Outcomes following liver transplantation for seronegative acute liver failure: Experience during a 12-year period with more than 100 patients



Seronegative hepatitis is a common cause of acute liver failure (ALF) requiring liver transplantation. The primary aim of this study was to examine outcomes following transplantation in this group and to identify factors associated with early (<2 months) mortality. Patients studied were 110 consecutive cases of seronegative ALF transplanted at the Queen Elizabeth Hospital, Birmingham, between January 1992 and January 2004. Univariate analysis of 44 pretransplantation recipient, donor, and operative variables was performed initially to identify factors associated with early posttransplantation mortality. Variables identified as significant or approaching significance were analyzed using stepwise multiple logistic regression analysis. Survival following transplantation for seronegative hepatitis was 83%, 81%, and 73% at 2, 12, and 60 months, respectively. The majority (71%) of deaths occurred within the 1st 2 months and sepsis / multiorgan dysfunction was the most common cause of early death. Univariate analysis revealed 9 variables predicting early death. Subsequent multivariate analysis identified high donor body mass index (BMI; a possible surrogate marker for hepatic steatosis) as the most important predictor of early death (P = .009; odds ratio, 1.2; 95% confidence interval, 1.0-1.3). Recipient age >50 (P = .015; odds ratio, 4.2; 95% confidence interval, 1.3-14.1) and non-Caucasian recipient ethnicity (P = .015; odds ratio, 4.9; 95% confidence interval, 1.2-19.2) were other variables associated with early death on multivariate analysis. This study specifically examined factors that determine the early outcome of transplanted seronegative ALF patients. In conclusion, we found that donor and recipient factors identify patients who have a high chance of early death after transplantation. (Liver Transpl 2005;11:27–34.)

Seronegative (also known as non-A, non-B or non-A, non-B, non-C or non-A-E, or non-A-E) acute liver failure refers to an idiopathic hepatitis resulting in acute liver failure (ALF). Well recognized causes of ALF are excluded and serological markers for hepatitis A and B serology are negative. Seronegative ALF is an important cause of liver transplantation for ALF. In the United Kingdom, seronegative ALF is the most common indication for liver transplantation in ALF, due to a steady decline in the number of patients transplanted following paracetamol (acetaminophen) overdoses.1

Despite the frequency of liver transplantation for seronegative ALF, the etiology of this condition remains unknown. A viral etiology has been assumed, however the discoveries of the hepatitis C and E viruses and the more recent discoveries of the hepatitis G and transfusion transmitted viruses have failed to account for the majority of patients with “non-A non-B” ALF.2–5 Other viruses (e.g., cytomegalovirus, herpes simplex virus, Epstein-Barr virus, parvovirus B19) are rare causes of ALF and are unlikely to be implicated in the majority of seronegative ALF cases.

Although the etiology of seronegative ALF remains obscure, a number of epidemiological observations have been made about this condition, including a predominance among females and the relatively young, prompting some investigators to speculate about an autoimmune rather than viral basis for the disease.6, 7 A further consistent observation has been the very poor prognosis of seronegative ALF in the absence of liver transplantation, with spontaneous recovery occurring in less than 20%.8, 9

Despite the prevalence of seronegative ALF as cause for liver transplantation in ALF, no studies have specifically analyzed the outcomes of this group following transplantation. The few large studies examining prognosis in ALF post–liver transplantation,10–12 have evaluated outcomes for all the heterogeneous causes of ALF, rather than analyzing diseases individually. These studies have not provided consistent findings and their interpretation is further hampered by a lack of evaluation of donor factors and a failure to differentiate between early deaths (the major contributor to posttransplantation mortality) and late deaths. To improve outcomes for ALF patients and the allocation of scarce donor organs, a better understanding of factors associated with early mortality is required. The primary aim of this study was, therefore, to determine pretransplantation donor and recipient variables that were predictive of early death in a large homogenous cohort of patients transplanted for seronegative ALF.


ALF, acute liver failure; BMI, body mass index.

Patients and Methods

Patients studied were 110 consecutive adult cases of seronegative ALF transplanted between January 1992 and January 2004 at the Liver Unit of the Queen Elizabeth Hospital, Birmingham, UK. Acute liver failure (which included hyperacute, acute, and subacute liver variants) was defined using criteria previously described by O'Grady et al.13 and in all patients the time from onset of jaundice to encephalopathy was less than 12 weeks. In addition, no patients had a history of chronic liver disease prior to liver transplantation and subsequent histopathological examination of liver explants was consistent with acute or subacute ALF in all patients.

Patients were transplanted according to King's College criteria, which have previously been found to be applicable in ALF patients at our institution.14 No identifiable cause for ALF, despite thorough pretransplantation evaluation including drug and toxin exposure, was found in any of the patient cohorts. All patients were routinely tested for viral (hepatitis A virus immunoglobulin M [IgM] antibody, hepatitis B virus core IgM antibody [IgM anti-HBcore], anti-hepatitis C virus) autoimmune (anti-nuclear antibody, anti-smooth muscle, anti–liver/kidney microsome antibody, immunoglobulins) and metabolic (copper, caeruloplasmin) liver diseases. Acute hepatitis B virus was initially excluded on the basis of a negative IgM anti-HBcore, as hepatitis B virus DNA and hepatitis B surface antigen may be absent from the serum in this setting. Positive IgM anti-HBcore were followed up with full serological and hepatitis B virus DNA studies. Due to the very low prevalence of acute hepatitis E in the United Kingdom and our prior failure to incriminate this virus in the etiology of seronegative ALF,15 testing for acute hepatitis E was not routinely performed. Serological testing for acute hepatitis E was, therefore, confined to those patients with a history of recent travel to endemic areas or to contact with a traveler to an endemic area. ALF due to autoimmune hepatitis was excluded on the basis of the lack of a past history of this condition, the absence of strongly positive autoantibodies (≥1 : 100), or a marked immunoglobulin G elevation and the lack of characteristic histological findings on the explanted liver.

For some analyses (survival, prevalence of chronic hepatitis on 1-year biopsies, rates of cirrhosis, and retransplantation) the seronegative cohort was compared to a “nonseronegative” group. This consisted of a maximum of 83 patients transplanted during the same time period for all other causes of ALF.

The main source of data collection was from the Birmingham Liver Unit database. Other sources for data collection were from the electronic medical records system of the Queen Elizabeth Hospital and patient case notes.

The primary endpoint of this study was patient death occurring within 2 months posttransplantation. The predictive value of 43 pretransplantation variables were investigated including recipient clinical, recipient laboratory, donor, and operative variables. These variables and categories are shown in Table 1.

Table 1. Pretransplantation Predictive Variables
Recipient Clinical VariablesRecipient Laboratory VariablesDonor and Operative Variables
  1. Abbreviations: ICP, intracranial pressure; INR, International Normalized Ratio; AST, aspartate-aminotransferase; ALP, alkaline phosphatase; ABO, ABO blood group.

AgeProthrombin time on admissionDonor age (years)
Age category (<50, >50)INRDonor body mass index (BMI)
GenderBilirubin (μmol/L)Cold ischemic time (minutes)
Race (Caucasian, non-Caucasian)Bilirubin on admission (μmol/L)Warm ischemic time (minutes)
Ventilation (yes, no)Creatinine (μmol/L)Total ischemic time (minutes)
Inotropes (yes, no)Creatinine on admission (μmol/L)ABO nonidentical grafts (yes, no)
Renal support (yes, no)Urea (mmol/L)Split or reduced grafts (yes, no)
Hematology support (yes, no)Albumin (gm/L)Estimated operative blood loss
ICP monitoring (yes, no)AST (U/L)Use of veno-venous bypass (yes, no)
Raised intracranial pressure (yes, no)ALP (U/L) 
Days on intensive carepH 
Prior significant comorbidity (yes, no)Platelet count (×109/L) 
Sepsis (yes, no)Hemoglobin (g/dL) 
Ascites (yes, no)White cell count (×109/L) 
Hepatic encephalopathy grade (0–4)  
Hepatic encephalopathy category (mild, severe)  
Days from jaundice onset to encephalopathy  
Days from admission to transplantation  
Days from admission to listing  
Days from listing to transplantation  

Recipient age was investigated as both a continuous and categorical variable (<50 years and >50 years). For the race variable, the non-Caucasian category consisted of Afro-Caribbean or Asian (Indian, Pakistani, and Bangladeshi) patients. Pretransplantation renal support consisted of either conventional hemodialysis or continuous veno-venous hemofiltration. Patients were considered to have had hematology support if they received either blood, platelets, freshly frozen plasma, or albumin prior to transplantation. Any patient with a clinically significant past medical history, prior to the onset of their presentation with ALF, was included as having prior significant comorbidity. Significant comorbidities in this category included cardiac disease, renal impairment, diabetes, and prior hematological malignancy. Hepatic encephalopathy grade was categorized as mild (grades 0, 1, 2) or severe (grades 3 or 4).

Recipient laboratory data were taken from the immediate pretransplantation time period, following appropriate resuscitation. For prothrombin, time data was only available at the time of admission. Creatinine and bilirubin were analyzed at both at the time of admission and immediately pretransplantation. Total ischemic time was the sum of the cold and warm ischemic times. ABO blood group nonidentical grafts signified the use of ABO blood group nonidentical but compatible grafts.

Analysis of data was performed using SPSS (Chicago, IL) software. Survival was calculated using the Kaplan-Meier method and life-table analysis. Univariate analysis was performed initially to identify variables associated with early (<2 months) death post–liver transplantation. Significance testing between groups (dead at 2 months, alive at 2 months) for each variable was performed using the Mann-Whitney test for continuous data. For categorical data, significance testing was performed using the Pearson chi-square test or Fisher's exact test if any cells had a count less than 5. Exact (2-sided) P values were used and values <.05 were considered significant. Variables found to be significant (P < .05) or approaching significance (P < .15) on univariate analysis were entered into a multivariate stepwise logistic regression analysis. For this analysis, all data was normally distributed or transformed into a normal distribution. Data from 102 of the 110 seronegative patients were able to be used for multivariate analysis. Where relevant, data were expressed as a mean ± 1 standard error.


The number of transplants performed from 1992 to 2003 for all causes of ALF was 193 and the mean number of transplants per year for ALF during this period was 16. Seronegative ALF accounted for the 110 (57%) of the 193 patients transplanted for ALF. There were 47 (24%) patients transplanted for paracetamol overdose, 19 (10%) for idiosyncratic drug reactions, 8 (4%) for Wilson's disease, 5 (3%) for acute hepatitis B, 2 (1%) for acute hepatitis A, and 2 (1%) for veno-occlusive disease.

During the study period, a total of 140 patients were admitted to our institution with seronegative ALF. Of these patients 123 / 140 (88%) achieved listing criteria. The majority (110 / 123; 89%) of listed patients were subsequently transplanted. The remaining 13 (11%) patients died on the waiting list for transplantation. Although seronegative ALF represented the largest (57%) etiological group transplanted for ALF at our institution, from previous reviews of our experience, the seronegative ALF subgroup represented only a minority (16%) of those presenting with ALF, reflecting the low likelihood of spontaneous improvement and the poor prognosis of seronegative ALF without transplantation.

The mean age of patients transplanted for seronegative ALF was 40 ± 1.2 years (range, 16-66). Females accounted for 65 (59%) of the seronegative cohort. The predominant ethnicity of transplanted seronegative patients was Caucasian (86%), with 9 (8%) Afro-Caribbean and 7 (6%) Asian patients also transplanted. Most patients had severe encephalopathy prior to transplantation with stage III–IV encephalopathy recorded in 54 (49%) of patients. Stage I–II encephalopathy was present in 43 (39%) patients and no overt clinical encephalopathy was identified in the remaining 13 (12%) patients in the study population. The mean waiting time from listing to transplantation was 2.9 days.

The Kaplan-Meier survival curves for seronegative and nonseronegative patients posttransplantation are shown in Figure 1. The survival for seronegative patients was 83%, 81%, and 73%, at 2, 12, and 60 months, respectively. The survival for the nonseronegative group posttransplantation was less at all time points: 77%, 74%, and 68% at 2, 12, and 60 months, respectively. These difference in Kaplan-Meier curves were not significant on Breslow (P = .08) or log-rank (P = .10) analysis.

Figure 1.

Cumulative survival by months posttransplantation.

A characteristic of the seronegative survival curve was the rapid early death rate followed by a plateau at approximately 2 months posttransplantation. Of the 28 seronegative deaths occurring within 5 years posttransplantation, 20 (71%) occurred within the first 2 months, and the next death occurred at 7 months posttransplantation. Therefore, 2 months was chosen as the cutoff point for early death vs. survival in the subsequent univariate and multivariate analysis.

Causes for early (<2 months) and late (>2 months) deaths in the seronegative group posttransplantation were investigated. The mean follow up period for patients was 56 ± 4 months (range, 0-145), which included 20 early and 11 late deaths. Sepsis / multiorgan dysfunction was the major cause of death in the early death group. In the early death, group sepsis / multiorgan dysfunction accounted for 12 (60%) of deaths. A bacterial cause for sepsis was found in 7 of these patients, whereas fungal (2 patients) and cytomegalovirus (1 patient) sepsis were less commonly identified. In 2 patients who died early from multiorgan dysfunction, no cause of sepsis was identified. Deaths from lung injury, which included pneumonia and adult respiratory distress syndrome (3 patients), raised intracranial pressure (2 patients), graft ischemia (2 patients), and primary graft nonfunction (1 patient) were less common. A different pattern of mortality was seen among those 11 seronegative patients who died at later time intervals posttransplantation. Sepsis / multiorgan dysfunction was relatively less important, responsible for only 4 (36%) deaths. Of these deaths, a bacterial cause for sepsis was identified in 2 patients, with no cause identified in the other 2 patients. There were 4 “miscellaneous” deaths that occurred in this group, due to suicide, myocardial infarction, lymphoma, and biliary complications. Death from chronic rejection occurred in 2 patients and 1 patient died from primary graft nonfunction following a regraft performed 21 months following initial grafting.

Univariate analysis of the variables shown in Table 1 was performed to identify factors predicting early (<2 months) death posttransplantation. Variables found to be statistically significant (<.05) or approaching significance (<.15) on univariate analysis are demonstrated in Table 2 and variables that did not approach statistical significance (P ≥ .15) are shown in Table 3. Only 2 variables were significantly associated with early death posttransplantation for seronegative ALF on univariate analysis (Table 2). These were non-Caucasian ethnicity and the use of inotropes prior to transplantation. Non-Caucasian ethnicity was the most significant of these variables and 35% of patients dying before 2 months had non-Caucasian ethnicity compared with only 10% of those who survived this initial time period. Of the 7 non-Caucasian patients that died early, 4 were Afro-Caribbean and 3 were of Asian ethnicity. Inotrope use prior to transplantation was also significantly more common in those patients dying early, with almost double the use of inotropes in those dying compared with early survivors (45% vs. 23%). Another 7 pretransplant variables (hemoglobin, donor body mass index (BMI), time from admission to transplantation, recipient age >50 years, donor age, warm ischemic time, and time from listing to transplantation) approached statistical significance (Table 2) and were also entered into the multivariate analysis.

Table 2. Pretransplantation Variables Significant or Approaching Significance on Univariate Analysis*
VariableEarly (<2 Months) DeathEarly (>2 Months) SurvivalP Value
  • *

    For continuous data, values represent the mean ± 1 standard error. The maximum number of observations for the early death group is 20 and for the early survival group is 90.

Non-Caucasian race7 / 20 (35%)9 / 90 (10%).001
Use of inotropes9 / 20 (45%)21 / 90 (23%).049
Hemoglobin (g/dL)10.4 ± .411.4 ± .2.06
Donor body mass index26.7 ± 1.223.9 ± .4.07
Days from admission to transplantation6.9 ± 1.25.8 ± .7.07
Recipient age >50 years8 / 20 (40%)20 / 90 (22%).10
Donor age (years)43 ± 337 ± 1.11
Warm ischemic time (minutes)58 ± 1251 ± 1.11
Days from admission to listing3.1 ± .63.1 ± .3.11
Table 3. Pretransplant Variables not Approaching Significance on Univariate Analysis*
VariableEarly Death*Early SurvivalP Value
  • *

    For continuous data, values represent the mean ± 1 standard error. The maximum number of observations for the early death group is 20 and for the early survival group is 90.

  • Abbreviations: LT, liver transplantation; HE, hepatic encephalopathy; ICP, intracranial pressure; ALP, alkaline phosphatase; INR, International Normalized Ratio; ABO, ABO blood group; AST, aspartate-aminotransferase; Early death, death occurring less than 2 months posttransplantation.

Use of split or reduced grafts2 / 20 (10%)2 / 90 (2%).15
Presence of sepsis prior to LT2 / 20 (10%)3 / 90 (3%).17
Use of renal support prior to LT5 / 20 (25%)12 / 90 (13%).19
Days from listing to LT3.7 ± 0.92.7 ± 0.4.19
Days from jaundice onset to HE25 ± 517 ± 2.21
Recipient age (years)43.4 ± 1.339.4 ± 3.0.27
Presence of prior comorbidity4 / 20 (20%)10 / 90 (11%).28
Total ischemic time (minutes)748 ± 53714 ± 20.30
Cold ischemic time (minutes)689 ± 50663 ± 20.33
Presence of raised ICP1 / 19 (5%)1 / 86 (1%).33
Days on intensive care3.1 ± 1.01.9 ± 0.3.34
Creatinine (μmol/L)154 ± 24134 ± 12.34
White cell count (×109/L)10.8 ± 0.910.5 ± 0.6.34
Presence of high grade IIE8 / 20 (40%)46 / 90 (51%).37
Platelet count (×109/L)136 ± 22150 ± 10.38
Bilirubin (μmol/L)451 ± 39486 ± 22.42
Presence of ascites prior to LT9 / 20 (45%)32 / 90 (36%).43
Urea (mmol/L)7.5 ± 1.75.6 ± 0.7.51
ALP (U/L)382 ± 46375 ± 29.51
pH7.47 ± 0.027.49 ± 0.04.54
Creatinine on admission (μmol/L)179 ± 41129 ± 11.55
Operative blood loss (units)6.3 ± 1.05.3 ± 0.4.57
HE grade2.2 ± 0.32.3 ± 0.1.58
ICP monitoring2 / 19 (11%)6 / 86 (7%).63
Female gender11 / 20 (55%)54 / 90 (60%).68
INR3.4 ± 0.53.5 ± 0.2.71
Albumin (gm/L)28 ± 229 ± 1.75
Prothrombin time on admission35 ± 931 ± 3.80
Hematology support prior to LT6 / 20 (32%)26 / 87 (30%).88
Use of ABO nonidentical grafts5 / 20 (25%)21 / 89 (24%).89
Bilirubin on admission (μmol/L)458 ± 32460 ± 19.92
AST (U/L)743 ± 317552 ± 78.92
Ventilation prior to LT10 / 20 (50%)43 / 88 (49%).93
Use of veno-venous bypass16 / 20 (80%)70 / 90 (78%)1.0

Only 3 of the 9 variables (donor BMI, recipient age >50 years, and non-Caucasian ethnicity) that were entered into the stepwise multiple logistic regression analysis were significantly associated with early death posttransplantation; these are shown in Table 4. Donor BMI was the most predictive variable on multivariate analysis, with an odds ratio of 1.2 for every unit increase in donor BMI. Recipient age >50 years was the second most predictive variable, followed by non-Caucasian ethnicity. These variables were associated with odds ratios of between 4 and 5 for death within the first 2 months posttransplantation.

Table 4. Pretransplantation Variables Significant on Multivariate Analysis
VariableP ValueOdds Ratio95% Confidence Intervals
Donor body mass index.0091.21.1–1.4
Recipient age >50 years.0154.21.3–14.1
Non-Caucasian ethnicity.0154.91.3–19.2

To investigate possible recurrence of seronegative hepatitis posttransplantation, the prevalence of chronic hepatitis on 1-year posttransplantation protocol biopsies was examined. This was compared to a control group of nonseronegative ALF patients transplanted during the same time period (Table 5). Patients transplanted for fulminant hepatitis B were excluded from this nonseronegative control group because effective prophylaxis protocols against recurrent hepatitis B were not in place during the early study period and a control group in which disease recurrence would not be expected was desired. There was a 3-fold higher prevalence of chronic hepatitis observed in the 1-year protocol biopsies of the seronegative group that was significant (Table 5). To assess for possible effects of disease recurrence on graft function, the prevalence of cirrhosis and retransplantation were also compared between these groups (Table 5). Patients who died before 12 months or were less than 12 months posttransplantation were excluded from the cirrhosis analysis and patients who were regrafted due to vascular events or primary nonfunction were excluded from the retransplantation analysis. The rates of cirrhosis (8%) and retransplantation (10%) in the seronegative group did not differ significantly from those of the control group (Table 5). Overall in the seronegative group there were 14 regrafts in 12 patients. The indication for regrafting was chronic rejection in 7 (50%) of cases. Other indications for regrafting included hepatic artery thrombosis (2 patients), biliary complications (2 patients), acute rejection (1 patient), primary graft nonfunction (1 patient), and recurrent subacute hepatitis similar to the disease present in the native liver (1 patient).

Table 5. Prevalence of Chronic Hepatitis, Cirrhosis, and Retransplantation in Seronegative and Nonseronegative Patients Posttransplantation
VariableSeronegative PrevalenceNonseronegative PrevalenceP Value
Chronic hepatitis on 1-year biopsies18 / 34 (53%)2 / 13 (15%).045
Cirrhosis7 / 86 (8%)3 / 44 (7%)1.0
Retransplantation11 / 109 (10%)5 / 80 (6.3%).35


Of the 43 variables examined in this study, donor BMI emerged as the most predictive of early mortality on multivariate analysis. To illustrate the importance of this variable, suggested by this study, the odds ratio of early death following transplantation with a liver from an obese donor (BMI = 35) relative to a normal donor (BMI = 25) is 1.2 to the power of 10 (1.2)10, which equals 6.2. Although donor BMI was likely to be a surrogate marker for hepatic steatosis in this study, histological classification of transplanted livers may have been a more robust variable to measure. Unfortunately, these data were unavailable; a prospective study evaluating histological grade of steatosis with outcomes following liver transplantation for ALF would be an important future study to confirm our finding.

The importance of graft steatosis as a predictor of poor outcomes post–liver transplantation has been previously described and may relate to the poor regenerative capacity and increased susceptibility to ischemic injury of steatotic hepatocytes.16–18 However, only 1 previous study has investigated the importance of graft steatosis in the specific setting of ALF.12 The association of steatotic grafts with poor patient survival on multivariate analysis from Bismuth's study12 is consistent with findings from our own study.

These data, therefore, support the concept that donor graft quality, rather than recipient factors, is the most important determinant of short-term survival following transplantation for ALF and that the use of marginal donors should be avoided in this setting. With respect to the use of split or reduced grafts in the setting of ALF, we were unable to draw firm conclusions from this study, as only 4 of 110 grafts used were nonwhole (2 split, 2 reduced). Despite this, a higher, nonsignificant prevalence of nonwhole grafts in the early death group was observed (10% vs. 2%; see Table 3), which also appears to support the concept of avoiding nonwhole / marginal grafts in the setting of ALF.

Recipient age >50 years was the second most predictive factor on stepwise logistic regression analysis. Age over 50 years is intuitively likely to be important in this relatively young cohort and may reflect impaired recovery capacity following the severe physiological insult associated with ALF and transplantation. The importance of recipient age is supported by 2 other studies that have found this variable to be associated with death following transplantation for ALF on univariate analysis.10, 19

The negative impact of non-Caucasian ethnicity on survival was the other significant variable on multivariate analysis. The ethnic origin of liver graft recipients has been previously recognized as a determinant of patient survival, with black and Asian populations having poorer long-term survival relative to Caucasians.20, 21 The reasons for these discrepancies are unknown. Relevant issues may be poorer histocompatibility antigens (HLA) matching of ethnic minorities and poorer compliance with immunosuppressive medication leading to increased rejection episodes.20 Lower socioeconomic status may also be a relevant contributor to poorer outcomes in these ethnic groups.20 Only 1 study has specifically examined the effect of race following transplantation in ALF, and Afro-American race predicted worse patient and graft survival on univariate analysis.10 In our study of seronegative ALF, the impact of non-Caucasian ethnicity was apparent at an early time point (2 months) posttransplantation, not examined in the above study, and was significant on both univariate and multivariate analysis. All of the 7 non-Caucasian patients who died within the 1st 2 months received grafts from Caucasian donors. However, it seems unlikely that “immunological” factors contributed to mortality in these patients, as there was no suspicion of acute or ductopenic rejection (histologically or clinically) in any of these patients. In addition, primary graft nonfunction, which may occur with an increased frequency in African American populations,20 was not seen in those non-Caucasian patients who died early. More aggressive disease or delayed presentation in non-Caucasians may be more important factors, and comparison of explant pathology together with other recipient clinical and laboratory variables between Caucasian and non-Caucasian groups would be an interesting further study.

Surprisingly, markers of pretransplantation renal function, which have frequently been identified as an important predictive variable in other studies of posttransplantation ALF outcomes,10, 11, 19 were not important on either univariate or multivariate analysis. Although the use of pretransplantation renal support was more frequent and the mean values for creatinine (on admission and prior to transplantation) and urea were higher in the poor prognosis group (Table 3), none of these variables approached significance on univariate analysis. Differences in study populations and design may account for these discrepancies. Our study examined a homogeneous population of seronegative ALF and it is possible that renal dysfunction is a more important predictive variable in nonseronegative causes of ALF. Another difference from other studies was the use of short-term mortality as an end point rather than all mortality, and it is possible that renal dysfunction pretransplantation is a better predictor of long-term rather than short-term mortality. Overall, this study suggested that the quality of the donor liver, rather than recipient comorbidities such as renal dysfunction, was the most important determinant of short-term mortality.

The demographic data of our seronegative population was consistent with that described in previous smaller studies, with a female predominance (59%) and disease onset at a relatively young age (mean of 40 years). Aplastic anemia was seen in only 1 patient. This was a 19-year-old male and occurred transiently following transplantation. The association of aplastic anemia with seronegative ALF has been described as occurring in up to 28% of seronegative ALF patients following transplantation and was attributed to acute parvovirus B19 infection.22 This larger study, with a less than 1% prevalence of aplastic anemia, does not support an important etiological role for parvovirus B19 in seronegative ALF.

Another interesting observation was the appearance, during posttransplantation follow-up, of strongly positive (≥1 : 100) antinuclear and / or anti–smooth muscle antibody titers in 11 (10%) patients. Immunological markers were performed in only 47 (43%) of the cohort and the real prevalence of such abnormalities may be higher. The significance of this finding is uncertain but it suggests that a subset of seronegative patients may have an underlying etiology similar to that of autoimmune hepatitis. The similar demographic features of autoimmune and seronegative hepatitis also supports this possibility. Poor sensitivity of autoantibodies in the setting of ALF may be another factor masking the diagnosis of autoimmune hepatitis in this setting.

Disease recurrence was strongly suggested by 1 case. This involved a 32-year-old woman with recurrent subacute hepatitis and postnecrotic collapse of her graft, occurring 9 months following initial transplantation. Retransplantation was required and the histology of the explant was similar to that of the native liver, which demonstrated massive hepatic necrosis. Autoantibodies and immunoglobulins were not available at the time of recurrence in this patient and a recurrent or de novo autoimmune hepatitis could not be definitively excluded. However, the subacute presentation and explant pathology did not support an autoimmune etiology.

In the remaining patients, evidence for disease recurrence was less conclusive. It seems plausible that disease recurrence may be altered by immunosuppressive therapy posttransplantation and we therefore examined for the prevalence of chronic hepatitis, which may be a manifestation of recurrence in this setting. The prevalence of chronic hepatitis on protocol 1 year posttransplant biopsies was significantly higher than that of a control group of predominantly drug-induced liver failure. This finding may provide indirect evidence of disease recurrence or of a more exaggerated inflammatory response in seronegative patients and confirms similar findings of a previous study by Mohamed et al.23 Although the significance of posttransplantation chronic hepatitis is uncertain, it did not appear to impair graft function or survival with similar rates of cirrhosis and retransplantation in the seronegative and control groups.

In summary, we describe a large single-center experience with transplantation for seronegative ALF. Good long-term outcomes (73% 5-year survival) were observed and factors that were most predictive of early mortality included donor BMI, recipient age >50 years, and non-Caucasian ethnicity.


We thank the Gastroenterology Society of Australia for financial support provided to A.W. and we thank Peter Nightingale for his expert statistical assistance.