Skip to main content

Foundational Trials

RELIEF

Image: Shutterstock / ChaNaWiT

Publication

  • Title: Extracorporeal Albumin Dialysis With the Molecular Adsorbent Recirculating System in Acute-on-Chronic Liver Failure: The RELIEF Trial
  • Acronym: RELIEF
  • Year: 2013
  • Journal published in: Hepatology
  • Citation: Bañares R, Nevens F, Larsen FS, Jalan R, Albillos A, Dollinger M, et al. Extracorporeal albumin dialysis with the molecular adsorbent recirculating system in acute-on-chronic liver failure: the RELIEF trial. Hepatology. 2013;57(3):1153-1162.

Context & Rationale

  • Background
    • Acute-on-chronic liver failure (ACLF) represents acute decompensation of cirrhosis with organ dysfunction and high short-term mortality, frequently driven by infection and/or haemorrhage, with deaths dominated by multi-organ failure rather than isolated hepatic biochemistry.1
    • Molecular adsorbent recirculating system (MARS) is an extracorporeal albumin dialysis platform designed to remove albumin-bound and water-soluble toxins, aiming to improve encephalopathy, renal dysfunction, and haemodynamics as a bridge to hepatic recovery or transplantation.
    • Pre-RELIEF evidence for MARS in ACLF was largely small, heterogeneous, and commonly focused on surrogate outcomes (laboratory indices, encephalopathy scores) rather than robust patient-centred endpoints.
  • Research Question/Hypothesis
    • Whether adding MARS to contemporary standard medical therapy improves short-term transplant-free survival in ACLF patients selected for severe hyperbilirubinaemia with major organ complications (hepatorenal syndrome and/or hepatic encephalopathy and/or rapidly progressive hyperbilirubinaemia).
  • Why This Matters
    • MARS is complex, costly, and resource-intensive; definitive evidence on hard outcomes is essential before widespread ICU/liver unit adoption.
    • A neutral mortality signal despite biochemical improvements would challenge the mechanistic assumption that toxin removal meaningfully alters the trajectory of ACLF multi-organ failure.

Design & Methods

  • Research Question: In adults with ACLF meeting severe hyperbilirubinaemia criteria plus major organ complications, does MARS + standard medical therapy increase 28-day transplant-free survival compared with standard medical therapy alone?
  • Study Type: Prospective, randomised, multicentre (19 European centres), open-label, parallel-group trial with stratified (MELD) permuted-block central randomisation; recruitment April 2003 to March 2009 (trial registration reported in manuscript).
  • Population:
    • Adults with cirrhosis and acute decompensation with an identifiable trigger; total bilirubin >5 mg/dL; after initial stabilisation and re-evaluation (minimum 24 hours after screening; randomisation within 24–48 hours after re-evaluation).
    • Required ≥1 of: hepatorenal syndrome (per International Ascites Club definition), hepatic encephalopathy grade ≥II, or rapidly progressive hyperbilirubinaemia (bilirubin >20 mg/dL at randomisation with >50% increase since admission).
    • Key exclusions included: progressive jaundice due to natural course or extrahepatic cholestasis; platelets <50,000/μL; INR >2.3 or disseminated intravascular coagulation; need for renal replacement therapy at inclusion or intrinsic renal disease; uncontrolled infection; active bleeding at inclusion; hepatocellular carcinoma >4 cm or portal vein thrombosis; severe cardiopulmonary disease; mean arterial pressure <60 mmHg despite vasopressors; major surgery within 4 weeks; HIV infection.
  • Intervention:
    • MARS + standard medical therapy.
    • Planned dosing: 4 consecutive daily sessions (days 1–4), then 3 sessions/week until “sustained improvement” or a maximum of 10 sessions within 21 days.
    • Session delivery: 6–8 h/day via central double-lumen catheter; blood flow 100–250 mL/min; albumin circuit flow 150 mL/min; anticoagulation according to local policy.
    • Stopping criteria for “sustained improvement” required simultaneously: creatinine <1.5 mg/dL; encephalopathy grade <I; bilirubin stable for 2 consecutive days without further treatment with >20% decrease from baseline.
  • Comparison:
    • Standard medical therapy alone, with protocolised guidance across centres for management of common precipitants and complications (e.g., antibiotics for infection/SBP, vasoactive therapy + albumin for HRS, standard therapies for variceal haemorrhage and encephalopathy, paracentesis for tense ascites).
    • Escalation/rescue therapies (including organ support) were permitted as clinically indicated.
  • Blinding: Unblinded (no sham extracorporeal therapy); primary outcome objective, but secondary outcomes (e.g., encephalopathy grading) vulnerable to detection bias.
  • Statistics: A total of 156 patients (78 per group) were required to detect a 20% absolute increase in 28-day survival (from 60% to 80%) with 80% power (beta 0.20) at the 5% significance level (alpha 0.05); target increased to 172 to allow ~10% dropouts. Primary analyses included both “intention-to-treat” (excluding major inclusion/exclusion violations) and per-protocol populations; survival compared with time-to-event methods and prespecified subgroup analyses.
  • Follow-Up Period: 90 days.

Key Results

This trial was not stopped early. Recruitment completed as planned across participating centres.

Outcome MARS + SMT SMT Effect p value / 95% CI Notes
Randomised (n) 95 94 Not applicable Not applicable 19 European centres; April 2003–March 2009
Included in “ITT” analysis set (n) 90 89 Not applicable Not applicable 10 excluded after randomisation due to major inclusion-criteria violations (5 per group)
Included in per-protocol analysis set (n) 71 85 Not applicable Not applicable Per-protocol exclusions: MARS 19 vs SMT 4; 12 MARS exclusions due to <3 sessions
Primary: 28-day transplant-free survival (ITT) 60.7% 58.9% Not reported P=0.79 Three liver transplants in each group within follow-up (ITT: 3.4% per group)
Primary: 28-day transplant-free survival (per-protocol) 60.0% 59.2% Not reported P=0.88 Only one transplant reported in per-protocol population (MARS arm; 1.4%)
90-day transplant-free survival (ITT) 46.1% 42.2% Not reported P=0.71 No survival separation on Kaplan–Meier analysis
90-day transplant-free survival (per-protocol) 44.7% 43.7% Not reported P=0.97 No survival separation on Kaplan–Meier analysis
28-day mortality (adjusted model) Not reported Not reported OR 0.87 95% CI 0.44 to 1.72; P=0.694 Adjusted for MELD >20 and spontaneous bacterial peritonitis as triggering event
Renal recovery at day 4 in baseline HRS (per-protocol): creatinine <1.5 mg/dL 16/34 (47.1%) 10/38 (26.3%) OR 0.40 95% CI 0.15 to 1.07; P=0.07 Early signal only; effect not sustained at day 21 (reported qualitatively)
Encephalopathy improvement at day 4 in baseline HE grade II–IV (per-protocol): improved to grade 0–I 15/24 (62.5%) 13/34 (38.2%) OR 0.37 95% CI 0.12 to 1.09; P=0.07 Early signal only; effect not sustained at day 21 (reported qualitatively)
Total bilirubin (day 4 laboratory change set): baseline → day 4 26.30 (11.66) → 17.58 (6.90) 26.65 (11.54) → 24.15 (11.30) Not reported P<0.001 % change: −26.4 (26.12) vs −8.92 (9.47)
Creatinine (day 4 laboratory change set): baseline → day 4 2.24 (2.01) → 1.43 (1.09) 2.13 (1.97) → 1.67 (1.29) Not reported P=0.022 % change: −20.04 (35.06) vs −6.43 (33.50)
Haemoglobin (day 4 laboratory change set): baseline → day 4 9.88 (1.59) → 8.99 (1.16) 10.26 (1.99) → 10.10 (1.79) Not reported P=0.009 Consistent with extracorporeal circuit-related blood loss/haemodilution and/or bleeding risk
Platelets (day 4 laboratory change set): baseline → day 4 133.17 (76.62) → 91.14 (65.05) 122.71 (73.90) → 114.08 (73.47) Not reported P<0.001 % change: −29.06 (29.37) vs −1.80 (40.59)
Mechanical ventilation during hospitalisation 22.5% 21.2% Not reported P=0.838 No between-group difference
Length of hospital stay (median; range) 24 (2–28) days 23 (1–28) days Not reported Not reported No evidence of shorter admission with MARS
Severe adverse event: significant bleeding (any source) 17/95 (17.89%) 9/94 (9.57%) Not reported P=0.145 Variceal bleeding: 5/95 vs 5/94 (P=0.892)
Severe adverse event: bacterial infection (where recorded) 32/54 (59.3%) 33/67 (49.3%) Not reported P=0.273 Death related to infection: 10/32 (31.3%) vs 9/44 (20.5%) (P=0.283)
  • Despite significant early improvements in bilirubin (17.58 vs 24.15 mg/dL at day 4; P<0.001) and creatinine (1.43 vs 1.67 mg/dL at day 4; P=0.022), MARS did not improve 28-day or 90-day transplant-free survival.
  • Signals for early renal recovery (HRS: 47.1% vs 26.3% at day 4; P=0.07) and encephalopathy improvement (62.5% vs 38.2% at day 4; P=0.07) did not translate into a definitive clinical outcome benefit.
  • Haematological effects were evident by day 4 with MARS (haemoglobin 8.99 vs 10.10 g/dL; P=0.009; platelets 91.14 vs 114.08 ×103/μL; P<0.001), and bleeding events were numerically higher.

Internal Validity

  • Randomisation and Allocation:
    • Central computerised randomisation with stratification by MELD and permuted blocks; allocation performed via secure randomisation site.
    • Stratification by disease severity (MELD) is methodologically appropriate given strong confounding by baseline risk.
  • Drop out or exclusions:
    • Post-randomisation exclusions from the “ITT” analysis set: 10/189 (5 per group) due to major inclusion-criteria violations, yielding ITT n=179.
    • Per-protocol exclusions were asymmetric (MARS 19 vs SMT 4), including 12 patients allocated to MARS who received <3 sessions, plus other deviations/withdrawals.
    • The differential per-protocol attrition risks biased estimates if exclusions correlate with prognosis (especially given feasibility constraints for extracorporeal therapy in the sickest).
  • Performance/Detection Bias:
    • Unblinded design introduces risk of co-intervention imbalance; however, the primary endpoint (death/transplant-free survival) is objective.
    • Secondary outcomes (encephalopathy grades, trigger identification, escalation decisions) are more vulnerable to subjective assessment and clinician behaviour.
  • Protocol Adherence:
    • Among per-protocol MARS patients, mean number of sessions was 6.5 (SD 3.1), with median session duration 6.8 h (range 2.4–10.5).
    • Delivered treatment intensity corresponds to MARS exposure on ~16.5% of days in the first 21 days (as reported by trialists), reflecting intermittent rather than continuous toxin clearance.
  • Baseline Characteristics:
    • Groups were broadly comparable at baseline in the ITT set: MELD 25.6 (8.2) vs 24.1 (9.6); bilirubin 27.2 (12.8) vs 26.9 (12.4) mg/dL; creatinine 2.42 (2.13) vs 2.24 (2.04) mg/dL.
    • Markers of phenotype balance showed some numerical differences (e.g., spontaneous bacterial peritonitis as trigger 14.4% vs 6.7%; HE grade >II 35.6% vs 43.8%), but without strong evidence of systematic imbalance in the ITT set.
  • Heterogeneity:
    • Multicentre delivery across 19 centres supports generalisability but increases between-centre variation (e.g., anticoagulation strategy, organ support thresholds), potentially diluting effect estimates.
    • Standard medical therapy was defined and periodically updated, which improves contemporaneity but introduces temporal heterogeneity over the prolonged recruitment interval.
  • Timing:
    • Randomisation occurred after a minimum 24-hour assessment and a subsequent 24–48 hour re-evaluation, selecting patients who remained eligible after early stabilisation and potentially delaying extracorporeal initiation.
  • Dose:
    • Planned maximum of 10 sessions over 21 days; actual mean 6.5 sessions in per-protocol recipients suggests moderate intensity relative to the severity and rapid trajectory of ACLF.
  • Separation of the Variable of Interest:
    • Bilirubin at day 4: 17.58 (6.90) vs 24.15 (11.30) mg/dL (P<0.001).
    • Creatinine at day 4: 1.43 (1.09) vs 1.67 (1.29) mg/dL (P=0.022).
    • Platelets at day 4: 91.14 (65.05) vs 114.08 (73.47) ×103/μL (P<0.001).
  • Outcome Assessment:
    • Primary outcome is patient-centred and objective; follow-up to 90 days reduces short-horizon interpretive ambiguity.
  • Statistical Rigor:
    • Prespecified power calculation and predefined subgroup analyses were reported; however, the assumed 20% absolute survival improvement is large and may be optimistic for a complex, multi-organ syndrome.
    • Use of both ITT and per-protocol as co-primary analysis sets complicates interpretability when protocol adherence differs materially by arm.

Conclusion on Internal Validity: Overall, internal validity appears moderate: randomisation and primary outcome objectivity support robustness, but open-label care and substantial differential per-protocol exclusions (with feasibility-related under-delivery of MARS) limit causal certainty regarding treatment intensity and subgroup responsiveness.

External Validity

  • Population Representativeness:
    • Predominantly European tertiary centres, with common ACLF precipitants (infection, GI bleeding, alcoholic hepatitis) and severe hyperbilirubinaemia; broadly representative of high-resource liver/ICU practice.
    • Exclusions (severe coagulopathy, dialysis requirement at inclusion, uncontrolled infection/bleeding, refractory shock) limit applicability to the sickest ACLF phenotypes seen in ICU.
  • Applicability:
    • MARS availability, staff expertise, and cost constrain implementation outside specialist centres; translation to resource-limited settings is limited.
    • Low transplantation rates in the trial reduce direct inference regarding “bridge-to-transplant” strategy settings.

Conclusion on External Validity: Generalisability is moderate: findings apply best to ACLF patients who are stable enough to receive intermittent extracorporeal therapy in specialist centres, and less well to more haemodynamically unstable, profoundly coagulopathic, or dialysis-dependent ICU populations.

Strengths & Limitations

  • Strengths:
    • Largest dedicated multicentre randomised evaluation of MARS in ACLF to date, across 19 European centres.
    • Clinically meaningful primary endpoint (28-day transplant-free survival) with 90-day follow-up.
    • Stratified randomisation by MELD and pre-specified subgroup analyses.
    • Documented biological separation (bilirubin/creatinine) demonstrating that the intervention had expected short-term physiological effects.
  • Limitations:
    • Open-label design without sham therapy; risk of co-intervention and assessment bias for secondary outcomes.
    • Modified ITT definition (excluding post-randomisation inclusion violations) and pronounced imbalance in per-protocol exclusions (MARS 19 vs SMT 4).
    • Under-delivery/feasibility constraints: 12 MARS patients received <3 sessions and were excluded from per-protocol analysis.
    • Prolonged recruitment (2003–2009) with evolving “standard medical therapy” could dilute or confound treatment effects over time.
    • Power calculation assumed a very large absolute survival benefit (20%), potentially underpowering detection of smaller but clinically relevant effects.

Interpretation & Why It Matters

  • Clinical practice
    MARS should not be expected to improve short-term transplant-free survival when applied as intermittent liver support in broadly-selected severe ACLF, despite early biochemical improvements.
  • Mechanistic inference
    The dissociation between toxin/laboratory improvement and survival suggests that ACLF mortality is not primarily reversible by toxin clearance alone, reinforcing the primacy of systemic inflammation, infection control, and multi-organ support strategies.
  • Trial design implications
    Future extracorporeal support trials require sharper phenotyping, earlier initiation, and clinically coherent endpoints (e.g., ACLF resolution, transplant-free survival in those listed, or organ support-free days), alongside rigorous strategies to ensure treatment delivery feasibility in the sickest.

Controversies & Subsequent Evidence

  • Concordance with other large device trials:
    • Another major multicentre extracorporeal liver support trial (Prometheus/FPSA; HELIOS) similarly failed to demonstrate an overall survival benefit in ACLF, reinforcing the broader uncertainty that extracorporeal detoxification alone alters hard outcomes in heterogeneous ACLF populations.2
  • Surrogate improvement versus patient-centred benefit:
    • High-level contemporary commentary around albumin “regeneration” devices emphasised the recurring pattern of biochemical improvement without definitive survival translation, arguing for better patient selection and endpoints aligned to realistic biological targets.3
  • Meta-analytic synthesis and treatment intensity:
    • Individual patient data meta-analysis has evaluated albumin dialysis across ACLF trials with specific focus on treatment intensity, highlighting that any putative benefit may be contingent on delivered dose and patient severity, but remains uncertain given heterogeneity and feasibility constraints in the sickest patients.4
    • Network meta-analysis has underscored persistent uncertainty and imprecision in estimates for liver support systems in ACLF, reflecting limited high-quality comparative evidence and variability in devices, populations, and co-interventions.5
  • Evolution of extracorporeal strategies:
    • Newer albumin-focused devices targeting dysfunctional albumin biology and inflammatory mediators (e.g., DIALIVE) have reported improvements in clinically relevant intermediate outcomes (e.g., ACLF resolution and organ dysfunction trajectories), supporting ongoing innovation but not superseding the need for adequately powered mortality-focused trials.6
  • Guidelines and practice positioning:
    • Recent European guidance on ACLF does not support routine use of extracorporeal liver support systems as standard therapy in ACLF, reflecting a persistent lack of consistent mortality benefit and resource implications.7
    • Contemporary American practice guidance similarly positions management around early recognition, trigger treatment, and organ support, with extracorporeal liver support generally confined to selected specialist-centre use and/or research contexts rather than routine adoption for mortality benefit.8

Summary

  • RELIEF was a 19-centre European randomised trial testing MARS + standard medical therapy versus standard medical therapy alone in severe ACLF (randomised n=189; ITT n=179).
  • MARS did not improve 28-day transplant-free survival (60.7% vs 58.9%; P=0.79) or 90-day transplant-free survival (46.1% vs 42.2%; P=0.71) in the ITT set.
  • MARS produced clear short-term biological effects at day 4 (bilirubin 17.58 vs 24.15 mg/dL; P<0.001; creatinine 1.43 vs 1.67 mg/dL; P=0.022).
  • Early signals of improvement in hepatorenal syndrome and hepatic encephalopathy did not reach conventional significance and did not translate into survival benefit.
  • Per-protocol analysis was complicated by substantially higher protocol non-adherence/under-delivery in the MARS arm (19 vs 4 exclusions), emphasising feasibility constraints of extracorporeal support in the sickest ACLF patients.

Further Reading

Other Trials

Systematic Review & Meta Analysis

Observational Studies

Guidelines

Notes

  • RELIEF demonstrates “biological plausibility without outcome translation”: early toxin/renal surrogate improvements did not change transplant-free survival.
  • Interpretation of per-protocol findings requires caution due to major asymmetry in feasibility-related exclusions in the extracorporeal therapy arm.

Overall Takeaway

RELIEF is a landmark negative multicentre trial showing that, in broadly-selected severe ACLF, intermittent MARS albumin dialysis improves short-term biochemical and organ-dysfunction surrogates but does not improve 28-day or 90-day transplant-free survival. Its enduring impact is methodological and conceptual: it re-anchors extracorporeal liver support from “promising physiology” to the requirement for demonstrable patient-centred benefit, shaping modern guideline caution and driving more targeted device development and endpoint selection.

Overall Summary

  • No transplant-free survival benefit at 28 or 90 days with MARS + SMT versus SMT alone, despite early improvements in bilirubin and creatinine.
  • Feasibility and treatment delivery (dose/intensity) are central threats to effect detection in extracorporeal liver support trials.
  • Subsequent evidence and guidelines support ongoing innovation but do not justify routine extracorporeal albumin dialysis for ACLF outside specialist contexts or trials.

Bibliography