of liver failure in general intensive care unit

OBJECTIVE
To produce French guidelines on Management of Liver failure in general Intensive Care Unit (ICU).


Introduction
Liver failure is a condition that can be managed or may occur in general critical care under two very different circumstances: on the one hand, acute liver failure (ALF) corresponding to a sudden deterioration in liver function due to toxic or infectious agents occurring in the absence of underlying liver disease. Although ALF is uncommon, it is mandatory to recognise it at an early stage so that a first-line diagnostic and therapeutic strategy can be adopted, and expert advice sought. On the other hand, the hospitalisation in critical care of cirrhotic patients is common, either for the complications of cirrhosis itself or for medical events likely to decompensate preexisting liver disease.
The French Society of Anaesthesia and Critical Care (Socié té française d'anesthé sie et de ré animation, SFAR) and the French Association for the Study of the Liver (Association française pour l'é tude du foie, AFEF) have joined forces to generate original guidelines dedicated to the management of liver failure in critical care setting. These guidelines are intended for a wide audience of intensivists. The aim is not to detail the specific measures adopted in specialised or dedicated liver ICU to manage these patients, but rather to specify the first-line diagnostic and therapeutic management that should be initiated in any critical care unit. The recommendations also focus on specifying the circumstances where expert advice from dedicated centres, most often backed by liver transplant units, is required. Twenty French-speaking experts were thus selected by an organising committee that had been appointed by the guideline committees approved by the executive boards of the two scientific societies. These experts were asked to produce guidelines covering two specific areas: acute liver failure and cirrhotic disease. Two bibliographic experts also used predefined key words to review all the literature in the field published during the past twenty years.

Aim of the guidelines
The aim of these guidelines is to provide a decision-making framework for physicians practicing in a general critical care setting who need to treat patients with suspected liver failure. The group tried to produce a minimum number of recommendations so as to highlight the important features to be retained in the two predefined fields. In situations of doubt, the weight of the literature was considered to be more important than the experts' opinions. The basic rules of universal good medical practice in intensive care were considered to be known and were excluded from the recommendations. The target audience is large, focusing on all professional intensivists with the exception of those working in structures dedicated to liver diseases, often backed by liver transplant centres.

Definitions
Two fields were defined: Acute Liver Failure (ALF) corresponding to a sudden deterioration in liver function due to the appearance of toxic or infectious agents in the absence of underlying liver disease; liver failure in patients with underlying chronic liver disease or cirrhosis, frequently associated with Acute on Chronic Liver Failure (ACLF).

Methods
These guidelines were drawn up by a group of experts acting on behalf of the Association française pour l'é tude du foie (AFEF) and the Socié té française d'anesthé sie et de ré animation (SFAR). The organising committee defined a list of questions to be addressed and designated experts to be responsible for each question. The questions were formulated using the Patient Intervention Comparison Outcome (PICO) model.
The Grade Method (Grade of Recommendation Assessment, Development and Evaluation) was used to compile these guidelines. Following a quantitative analysis of the literature, the method can be used to separately determine the quality of the evidence available, i.e. to estimate the level of confidence required to analyse the effects of the quantitative intervention, and the level of recommendation. The quality of evidence is rated as follows: high quality of evidence: further research is very unlikely to affect confidence in the estimate of the effect; moderate quality of evidence: further research is likely to have an impact on confidence in the estimate of the effect and could change this estimate of the effect; low quality of evidence: further research is very likely to have an impact on confidence in the estimate of the effect and is likely to change this estimate of the effect; very low quality of evidence: any estimate of the effect is very unlikely.
The level of recommendation is binary (either positive or negative) and strong or weak: strong recommendation: we recommend (grade 1 + ) or do not recommend (grade 1 À ) this action; weak recommendation: we suggest (grade 2 + ) or do not suggest (grade 2 À ) this action.
The strength of the recommendations was determined according to key factors and validated by the experts after a vote using the Grade Grid method [1].
The compilation of a guideline required that at least 50% of voting participants had an opinion and that fewer than 20% of participants voted for the opposite proposal. The compilation of a strong agreement required the approval of at least 70% of the voting participants.

Results
Nineteen experts and four coordinators agreed to address questions concerning the treatment of liver failure and its practical implementation in the general intensive care setting in two specific contexts: acute liver failure (ALF) defined as the development of hepatocellular dysfunction in patients without preexisting liver disease; liver failure in patients with underlying chronic liver disease or cirrhosis, frequently associated with acute on chronic liver failure (ACLF).
ACLF is defined as a clinical syndrome of sudden hepatic decompensation observed in patients with chronic liver disease and associated with the failure of one or more extrahepatic organs.
The experts summarised the work and applied the Grade methods, which resulted in 18 recommendations and three management algorithms. Six of these recommendations were strong (grade 1), six were weak (grade 2) and six were expert opinions. No recommendation was formulated in response to one question. After two rounds of scoring and various amendments, a strong agreement from all voting participants was obtained for all recommendations (100%).

First area. Acute liver failure
Acute liver failure (ALF) is a rare disease (fewer than 10 cases per million persons per year in the developed world), characterised by rapidly progressive liver dysfunction associated with a fall in prothrombin time (PT) ratio levels observed in patients without any preexisting liver disease and in less than 26 weeks.
Severe ALF defines a syndrome characterised by a PT ratio less than 50%.
Serious ALF defines a syndrome characterised by a PT ratio less than 50% in combination with encephalopathy.
There are also several definitions based on the time elapsing between jaundice and the development of encephalopathy. Fulminant hepatitis refers to the occurrence of encephalopathy within 15 days of the onset of jaundice, although the original term ''fulminant hepatitis'' is classically used to describe severe ALF. The terms ''severe'' or ''serious'' ALF are used in the remainder of this text. The prognosis of ALF has gradually improved in recent years, with a 2-year survival rate of about 90% in the event of transplantation and of 90% after severe ALF due to paracetamol without transplantation [2].
An algorithm (Fig. 1) and a table (Table 1) summarising ALF management are proposed.
6. Question 1: In patients with acute liver failure, which etiological exams should be performed to reduce morbidity and mortality?
R1 -In patients with severe acute liver failure, we recommend the determination of serum acetaminophen levels, serology for Hepatitis A (IgM VHA) and Hepatitis B (HBsAg and anti-HBc IgM) viruses, urinary toxins (amphetamine, cocaine), and the performance of an echocardiography and hepatic echo-Doppler.

Rationale
Among the most common causes of ALF, acetaminopheninduced hepatotoxicity (whether intentional or not) is the most common cause of severe ALF and an indication for emergency liver transplantation (22% in France) [3,4]. ALF may also be caused by Hepatitis A and B infections (14.6%), other drugs (antimicrobial and antiepileptic agents, statins) (9.4%) or other various toxic agents (herbal supplements, cocaine, ecstasy or the ingestion of mushrooms) (4.8%).
Less frequently, others causes of ALF may be identified such as autoimmune hepatitis, Wilson's disease, metabolic disease, etc. These account for 28.8% of patients on the liver transplant waiting list. Noteworthy, the cause of ALF remains unknown in 25% of patients, despite intensive investigation. Clinical assessment and extensive search in patient's medical history are important elements to determine the cause of ALF.
Early recognition of the reason for ALF is crucial to guide potential specific management and predict its outcome [3]. The transplant-free survival of patients with ALF related to acetaminophen, hepatitis A, hypoxic hepatitis or pregnancy is 50%. Conversely, when ALF is associated with hepatitis B, drugs other than acetaminophen, autoimmune hepatitis, Wilson's disease or Budd-Chiari syndrome, the transplant-free survival rate is lower than 25%.
Abdominal Doppler ultrasound should be performed rapidly to exclude chronic liver disease suggested by ascites, hepatomegaly, dysmorphic liver or malignant infiltration of the liver, and to verify the permeability of vessels (hepatic veins, portal vein). Echocardiography should also be performed when acute ischemic hepatocellular injury is suspected [critically ill patients with primary cardiac or circulatory failure, elderly individuals, underlying history of heart disease or arrhythmia, patients with acute renal failure, serum level of aspartate transaminase (AST) exceeding those of alanine aminotransferase (ALT)] [5].
7. Question 2: In patients with acute liver failure, which specific treatments should be initiated rapidly to reduce morbidity and mortality?
R2.1 -In patients with acetaminophen-induced acute liver failure, we recommend the initiation of N-acetylcysteine therapy without waiting for, and regardless of, the results of serum acetaminophen determinations.
(GRADE 1+) STRONG AGREEMENT R2.2 -In patients with acute liver failure whatever the aetiology, we suggest the initiation of N-acetylcysteine therapy to improve morbidity and mortality.
It has been suggested that the therapeutic use of acetylcysteine may also benefit patients with non-acetaminophen-related ALF. A recent meta-analysis analysed the results of four prospective clinical trials regarding the safety and efficacy of acetylcysteine in patients (both adults and children) with ALF not related to acetaminophen poisoning [9]. A total of 331 patients received treatment with acetylcysteine as compared to 285 patients in control group [10][11][12]. No statistical difference was seen between the two groups in terms of overall mortality. However, liver transplant-free survival (41% versus 30%, OR = 1.61, 95% CI 1.11-2.34, P = 0.01) and post-transplant survival (85.7% versus 71.4%, OR = 2.44, 95% CI 1.11-5.37, P = 0.03) were significantly greater in treated patients. We performed a new meta-analysis including adult patients only and two recent studies [13,14] [10,15], which displayed improvements in overall survival (four studies with 499 patients: 76% versus 59%, OR = 2.30, 95% CI 1.54-3.45 P <0.0001) and liver transplant-free survival (three studies with 419 patients: 64% versus 26%, OR = 4.81, 95% CI 3.22-7.18, P < 0.0001). However, because of methodological bias, we have produced a lower grade (GRADE 2) recommendation for the use of acetylcysteine in non-acetaminophen-related ALF.
In the prospective double-blind trial performed by Lee et al. [10], the beneficial effects of acetylcysteine on transplant-free survival were confined to patients with grades I-II coma. This suggests that the interval between drug ingestion and treatment with acetylcysteine is closely related to the outcome, so that acetylcysteine should be started as early as possible. Acetylcysteine has complex antioxidant and immunologic effects whose mechanisms of action are not completely understood.
R2.3 -In patients with acute liver failure, whatever the aetiology, the experts suggest that advice should be obtained from a liver transplantation centre in order to discuss: 1. Antinuclear antibodies, antimitochondrial antibodies, anti-LKM antibodies, smooth muscle antibodies (autoimmune hepatitis); anti-HEV IgM (hepatitis E virus); Anti-HSV IgM (herpes simplex virus types 1 and 2); serum and urinary copper, serum caeruloplasmin, (Wilson's disease). Pregnancy related ALF: there are two hepatic emergencies which can occur during the third trimester of pregnancy: HELLP syndrome and acute fatty liver of pregnancy.

Liver transplantation
Severe ALF defines a syndrome characterised by a PT ratio < 50%. There should be a contact with a liver transplantation centre for each patient with severe ALF allowing discussion about diagnostic and consideration for transfer in a dedicated centre.
Hepatic encephalopathy (HE) is of key prognostic importance in ALF. Neurological symptoms are associated with overall survival (90.1% in ALF patients without HE versus 37.8% in ALF patients with HE; P < 0.0001) [16]. HE grades are associated with outcome. Short-term transplant-free survival (three weeks) has been shown to vary considerably, from 52% in patients with grade 1-2 HE to 33% in those with grade 3-4 HE [3]. The development of grade 3-4 HE is associated with brain oedema and intracranial hypertension in 38% to 81% of patients [17,18]. HE is the key indicator for LT. This criterion is used to select candidates for LT (King's College criteria in ALF patients related or not to paracetamol; Clichy-Villejuif criteria) [4,19]. These criteria are prognostic indicators of transplant-free survival and are used in the selection of patients for LT. The historical criteria are those developed by King's College Hospital: in acetaminophen-induced liver failure a poor prognosis is correlated with a pH lower than 7.3, prothrombin time longer than 100 s, a creatinine level higher than 300 mol/l and HE more severe than grade 3. Furthermore, blood lactate levels higher than 3.5 mmol/L after 4 hours, or 3.0 mmol/L after 12 hours of management and the early restoration of intravascular volume, are indicative of a poor prognosis. Recent studies have confirmed the clinically acceptable specificity but more limited sensitivity of these systems (50% to 60%) [20][21][22][23]. In order to address these limitations, a wide variety of alternative prognostic systems and markers have been proposed to replace or supplement existing criteria (for example, to identify a patient for transfer to an LT centre). These include the use of factor V levels less than 20% [4], or the combination of criteria: vasopressors need in a context of organ failure [20,24], the routine laboratory measurements including those such as ammonia (with a threshold of 100 mmol/L) [17,18,25,26] or bilirubin (threshold ranging from 140 mmol/L to 200 mmol/L [4,27]), or composite laboratory determinations such as the MELD score [22,27,28].
There is a close relationship between elevated arterial ammonia levels and the development of encephalopathy, there being a greater risk of intracranial hypertension when ammonia levels are sustained between 150 and 200 mmol/L [17,18,26].
All these clinical and biological criteria are correlated to the severity of liver failure but they nevertheless have an independent prognostic value [4,17,22,[26][27][28].

Question 3:
In patients with acute liver failure, which symptomatic treatment should be initiated to reduce morbidity and mortality?
R3 -In order to reduce morbidity and mortality in patients with acute liver failure, the experts suggest that extrahepatic organ failure should be treated early and any aggravating factors should be prevented, as shown in the Table below. EXPERT OPINION, STRONG AGREEMENT

Rationale
Patients with ALF develop hypotension with systemic vasodilation and volume depletion that reflect the severity of the underlying liver failure [29]. Most patients with ALF will have a hyperdynamic circulation. Haemodynamic assessment is necessary to determine volume depletion and both right and left-sided cardiac function, because some patients can develop right or left cardiac dysfunction [1]. There is no specific literature to guide the type of fluids to use or the choice of vasopressor agents [30,31].
There is an evidence of adrenal dysfunction in 50% of patients with ALF, which is at least relative [32,33]. Only one retrospective study including 40 patients, reported that the use of hydrocortisone (300 mg per day) could reduce the need for vasopressors [34]. In the ICU population, the diagnosis and management of critical illness-related corticosteroid insufficiency remains a matter of debate [35][36][37].
Tracheal intubation is usually indicated when Glasgow coma score is less than 8. Mechanical ventilation settings should be protective, as stipulated in guidelines from specialist critical care societies [38]. There have not been any specific studies on ALF populations [38,39]. High levels of PEEP (> 10 cmH 2 O) could be associated with a potential risk of hepatic congestion [40][41][42].
No randomised controlled trials (RCT) have evaluated sedation practices in patients with ALF. In the ICU population, studies have shown that protocol-based sedation (to minimise its depth and duration) seems to reduce overall morbidity and mortality. The use of benzodiazepines should be avoided [43,44]. In patients with acute or chronic encephalopathy, a meta-analysis (8 RCTs; n = 736 patients) showed that flumazenil lowered the encephalopathy score, also suggesting a deleterious effect of benzodiazepines in this population [45]. Dexmedetomidine should be used with caution, as its metabolism is exclusively hepatic [46]. No studies have recommended or not the systematic use of sedation in patients with an altered Glasgow Coma score without intracranial hypertension (ICH), a classically described complication of HE that affects 20% of patients with ALF [16]. Regular monitoring must be ensured of patients with high-grade encephalopathy (grades 3 and 4). Two multicentre observational studies in patients with severe encephalopathy did not demonstrate a significant difference in 1-month mortality, whether the patients were monitored using ICP devices or not (pooled data: RR = 0.79, 95% CI 0.61-1.02) [47,48]; ICP devices have been associated with haemorrhagic complications (7% to 20% of cases in patients with ALF) [47][48][49]. Transcranial Doppler ultrasound is a useful monitoring tool that could be used first-line in this context. There are no specific therapeutic options for raised intracranial pressure in patients with ALF [50][51][52][53][54][55].

Coagulation
A recent multicentre observational study including 1770 patients with ALF reported bleeding complications in only 187 patients (10%), which included 22 (1.5%) post-procedural bleeding episodes [56]. Eighty-four per cent of spontaneous bleeding episodes originated from the upper gastrointestinal tract. An in-depth analysis of coagulation abnormalities (thrombocytopenia and prolongation of the INR) in ALF patients suggested that most patients had rebalanced haemostasis between pro-and anticoagulant factors [57]. Prophylactic administration of coagulation factors precludes the assessment of the natural evolution of the disease. There is no support for the use of coagulation factors and this should be limited to active bleeding or invasive procedures with a high risk of complications.

Renal replacement therapy
No RCT have been performed specifically to evaluate the strategy and timing of its initiation in patients with ALF [16]. Regional citrate anticoagulation should be monitored because of its potential metabolic effects in patients with ALF [58].

Liver support devices
Two well-designed RCT including 115 patients with ALF (not related to hypoxic hepatitis) failed to demonstrate a significant reduction in mortality (pooled data: RR = 0.82; 95% CI 0.42-1.59) [59,60]. The place of liver support systems in the management of patients with ALF need to be better defined. In any case, these techniques should not delay transfer to a liver transplantation centre.
ALF is frequently associated with electrolyte and metabolic disturbances, particularly in patients with hyperacute ALF or when it is associated with acute kidney injury [61]. Hypoglycaemia is a well-known complication of severe ALF [2,61]. Its clinical features can be confused with those of hepatic encephalopathy. Blood glucose parameters should therefore be monitored at least every 2 hours. No studies have evaluated the optimal target for blood glucose levels. Hyponatremia (sodium < 130 mmol/L) is common in patients with ALF [62]. There is a correlation between hyponatremia and intracranial pressure. It has been shown that the infusion of hypertonic saline to maintain serum sodium levels at between 145 and 155 mmol/L significantly decreased ICP [54]. However, a serum sodium level above 150 mmol/L is deleterious and should be avoided. In practice, we recommend targeting the sodium level at between 140 and 145 mmol/L. Any corrections should not exceed 10 mmol/L per 24 h [63]. Electrolyte disturbances, such as serum phosphate, are commonly observed in patients with ALF, and should be monitored and corrected.
Patients with ALF and organ failure have increased energy expenditure similar to that of other ICU patients. There are no specific nutritional guidelines in patients with ALF.
The use of osmotic laxatives (lactulose) or non-absorbable antibiotics (rifaximin) to lower ammonia levels is not recommended [61].
Although there are no data to support its use, stress ulcer prophylaxis is usually recommended in this at-risk population [61].
Patients with ALF have increased susceptibility to infections. Bacterial infections have been documented in 60%-80% of patients with ALF, and fungal infections occur in one third of patients [64]. Empirical broad-spectrum antibiotics should be administrated to ALF patients if there are signs of sepsis and/or of worsening encephalopathy [65]. These broad-spectrum antibiotics should cover common organisms such as enterobacteria, staphylococcal or streptococcal species, or as a function of the ecology in the unit [64].

Second area. Cirrhotic ICU patients and acute-on-chronic liver failure
Cirrhosis is an end-stage form of chronic liver disease. The prevalence of cirrhosis is estimated at between 2000 and 3300 people per million inhabitants. There are approximately 700,000 patients with cirrhosis in France, 10,000 to 15,000 of whom die every year. A wide range of diseases and conditions can damage the liver and lead to cirrhosis. Ninety per cent of cirrhosis cases are related to chronic alcohol abuse, chronic viral hepatitis (hepatitis B and C) and fatty liver disease in a context of metabolic syndrome. In terms of its pathophysiology, cirrhosis consists in the gradual replacement of healthy liver tissue by annular fibrosis that is responsible for destroying the architecture of the liver parenchyma and blocking intrahepatic portal blood flow. These architectural changes are associated with life-threatening complications related to portal hypertension and/or hepatic dysfunction that may require ICU hospitalisation. These complications may be either specific to liver disease (ascites, spontaneous bacterial peritonitis, variceal haemorrhage, hepatic encephalopathy or hepatorenal syndrome) or non-specific (infection, acute kidney failure). In addition, given the prevalence of this disease, it is frequent that the reason for initial ICU admission is not a complication of cirrhosis but rather another condition likely to lead to organ failure such as pneumonia-related septic shock, community-acquired peritonitis, stroke, trauma, etc. Under these circumstances, even if cirrhosis was compensated or even often not diagnosed at admission, the precipitating event is likely to decompensate the liver disease.
Cirrhotic patients requiring ICU hospitalisation have a high rate of in-hospital mortality (ranging from 30% to 50%). The severity of cirrhosis was initially established by the Child-Turcotte score [66] subsequently modified by Pugh [67]. A new entity named ACLF has recently been described [68] as a combination of acute decompensation of cirrhosis, one or several extrahepatic organ failures and high short-term mortality [68]. ACLF is a dynamic syndrome and can occur in patients with or without a prior history of acute decompensation. The diagnosis of ACLF and its classification in four grades of severity based on the number of organs failing is made using the CLIF SOFA score Supplementary Tables 1 and 2, [68][69][70]. It should also be noted that an ACLF grade based on organ failure has been shown to better predict the outcomes of ICU cirrhotic patients than standard prognostic methods such as the MELD or Child-Pugh scores.
These guidelines will focus on the management of cirrhosis and its complications requiring ICU hospitalisation, and provide recommendations based on the best evidence available in the literature.

Question 4:
In patients with cirrhosis, which criteria should be used to guide admission or non-admission to an ICU to improve their prognosis?
R4 -We do not suggest denying the admission of patients with cirrhosis to the ICU solely because of their underlying cirrhotic condition.

Rationale
The proportion of cirrhotic patients admitted to ICUs is increasing [71][72][73][74]. However, to date, there have been no objective and specific criteria to guide the admission of these patients to critical care.
The recent definition of a new clinical entity, Acute-on-Chronic Liver Failure (ACLF), as an acute decompensation of cirrhosis associated with organ failure(s) defined using the CLIF-SOFA score could be useful to ensure the early identification of patients who could benefit from ICU admission. Indeed, ACLF is associated with a 28-day mortality rate of at least 15% [68].
The number of organ failures at admission to the ICU is also associated with short-term mortality [69,70,74,77,82,112,113]. A recent and large multicentre study conducted on 17,044 cirrhotic patients in intensive care reported a hospital mortality rate lower than 50% in patients with three or more organ failures, thus delegitimising decisions not to admit a cirrhotic patient to the ICU based on liver disease alone [113].
In addition, the reason for ICU admission also influences prognosis, with digestive haemorrhage classically being associated with a better prognosis and septic shock with a poorer one [68,70,105,114]. Finally, among cirrhotic patients admitted to an ICU, those who have been directly admitted have a better prognosis than those who were initially admitted to the Hepatology Unit, thus highlighting the value of early admission to ICU [68,69,72,115].
As with non-cirrhotic patients, the prognosis for cirrhotic patients at ICU admission is therefore very largely associated with the presence of organ failure(s) graded using different scores. However, none of these scores has sufficient predictive power to guide individual patient prognosis. There is therefore no reliable indicator of the futility of resuscitation in cirrhotic patients, regardless of the severity of cirrhosis or number of organ failures [116].
Finally, several studies have shown that the ICU prognosis of cirrhotic patients has improved in recent years (period effect independently associated with mortality), [72,89,100,108,113,114,117,118] whether the reason for admission was variceal bleeding [119] or septic shock [74,120]. The maintenance of life-sustaining therapy during the ICU stay is considered in question 10.
10. Question 5: In patients with cirrhosis hospitalised in an ICU, which specific measures to manage acute kidney injury should be used to reduce morbidity and mortality?
R5.1 -To define and assess the severity of acute kidney injury (AKI) in patients with cirrhosis, the experts suggest: 1. Using the modified KDIGO criteria specifically refined by the International Club of Ascites 2. Managing AKI according to its severity and to the algorithm proposed (Fig. 2) 3. Not contraindicating renal replacement therapy in these patients just because of their underlying cirrhotic condition.
EXPERT OPINION, STRONG AGREEMENT

Rationale
In non-cirrhotic patients, the reference for diagnosing and assessing the severity of acute kidney injury (AKI) is the KDIGO classification [121,122]. The limitation of this classification in patients with cirrhosis is the use of urinary flow. Indeed, in advanced stages, cirrhosis is often associated with oliguria and water and salt retention, oedema and ascites, while renal function remains normal. Conversely, the use of diuretics to treat ascites and oedema can lead to an abnormally high urine output with impaired renal function. Experts in the International Ascites Club have therefore proposed using the KDIGO classification to diagnose and assess the severity of AKI, based on the same criteria as for serum creatinine without the urine output criterion (Table 2) [122]. Several studies have shown that this modified classification has a good prognostic value with early mortality being well correlated with the severity of AKI [123][124][125][126][127]. It is important to note that the definition is based on the dynamic change in serum creatinine from baseline and no longer on a fixed threshold for serum creatinine. Indeed, as a marker of renal function, serum creatinine has many limitations in patients with cirrhosis, including: a decrease in muscle creatine, a precursor of creatinine, related to frequent sarcopenia in these patients [128]; an increase in the distribution volume related to ascites and oedema; interference with creatinine measurement techniques related to elevated serum bilirubin levels [129,130].
Use of the modified KDIGO classification (Table 2) is important because it enables the management of acute renal failure to be adapted as a function of its severity (Fig. 2). So, although the eviction of risk factors may be sufficient at an early stage, albumin fluid therapy is necessary in the event of progression to stages 2 or 3 [121]. Finally, although the prognosis with renal replacement therapy is very poor in cirrhotic patients with AKI, it is not contraindicated provided that its duration is planned to be short and that it is integrated in a therapeutic plan, such as liver transplantation, or in a context of reversible precipitating events, such as sepsis (Table 3) [131].
R5.2 -In patients with cirrhosis hospitalised in the ICU, we suggest treating hepatorenal syndrome with a vasoconstrictor agent (terlipressin as first-line therapy) and concentrated albumin GRADE 2+, STRONG AGREEMENT

Rationale
Several meta-analyses have suggested a beneficial effect of the combination of a vasoconstrictor and albumin on short-term survival and hepatorenal syndrome (HRS, defined in Table 2) regression [131][132][133].
The objective of vasoconstriction is to counterbalance splanchnic arterial vasodilation in order to improve renal perfusion. The reference treatment is terlipressin. A Cochrane meta-analysis that included nine randomised studies (394 patients) comparing different vasoconstrictors in combination with albumin concluded that terlipressin was the most effective in treating HRS [134]. Most recent studies have suggested that the response rate in terms of an improvement in renal function (complete or partial response) ranges from 64% to 76% [134]. However, these rates are significantly lower in the event of recurrent HRS (20% of cases). Terlipressin is usually administered intravenously (bolus injections) at an initial dose of 0.5 mg to 1 mg every 4 to 6 hours. The dose can then be gradually increased to a maximum of 2 mg every 4 to 6 hours if the serum creatinine level does not fall by more than 25%. Treatment should be maintained until a complete response is obtained or for a maximum of 14 days in the case of a partial response. The continuous administration of terlipressin at the same daily doses (2 to 12 mg/24 h IVSE) appears to be an interesting alternative as it remains as effective as boluses but enables lower daily doses, thereby reducing the risk of adverse events such as cardiac or intestinal ischemia, pulmonary oedema, or distal necrosis [135]. More recently, another meta-analysis pointed out that norepinephrine might be a reliable alternative in patients with a central venous catheter (terlipressin can be administered via a peripheral vein) [133]. Indeed, no difference in the reversibility or relapse of HRS was found between the terlipressin + albumin and norepinephrine + albumin arms, although the numbers of patients included in these studies remain small. Conversely, the combination of midodrine and octreotide is less effective and should not be used [135,136].
The addition of albumin can be discussed insofar as its beneficial effects have only been demonstrated in combination with vasoconstrictors. Indeed, while the administration of albumin might improve systemic haemodynamics by increasing cardiac output and through its antioxidant and anti-inflammatory properties [137], no study has ever compared a strategy combining vasoconstrictors and albumin with another that only used vasoconstrictors. Therefore, the doses usually recommended for albumin (1 g/kg before the initiation of vasoconstrictor treatment Table 2 Modified KDIGO criteria for definition of acute kidney injury in patients with cirrhosis.

Baseline serum creatinine
A value of serum creatinine obtained in the previous three months, when available, can be used as baseline serum creatinine In patients with more than one value within the previous 3 months, the value closest to the admission time to the hospital should be used In patients without a previous serum creatinine value, the value on admission should be used as baseline Definition of acute kidney injury Increase in serum creatinine ! 26.5 mmol/L within 48 hours or A percentage increase serum creatinine ! 50% from baseline, which is known or presumed to have occurred within the prior 7 days Staging of acute kidney injury Stage 1: increase in serum creatinine ! 26.5 mmol/L or an increase serum creatinine ! 1.5-fold to 2-fold from baseline value Stage 2: increase serum creatinine ! 2-fold to 3-fold from baseline value Stage 3: increase serum creatinine ! 3-fold from baseline value or serum creatinine ! 353.6 mmol/L with an acute increase ! 26.5 mmol/L or initiation of renal replacement therapy Table 3 Diagnostic criteria of hepatorenal syndrome (all the criteria must be respected to retain the HRS).

Diagnosis of cirrhosis and ascites
Diagnosis of acute kidney injury stage 2 or 3 according to KDIGO criteria No response after 2 consecutive days of diuretic withdrawal and plasma volume expansion with albumin (1 g/kg of body weight) Absence of shock No recent use of nephrotoxic drugs (NSAIDs, aminoglycosides, iodinated contrast. . .) No macroscopic signs of structural injury defined as Absence of proteinuria (> 500 mg/day) Absence of microhaematuria (> 50 RBCs per high power field) Normal findings on renal ultrasonography NSAIDs: non-steroidal anti-inflammatory drugs; RBCs: red blood cells. and then 20 to 40 g/day) [121] and the duration of treatment remain empirical, as do the haemodynamic objectives that are not defined. Uncertainties regarding the specific effect of albumin, and recent changes to the definition of HRS may modify the interpretation of the results of previous studies and result in downgrading the level of evidence of the recommendation.
11. Question 6: In patients with cirrhosis hospitalised in the ICU, which specific management of sepsis should be used to reduce morbidity and mortality?
R6 -In order to reduce the morbidity and mortality of critically ill patients with cirrhosis, whatever the symptoms and organ failure(s), we suggest: 1. Performing a systematic search for infection that should include microbiological and cytological examination of ascites fluid (a concentration of polymorphonuclear cells > 250/mm 3 in the ascitic fluid will confirm the diagnosis of spontaneous bacterial peritonitis), 2. Initiating early empirical antibiotic therapy that should be tailored to the suspected site of the infection, the causative pathogen once it has been identified, and the local ecology GRADE 2+, STRONG AGREEMENT

Rationale
One third of cirrhotic patients develop an infection at admission or during hospitalisation, which is four to five times higher than the incidence in the general population. Infection is an independent risk factor for mortality, accounting for one third to one half of the causes of death in cirrhotic patients [138,139]. Several retrospective observational studies have shown an association between delayed antibiotic therapy and mortality [140][141][142][143][144]. In a study including 126 cirrhotic patients with spontaneous bacterial peritonitis-related septic shock, the adjusted OR associated with mortality was 1.86 (95% CI 1.10-3.14) per hour of delay (P = 0.02) [142]. Another study of 852 cirrhotic patients with bacteraemia showed that a delay of more than 24 hours between the onset of bacteraemia and the initiation of appropriate antibiotic therapy was associated with a six-fold increase in mortality [OR = 6.06 (90% CI 4.28-8.58; P < 0.0001)] [143]. Antimicrobial treatment should therefore be started early when there is any clinical or biological suspicion of sepsis. Worthy of note is a worsening of encephalopathy should suggest an infection as its triggering factor [145,146]. CRP and PCT serum levels were reported as being reliable biomarkers of infection in cirrhotic patients in a meta-analysis that included 10 studies (n = 1144 patients) [147]. However, it should be remembered that the diagnostic threshold for CRP decreases with the severity of cirrhosis (10 mg/L on average in mild cirrhotic patients, 5 mg/L for Child C patients) [148]. Initial antibiotic therapy should target enterobacteria and Gram+ Cocci, which account for the majority of causal pathogens in this population in whom more than half of infections are spontaneous bacterial peritonitis and urinary tract infections [139]. The local microbiological ecology should also be considered when choosing empirical antibiotic therapy. In a recent meta-analysis (eight studies, 1074 positive ascites cultures) [149], the average resistance rate to 3 rd generation cephalosporins (C3G) was one third in community-acquired SBP and two thirds in healthcare-associated infections.
Concerning haemodynamics, cirrhotic patients are prone to develop hypoxic hepatitis that can worsen liver failure. The latter may be multifactorial [29]: hypovolaemia and vasoplegia, as well as right or left heart failure, pulmonary arterial hypertension associated with portal hypertension, hepatopulmonary syndrome and acute pulmonary oedema associated with volume overload. There is a lack of specific data on the management of sepsis in cirrhotic patients, and the approach is currently based on guidelines for the general ICU population [150]. No specific data on the use of vasopressors are available outside observational studies [151]. A meta-analysis (16 studies, 1507 patients) showed that half of cirrhotic patients with septic shock, had, at least a relative, corticosteroid insufficiency, which was associated with higher rates of sepsis and hospital mortality [152]. An observational study [153] and a double-blind RCT [154] included 150 cirrhotic patients with septic shock in order to assess the impact of 50 mg hydrocortisone every 6 h on mortality (very low level of evidence). There was no significant difference in ICU mortality (pooled data: RR = 0.78; 95% CI 0.58-1.05). However, hydrocortisone was associated with a higher rate of shock recovery (pooled data: RR = 1.49; 95% CI 1.17-1.92; P = 0.001). The study with the highest level of evidence [154] reported a significantly higher rate of gastrointestinal bleeding in the hydrocortisone arm (RR = 3.00; 95% CI 1.08-8.36; P = 0.02). Overall, we are not able to produce a recommendation regarding the use of hydrocortisone in cirrhotic ICU patients; their management needs to be evaluated on a case-by-case basis in a context of refractory shock.

Question 7:
In patients with cirrhosis hospitalised in ICU, when concentrated albumin should be administered to reduce morbidity and mortality?
R7.1 -In cirrhotic patients hospitalised in the ICU, we recommend the use of a concentrated albumin infusion after highvolume paracentesis (more than 4 to 5 litres of ascites fluid removed) GRADE 1+, STRONG AGREEMENT

Rationale
Large volume paracentesis (> 5 litres) associated with plasma volume expansion has been shown to be more effective than diuretic therapy in eliminating ascitic fluid and was associated with a lower incidence of complications [155]. In patients treated with paracentesis without plasma expansion, paracentesis-induced circulatory dysfunction (PICD) may be present in 70% of cases. The diagnosis of PICD is based conventionally on an increase in plasma renin activity of > 50% of the pre-treatment value to above 4 ng/mL/h on the sixth day after paracentesis. PICD is associated with an increased rate of recurrent ascites, the development of hepatorenal syndrome, hyponatremia and reduced survival [156]. A recent meta-analysis assessing 17 randomised studies (1225 patients) provided evidence that albumin infusion (8 g/L of ascites fluid removed) after paracentesis is the most effective therapy in the prevention of PICD, as compared to alternative treatments, and reduced the odds of PICD by 61% (OR = 0.39, 95% CI 0.27-0.55) [157]. The ability of albumin to reduce this complication was also shown in a subgroup analysis versus each of the other volume expanders tested (e.g. dextran, gelatin, hydroxyethyl starch, hypertonic saline). With albumin, the odds of hyponatremia and mortality were reduced by 42% (OR = 0.58, 95% CI 0.39-0.87) and by 36% (OR = 0.64, 95% CI95% 0.41-0.98), respectively [157]. The findings of two other recent randomised studies [158,159] did not alter the conclusions of the meta-analysis [160]. Another recent meta-analysis contested the effect of albumin on mortality [161]; however, this finding was dependent on the inclusion of two studies with unsuitable controls (no treatment in one and mannitol in the other). If these two studies were excluded, the benefit of albumin on mortality remained significant in the whole cohort of patients (OR = 0.58, 95% CI 0.40-0.86) and in the subgroup of patients without hepatocellular carcinoma (OR = 0.60, 95% CI 0.39-0.91). The administration of albumin (6 to 8 g per L ascites fluid removed) is recommended to prevent PICD after large volume paracentesis (> 5 litres) [162,163]. In patients whose paracentesis produces less than 5 L ascites, the risk of developing PICD is low. However, EASL clinical practice guidelines state that these patients should still be treated with albumin because of concerns about the use of alternative plasma expanders such as Dextran.
One controversial issue remains the dose of albumin that should be administered, but only a few studies addressed this question. The results of an unblinded randomised pilot study showed in 70 patients with low severity cirrhosis (mean MELD Score at 16 to 17) that treatment with a half dose of albumin (4 g per L versus 8 g per L ascites fluid removed) was effective and safe in preventing PICD [164]. As the PICD risk is greater when more than 8 litres of ascites fluid are removed, it seems preferable to limit ascites removal of less than 8 litres during a paracentesis procedure [165]. Finally, although there have been no studies, it is recommended that the albumin infusion should be done slowly in order to prevent any potential cardiac overload promoted by preexisting cardiomyopathy.

R7.2 -In patients with cirrhosis hospitalised in the ICU, we
suggest that concentrated albumin infusions should be used in the event of spontaneous bacterial peritonitis (SBP) GRADE 2+, STRONG AGREEMENT

Rationale
Spontaneous bacterial peritonitis (SBP) is an acute bacterial infection of ascitic fluid common in patients with cirrhosis (prevalence: 10% to 30%). Diagnosis requires paracentesis, or sampling of the peritoneal fluid from the peritoneal cavity. When fluid contains large numbers of white blood cells (neutrophils) (> 250 cells/mm 3 ), the infection is confirmed.
Sort et al. showed that treatment with combination of intravenous albumin (1.5 g/kg on day 1 and 1 g/kg on day 3) and antimicrobial therapy could reduce the incidence of renal impairment (10% versus 33%; P = 0.002) and death (22% versus 41%; P = 0.03), as compared to antibiotic alone [166]. Since this publication, numerous studies about fluid resuscitation in septic context affecting cirrhotic patients have been published. Six randomised controlled trials (RCTs) including 577 patients with cirrhosis and SBP (4 studies) or other types of bacterial infection (2 studies), compared patients assigned to receive antibiotic or antibiotics plus albumin (0.14 to 1.5 g/kg on days 1 and 3) (5 studies) or antibiotics plus hydroxyethyl starch (one study) [166][167][168][169][170][171]. None of these studies were double-blinded and included patients hospitalised in ICU. We performed a metaanalysis including additional data [172] showing that when compared to standard antibiotic therapy alone, treatment involving albumin plus antibiotics improved survival at 3 months (OR = 0.66; 95% CI 0.45-0.96, P = 0.03). Albumin has beneficial effects on renal function involving lower incidence of renal impairment (6 RCTs: OR = 0.46, 95% CI 0.30-0.71, P < 0.001). Since none of these studies have been performed in ICU patients, the level of the recommendation is lowered because it is based on indirect evidence.
In addition, only the oldest studies, which included only SBP, showed that albumin infusions significantly improved survival at 3 months (OR = 0.36, 95% CI 0.21-0.61; P = 0.0001) and renal impairment (OR = 0.21, 95% CI 0.11-0.42; P < 0.0001). These data restricted the recommendation to patients with SBP (studies carried out between 1999 and 2009). The benefits of albumin infusions in patients with SBP could be explained by haemodynamic effects after paracentesis, although they may merely be due to these products' expanding properties. Albumin is believed to be effective in patients with SBP because of its ability to improve intravascular volume (altered by paracentesis) and to bind proinflammatory molecules. However, the ideal dose has yet to be determined and an albumin infusion should be associated with haemodynamic monitoring (volume status, cardiac output).
In cirrhotic patients with an infection other than SBP, two RCTs failed to show that albumin infusion improved the survival, although it delayed the onset of renal failure [169,170]. The use of HES should be proscribed in patients with cirrhosis. The administration of HES may favour liver failure, particularly in cirrhosis setting. The rational for these adverse effects is the lysosomal storage of HES in Kupffer cells and hepatocytes.

Rationale
Splanchnic vasoconstrictor agents such as somatostatin, octreotide and terlipressin exert their vasoactive effects and decrease splanchnic blood flow and portal pressure. In patients with acute variceal bleeding who are undergoing endoscopic sclerotherapy, several studies showed that the early intravenous administration of a vasoactive agent was more effective than placebo in the overall control of haemorrhage [173][174][175]. In one meta-analysis published in 2012 [176], the use of vasoactive agents was associated with a significantly lower risk of mortality at 7 days (RR 0.74; 95% CI 0.57-0.95; P = 0.02), fewer transfusion requirements and a shorter duration of hospitalisation. This metaanalysis included 19 studies with several vasoactive medications (octreotide, n = 9; somatostatin, n = 3; terlipressin, n = 4; vasopressin, n = 2; and vapreotide, n = 1). In a multicentre, randomised trial that included 780 patients, the proportion of treatment failure within the first 5 days, was not different with three well-known vasoactive drugs (octreotide, somatostatin or terlipressin) as adjuvant therapy to standard endoscopic treatments [177].
Bacterial infections are a common cause of mortality in patients with cirrhosis and upper gastrointestinal bleeding. A metaanalysis including 12 controlled trials concluded that antibiotic prophylaxis was associated with a significant reduction in overall mortality (RR 0.79, 95% CI 0.63-0.98), in mortality from bacterial infections (RR 0.43, 95% CI 0.19-0.97) [178] and a marked reduction in overall re-bleeding episodes among patients under antibiotic prophylaxis (RR 0.53, 95% CI 0.38-0.74). Trials to evaluate length of hospital stay have shown that patients receiving antibiotic prophylaxis tended to be admitted for shorter periods [178]. In the meta-analysis of these trials, twelve studied antibiotic prophylaxis using quinolones (five trials), quinolones plus betalactams (two trials), cephalosporins (three trials), carbapenems (one trial) and non-absorbable antibiotics (one trial) versus no intervention or a placebo. All the antibiotics tested achieved a beneficial effect on the bacterial infection.
One study found that a high-dose infusion of a proton-pump inhibitor before endoscopy accelerated the resolution of signs of bleeding in ulcers and reduced the need for endoscopic therapy [179]. This reaffirms that optimal acid suppression facilitates clot formation over arteries in bleeding peptic ulcers. Fewer than 5% of the patients included in this study had cirrhosis. However, because the bleeding is not related to portal hypertension in 30% of patients with cirrhosis, the expected benefit of proton-pump inhibitors prior to endoscopy in these patients led the experts to overstate the level of evidence for proton-pump inhibitors in this indication.
Upper gastrointestinal endoscopy enables the examination and also the treatment of most upper gastrointestinal bleeding. A combination of banding ligation and vasoactive therapy for 2 days was superior to the infusion of vasoactive therapy for 5 days alone in reducing very early re-bleeding [180]. Per-endoscopic treatments usually consist of banding ligation of oesophageal varices and glue therapy for gastric varices. Patients with upper gastrointestinal bleeding and features suggestive of cirrhosis should undergo endoscopy within 12 hours of presentation [180][181][182]. This is supported by data from a non-randomised study, which reported less recurrence of haemorrhage and better survival when endoscopy was performed within the first 12 hours [183]. This period also allows for a rapid discussion on the advisability of TIPS in the event of variceal haemorrhage uncontrolled by endoscopy and drug therapy (refractory bleeding).
R8.3 -In patients with cirrhosis and acute upper gastrointestinal bleeding, we suggest to consider Transjugular intrahepatic portosystemic shunt (TIPS) with a covered stent within 24 to 72 hours of the bleeding for episodes in Child-Pugh class C patients (< 14) or in Child-Pugh class B patients with initially active bleeding at endoscopy (early TIPS).

Rationale
A randomised controlled trial evaluated the early use of TIPS in cirrhotic patients in Child-Pugh class C or class B [182]. A total of 63 patients with acute variceal bleeding were included. This study showed that the early use of TIPS was associated with a significant reduction in mortality at 2 years. However, this study suffered from several limitations that deserve consideration: the main endpoint was the rate of failure to control bleeding within 5 days; seventeen patients were treated with sclerotherapy, which is not the recommended endoscopic treatment, and the recruitment delay was long as 359 patients were screened and 63 were included between 2004 and 2007. The two other studies did not have a controlled design (low level of evidence) [187,188]. Another old randomised controlled study specifically analysed early TIPS in this indication, but with an uncoated stent and the selection of high-risk patients, using the hepatic venous pressure gradient [189]. Finally, a meta-analysis assessed the effects of early TIPS on patient prognosis, but it was of poor quality because of large heterogeneity between the types of stents used (coated or not) or the selection criteria for patients [190].
R8.4 -In patients with cirrhosis and acute upper gastrointestinal bleeding, the experts suggest considering emergency Transjugular intrahepatic portosystemic shunt (TIPS) with a covered stent in the event of variceal bleeding refractory to endoscopic treatment (salvage TIPS).
EXPERT OPINION, STRONG AGREEMENT

Rationale
In a study of patients with cirrhosis and variceal haemorrhage uncontrolled by endoscopy and drug therapy, TIPS (salvage TIPS) was able to halt the haemorrhage in 80% of cases [191]. In this setting of active bleeding and a failure of standard medical and endoscopic haemostasis, alternative measures include balloon tamponade (Blakemore or Linton tube) or a self-expandable oesophageal stent. Both these treatments could be recommended as a bridge to definitive therapy (TIPS therapy) in patients with cirrhosis and massive or refractory oesophageal variceal bleeding. There has only been one randomised controlled trial, which included 28 patients to compare self-expandable oesophageal stents and balloon tamponade in the treatment of acute and refractory bleeding from oesophageal varices [192]. There were no significant differences in 15-day survival rates between the oesophageal stent and balloon tamponade groups. The availability of oesophageal stents is limited. Consequently, the experts were not able to make a recommendation about their use.
14. Question 9: In patients with cirrhosis hospitalised in the ICU, which measures to manage haemostasis should be initiated to reduce morbidity and mortality?
R9 -The experts suggest that routine prophylactic fresh frozen plasma, platelets or fibrinogen concentrates should not be systematically administrated before invasive procedures in cirrhotic patients in order to reduce bleeding.
EXPERT OPINION, STRONG AGREEMENT

Rationale
Although conventional coagulation tests are altered in patients with cirrhosis, the balance between pro-and anticoagulant factors is respected and the generation of thrombin is normal [193,194]. Post-procedural bleeding is not predicted by these tests (platelet count or international normalised ratio) [195]. Nor is there consensus regarding the platelet or plasma transfusion thresholds or any evidence of their clinical efficacy [196]. The French Medicines Agency (Agence nationale de sé curité du mé dicament et des produits de santé ) and the European Society of Hepatology (EASL) recommend an infusion of platelets when the count is < 50 G/L but this has not been substantiated by any publication [197,198].
Standard doses of fresh frozen plasma rarely correct coagulopathy in patients with cirrhosis. To achieve that, large amounts of plasma need to be transfused, associated with adverse effects such as increased vascular volume and portal pressure, or nonspecific blood products adverse effects [199].
Although the optimal fibrinogen level is uncertain, the preventive administration of fibrinogen is also a matter of considerable debate. A benefit/risk assessment should be performed on a case-by-case basis, depending on the patient's haemostatic parameters.
Use of a thrombopoietin receptor agonist (eltrombopag) could raise the platelet count and avoid the use of platelet transfusions in cirrhotic patients prior to an invasive procedure. However, other reports have suggested an increased risk of thrombosis. The risk/ benefit ratio is therefore unfavourable [200].
No recommendation: the experts were not able to produce recommendations concerning thromboprophylaxis in critically ill patients with cirrhosis.

Rationale
Despite their low levels of coagulation factors, thrombin formation and clot formation are normal in cirrhotic patients [193,206]. These patients have a frequent imbalance between procoagulant and anticoagulant activity, which frequently results in a hypercoagulable state [207,208]. Several studies have demonstrated an increase in thrombotic events in cirrhotic patients versus a control population [209,210]. Very few studies have examined the value of thromboprophylaxis in patients with cirrhosis; the level of evidence is low and only based on one metaanalysis of non-randomised retrospective series [211]. The heterogeneity of the studies in this meta-analysis did not allow to formally conclude as to the effectiveness of thromboprophylaxis to prevent venous thromboembolism (pooled OR 0.87 95% CI 0.34-2.18) [211]. In the same meta-analysis, thromboprophylaxis was not associated with a higher rate of bleeding in patients with cirrhosis. Other retrospective non-randomised cohort studies confirmed no extra risk of bleeding. Only one retrospective study including 256 cirrhotic patients showed an increase in overall haemorrhages but not an increased risk of major haemorrhage [212,213].
Prophylactic treatment with unfractionated heparin has been shown to be associated with an increased risk of bleeding events [214]. We recommend assessing the individual risk of venous thromboembolism, particularly in patients with cirrhosis who are undergoing surgery. For patients at high risk of thromboembolism and with a high bleeding risk, use of an intermittent pneumatic compression device as prophylaxis for venous thromboembolism may be an alternative [215,216], but it has not been evaluated in patients with cirrhosis.
15. Question 10: In patients with cirrhosis hospitalised in the ICU, when should expert advice be obtained in order to reduce morbidity and mortality?
R10 -The experts suggest that expert advice should be sought regarding patients with cirrhosis who are hospitalised in the ICU: 1. At admission, if the patient is already on the liver transplantation waiting list.
2. Soon after ICU admission to discuss the intensity of care to be provided as a function of the number of organ failures and their course.
3. In order to discuss the benefits of a liver support technique. 4. At discharge from the ICU in order to organise hepatology management, with a view to potential possible liver transplantation EXPERT OPINION, STRONG AGREEMENT

Rationale
1. The onset of complications in patients with cirrhosis who are on the liver transplant waiting list should be reported to the LT centre. Indeed, acute decompensation does not contraindicate liver transplantation and might even favour it by prioritising the patient on the list.
2. The need for ICU admission, whatever the reason, is associated with high mortality rates in patients with cirrhosis (ICU and 6month mortality rates of respectively 42.7% and 75%) [114]. The prognosis for these patients depends on the severity of both hepatic failure (evaluated using the MELD score at ICU discharge) and extrahepatic organ failures (respiratory, neurological, haemodynamic, renal, coagulation, hepatic) [104,114,217]. Depending on the number of organ failures, 28-day mortality rate ranges from 4.7% in patients without organ failure to 76.7% in patients with ACLF grade 3 [68]. Evolution of the SOFA score and/or the number of organ failures during the ICU stay is a crucial prognostic factor. A retrospective study that included 138 patients with cirrhosis supported the conclusion that the most important risk factor for inhospital mortality was the modified SOFA score (without haematological failure) assessed after 3 days [69]. In another study, the SOFA score proved to be the most accurate, with a score > 10 predictive of mortality in 93% of cases [90]. However, the delta SOFA score (difference between baseline and at 48 h or 3-day) seems to display various prognostic values when compared to the score value at 48 h or 3-day with respect to short-term mortality [69,90,217]. Persistence of 3 or 4 non-haematological organ failures at 3-day [69] or between 3-day and 7-day  predicted inhospital mortality with a specificity of 93% or 100%. The same results were observed regarding the use of organ replacement methods (vasopressors, mechanical ventilation, renal replacement therapeutic or artificial liver support systems) during the stay of patients in intensive care [77,115]. In the CANONIC study, assessment of the ACLF grade variation between J3 and J7 after diagnosis was significantly better able to predict 28-day and 90-day mortality rates than the ACLF grade at diagnosis (AUROC 0.86 versus 0.65 and 0.81 versus 0.62 P < 0.0001, respectively) [220].
Prognosis assessment for cirrhotic patients admitted in ICU is more accurate if performed with analysis of the number of failing organs, or the SOFA score a few days after ICU admission than on ICU admission. The number of organ failures must be taken into account when identifying a patient in whom aggressive treatment may offer recovery or determine candidacy for transplantation, in order to optimise care [69,116,221]. In recent years, liver transplantation in patients with ACLF has been considered to be feasible. Retrospectives studies that included patients with ACLF grade > 2 showed that liver transplantation improved the poor prognosis of the most severely ill patients with cirrhosis [220][221][222][223][224][225][226][227][228][229]. The probability of survival was 78% at one year in patients receiving an early transplant compared to less than 10% in those not transplanted [222][223][224][225][226][227][228][229]. However, in a context of organ shortage, individual benefice should be considered together with the general interest, since an increased postoperative mortality has also been reported after transplantation for the most severe patients. Numerous studies have tried to identify factors predictive of early post-LT mortality in order to avoid futile LT. Thus, several specific scoring systems based on pre-LT variables (organ failure, organ support, sepsis, comorbidities, MELD score, SOFA score, etc.) have been developed [223,[227][228][229]. However, none of them have been included as delisting criteria in current allocation systems. To obtain good results, rapid decisionmaking by a multidisciplinary specialised team (including societal and clinical criteria) is necessary because the ''transplantation window'' is often narrow in these patients [221].
3. Appropriate artificial liver supports use in the context of ACLF might be: (1) Allowing the liver to regenerate (bridge to recovery), or (2) use of supportive therapy until liver transplantation (bridge to transplant).
To date, the best-known devices are based on the principle of albumin dialysis. The MARS 1 and Prometheus 1 systems have been the most widely studied in France. Two multicentre randomised European studies in patients with acute decompensation compared MARS 1 or Prometheus 1 with standard medical treatment. These studies did not demonstrate any benefit on survival at 28 and 90 days [230,231]. However, a significant improvement in hepatic encephalopathy and hepatorenal syndrome was seen in a post-hoc analysis using albumin dialysis with MARS versus standard medical therapy. It should be noted that the inclusion criterion for these two studies was decompensated cirrhosis, but the presence of ACLF was not known. These studies were therefore able to include patients with acute decompensation without organ failure (ACLF grade 0) or with multiple organ failures (ACLF grade 3) whose prognosis and mortality differed markedly, ranging from 4% to 80% [68]. Artificial support systems were originally used in patients with a severe stage of liver disease and multiorgan failure. More recently, these systems have been used as supportive therapy to await liver transplantation. An observational study [232] and a meta-analysis [233] showed that use of an artificial liver support system was associated with improved short-term survival (14-day and 28-day) in patients with ACLF and multiple organ failure. This short-term improvement could allow these patients to access liver transplantation, which remains the essential therapy in patients with end-stage liver disease [225,227,228]. Current data are pointing to potential interest in liver support systems to provide a ''bridge'' to final treatment. The indications need to be explored and the future of these devices requires new research protocols. The experts therefore suggest that patients should be referred to an expert centre at an early stage after decompensation of their cirrhosis. 4. Patients with cirrhosis alive at discharge from ICU have a poor prognosis, with a 1-year survival rate of less than 25% without transplantation [114]. These patients should therefore be referred systematically to a liver transplant unit.

Disclosure of interest
Catherine Paugam-Burtz and Emmanuel Weiss report speaker fees from Baxter.
Lionel Velly reports consulting and speaker fees from Baxter. The other authors declare that they have no competing interest.

Appendix B
Supplementary Table 2 Diagnostic and severity criteria of Acute-on-Chronic-Liver-Failure (ACLF) based on the organ failure (predefined by the CLIF-SOFA). Patients with three organ failures or more Grade 3