Bioartificial liver in acute liver failure: Impostor or simply misunderstood?


Demetriou AA, Brown RS, Busuttil RW, Fair J, McGuire BM, Rosenthal P, et al. Prospective, randomised, multicenter, controlled trial of a bioartificial liver in treating acute liver failure. Ann Surg 2004; 239: 660670. (Reprinted with permission by Lippincott, Williams & Wilkins.)


Objective: The HepatAssist liver support system is an extracorporeal porcine hepatocyte-based bioartificial liver (BAL). The safety and efficacy of the BAL were evaluated in a prospective, randomized, controlled, multicenter trial in patients with severe acute liver failure. Summary Background Data: In experimental animals with acute liver failure, we demonstrated beneficial effects of the BAL. Similarly, Phase I trials of the BAL in acute liver failure patients yielded promising results. Methods: A total of 171 patients (86 control and 85 BAL) were enrolled. Patients with fulminant/subfulminant hepatic failure and primary nonfunction following liver transplantation were included. Data were analyzed with and without accounting for the following confounding factors: liver transplantation, time to transplant, disease etiology, disease severity, and treatment site. Results: For the entire patient population, survival at 30 days was 71% for BAL versus 62% for control (P = 0.26). After exclusion of primary nonfunction patients, survival was 73% for BAL versus 59% for control (n = 147; P = 0.12). When survival was analyzed accounting for confounding factors, in the entire patient population, there was no difference between the 2 groups (risk ratio = 0.67; P = 0.13). However, survival in fulminant/subfulminant hepatic failure patients was significantly higher in the BAL compared with the control group (risk ratio = 0.56; P = 0.048). Conclusions: This is the first prospective, randomized, controlled trial of an extracorporeal liver support system, demonstrating safety and improved survival in patients with fulminant/subfulminant hepatic failure.


The comparison of the search for an effective liver support device to that of the endeavor to find the “holy grail” is clichéd and parochial, but it effectively articulates the importance of such a device to liver disease. The clinical burden of liver disease is increasing and is anticipated to continue to do so for at least another 10 to 15 years. The management of liver failure is becoming progressively more resource intensive and expensive, as futility is replaced by optimism for positive clinical outcomes in these settings. Liver transplantation is pivotal to the management of patients with both acute and chronic liver failure. However, only a minority of patients with liver failure get access to a liver transplant and then with limited control over the timing of the intervention and the clinical condition of the patient at the time of surgery. An effective liver support device would stabilize critically ill patients, optimize patients for liver transplantation, and allow some patients to escape the need for transplantation altogether.

Although enthusiasm for liver support devices has waxed and waned over the last 40 years, no device has established itself in clinical practice outside the realm of the enthusiasts. The initial enthusiasm generated by the early studies of charcoal hemoperfusion in the 1970s and 1980s was dashed by the apparent negativity of the pivotal trial.1, 2 This immediately evoked concerns about trial design and interpretation and concerns “that the baby was being thrown out with the bathwater.”3, 4 This was to become a recurring theme. The systems that have been or are currently being evaluated are classifiable as biological, nonbiological, or hybrid systems. The biological systems include bioartificial livers (BALs) utilizing cell-based therapies including porcine cells, human hepatoblastoma cells (C3A cell line), or human hepatocytes. The nonbiological systems include charcoal hemoperfusion, hemdiabsorption using powdered-activated charcoal, high volume plamapheresis and albumin dialysis (including MARS). A systematic review of trials published using these devices up to September 2002 evaluated the outcome in 353 patients with acute liver failure and 130 patients with acute-on-chronic liver failure.5 It was concluded that these systems had no effect on mortality in randomized trials of patients with acute liver failure, but a 33% reduction in mortality was seen in patients with acute-on-chronic liver failure. No significant benefit in bridging patients to transplantation was identified. A significant improvement in encephalopathy was identified but this was considered a “soft” outcome measure prone to observer bias.

A challenge in the development of a liver support device is the reality that good trials are difficult to design and execute in acute liver failure, which has been the conventional testing ground for these devices. Acute liver failure is a heterogenous condition, with etiology, age, and the tempo of disease progression all significantly influencing the capacity for spontaneous regeneration. The latter is a clearly a prerequisite for studies that assess the ability of a device to bridge patients to transplant-free survival, but it is probably also relevant to studies evaluating devices as bridges to transplantation. This has become the leading objective of studies over the last decade or so, and there is little evidence of confidence in testing the ability of current devices in bridging patients to transplant-free survival. The profile of patients with the best prospects of recovery is of young patients with rapidly progressive disease who have acetaminophen induced liver failure and hepatitis A or B.6 The patients least likely to reflect a benefit from liver support are older, have subacute liver failure, and have seronegative (or indeterminate) hepatitis or idiosyncratic drug reactions. There is also considerable heterogeneity with respect to the clinical complications that arise in these patients, particularly with regard to intracranial hypertension, SIRS/sepsis, hemodynamic instability, and renal failure, and these too need to be considered at the stage of trial design.2

Modern trials of liver support devices in acute liver failure are “hostages to fortune” by virtue of the practice of diverting these patients to liver transplantation once a donor organ has been allocated. Reform of the United Network for Organ Sharing allocation system has advantaged patients with acute liver failure in the United States, and now the majority of patients are receiving transplants within 48 hours of being wait-listed, a situation that is comparable to that of the United Kingdom. This results in an inadequate period of time available to appropriately assess the efficacy of a device in a substantial cohort of a study population.

Demetriou and his colleagues have recently reported the outcome of a controlled trial of a porcine hepatocyte–based BAL in 171 patients with acute liver failure or primary nonfunction after liver transplantation.7 With this device, separated plasma is initially pumped through a charcoal column before passing through a cartridge with hollow perforated fibers (pore size, 1.5 μm) that facilitate contact with the porcine hepatocytes. Twenty centers across the United States and Europe recruited 171 patients over a 3-year period, including 24 patients with primary nonfunction, 121 with fulminant hepatic failure, and 26 with subacute liver failure. Of these, 85 were randomized to BAL and 86 were controls. The primary end point was 30-day survival. The average number of treatments was 2.9 (range, 0-9), and the results were analyzed on an intention-to-treat basis.

The data and safety monitoring board terminated the trial after 171 patients had been recruited “because it determined that trial continuation, under the protocol in place and using this type of data analysis, was likely to be futile for the primary end point of 30-day survival.” The 30-day survival was 71% in the BAL group and 62% in the control group (P = .26). Subset analyses showed no difference in 30-day survival in the patients with primary nonfunction (75% vs. 58%; P = .667) or acute liver failure (73% vs. 59%; P = .117). A conclusion that can, and has been, drawn is that this is a negative study.

However, further analyses, although possibly post hoc, have identified a number of observations that deserve further consideration before the baby disappears with the bathwater. The first of these observations is that survival benefit was seen in the subgroup of 80 patients with acute liver failure with an identifiable etiology. This cohort had a 44% reduction in mortality at 30 days and a significant delay in the time to death. This observation could be a quirk of overanalysis, but equally it could be a reliable observation given that this group is very similar to the patient profile outlined above as most likely to have early liver regeneration, and hence it is very credible that these patients would show benefit from the device when others might not.

The study was clear in demonstrating the safety of the BAL device in a population where that could not be taken for granted. It also found no evidence of transmission of porcine endogenous retrovirus, an important theoretical risk that could limit the acceptance of this device. The device also appeared to reduce serum bilirubin levels, but the contribution of the different components of the device to this phenomenon remains uncertain. However, the study failed to shed light on one of the most intriguing issues—i.e., the ability of liver support devices to prevent or treat intracranial hypertension. An early study with BAL did show a reduction in intracranial pressure, but from levels that were not particularly high (mean, 17 mm Hg) and certainly not in the range that would lead to brainstem herniation or ischemic/hypoxic brain injury.8 The incidence of recognized intracranial hypertension in the latest study was low in both groups (11.6% vs. 7.1%), but these figures are almost certainly underestimates as systematic use of intracranial pressure monitoring was not utilized in this study (mainly reflecting the challenge of multicenter studies). This important issue remains to be resolved.

For the regulators and proponents of evidenced-based medicine, this trial probably represents a “full stop.” To the community of clinicians practicing hepatology and in need of an effective device it is probably a “semicolon.” However, to the clinical researcher with a commitment to developing an effective liver support device it probably represents only a “comma.” It is tempting to speculate that a different trial design and patient selection could have given a positive outcome and totally altered the climate for the development of liver support devices. Enthusiasts will hope that the positive messages of this trial, together with the lessons learned, will galvanize the search for the “holy grail.” The need for clinical trials persists, but these trials must set achievable targets for the device, use appropriate clinical end points, and be conducted in a patient population likely to demonstrate the benefits of the device.