Changing etiologies and outcomes of acute liver failure: Perspectives from 6 transplant centers in Argentina
Hepatology and Liver Transplant Unit, Hospital Universitario Austral, Pilar, Argentina
Address reprint requests to Manuel Mendizabal, M.D., Hepatology and Liver Transplant Unit, Hospital Universitario Austral, Avenida Juan D. Perón 1500, Pilar, Argentina B1629AHJ. Telephone: +54 2304482884; FAX:+54 2304482236; E-mail: firstname.lastname@example.org
Acute liver failure (ALF) is a complex medical condition characterized by sudden and severe liver cell dysfunction leading to coagulopathy and hepatic encephalopathy (HE) in people not known to have an underlying liver disease. The advent of orthotopic liver transplantation (OLT) and improvements in overall medical management have dramatically improved the prognosis for this serious disorder, with survival increasing from 30% in the 1970s to 76% in the present.[2, 3]
Numerous epidemiological studies have described ALF as a heterogeneous syndrome with different patterns, prognoses, and outcomes.[2, 3] Additionally, great geographical differences have been recognized, and this makes the translation of data from one region to another difficult. Knowledge of current etiologies and outcomes of ALF may help to prevent the onset of the disease and lead to more selective treatment of ALF and thus lower the need for OLT. Information about the causes and outcomes of patients with ALF in Argentina is limited to a retrospective, single-center experience that included 64 adult patients. The most prevalent causes of ALF were indeterminate causes, viral hepatitis, autoimmune hepatitis (AIH), and drug toxicity. Although this study provided new data on management, outcomes, and the performance of prognostic scores in Argentina, the insight gained was limited because of the single-center experience. Furthermore, hepatitis B virus (HBV) and hepatitis A virus (HAV) vaccines were included in the national vaccination program in 2002 and 2005, respectively, and their impact on ALF epidemiology in the adult population has not yet been evaluated.
The aims of the present study were to investigate the etiologies and short-term outcomes of patients with ALF, to evaluate the performance of different prognostic criteria, and to identify clinical prognostic factors of death in what is, to our knowledge, the most extensive series of adult patients with ALF ever studied in South America.
PATIENTS AND METHODS
The liver transplantation (LT) programs involved in the survey belong to Hospital Universitario Austral, Hospital Italiano de Buenos Aires, Fundación Favaloro, Hospital Privado de Córdoba, Hospital Alemán and Sanatorio Allende. The survey was previously discussed by all the participating centers. The local institutional review board of each center approved the study. All the centers involved had an LT program with more than 10 years of experience. The study included adult patients (16 years or older) hospitalized with ALF at any of the participating centers between June 1, 2005 and December 31, 2011. Patients who received extracorporeal liver support were excluded from the analysis. A physician from each liver unit collected the data from all the cases handled in the unit during the study period, and each patient was identified with a code. All enrolled patients met the entry criteria for ALF: the presence of coagulopathy [international normalized ratio (INR) ≥ 1.5 or prothrombin index < 50%] and any grade of HE within 26 weeks of the first symptoms without a known underlying liver disease. Patients with Wilson's disease, vertically acquired HBV, or AIH with cirrhosis were included if their disease had been recognized for <26 weeks. Each center provided detailed demographic, laboratory, clinical, and outcome information for all enrolled patients. ALF was classified as hyperacute, acute, or subacute impairment according to O'Grady's classification. An etiological diagnosis was made at each study center on the basis of the clinical history, imaging studies, laboratory values, and, in some cases, histopathological characteristics. To confirm the diagnosis of AIH, we retrospectively applied the simplified AIH criteria: those patients who presented with a score ≥ 6 were identified as having AIH. However, because autoantibodies are common in ALF, we considered them positive only with significant titers (≥1/80). ALF was considered to be indeterminate when extensive clinical and laboratory evaluations (including toxicology screens, autoantibodies, and serological markers for viral hepatitis) as well as imaging studies were inconclusive.
Even though patient management and candidacy for OLT were determined at each site, the criteria were uniform and followed published clinical guidelines. HE was graded from 1 to 4 according to the West Haven criteria. Intracranial hypertension (ICH) was diagnosed on the basis of either combined clinical and computed tomography data or intracranial pressure monitoring. The King's College criteria (KCC), the Model for End-Stage Liver Disease (MELD) score, and the Clichy criteria were determined as described.[10-12]
All data records were checked for missing values and inconsistencies, queries were referred to the participating institution, and corrections were made at the data coordinating center.
Data are presented as percentages or as means and SD with 95% confidence intervals (CIs). The normality distribution of different variables was tested with the Shapiro-Wilk test. A comparison of baseline parameters between patient groups was performed with the Student t test, the Mann-Whitney or median nonparametric test, and the χ test for categorical parameters. To identify independent predictors of death for patients who did not undergo OLT (versus spontaneous survivors), a logistic regression analysis was performed for the variables that showed a level of significance of P < 0.1 in the univariate analysis. Receiving operating characteristic (ROC) curve analysis was used for a comparison of the prognostic accuracy of different outcome prediction models. Tests were 2-sided, and significance was accepted at P < 0.05. The statistical analysis was performed with R 2.12.0.
Demographic Characteristics and Clinical Data
Table 1 provides an overview of relevant baseline parameters. One hundred fifty-four patients with ALF were included in the analysis, the mean age at diagnosis was 45.2 ± 15.2 years, and 95 of the patients (61.7%) were female. The mean baseline parameters were as follows: INR, 3.8 ± 2.5; factor V, 33% ± 21%; creatinine, 1.32 ± 1.35 mg/dL; and total bilirubin, 19.4 ± 9.5 mg/dL. The MELD scores and the Clichy criteria on admission were calculated for 147 patients, whereas the KCC were calculated for 144 patients.
Table 1. Baseline Parameters for Patients With ALF
Overall (n = 154)
The data are presented as means and standard deviations.
According to O'Grady's ALF classification, 40% were hyperacute, 42% were acute, and 18% were subacute. The time from jaundice to encephalopathy could not be accurately established for 19 cases (12.3%). At presentation, grade II HE was the most frequent finding (32%), whereas 28% of the patients presented with a deep coma (grade IV HE). The mean interval from jaundice to encephalopathy was 18.7 days (range = 0-140 days). It is noteworthy that 18 patients (12%) had coagulopathy and no HE when hospitalized but developed it during admission.
Extrahepatic organ dysfunction was frequently reported during hospitalization. Ventilatory support was required by 84 ALF patients (54.5%), vasopressors were required by 63 (41%), and renal replacement therapy was required by 39 (25.3%) during the intensive care unit stay. Cerebral edema was found in 33 patients (21%), and intracranial pressure monitoring was used for 54 patients (35.1%), with only 4 patients presenting with bleeding or an infection related to the procedure.
Distribution of Etiologies and Outcomes of ALF
The etiology of ALF in the patient population is presented in Fig. 1. The majority of the cases were due to viral hepatitis involving 49 patients (32%). Among these patients, 46 (30% of the studied population) had an acute de novo HBV infection, only 2 had acute HAV, and 1 had herpes simplex virus. No cases of ALF due to hepatitis C, D, or E were identified; however, hepatitis D and hepatitis E were not evaluated for all cases. AIH and indeterminate causes were the second most common causes of ALF, with each accounting for 40 cases (26%). The simplified AIH criteria were retrospectively applied to all AIH cases presenting with a score ≥ 6. Titers for anti-nuclear antibodies and anti-smooth muscle antibodies were ≥1/80 in 72% and 51% of the AIH patients, respectively; anti–liver-kidney microsome antibodies were present in only 1 patient. The median level of γ-globulins was 2.4 ± 0.9 g/dL. A histological analysis was performed for 30 patients: 11 had no fibrosis, 8 patients were found to be cirrhotic, and the 11 remaining patients exhibited some degree of fibrosis. Indeterminate liver failure was responsible for 40 cases (26%), and drug or toxic reactions were described for 16 patients (10.4%); these cases included Amanita phalloides poisoning in 4 cases and nonsteroidal anti-inflammatory drugs in 3 cases, and the remaining 9 cases were due to different drugs or toxic substances. Notably, no acetaminophen (ACM) intoxication was reported.
The risk of death or OLT appeared to be particularly high for patients with ALF due to an indeterminate cause (92%), AIH (75%), or drug toxicity (75%).
Outcomes of ALF
The overall outcomes of the 154 patients with ALF are shown in Fig. 2. Altogether, 121 patients (79%) were listed for OLT, and 83 patients (54%) received a liver graft; 9 of the 83 OLT recipients (11%) died after surgery, and 33 patients (21%) died without OLT. Only 38 patients (25%) recovered spontaneously without OLT. Only 38 patients (25%) recovered spontaneously without OLT. The mean time from transplantation listing to actual transplantation was 3.8 days (range = 0-8 days). The mean duration between admission and death was 5.2 ±4.2 days. Death was attributed to infectious complications (31%), ICH (31%), and multiorgan failure (38%). Overall, 111 ALF patients (72%) were discharged from the hospital alive.
Prediction of OLT or Death for Patients With ALF
In the univariate analysis, the predictors of death for patients who did not undergo OLT (versus spontaneous survivors) were an encephalopathy grade ≥ 3 on admission, INR, creatinine, bilirubin, aspartate aminotransferase, a need for vasopressors or ventilatory support on admission, renal replacement therapy, development of ICH, MELD score, a MELD score ≥ 29, Clichy criteria, and KCC (Table 2). A MELD score ≥ 29 was found to be the best discriminant between survivors and nonsurvivors through the construction of ROC curves. In the multivariate logistic regression, a backward elimination selection procedure selected a MELD score ≥ 29 [odds ratio (OR) = 14.2, 95% CI = 3.37-60.1, P < 0.001] and an encephalopathy grade ≥ 3 on admission (OR = 4.49, 95% CI = 1.27-15.8, P = 0.002) as predictors of adverse outcomes on admission.
Table 2. Univariate Analysis of Patients With ALF
Spontaneous Survivors (n = 38)
Death Without OLT (n = 33)
The data are presented as means and standard deviations. MELD score ≥29 and MELD score were available for 35 and 32 patients in the spontaneous survivors group and in the death without OLT group, respectively Clichy criteria were available for 37 and 31 patients in the spontaneous survivors group and in the death without OLT group, respectively KCC were available for 36 and 33 patients in the spontaneous survivors group and in the death without OLT group, respectively. Percentages where calculated over available patients.
In a direct comparison using an ROC curve analysis, the MELD score [area under the curve (AUC) 0.830, 95% CI = 0.73-0.93] had better prognostic accuracy for predicting outcomes (OLT-free survival versus death without OLT) than the Clichy criteria (AUC = 0.719, 95% CI = 0.58-0.85) or the KCC (AUC = 0.631, 95% CI = 0.49-0.77; Fig. 3). The AUCs estimated for the Clichy criteria, KCC, and MELD score were compared in pairs and were significantly different for all cases (P < 0.001).
Different terms have been proposed in the literature for ALF; however, no consensus exists on the severity of either encephalopathy or coagulopathy that marks the transition from liver injury to frank liver failure. In a recent systematic review of 41 definitions of ALF, 4 components were identified that underlined the definition of ALF: the presence of HE, the interval between the onset of disease and the occurrence of HE, the presence of coagulopathy, and preexisting liver disease. We considered the definition of ALF proposed by Polson and Lee and taking into account these 4 variables to be the most representative.
Large epidemiological national surveys of ALF involving more than 1 center are scarce in South America. Information about the etiologies, management, and outcomes of ALF is usually taken from other regional studies (especially US and European populations).[2, 4, 14] The current study included patients from 6 transplant centers: 4 located in the metropolitan area of the city of Buenos Aires and 2 located in the province of Córdoba. We cannot say that the sample represents the whole ALF population, but it reflects the overall spectrum of patients with ALF in the country. The most common etiologies of ALF in our series were HBV infection, AIH, and indeterminate causes. Remarkably, no ACM toxicity was reported. These findings clearly contrast with reports from the United States and Europe, where ACM is one of the most common etiologies of ALF.[2-5, 15] Our findings are even more striking if we take into account the fact that ACM is an over-the-counter drug in Argentina. Possible explanations for our null incidence of ACM toxicity could be that it is rarely used as a method of suicide; it is infrequently combined with other pain killers; and finally, gene polymorphisms (not studied in our population) might have an important role in determining patients at risk of developing ALF due to ACM.
Another surprising finding is that acute AIH was the second most common etiology diagnosed (26%) in our series. The described frequency of ALF ascribable to AIH varies from 2% to 19%.[5, 17] The diagnosis of AIH is challenging, especially when the patient presents with ALF. Recently, clinical and histological criteria have been proposed for the fulminant presentation of AIH. Eight of the patients with acute AIH in our series presented with cirrhosis. The question of whether an acute presentation of AIH reflects acute disease or, rather, a flare of a chronic illness is still a matter of debate. The prompt diagnosis of AIH is relevant because the initiation of steroid therapy at an early stage can be beneficial. In our study, the diagnosis of fulminant AIH was established with the simplified AIH criteria, which are a combination of laboratory and histological findings.
Acute viral hepatitis is the most common cause of ALF in developing countries. Almost one-third of our cases of ALF were due to HBV infections, and this rate is considerably higher than the rate of 11% reported for an Argentinean single-center experience. Interestingly, all the patients presented with a de novo HBV infection. For those patients who underwent transplantation, the explant showed no histopathological evidence of chronic disease, and for those who did not undergo OLT, the hepatitis B surface antigen was cleared. On the other hand, the rate of fulminant HAV diagnoses decreased from 12% to 1% in comparison with the same study. Universal and mandatory single-dose HAV vaccination of 1-year-old toddlers was introduced in 2005, and no new pediatric ALF cases due to HAV have been reported since 2007. This measure has indirectly affected the adult population as well. Universal HBV vaccination of all newborns was introduced in 2002, so the incidence of acute HBV infections in adults would be expected to decrease in coming years.
A large percentage of our patients (54%) underwent OLT. We think that a major cause of our high rate of transplantation was the absence of ACM-induced ALF. These results agree with those reported recently by King's College Hospital for a non-ACM group. When we compared our series with the US ALF study group, we found similar listing times (3.8 versus 3.5 days) but a considerably lower death rate on the waiting list (19% versus 30%). One reason for this difference is that ACM is responsible for 45% of US cases, and these patients are often refused transplantation because of psychosocial disorders or substance abuse. Another explanation is that ALF secondary to ACM progresses very rapidly, whereas in idiosyncratic or AIH cases, the disease evolution is usually more prolonged, and the likelihood of finding a suitable organ is greater. Early referral to a transplant center when acute liver injury and coagulopathy present remains a key factor for improving survival. Almost 40% of the patients in our study presented with grade III/IV HE on admission to the transplant center, and this indicates late referral to transplant centers in some cases. It should also be noted that Argentina is a vast country with a single transplant list, and most transplant centers are located in the center of the country; this makes it sometimes difficult to transfer these critical patients at earlier stages. On the other hand, 12% of the patients showed no HE on admission but developed it when they were hospitalized.
Currently, the KCC and, to a lesser extent, the Clichy criteria are most widely used as selection criteria for OLT. However, the major limitations of these criteria are their low negative predictive values.[22-25] Recently, interest has been focused on the application of the MELD score to predicting ALF outcomes in both ACM and non-ACM groups, but any consistent advantage has failed to be conclusively demonstrated.[5, 24, 26, 27] In our population, the concordance statistic for the MELD score (with the exclusion of transplant patients) was significantly higher than the concordance statistics for the KCC and Clichy criteria. Noticeably, the Clichy criteria were superior to the KCC. We acknowledge that the MELD score discards 3 significant concepts of the natural history of ALF: age, underlying etiology, and the interval between the development of jaundice and the onset of encephalopathy. Rather than proposing the MELD score to be the best prognostic model for ALF, we think that it can be used as an additional tool to decide which patients will require OLT. In our series, we also identified 2 independent predictors of outcomes on admission: a MELD score ≥ 29 and an HE grade ≥ 3. Recently, some studies have included apoptosis and necrosis markers as predictors of outcomes of ALF with promising results.[25, 26] A modification of the classic MELD score with the M-65 antigen more sensitively and specifically predicted outcomes than the MELD score or the KCC. The US ALF study group proposed an index that combines different clinical markers on admission (coma grade, INR, bilirubin, and phosphorus) with an apoptosis marker, M-30, and this predicted outcomes of ALF more accurately than the KCC and the MELD score. Markers of cell death or apoptosis are encouraging, but their use is technically limited and should be confirmed in further studies.
In summary, we have presented data on the epidemiology, prognosis, and outcomes of patients with ALF at 6 experienced LT centers in Argentina. HBV and AIH were the most common causes of ALF in our series. The MELD score predicted outcomes of ALF more accurately than the KCC and the Clichy criteria in a population with no reported ACM toxicity. Nevertheless, we recommend the careful interpretation of these results because not all the country's LT programs were included in the survey. We also acknowledge that given the retrospective design of our study, we cannot account for inherent and undocumented patient characteristics. Additionally, the number of patients who survived or died without OLT was rather small. Future prospective studies should be conducted in Argentina to confirm our results and to further improve the outcomes of ALF patients.