Management of acute renal replacement therapy in critically ill cirrhotic patients.

Renal replacement therapy (RRT) in cirrhotic patients encompasses a number of issues related to the particular characteristics of this population, especially in the intensive care unit (ICU) setting. The short-term prognosis of cirrhotic patients with acute kidney injury is poor, with a mortality rate higher than 65% in patients with RRT requirement, raising questions about the futility of its initiation. Regarding the management of the RRT itself, there is still no consensus with respect to the modality (continuous versus intermittent) or the anticoagulation required to improve the circuit life, which is shorter than similar at-risk populations, despite the altered haemostasis in traditional coagulation tests frequently found in these patients. Furthermore, volume management is one of the most complex issues in this cohort, where tools used for ambulatory dialysis have not yet been successfully reproducible in the ICU setting. This review attempts to shed light on the management of acute RRT in the critically ill cirrhotic population based on the current evidence and the newly available tools. We will discuss the timing of RRT initiation and cessation, the modality, anticoagulation and fluid management, as well as the outcomes of the RRT in this population, and provide a brief review of the albumin extracorporeal dialysis from the point of view of a nephrologist.

INTRODUCTION AND SCOPE OF THE PROBLEM

Renal replacement therapy (RRT) in cirrhotic (also known as chronic liver disease) patients encompasses several issues related to the particular characteristics of this population, especially in the intensive care unit (ICU) setting. Despite the increasing literature regarding this topic, currently there is no consensus about when to initiate RRT or how to deal with dialysis in terms of modality, anticoagulation and volume management. Furthermore, the short-term prognosis of cirrhotic patients who develop multiple organ failure is poor, raising doubts regarding futility.

This review attempts to shed light on the management of acute RRT in this complex population based on the current evidence and new tools available.

Epidemiology and definition of acute kidney injury in liver cirrhosis

Acute kidney injury (AKI) occurs in about 20% of hospitalized cirrhotic patients [1] and in about 40% of those admitted to the ICU [2, 3]. Ascites, elevated bilirubin, spontaneous bacterial peritonitis (SBP) and use of aminoglycoside antibiotics have previously been identified as significant risk factors for renal failure in cirrhotic patients [1].

The patient with liver cirrhosis is prone to requiring ICU admission due to their elevated risk of acute decompensation and extra-hepatic organ failure [4]. The prevalence of cirrhotic patients in ICUs is less than 5% in the reported series [4], of which up to 20–30% require RRT [4, 5].

Although one of the most known cause of AKI is hepatorenal syndrome (HRS), it actually is not the most common cause of AKI requiring RRT in the ICU, with acute tubular necrosis (ATN) in the context of sepsis, followed by hypovolemic shock the most common ones [2]. HRS is a specific type of AKI seen in patients with advanced cirrhosis characterized by impairment of kidney function because of vasoconstriction of the renal arteries due to marked splanchnic arterial vasodilation in the absence of substantial abnormalities in kidney histology, with a potential reversibility. It remains a diagnosis of exclusion, but the criteria of cirrhosis and ascites plus no response after 2 days of diuretic withdrawal and volume expansion with albumin are needed for the diagnosis. Biomarkers such as neutrophil gelatinase–associated lipocalin (NGAL) are recommended in current European Association for the Study of Liver (EASL) disease guidelines to distinguish between ATN and HRS [6].

Martin-Llahí et al. [7] investigated the predictive value of the cause of AKI in hospitalized patients with cirrhosis. Causes were classified into four groups: associated with bacterial infections, hypovolemia, HRS and parenchymatous AKI, being the most frequent cause. Three-month probability of survival ranged from 73% in patients with parenchymatous AKI to 15% for HRS.

In 2015, the International Club of Ascites (ICA) published a new definition of AKI in patients with cirrhosis, based on the Kidney Disease Improving Global Outcomes (KDIGO) criteria but without considering urinary output (Table 1). Thus, based on baseline serum creatinine (sCr) (a value obtained within the previous 3 months), AKI is defined by an absolute increase of sCr ≥0.3 mg/dL within 48 h or a percentage increase of sCr ≥50% from baseline. Three stages are defined: stage 1 when the previous criteria are met (a relative increase of sCr 1.5– to 2-fold from baseline), stage 2 when the increase in sCr is >2-fold to 3-fold from baseline and stage 3 when there is an increase of sCr >3-fold from baseline or sCr is ≥4.0 mg/dL with an acute increase of ≥0.3 mg/dL or initiation of RRT [8].

Table 1.

International Club of Ascites (ICA-AKI) classification of AKI in patients with cirrhosis, with AKI 1 sub-classification (adapted from reference number [8])

Subject	Definition
Baseline sCr	A value of sCr obtained in the previous 3 months when available can be used as baseline sCr. 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 sCr value, the sCr on admission should be used as baseline.
Definition of AKI	• Increase in sCr ≥0.3 mg/dL (26.5 umol/L) within 48 h or,• A percentage increase in sCr ≥50% from baseline, which is known, or presumed, to have occurred within the prior 7 days
Staging of AKI	Stage 1: Increase in sCr ≥0.3 mg/dL (26.5 umol/L) or an increase in sCr ≥1.5-fold to 2-fold from baseline- 1A: sCr <1.5 mg/dL- 1B: sCr ≥1.5 mg/dLStage 2: Increase in sCr >2-fold to 3-fold from baselineStage 3: Increase of sCr >3-fold from baseline or sCr ≥4 mg/dL (353 umol/L) with an acute increase ≥0.3 mg/dL (26.5 umol/L) or initiation of renal replacement therapy
Progression of AKI	Progression: progression of AKI to a higher stage and/or need for RRT		Regression: regression of AKI to a lower stage
Response to treatment	No response: no regression of AKI	Partial response: regression of AKI stage with a reduction to a value of sCr ≥0.3mg/dL (26.5 umol/L)		Full response: return of sCr to a value within 0.3 mg/dL (26.5 umol/L) of the baseline value

More recently, stage 1 has been subdivided into two subgroups based on serum levels of sCr: stage 1A (sCr <1.5 mg/dL) and stage 1B (sCr ≥1.5 mg/dL). This sub-classification is supported by the difference in survival and AKI progression or resolution rates between both groups. Also, the aetiology of AKI differs between AKI 1A and 1B, with HRS and ATN being uncommon in AKI 1A, but accounting for 48% of AKI 1B [9].

Outcomes of cirrhotic patients with need of acute RRT

The low prevalence of cirrhotic patients in the ICU is related to the frequent exclusion for admission based on the high ICU mortality rate (>50%) [10, 11], especially in patients with certain characteristics such as AKI, infection at admission or the need of vasopressor therapy, although a slight improvement has been achieved in recent years [4, 12] (Figure 1). This improvement may be explained by a reduction in the organ failure scores at admission due to earlier referral to ICU and/or organizational improvements such as access to tertiary liver centres with access to liver transplant [4].

FIGURE 1:

Kaplan– Meier curve for 28-day mortality in critically ill cirrhotic patients requiring RRT compared with patients without RRT. Source: Reproduction of Figure of Staufer et al. [5]. Used with permission of the publisher.

Several predictive scoring systems for mortality have been developed in general ICU populations to evaluate the severity of illness and prognosis, but they are not specific for cirrhotic patients. Cholongitas et al. described that organ system failure (OSF) and Sequential Organ Failure Assessment (SOFA) had the best predictive ability in this population compared to Acute Physiology and Chronic Health Disease Classification System II score (APACHE) II [13, 14]. For the subgroup of patients with AKI, the SOFA score had the best discriminative accuracy [3]. Regarding specific cirrhosis scores, the model for end-stage liver disease (MELD) is currently used to predict short-term survival in cirrhotic patients and to prioritize recipients for transplantation [15]. In a single study evaluating MELD score in critically ill cirrhotic patients, this model had almost the same discriminative capacity as the SOFA score [area under the curve (AUC): 0.81 versus 0.83] and was better than APACHE II [14].

More recently, Moreau et al. [16] used for the first time a modified SOFA score called Chronic Liver Failure Consortium-SOFA (CLIF-SOFA), which was designed specifically for the Consortium Acute-on-Chronic Liver Failure in Cirrhosis (CANONIC) Study Investigators of the European Association for the Study of the Liver (EASL)–CLIF Consortium. This score was as accurate as MELD in the prediction 28-day mortality and based on it, acute-on chronic liver failure was divided in several stages. This score is based on the number of organ failures, but, interestingly, kidney failure is counted separately, increasing the score itself. Later, two independent predictors of mortality (age and white blood cell count) were added to CLIF-SOFA score to create the Chronic Liver Failure-Sequential Organ Failure Assessment Consortium acute-on-chronic liver failure (CLIF-C-ACLF) score [17]. The reader can find an online calculator following this link https://www.efclif.com/scientific-activity/score-calculators/clif-c-aclf and the manual formula in Supplementary data, Table S1.

Regarding RRT, the largest retrospective study including cirrhotic patients requiring acute RRT was performed by Allegretti et al. [18]. It included 472 patients between 2005 and 2015 in a network of five acute care hospitals (Partners Healthcare, Massachusetts). The reasons for RRT were HRS or ATN, and patients were stratified by liver transplant listing status. The authors found that the cause of AKI (HRS versus ATN) was not significantly associated with mortality in patients with cirrhosis who required RRT within listed and non-listed groups. On the contrary, liver transplantation waiting list status or MELD score was associated with mortality and therefore may be emphasized in the decision to initiate RRT. Among those not listed for liver transplantation, mortality rates were extremely high compared with the listed ones (85% versus 46%), but the authors suggest that there may be a selected group of non-listed patients where a time-limited trial of RRT may be appropriate, particularly those who have fewer signs of critical illness.

Another contemporary study conducted by Staufer et al. [5] tried to identify the selected group of patients in which RRT would not be futile, based on calculation of CLIF-C-ACLF, CLIF-SOFA, SOFA and MELD scores on admission, 24 h prior to RRT and 24–48 h after start of RRT. The study included 78 cirrhotic patients requiring acute RRT in ICU and showed that the critical illness score CLIF-C-ACLF calculated 48 h after starting RRT was the best predictor of ICU mortality in RRT patients regardless of liver transplant options (AUC: 0.866). Interestingly, scores calculated on admission had low accuracy for the prediction of ICU mortality, except in patients with five or more organ failures assessed by CLIF-SOFA score (100% mortality), but after 48 h, a CLIF-C-ACLF score of ≥59.5 could predict ICU mortality with a sensitivity of 83.3% and a specificity of 85.7%, and could help to identify futility (Figure 2) [5]. To our knowledge, this has been the first study to assess CLIF-C-ACLF and CLIF-SOFA score in patients with AKI and need of RRT.

FIGURE 2:

Components of CLIF-C-ACLF score, adjusted by age and white-cell count. Forty-eight hours after admission, a CLIF-C-ACLF score of ≥59.5 can predict ICU mortality with a sensitivity of 83.3% and a specificity of 85.7% [5] INR: international normalized ratio.

The described studies encourage clinicians not to discard acute RRT in the non-listed cirrhotic population despite the poor prognosis in this group. There may be a selected subgroup of patients with less signs of critical illness where a time-limited trial of RRT may be appropriate.

MANAGEMENT

Those who require RRT represent the most morbid subgroup of the cirrhotic population [5, 19], with specific characteristics compared with patients without liver disease (Table 2). There is an added difficulty in the volume management due to the decreased effective circulating volume because of splanchnic arterial vasodilatation and hypoalbuminemia [20] frequent in these patients. The problems with volume management can lead to intradialytic hypotension and cardiac events. In addition, the haemorrhagic risk in these patients is high and must be carefully assessed to prioritize systemic anticoagulation, regional citrate anticoagulation (RCA) or non-anticoagulation in the RRT prescription. Hyponatremia is frequent in this population. Sodium represents 95% of ionic osmolality and so the main determinant of plasma conductivity. Central pontine myelinolysis is a rare neurologic condition most frequently caused by the rapid correction of hyponatremia, thus in order to avoid it, dialysate conductivity should be adapted. Since ammonia is similar to urea in terms of diffusive clearance, both modalities of continuous renal replacement therapy (CRRT) and intermittent haemodialysis (IHD) should be effective in clearing ammonia, differing only by the rate of removal and the rebound after IHD attributed to delayed ammonia shifts from secondary compartments. IHD removes ammonia and urea more slowly from the cerebrospinal fluid than from plasma, creating an osmotic gradient that could result in cerebral oedema. In patients with acute liver failure, the dialysis can worsen the cerebral oedema secondary to hyperammonemia and thus CRRT should be considered in these situations; the cirrhotic patients are less labile to these changes and its impact is minor but not negligible. It should be pointed out that to date there is not consistent evidence that supports the use of RRT in isolated hepatic encephalopathy [21].

Table 2.

Specific challenges of acute haemodialysis in cirrhotic patients

Characteristics	Problems
Decreased effective circulation volume secondary to splanchnic arterial vasodilatation and hypoalbuminemia	Problems with volume management can lead to intradialytic hypotension and cardiac events
Altered haemodynamic and haemostatic pathways	Increased haemorrhagic and/or prothrombotic riskLower circuit lifespan
Hyponatremia and hyperammonemia	Risk of central pontine myelinolysisRisk of cerebral oedema in acute liver failure patients

In this section, we will discuss the key points for RRT management in the critically ill cirrhotic patient, summarized in Figure 3.

FIGURE 3:

Summary of RRT management in the critically ill cirrhotic patient. RRT: renal replacement therapy, sK: serum potassium, sHCO3: serum bicarbonate, AKI: acute kidney injury, uNGAL: urinary neutrophil gelatinase–associated lipocalin, PaFi: ratio of the partial pressure of arterial oxygen to the fraction of inspired oxygen, HRS: hepatorenal syndrome, BIS: bioimpedance spectroscopy, RBVM: relative blood volume monitoring, IRRT: intermittent renal replacement therapy, CRRT: continuous renal replacement therapy, INR: international normalized ratio, RCA: regional citrate anticoagulation.

Timing of RRT initiation

In the critically ill population, the timing of RRT initiation in the absence of life-threatening complications (e.g. hyperkalemia, pulmonary oedema or severe metabolic acidosis) used to generate a dilemma between clinicians because of conflicting evidence regarding clinical outcomes and survival in an early versus delayed RRT initiation [22]. Nowadays, several clinical trials have shown no benefit of an early start of RRT and are summarized below.

The Artificial Kidney Initiation in Kidney Injury (AKIKI) Study Group has recently published two randomized controlled trials where they studied the differences between outcomes in an early, delayed and more-delayed initiation of RRT. In the first one, they compared the early versus delayed initiation of RRT in patients with AKI stage 3 who were receiving mechanical ventilation and/or catecholamine infusion that who not have an urgent indication of RRT [23]. Early strategy consisted of RRT initiation immediately after randomization. They defined the delayed strategy group as initiation of RRT with the onset of blood urea nitrogen (BUN) level higher than 112 mg/dL, a serum potassium concentration greater than 6 mmol/L, metabolic acidosis with pH <7.15 or acute pulmonary oedema responsible for severe hypoxemia despite diuretic therapy. One interesting finding was that the need for RRT could be obviated in almost 50% of cases in the delayed-strategy group, resulting in 17 RRT-free days at Day 28. In addition, the recovery of renal function assessed by diuresis was faster in the delayed-strategy group, with no difference in mortality between both groups.

In the second study [22], they focused on two different strategies for a delayed-RRT initiation: in the first (delayed), RRT was initiated immediately after randomization (defined as oliguria for more than 72 h or a BUN concentration higher than 112 mg/dL) and in the second one (more-delayed), it was initiated if BUN concentration was higher than 140 mg/dL or mandatory indication. This resulted in no difference in RRT-free days between groups, and in a multivariable analysis, the 60-day mortality was higher in the more-delayed strategy. However, it is important to highlight that in these two trials, patients with diagnosis of Child C liver cirrhosis were excluded.

Similar results were found in the Standard versus Accelerated Initiation of Renal Replacement Therapy in Acute Kidney Injury (STARRT-AKI) trial which did include patients with liver disease [24], where the patients received an accelerated strategy for RRT (defined as initiation of RRT within 12 h after reaching AKIN2 or AKIN3 stage) in contrast to a standard strategy {defined as avoiding RRT until the development of a serum potassium level ≥6.0 mmol, a pH ≤7.20, a serum bicarbonate level ≤12 mmol/L, clinical perception of volume overload, evidence of severe respiratory failure [ratio of the partial pressure of arterial oxygen to the fraction of inspired oxygen (PaFi) of 200 or less], or persistent AKI for at least 72 h after randomization}. This study included a total of 337 patients who had liver disease as a pre-existing condition, 172 in the accelerated strategy and 165 in the standard strategy. The aetiology of the liver disease was not specified. They found that an accelerated strategy for the initiation of RRT did not have a lower risk of death at 90 days, with a mortality of 43.9% and 43.7%, respectively. Among the patients who were alive at 90 days, dependence on RRT was greater in the accelerated strategy group {10.4% versus 6.0%, risk ratio (RR) 1.74 [95% confidence interval (CI) 1.24–2.43]}, suggesting that greater exposure to RRT or mediated by iatrogenic factors may compromise kidney function recovery. It is also important to note that in the standard strategy group, the RRT was initiated only in 61.8% patients, with a median of 31.1 h. Of note, there was no sub-analysis to assess the outcome of the patients with liver disease.

Nevertheless, there is no consensus about the ideal timing for initiation of RRT in the critically ill cirrhotic population [22, 24, 25]. Although the accepted definition of AKI in cirrhosis is based on sCr levels [26], it is important to take into account the limitations of this biomarker in this particular population. Creatinine production depends on muscle mass and physical activity, often reduced in the cirrhotic patient. Furthermore, the tubular secretion of creatinine is increased in these patients. All this, added to the fact that the sCr value may be affected by volume overload, leads to a potential underestimation of the real sCr [27].

In this scenario, some studies have begun to use different biomarkers to help define established kidney injury. The Early Versus Late Initiation of Renal Replacement Therapy In Critically Ill Patients With Acute Kidney Injury (ELAIN) clinical trial proved that the use of the KDIGO classification together with the use of plasma NGAL can reliably detect patients with progressively deteriorating AKI [25].

NGAL is one of the most studied biomarkers associated with AKI. NGAL is a low molecular weight protein produced by the thick ascending loop and collecting duct cells, and is a marker of kidney injury. Both serum and urine NGAL have performed well in detecting AKI across a variety of ICU populations [28, 29]. However, NGAL and other similar biomarkers, the description of which is beyond the scope of this review, may not be specific for kidney disease and should accordingly be subjected to interpretation depending on the clinical condition [27, 30, 31].

In recent studies, NGAL has been proposed to correctly assess the difference between ATN and other causes of AKI in the cirrhotic population [28, 32, 33]. The clinical significance of establishing this difference lies in the completely different management of these two conditions: treatment of ATN is supportive, i.e. to remove risk factors, treatment of possible infections, haemodynamic and volume monitoring and correction if impaired and RRT if needed; on the contrary, treatment of HRS is based on volume expansion with albumin and administration of vasoconstrictors, specifically terlipressin, and evaluation for liver transplantation [23].

Huelin et al. [28] conducted a prospective study of consecutive patients with cirrhosis admitted to a Liver Unit of a tertiary hospital for acute decompensation of the disease who either had AKI at admission or developed it during hospitalization. The primary objective was to investigate the accuracy of three urinary biomarkers [NGAL, monomeric NGAL (mNGAL) and IL-18] in the prediction of ATN versus other causes of AKI. They found that at enrollment, patients with ATN had NGAL between 153 and 2114 µg/g (mean 799 µg/g), mNGAL between 108 and 1838 µg/g (mean 543 ug/g) and IL-8 between 11 and 107 pg/g (mean 33 pg/g), while hypovolemia-induced AKI and HRS-AKI had lower values, with a mean NGAL of 48 and 76 µg/g, respectively. They concluded that in this cohort, the urinary NGAL has a high accuracy in the differential diagnosis between ATN and other types of AKI, including HRS and hypovolemia-induced AKI. In addition, urine NGAL was also associated with progression of AKI during hospitalization, need for RRT and 28-day mortality [28]. These findings are in concordance with another study of Allegretti et al. [32] which found that the urinary NGAL correctly identified ATN from other types of AKI in cirrhosis: at enrollment, urinary NGAL levels were lowest in prerenal AKI [(45 [0, 154] µg/g)], higher in HRS [(110 [50, 393] µg/g)], and highest in ATN [(344 [132, 1 429] µg/g)]. They defined that the optimal NGAL cut point to discriminate ATN from non-ATN kidney injury could be of 244 µg/g. This biomarker also held value in improving prediction of short-term mortality and AKI stage progression.

More recently, a novel biomarker, urinary C-C motif chemokine ligand 14 (CCL14), has been identified as a predictor of persistent AKI [34], a clinical situation more likely to require RRT. It would be interesting to test the efficacy of this biomarker in the specific population of patients with liver cirrhosis.

In conclusion, there is still no consensus on the ideal timing for initiation of RRT in the critically ill cirrhotic patients. Clinical judgment with careful evaluation of a patient's clinical situation and prognosis continues to be the main determinant as to whether and when to initiate RRT [35]. Novel biomarkers seem to help differentiate acute tubular necrosis (ATN) from other causes of AKI in the cirrhotic population, as well as to predict the need of RRT initiation, and could be a useful tool for clinical practice.

Modality

There are two well-known modalities of acute RRT: intermittent (IRRT) and continuous (CRRT), each with their own benefits and drawbacks. The IRRT is provided generally every 2 days and each treatment typically lasts between 3 and 5 h, achieving a fast removal of small molecules and volume, this last one adjusted to the patient's haemodynamic tolerance. On the other hand, CRRT is intended to last 24 h or longer, providing a slower removal of fluids and solutes. Hybrid haemodialysis modalities [e.g. sustained low efficiency dialysis (SLED) or slow extended daily dialysis (SLEDD)] generally last between 8 and 12 h, with intermediate rates of ultrafiltration and clearance [26, 36, 37].

The initial modality remains the subject of debate. There are several controlled trials and meta-analysis that compare the benefits of CRRT and IRRT in the general critically ill population [38-42]. Vinsonneau et al. [42] conducted a multicentre randomized clinical trial comparing IRRT and CRRT, finding no differences between both techniques in survival rates, renal function recovery or adverse events.

A meta-analysis conducted by Bagshaw et al. [38] studied the clinical outcomes of patients in the ICU with AKI that underwent CRRT and IRRT. They found that CRRT was associated with fewer episodes of haemodynamic instability and arrythmias. However, there were no differences in mortality or recovery of renal function at hospital discharge [38]. Another meta-analysis conducted by The Cochrane Group, also found a higher mean arterial pressure in CRRT versus IRRT, but with similar number of hypotension episodes [39]. Vanholder et al. [41] concluded that if there is a haemodynamic benefit for CRRT, this does not translate into differences in survival.

In the critically ill cirrhotic patients there have been no controlled trials to demonstrate superiority of one modality over the other. CRRT may be better tolerated not only because of its alleged haemodynamic benefit, but also because of the slower correction of hyponatremia [43]. Wong et al. [44] studied the survival of liver transplant candidates with acute kidney failure and requirement of RRT. She found an overall survival of 33%, with an increased mortality in the patients who received CRRT compared to IHD (P = 0.02). However, the patients who were placed on CRRT were more severely ill and with greater haemodynamic instability, with a higher APACHE II score and lower mean arterial pressure, and were more likely intubated or on vasopressors. This difference in mortality is more likely to be related to the patient's characteristics in each group rather than the modality itself, as has been observed in previous studies of the general population [45].

Given the lack of evidence, it is accepted that the choice of RRT modality is determined by the patient's haemodynamic stability and severity of illness [43]. CRRT is preferred in haemodynamically unstable patients, in multi-organ failure, in severe hyponatremia, probably in patients with severe sepsis and hyperammonemia, and in patients with acute liver failure.

Anticoagulation

Patients with cirrhosis have profoundly altered haemodynamic and haemostatic pathways (i.e. reduction in specific coagulation factor V, FVII, FIX, FX, FXI, prothrombin, protein C and protein S, and increase of FVIII and von Willebrand factor activity, thrombocytopenia, etc.), resulting in a procoagulant, anticoagulant or mixed phenotype [46]. In addition, the INR is not a reliable marker to assess haemostasis, as it depends on some procoagulant factors (I, II, V, VII and X), but does not reflect the deficiencies of the anticoagulation system (low C protein), leading to a hypercoagulable state despite the prolonged INR [47, 48]. To delve into the pathophysiology of haemostasis in patients with liver disease, we recommend the article of Zermatten et al. [49].

CRRT benefits are compromised with premature interruption of the technique due to inadequate anticoagulation [50]. Also, in the cirrhotic population, the CRRT circuits seem to have a shorter life span compared with other ICU populations, with a median circuit life span of 12 h without anticoagulation [51-53]. Premature filter loss secondary to clotting can reduce the clearance and ultrafiltration rate and contribute to blood loss [50]. There are conflicting data regarding the characteristics that associate with circuit survival times. It has been described that higher activated partial thromboplastin time (APTT), serum bilirubin levels on ICU admission and thrombocytopenia (especially with platelet count ˂50 × 109/mm3) are associated with longer circuit lifespan [52]. However, other studies show no relationship between laboratory clotting times or peripheral platelet count with the duration of the circuit lifespan [53].

Systemic heparin and RCA remain as the two main anticoagulation strategies used in CRRT. According to the KDIGO Guidelines [54], citrate is preferred over heparin for anticoagulation in CRRT in patients with an increased bleeding risk or impaired coagulation and not already receiving effective systemic anticoagulation, but with a low level of evidence (2B).

Regarding anticoagulation with heparin, it has been described that systemic or regional anticoagulation for CRRT in patients with liver disease did not prolong circuit life span, despite the significantly elevated systemic APTT with systemic heparinization, higher intravenous heparin dose with regional heparinization and greater proportion of circuits performed using haemodiafiltration versus haemofiltration. Given these results and the possible complications related to the use of heparin (e.g. thrombocytopenia), heparinization for CRRT in this cohort may not be justifiable [52].

In patients like the cirrhotic population in whom bleeding risk is high, regional anticoagulation with citrate (RCA) may be indicated. This technique consists of citrate infusion to the blood before the extracorporeal circuit that chelates ionized calcium, a mandatory cofactor of many enzymes of the coagulation cascade, thus avoiding coagulation of the circuit. Afterwards the calcium–citrate complexes (CCC) are mainly removed by the dialysis itself (up to 60%) [55, 56] and systemic serum calcium is restored by a post-filter calcium infusion. The remaining CCC is metabolized to bicarbonate by the liver and to a lesser extent by the skeletal muscle and kidney, and a potentially altered metabolism in patients with hepatic impairment has been the matter of concern in past years [57]. When citrate catabolism is normal, RCA administration leads to plasma alkalinization. However, if the citrate administration exceeds the body's capacity to metabolize it, the CCC accumulates in the peripheral blood, leading to metabolic acidosis, an increased total blood calcium to ionized calcium (Catot/Cai) ratio (>2.5) and severe ionized hypocalcemia. The Catot/Cai ratio reflects on plasma citrate concentrations since total calcium is the sum of albumin and citrate-complexed calcium and ionized calcium is reduced by its union to citrate. This could cause decreased myocardial contractility and vasoplegia, and could be potentially lethal [56]. This constitutes the main reason why historically RCA has been contraindicated in patients with altered liver function.

Current knowledge about the safety profile of citrate in critically ill patients with altered hepatic function is improving. Klingele et al. [58] carried a retrospective study where 69 patients of these characteristics underwent CRRT with RCA. Contrary to what was expected, incidence of metabolic alkalosis and accumulation of citrate was less than expected (24.6% and 23.2%, respectively) (Figure 4). These alterations were corrected with changes of the dialysate and blood flows. No RCA had to be stopped because of hypercalcemia or accumulation of citrate [58]. Furthermore, Zhang et al. [59] conducted a meta-analysis regarding the safety profile of RCA for CRRT in liver failure patients, finding no increased risk of citrate accumulation and no difference in pH, serum lactate level and Catot/Cai ratio between liver and non-liver failure patients; also, the circuit life was longer. When comparing circuit life span between RCA-CRRT and systemic heparin anticoagulation, the first one is better than the latter (circuit life span 47 h versus 27 h) [50].

FIGURE 4:

Differences between expected and occurred complications in patients with impaired liver function undergoing RRT with regional citrate anticoagulation. The incidence of the expected complication was based on the current knowledge of the metabolism of citrate, and was not estimated. Source: Adaptation of Figure of Klingele et al. [58].

To help predict which patients are at risk of citrate accumulation, a prospective observational study was conducted in 28 critically ill patients with decompensated liver cirrhosis or acute liver failure, performing 43 CRRT runs using RCA [60]. They identified that serum lactate ≥3.4 mmol/L and prothrombin time ≤26%, measured before the onset of the RRT, can predict an increase in the Catot/Cai ratio ≥2.5 with high sensitivity and specificity. On the contrary, liver function showed poor predictive capabilities.

Another strategy to address the circuit life span in patients with liver impairment is to start without any anticoagulation whatsoever [52]. In patients prone to bleeding complications, we recommend to start the CRRT without anticoagulation and if the circuit clots faster than anticipated, RCA-CRRT constitutes a safe alternative to systemic heparin, remarking the need of caution and close monitoring of calcium and metabolic disorders measuring the Catot/Cai ratio at least twice a day for the first 48 h.

Fluid management

Fluid overload is associated with worse outcomes in critically ill patients and in cirrhotic cohorts as well [61, 62]. These patients have a haemodynamic pattern of hyperdynamia and vasoplegia in the setting of relative hypovolemia. These characteristics worsen with cirrhosis progression, making it difficult to maintain the necessary negative hydric balance in critically ill patients.

Volume assessment requires an integration of different clinical, analytical and technological data [63]. Although physical exam including ultrasound is mandatory, it is usually not precise enough, so new tools such as bioimpedance spectroscopy (BIS) and the relative blood volume monitoring (RBVM) may provide useful information for management.

BIS provides an estimation of intracellular water (ICW) and extracellular water (ECW), which can help to identify the total fluid overload of the patient [64]. One disadvantage is that volumes that are surrounded by a tissue layer such as peritoneum are not well detected, so it is recommended to evacuate the ascitic fluid before using BIS.

The RBVM tool is based on sensors placed in the intermittent dialysis machine [65]. The volume of plasma water is reduced during ultrafiltration, but the intravascular protein mass remains constant even when there are changes in blood volume due to ultrafiltration and refilling. The result is a change in the concentration of proteins and thus in the density of the blood. The initial relative blood volume (RBV) of 100% is reduced to approximately 80% on completion of a dialysis session. The results of continuous measurement of RBV can be used to ensure that the reduction of blood volume is restricted to a level that is tolerable for the patient. While it has proved to be a useful tool in stable chronic haemodialysis patients [66], evidence is still conflicting regarding whether it can predict intradialytic hypotension in the critical care setting [67].

Another possible tool used to avoid hypotension during dialysis treatment is albumin infusion during the therapy. In a recent study, Macedo et al. [68] demonstrated that in hypoalbuminemic patients (defined as serum albumin level < 3 g/dL at the initiation of dialysis), albumin administration results in fewer episodes of hypotension and improved fluid removal, in comparison with sodium chloride 0.9% administration. However, the population was not critically ill and there is a possible confounding factor because the albumin infusion used also contains sodium. Moreover, the study did not specify if cirrhotic patients were included, or the cause of hypoalbuminemia.

Sodium modelling has also been described as a tool for mitigation of intradialytic hypotension in the setting of routine ambulatory dialysis for end-stage renal disease. However existing data suggest that this is not reproducible in a critically ill patient. Lynch et al. [69] studied 191 patients with AKI who underwent IHD in the ICU, finding no association between the use of sodium modelling, attainment of ultrafiltration goal, total ultrafiltration volume or the composite outcome of all-cause mortality and dialysis dependence at the time of discharge.

For patients with indication of liver transplantation, CRRT has been used not only before and after the surgery, but also as part of the intraoperative management, especially due to high portal venous pressures and hypervolemia itself, to help in fluid management. It has been also considered a helpful tool in patients at high risk of cerebral oedema. The use of intraoperative CRRT may help sicker patients tolerate the surgery and achieve short-term clinical outcomes comparable to less ill patients without intraoperative RRT, but with no impact on the short- or long-term mortality rate, hospital length of stay and post-liver transplant complications [70].

Cessation of RRT

It is equally important to correctly identify not only the adequate timing to initiate the RRT, but also when to discontinue it. However, there is a lack of evidence on the RRT weaning. SCr decreases because of the RRT and not because of the recovery of kidney function for itself, and therefore it is not a reliable factor for RRT discontinuation. The urine output could be used as a marker with several limitations, such as the potential preservation of urine output during AKI, the occurrence of urine output before renal recovery, and the negative impact of diuretic use on its predictive ability. Aniort et al. [71] measured the daily urinary urea excretion in ICU patients with AKI undergoing IHD weaning. The inclusion criteria were patients older than 18 years treated with IHD for at least 7 days and four IHD sessions. The exclusion criteria were patients with a decision to forgo life-sustaining treatment, patients who had undergone renal transplantation, patients treated with continuous RRT at the time of weaning, chronic dialysis patients, and patients with urine output less than 100 mL/24 h before the last dialysis session in the ICU. Liver failure was not an exclusion criterion, but it is not specified. They found that values greater than 1.35 mmol/kg/24 h could have a better predictive value than urine output or urinary urea concentration alone.

The BEST study [72] focused on patients discontinuing CRRT in an ICU setting, which included patients with HRS. They found that urine output at the time of cessation of CRRT was the most important predictor of successful weaning and that diuretic usage negatively affected the predictive ability of urine output. In terms of survival, patients whose CRRT was discontinued successfully had better outcome with an intrahospital mortality of 28.5%, while patients who needed to be retreated with RRT had a mortality of 42.7%.

Another study that focused on CRRT weaning that also included patients with history of liver disease and acute liver failure [73] measured 2-hourly creatinine clearances within the 12-h time period preceding CRRT cessation, finding that a level of 23 mL/min was shown to have optimal sensitivity and specificity, and was better predictor than the sCr and urine output alone.

Regarding the factors associated with failure of discontinuation of RRT, Wu et al. [74] found that the independent factors for restarting RRT within 30 days were longer duration of RRT, higher SOFA score, oliguria and age above 65 years. This study included 12 patients with liver failure, of which 10 remained free of dialysis for at least 30 days. Therefore, liver failure on its own does not appear to associate with failure of discontinuation of RRT.

Once the cessation of the CRRT is decided, SLED constitutes an attractive alternative to the regular IHD. It is a hybrid technique that has advantages compared with both CRRT and IHD, and is able to provide satisfactory metabolic control with few episodes of hypotension [75]. Therefore, in the critically ill patients, it could be a useful bridge technique between CRRT and IHD.

ALBUMIN EXTRACORPOREAL DIALYSIS—WHAT THE NEPHROLOGIST NEEDS TO KNOW

In the last decades, extracorporeal systems that clear hepatic toxins have been investigated in order to create a window of opportunity for liver transplantation in acute liver failure or acute-on-chronic liver failure. However, since there are not robust randomized studies, it is difficult to develop evidence-based clinical and biochemical criteria to define the need to initiate therapy and the optimal duration of therapy.

Currently, there are mainly two systems that are used in this context: Molecular Adsorbent Recirculating System (MARS®) (Gambro, Lund, Sweden) and PROMETHEUS® (Fresenius AG, Hamburg, Germany). The main difference remains on the mechanism of albumin-bound toxin removal [76, 77].

In MARS®, blood is dialysed across an albumin-impregnated high-flux dialysis membrane with a cutoff of 50 kDda [78]. The dialysate consists of 20% human albumin in which albumin-bound toxins in blood are released to the membrane and picked up by albumin in the dialysate, whereas water-soluble solutes diffuse across the membrane. The albumin dialysate goes to a high-flux dialysis filter that cleans the water-soluble toxins and subsequently goes through two sequential adsorbent columns containing activated charcoal and anion exchange resin that remove the albumin-bound toxins. In the PROMETHEUS ® system, patient plasma is fractionated through an albumin-permeable filter with a cutoff of 250 kDa. Albumin and other plasma proteins cross the membrane and pass across two columns in series: one an anion-exchange column and another a neutral resin adsorber. The cleaned albumin/plasma is returned to the standard blood pool circuit where it is then treated with conventional high-flux haemodialysis that cleans water soluble toxins (creatinine, urea, etc.) [79, 80].

From the nephrologist’s point of view, both systems can depurate water-soluble toxins and electrolytes such as potassium, but only MARS has specific studies that demonstrate a meaningful degree of urea reduction [81]. If ultrafiltration for volume control is needed during the technique, only PROMETHEUS can be used. On the other hand, RRT can be performed alone once the session is completed using the same machine, which allows volume management since both techniques are used with monitors of continuous RRT.

CONCLUSION

RRT in cirrhotic patients involves a number of controversial issues related to the particular characteristics of this population. Regarding prognosis, ICU mortality rate is high, especially in those patients not listed for liver transplantation, raising questions regarding the futility of its initiation. Scores of critical illness like CLIF-C-ACLF can help to identify a selected group of patients where a time-limited trial of RRT may be appropriate.

Once an RRT is decided, the timing of initiation should be considered according to the current evidence in general cohorts that show no benefit in an early initiation in a clinical scenario where no urgent dialysis criterion is met. EASL guidelines support using biomarkers like NGAL to distinguish ATN from HRS; however, more studies about other novel biomarkers are needed to support its use in prediction of RRT initiation.

The choice of RRT modality remains a subject of debate, where continuous techniques are associated with fewer hypotension episodes and slower correction of hyponatremia, but show no benefit in terms of survival. It is accepted that the modality should be determined by the patient's haemodynamic stability and severity of illness.

The importance of an adequate anticoagulation during CRRT is that the benefits of this therapy are best achieved when there are no interruptions, and in the cirrhotic patient the CRRT circuits seem to have a shorter life span compared with other ICU populations despite the altered coagulation pathways. The decision of the anticoagulation strategy should be individualized for each patient. In patients with severe thrombocytopenia or with bleeding complications, it is reasonable to start without any anticoagulation whatsoever and if the circuit life span is shorter than anticipated, RCA constitutes a safe alternative, remarking the need of close monitoring of calcium and metabolic disorders.

Volume management is one of the most complex issues in this special population and new tools like BIS and RBVM could be of great help, but more studies are needed in the critical care setting. On the other hand, sodium modelling has not proved to reduce the incidence of intradialytic hypotension or better outcomes.

Despite the increasing literature regarding this topic, the majority of studies about the critically ill patient do not consider the cirrhotic population in this clinical setting. More studies and clinical trials are needed, and a multidisciplinary team with nephrologists, intensivists and hepatologists is recommended in order to handle these complex patients.

DATA AVAILABILITY STATEMENT

The authors confirm that the data supporting the findings of this study are available within the article and its supplementary materials.

Click here for additional data file.