SPAD employs the same blood membrane exchange mechanism as MARS but uses the albumin-containing dialysate in a single-pass mode. This allows for a technical simplification of the system if compared to MARS. On the other hand, for cost reasons the dialysate-albumin concentration needs to be kept low in the range between 1 % and 5 %. Typically, SPAD is used with low blood and dialysate flow rates, comparable to those used in continuous veno-venous hemodialysis or hemodiafiltration (Mitzner et al. 2006) (Fig. 2).

The Prometheus system (also fractionated plasma separation and adsorption, FPSA) consists of the AlbuFlow membrane that allows the passage of a plasma fraction containing patients’ albumin. This fraction is passed over two sorbent columns (a neutral resin and an anion exchanger) to achieve removal of albumin-bound toxins from the albumin. The plasma fraction is then passed back to the blood circuit. The whole blood is then dialyzed while passed through a high flux dialyzer and returned to the patient (Fig. 3) (Falkenhagen et al. 1999; Rifai 2011). The preferred treatment time is approximately 6 h. However, technically longer treatments are possible and were clinically used at least in single cases. The flow rates used are 150–300 mL/min blood flow depending on the hemodynamic status of individual patients, 300 mL/min FPSA circuit flow rate, and 500 mL/min dialysate flow rate (Mitzner et al. 2006).

Various anticoagulants can be used for all three albumin-cleansing methods. However, typically best clinical performance is reached with citrate.

TPE is based on filter or centrifuge separation of patient’s plasma with 1:1 volume exchange with either fresh frozen plasma (FFP) or, e.g., a Ringer-lactate solution containing 5 % human albumin. Advantages are the instant availability and easy technical performance. Potential disadvantages are allergic reactions against FFP. Exchange against FFP can help stabilize the coagulation situation of patients with a high risk of bleeding (Stenbøg et al. 2013).

These devices utilize hepatocytes from animal or human sources. The most developed systems that have been tested clinically so far are the HepatAssist and the extracorporeal liver assist device (ELAD) . The HepatAssist that is not any longer in clinical use utilized a two-step approach consisting of plasma separation and charcoal perfusion followed by perfusion of the plasma through a cell module with porcine hepatocytes immobilized on the outside of a plasma separator membrane. ELAD uses a selective plasma filtration with passage of the resulting plasma fraction through a hollow fiber module carrying up to 450 g of human C3A-hepatoblastoma cells placed on the outside of the hollow fibers (Wertheim et al. 2012).

Many substances accumulate in the plasma during the course of liver failure. If they can be linked to the development, maintenance, or further aggravation of liver failure, they qualify as “liver failure toxins”. One of the key advantages of liver support methods over renal dialysis techniques is that the pattern of harmful substances that can be removed from blood is much broader. Especially a significant removal of albumin-bound metabolites and drugs that accumulate in liver or kidney failure, enzyme defects such as protoporphyria, or drug overdose belongs to this pattern. Substances that are bound to serum albumin and exert damaging effects in higher concentrations are termed albumin-bound toxins (ABT). Rather different groups of biochemicals belong to this group, including steroid acids (e.g., bile acids), open and closed tetrapyrroles (e.g., bilirubin or protoporphyrin), amino acids (especially aromatic amino acids), glycoside derivatives (e.g., indoxyl sulfate), phenols (e.g., para-cresol), lipids (short- and medium-chain fatty acids such as octanoate), and heterocyclic organic compounds (such as furancarboxylic acid). For MARS the range of clearances for ABT was found to be in between 10 and 60 ml/min (Mitzner et al. 2001). Moreover, albumin-cleansing methods allow for removal of water-soluble and thus dialyzable substances such as smaller proteins (e.g., cytokines like interleukin-6 or tumor necrosis factor alpha), ammonia, creatinine, or urea (Gaspari et al. 2006; Stefoni et al. 2006; Yuan et al. 2006; Nadalin et al. 2007; Novelli et al. 2005; Heemann et al. 2002; Lisboa et al. 2012).

The clinical relevance of ABT removal was investigated in detail in a number of animal and clinical trials. Plasmatic nitric oxide (NO), bound to albumin as a nitrosothiol, is responsible for the typical hemodynamic changes of liver failure (hyperdynamic hypotension ). NO removal by MARS was demonstrated in several clinical investigations (Guo et al. 2003; Sen et al. 2004a; Kurtovic et al. 2004; Laleman et al. 2006). Capability to remove inducers of hepatic encephalopathy such as ammonia, tryptophan, and endogenous benzodiazepines renders albumin dialysis a valuable tool for this major complication of liver failure (Mitzner et al. 2001; Parés et al. 2009; Donati et al. 2014; Rustom et al. 2014). The Fisher index as the ratio of branched-chain and aromatic amino acids is increasing during MARS treatments (Mitzner et al. 2001; Parés et al. 2009; Rustom 2014). A constant finding is the removal of bilirubin and bile acids (Huang et al. 2012; Lisboa et al. 2012; Cisneros-Garza et al. 2014; Donati et al. 2014; Rustom et al. 2014). Both fractions the conjugated and, to a lesser extent, the unconjugated bilirubin are removed (Mitzner et al. 2001; Donati et al. 2014). It was found that MARS changes the plasma bile acid composition toward hydrophilic bile acids (Stadlbauer et al. 2007). Moreover, significant clearance of proinflammatory and anti-inflammatory cytokines was observed (Guo et al. 2003; Kurtovic et al. 2004; Auth et al. 2005; Di Campli et al. 2005; Isoniemi et al. 2005; Yuan et al. 2006). However, this did not always result in decrease of blood cytokine levels (Sen et al. 2004a; Ilonen et al. 2006; Stadlbauer et al. 2006). MARS removes copper in the setting of acute Wilson’s disease (Mitzner et al. 2001; Rustom et al. 2014). A probably very important effect of albumin dialysis is an increase of the binding capacity of patient’s albumin. In a group of patients with acute decompensation on top of chronic liver failure (AoCLF), the median binding capacity was 63 % (compared with healthy controls 98 %, p < 0.001). MARS treatments resulted in a significant increase (Klammt et al. 2007, 2008). The impact of this effect remains to be investigated. However, better drug-binding capacity and internal clearance of ABT can be assumed.

Data situation regarding clearance performance of Prometheus, SPAD, and TPE is less complete than for MARS. SPAD removes, among others, bilirubin and copper (Kreymann et al. 1999). Prometheus has proven effect on ammonia, bilirubin, and bile acids (Rifai et al. 2006; Kribben et al. 2011; Krisper et al. 2011). TPE, by definition, can remove virtually every plasma compound. However, the clinical effect is limited by the typically short treatment times. Clinical reports describe among others lowering of ammonia, copper, various exogenous toxins, and drugs (Hilal and Morehead 2014; Stenbøg et al. 2013; Ye et al. 2014).

There is scare data for substance clearances during cell bioreactor treatments. However, removal of bile acids and, to a lesser extent, bilirubin was reported for the HepatAssist system (Krisper et al. 2011; Demetriou et al. 2004).

Hepatic encephalopathy (HE) is a major complication of both chronic and acute liver failure. MARS can improve HE grade and Glasgow Coma Scale (for review, see Mitzner et al. 2002). A multicenter randomized clinical trial studying MARS in 70 AoCLF patients with HE grades III and IV showed significant advantages of MARS versus standard therapy with regard to time to improve and grade of improvement (Hassanein et al. 2007). This was confirmed by other randomized clinical trials (Sen et al. 2004a; Huang et al. 2012; Banares et al. 2013), in several case series (Heemann et al. 2002; Gaspari et al. 2006; Hetz et al. 2006; Stefoni et al. 2006; Yuan et al. 2006; Camus et al. 2009; Parés et al. 2009; Cisneros-Garza et al. 2014), and a meta-analysis (Vaid et al. 2012). Generally, MARS is regarded as a valuable treatment option for HE (Kobashi-Margáin et al. 2011; Leise et al. 2014).

There exist single case reports on positive impact on HE with SPAD (Kreymann 1999) and TPE (Liu 2013; Stenbog 2013). The authors are unaware of significant improvements reported for Prometheus or cell bioreactor treatments.

A drop in intracranial pressur e (ICP) during clinical use of MARS was reported by different groups (Mitzner et al. 2001). No randomized clinical trial has investigated this phenomenon so far. However, in a controlled animal study using an ALF-pig model based on devascularization of the liver, MARS, initiated two hours after clamping, significantly attenuated the ICP increase. The MARS group had a significantly lower brain water content and brain ammonia concentration (Sen et al. 2006). Similar results from an animal model of increased ICP were obtained for Prometheus (Ryska et al. 2012). No reports were found for impact on ICP by SPAD, TPE, or cell bioreactors.

Several groups reported improvement of kidney function during MARS treatments. This included decrease in creatinine and urea, increase in urine output, and resolution of HRS (Mitzner et al. 2001; Heemann et al. 2002; Saich et al. 2005; Hetz et al. 2006). In a recent study, of 32 HRS type 1 patients, 13 (40 %) had improved renal function. Among these, nine (28 %) had complete renal recovery. The 28-day survival rate was 47 % (Lavayssière et al. 2013). The positive impact on renal function in HRS type 1 was confirmed in a controlled randomized trial (Mitzner et al. 2000). A possible mode of action is improvement of renal blood flow with subsequent reuptake of organ function (Mitzner et al. 2001). A significant decrease in plasma renin was found in HRS patients treated with MARS that might reflect improved renal blood perfusion (Schmidt et al. 2001; Catalina et al. 2003; Laleman et al. 2006). However, positive effects on kidney function will be most likely if therapy is initiated prior to irreversible ischemic damage to the organs (Wong et al. 2010). MARS is considered as a valuable treatment option for HRS (Cárdenas and Ginès 2006; Moreau and Lebrec 2007).

There are case reports suggesting efficacy of TPE for HRS (Hilal and Morehead 2014; Yu et al. 2014). For all other liver support methods, no clear signals with regard to impact on HRS are available.

Accidental or suicidal drug overdose resulting in life-threatening intoxications represents an indication for MARS. The therapeutic goal is either secondary drug removal, if the drug in question is albumin bound and present in the blood circulation, or, more frequently, to treat drug-induced liver failure (for review, see Wittebole and Hantson 2011). Intoxications and liver failure cases induced by various drugs, e.g., acetaminophen or natural toxins, such as amanita toxin, were successfully treated (Koivusalo et al. 2005; Braun et al. 2006; Pichon et al. 2006; Lee et al. 2005; Sorodoc et al. 2010; Swarnalatha et al. 2013) (Please see list below).

Use of MARS in drug overdose and poisoning caused by various drugs, toxins, chemicals, and other substances (for review, see Wittebole 2011; Mitzner et al. 2001, 2002; Prokurat et al. 2002; Lee et al. 2005; Braun et al. 2006; Pichon et al. 2006; Sorodoc et al. 2010; Swarnalatha et al. 2013; Rustom et al. 2014):

With regard to unintended drug removal, there were case reports of relevant removal of piperacillin/tazobactam (Ruggero et al. 2013) and of moxifloxacin and meropenem in an in vitro model (Roth et al. 2013). Only mild impact was found on amphotericin B formulations (Weiler et al. 2011), while no impact on tacrolimus plasma levels was observed (Personett et al. 2014). Accordingly, dose adjustments of affected drugs may be required.

Prometheus was found to be effective in case series of mushroom and ecstasy/cocaine poisonings (Vardar et al. 2010; Kramer et al. 2003). For TPE two cases of drug-induced liver injury following multiple antibiotics and tuberculostatic therapy showed positive results (Liu et al. 2013).

Patients with unbearable pruritus resistant to medical therapy respond well to MARS treatments. Underlying liver diseases were cholestatic forms of liver disease such as PBC or primary sclerosing cholangitis as well as chronic viral hepatitis. Typically, two single treatments lowered pruritus impressively as was documented by visual analog scale. The relief lasted between several weeks up to 3 months. However, a number of cases did not respond (Saich et al. 2005; Bellmann et al. 2004; Gaspari et al. 2006; Montero et al. 2006; Parés et al. 2010). MARS was found to be effective as a repeated outpatient treatment (Leckie et al. 2012). It appeared to be safe and effective in children with repeated long-term uses in cases of cholestatic pruritus (Schaefer et al. 2012). The positive clinical effect of MARS on pruritus cannot be explained fully today. However, selective removal of hydrophobic bile acids leading to a longer-lasting shift in the bile acid pattern of the patients was suspected to be a potential mechanism (Stadlbauer et al. 2007; Parés et al. 2010). Protein analysis from MARS column posttreatment revealed a specific removal pattern that might hint at pathophysiologically new traces regarding the cause of hepatic pruritus (Gay et al. 2011). Gene profiling microarray analysis of cytokines revealed the development of an anti-inflammatory pattern resulting from MARS (Lisboa et al. 2012).

Also for the Prometheus system, effective treatment of hepatic pruritus was reported (Rifai et al. 2006). No reports were found describing clinical use of SPAD or cell bioreactors for hepatic pruritus.

Hypoxic situations resulting in ischemic liver failure have been treated with MARS in the context of cardiogenic shock with low-output failure (El Banayosy et al. 2004; Drolz et al. 2011; Zittermann et al. 2013). No reports are available for the other liver support methods.

In ALF patients listed for liver transplantation, MARS can be applied as a bridging method to stabilize the patient’s condition. Not only was the treatment reported to be safe, but patient’s condition improved markedly in a substantial number to such an extent that sustained liver regeneration was achieved. Koivusalo et al. (2005) report 56 patients with ALF (29 toxic, 22 unknown, 5 other). All fulfilled liver transplantation criteria or had ingested a lethal dose of a known toxic agent (e.g., paracetamol, Amanita phalloides). A mean number of 3 MARS treatments were performed per patient; target treatment duration was 22 h/session. The 1-year survival was 84 %. Recovery of native liver function occurred in 30 pats (1-year survival: 79 %). In the transplanted group, 1-year survival was 94 %. In the subgroup of toxic ALF, the recovery rate was 76 % and 23 % in the ALF of unknown origin. Camus et al. (2009) found similar results in their liver transplantation candidates. They treated two times/pat. for 8 h/session and found a transplantation-free survival of 29 %. A number of other groups reported safe and successful bridging to liver transplantation, including anhepatic phases, or even recovery of native liver function (Liu et al. 2004; Choi et al. 2005; Doria et al. 2006; Gaspari et al. 2006; Yuan et al. 2006; Chen et al. 2011; Pőcze et al. 2013), among others in infants and children (Trittenwein et al. 2006; Nadalin et al. 2007; Rustom et al. 2014). However, not all groups saw native liver recovery (Gaspari et al. 2006; Wai et al. 2007). In 2008 a multicenter randomized controlled trial in 102 ALF patients in France investigated the role of MARS as a bridge to liver transplantation. It found the method to be safe and potentially helpful to improve transplant-free 6-month survival (Saliba et al. 2013).

Prometheus was used safely in larger case series of ALF and bridging to transplantation (Grodzicki et al. 2009; Sentürk et al. 2010). For SPAD and TPE, successful cases of bridging to LTx in acute Wilson’s disease were described (Kreymann et al. 1999; Hilal and Morehead 2014).

For bioartificial systems, the largest reported randomized controlled trial investigating the impact of the HepatAssist system on the course of ALF and primary non-function after liver transplantation, the bioreactor improved survival in a subgroup analysis (Demetriou et al. 2004). There is no information on the use of ELAD as a bridging tool.

Both MARS and TPE seemed to be safe and feasible for the treatment of post-liver transplant graft dysfunction (Lee et al. 2010).

In children, Lexmond et al. (2015) found MARS safe and efficient even in very sick children. Only the sickest subgroup was bridged to transplantation. They had an outcome comparable to a less severely diseased subgroup. Another study suggested that in children with ALF, TPE combined with hemodialysis may be more advisable than MARS. However, all treatments were tolerated well (Schaefer et al. 2011).

Influence on survival was evaluated in a number of controlled randomized trials so far. In an HRS type I trial including 13 patients, significant improvement in survival in the MARS group was reported. Seven-day survival was 67 % in the MARS versus 0 % in the control group. Thirty-day survival was 25 % in the MARS group (Mitzner et al. 2000). In another study, in 24 AoCLF patients with severe cholestasis (mean bilirubin higher than 30 mg/dl), a significant improvement of 30-day survival was found (92 % in the MARS group vs. 50 % in the control group, p < 0.05) (Heemann et al. 2002). In the 3-year follow-up of a larger patient group of 149 patients with alcohol-induced AoCLF, a significant survival advantage (33 % vs. 15 %) was found as compared to standard of care (Hessel et al. 2010). A Cochrane Biliary Group analysis of liver support systems from 2003 found a significant 33 % reduction in mortality in AoCLF. This effect was mainly carried by the participating MARS studies (Kjaergard et al. 2003). However, the so far largest study performed with MARS in 189 AoCLF patients did not find a difference in survival (Banares et al. 2013). Also, the HELIOS trial investigating the impact of Prometheus on survival in 145 AoCLF patients found no overall survival benefit. However, in the subgroup of patients with MELD, >30 survival was significantly improved (Kribben et al. 2012).

Regarding acute liver failure, a multicenter randomized trial of MARS in 102 ALF patients fulfilling high-urgency liver transplant criteria in France found a nonsignificant trend toward improved survival in the MARS group and a significantly improved transplant-free 6-month survival in those patients treated with at least three sessions of MARS (Saliba et al. 2013). These results confirm smaller studies that have reported improvement in transplant-free survival in ALF patients treated with MARS (Koivusalo et al. 2005; Camus et al. 2009; Cisneros-Garza et al. 2014). The Helsinki transplant center has reported experience from over 150 ALF patients being treated with MARS. Authors concluded that the implementation of MARS has likely contributed to improve 6-month survival in both non-transplanted (40 % before vs. 66 % after MARS, P = 0.03) and transplanted (77 % vs. 94 %, n.s.) ALF patients (Kantola et al. 2011). A meta-analysis found significant impact of extracorporeal liver support on ALF survival (Stutchfield et al. 2011).

There is limited information on survival data for SPAD, TPE, and cell bioreactors. It appears that high-volume TPE can have a positive impact on survival in acute liver failure patients (Larsen et al. 2010). Bioartificial liver support systems have not demonstrated a convincing survival benefit to date (Demetriou et al. 2004). However, clinical trials especially with the ELAD system have been initiated in the last few years (Wertheim et al. 2012).

The use of MARS both in ALF and in AoCLF was analyzed with regard to its cost utility ratio. In the 3-year follow-up of 149 AoCLF patients, a significant survival advantage (33 % vs. 15 %) was found as compared to standard of care with a favorable cost-benefit ratio (Hessel et al. 2010). Kantola et al. (2010) compared 90 ALF patients treated with MARS from 2001 to 2005 and a historical control group of 17 ALF patients treated from 2000 to 2001. The 3-year outcomes and number of liver transplantations were recorded. Compared to the controls, the average cost per quality-adjusted life year (QALY) saved was considerably lower in the MARS group (64,732 euros vs. 133,858 euros) within a time frame of 3.5 years. The authors concluded that MARS treatment combined with standard medical treatment for ALF in an ICU setting is more cost-effective than standard medical treatment alone.

Last but not least, in an effort to lower treatment costs, Drexler et al. (2009) determined the optimal dialysate albumin amount to be 100 g rather than 120 g per session, as is the clinical standard today.

Liver support should be considered in AoCLF patients not responding to standard of care within several days. In ALF with a high expected mortality rate, commencement of liver support treatment is recommended as soon as the diagnosis is made. ALF and AoCLF represent rather different indications for liver support, and therefore, different inclusion and exclusion criteria need to be applied. The absence or presence of sepsis and severe disseminated intravascular coagulation seem to divide AoCLF patients in good and bad candidates for MARS. We recommend early and sufficiently aggressive antibiotic treatment of infections as well as antibiotic prophylaxis in those not infected. In AoCLF, very low platelet count (<50 Gpt/l), high INR (>2,3), and advanced kidney failure needing dialysis or hemofiltration represent high-risk patients that might not take advantage from treatment. An Italian study found that age, male gender, and sequential organ failure assessment score (but not model for end-stage liver disease score) were factors predicting death, whereas the number of MARS sessions and the increase in hepatocyte growth factor proved protective factors (Donati et al. 2014). Inderbitzin et al. (2005) found a critically low plasma disappearance rate of indocyanine green of ≤5 %/min at baseline to be correlated with unfavorable outcome.

In AoCLF, total dosage of treatment should be handled flexible with days of pausing in between, especially if the platelet count is decreasing to values below 50 Gpt/l or INR going above 2.3. The mode should be rather intermittent than continuous with treatment lengths of 6–8 h per day. In ALF the need for treatment is much bigger and probably continuous treatment with few breaks is most efficient. In ALF much worse INR values can be tolerated than in AoCLF, probably due to the different pathogenesis of INR increase (synthetic defect vs. hypercoagulation). Cautious anticoagulation preferably with citrate or small doses of heparin is recommended, whereas no anticoagulation is not advisable for most patients (Faybik et al. 2006; Tan et al. 2007; Yuan et al. 2011; Meijers et al. 2012). Thromboelastography reveals patients at risk for bleeding (via detection of fibrinolysis), while absent thrombocytopenia and elevated plasma fibrinogen predicted clotting of the MARS system (Bachli et al. 2011). For Prometheus, problems with clotting due to direct adsorption of protein C and S to anion exchanger column were described (Meijers et al. 2007). While heparin anticoagulation is not advisable, the use of citrate was found to be safe (Rifai et al. 2008). In patients with high risk of bleeding, addition of TPE before, e.g., albumin-cleansing methods should be considered (Huang et al. 2012; Ince et al. 2013).