Abstract
Extracorporeal albumin dialysis (ECAD) may improve severe hepatic encephalopathy (HE) in patients with advanced cirrhosis via the removal of protein or non–protein-bound toxins. A prospective, randomized, controlled, multicenter trial of the efficacy, safety, and tolerability of ECAD using molecular adsorbent recirculating system (MARS) was conducted in such patients. Patients were randomized to ECAD and standard medical therapy (SMT) or SMT alone. ECAD was provided daily for 6 hours for 5 days or until the patient had a 2-grade improvement in HE. HE grades (West Haven criteria) were evaluated every 12 hours using a scoring algorithm. The primary endpoint was the difference in improvement proportion of HE between the 2 groups. A total of 70 subjects [median age, 53; 56% male; 56% HE grade 3; 44% HE grade 4; median model for end-stage liver disease (MELD) 32 (11–50) and CPT 13 (10–15)] were enrolled in 8 tertiary centers. Patients were randomized to ECAD + SMT (n = 39) or SMT alone (n = 31). Groups were matched in demographics and clinical variables. The improvement proportion of HE was higher in ECAD (mean, 34%; median, 30%) versus the SMT group (mean, 18.9%; median, 0%) (P = 0.044) and was reached faster and more frequently than in the SMT group (P = 0.045). Subjects receiving ECAD tolerated treatment well with no unexpected adverse events. Conclusion: The use of ECAD may be associated with an earlier and more frequent improvement of HE (grade 3/4). Because this 5-day study was not designed to examine the impact of MARS on survival, a full assessment of the role of albumin dialysis awaits the results of additional controlled trials. (HEPATOLOGY 2007.)
Hepatic encephalopathy (HE) is a complex neuropsychiatric syndrome commonly seen in patients with advanced liver disease. HE arises from the effects of circulating toxins on cerebral functions. Putative toxic molecules accumulate in patients with liver decompensation because of increased production, portal-systemic shunting, or lack of hepatic detoxification.1 Patients with advanced cirrhosis and a superimposed acute liver injury often decompensate and present with manifestations of hepatic failure, including worsening HE and coma.2.
Current therapy for HE includes the use of nonabsorbable disaccharides or poorly absorbable antibiotics, as proposed more than 3 decades ago.3 However, standard therapy is less effective in patients with severe degrees of liver failure.4, 5 Under these circumstances, the concept of supporting the failing liver for a time while correcting the precipitating event might help patients recover from HE or be stabilized until they receive a liver transplant.6 Extracorporeal albumin dialysis (ECAD) using the molecular adsorbent recirculating system (MARS) is a new method of hemodiafiltration whereby blood is dialyzed against an albumin-containing solution across a high-flux membrane.7 The technique allows combined removal of albumin-bound and water-soluble toxins.8-11
In uncontrolled trials of ECAD using the MARS device, patients had a reduction in ammonia levels, clearance of aromatic amino acids, and improvement in systemic hemodynamics. These effects might explain its potential benefits in improving HE in patients with liver decompensation.8-11 However, none of the published trials addressed patients with severe HE (grade 3 and 4) or standardized the approach of ECAD treatment. In addition, these trials did not clearly define the level of improvement in HE or standardize the method of assessing HE or report the time spent in the improved mental state in a randomized controlled manner.
The goal of this study was to establish in a randomized controlled multicenter study the efficacy, safety, and tolerability of ECAD using the MARS device in improving grade 3 and 4 HE in patients with acute decompensation complicating chronic end-stage liver disease (ESLD).
Abbreviations
CTP, Child-Turcotte-Pugh; DSMB, data safety monitoring board; ECAD, extracorporeal albumin dialysis; ESLD, end-stage liver disease; GCS, Glasgow Coma Scale; HE, hepatic encephalopathy; IP, improvement proportion; MARS, molecular adsorbent recirculating system; MELD, model for end-stage liver disease; SD, standard deviation; SMT, standard medical therapy.
Patients and Methods
Participants
Seventy patients, 18 years of age or older, presenting with manifestations of cirrhosis and HE grade 3 or 4, were enrolled into the study. Cirrhosis was determined by the medical history and was confirmed clinically, biochemically, and radiologically. Exclusion criteria included patients with active hemorrhage, hemodynamic instability, acute cardiopulmonary complications (pulmonary edema, massive aspiration pneumonia, heart failure), pregnancy, and those who were on active renal replacement therapy. Patients presenting with drug intoxication or irreversible brain damage or nonhepatic causes of altered mental status were also excluded as well as patients with acute liver failure, hepatocellular carcinoma, or who had received a liver transplant.
Study Design
Patients were randomized to standard medical therapy (SMT) with or without ECAD using the MARS device. SMT included treatment of the precipitating events of the acute episode of HE; oral lactulose titrated to achieve 2 to 3 daily bowel movements; oral neomycin or metronidazole, and daily zinc sulfate. Most patients (96%) received systemic antibiotics. All centers applied the same SMT proposed by the protocol.
ECAD was performed using the MARS device (Teraklin AG, Rostock, Germany) attached to a standard hemodialysis machine (Fresenius 2008 or AK 100/200 Gambro) (Fig. 1). ECAD treatments were done every day for 6 hours until the patient responded with an improvement in HE of at least 2 grades from the randomization grade; or had 5 treatments; or dropped out of the study for liver transplantation, withdrawal of consent, or death. Patients were followed for 180 days after the end of the 5-day study period. During that time, patients had 6 visits. Management during the follow-up period was according to standard medical care for patients with ESLD.
Molecular adsorbent recirculating system (MARS) device. The blood circuit driven by the dialysis machine passes through the MARS® FLUX dialyzer at a median speed of 210 (170–500) mL/minute before returning to the patient via the intravenous catheter. While passing the MARS®FLUX dialyzer, water-soluble and albumin-bound molecules of small molecular weight are removed from the blood by diffusion following a concentration gradient through the membrane into a standard dialysate solution containing 16.6% albumin. The albumin dialysate circulates the dialyzer at a median speed of 200 (100–250) mL/minute. The albumin dialysate is recycled by consecutive conventional dialysis (diaFLUX); to remove unbound water-soluble molecules and over uncoated charcoal (AC 250) and ion exchange resin (IE 250) to adsorb albumin-bound molecules. The flow of the dialysate is 500 (300–800) mL/minute dialysate flow provided by the standard dialysis machine.
The study was approved by all Institutional Review Boards/Ethics Committees, and written informed consents were obtained from all participants' next of kin.
Objectives
The primary objective of the study was to compare the efficacy, safety, and tolerability of ECAD using the MARS device in improving severe HE by 2 grades compared with standard medical therapy alone in patients with chronic ESLD during the 5-day study period. The secondary objectives were to assess the time to improvement of HE by 2 grades and the effects of ECAD on laboratory parameters.
Assessment of Efficacy
An evaluation of the efficacy of ECAD was obtained by comparing the improvement proportion (IP) between the 2 groups. IP roughly indicates the amount of time in an improved mental state during the study period. It was calculated as the number of improved HE assessments (≥2 grade improvement in HE from baseline) divided by the total number of assessments done during the study. The patient's HE grade was assessed twice a day (every 12 hours) using the HE scoring algorithm. During the 5-day study period, up to 10 assessments after the baseline assessment were available to calculate the IP for each patient who completed the 120-hour study period. Patients who received a liver transplant or withdrew consent before the end of the 120-hour study period had their IP calculated by dividing the number of assessments in the improved state over the total number of assessments done before exiting the study. The time to first improvement was calculated as the time from randomization to the first assessment indicating improvement in HE of 2 grades from baseline. A patient was considered a “responder” if he or she had at least one 2-grade improvement in HE anytime during the 5-day study period. In addition, the cumulative number of improvements per person over time is given.
Assessment of HE
HE was assessed using HE scoring algorithm, an adaptation of the West Haven Criteria that involves the assessment of HE grade both clinically and via a battery of validated neuropsychological tests (Table 1).12 The 4 areas of brain function delineated in the West Haven criteria were assessed, including state of consciousness, intellectual function, personality behavior, and neuromuscular abnormalities. Assignment of HE grade depended on both the clinical state of the patient and the results of the neuropsychological tests. The Glasgow Coma Scale (GCS) also was used to assess all patients during the study period.
| HE Grade | Clinical Assessments | Neuropsychological Assessments | HE Grade Determination |
|---|---|---|---|
| IV | ○ No eyes open | □ Not applicable | All 3 tests impaired |
| ○ No motor response | |||
| ○ No verbal response | |||
| III | ○ Somnolence | □ Mental control = 025 | At least 3 tests impaired: clinical or neuropsychological |
| ○ Confusion | |||
| ○ Disoriented to place | |||
| ○ Bizarre behavior/ anger/rage | |||
| ○ Clonus/rigidity/ nystagmus/Babinsky | |||
| II | ○ Lethargy | □ Slow responses25 | At least 2 clinical and 3 neuropsychological Impaired |
| ○ Loss of time | □ Anxiety26 | ||
| ○ Slurred speech | □ Amnesia of recent events27 | ||
| ○ Hyperactive reflexes | □ Amnesia of recent events27 | ||
| ○ Inappropriate behavior | |||
| I | □ Complex computations28 | At least 4 tests impaired: clinical or neuropsychological | |
| ○ Sleep disorder | □ Construction ability29 | ||
| ○ Tremor | □ Shortened attention span30 | ||
| □ Depression26 |
- Algorithm used in the study for scoring the HE grades (HESA). Assessment using HESA starts by scoring the highest HE grade (grade 4) and going downward through the algorithm. Clinical assessments (○) were performed by a hepatologist. Neuropsychological assessments (□) were performed separately by a neuropsychologist based on validated tests. The final “HE Grade Determination” was calculated by an independent coordinator. HE grades 1 and 2 required more neuropsychological testing than grades 3 and 4, which were clinically assessed by level of consciousness.
Assessment of Safety and Tolerability
All adverse events reported during the study were reviewed by an independent data safety monitoring board (DSMB) with regard to their potential relationship to the device. Tolerability was assessed by the ability of the patients to sustain therapy for at least 4 hours on the device and by the hemodynamic changes encountered during each treatment.
Laboratory Testing
Standard blood testing was done by certified laboratories. Amino acids were determined chromatographically (LC3000, Eppendorf, Hamburg). Bile acids were measured by a spectrophotometrical assay (Wako Chemicals USA, Inc. in-vitro diagnostic kit) and benzodiazepines were measured with an enzyme immunoassay (Emit Tox: Syva Company, Dade Behring Inc., Cupertino, CA).
Randomization
Randomization was carried out, using a ratio of 1:1 for ECAD to control in blocks of 4 patients. Separate randomization schedules (stratified randomization) were made for the 4 combinations of Child-Turcotte-Pugh (CTP) classification (B, C) and HE grade strata (HE3, HE4) by an independent clinical research organization. The allocation concealment was a set of 4 groups of consecutively numbered blinded envelopes given to each center. Once a subject was enrolled in the study, the principal investigator informed the study coordinator, who then opened the envelope with the lowest numerical number available within the assigned CTP/HE set.
Statistical Methods
Power Calculation.
Based on prior experiences with ECAD and nonabsorbable disaccharides in improving HE in patients with chronic liver diseases, we expected the mean time in an improved state in the control group to be approximately 20 hours (SD = 16.0 hours) and that for the ECAD group to be approximately 40 hours (SD = 34.0 hours)8, 13 out of a 120-hour surveillance period. Thus, a mean 20-hour difference corresponds to a mean difference of 20/120 = 16.7% and the improvement proportion standard deviation (SD) corresponds to SD = 16/120 = 13.3% for the control group and SD = 34/120 = 28.3% for the MARS group. Based on this, using the nonparametric Mann-Whitney test and a 2-sided alpha = 0.05 significance criterion, a sample size of 31 patients per group (total 62) provides 80% power. The sample size was increased to 70 patients to adjust for a possible loss of up to 10% of the patients. No interim analysis was planned.
Primary Analysis.
The IP of all subjects was compared between the groups using nonparametric Wilcoxon rank-sum test (Mann Whitney U statistics). The cumulative number of improvements was computed over time for each subject using reliability methods (SAS Procedure RELIABILITY, SAS Inc., Cary, NC). The mean cumulative number of improvements was compared between the 2 groups over time using repeated-measure analysis of variance methods.
Secondary Analyses.
The time from randomization to first improvement was estimated using Kaplan-Meier methods and compared between the groups using the log-rank test. The course of biochemical parameters was compared on their original scale using the nonparametric analog (Wilcoxon rank-sum test = Mann-Whitney U test). Exact chi-squared methods were used to compare the percentage of patients who were responders between groups (StatXact, Cytel Inc). Patient survival was estimated using Kaplan-Meier survival methods and compared using the log-rank test.
Comparability of Groups.
Demographics and baseline parameters were compared between the 2 groups by exact chi-squared methods for discrete variables and parametric (t test) or nonparametric (Mann-Whitney U) methods for continuous variables, to determine whether the stratified randomization was successful and the groups were comparable.
Bivariate Covariate Assessment.
The relations between continuous potential covariates such as age and continuous or ordinal outcomes such as IP were assessed using nonparametric Spearman correlations (rs). The relation between discrete potential covariates such as sex and continuous outcomes was assessed using the Mann-Whitney U test.
Multivariate Responder Model.
Potential covariates that could affect response were simultaneously assessed using backward stepdown multiple logistic regression methods with a liberal P < 0.15 variable retention criterion. Covariates that were significantly related to response were used to compute adjusted response odds ratios.
Study Populations.
The primary analyses included all 70 patients randomized in the trial. The “per-protocol analysis” included 62 patients after the DSMB assessed blindly the eligibility of patients based on inclusion and exclusion criteria and having sufficient data for evaluation in the case of early withdrawal.
Results
Participants
Between August 2000 and November 2003, 82 patients were screened and 70 randomized into the study. Thirty-one patients were assigned to the SMT arm and 39 to SMT and ECAD (Fig. 2). The characteristics of the 70 patients in the 2 treatment groups, the primary liver disease, and the precipitating factors of HE are presented in Table 2; 56% and 44% exhibited grade 3 or 4 HE, respectively. The median CTP score was 13 (range, 10–15), the model for end-stage liver disease (MELD) score was 32 (range, 11–50) and the GCS was 6 (range, 3–15). At baseline, the groups were statistically comparable. Mechanical ventilation occurred in 35.5% of the SMT and 51.3% of the ECAD groups. For ventilation and control of agitation, sedation (propofol in 75% and benzodiazepines in 25%) was used in 10% (3/31) and 13% (5/39) of the groups, respectively. At baseline, circulating benzodiazepines were detected in 1 sample. The median time to randomization from the first presentation with severe HE (3/4) was 2 days in both groups, and the mean time was 5.6 ± 11.0 days and 2.6 ± 2.5 days in the ECAD and SMT, respectively (P = 0.07, mean ± SD). During this time, patients were managed with their respective local standard of care for patients with HE. Although there were statistically significant differences in age and serum albumin and a borderline difference in time to randomization between the 2 groups (P = 0.07), these 3 factors did not have a significant correlation with IP, the primary endpoint (rs = −0.14, P = 0.26 for age, rs=−0.09, P = 0.45 for albumin, rs=−0.095, P = 0.44 for time to randomization). Thus, there was no need to adjust for these covariates because the unadjusted and adjusted values are virtually the same.
Study flow chart.
| ECAD n = 39 | SMT n = 31 | P Value | |
|---|---|---|---|
| Age† | 49 (20–67) | 56 (32–76) | 0.019 |
| Sex | 24 Males 15 Females | 15 Males 16 Females | NS |
| HE grade 3 | 20 (51%) | 19 (61%) | NS |
| HE grade 4 | 19 (49%) | 12 (39%) | |
| Ascites | 5 None 22 Mild, moderate 12 Severe | 3 None 20 Mild, moderate 8 Severe | NS |
| Creatinine† (0.5–1.5 mg/dL) | 1.7 (0.4–5.6) | 1.7 (0.6–5.0) | NS |
| Total bilirubin† (<1.2 mg/dL) | 15.8 (1.8–54.5) | 12.2 (2.3–58.9) | NS |
| Albumin† (3.5–5.0 g/dL) | 2.4 (1.3–4.1) | 3.0 (1.8–4.5) | 0.004 |
| INR† | 2.1 (1.2–5.9) | 1.9 (1.3–5.4) | NS |
| Primary liver disease (cirrhosis) | |||
| Alcoholic liver disease (ALD)* | 15 (39%) | 12 (39%) | NS |
| Viral hepatitis (HCV/HBV) | 11 (28%) | 10 (32%) | |
| ALD/viral hepatitis | 6 (15%) | 3 (10%) | |
| Cryptogenic cirrhosis | 4 (10%) | 3 (10%) | |
| AIH/PSC | 3 (8%) | 2 (6%) | |
| Drug induced | 0 (0%) | 1 (3%) | |
| HE precipitating factor | |||
| Infection | 10 (26%) | 10 (32%) | NS |
| Bleeding | 6 (15%) | 3 (10%) | |
| Electrolyte imbalance | 3 (8%) | 6 (19.3%) | |
| Other | 7 (18%) | 6 (19.3%) | |
| Unknown | 13 (33%) | 6 (19.3%) | |
| CTP score | 13 (10–15) | 12 (10–15) | NS |
| MELD score | 33 (11–49) | 28 (15–50) | NS |
| SOFA score ≥ 9 | 88% | 81% | NS |
| SIRS score · 2 | 51% | 45% | NS |
| Glasgow coma score† | 6 (3–12) | 8 (3–15) | NS |
- * Acute alcoholic hepatitis was reported in 4 patients (2 in each group).
- † Presented as median (ranges).
- Abbreviations: AIH, autoimmune hepatitis; PSC, primary sclerosing cholangitis; CTP, Child-Turcotte-Pugh; MELD, Model for End Stage Liver Disease; SOFA, Sequential Organ Failure Assessment; SIRS, Systemic Inflammatory Response Syndrome; NS, not significant; Other, drug reaction, constipation, post portal-systemic shunts, and non-adherence to medications.
Improvement in HE
The median IP for the ECAD group was significantly higher than in the SMT group (30 versus 0) (P = 0.044) (Table 3). At any given time during the 5-day study, the mean cumulative number of improvements per patient was higher in the ECAD group than in the SMT group (P < 0.01) (Fig. 3A). In the ECAD group, 62% (24/39) were responders compared with 40% (12/30) in the SMT group (P = 0.076). At the time of assessment during the study, the percentage of patients showing a response was higher in the ECAD group than the SMT group (Table 4). None of the surviving patients who improved their HE by 2 stages returned to their pre-study HE grade during the 5-day study period
| Group | N | Min | Max | Mean | Median | SD | SEM |
|---|---|---|---|---|---|---|---|
| ECAD | 39 | 0 | 100 | 34.0 | 30* | 34.2 | 5.5 |
| SMT | 31 | 0 | 80 | 18.9 | 0 | 26.0 | 4.7 |
- * P = 0.044.
- The improvement proportion is the number of HE assessments with at least 2 grades improvement in HE from baseline divided by the total number of assessments done during the study. The ECAD group had a significantly higher improvement proportion than the SMT group (*P = 0.044).
(A) Mean cumulative number of improvements per person (ITT). Mean cumulative number of improvements per person in ECAD (MARS) versus SMT. At any given time during the 5-day study the mean cumulative number of improvements per patient was significantly higher in the ECAD group than in the SMT group (P < 0.01). By 120 hours, the ECAD group mean is almost twice the SMT mean. (B) Secondary endpoint: time to first improvement (ITT). Kaplan-Meier estimated time to first improvement in ECAD versus SMT. Patients in the ECAD group responded faster and at a higher rate than patients in the SMT group (Log-rank test P = 0.045).
| n (PP) | 24 Hours | 48 Hours | 72 Hours | 96 Hours | 120 Hours | Kaplan Meier |
|---|---|---|---|---|---|---|
| SMT % 31 (29) | 0 (0)21 (22)37 (34)37 (34)45 (43) | ITT P = 0.045 | ||||
| ECAD % 39 (33) | 13 (15) | 30 (34) | 58 (64) | 61 (67) | 72 (74) | PP P = 0.017 |
- Values are percentage of patients reaching a 2-grade improvement at each time after randomization. () represent PP analysis. At any time during the study, patients receiving ECAD achieved a higher response rate than the SMT group.
Time to First HE Improvement
Patients in the ECAD group responded significantly faster and at a higher rate than patients in the SMT group (log-rank P = 0.045) (Fig. 3B). This difference occurred in spite of the delay in starting ECAD therapy after randomization by a median of 5.25 (1–23) hours. The median time to the first 2 grade improvement in HE was 72 hours in the ECAD group versus 108 hours in the SMT group.
Tolerability and Safety of ECAD
The mean number of ECAD treatments for the 39 patients was 2.7 ± 1.5 per patient. A total of 108 treatment sessions were done, of which 94% (n = 102) completed more than 4 hours. In the 6 sessions of less than 4 hours, treatment was discontinued because of hemodynamic instability (n = 2), gastrointestinal bleeding (n = 1), inadequate intravenous access (n = 1), malfunction of the dialysis machine (n = 1), and 1 patient went to liver transplantation. Hemodynamically, patients tolerated the treatment without significant changes in blood pressure or heart rate. The mean arterial pressure decreased to less than 60 mm Hg by the end of treatment in only 9 (8%) ECAD sessions. The median platelet counts in the ECAD group (86 ± 51 × 10/mm3) decreased by 21% (range, 0%–86%) after each treatment session (70 ± 43 × 10/mm3). Platelet transfusion was required in 49% and 32% of patients in the ECAD and SMT groups, respectively. The median (range) platelets required for the ECAD group was 0 (0–4256) cc per patient versus 0 (0–900) cc in the SMT group (P = 0.04). Heparin for anticoagulation was used in 26% (n = 28) of the treatment sessions. Heparin was given initially as an intravenous bolus at a median dose of 500 IU (0–1000) and then as a continuous infusion at a median hourly dose of 500 IU (0–700). One patient on heparin experienced a hematoma at the access site.
During the 5-day study period, 20 (6 related to dialysis catheter insertion) and 8 serious adverse events were reported in the ECAD and SMT groups, respectively (Table 5). During the follow-up period, there were 7 patients with gastrointestinal bleeding (days 6, 8, 8, 19, 24, 31, and 92) in the ECAD versus only 1 patient in the SMT group (day 80). None of these bleeding events were deemed related to ECAD by the DSMB.
| Complications | Type of Event | Days 1–5 | Days 6–10 | Days >11–180 | |||
|---|---|---|---|---|---|---|---|
| ECAD | SMT | ECAD | SMT | ECAD | SMT | ||
| n = 39 | n = 31 | n = 34 | n = 26 | n = 30 | n = 24 | ||
| Neurological | Hepatic encephalopathy | 2 | 2 | ||||
| Coma | 1 | ||||||
| Brain edema | 1* | ||||||
| Brain bleeding | 1* | ||||||
| Gastrointestinal and hepatic | GI bleeding | 3 + | 3 (2*) | 4 (1*) | 1 | ||
| Nausea, vomiting | 1 | 2 | |||||
| Progressive jaundice | 1 | 1 | |||||
| Progressive ascites | 2 | ||||||
| Ltx-transplant rejection | 1 | ||||||
| Liver failure/decomp. | 1* | 1* | 3* | 6 (5*) | |||
| Cardiovascular | Hypotension/hemodynamic instability/shock | 3(2+, 1*) | 2 (1*) | 3 (2*) | |||
| Supra ventricular tachycardia | 1+ | ||||||
| Atrial fibrillation | 1 | ||||||
| Cardiac arrest | 1 | 1* | |||||
| Respiratory | Pulmonary bleeding | 1 | |||||
| Pulmonary infection/pneumonia | 1* | 1* | |||||
| Aspiration | 1 | ||||||
| Respiratory failure | 1* | 1* | 1* | ||||
| Hematologic | Leukopenia/pancytopenia | 1+ | |||||
| Thrombocytopenia | 2+ | ||||||
| Renal | Progressive renal dysfunction | 1 | |||||
| Acute renal failure | 1† | 1 | 2 | 1* | |||
| Hepatorenal syndrome | 1* | ||||||
| Systemic | Multi organ failure/sepsis other than septic shock | 1* | 4* | 1* | 8* | 5* | |
| Catheter related | Arterial mispuncture/bleeding hypotension | 3 | |||||
| External bleeding/Hematoma from Catheter insertion | 2 | ||||||
| Retroperitoneal bleeding from accidental femoral arterial slicing | 1* | ||||||
| Total SAEs | 20 (51%) | 8 (26%) | 7 (21%) | 3 (12%) | 29 (97%) | 24 (100%) | |
| Total death | 5 (13%) | 5 (16%) | 4 (12%) | 2 (8%) | 10 (33%) | 10 (42%) | |
| SAEs possibly related to ECAD | 9 | 0 | 0 | ||||
| SAEs related to catheter | 6 | 0 | 0 | ||||
- * (death).
- † (Possibly related to ECAD).
Laboratory Results
The single treatment effect for the 108 ECAD treatments resulted in the decrease of median values for: serum ammonia [72 (18–297) to 65 (9–313) μmol/L (P < 0.01)]; total serum bilirubin [16.4 (0.2–54.5) to 16.1 (1.9–42.1) mg/dL (P < 0.01)]; total bile acids [53 (11–249) to 41 (7–207) μmol/L (P < 0.01)]; blood urea nitrogen [33 (1–171) to 19 (2–114) mg/dL (P < 0.01)] and creatinine [1.7 (0.4–5.6) to 1.2 (0.3–7) mg/dL (P < 0.01)].
Patients who received ECAD had significant improvement in their laboratory values compared with baseline for creatinine, blood urea nitrogen, bile acids, branched chain/aromatic amino acid ratio, and serum ammonia levels, whereas the SMT group had an improvement in only the bile acid levels (Table 6).
| Test | SMT | ECAD | ||||||
|---|---|---|---|---|---|---|---|---|
| Baseline | EOS | P | % Change | Baseline | EOS | P | % Change | |
| Creatinine mg/dL | 1.7 (0.6–5) | 1.4 (0.4–5.7) | 0.096 | −13 (−77–67) | 1.7 (0.4–5.6) | 1.4 (0.4–4.5) | 0.001 | −18 (−68–133) |
| BUN mg/dL | 42.5 (2–136) | 48 (3–147) | 0.927 | −1 (−68–229) | 40 (6–171) | 20 (4–84) | 0.0001 | −38 (−88–217) |
| Bilirubin mg/dL | 12.2 (2.3–58.9) | 12.8 (3–57.4) | 0.134 | 10 (−79–91) | 15.8 (1.8–54.5) | 16.1 (3–38.5) | 0.064 | −7 (−60–352) |
| Bile acids μmol/L | 65.4 (12.2–247.1) | 54.5 (2–230) | 0.008 | −30 (−85–9) | 65.2 (38.1–249) | 61 (11–207) | 0.003 | −35 (−79–51) |
| BCAA/AAA [1] | 1.175 (0.62–2.49) | 1.04 (0.35–5.5) | 0.208 | 10 (52–378) | 0.96 (0.49–2.98) | 1.44 (0.57–3.37) | 0.031 | 26 (−30–271) |
| Ammonia μmol/L | 90.5 (34–786) | 63 (32–308) | 0.307 | −24 (−74–106) | 104 (43–449) | 60.5 (22–182) | 0.001 | −35 (−84–30) |
- Abbreviation: EOS, end of study.
- ECAD was done at a mean of 2.7 ± 1.5 sessions per patient. Although ECAD was offered at a mean of 2.7 sessions per patients, it resulted in significant improvement in ammonia, bile acids, branched chain/aromatic amino acid ratio (BCAA/AAA), creatinine, and BUN. However, SMT patients had significant improvements in their bile acids only. Values are expressed as medians and ranges. The P value is determined using the Wilcoxon test.
Per Protocol Analysis
The DSMB excluded 8 patients from per-protocol analysis; 5 patients did not meet protocol criteria and 3 patients had fewer than 3 HE assessments (2 due to early transplantation and 1 withdrew consent). Hence, 29 patients in the SMT group and 33 patients in the ECAD group qualified for per-protocol analysis. The IP for the ECAD group (median, 50; range, 0–100) was significantly higher than in the SMT group (median, 0; range, 0–80) (P = 0.007) as well as the time to first improvement (P = 0.017).
Predictors of Response
In patients with MELD ≥ 30, 15% of the SMT group had a 2-grade improvement in HE versus 68% of the ECAD group (P = 0.005). However, in patients with MELD scores less than 30, there was no relation between HE improvement response and treatment group (62.5% versus 53% in SMT and ECAD, respectively). In a post-hoc logistic analysis, we simultaneously considered treatment (SMT or ECAD), sex, race, age, and baseline CTP, ascites, HE, GCS, creatinine, total bilirubin, albumin, international normalized ratio, and MELD score as potential predictors of 2-grade response (yes/no). Of these 13 potential predictors, 3 baseline predictors were identified as simultaneously significant: receiving ECAD, CTP, and GCS scores. ECAD increased the response, whereas a high CTP and a low GCS score at baseline decreased the response (odds ratio = 0.56 for high CTP, odds ratio = 0.18 for low GCS). After controlling for these 2 predictors, the adjusted odds ratio for ECAD in achieving a 2-grade improvement in HE was 3.78 (P = 0.024). The other 10 factors including MELD were not statistically significant predictors after controlling for these 3.
Survival of the Study Population
During the 180 days of follow-up, 71% and 64% of the patients died, and 23% and 26% had liver transplantation in the SMT and ECAD groups, respectively. Responding to therapy was the only predictor for transplant-free survival in the first 4 weeks (P < 0.05) (Table 7). Patients who responded to therapy had a 4-week transplant-free survival of 47% versus only 20% in patients who failed to respond to any treatment during the 5 days.
| 2-Week Survival (Without Transplant) | 4-Week Survival (Without Transplant) | 4-Week Survival (With Transplant) | |||||
|---|---|---|---|---|---|---|---|
| Transplant | NA | NA | 14/15 | P < 0.001 | |||
| No transplant | NA | NA | 19/54 | ||||
| HE responders | 26/32 (81%) | P < 0.001 | 14/30 (47%) | P < 0.05 | 20/36 (56%) | NS. | |
| HE nonresponders | 10/29 (34%) | 5/25 (20%) | 13/33 (39%) | ||||
| MELD ≥ 30 | ECAD | 9/17 (53%) | P = 0.066 | 4/16 (25%) | P = 0.30 | 7/19 (37%) | P = 0.41 |
| SMT | 2/11 (18%) | 1/11 (10%) | 3/13 (23%) | ||||
| MELD < 30 | ECAD | 11/12 (92%) | P = 0.36 | 4/8 (50%) | P = 0.51 | 10/14 (71%) | P = 0.87 |
| SMT | 11/14 (79%) | 9/14 (64%) | 11/16 (69%) | ||||
- Abbreviations: NA, not applicable; NS, not significant.
- Responders are patients who had an improvement proportion > 0, whereas nonresponders had an improvement proportion = 0 within the 5-day study period.
Discussion
Severe HE is a major cause of morbidity and mortality in ESLD.14 Improvement of HE depends on correction of precipitating factors and the liver's functional capacity. In this prospective, randomized, controlled multicenter trial, performed in patients with ESLD and acute decompensation, severe HE responded poorly to conventional therapy and was associated with a poor prognosis, as previously noted.6, 15 All patients received standard medical therapy using nonabsorbable disaccharides or antibiotics, an approach under scrutiny itself.5 Our cohort had cirrhosis and a high MELD score, a marker of advanced liver failure and high short-term mortality.16
The study design followed the recommendations of a working party on studies in HE.17 It defined the study endpoints, included a control group, and established a methodology to quantify therapeutic effects considering the multiaxial definition of HE. The algorithm used to quantify improvement in HE grade (HE scoring algorithm) included clinical assessments and well-established simple neuropsychological tests that had a greater weight on the algorithm than the clinical tests as HE improved.
The endpoint of the study was improvement in HE of 2 grades from baseline. The time spent in an improved mental state during the 5 days of the study was measured using the improvement proportion for each group (Table 3). These strict definitions of response contrast with earlier pilot studies by Stange et al.,8 Heemann et al.,18 and Novelli et al.,19 which did not define the extent of HE improvement or the time spent in the improved state. In this study, although not all patients had improvement in their HE, ECAD was more effective than the SMT group in resolving advanced stages of HE. Furthermore, time to improvement was significantly shortened in the ECAD group compared with the SMT group (P = 0.044) in spite of the 5.25 hours' delay in starting ECAD after randomization. None of the surviving patients who improved their HE by 2 stages returned to their pre-study HE grade during the 5-day study period. The improvement of HE could be explained by the removal of both water-soluble and protein-bound putative toxins, reductions of ammonia concentration, bile acid levels, change in the ratio of branched/aromatic chain amino acids, or improvement in systemic hemodynamics.20 The significant decrease in ammonia concentration in this study could be related to the high flow rates used during ECAD therapy (Fig. 1).
Despite the disease severity of our study population, patients tolerated ECAD well using the MARS device. Of note, only 2 patients had their ECAD therapy discontinued because of hemodynamic instability. Hemodynamic effects of MARS have been reported by Schmidt et al.,21 who showed a significant increase in systemic vascular resistance during ECAD therapy. They proposed that removal of vasodilatory substances was a possible explanation for this finding. Thrombocytopenia is common in patients with decompensated ESLD and is perpetuated by extracorporeal therapies. In our study, patients who received ECAD required significantly more platelet transfusion than patients who only had SMT. Bleeding complications were reported at a lower rate in this study compared with other published data, possibly because only 26% of the sessions used low-dose heparin, and the treatment duration was limited to 6 hours.22. Three patients in the ECAD group experienced gastrointestinal bleeding in the first week after therapy, an experience not seen in previous studies.18, 19
Limitations of the study should be noted. The control group did not receive the ideal comparison, standard albumin-free dialysis, because this approach could not be justified based on the lack of supportive data. Evaluators were not blinded to the therapy received, but separate clinical and neuropsychological assessments were finalized by an independent coordinator to determine the grade of HE. There was a difference in the time to randomization, more prolonged in the MARS group (P = 0.07); although this difference could be a factor influencing results, a covariate analysis did not indicate an independent influence on the study. Finally, the specificity of the effects of MARS should be viewed with caution: A systematic review of 6 randomized controlled trials of different types of artificial devices for acute-on-chronic liver failure (including MARS) indicate an overall beneficial effect on survival.23 However, another more recent meta-analysis did not suggest improved survival with MARS in patients with acute-on-chronic failure.24 Therefore, the benefit of MARS in improving survival needs to be addressed further.
The MELD score of the cohort, 31 ± 10 (range, 11–50), predicted 77% mortality at 90 days.16 The SOFA score was 9 or greater in 85% of the patients, a factor associated with a mortality of greater than 80% during hospitalization.6 A post-hoc analysis examined the influence of treatment on survival. The results should be interpreted with caution, because the study was not designed to investigate the impact of ECAD on survival. Nonetheless, 2 factors were predictors of a 4-week survival: Performance of liver transplantation and improvement of HE by 2 grades. In addition, there was a trend for better 2-week transplant-free survival in patients with MELD of 30 or greater who received ECAD. This trend should be examined further in studies designed and powered to evaluate the impact of ECAD on survival with or without liver transplantation.
In conclusion, the role of albumin dialysis in patients with liver failure continues to be defined. In this study, the ECAD group using the MARS device responded earlier in terms of improvement in severe HE than SMT. This study introduces ECAD as a potentially effective modality for management of severe HE in patients with acute decompensation complicating ESLD who fail to respond to conventional therapy. The full role of MARS in the treatment of ESLD, with survival as an end-point, awaits the completion of ongoing controlled clinical trials.
Acknowledgements
In addition to the authors, the following investigators participated in this study:
Drs. Frederik Nevens and Alexander Wilmer of UZ Gasthuisberg, Leuven, Belgium
Dr. Robert Fontana of University of Michigan, Ann Arbor, Michigan
Dr. Stephen Caldwell of the University of Virginia, Charlottesville, Virginia.
The authors thank Deanna L. Oliver for her support in preparation of the manuscript. The authors also thank Grifols Inc. Los Angeles for supplying the albumin solution.
