Review article: liver support systems in acute hepatic failure


Hodgson Department of Gastroenterology, Imperial College School of Medicine, Hammersmith Hospital, Du Cane Road, London W12 0NN, UK. E-mail:


The treatment of acute hepatic failure has developed rapidly over the last 40 years, reducing morbidity and mortality from this syndrome. Whilst this has been partly attributed to significant improvements in the specialist medical management of these patients, advances in surgical techniques and pharmaceutical developments have led to the establishment of successful liver transplantation programmes, which have improved mortality significantly.

This review will examine the clinical impact of alternative methods that have been used to provide extra-corporeal hepatic support. Non-biological, bio- logical and hybrid hepatic extra-corporeal support will be explored, offering a comprehensive historical overview and an appraisal of present and future advances.


Hepatic failure has long been a challenge to the physician. Both acute and chronic hepatic failure present a spectrum of clinical problems associated with high morbidity and mortality. This review surveys changing perspectives in the treatment of acute hepatic failure (AHF).

The development of hepatic encephalopathy, jaundice and coagulopathy defines AHF. Following Lucke and Mallory’s description of two distinct clinical patterns observed in acute hepatitis, Trey & Davidson1, 2 introduced the terms fulminant and sub-fulminant hepatic failure. These have since been modified to make the classification more accurate and universal. Their series and other similar reports have helped to inform and familiarize clinicians. Early diagnosis, development of specialist centres and better understanding has reduced mortality. Advances in intensive care monitoring, management and pharmacological therapy have made a significant impact on survival. Whilst liver transplantation remains the only definitive treatment for patients, improved techniques and immuno-suppressive regimes have improved long-term survival post-transplantation. Organ availability, however, remains a limiting factor. The clinical aspects of the management of AHF have recently been reviewed in this journal.3

Over the last 50 years there have been advances in the understanding of the deranged pathophysiology encompassing the clinical features of AHF: encephalopathy, cerebral oedema, haemorrhage, electrolyte and metabolic disturbance, renal failure, cardiovascular instability and increased risk of infection. The ‘toxin hypothesis’ and the ‘critical mass theory’, originally thought to explain the changes, have been modified by the realization that synergistic forces (endotoxin and cytokines) are also involved in AHF, and these views have led to a range of approaches to artificial liver support.

Advances taking place in the treatment of chronic renal failure were trialled in patients with chronic liver disease and began an era of ‘non-biological liver support’. Areas investigated include haemodialysis, haemofiltration, the use of adsorbents, plasma exchange and later plasmapheresis. With the limited impact that these systems provided, alternative approaches were also investigated. The use of extra-corporeal perfusion of xenogeneic and allogeneic cadaveric organs marked the advent of ‘biological liver support’. The effects of xenogeneic liver homogenate, fresh liver slices, freeze-dried liver granules and whole organs on patients in AHF have been reported. Subsequently, technological advances in cell biology and biotechnology allowed more sophisticated systems to be developed. More recently a class of ‘hybrid liver support’ has undergone clinical trials involving the use of biological tissue with non-biological materials. Several investigators have devised complex systems that combine human and pig cryopreserved hepatocytes, providing the synthetic, metabolic and excretory functions of the failing liver with advanced biotech constructs. Preliminary reports have identified the feasibility of the approach, but effectiveness has not yet been proven.

Finally, also in its infancy, ‘hepatocyte transplantation’ has been shown in animal studies to be effective in the treatment of AHF.

This review will summarize previous and current endeavours in artificial liver support.


The unique and complex architecture of the liver goes some way to explaining its diverse involvement in maintaining metabolic homeostasis within the body. AHF leads to deranged intracellular metabolism and failure of interconversion of carbohydrates, lipids and amino acids, and reduced synthesis of plasma proteins, coagulation factors and lipoproteins. The loss of detoxification and biotransformation increases susceptibility to further damage and may lead to an increased incidence of infection. These metabolic abnormalities are thought to cause the unique clinical features seen in AHF.

Decades of investigation have focused on two hypotheses that attempt to link the metabolic changes to the clinical features of AHF. The ‘toxin hypothesis’ suggests that the failing organ is unable to clear toxins normally processed by the healthy liver from the bloodstream. Animal and human studies of AHF have identified the presence and deleterious effects of ammonia, phenols, mercaptans, aromatic amino acids, fatty acids, cytokines and nitric oxide moeities. The ‘critical mass hypothesis’ suggests that a profound loss of hepatocellular metabolic capacity below a critical threshold leads to end organ dysfunction and failure to support peripheral organ function. More recent investigation suggests that the hepatocyte itself, once injured, contributes to the amplification of liver injury. The reduced integrity of the cell membrane, the loss of intracellular homeostasis, imbalance of pro-oxidant and antioxidant pathways, mitochondrial damage and depletion of ATP all contribute to the generation of toxic species, decreased cytoprotective capabilities and alterations in cell-to-cell interactions.

Both animal and human studies have shown that, under extreme circumstances, total hepatectomy improves the clinical stability of the subject in AHF.4

The role of artificial hepatic support is therefore a complex one. It must encompass several roles, i.e. the removal of toxins incriminated in the pathobiology of AHF, the synthesis of products such as coagulation factors, albumin and other plasma proteins, and it must also attempt to reverse the massive inflammatory process taking place within the failing organ. Ideally, hepatic support should provide metabolic, synthetic and detoxification functions, allowing time for recovery and regeneration of the host organ. Where regeneration is not possible, hepatic support may allow time for organ transplantation to take place.



Inspired by successful advances taking place in the management of chronic renal failure in the 1950s, extra-corporeal technology was applied to patients with chronic liver disease. Haemodialysis requires a semi-permeable dialysis membrane through which fluid and small solutes may pass, whilst exchange occurs against dialysis fluid. Transfer of solutes and molecules occurs by diffusion.

Haemodialysis had been shown to improve survival in patients with chronic renal failure and other diseases such as porphyria.5, 6 Zysno et al.7, 8 demonstrated improvements in EEG recordings that correlated with clinical improvement in severely uraemic patients. This principle was applied to a group of chronic cirrhotic patients, demonstrating improved encephalopathy and an associated decrease in arterial ammonia levels.9 In 1968 a patient in acute liver failure was treated with haemodialysis and showed temporary improvement in their clinical status following treatment.10 (Table 1 illustrates the various studies discussed.)

Table 1.  . Non-biological hepatic support in man Thumbnail image of

Success was also reported by Oules et al.,11 who suggested that removal of soluble, uncharacterized compounds in the ‘middle’ molecular weight range (400–2000 Da) improved encephalopathy in patients with chronic liver disease and also the clinical status of patients with chronic renal failure.12 However, these observations were challenged by several investigators, who suggested that the toxaemia of AHF was not due to middle molecules, based on the results of a larger trial showing minimal improvement in clinical state.

Application of haemodialysis became more widespread in the early 1970s and reports of improvement in lactic acidosis13 and of improved survival (individual case reports) following paracetamol, salicylate and benzodiazepine overdoses also appeared. Advances in membrane technology allowed Opolon et al.12 to demonstrate improved neurological status in animal models of AHF using polyacrylonitrile (PAN) membranes. This was followed by a clinical trial using PAN membranes in 1975,14 in which five complete neurological and two partial recoveries were reported in a group of nine non-decerebrate patients in coma due to ‘acute hepatic atrophy’. No improvement in survival was demonstrated. Similar findings were reported by Denis et al.15 in 41 patients with fulminant hepatic failure who had between them 180 treatment periods. Seventeen patients had complete neurological recovery from coma and seven had partial recovery. Of those that had complete neurological recovery, nine survived.

Whilst PAN membranes allowed increasing solute removal, it and other similar polymers such as polysulphone were unable to remove protein-bound and lipophilic substances. An alternative approach was developed using a liquid membrane that allowed the removal of lipophilic toxins but prevented the removal of physiological substances such as hormones. The structure consisted of a hydrophobic polysulphone membrane with large voids containing paraffin oil. Blood passed through the liquid membrane filter with a sodium hydroxide acceptor solution on the other side. Protein-bound toxins contacting the membrane were released into the oily layer, diffused through the membrane and became water-soluble due to a reaction with the accepting solution. Application in pigs has shown improvement in neurological function.16

Haemodialysis was further adapted by Stange & Mitzner,17 who introduced the Molecular Adsorbents Recirculating System (MARS) in 1993. It had been well known that some toxins were avidly protein bound and that standard haemodialysis membranes did not remove these. The polysulphone membrane was impregnated on both sides with albumin and dialysis took place against a closed loop dialysate also containing 10–20% albumin. In vitro studies have shown enhanced removal of protein-bound toxins such as bilirubin, bile acids, tryptophan and fatty acids. The success of this technique is dependent on complex physiochemical processes involving specific trilateral interactions between ligands, binding proteins and the polymer. More recent clinical studies have demonstrated improved clinical parameters (encephalopathy and renal function) with reductions in bilirubin, urea and creatinine in eight patients with decompensated chronic liver disease following a total of 47 single treatments.18, 19

Despite the improvements in technology and apparent effects on encephalopathy, there is no proof of effectiveness in altering survival. Haemodialysis is therefore not currently recommended for use in the treatment of AHF. Haemodynamic instability may occur, as well as electrolyte and fluid shifts, and these changes can exacerbate cerebral oedema, leading to increased intracranial pressure; they may also have adverse effects on cardiac output and other haemodynamic parameters.20, 21 High-volume haemodialysis has, however, been used together with plasma exchange in AHF, as discussed below.


Haemofiltration has been shown to be more suitable than haemodialysis for the treatment of some patients in renal failure following sepsis, trauma and liver failure. Haemofiltration uses a permeable membrane and relies on continuous convective solute removal, avoiding large volume and electrolyte shifts. There is no dialysate fluid, only a substitution solution replacing the ultrafiltrate. Convective removal of solutes has been shown to be more efficient than haemodialysis for the removal of molecules with molecular weights of 2000–3000 Da. This range corresponds to the ‘middle’ molecules thought to be involved in encephalopathy and liver failure.

Continuous venous–venous haemofiltration is preferred in AHF because it offers haemodynamic stability and allows predictable and gradual control of metabolic disturbance.21, 22 The substitution solution for AHF can be lactate-free and benefits have been observed using a bicarbonate-buffered solution.23 Venous–venous circuits are commonly used and require anticoagulation with heparin, low molecular weight heparin or prostacyclin if the platelet count is low.24 Potentially, the beneficial effects include efficient removal of middle and indeed larger molecules, allowing immuno-modulation with the removal of vaso-active substances from the circulation, including immunoglobulins, IL-1, IL-6, TNF-α, C1q, C3a, C5a and platelet activating factor.25[26]–27 Removal of these may be advantageous in sepsis, multi-organ failure and systemic inflammatory response syndrome, all variants of that seen in AHF. Bellomo et al. described high-volume exchange haemofiltration of up to 6 L/h, which has demonstrated improved survival in animal studies of sepsis.27 This has recently been demonstrated in a clinical setting.28


There are advantages also in haemodiafiltration, which combines the advantages of both convection (removal of larger molecules, 2000–3000 Da) and diffusion (removal of smaller molecules, 400 Da or less). Yoshiba et al.29 used this technique in an uncontrolled series of patients with AHF, sub-acute hepatic failure and late onset hepatic failure, demonstrating survival rates of 85, 54 and 50%, respectively. Adverse effects included complement activation, activation of the coagulation cascade and release of vaso-active and chemo-attractant fragments.


There has been considerable focus on the use of different adsorbents as a part of non-biological hepatic support systems to remove toxins or middle molecules thought to contribute to the clinical manifestations of AHF. Most work in this area has involved the study of three major types of sorbents, including activated charcoal in various forms, synthetic neutral resins (XAD-2,4,7) and anion exchange resins (Dowex-1). This work has recently been further developed with the combination of charcoal and a cation exchange resin in the form of the Biologic-DT system.55[56][57]–58


Yatzidis30 demonstrated the adsorbent properties of coconut charcoal in a haemoperfusion circuit. This provided effective removal of creatinine, uric acid and guanidine from uraemic dogs. Removal of phenobarbitone in dogs and in vitro and in vivo work in pigs demonstrating the removal of paracetamol using charcoal haemoperfusion led to isolated clinical case reports.31[32]–33 Problems with biocompatibility of charcoal encouraged adaptations of its use. Chang34 used microencapsulated charcoal for haemoperfusion in AHF, recording complete recovery of consciousness from hepatic coma in a patient with alcoholic hepatitis. Microencapsulation stops fine particles of charcoal escaping into the circulation, thus improving its biocompatibility. Gazzard et al.35 reported the use of charcoal haemoperfusion in 22 cases of AHF with grade IV hepatic coma in an uncontrolled trial. Recovery of consciousness and improved survival was demonstrated. Subsequently several groups attempted to repeat these observations, but were unsuccessful, encountering problems such as intractable hypotension, charcoal particle emboli, and thrombocytopaenia.36, 37

To improve the biocompatibility of charcoal Weston et al.38 examined the advantages of coated and uncoated charcoal. However, little difference in leucocyte and platelet loss was demonstrated. Gelfrand et al.39 examined the use of charcoal coated with acrylic hydrogel, which resulted in improved clinical tolerance to haemoperfusion. Uncontrolled clinical trials showed improved levels of consciousness in 9 of 10 patients in hepatic coma and complete recovery in 40%.

Gimson et al.40, 41 used prostacyclin to prevent platelet activation in clinical trials of charcoal haemoperfusion. Two groups were studied, those in grade III and those in grade IV encephalopathy. The use of charcoal haemoperfusion in the grade III group led to a better rate of survival (65%) and a reduced incidence of cerebral oedema (49%) as opposed to those in the grade IV group (20% survival, 78% cerebral oedema). Contrasting findings were reported in the largest controlled trial of charcoal haemoperfusion undertaken by O’Grady et al.42 One hundred and thirty-seven patients were entered into two controlled trials run concurrently. Trial A randomized 75 patients in grade III encephalopathy to 5 or 10 h of charcoal haemoperfusion. There was no significant benefit seen in survival, incidence of renal failure or cerebral oedema in the two groups. Trial B randomized 62 patients in grade IV encephalopathy to charcoal haemoperfusion or no haemoperfusion. Survival rates were similar (39.3 and 34.5%, respectively). The conclusion was that charcoal haemoperfusion had little influence on survival or incidence of cerebral oedema and renal failure. The additional conclusion was reached that the apparent effectiveness of the approach, noted in uncontrolled trials during the 1970s and early 1980s, was due to the use of historical controls and in particular reflected the progressive general enhancement of intensive care of patients in severe liver failure.

Other adsorbents.

The properties of other adsorbents have been explored in an attempt to overcome the biocompatibility problems encountered with charcoal. The use of Dowex 50-X8, a cation exchange resin, was reported in dogs with hepatic coma.43 Blood ammonia levels were reduced by 50% and concurrent clinical improvement encouraged its use in humans. Small uncontrolled trials showed reversal in coma and reduction of ammonia levels in 20% of patients.44 Dowex 50-X8 and Amberlite IR 120 (another cation exchange resin) were trialled in 13 patients with hepatic failure, with reports of improvement in conscious level in 54% of patients.45[46]–47 Amberlite XAD-7, an albumin-coated macro-reticular resin, was shown to reduce plasma total bilirubin in 19 patients with AHF, of whom eight left hospital.48 Its ability to remove TNF-α, IL-6 and IFN-α, as compared with charcoal, was also demonstrated.49, 50

More specific resins have been developed and tested with recorded improvement in clinical parameters and survival. Examples include polyamine triglycidylisocyanurate (PAT) resin, polylysine-immobilized chitosan beads for removal of bilirubin, and polymyxin B immobilized on polystyrene fibres for the removal of endotoxin.51[52][53]–54 The last two in particular have been associated with biocompatibility problems and despite, or perhaps reflecting, the multiple types trialled, no adsorbents have come into widespread clinical usage.


The Biologic-DT system55 is currently in use at a small number of centres. This system combines charcoal and cation exchanger in a system called ‘haemodiabsorption’. Blood is dialysed across a parallel plate dialyser with a cellulose membrane that has sorbent present at the membrane surface. The sorbent contains powdered charcoal, sodium and a loaded cation exchange resin.

Animal studies have examined the effect of the Biologic-DT system in dogs with AHF induced by a two-step devascularization procedure. Following 6 h of treatment the treated animals were more physiologically stable, developed less lactic acidosis, had reduced transaminase increases and also maintained higher blood glucose levels than the untreated controls.56 Uncontrolled clinical trials in patients with hepatic failure have recorded improved neurological status and normalization of diastolic blood pressure in treated patients.57, 58 Controlled trials, however, have shown no benefit in survival in AHF but plasma lactate, creatinine and bilirubin were reduced. Increased thrombocytopaenia, decreased fibrinogen and an increase in the activated clotting time were, however, seen in the Biologic-DT treated group. Neurological improvement has been recorded in chronic decompensated patients.59

Plasma exchange and plasmapheresis

The removal of toxins from plasma and replacement with plasma from healthy individuals was first performed by Lepore et al.60, 61 They reported neurological improvement prior to death in two patients out of nine with AHF. This followed one to 12 treatments with 10–83 L of plasma exchanged. Patient survival was demonstrated by Buckner et al.,62 who treated four patients with 10 L exchanges per day for 3–36 days. Three patients survived in this uncontrolled trial. The development of membrane separators led to several groups claiming increased survival.63, 64

In 1985, Winikoff et al.65 suggested that plasmapheresis was only a bridge to liver transplantation (OLT). Berk53 demonstrated the theoretical advantage of high-volume plasmapheresis, suggesting that the volume of distribution of toxins would correspond to the extracellular space. The development and subsequent use of high-volume plasmapheresis (HVP) for the treatment of AHF has since been refined.

Kondrup et al.66 investigated the effect of HVP in 11 patients with AHF, all initially in grade III or IV encephalopathy. An average of 2.6 exchanges, each with a mean volume equivalent to 16% of body weight, were performed. Five of the 11 patients, all with acetaminophen poisoning, survived. The six non-survivors remained haemodynamically stable during treatment for a mean of 6.9 days. The conclusion was that HVP could be used as a bridge to transplantation, and also that patients who have residual liver function may be supported until their own organs recover.

A larger series of 40 patients (mixed aetiologies) receiving HVP reported 28 survivors (54%). Seventeen patients received OLT, of whom three later died. The documented improvement was in parameters of cerebral blood flow (CBF), cerebral perfusion pressure (CPP), cerebral metabolic rate for oxygen (CMRO2), and mean arterial pressure (MAP). No episodes of raised intracranial pressure (ICP) were reported. Systemic vascular resistance (SVR) increased and cardiac output (CO) appropriately decreased, an effect lasting 12 h. These changes imply improved tissue oxygen extraction systemically and in the brain.67 Similar haemodynamic findings were reported by Clemmesen et al.20 who suggested that the removal of a humoral factor by HVP may explain these results.

The effect of HVP on intracranial parameters, CBF, CMRO2 and oxygen extraction has been reported in 12 patients with AHF. Encephalopathy was seen to improve in four patients and improvements in CBF and CMRO2 were statistically significant.68 This improvement was thought to reflect partial removal of neuroinhibitory plasma factors.

There is therefore currently some evidence that HVP improves haemodynamic parameters, reduces cerebral oedema and prolongs survival in those awaiting OLT. It may be that it can be effectively used to sustain life in a small group of patients that have some recovering hepatic function, and might be of benefit in patients who are not awaiting transplantation.


Xenogeneic and allogeneic extra-corporeal support have been investigated for 50 years. Several techniques have been used, all of which have had only limited success (Table 2). Approaches that have been explored include cross-haemodialysis with animals, and extra-corporeal haemoperfusion with pig, baboon and human livers. Haemoperfusion has been extended to pass blood over liver fragments and also liver cells. All techniques performed have shown some improvement in clinical and/or physiological parameters but not improvement of survival in controlled trials.

Table 2.  . Biological hepatic support in man Thumbnail image of

Xenogeneic support has the disadvantage that naturally occurring antibodies in the human recipient will react with endothelial expressed antigens in the perfused liver. This rapidly leads to complement activation, activation of the clotting cascade, and haemodynamic instability.


Kimoto performed a technique called cross-heterohaemodialysis in 1957, which involved haemodialysis in a man in hepatic failure against the circulation of a live dog.85 This resulted in improvement of ammonia levels. This technique was further adapted with the addition of an ion exchange column by Hori et al.69, 70 who performed cross-haemodialysis between a man in AHF and a dog separated by a semi-permeable membrane. This was associated with recovery from hepatic coma and a corresponding reduction in blood ammonia.

Extra-corporeal haemoperfusion

Eiseman et al.71 in 1965 described the first experience of heterologous liver perfusion in the treatment of liver failure. This technique demonstrated bile flow, galactose elimination, clearance of ammonia and bilirubin in a pig liver extra-corporeally perfused with human blood. Abouna72, 73 repeated this technique using pig livers on 10 patients with AHF, seven with acute hepatitis and liver cell necrosis and three with pre-existing liver disease. Improvements in conscious state and survival greater than 4 days was recorded, but only two patients survived.73 Waldschmidt et al. haemoperfused nine patients with AHF using isolated pig livers and reported one survivor, and transiently improved conscious levels of the other patients who later died.74

Isolated baboon livers were also used to provide extra-corporeal support in the treatment of 14 grade IV encephalopathic patients. Thirteen patients had viral hepatitis and one had developed drug-induced AHF. Thirty perfusions were performed (29 baboon livers and 1 human cadaveric liver) each lasting 5–27 h, with 1–4 perfusions per patient. Bile excretion peaked at 7–8 h and was used as a marker for successful perfusion. Lie et al.75[76]–77 claim a 50% survival rate of patients with AHF, when perfused by this technique. Baboon livers were also used by other groups who have claimed survival rates of 34–45%. No controlled trials exist to compare treatment regimes.78

Extra-corporeal liver perfusion using human livers procured and unused by the United Network for Organ Sharing was reported in three patients with AHF in 1993.79 All patients showed improved serial serum bilirubin and arterial ammonia values, while two of three patients also showed marked neurological improvement. These patients were later successfully transplanted. The third patient failed to show clinical improvement and died 7 days after the treatment was discontinued.

Common problems encountered with xenogeneic haemoperfusion were: activation of the clotting cascade, antibody and immune complex formation, and the expected sequelae. This limited the length of perfusion time per session and its repeated use. Technical problems also include the requirements of large numbers of trained surgical, nursing, technical and ancillary staff and the supply of animals for each haemoperfusion session. In the case of cadaveric human liver extra-corporeal perfusions, organs and transplant surgeons are of limited availability.

Transgenic organ transplantation

Despite problems with organ availability, immunological incompatibility, hyperacute rejection and ethical dilemmas, the role of xenogeneic transplantation is an area of current interest. The development of transgenic animals with genetically modified immune expression to reduce incompatibility for use in organ transplantation is now the subject of ethical debate and approval. Manipulation of the animals’ genome involves the insertion of human anti-complement genes and modification of their xenoantigens. Preliminary studies have perfused human blood into isolated transgenic pig livers expressing human complementary–regulatory protein, human CD59 and human decay-accelerating factor (hDAF). Results show reduced complement and TNF-α expression and reduced activation of the hyperacute rejection process.80[81]–82 Expression of porcine MHC class II is a potent stimulator of human CD4+ T-cells and is thought to be dependent on class II transactivator (CIITA), a bi- or multifactorial domain protein. Manipulation of this protein has successfully demonstrated potent suppression of MHC class II expression and may lead to prolonged graft survival.83

Practical use at present still remains some way off. Even without the complexities that xenotransplantation introduces (such as the potential for spread of viruses from animal to man), there is much that needs to be addressed with respect to perfusion of whole livers extra-corporeally, necrosis, apoptosis, ischaemia and reperfusion injury, and allogeneic rejection.84


Hybrid hepatic support combines the use of biological tissue with the use of non-biological materials. The use of hepatic tissue may provide synthetic, excretory and biotransformational functions, which combined with membranes removing cytokines and other toxins, is thought to be beneficial in AHF.

Recent advances in the isolation and characterization of hepatocyte function and growth, and the requirements for in vitro function, have allowed the use of colonies of cells in hybrid liver support. Bioreactor designs have changed to allow for optimum integrity and function of cells taking into account the complex biophysics involved in oxygenation, removal of waste products and flow dynamics.

The concept of hybrid support was introduced in the last section, and was developed by the pioneers Kimoto and Hori.85, 86 Their techniques were used by others and cleverly adapted by Mikami et al.87 and Nose et al.88 An extra-corporeal circuit was used to assess the function of liver tissue homogenate, fresh liver slices and freeze-dried granules of liver tissue in clinical trials. Patients were connected to a dialysis circuit, and blood was then passed to a ‘metabolic circuit’ which contained a gel-type cellulose membrane, a bubble oxygenator and a chamber or ‘bioreactor’ containing either liver homogenate, liver slices or freeze-dried granules. Limited clinical application demonstrated stable glucose and lactate levels and a reduction in ammonia levels. The use of liver fragments has been further developed and now colonies of cells are being used in similar but more sophisticated artificial liver support systems. Parallel advances have been taking place in biotechnology, study of cell growth, function, integrity and manipulation. Cells may now be genetically manipulated, cell lines immortalized and cryopreserved.


Ideally, liver cells maintained in an extra-corporeal environment should express the full functional repertoire of the normal liver. As the liver contains various sub-populations, not only the major metabolically active cells, i.e. the hepatocytes, but also Kupffer cells, sinusoidal endothelial cells and stellate (mesenchymal-derived) cells providing extracellular matrix and growth factors should ideally be present. This is clearly a major challenge. However, there is general consensus that the most important functions of an extra-corporeal liver circuit, i.e. detoxification and synthesis, are largely (although not totally) fulfilled by the hepatocytes.

Calculations indicate that there are of the order of 1–2 × 1011 hepatocytes in a normal adult human liver. Data from surgical resections indicate that probably one-third of this number is sufficient for normal survival, but of course a larger number may be required to reverse the changes of hepatic failure. Of course, data from surgical experience also defines the numbers of cells that are adequate when they are in the optimum, i.e. natural, configuration as components of a normal liver.

The first major issue in providing hepatocytes for such systems is that the mature adult hepatocyte is a non-dividing cell. It is capable of entering DNA synthesis and dividing in vivo, but in vitro there are only specialized circumstances in which limited rounds of cell division can be achieved. The option of taking a limited number of adult human hepatocytes and letting them proliferate in culture to provide the number required, is currently not available. Scientific approaches to the limited replicative potential of adult hepatocytes in vitro involve identifying growth factors and media that will allow proliferation, and/or genetically manipulating cells to remove the normal checks to division in the adult cell.89 The use of foetal or neonatal hepatocytes with a greater proliferative rate is also being explored.90 However, until such techniques are truly successful, workers have to use one of two options—either utilizing cell lines from human livers, or using primary cells from other species. Current evidence indicates that any proliferating hepatocyte cell line from a human liver falls short in some respects from a fully functional repertoire. On the other hand the use of xenogeneic cells has the same problems as have been alluded to under xenogeneic haemoperfusion, due to naturally occurring antibodies. It is possible, however, with the use of diffusion barriers to prevent direct exposure of xenogeneic cells to human antibodies and cells. This should delay immediate deleterious immune responses whilst a patient is connected to an extra-corporeal circuit, but the diffusion barriers which prevent exchange of proteins and protein-bound small molecules will also limit the effectiveness of functional replacement.

The second set of issues in providing adequately functioning hepatocytes derives from our expanding knowledge of the requirement for maintenance and expression of fully differentiated function. Primary hepatocytes in the most straightforward configuration of tissue culture—as a monolayer—rapidly lose differentiated function. A variety of techniques can maintain function: co-culture with non-parenchymal cells, culture as multilayer spheroids and exposure of cells to extracellular matrix proteins. In essence, maintaining hepatocytes with normal occupancy of their surface integrin receptors,91 and normal cell-to-cell contact, is probably the critical feature for maintaining function. This can be achieved by, for example, encapsulating cells in substances such as alginates;92 such techniques also tend, by producing compact cell masses, to reduce the volume required to hold 3 × 1010 cells and thus render extra-corporeal circuits practicable. However, too compact a cell mass imperils transfer of nutrients, cell products, and, perhaps most importantly, oxygen. Thus the micro-design of the cell aggregates, and the macro-design of the bioreactor, are of fundamental importance.

Bioreactor design

The successful development of the artificial liver is dependent on the hepatocyte component, matrix support and bioreactor design. Much has been invested in the design of the ‘bioreactor’, which will allow optimum cell culture, storage and cell integrity allowing its practical use in the circuit designed for support.

Three general forms of hepatocyte culture have been identified: suspension culture, attachment culture and hepatocyte spheroid culture. Suspension culture is the least effective method as hepatocytes lose function rapidly.93 Advantages, however, include low gradients of nutrients, metabolites, toxins and oxygen, allowing high transfer efficiency.94, 95

Attachment culture has been shown to maintain cell integrity and function.96 This has been used extensively in artificial liver support systems that use hollow fibre ultrafiltration cartridges. These devices rely on transmembrane diffusion for exchange of metabolites and this may lead to reduced efficiency.

Hepatocyte spheroid culture allows optimal distribution of media around hepatocytes, allowing high mass transfer efficiency. This may be reduced when cells have been encapsulated and/or coated.93, 97, 98

The choice of culture suspension will influence the design, materials and construction of the bioreactor as efficient attachment and growth will be required. Three basic designs have emerged; bioreactors for use with suspension culture, bioreactors based on cell immobilization and those with membranes.

Most bioreactors have used capillary membranes within a cartridge for cell attachment. Capillary membranes allow a number of other functions to take place (gas exchange, substrate supply and waste removal) efficiently and with practical ease. Hepatocytes may be seeded, cultured and grown within capillary membranes and perfused in the extra-capillary space providing mechanical and physiological protection from toxic blood or plasma. This approach has been used by Nyberg et al.99, 100 in the three-compartment gel entrapment bioreactor. This entraps porcine hepatocytes in a collagen matrix inoculated into the capillary lumen spaces of two 100 kDa molecular mass cut-off hollow fibre bioreactors.

An alternative approach is to construct a bioreactor with hepatocytes in the extra-capillary space with capillary membranes providing the in-flow and out-flow of media, oxygen, nutrients, toxins and waste. Capillary membrane constructions rely on transmembrane diffusion for mass transfer and so choice of materials is also of paramount importance.

Construction of the three artificial liver support devices that have had clinical exposure will be reviewed: Sussman’s Extra-corporeal Liver Assist Device (ELAD), Demetriou’s Bioartificial Liver (BAL) and Gerlach’s hybrid liver support system.

ELAD, developed by Sussman et al.101 in 1992, incorporates the C3A cell line. This is a highly differentiated clonal population isolated from a human hepatoblastoma cell line, HepG2. Two hundred grams of cells were originally seeded and grown in the extra-capillary space of a haemodialysis cartridge containing ≈ 10 000 hollow fibres with a surface area of 2 m2. Two to four weeks is required for adequate numbers of cells to have grown for clinical use. They may then be stored and studies have shown good function (glucose, albumin secretion) after 8 months. The intra-capillary space is used in the growth period for culture medium and oxygen supply and later in clinical use this space is used for perfusion of blood. Diffusion gradients and mass transfer is felt to be efficient as fibres are approximately four cells apart, allowing adequate oxygenation and waste removal. The membrane has a molecular weight cut-off of 70 000 kDa, protecting hepatocytes from flow trauma, white cells and immunoglobulins, but allowing middle molecules and ammonia to pass across the membrane.

Recent modifications include two main design changes, i.e. increased cell numbers used per cartridge (700 g) and adaptation of the circuit so that it can be perfused with plasma and not blood (1999).

The BAL system, developed by Demetriou and co-workers, is conceptually very similar to ELAD but originally had three major differences: the cell source, i.e. primary pig hepatocytes rather than a human tumour derived cell line; the perfusate, which is plasma rather than blood; and the presence of a charcoal column filtering the plasma prior to its entry to the bioreactor.102, 103

Hepatocytes are isolated from pigs and attached to collagen-coated dextran microcarriers. More recently cryopreserved cells have been used. The hollow fibre bioreactor consists of a polycarbonate cylinder containing cellulose nitrate/cellulose acetate porous fibres. Fibres have a pore size of 0.2 μm and a total internal surface area of about 6000 cm2. The total extra-capillary surface area is 7000 cm2.

The BAL system comprises a plasma separator, generating plasma (80–105 mL/min) from venous blood and passing this to the charcoal column. The plasma is then directed across the bioreactor at high flow rates (220–500 mL/min) which allows several passes before it is passed back to the individual.

Gerlach et al.97, 98, 104 described a more sophisticated hybrid liver support system. The structure is housed in a polyurethane PUR 725 case. The bioreactor is made up of several interwoven, independent polyurethane capillary systems, entering and leaving the bioreactor in four discrete bundles and each serving a different function. The four capillary bundles provide plasma in-flow, oxygen supply and carbon dioxide removal, plasma out-flow, and sinusoidal endothelial co-culture. Hepatocytes (pig hepatocytes) are seeded in the extra-capillary space and find all types of bundles locally, thus reducing transmembrane diffusion gradients. The design can be adapted by the addition of further bundles, allowing additional functions to be incorporated. The design is such that many identical capillary units supply only a few hepatocytes, thus allowing small diffusion gradients.


Both ELAD and BAL have been trialled extensively in vitro, in vivo animal work and also in clinical trials. (Table 3 illustrates the clinical trials.)

Table 3.  . Hybrid hepatic support in man Thumbnail image of


Kelly et al.105 performed the first in vivo trials in 1992. Six male beagles had portacaval shunts inserted and then underwent a total hepatectomy. Dogs were kept sedated and were given sodium bicarbonate and glucose maintenance infusions to avoid hypoglycaemia. ELAD was perfused by vascular access via the internal carotid artery and the external jugular vein. Blood was driven through ELAD solely by arterial pressure and maintained at a flow rate of 80–100 mL/min. The circuit was heparinized to avoid clot formation. Of the six beagles, three control dogs lived for 3[4]–5 h after surgery but did not awake from the anaesthesia. Two of the three hepatectomized dogs also died within 3[4]–5 h of surgery but some did require extra sedation. The third hepatectomized dog survived 125 h post-surgery. Plasma ammonia levels were seen to increase following surgery in control animals, and this was felt to be a pre-terminal event. This was not seen in the ELAD-treated dogs.

Further work was carried out on an improved model of AHF using intravenously administered acetaminophen.106 Control animals developed AHF with worsening encephalopathy, hypoglycaemia, prolonged prothrombin time and severe transaminitis. Death occurred in all controls at 15–30 h. Dogs treated with ELAD for 42–48 h had a survival rate of 80%.

Technical modifications were made before the system underwent trials in man. These included a switch to a venous–venous circuit and also an increase in flow rate that allowed the formation of an ultrafiltrate, which would then bathe the cells within the extra-capillary space. Filters were inserted to stop any cell debris entering the patient’s circulation, as some concern was raised over the use of a hepatoblastoma cell line metastasizing in the patient.

Initial trials between 1991 and 1993107[108]–109 were carried out in four centres and 11 AHF patients were treated with ELAD. All patients were in grade III/IV encephalopathy, nine of whom required mechanical ventilation. One patient had been made anhepatic following primary graft failure. The age range was 9–63 years old and treatment times ranged between 9 h and 144 h. Encephalopathy improved in eight out of 11 patients and renal function was preserved in those who were not anuric at the start of treatment. Improvement in the galactose elimination capacity was demonstrated. Of the 11, six patients died, four received liver transplantation and one survived without transplantation.

Following the initial studies, a controlled clinical trial containing 24 patients in two risk-stratified groups was initiated. Groups I and II had predicted survivals of 50% and less than 10%, respectively. Patients were then randomly allocated intensive care only or intensive care and ELAD treatment within each group. Patients were treated for a mean period of 72 h  (3–168 h). Results showed improved galactose elimination time, and encephalopathy, but no survival changes; Group I had 78% survival compared with 75% in the control group; group II also showed no difference, with 33% survival compared to 25% in controls.110


Rozga et al. reported the use of BAL in the treatment of an animal model of AHF. Dogs were ventilated, had portacaval shunts inserted and then hepatic and gastroduodenal arteries were occluded. They were divided into three groups. Group 1, the control, received treatment with BAL containing no cells. Group 2 received BAL containing dog hepatocytes and group 3 received BAL with pig hepatocytes. Within 5 h of devascularization all animals in the control group demonstrated increases in lactate, transaminases, lactate dehydrogenase and a significant decrease was seen in the blood glucose, pH and mean arterial blood pressure as compared to the BAL/hepatocyte treated animals.102

This study demonstrated that there was no significant difference in the use of allogeneic or xenogeneic cells within the BAL bioreactor. However, only single treatments lasting 4–6 h were performed and the effect of xenogeneic hepatocytes might have become more apparent upon subsequent use. BAL and charcoal perfusion were compared in a canine model of hepatic ischaemia in 1993. Group 1 (n=7) was treated with BAL; group 2 was plasmapheresed and treated via a charcoal column only. After 6 h group 1 had significantly lower blood ammonia and normalized prothrombin times. Blood glucose was also significantly higher in group 1 than in group 2.103

BAL was soon clinically applied and an early case report demonstrated significant improvements in haemodynamic stability, increased presence of clotting factors, reduction in serum ammonia levels and also improved mental state following a single period of BAL treatment lasting 6 h in a patient with severe decompensated alcoholic liver disease.111 This demonstrated that BAL could be used safely in man and in 1993, 10 patients were treated with BAL for AHF. Eight patients showed improvement in ICP, CPP, transaminases, ammonia levels and clinical encephalopathic state leading to OLT.112

In 1996 Chen et al.113 examined the effect of BAL on two groups of patients. Group 1 consisted of 12 patients with AHF all of whom were bridged to OLT. Biochemical parameters, blood glucose, ammonia and bilirubin improved significantly in this group, as did ICP and CPP. Group 2 contained eight patients with acute decompensation of chronic liver disease of whom six patients died.

Watanabe et al.114 presented similar findings with three groups of patients treated with BAL: group 1, patients with AHF (n=18); group 2, three patients with primary non-function of transplanted grafts; and group 3, 10 patients with decompensation of chronic liver disease (not regarded as candidates for OLT). Analysis of the results revealed significant improvements in biochemical and clinical parameters as seen previously (Chen et al.113). All but two patients in groups 1 and 2 went on to have OLT. Two surviving patients from group 3 also went forward for OLT at a later date.

A multicentre controlled trial of the use of BAL in AHF is currently taking place, the results of which are eagerly awaited.


The use of isolated primary cryopreserved hepatocytes presents some immunological and microbiological concerns. Immunological attack will come from the host, by both cellular and humoral mechanisms. Naturally occurring antibodies will attack the pig cells. The synthesis of foreign plasma proteins, coagulation factors and transport factors by the pig hepatocytes have significant immunological effects on the host, including antibody formation. Patients have been shown to develop type III hypersensitivity (serum sickness). Coagulation factors produced by the pig hepatocytes used in the BAL have been shown to cause immune complex deposition in animal and human studies and may contribute to end organ dysfunction.115 Some investigators have avoided this problem by manipulating cell lines and abrogating protein production.116, 117

The use of xenogeneic material also raises the possibility of the transfer of viral or prion disease to the host. Recent reports of pig endogenous retrovirus (PERV) infections in man led to concern over the use of pig hepatocytes in bioartificial support systems. Retrospective studies of patients treated with BAL have shown no evidence of PERV DNA in blood samples. Similar results have been found in patients who have received other forms of porcine tissue (pancreatic islets and heart valves).118[119]–120 However, the long-term aetiology of viral or prion transfer in immuno-compromised patients is still an area that is unclear and highly controversial.


Injection of hepatocytes into an individual with AHF has the potential to replace detoxifying and synthetic function, provided cells access blood or tissue fluid in equilibrium with plasma. Sutherland et al.121 in 1977 demonstrated improved survival following transplantation of 1.5–2.0 g hepatocytes intra-portally or intra-peritoneally after dimethylnitrosamine toxicity in rats as compared to controls. Sommer et al.122 used D-galactosamine in rats as the AHF model and showed similar survival. However, the benefits of transplantation could also be obtained by injection of irradiated hepatocytes, hepatocyte supernatants and homogenates, suggesting that substances other than the hepatocyte itself may be eliciting beneficial effects.

Surgical models of AHF (90% hepatectomy) in animals were also used to demonstrate improved survival.123, 124 As with previous experiments, improved survival was evident if transplantation was carried out prior to or at the same time as the original insult. Advances in biotechnology have led to the concept of encapsulation of hepatocytes. This protects the hepatocytes by reducing immunogenicity, mechanical trauma and also allows three-dimensional structures to develop within the semi-permeable capsules.125 Improved results are also found if the hepatocytes are co-cultured and transplanted with non-parenchymal cells. These developments may enhance cell-to-cell interactions, improve cell viability and may reduce the incidence of host rejection.

The clinical use of hepatocyte transplantation has been limited due to the difficulty of providing adequate and as yet unknown quantities of hepatocytes. These transplanted cells are required to function optimally in a toxic environment. Both Mito and Kusano126, 127 demonstrated little benefit from intrasplenic transplantation of hepatocytes isolated from nine chronic hepatitics and cirrhotics (their own left lateral segments). Habibullah et al.128 delivered 60 × 106 cells/kg body weight of human foetal hepatocytes (isolated from 2 to 34 week gestational foetuses) intra-peritoneally into seven AHF patients. Overall survival was 43% compared with 33% in matched controls. Further small studies have been reported. Soriano et al.129 reported three children in AHF who were treated with intra-portal injection of cryopreserved hepatocytes taken from unused donor segments. One out of three children survived and some biochemical parameters improved post-transplantation. Bilir et al.130 showed improved encephalopathy, serum ammonia levels and prothrombin times, following percutaneous cryopreserved human hepatocyte transplantation in three patients with AHF. Strom et al.131 performed a prospective controlled trial of transplanted isolated fresh and cryopreserved human hepatocytes as a bridge to transplantation. Five hepatocyte transplant recipients with grade IV encephalopathy and multi-organ failure and four patients of equal illness severity with AHF were studied. Medical treatment resulted in significant improvements in the biochemical markers of AHF, including blood ammonia. No improvement was seen in haemodynamic stability or cerebral stability and all died within 3 days. Those receiving hepatocyte transplants maintained normal cerebral perfusion and haemodynamic stability with significant reductions in blood ammonia and liver injury markers. All were transplanted within 2–10 days.

This area requires further refinement and investigation to demonstrate definite clinical impact. Overall, it seems challenging to develop techniques for acute failure, if the estimate of ≈ 3 × 1010 fully functioning hepatocytes are required is correct, due to the difficulties of establishing the conditions for effective function immediately. On the other hand the technique may be of greater potential in chronic disease, and can also be modified for gene transfer.


The clinical sequelae of AHF reflect the multi-faceted dysfunction that takes place once the liver fails. Recognition of the extreme nature of these problems has led to the development of specialist centres, which have been instrumental in reducing the mortality associated with this devastating syndrome. Treatment of clinical problems in AHF such as cerebral oedema, systemic hypotension and renal failure may be temporarily valuable until either transplantation takes place or the organ begins to regenerate. Greater understanding of the biochemistry of systemic inflammatory response syndrome, sepsis and AHF have led to evidence-based treatment of clinical problems. The complexities of liver function and dysfunction are still not completely understood and therefore it is naive to expect artificial support systems to correct and/or replace the synthetic, metabolic and excretory functions that are associated with a healthy organ.

Currently, artificial support systems demonstrate improvements in some clinical and biochemical parameters. Effects on survival are still not clear and the results of multicentre randomized controlled trials are awaited. The design of such trials poses several problems. AHF is a rare disease and therefore to recruit the appropriate number of patients into both the control and active treatment arms will take some time unless the multicentre approach is used. Problems will arise in the matching of patients within the trial, as the syndrome is both disparate and varied in its pattern and presentation. Present treatment for AHF is liver transplantation, which is well established with low morbidity and mortality. Following patient recruitment, it would be unethical to deny or delay transplantation if available. This may result in significant loss of numbers. The varying aetiology of AHF has a great influence on patient survival. Therefore groups will have to be subdivided according to aetiology and controlled studies will have to examine the differences within each group. Separate studies will have to take place to look at the decompensated chronic liver disease group. These patients may also have to be studied according to the aetiology of their disease.

The nature of international multicentre trials raises issues such as the fact that AHF classification differs between countries, as does the incidence, epidemiology and treatment of different aetiologies. Intensive care units and specialist centres have differences in medical and nursing standards as well as marked differences in management of the critically ill. Evidence exists that genotypic variation amongst patients may introduce degrees of susceptibility to the immune and physiological response seen in AHF and sepsis. This may have an as yet unexplained effect on individual patient survival.132

It is clear that precise end-points are required for such a trial. Design of a multicentre controlled trial will be a difficult task, the results will take time to accrue and must in the first instance be carefully interpreted.