The pathophysiological basis of acute-on-chronic liver failure (ACLF) is unclear but systemic inflammatory response is thought to be important. In patients with ACLF, the molecular adsorbents recirculating system (MARS) improves individual organ function, but the effect of MARS on the proposed mediators of systemic inflammatory response is unclear. The present study was designed to determine the effect of MARS on the cytokine profile, oxidative stress, nitric oxide, and ammonia. A total of 18 patients with alcohol-related ACLF due to inflammation-related precipitants were randomized to receive standard medical therapy (SMT) alone, or with MARS therapy over 7 days. Plasma cytokines, malondialdehyde (MDA), free radical production, nitrate / nitrite (NOx), and ammonia were measured. Encephalopathy improved significantly with MARS (P < .01), but not with SMT. Mean arterial pressure and renal function remained unchanged. No significant change of plasma cytokines and ammonia levels were observed in either group. Plasma MDA levels did not change either. There was a fall in NOx (P < .05) with MARS, but not with SMT. In conclusion, in inflammation-related ACLF patients, albumin dialysis using MARS results in improvement of encephalopathy, independent of changes of ammonia or cytokines, without improving blood pressure or renal function. These results should temper the liberal use of MARS until further data is available. (Liver Transpl 2004;10:1109–1119.)
The pathophysiological basis of acute-on-chronic liver failure (ACLF) is uncertain but current hypotheses suggest that systemic inflammatory response may underlie the transition of a patient from a stable cirrhotic state to developing progressive liver injury and end-organ failure. Cytokines are believed to be important in the pathogenesis of decompensated alcoholic liver disease.1 A predominantly proinflammatory cytokine profile might cause hepatic inflammation and necrosis. Mitochondria are a target for tumor necrosis factor (TNF) initiated death signals, releasing reactive oxygen species, leading to apoptosis and cell death.2, 3 Inflammation and oxidative stress also induce production of nitric oxide (NO), which appears to cause the circulatory4–10 and renal11–13 disturbances of liver failure. There is increasing evidence that the mediators of inflammation (e.g., proinflammatory cytokines, NO, and oxidative stress) could modulate14, 15 the effect of hyperammonaemia16–18 in precipitating encephalopathy.
The molecular adsorbents recirculating system (MARS) (Teraklin AG, Rostock, Germany) is an albumin dialysis-based extracorporeal device that removes protein-bound toxins from the blood.19, 20 Several studies, including two small randomized controlled trials,21, 22 in patients with acute decompensation of cirrhosis, have demonstrated improvement of serum bilirubin,20, 21, 23–25 hepatic encephalopathy,23, 26, 27 systemic hemodynamics,21, 26–28 renal function,23, 26, 27 and short-term survival.23, 26, 27, 29 The hypothesis is that accumulation of nondialyzable “toxins” in liver failure contributes to the pathogenesis of end-organ dysfunction. Their removal would interrupt the progressive clinical worsening, allowing recovery from the acute episode. However, which toxins are important remain undefined.
Therefore, removal / reduction of these toxins could improve the outcome of ACLF. In a recent uncontrolled study,25 we demonstrated an apparently improved survival among severe alcoholic hepatitis patients following MARS therapy. The present study was designed as a follow-up to determine whether and how MARS affects the cytokine profile, markers of oxidative stress, NO, and ammonia.
ACLF, acute-on-chronic liver failure; MARS, molecular adsorbents recirculating system; SMT, standard medical therapy; MDA, malondialdehyde; HRS, hepatorenal syndrome; MAP, mean arterial pressure; TNF, tumor necrosis factor; IL, interleukin; NO, nitric oxide; NOx, nitrate / nitrite.
Patients and Methods
This prospective randomized controlled study was conducted on 18 patients at University College London Hospital, with informed, written consent from all patients with encephalopathy less than grade 2, or assent from next of kin for all other patients, with approval of the local research ethics committee in accordance with the Declaration of Helsinki (1989).
Patients included had alcoholic liver disease with ACLF, defined as acute deterioration in liver function over 2–4 weeks with a defined inflammation-related precipitant (infection or alcoholic hepatitis) leading to severe progressive clinical deterioration despite supportive care (over 48 hours) with increasing jaundice (serum bilirubin > 100 μmol/L) and either encephalopathy (≥ Grade 2) or hepatorenal syndrome (HRS) in a patient with clinical, radiological, biochemical, and histological evidence of cirrhosis. Variceal bleeding or infection (if present) was controlled for at least 48 hours before inclusion.
Age <18 or >75 years, lack of consent / assent, prior enrollment in another study, known hepatic / extrahepatic malignancy, uncontrolled infection or upper gastrointestinal bleeding over the previous 48 hours, pregnancy, prior treatment with terlipressin for HRS, coexisting HIV infection, or severe cardiorespiratory disease.
Timing of Randomization
Following initial evaluation, patients received standard medical therapy (SMT) and the precipitating factors were treated appropriately. If there was no improvement over 48 hours, they were randomized for the study.
Randomization was performed by opening sealed prenumbered envelopes. One group received SMT alone (SMT group), while the other received MARS treatment plus SMT (MARS group). All patients were monitored over 7 days from inclusion.
A documented biochemical result of MARS treatment is reduction of serum bilirubin,21, 22, 25 which we used as a surrogate marker for its pathophysiologic effects. Sample size was calculated assuming a reduction in bilirubin of 40% after 4 MARS treatments.21, 22, 25 With an α = .05 and β = .2, 17 patients were required to show a difference between MARS-treated and non-MARS-treated patients. A total of 20 patients with ACLF were evaluated for study, of which 2 had nonalcoholic etiologies. A total of 18 were finally included.
The MARS system (Teraklin AG, Rostock, Germany) has been described elsewhere.19, 20 Briefly, it consists of a blood circuit, an albumin circuit, and a renal circuit, in our case using continuous veno-venous hemofiltration (Hospal BSM 22c, Lyon, France). Blood is dialyzed across an albumin-impermeable high-flux dialysis membrane (MARS Flux; Teraklin AG). The albumin circuit contains 600 mL of 20% human albumin, which passes through the dialyzate compartment of the blood dialyzer. It subsequently undergoes hemofiltration and passage through activated charcoal and anion exchange resin columns to remove acquired toxins. Heparin was used as required to prevent clotting in the extracorporeal circuit. Patients were treated for 4 sessions of 8 hours, over the 7-day study period.
The patients were evaluated clinically and biochemically before and after each MARS treatment. The Child-Turcotte-Pugh30 and Model for End-Stage Liver Disease (MELD)31 scores were calculated. Encephalopathy was assessed by West Haven criteria32 by 2 independent clinicians not directly involved with the study, and HRS was defined by International Ascites Club criteria.33 Patients were mechanically ventilated if hypoxemic. Mean arterial pressure (MAP), electrocardiogram, heart rate, and temperature were monitored continuously. Patients were actively warmed to a core temperature of 37°C if required during MARS therapy. Intravascular volume was maintained using crystalloids, colloids, or red cells, as appropriate, to maintain central venous pressure between 8–10 mm of Hg. Renal support was instituted with continuous veno-venous hemofiltration, where appropriate. Noradrenaline was used to maintain MAP < 55 mm of Hg where necessary. Blood glucose was maintained between 5–7 mmol/L. Tense ascites was treated with paracentesis and volume replacement. Concomitant infection / spontaneous bacterial peritonitis was treated with antibiotics, variceal bleeding with endoscopic band ligation, and encephalopathy with lactulose and bowel enemas.
Blood samples were collected at baseline and on day 7 from all patients. In the MARS group, samples were also collected before and after each MARS treatment. Following each MARS treatment, albumin dialyzate samples were taken from the different segments of the circuit (segment 1: between blood and hemofilter columns; segment 2: between hemofilter and charcoal columns; segment 3: between charcoal and resin columns; segment 4: between resin and blood columns). A complete set of pre- / post-MARS samples was available for 20 sessions.
Venous plasma samples from 8 healthy volunteers were used to obtain control values for the various laboratory measurements described below.
Cytokines were determined from ethylene diamine tetra acetate–anticoagulated plasma samples. All assays were performed as a single batch using commercially available sets (BioSource International, Nivelles, Belgium) following the manufacturer's instructions. Optical density was measured on a Sunrise 96-well plate reader with accompanying Magellan 3 software (Tecan, Salzburg, Austria) at 450 nm referenced against 630 nm. Individual standard curves were fitted and individual values were extrapolated. The lower limit for the detection of the cytokines was 3 pg/mL. The intraassay coefficient of variation was 5.4% to 6.4%. Control values for our laboratory were as follows: 1) interleukin (IL)-6; IL-8; IL-10; and TNF-α were undetectable; 2) TNF-R1 was 1.59 ± .13 ng/mL; and 3) TNF-R2 was 1.88 ± .25 ng/mL (values are means ± standard deviations).
Nitrite and Nitrate
Nitrite and nitrate were determined from heparinized plasma samples, by a modified Greiss test, using methods described previously. Briefly, the plasma / albumin samples were diluted and centrifugally separated to remove proteins though a 12-kd cutoff filter (Vectaspin, Whatman, Maidstone, UK) as described by Giovannoni et al.34 The nitrite and nitrate in the filtrate were determined against a standard curve measured at 550 nm as above, utilizing methods described by Miranda et al.35 Control values were 40 μmol/L (median; range: 30–50 μmol/L) for nitrate and 2 μmol/L (median; range: 1–3.5 μmol/L) for nitrite.
Ammonia was measured in the filtrate obtained above (from plasma / dialyzate) using an indophenol detection method. Briefly, the filtrate (50 μL) was incubated in a 96-well plate at 30°C for 3 hours in the presence of phenol (10 mmol/L), nitroprusside (10 mmol/L), sodium hypochlorite (10 mmol/L), and sodium hydroxide (0.5 mmol/L) (total volume 200 μL). Concentration was determined against a standard curve measured at 630 nm as above. Control values were 20 μmol/L (median; range: 9–30 μmol/L).
Malondialdehyde (MDA) was determined using a modified thiobarbituric acid reactive substances assay described by Lapenna et al.36 wherein the major interfering / oxidizable component in the plasma is inhibited by addition of sodium sulfate. Control values were 1.8 μmol/L (median; range: 1.2–2.5 μmol/L).
Results were expressed as median (range). Significance of difference within a group was tested by paired t-test or Wilcoxon's matched pair test, and between groups by single factor analysis of variance or Mann-Whitney test, as appropriate. Linear regression was used to determine the relationship between variables. A Kaplan-Meier–style plot was used to study the time course of change in encephalopathy between the two groups, and a log-rank test was used to test significance between them. P < .05 was considered statistically significant.
The demographic profile on inclusion is described in Table 1.
Table 1. Baseline Demographic Profile of the 18 Patients Included in the Study
MARS Group (n = 9)
SMT Group (n = 9)
Ethanol + hepatitis C
Continuous veno-venous hemofiltration requirement
Clinical and Biochemical Changes
Clinical, biochemical and prognostic parameter changes over the 7-day period are described in Table 2. Table 3 describes changes of encephalopathy and MAP over 20 individual sessions of MARS in the 9 MARS-treated patients.
Table 2. Change of Clinical, Biochemical, and Prognostic Parameters, and Plasma Values of Cytokines, Malondialdehyde, NOx (Nitrate and Nitrite) and Ammonia Over 7 Days in the MARS and SMT Group of Patients
MARS Group (n = 9)
SMT Group (n = 9)
Abbreviations: MAP, mean arterial pressure; INR, international normalised ratio; MELD, model for end-stage liver disease; TNF, tumour necrosis factor; TNF-R, TNF-receptor; IL-interleukin; NH3, arterial ammonia; MDA, malondialdehyde.
P value (compared to baseline): *<.05, †<.01, ‡<.001.
MAP (mm Hg)
Serum bilirubin (μmol/L)
Serum creatinine (μmol/L)
Prothrombin time (second)
Serum albumin (g/L)
Table 3. Change of Hepatic Encephalopathy, Mean Arterial Pressure, and Plasma Values of NOx (Nitrate and Nitrite), Ammonia, Malondialdehyde and Cytokines Over 20 Individual Sessions of MARS Treatment
Abbreviations: MAP, mean arterial pressure; TNF, tumour necrosis factor; TNF-R, TNF-receptor; IL, interleukin; NH3, arterial ammonia; MDA, malondialdehyde.
MAP did not change significantly in either group. Two patients in each group received inotropes. However, inotrope requirement did not change with MARS therapy. Even excluding these 4 patients from the analysis, no significant change of MAP was seen in either group.
Encephalopathy improved significantly in the MARS group over individual MARS sessions as well as over the 7-day study period, but not in the SMT group. Figure 1A depicts the percentage of patients without improvement of encephalopathy (defined as reduction by ≥1 grade) over the 7-day period in both groups. Taking day 0 to be the start of the study, in the MARS group only 11.1% had not shown an improvement by day 2, while in the SMT group 38.1% had not improved even at the end of the study period (P = .01). The rapid improvement of encephalopathy in the MARS group was not matched by a corresponding improvement of arterial ammonia (Fig. 1B). Two patients in the MARS group showed a subsequent worsening of encephalopathy by 1 grade, related to infection in 1 and variceal bleeding in the other.
No significant change of renal function (serum creatinine, urine output) was observed in either group. Of the 5 patients with HRS in the MARS group, 3 (2 of type I; 1 of type II) did not survive (1 death within the 7-day study period). In the SMT group, there were 5 HRS patients as well, of whom 3 (1 of type I; 2 of type II) died while in the hospital (none in the study period).
A significant improvement of serum bilirubin over 7 days was seen in the MARS group, but not in the SMT group. Serum albumin and prothrombin time remained unchanged.
Child-Turcotte-Pugh score improved significantly only in the MARS group, while the MELD score improved significantly in both groups over 7 days.
The associated changes of plasma cytokines, MDA, nitrate and nitrite (NOx), and ammonia (arterial) over the 7-day study period are described in Table 2. Table 3 describes changes of plasma cytokines, MDA (Fig. 2), NOx (Fig. 3), and ammonia (arterial) observed over 20 individual MARS sessions in the 9 MARS-treated patients. NOx, ammonia, MDA, TNF-R1, and IL-8 were all detectable in the MARS albumin dialyzate (segment 1).
Plasma levels of proinflammatory cytokines (TNF-α, IL-6, and IL-8) and their receptors (TNF-R1 and TNF-R2) were elevated in most patients in both groups, along with elevated anti-inflammatory cytokines (IL-10). However, there were no significant changes in either group, either over 7 days or over individual MARS sessions. TNF-R1 and IL-8 were detectable in the MARS albumin dialyzate (segment 1). TNF-R1 was removed across the hemofilter column (between segments 1 and 2), while IL-8 was removed across the charcoal column (between segments 2 and 3) (Fig. 4).
Thiobarbituric Acid Reactive Substances Assay for MDA
MDA level was elevated 4-fold in both groups of patients at baseline, and did not change significantly in either group (Fig. 2). The reduction of plasma MDA (relative to baseline) in the MARS group (mean: 15.9%; range: −39.7% to 51.1%) was significantly greater than in the SMT group (mean: −18.9%; range: −131.0% to 15.1%; P < .05), but this was unlikely to be of clinical significance. MDA was detectable in the MARS albumin dialyzate (segment 1), but did not appear to be removed by any particular part of the circuit.
Nitrate / Nitrite (NO / NO)
Plasma nitrate and nitrite, which were elevated 2- to 3-fold in both groups, reduced significantly in the MARS group over 7 days, but not in the SMT group (Fig. 3). A significant reduction in nitrate was also observed over individual MARS treatments. Within the MARS circuit, nitrite was removed across the hemofilter column (between segments 1 and 2) and nitrate across the anion exchange resin column (between segments 3 and 4) (Fig. 4).
Plasma ammonia was elevated 4- to 5-fold and did not change significantly in either group. No difference in concentration was detected either over 7 days or over individual MARS sessions even though ammonia was detected in the dialyzate of the MARS circuit (segment 1), from which it was found to be removed across the hemofilter column (between segments 1 and 2) (Fig. 4).
There were 5 in-hospital deaths in both groups, while 4 patients from each group could be discharged (all alive after 3 months follow up). In the MARS group, 2 patients died due to variceal bleeding and 3 due to multiorgan failure (2 with associated sepsis). In the SMT group, 1 patient died due to variceal bleeding and 4 due to multiorgan failure (2 with associated sepsis). There was 1 death within the 7-day study period in the MARS group (due to multiorgan failure on day 7; the data for day 6 for this patient [prior to the onset of the final acute deterioration] has been used in the analysis) and none in the SMT group. Neither clinical, nor biochemical or pathophysiological parameters over the 7-day study period were significantly different between those who ultimately survived and those who did not in either group (data not shown).
Although several investigators have described the clinical effects of MARS,22, 23, 26, 27, 29 this study is the first to investigate the pathophysiological alterations produced in a relatively homogeneous group of patients with ACLF due to inflammation-related precipitants in a randomized controlled manner. It is important to note that our study was not designed to look at mortality but at the pathophysiologic basis of MARS treatment. The main findings were improvement of encephalopathy without improving blood pressure or renal function, which was accompanied by a significant reduction of NOx, without significant change of cytokine profile, MDA levels, or plasma ammonia.
The primary hypothesis tested in the present study related to whether MARS treatment could alter the cytokine profile and / or reduce oxidative stress (as measured by plasma MDA levels) in patients of ACLF due to inflammation-related precipitants. We were unable to detect any significant changes in either parameter over the 7-day period. While the reduction of MDA levels in the MARS group was significantly more than in the SMT group, MARS therapy failed to achieve a significant reduction of baseline levels either over individual treatment sessions, or over the 7-day study period. Therefore, it is unlikely that change of oxidative stress is an important factor in MARS therapy. It was however surprising that in spite of an observed removal of cytokines and their receptors (IL-8, TNF-R1) by MARS, plasma levels did not fall, suggesting that concurrent cytokine production due to the disease process itself balanced any removal.
In keeping with previous observations,23, 26, 27 encephalopathy improved significantly and rapidly with MARS treatment, but without any associated significant change in ammonia. While ammonia was detectable in the MARS dialyzate, this removal was not enough to reduce arterial levels, suggesting that improvement of encephalopathy due to MARS treatment is probably independent of improvement of hyperammonemia. Inflammation is also believed to be important in the pathogenesis of encephalopathy.14 A reduction of proinflammatory cytokines could have been responsible for improvement of encephalopathy, but this was not observed. Oxidative stress has increasingly been thought to be important in the pathogenesis of encephalopathy,37–40 through its effects upon mitochondria, oxidation of membrane phospholipids, and various enzymes involved in energy metabolism.40 However, a clear improvement of oxidative stress was not observed either. Another potential underlying mechanism may be the reduction of plasma NOx levels.40 The role of NO in encephalopathy is unclear, but evidence is emerging that nitrosation of critical proteins may contribute by altering their function or causing injury by formation of nitrotyrosine.41 Thus, one might hypothesize that in the presence of hyperammonemia, the mediators of inflammation may contribute to its adverse neuropsychological effects.15 It is also important to note that in the group of patients studied, encephalopathy was not present as a distinct entity, but rather as a component of ACLF. Thus, it is difficult to look at the change of encephalopathy in isolation without regarding the entire clinical picture as a whole. The same principle applies to any underlying pathophysiological basis, which one would expect to differ in ACLF from that of “pure” hepatic encephalopathy.
It was surprising to find no change in MAP following MARS therapy, in contrast to some previous studies21, 25–28 (which also demonstrated MARS-related improvement of systemic vascular resistance and cardiac output). However, most of these studies were uncontrolled. The effects of hemofiltration, inadequate resuscitation, and hypothermia also need to be considered, as these can affect MAP. It must be pointed out that most of our patients were not substantially hypotensive at baseline, which might explain the lack of subsequent rise of MAP post-MARS. At least 1 other study has reported no change of MAP following MARS in a group of patients who were hemodynamically stable pretreatment.42 The substantial reduction of plasma NOx in the MARS group, both as an immediate effect of treatment, and over the 7-day study period, demonstrates a possible mechanism of improving systemic hemodynamics. A reduction of plasma NOx indicates either direct removal by MARS, or reduction of NO synthesis (with reduced production of its metabolites), or a combination of both. Fairly high levels of NOx in the dialyzate indicate direct removal. Whether there is an additional reduction of NO synthesis by altering factors favoring its production in ACLF, e.g., by removing endotoxins,6, 43 needs further study. The biological significance of removal of NOx by MARS is unclear, but nitrite (NO) has been implicated in the formation of nitrotyrosine residues under oxidative stress conditions.44
It is difficult to explain the lack of any significant change in MAP despite a reduction of NOx. To clarify this apparent discrepancy, we investigated the alterations in the vasoconstrictor neurohormonal profile in 3 other patients with alcohol-related ACLF being treated with MARS.5, 45 Plasma renin activity, angiotensin II, noradrenaline, and endothelin-1 were substantially elevated in the patients prior to MARS, and were reduced markedly at the end of treatment and sustained 12 hours after MARS (Fig. 5; and unpublished data). It is possible that the lack of improvement in MAP observed in our study despite a reduction in NOx may be because of concurrent removal of the vasoconstrictor neurohormones. However, these data need to be confirmed by other studies designed for this purpose.
The lack of improvement of renal function following MARS therapy was also unexpected, and apparently contradicts some previous studies that have reported improvement in patients with HRS following MARS treatment. While most reports did not have a control arm,23, 26, 27 the 1 study that did have a control arm21 had a small sample size (n = 13). One more death in the MARS arm would have meant that the significance of the observation would have been lost. In addition, it should also be noted that rather than looking at “pure” HRS, the present study included ACLF patients who had renal dysfunction as a component of their clinical picture. Thus, any observed clinical change, or the lack of it, and the underlying pathophysiological basis is not necessarily comparable with other such studies. An apparent reduction of serum creatinine was observed over 7 days in the SMT group, which was not statistically significant, but which probably accounted for the significant improvement of the MELD score in this group (while the significant reduction of serum bilirubin probably achieved the same in the MARS group).
We also examined the role of the different components of the MARS circuit in the removal of toxins. Ammonia and nitrite (highly water-soluble) are removed across the hemofilter column, suggesting clearance by the ultrafiltrate. IL-8 is removed by the activated charcoal. Nitrate (NO, the major component of NOx), though also water soluble, is known to associate with albumin and was removed across the anion exchange resin but not across the hemofilter. Though TNF-R1 decreased across the hemofilter column, its molecular weight (55 kd)46 exceeds the pore cutoff size on the hemofilter membrane (30 kd), suggesting adsorption onto the membrane, rather than removal into the ultrafiltrate. This is supported by similar observations in oliguric patients undergoing hemofiltration, where a high plasma clearance of TNF-R1 was not matched by clearance into the ultrafiltrate.47 MDA, while measurable in the dialyzate, did not appear to be removed by any specific component, suggesting that the albumin in the dialyzate itself acted as a sump.
Finally, it must be reemphasized that while the present study looks at encephalopathy, renal dysfunction, and circulatory disturbances in liver disease, it does not look at them in isolation, but rather as part of the overall clinical picture of ACLF. This might explain the difference in clinical changes observed in this study, as well as the underlying pathophysiological basis, compared to other studies performed in the past that did not evaluate ACLF as a whole.
The main clinical change with MARS therapy observed in our study was an improvement of encephalopathy. The inability to reproduce some of the other clinical benefits previously reported (such as improvement of blood pressure or renal function) suggests that results of treatment may be determined by the population of patients being studied, timing of intervention, precipitating factor of ACLF, and the baseline hemodynamics. For instance, data from the International MARS Registry indicates that hemodynamic improvements with MARS therapy is predominantly seen in patients with a lower baseline MAP than was seen in the present study.48 Given the variability in the observed clinical responses, further studies need to be performed to obtain more definitive data. Our results certainly suggest that the liberal unrestricted use of MARS outside of a research setting should be discouraged until such data is available.
Our results also show that the clinical changes observed with MARS treatment in patients with ACLF precipitated by inflammation are independent of changes in plasma ammonia or cytokine levels, or of oxidative stress. The predominant pathophysiologic effect of MARS is on other mediators of inflammation like NOx, which are reduced with MARS treatment, possibly due to a combination of removal and reduced production. These findings highlight the importance of factors other than ammonia in the pathogenesis of encephalopathy, but the identity of all the factors involved remains unknown.
Teraklin AG, Germany, provided the MARS kits for the study, free of cost.