Article Text

Download PDFPDF

Early features of acute-on-chronic alcoholic liver failure: a prospective cohort study
  1. Aezam Katoonizadeh1,2,
  2. Wim Laleman1,
  3. Chris Verslype1,
  4. Alexander Wilmer3,
  5. Geert Maleux4,
  6. Tania Roskams2,
  7. Frederik Nevens1
  1. 1Department of Hepatology, University Hospital Gasthuisberg, KU Leuven, Leuven, Belgium
  2. 2Morphology and Molecular Pathology, University Hospital Gasthuisberg, KU Leuven, Leuven, Belgium
  3. 3Medical Intensive Care Unit, University Hospital Gasthuisberg, KU Leuven, Leuven, Belgium
  4. 4Interventional Radiology, University Hospital Gasthuisberg, KU Leuven, Leuven, Belgium
  1. Correspondence to Dr Wim Laleman, Department of Liver and Pancreaticobiliary Disorders, University Hospital Gasthuisberg, KU Leuven, Herestraat 49, 3000 Leuven, Belgium; wim.laleman{at}


Background ‘Acute-on-chronic liver failure’ (ACLF) is characterised in a more advanced stage by liver failure associated with multiple other end-organ failure. The global clinical characteristics of this entity remain, however, ill-defined.

Objective To characterise and evaluate the clinicopathological features of patients with ACLF compared with patients with chronic decompensated cirrhosis (CHD) in a prospective, homogeneous cohort of patients with histologically proven alcoholic cirrhosis from 2002 to 2007.

Results In total 250 patients were screened (ACLF (n=70, 28%) and CHD (n=180, 72%)). Alcoholic liver disease was observed in respectively 61/70 (87%) of patients with ACLF and 72/180 (40%) of patients with CHD. After exclusion of 31 patients, 102 patients were studied: 54 with ACLF (median age 51 years; Child–Pugh 12±2; in-hospital mortality 46% (25/54)) and 48 patients with CHD (median age 53 years; Child–Pugh 10±2; in-hospital mortality 10% (5/48)). In the patients with ACLF who survived the hospital stay, the difference in transplant-free survival compared with patients with CHD tended to attenuate with time. At admission the apparent infection of patient groups was comparable but during hospitalisation infection occurred more frequently in patients with ACLF (31/53 (58%)) than in those with CHD (12/47=26%) (p=0.007). Early signs of infection, positive systemic inflammatory response syndrome (SIRS) criteria at admission and ductular bilirubinostasis (p=0.04), were early features that predicted outcome in ACLF.

Conclusion Patients with ACLF have a high short-term mortality but those who survived the acute exacerbation show a long-term outcome comparable to that of patients with CHD. Infection is the most common cause of mortality in these patients. Positive SIRS criteria and ductular bilirubinostasis are early markers of ACLF and might allow more rapid identification of high-risk patients.

  • Acute-on-chronic liver failure
  • infection
  • ductular bilirubinostasis
  • SIRS
  • sepsis
  • alcoholic cirrhosis
  • alcoholic liver disease
  • bacterial infection
  • cirrhosis
  • liver failure

Statistics from

Significance of this study

What is already known about this subject?

  • ‘Acute-on-chronic liver failure’ (ACLF) is described as an acute and rapid deterioration of liver function accompanied by subsequent multiple end-organ failure in a patient with previously well-compensated liver disease owing to the effects of a precipitating event.

  • The mortality of ACLF is know to be high in hospital but no long-term data are available.

  • The global clinical and histological characteristics of ACLF are ill-defined and no prospective, comparative data are available.

What are the new findings?

  • This is the first prospective comparative study of the clinicopathological characteristics of ACLF.

  • Patients with ACLF have a high-short term mortality but in those who survived the hospital stay, the difference in transplant-free survival from that of patients with chronic decompensated cirrhosis tended to attenuate with time.

  • Infection is the most common cause of mortality in patients with ACLF.

  • The presence of a temporary immune dissonance with early features of systemic inflammatory response syndrome (SIRS) and ductular bilirubinostasis as early signs of infectious complications is confirmed.

How might it impact on clinical practice in the foreseeable future?

  • The identified early markers, positive SIRS criteria and ductular bilirubinostasis, might be helpful in detecting ACLF at a stage of the disease when therapeutic interventions might still be effective.


Liver failure, whether liver insufficiency in chronic disease or occurring de novo in acute disease, remains a condition with poor outcome and high mortality.1 2 For chronic liver insufficiency the term ‘acute-on-chronic liver failure’ (ACLF) and chronic hepatic decompensation (CHD) were introduced several years ago in an attempt to subtype different clinical courses in patients with cirrhosis and liver insufficiency.3 In its first working definition ACLF was described as an acute and rapid deterioration of liver function in a patient with previously well-compensated liver disease owing to the effects of a precipitating event. These precipitants might involve either direct hepatotoxic factors (such as alcohol, drugs, etc) or indirect injury-inducing events (such as infection, variceal bleeding, surgery, ….). Clinically, this syndrome is characterised by jaundice, hepatic encephalopathy and/or hepatorenal syndrome and a rapidly evolving multiorgan failure. ACLF leads to a short-term mortality of 50–90%.1 3 4 The importance of this syndrome resides in the conceptual suggestion of reversibility or recompensation since a patient who can survive an episode of liver failure and associated subsequent multiorgan dysfunction, may re-emerge above the critical threshold of functional liver cell mass, thus precluding the need for transplantation and having an acceptable long-term prognosis.

ACLF should be distinguished from chronic relentless hepatic decompensation (CHD), which usually occurs in patients with end-stage cirrhosis as a result of progression of their underlying primary liver disease. Because this is deemed irreversible owing to loss of regeneration potential, liver transplantation is considered the only therapeutic option in these patients.1–3 Although in theory, these different clinical concepts seem straight forward, in practice, the clinical picture is often complex, with lack of identification of a clear precipitant and problems in determining the ideal moment and/or necessity or usefulness of a certain therapeutic intervention.1 4 Moreover, the pathophysiological pathway leading from compensated to a suddenly decompensated state remains not entirely elucidated. Ample evidence suggests that dysregulated inflammation has an important role owing to an imbalanced host response to injury with a self-perpetuating effect of liver insufficiency.5–14

In this study, we aimed to characterise patients with ACLF, identified by a clinical working definition3 (see ‘Patients and Methods’), in a homogeneous group of patients with histologically proven alcoholic cirrhosis, and to determine early clinical and histological features which might allow earlier recognition of the syndrome.

Patients and methods

Study population and treatment

This is a prospective single-centre study, conducted at a tertiary liver unit from 2002 until 2007, investigating the characteristics and outcome of patients with ACLF and their relative differences from patients with CHD. In order to study a homogeneous group of patients, only those with histologically proven alcoholic cirrhosis were included.

A diagnosis of ACLF was made if patients, previously known to have compensated chronic liver disease (as defined by Ginés et al15), developed intrahepatic cholestasis (serum bilirubin >5 mg/dl without radiological evidence of extrahepatic origin) and had at least one of the following complications within a period of 4–8 weeks after an identifiable acute hepatic insult: (a) a progressive hyperbilirubinaemia defined as a >50% increase of bilirubin or up to a level of >20 mg/dl; (b) hepatic encephalopathy ≥II; (c) de novo development of ascites and/or (d) hepatorenal syndrome, defined according to the current criteria of the International Ascites Club16 (Major criteria: (a) the presence of chronic liver disease and portal hypertension; (b) low glomerular filtration rate indicated by a creatinine >1.5 mg%; (c) the absence of shock, gastrointestinal or renal fluid losses, recent treatment with nephrotoxic drugs; (d) lack of effect after expansion of plasma volume with isotonic saline or following diuretic withdrawal and (e) lack of proteinuria >500 mg/dl and ultrasonographic evidence of obstructive uropathy or parenchymal renal disease).

This overall definition corresponds to the previously published suggested working definition by Jalan and Williams,3 which was recently accepted at the consensus meeting on ACLF of the Asian Pacific Association for the study of the liver.17 Special consideration was directed towards detection of precipitating events. Thus an extensive history was taken for intake of different drugs or any other xenobiotic agents over the past 3 months, bleeding or infection. A superimposed viral infection was ruled out by appropriate serology (eg, HAV, HBV, HCV,…).

If during hospitalisation patients fulfilled the above-mentioned criteria for ACLF, a transjugular liver biopsy was performed within 48–72 h after the suggestion of potential eligibility. For the purpose of this study time zero was not the time of histologically proven diagnosis but the moment of admission at the hospital. All patients were treated with standard of care, including high caloric nutrition (at least 30 kcal/kg/day). Hepatic encephalopathy was treated with lactulose in order to obtain two to three bowel movements a day. Protein was restricted (60 g/day) only for a limited period in patients with stage 3 or more encephalopathy. Hepatorenal syndrome was treated with albumin (1 g/kg on the first day and 20–40 g/day thereafter) and terlipressin (with a starting dose of 0.5 mg every 4 h, increasing to 2.0 mg after 48 h). For patients with severe alcoholic steatohepatitis (ASH) (Maddrey score >32), and after excluding infection, prednisolone 40 mg/day was initiated. If no significant fall in bilirubin was seen (‘early change in bilirubin level’ (ECBL)) after 1 week, treatment was discontinued.18 During treatment with prednisolone, all patients received prophylaxis with quinolones and gastric acid suppression by proton pump inhibitors. Renal replacement treatment was offered for patients with uraemic symptoms, severe electrolyte disturbances or uncontrollable volume excess. Variceal bleeding was treated with a combined medical and endoscopic approach and if this failed, a rescue transjugular intrahepatic portosystemic shunt was used.19 20 Ascites was treated accordingly with low salt intake, spironolactone and/or furosemide and large-volume paracentesis with adequate albumin substitution.21 22 Prophylactic antibiotics (quinolones) were given, both to patients with CHD and ACLF if they had tense ascites and Child's class C cirrhosis, as primary prevention of spontaneous bacterial peritonitis.23 24 Patients who were treated with albumin dialysis (MARS or Prometheus) were excluded in order to eliminate an interventional bias.

The characteristics and natural history of these patients was compared with those of patients with chronically decompensated alcoholic cirrhosis (CHD) referred during the same period and evaluated for eligibility for liver transplantation. In this group of patients, transjugular biopsy was equally routinely performed in order to assess cirrhosis, hepatic venous pressure gradient and to rule out continuing alcohol abuse. As for the patients with ACLF, time zero was not the moment of biopsy but the moment of admission.

In both groups, patients with hepatocellular carcinoma, extrahepatic cholestasis, severe intrinsic renal disease or without histologically proven cirrhosis were excluded.

The local ethics committee approved this prospective registration of data of patients with complicated alcoholic liver disease for whom transjugular liver biopsy is part of our routine evaluation and in whom no other specific invasive procedure was performed for the purpose of this study.

Clinical parameters

Within the first 48 h of admission several clinical parameters, including systemic inflammatory response syndrome (SIRS)25 and infectious parameters, were systematically registered. All patients were intensively followed up by subsequent daily clinical assessment and blood examination during hospital stay. Thereafter, patients were re-evaluated at the outpatient clinic every 2– 4 weeks until death, transplantation or resolution of decompensation. Since infections usually start off subclinically in these patients, systemic screening for infection was routinely performed in all patients at admission. Patients were considered to have an infection if they met one or more of the following conditions: (a) ascitic polymorphonuclear cells >250×106/l; (b) clinical findings of infection confirmed by culture positivity (urine, blood or ascites) or (c) radiological confirmation for cases of respiratory infection. SIRS and sepsis were defined according to the American College of Chest Physicians/Society of Critical Care Medicine consensus conference.25 SIRS was considered to be present if two or more of the following conditions were met: temperature >38°C or <36°C, heart rate >90 bpm, respiratory rate >20 breaths/min or arterial carbon dioxide tension <32 mm Hg, white blood cell count >12×109/l or <4×109/l. The parameters of the different clinical prognostic models (Child–Pugh, MELD, Maddrey and SOFA)26–28 were also systematically registered.


The slides were reviewed by two observers (TR and AK) in consensus using a multiheaded microscope. Liver specimens were assessed for a wide spectrum of histological features: steatosis; ballooning degeneration; lobular and portal inflammation; Mallory bodies; satellitosis; apoptosis megamitochondria; parenchymal and ductular bilirubinostasis; fibrosis; amount of the residual parenchyma based on the thickness of septa; iron deposition; ceroid macrophages; cholangiolitis; number of proliferating hepatocytes, intermediate hepatocytes and hepatic progenitor cells. These last cells represent a major cellular source of regeneration when the regenerative capacity of mature cells is impaired29

(Immuno)histochemical staining and histopathological evaluation and scoring

Formalin-fixed, paraffin-embedded liver specimens were used for routine histochemical and immunohistochemical staining. Histochemical staining included (a) haematoxylin and eosin (H&E) to evaluate necroinflammation (steatosis, lobular inflammation, cell death and ballooning degeneration) based on a recent proposed scoring system for alcoholic liver disease by Yip and Burt30 and to detect megamitochondria, satellitosis and cholangiolitis (each was scored as absent or present) and Mallory bodies (scored as 0=absent, +1=a few and +2=many); (b) periodic acid–Schiff after predigestion with amylase to detect ceroid macrophages (scored as 0=absent, +1=a few and +2=many) and cytoplasmic inclusions in hepatocytes like α1 antitrypsin; (c) Sirius red to evaluate thickness of septa and the amount of residual liver parenchyma (assessed as being <50% or >50% compared with normal liver). Pericellular fibrosis (scored as absent or present) was also assessed by this staining; (d) Perl's staining to evaluate the severity of hepatocyte iron deposit. A semiquantitative method was adapted from Scheuer31 as follows: 0=no hepatocellular iron, +1=staining present in a very few hepatocytes, +2=staining present in 5–25% of the hepatocytes, +3=staining present in 25–50% of hepatocytes and +4=staining present in >50% of hepatocytes; (e) Hall's staining to detect bilirubinostasis both in parenchyma and in ductules (scored as 0=absent, +1=focal and +2=marked).

Immunohistochemical staining included CK7 (dilution 1/50, Dako, Gllstrup, Denmark), CK19 (dilution 1/20, Dako) to count the number of hepatic progenitor cells (HPCs) and intermediate hepatocytes, Ki 67 (Mib-1) (dilution 1/100, Dako), to count the number of proliferating hepatocytes and ubiquitin (dilution 1/100, Dako) to confirm the Mallory bodies which were seen on H&E stained sections. The number of HPCs and intermediate hepatocytes was assessed by calculating the average number of HPCs and intermediate hepatocytes per high-power field based on a count of these cells in three to five non-overlapping high-power fields (depending on available portal tracts or parenchyma), using a 40× objective. For proliferating hepatocytes this assessment was made by counting the number of positive nuclei using a 40× objective in three to five non-overlapping fields depending on the remaining parenchyma. The average of these numbers was then taken.

The detailed histological features of ASH were defined by the presence of lobular inflammation, cell death, steatosis and ballooning degeneration, proposed as a working definition by Yip and Burt.30 In this model each feature was scored from 0 to 3 except for ballooning which was scored from 0 to 2, with a total score ranging from 0 to 11. We developed the following four-grade scoring system to evaluate the necro-inflammation: 0=no ASH; 1–3=mild; 4–6=moderate; >6=severe.

Statistical analysis

Data are expressed as median ± interquartile range (IQR) for continuous variables, and frequency with percentage for categorical variables. Continuous variables were compared using an unpaired t test (for normally distributed ones) or non-parametric test such as the Wilcoxon test (for those with skewed distribution). The categorical variables were analysed by χ2 tests (χ2 Pearson or Fisher exact test). In-hospital mortality was considered as a dependent variable and variables such as age, sex, serum bilirubin and creatinine levels, C-reactive protein, SIRS, etc, were considered as independent variables. To identify independent predictors of in-hospital mortality, multivariate logistic regression analysis was performed. Variables with statistical significance on univariate analysis were used for multivariate analysis. The discrimination ability of different clinical prognostic models (MELD, SOFA, Child–Pugh and Maddrey) to predict the outcome was evaluated by using the area under a receiver operating characteristic curve. The Youden index (sensitivity + specificity−1) was used to select the best cut-off point. A p value ≤0.05 was considered as statistically significant. JMP 7.0 statistical software (from SAS) was used for analysis.


Characteristics of the two patient groups and outcome

In total 250 patients were screened for inclusion (figure 1): 180 had CHD and 70 ACLF. Thus, ACLF accounted for 28% of patients admitted to our unit with liver insufficiency associated with chronic liver disease. Alcoholic liver disease was seen in 72/180 (40%) patients with CHD and in 61/70 (87%) patients with ACLF. The underlying liver disease in the other patients with ACLF were: non-alcoholic steatohepatitis (n=3), HBV (n=3), HCV (n=2) and autoimmune hepatitis (n=1). In the ACLF group, seven patients (7/61, 11%) were additionally excluded because of the absence of cirrhosis on histology. The characteristics of these seven patients, with a median (IQR) age of 49 years (11), Child–Pugh of 12 (2), MELD of 24 (5) and Maddrey of 44 (23), did not differ from those of the other patients with ACLF, but they all survived to hospital discharge. Forty-five of the 54 remaining patients with acute-on-chronic alcoholic liver failure (83%) experienced a first episode of liver decompensation, whereas nine (17%) had experienced a previous episode of decompensation but had become compensated for at least 6 months before entry to the study.

Figure 1

Overview of patient selection (*four patients met two exclusion criteria). HCC, hepatocellular carcinoma.

Cirrhosis was histologically confirmed in all the patients with CHD. Twenty-four patients in this group were excluded because of either the presence of hepatocellular carcinoma (n=11), intrinsic renal disease (n=2), treatment with liver-assist devices (n=13) or concomitant aetiology (viral+alcohol) of the underlying chronic liver disease (n=2).

Consequently, the study group comprised 102 patients: 54 with ACLF and 48 with CHD. The median interval between the moment of admission and histologically proven diagnosis of cirrhotic ACLF or chronically decompensated alcoholic cirrhosis was 7 days (IQR 9). Potential precipitating events inducing ACLF included variceal bleeding (n=5), apparent infection (n=12) and excessive alcohol intake (n=13). Spontaneous bacterial peritonitis (n=4), urinary tract infection (n=4), pneumonia (n=2) and cryptogenic haematogenic infection (n=2) were identified as a focus of infection. Multiresistant bacteria, involving methicillin-resistant Staphylococcus aureus, were diagnosed in one patient in both groups. There were no cases of viral (eg, HAV, HBV,…), drug-induced or postoperative acute liver impairment. For the remaining 24 patients, no clear precipitating event could be recognised when ACLF was diagnosed.

Table 1 shows the demographic and clinical characteristics of the two patients groups at admission. Patients with ACLF were proportionally more actively drinking and had consumed significantly more units of alcohol in the last month before admission than those with CHD (192±63 vs 43±8 g/day, respectively). Since patients with ACLF were selected on the presence of rapidly increasing jaundice and/or encephalopathy and/or renal impairment, these complications were more frequently found in this group.

Table 1

Outcome and clinical characteristics of patients with ACLF and CHD within 48 h of admission

During hospitalisation, there was no significant difference in gastrointestinal bleeding between the patient groups. The cause of gastrointestinal bleeding in the ACLF group was variceal bleeding (n=10), gastritis (n=1) and oesophageal ulcer (n=1). In the CHD group the causes of GI bleeding included varices (n=9), portal hypertensive duodenopathy/bulbar ulcer (n=1) and cardia ulcer (n=1).

Patients with ACLF were more frequently admitted to the intensive care unit (59% vs 23% for CHD) and were in need more often in need of mechanical ventilation (31 vs 10%). Renal replacement was not statistically more frequently needed in the ACLF group.

In the ACLF group, treatment with prednisolone 40 mg/day of 9/13 patients with severe ASH (Maddrey score >32) was started. Four patients were not treated because of concurrent infection, of whom three died from this infection. Of the nine treated patients, four achieved ECBL (44% responders) and continued to receive steroids for 28 days. Of those, three survived and one patient died of variceal bleeding. In the non-responder group (5/9), three patients died because of either multiple organ failure (n=1) or sepsis (n=2) within 10 days after discontinuation of steroids. In the CHD group, treatment with steroids of three patients with superposed severe ASH was started. None of them realised ECBL and all died of sepsis during hospitalisation.

In-hospital mortality was significantly higher in the ACLF group (25/54, 46%) than in the CHD group (5/48, 10%) (p=0.0001). In ACLF, mortality was attributed to hepatorenal syndrome (n=2), variceal (re-)bleeding (n=5), multiorgan failure without obvious infectious trigger (n=6) and sepsis (culture proven)-related complications (n=12). In the CHD group, variceal bleeding was the immediate cause of death in two patients. The other causes of death in this group included multiorgan failure without apparent infection in two patients and sepsis-related complications (culture proven) in one patient.

During follow-up 7% (4/54) patients with ACLF and 29% (14/48) of patients with CHD received a liver transplant. All the patients who received a transplant survived. The transplant-free survival of both groups of patients is given in figure 2. Two patients in both groups were lost from follow-up. Patients with ACLF who survived the hospital stay had a less ominous prognosis. The difference in transplant-free survival attenuated after 1 year (survival at 3 months: 44.2% vs 82.2% (ACLF vs CHD)—p=0.0002 (95% CI 19.4% to 52.9%); at 6 months: 36.5% vs 60%—p=0.029 (95% CI 4.1% to 40.6%); at 12 months: 36.5% vs 53%—p=0.132 (95% CI −2.4% to 34.4%) and at 5 years: 29.7 vs 34.4%—p=0.0.767 (95% CI −13.0% to 22.3%).

Figure 2

Transplant free survival of patients with acute-on-chronic liver failure (ACLF) and chronic hepatic decompensation (CHD) during 5 years of follow-up. Patients with ACLF who survived hospital-stay had favourable long-term prognosis. The difference in transplant-free survival disappeared after 12 months. p Values are illustrated for each demarcated time point.

Parameters of infection at admission and outcome

Within 48 h of admission there was no difference between the patient groups in apparent infection (28% in ACLF vs 19% in CHD) (table 2). However, during hospitalisation, the presence of infection was significantly higher in patients with ACLF (58%) (n=31/53) than in those with CHD (26%) (n=12/47) (p=0.007).

Table 2

Comparison of infectious parameters of patients with ACLF and CHD

In contrast, within the first 48 h of admission, patients with ACLF showed a significantly higher white blood cell count and serum C-reactive protein level (11.5±8×109/l and 40±22 mg/l, respectively) than patients with CHD (7.5±3×109/l and 13.5±18 mg/l). SIRS within the first 48 h of admission was present in 69% of patients with ACLF compared with 23% of the CHD group (p=0.0001) (table 2). In the ACLF group with positive SIRS criteria within the first 48 h of admission, 60% (n=21) had no culture-proven infection. Twenty of 34 patients with positive SIRS criteria (59%) died compared with three out of 17 patients (18%) without SIRS (p=0.01).

Histological characteristics of the patient groups

The transjugular-guided liver biopsy procedure using three passes was successful in all patients and 98% (n=100) of samples were satisfactory for histological analysis. No major complications were seen. The median length was 1.3 cm (IQR=1) and the median number of portal tracts was 14 (IQR=10), which is considered valid.32

Table 3 shows detailed histological characteristics of the patient population. Of the regenerative parameters, hepatic progenitor cells were highly activated in both groups. In contrast, the number of proliferating (mib-1 positive nuclei) hepatocytes was equally low in the two groups.

Table 3

Comparison of histological characteristics of patients with ACLF and CHD

The infectious parameters such as ductular bilirubinostasis and cholangiolitis were significantly more frequent in ACLF (75% and 64% respectively) than in CHD (30% and 24%, respectively). Also, Mallory bodies were more common in patients with ACLF than in those with CHD (70% vs 40%). Although some features of alcoholic steatohepatitis (mainly ballooning and steatosis) were significantly more frequent in ACLF (94% and 70% respectively), the presence of other features such as lobular inflammation was similar in both groups (96% and 86%, respectively).

Predictors of in-hospital mortality in ACLF

Individual parameters

Clinical parameters predictive of in-hospital mortality in ACLF within 48 h of admission and histological features predictive for mortality at the moment of diagnosis are listed in table 4. Univariate analysis of a wide spectrum of clinical parameters showed age, INR and SIRS as predictors of in-hospital mortality. At the moment of diagnosis (around 1 week after admission) serum creatinine (p=0.008) and bilirubin (p=0.01) levels also predicted outcome. Among histological features marked ductular bilirubinostasis (p=0.003) (figure 3) and Mallory bodies (p=0.002) predicted in-hospital mortality.

Table 4

Clinical predictors of in-hospital mortality (within 48 h of admission) and histological predictors in patients with ACLF by univariate analysis

Figure 3

Ductular bilirubinostasis. (A) Immunohistochemistry for ubiquitin in a patient with acute-on-chronic alcoholic liver failure showing Mallory bodies in hepatocytes, original magnification ×400. (B) Haematoxylin and eosin (H&E) and (C) Hall's staining in a patient with acute on chronic alcoholic liver failure illustrate presence of bile plugs in dilated ductules at the interface between portal tract and parenchyma, original magnification ×400.

Multivariate analysis showed that the following parameters were related to mortality: age (p=0.0003), SIRS within the first 48 h of admission (p=0.05) and the histological presence of marked ductular bilirubinostasis (p=0.04) and Mallory bodies (p=0.01). In the CHD group, the small number of deaths did not allow us to establish significant differences for in-hospital mortality in comparison with the ACLF group.

Prognostic scores

Using the area under the receiver operating characteristic curve, we further evaluated the discrimination ability of different clinical prognostic models in predicting in-hospital mortality in patients with ACLF at admission (figure 4). The SOFA score had the best accuracy to discriminate between hospital survivors and non-survivors (95% CI 1.6 to 7.07), followed by the Maddrey score (95% CI −0.07 to −0.01). Using the Youden index, the best cut-off point for SOFA was 9 with a sensitivity and specificity of 84% and 68%, respectively. Although the MELD score had a very high specificity to discriminate between hospital survivors and non-survivors (93%), its sensitivity was rather low (36%).

Figure 4

Prediction of in-hospital mortality in 54 patients with acute-on-chronic liver failure (ACLF) at admission using area under receiver operating characteristic (ROC) curves. SOFA score had the best accuracy to discriminate between hospital survivors and non-survivors (area under the curve (AUC)=80%, p=0.0001), followed by Maddrey—(AUC=73%, p=0.008) and MELD score (AUC=67%, p=0.01). The Child–Pugh score did not predict survival (AUC=61%, p=0.1).

Correlation between ductular bilirubinostasis and infection during hospitalisation

Since ductular bilirubinostasis was highly predictive of in-hospital mortality, we further investigated the correlation between this early histological feature of sepsis in the total study population. The prevalence of ductular bilirubinostasis in biopsy specimens from patients who developed infection during hospitalisation was indeed significantly higher (72%) (n=31) than of that in those who did not develop infection (39%) (n=22) (p=0.01). To find out if ductular bilirubinostasis was an early sign of infection detection, we re-evaluated our cohort for this association. A positive correlation between infection during hospitalisation and ductular bilirubinostasis was seen. Of 18 patients who developed infection during hospitalisation 12 (67%) had ductular bilirubinostasis on their biopsies, while of 35 patients who did not develop infection only eight (23%) had ductular bilirubinostasis (p=0.009).


This prospective study assessed the validity of the clinical working definition of ACLF and investigated the importance of early clinical characteristics. ACLF appeared to be a separate clinical identity with high short-term mortality. However, in patients with ACLF who overcame the acute decompensation the difference in survival tended to attenuate over time to the same level of patients with CHD. Our investigations further showed that in patients with ACLF, features of infection such as SIRS and ductular bilirubinostasis at biopsy were early characteristics of ACLF.

The finding that features of severe infections are also features of ACLF is not surprising since a dysregulated inflammation is considered as one of the hallmarks of this syndrome.1–4 14 17 Immune dysfunction in a cirrhotic patient is complex, highly interactive and multifactorial problem.7 33 On the one hand, it starts with a decreased opsonisation capacity, which is crucial in bacterial phagocytosis, and bactericidal activity of these same phagocytic cells. On the other hand, the function of Kupffer cells, which represent major effectors of the reticuloendothelial system, is seriously impaired owing to portosystemic shunting, which leads to a lack of clearance of portal and systemic bacteria by the action of the reticuloendothelial system. The latter also explains why other bacterial products such as endotoxins and cytokines fail to be cleared.

Although the persistence of microbes, their toxins or injury are important triggers of an inflammatory reaction, the constellation of sepsis, or a syndrome resembling it, is primarily caused by the host's imbalanced reaction to these initiating injuring factors.8–10 12–14 More specifically, if microbes or any form of tissue damage (hypoxia, ischaemia, toxic agent, ….) are present, both a local proinflammatory and anti-inflammatory reaction are simultaneously initiated in an attempt to confine and control invading microbes locally, destroy damaged tissue and repair damage. If the initiating infectious or non-infectious factors overwhelm the local response, or the local response becomes exaggerated, proinflammatory mediators enter the systemic circulation, resulting in recruitment of a systemic response, which we recognise clinically as SIRS.8–10 The severity of the disease syndrome resulting from this cascade of systemic mediators depends on the balance of SIRS and the compensatory anti-inflammatory syndrome (CARS).8 10

In ACLF, SIRS and CARS have been shown to be disproportionally present. SIRS is amplified by increased levels of proinflammatory cytokines (interleukin 1, interleukin 6 and tumour necrosis factor α) that activate white blood cells.11 12 Defective CARS is shown by decreased HLA-DR expression on monocytes, leading to overwhelming infections.13 14 This latter is also called ‘immune paralysis’. When the equilibrium between proinflammatory and anti-inflammatory forces is seriously compromised, by overwhelming SIRS, defective CARS or an oscillation between severe inflammation and immunosuppression, a condition called ‘immunological dissonance’ arises as a result of these two forces ultimately reinforcing each other instead of maintaining a balance.8–10 This ultimately leads to a multiorgan dysfunction syndrome, septic shock or anergy.

In our cohort of patients with ACLF, we considered infection to be a major precipitating factor of SIRS. Although the high incidence of SIRS at admission (69%) initially did not similarly amount to the number of overtly documented bacterial infections (28%), later on during admission the number of patients with previously documented infections (58%) paralleled the number of SIRS-positive patients (67%). This finding favours the premise that a number of patients had an occult infection at admission and developed overt infection soon after admission.

Moreover, we also found a clear correlation with another early marker of sepsis and multiorgan failure: ductular bilirubinostasis and cholangiolitis. Ductular bilirubinostasis is the descriptive term for the presence of bile plugs in dilated ductules at the interface between the portal tract and parenchyma. Ductular epithelial cells are often damaged and/or vacuolated. It has been shown earlier that ductular bilirubinostasis represents an early—even subclinical—sign of septicaemia, which is a prognostically ominous sign (cholangitis lenta).34–37 Since SIRS and ductular bilirubinostasis were found to be independent predictors of outcome in ACLF, these factors might be considered auxiliary tools in the characterisation and prognostic assessment of these patients.

As mentioned earlier, this study showed that for patients with ACLF who recovered from the acute decompensation, the long-term, transplant-free outcome became similar to that of patients with CHD. We focused further on the histological features of regeneration. Our results showed that hepatic progenitor cells were equally activated in both groups, while the number of proliferating hepatocytes were comparably low, which might suggest that the liver reached its maximal level of regeneration. On the other hand, a high level of activation of progenitor cells and a low degree of hepatocyte replication, might also suggest that there is a replication inhibition, which was previously reported in subacute liver failure29 and chronic hepatitis.38 A potential pathophysiological explanation for this dual finding might involve increased cellular production of reactive oxygen species, including H2O2, due to for example chronic consumption of ethanol, leading to an increased death rate of hepatocytes while inhibiting the proliferative activity of mature hepatocytes.39–41 This may explain why in our series of patients the presence of marked Mallory bodies was considered to be a predictor of outcome since these latter result from the presence of oxidative stress.

Our study further suggests that liver biopsy has an important role because it recognises early features of sepsis and also confirms the presence of cirrhosis. This latter is of importance as 10% of the screened patients did not have cirrhosis on biopsy. Although their characteristics did not differ from those with cirrhosis, all of them survived, which reconfirms the fact that the ‘true cirrhotic state’ is crucial for a patient's prognosis.42

Furthermore, we challenged the ability of currently available prognostic scores to predict in-hospital mortality in patients with ACLF at admission to hospital. This question was asked to facilitate decisions about, for example, the early/appropriate use of liver-assist treatment. In our cohort the SOFA score, with a cut-off point of 9, had the best accuracy (sensitivity and specificity of 84% and 68%, respectively) to discriminate between hospital survivors and non-survivors (area under the curve=80%) within 48 h after admission. However, whether this cut-off point is useful in deciding whether to use additional invasive treatments has to be further investigated.43 44

Our study has some limitations. We opted to study a homogeneous group of patients with alcoholic cirrhosis. In Belgium, as in our neighbouring countries,17 this is the major group of patients with ACLF. Whether these findings also apply to patients with other causes of cirrhosis remains to be validated. Additionally, cautious interpretation of the comparison of long-term, transplant-free survival is in place since our cohort, although the largest one so far described, might be underpowered for the evaluation of differences in survival at 1 and 5 years owing to the high short-term in-hospital mortality. The last concern relates to potential overfitting in the multivariate analysis, which was largely attenuated by considering only the strongly significant results at multiple testing to avoid overinterpretation.

In conclusion, using the current clinical working definition of ACLF, we selected a group of patients with a high hospital mortality but with a comparable outcome to patients with CHD if they recovered from the acute failure component. We also confirmed in this group of patients the presence of a temporary immune dissonance with early features of SIRS and ductular bilirubinostasis, as early signs of infectious complications. These factors might be helpful for detecting ACLF at a stage when therapeutic interventions might still be effective.



  • Linked article 214627.

  • AK and WL contributed equally and share first authorship.

  • Funding WL and FN were supported by a grant from the Fund for Scientific Research - Flanders (Fundamenteel klinisch mandaat FWO Vlaanderen).

  • Competing interests None.

  • Ethics approval This study was conducted with the approval of the local ethics committee, University Hospital Gasthuisberg, K.U. Leuven, Leuven, Belgium.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Linked Articles