Antituberculosis therapy–induced acute liver failure: Magnitude, profile, prognosis, and predictors of outcome


  • Potential conflict of interest: Nothing to report.


Antituberculosis therapy (ATT)–associated acute liver failure (ATT-ALF) is the commonest drug-induced ALF in South Asia. Prospective studies on ATT-ALF are lacking. The current study prospectively evaluated the magnitude, clinical course, outcome, and prognostic factors in ATT-ALF. From January 1986 to January 2009, 1223 consecutive ALF patients were evaluated: ATT alone was the cause in 70 (5.7%) patients. Another 15 (1.2%) had ATT and simultaneous hepatitis virus infection. In 44 (62.8%) patients, ATT was prescribed empirically without definitive evidence of tuberculosis. ATT-ALF patients were younger (32.87 [±15.8] years), and 49 (70%) of them were women. Most had hyperacute presentation; the median icterus encephalopathy interval was 4.5 (0-30) days. The median duration of ATT before ALF was 30 (7-350) days. At presentation, advanced encephalopathy and cerebral edema were present in 51 (76%) and 29 (41.4%) patients, respectively. Gastrointestinal bleed, seizures, infection, and acute renal failure were documented in seven (10%), five (7.1%), 26 (37.1%), and seven (10%) patients, respectively. Compared with hepatitis E virus (HEV) and non-A non-E–induced ALF, ATT-ALF patients had nearly similar presentations except for older age and less elevation of liver enzymes. The mortality rate among patients with ATT-ALF was high (67.1%, n = 47), and only 23 (32.9%) patients recovered with medical treatment. In multivariate analysis, three factors independently predicted mortality: serum bilirubin (≥10.8 mg/dL), prothrombin time (PT) prolongation (≥26 seconds), and grade III/IV encephalopathy at presentation. Conclusion: ATT-ALF constituted 5.7% of ALF at our center and had a high mortality rate. Because the mortality rate is so high, determining which factors are predictors is less important. A high proportion of patients had consumed ATT empirically, which could have been prevented. Hepatology 2010

Isoniazid, rifampicin, pyrazinamide, and ethambutol are usual first-line combination chemotherapy in active tubercular diseases.1, 2 Hepatotoxicities of the former three drugs are well documented.3-5 A meta-analysis of antitubercular therapy (ATT)–associated hepatotoxicity reported that the frequency of overt clinical hepatitis caused by isoniazid, rifampicin, or both together was 0.6%, 1.1%, and 2.6%, respectively.6 The spectrum of ATT hepatotoxicity is diverse, ranging from asymptomatic rise in aminotransferases (to fivefold) in 2.3% to 28% to acute liver failure (ALF) in approximately fewer than 0.01% of the individuals.7-10 However, the frequency of ATT-induced hepatotoxicity among Indian seems to be higher (11.5%) than that for the white population (4.3%).6

Approximately 20% of the global populations with tuberculosis (3.32 million) live in India.11 Therefore, number of patients with ATT hepatotoxicity in India may be substantial. A prospective study reported that seven (9.72%) of 72 patients diagnosed as ATT hepatotoxicity developed ALF (ATT-ALF) and had a mortality rate of 85.7%.12 Therefore, ATT-ALF may have a more aggressive natural course than in hepatitis virus(es)–associated ALF, the frequent cause of ALF in South Asia.13 ATT is the commonest drug-induced ALF in south Asia, where tuberculosis is endemic.13, 14 Whereas clinical profile, natural course, prognostic model, and therapeutic strategies have been well described for paracetamol-induced ALF,15-17 information on ATT-ALF in the English literature is scant and consists of reports of only few patients.18-21 Therefore, this prospective study was conducted to evaluate magnitude, course, complications, and prognostic model in patients with ATT-ALF in an endemic region of tuberculosis.


ALF, acute liver failure; ATT, antituberculosis therapy; HAV, hepatitis A virus; HBsAg, hepatitis B surface antigen; HCV, hepatitis C virus; HEV, hepatitis E virus; IEI, icterus-encephalopathy interval; IgM, immunoglobulin M; KCH, Kings College Hospital; MELD, model for end-stage liver disease; PT, prothrombin time.

Patients and Methods

Consecutive patients diagnosed as ATT-ALF between January 1986 and January 2009 were included in the study. During this period, 1223 patients (≥13 years) with ALF were admitted to the Department of Gastroenterology, All India Institute of Medical Science, New Delhi. ATT was implicated as a cause of ALF in 70 (5.7%) patients. Consent to include the data and analysis for the purpose of the study was obtained from the nearest relative of the each patient at the time of enrollment. The study was approved by the ethics committee of our Institute, and it is in agreement with the Helsinki Declaration.

Study Variables


ALF was defined by occurrence of encephalopathy within 4 weeks of symptoms in absence of preexisting liver disease.22


ATT-ALF was diagnosed if the patient with ALF had a history of consumption of at least two of the three first-line hepatotoxic drugs (isoniazid, rifampicin, and pyrazinamide) for a minimum period of 1 week and if the patient's sera tested negative for evidence of known hepatitis virus(es) infection (hepatitis A virus [HAV], hepatitis B virus, hepatitis C virus [HCV], and hepatitis E virus [HEV]), and absence of any other identifiable cause of acute liver injury.

Grading of Encephalopathy23

  • Grade 1: Loss of sleep rhythm, drowsiness, confusion, and flapping tremors

  • Grade 2: Features of grade 1 encephalopathy with loss of sphincter control as well

  • Grade 3: Unconsciousness with no response to oral commands, but responding to painful stimuli

  • Grade 4: Deep unconscious state, with no response to pain

The icterus-encephalopathy interval (IEI)13 was defined as the intervals from the onset of jaundice to the onset of hepatic encephalopathy.

The ATT-ALF interval was defined as interval from the day of initiation of antitubercular drugs to the onset of hepatic encephalopathy.

Cerebral edema was defined clinically by the presence of spontaneous or inducible decerebrate posturing or by the presence of any of two: hypertension (blood pressure ≥150/90 mmHg), bradycardia (heart rate < 60/minute), pupillary changes or neurogenic hyperventilation.13, 24 Neurogenic hyperventilation was diagnosed by the presence of hyperventilation in the absence of any metabolic or respiratory causes of hyperventilation.13, 24

Infection was diagnosed by the presence of pyrexia (temperature > 101°F) or hypothermia (temperature < 98°F) and neutrophil leukocytosis (total leukocyte count >15,000/mm3, with ≥80% polymorphs), and one or more of: positive blood culture, positive urine culture, or radiological evidence of pneumonitis.13, 24, 25

Renal failure was diagnosed if patients developed decreased urine output (<400 mL in 24 hours), with serum creatinine greater than 1.5 mg/dL and blood urea greater than 40 mg/dL, despite hydration, objectively assessed by central venous pressure of 10 cm saline or more.13

Management Protocol

All patients were admitted to the intensive care unit. History was obtained from the patient's attendants, and a detailed clinical examination was performed at admission. History regarding the possible implication of drug(s) was sought in all cases.

A uniform management protocol was implemented in each case, which included stress ulcer prophylaxis, monitoring and correction of blood sugar levels, inotropic support to maintain mean arterial pressure above 60 mm Hg, and elective ventilation in grade IV encephalopathy and in grade III encephalopathy with cerebral edema. Intravenous mannitol was used to control cerebral edema. Prophylactic antibiotics were started at presentation in all cases (such as piperacillin-tazobactam, vancomycin, and fluconazole). Antibiotic therapy was modified if necessary, based on culture reports. Antibiotics were continued until neurological recovery and resolution of evidence of infection. Renal replacement therapy (hemodialysis) was used whenever required. All hepatotoxic antitubercular drugs were discontinued, and a modified regimen consisting of ethambutol and ofloxacin with or without streptomycin was used if active tuberculosis was present. Adverse events, such as gastrointestinal bleeding, renal failure, seizure, and death that occurred during hospitalization were recorded. Liver transplantation for ALF is not available at our center, and each patient was followed up until recovery or death.

Prothrombin time (PT) prolongation over control was estimated every day in each patient. It was calculated on freeze (−80°C)-thawed (37°C) plasma by the manual tilt-tube technique. The clotting time in seconds was recorded after adding 200 μL thromboplastin reagent (Liquiplastin—rabbit brain thromboplastin; Tulip Diagnostic (P) Ltd., India; ISI = 1.5) to 100 μL plasma.

Viral Markers

Serum from each patient was tested for hepatitis B surface antigen, immunoglobulin M (IgM) anti-hepatitis B core and IgM antibody against HAV using commercial enzyme-linked immunosorbent assay (ELISA) kits (Organon, Teknika, Netherlands). IgM antibody to open reading frame (ORF)-1, ORF-2, and ORF-3 of HEV and HEV RNA were tested by methods described by us.26, 27 Anti–hepatitis C antibody was tested by using a third-generation commercial enzyme-linked immunosorbent assay (Xcyton, Bangalore, India). Hepatitis C virus RNA was tested if needed, by reverse transcription nested polymerase chain reaction.13

Appropriate tests for autoimmune hepatitis (anti-nuclear antibody, anti-smooth muscle antibody, anti-liver kidney microsomal antibody) and Wilson's disease (serum ceruloplasmin, 24-hour urinary copper, and serum copper) were performed in each case. Each patient had bedside ultrasonography.

Non-A non-E virus–induced ALF was diagnosed if acute hepatitis B virus, HEV, HAV, and HCV infection were absent and other known causes of ALF were excluded.

Statistical Analysis

Normally distributed continuous variables were expressed as mean (SD), and the continuous variables with skewed distribution were expressed as median (range). Categorical data were presented as proportions. For the comparison of continuous covariates between three groups, namely, ALF caused by ATT, HEV, and non-A non-E hepatitis, a one-way analysis of variance with Bonferroni correction as a posthoc test was used. Similarly, the comparisons between these groups for skewed data were performed by the Kruskal-Wallis test followed by the Mann-Whitney test with adjusted P-values.

Univariate analysis was performed to compare survivors and nonsurvivors using an independent t test or Mann-Whitney U test for continuous variables, and a chi-squared test or Fisher's exact test for categorical variables, wherever applicable. A P-value of less than 0.05 was taken as significant. The significant variables were dichotomized using discriminant values derived by constructing receiver operating characteristic curves for most relevant variables. Multivariable logistic regression was performed to identify significantly important variables associated with mortality. The variables with significance P ≤ 0.10 in the univariate analysis were taken in the multivariable analysis. The variables were selected by using stepwise selection procedure in the multivariable analysis with entry probability 0.05 and removal probability 0.1. Odds ratios with 95% confidence intervals for each variable in the final multivariable analysis were reported. Data were analyzed by using SPSS software verson 15.0 (SPSS, Chicago, IL) and STATA (version 9).


One thousand two hundred twenty-three (n = 1223) ALF patients were hospitalized during the study period. The etiological breakup has been depicted in Fig. 1. ATT was the cause of ALF in 70 (5.7%) cases. HEV and a non-A non-E virus were implicated in 364 (29.8%) and 468 (38.3%) patients, respectively. We identified another 15 patients (1.2%) with ALF who, while on ATT, also had positive serological markers of hepatitis virus infection (Fig. 1).

Figure 1.

Cause of all cases of acute liver failure (n = 1223). *Seven patients IgM anti-HEV positive; three patients IgM anti-hepatitis B core positive; four patients HBsAg positive, and one patient anti-HCV positive (n = 15; ATT+ viral infection). #One hundred six patients HBsAg positive; eight patients anti-HCV positive; three patients both HBsAg & anti-HCV positive. ¶Negative serology for hepatitis B virus, HEV, HAV, and HCV infection and absence of any identifiable cause of ALF (non-A non-E ALF) ¥Etiological evaluation was not available for 29 patients.

Magnitude of AT-Induced ALF

During the study period, 91(7.44%) patients with ALF had consumed ATT for more than 1 week. In six of them, the detailed compositions of ATT were unavailable and were excluded. The remaining 85 patients had taken isoniazid and rifampicin with or without pyrazinamide for at least 7 days. Fifteen of these 85 patients had associated hepatitis virus infection (seven patients: IgM anti-HEV positive; three: IgM anti-hepatitis B core positive; four: hepatitis B surface antigen (HBsAg) positive and one: anti-HCV positive). Therefore, 70 (5.7%) cases of ALF were attributable to ATT alone, and 15 (1.2%) in addition had hepatitis virus infection. The magnitude of ATT-ALF apparently declined during the last decade. Although 74%(52/70) of patients with ATT-ALF were admitted during 1986 to 1997, only 26%(18/70) were hospitalized between 1998 and 2008.

Baseline Characteristics, Laboratory Parameters, and Complications in Patients with ATT-ALF (n = 70)

The mean age (±SD) was 32.87 (15.8) years, whereas the median (range) age was 27 (13-80) years (Table 1). Forty-nine (70%) patients were younger than 35 years. In a large proportion of patients (n = 44, 62.8%), ATT was started on the presumptive diagnosis of tuberculosis. Only 26 of 70 (37.2%) patients had a definite diagnosis of tuberculosis (Table 1). There were 49 (70%) women or girls, and five (10%) of them were pregnant. The median IEI was 4.5 days (range, 0-30). The median interval of hospitalization after onset of ALF was 2 (0-9) days. Advanced encephalopathy (grades III-IV) and cerebral edema at presentation was documented in 57 (75.5%) and 29 (41.4%) patients, respectively. The mean PT prolongation was 27.68 ± 17.05 seconds. The details of other laboratory parameters are presented in Table 1. During hospitalization, complications such as gastrointestinal bleed, seizures, infection, and renal failure occurred in seven (10%), five (7.1%), 26 (37.1%), and seven (10%) patients, respectively. The corresponding parameters in ALF caused by ATT with hepatitis virus co-infection were similar except for the ATT-encephalopathy interval, which was significantly longer than in the ATT-ALF group (median, 120 versus 30 days, P = 0.002) (Table 1).

Table 1. Baseline Characteristics, Laboratory Parameters, and Complications in ALF Associated with ATT With or Without Hepatitis Virus Coinfection (N = 85)
VariablesATT (N = 70)ATT + Hepatitis Virus (N = 15)P
  • Variables are expressed as median (range), unless stated otherwise.

  • *

    Tubercular involvement of two or more noncontiguous sites.

  • One patient each for bone, meningeal, and pericardial tuberculosis.

  • Started on ATT on the basis of presumptive diagnosis of tuberculosis without any definite evidence (In patients with pyrexia of unknown origin, unexplained weight loss, pleural effusion, ascites, lymphadenopathy).

  • Abbreviations: AST, aspartate transaminase; ALT, alanine transaminase.

Mean age (years) mean ± SD, (range)32.87 ± 15.13, (13-80)29.13 ± 12.99, (17-68)0.39
Female, n (%)49 (70)07 (46.6)0.08
Pregnant, n (% of female)05 (10)001.00
Icterus—encephalopathy interval (days)4.5 (0-30)10 (0-30)0.11
ATT—encephalopathy interval (days)30 (7-350)120 (60-210)0.002
Sites of tuberculosis: (n %)  0.27
 Lymph node04 (5.7)02 (13.3) 
 Pleuropulmonary13 (18.5)05 (33.3) 
 Abdominal04 (5.7)02 (13.3) 
 Disseminated*02 (2.8)00 
 Others03 (4.2)00 
 Empirical44 (62.8)06 (40) 
Encephalopathy grade, n (%)   
 I-II17 (24.3)3 (20) 
 III-IV51 (75.7)12 (80)0.78
Cerebral edema at admission, n (%)29 (41.4)07 (46.7)0.72
Bilirubin level (mg/dL)12.5 (2.5-52.3)15 (3-31)0.61
Peak bilirubin level (mg/dL)14.5 (6.5-52.3)21.8 (6.6-36)0.36
AST (IU/L)258 (93-3873)244 (100-1980)0.60
ALT (IU/L)298 (91-2920)248 (69 -2865)0.70
Peak ALT (IU/L)783 (170-2920)583 (185-2856)0.26
Alkaline phosphatase (U/L)127 (9-574)141 (11-368)0.77
Serum albumin (mg/dL), mean ± SD2.96 ± 0.612.90 ± 0.290.72
Prothrombin time prolongation (seconds)27 (4-60)37 (10-60)0.71
Peak prothrombin time prolongation (seconds)38 (12-60)48 (24-60)0.23
Blood urea (mg/dL)23 (12-177)28 (14-90)0.49
Serum creatinine (mg/dL)0.7 (0.2-4.0)0.70 (0.2-2.0)0.25
Complications observed during hospitalization, n (%):   
 Gastrointestinal bleed7 (10)000.34
 Seizures5 (7.1)2 (13.2)0.62
 Infection26 (37)6 (40)0.83
 Renal failure7 (10)000.34
Duration of hospitalization (days)   
 Admission to death time (patients who died)3 (1-27)6 (3-17)0.14
 Admission to discharge time (patients who survived)8.17 (2-24)9 (4-21)0.68

Comparisons of Clinical Characteristics, Laboratory Parameters, Complications, and Survival Rates Between ALF Attributable to ATT, HEV, and Non–A, Non–E Hepatitis

The two most frequent causes of ALF at our center were non-A non-E hepatitis (38%, n = 468) and acute hepatitis E (30%, n = 364), respectively (Fig. 1). The mean age of patients with ATT-ALF was higher than that of patients with ALF attributable to the former two causes (32.87 versus 26.8 and 29.08 years; P < 0.001, Table 2). The proportion of women with ATT-ALF was similar to that in HEV-ALF (70% in each group). Among women, the pregnancy rates were significantly lower in ATT-ALF than in the other two etiological groups (P < 0.001). The grades of encephalopathy, IEI, and frequency of cerebral edema at presentation were similar between the three groups (Table 2). The degree of liver enzymes elevation was significantly lower in ATT ALF compared with the other two causes (Table 2). The proportion of patients with a PT prolongation longer than 25 seconds among patients with ATT-ALF was similar to that of HEV-ALF (57% and 55%, P = 0.67), but was significantly higher than in non-A non-E ALF (57% and 41%, P = 0.014). The complications encountered in each group of ALF patients have been depicted in Table 2. The survival frequency among patients with HEV-ALF (53.8%) was significantly higher than that in ATT-ALF (32.8%) (P < 0.001). However, the survival rates were similar between ATT-ALF and non-A non-E ALF (32.8% and 38.2%; P = 0.385) (Table 2). Therefore, survival among patients with ATT-ALF and non-A non-E ALF was significantly lower than in patients with HEV-ALF.

Table 2. Comparisons of Baseline Characteristics, Laboratory Parameters, Complications and Survival Rates Between Patients with ALF Due to ATT, HEV, and Non-A, Non-E Hepatitis (n = 902)
VariablesATT N = 70HEV N = 364Non-A, Non-E N = 468P
  • All variables are expressed as median (range), unless stated otherwise.

  • *

    Significant between all three combination (one-way ANOVA).

  • Significant between ATT and non-A, non-E (chi-squared test).

  • Significant between ATT and either of two (one-way ANOVA with Bonferroni correction).

  • §

    Significant between ATT and HEV (Kruskal-Wallis followed by Mann-Whitney test).

  • Significant between ATT and other two (Kruskal-Wallis test followed by Mann-Whitney test)

  • Significant between HEV and non-A, non-E (one-way ANOVA with Bonferroni correction).

  • **

    Significant between ATT and HEV (chi-squared test).

Mean age (years) ± SD32.87 ± 15.1326.8 ± 9.1929.08 ± 12.08<0.001*
Women, n (%)49 (70)255 (70.05)223 (47.64)<0.001
Pregnant, n (% of women)5 (10)145 (56.8)72 (32.28)<0.001
Icterus—encephalopathy interval (days)4.5 (0-30)4.0 (0-30)4.0 (0-30)0.081
Encephalopathy grade, n (%)   0.567
 I-II17 (24.3)81 (22.2)93 (19.9%) 
 III-IV51 (75.7)283 (77.8)375 (80.1%) 
Cerebral edema at admission, n (%)29 (41.4)160 (44)242 (51.7)0.060
Bilirubin level (mg/dL)12.5 (2.5-52.3)13.5 (3-73)14.3 (2.1-73)0.390
AST (IU/L)258 (93-3873)442 (75-6060)303 (79-4815)0.018§
ALT (IU/L)298 (91-2920)742 (71-5840)580 (77-7590)<0.001
Prothrombin time prolongation, n (%)   <0.001
 <25 seconds30 (43)165 (45)275 (59) 
 25- <60 seconds31 (44)128 (35)152 (32) 
 ≥ 60 seconds09 (13)70 (20)42 (09) 
Alkaline phosphatase (U/L)127 (9-574)239 (12-1544)181 (6-1657)<0.001
Serum albumin (mg/dL), mean ± SD2.96 ± 0.612.88 ± 0.662.304 ± 0.690.005¥
Blood urea (mg/dL)23 (12-177)18 (10-144)20 (10-208)<0.001
Serum creatinine (mg/dL)0.7 (0.2-4.0)0.96 (0.2-11)0.7 (0.3-11)0.700
Complications observed during hospitalization, n (%):    
 Gastrointestinal bleed7 (10)20 (5.5)33 (7.1)0.338
 Seizures5 (7.1)46 (12.6)37 (7.9)0.060
 Infection26 (37)147 (40.4)99 (21.8)<0.001
 Renal failure7 (10)37 (10.2)55 (11.8)0.740
Survival, n (%)23 (32.85)196 (53.85)179 (38.24)<0.001**


Forty-seven patients (67.1%) died, and 23 (32.9%) survived. None underwent liver transplantation. The median (range) time of death from day of hospitalization was 3 (1-27) days. The numbers of patients who died at the end of days 3, 5, 10, and 30 of hospitalization were 25 (35.7%), 36 (51.4%), 43 (61.4%), and 47 (67.1%), respectively, indicating that 76% (36/47) of deaths occurred within 5 days of hospitalization. The mortality rates among patients with ALF caused by ATT with hepatitis virus co-infection was markedly high (12/15, 80%); however, it was not significantly different (P = 0.53) from the mortality rate in patients with ATT-ALF alone. Among all causes, HEV-ALF had the best survival rates (Fig. 2). Although there has been a decrease in the cases of ATT-induced ALF during last decade, the mortality remained the same (63.07% [35 of 52 patients]) during 1986-97 and 66.7% [12 of 18 patients] during 1998-2008). Postmortem liver biopsy using Menghini's needle was obtained in 38 of the 47 patients who died. Consent for liver biopsy was provided by the nearest relative. Histological evaluation revealed morphology of acute hepatitis in all, with massive necrosis (n = 12) or submassive necrosis (n = 16) or bridging necrosis (n = 10). These histological features were not specific to ATT-ALF, and similar histological features were documented in the postmortem liver biopsy specimens of patients with ALF in which the ALF was attributable to other causes.

Figure 2.

Outcome of ALF patients as per cause (n = 1194). All recovered spontaneously; none underwent liver transplantation. Mortality of ATT-ALF was significantly higher only in comparison with that of HEV-ALF (P = 0.001). However, in comparison with other causes, the mortality attributable to ATT-ALF was not significantly different (data not shown). Percentage of surviving patients is given above each bar.

Predictors of Outcome

Univariate analysis showed that the proportion of patients with advanced hepatic encephalopathy (grades III-IV), cerebral edema, median serum bilirubin, and PT prolongation at presentation were significantly higher among nonsurvivors than in survivors (Table 3). All continuous variables that were found to be significant by univariate analysis were dichotomized by the construction of receiver operating characteristics curves for better discrimination of the cutoff values between survivors and nonsurvivors. The curve was constructed using a value of serum bilirubin and prothrombin time prolongation for each patient as independent variables and mortality as outcome (Fig. 3A, B). The areas under the curve were 0.737 (95% CI, 0.611-0.863) and 0.717 (95% CI, 0.57-0.86), respectively. A cutoff of serum bilirubin at 10.8 mg/dL was found to be 67.39% sensitive and 73.9% specific for predicting mortality. The PT prolongation greater than 26 seconds yielded 71.05% sensitivity and 68.18% specificity for predicting mortality. At a level of serum bilirubin at least 10.8 mg/dL, 84% of patients died and 16% survived; whereas below this cutoff, the corresponding figures were 46% and 53% (P = 0.001). Similarly, at PT prolongation of 26 seconds or more, the percentage of nonsurvivors and survivors were 80% and 20%, whereas at a level below 26 seconds, they were 42% and 58%, respectively (P = 0.003).

Figure 3.

(A) Receiver operating characteristic curve of serum bilirubinlevels and mortality and (B) of prothrombin time prolongation and mortality.

Table 3. Comparison of Variables Between Survivors and Nonsurvivors in Patients with ATT-ALF (n = 70)
VariablesSurvivors (N = 23)Nonsurvivors (N = 47)P
  1. All variables are expressed as median (range) unless stated otherwise.

Age (mean ± SD)28.78 ± 12.9534.87 ± 16.820.132
Sex, men:women7:1614:330.956
Pregnant (%)2 (8.6)3 (6.38)1.00
Icterus—encephalopathy interval (days)3 (3-30)5 (1-30)0.647
Encephalopathy grade (%)   
 I-II9 (52.94)8 (47.06) 
 III-IV14 (26.42)39 (73.5)0.043
Cerebral edema at admission (%)5 (17.24)24 (82.76)0.019
ATT–ALF interval (days)17.5 (7-330)41 (7-350)0.121
Sites of tuberculosis: (n %)  0.440
 Lymph node0004 (8.5) 
 Pleuropulmonary06 (26.08)08 (17.02) 
 Abdominal02 (8.6)02 (4.2) 
 Disseminated0002 (4.2) 
 Others0003 (6.3) 
 Empirical15 (65.2)29 (61.7) 
Bilirubin level (mg/dL)6.5 (1.5-34.5)16.85 (3.2-52.3)0.001
AST (IU/L)169.5 (73-980)296 (90-912)0.189
ALT (IU/L)240 (78-980)250 (71-938)0.394
Alkaline phosphatase (U/L)137 (10-574)122.5 (9-437)0.994
Albumin (mg/dL) mean ± SD2.97 ± 0.71)2.96 ± 0.560.969
Prothrombin time prolongation (seconds)18.5 (4-60)31.5 (6-60)0.005
Blood urea (mg/dL)23 (12-102)23 (12-177)0.707
Serum creatinine (mg/dL)1.0 (0.2-3.0)1.0 (0.3-4.0)0.807

A multivariate analysis with logistic regression showed that serum bilirubin (≥10.8 mg/dL), PT prolongation (≥26 seconds), and grade III/IV hepatic encephalopathy at presentation were independent predictors for mortality; odds ratios were 13.08, 6.81, and 6.08, respectively. Subsequently, sensitivity, specificity, positive predictive value, and negative predictive value of one or more of the three adverse factors were calculated. The mortality rate was found to be increased with the increasing number of adverse prognostic factors (Table 4).

Table 4. Sensitivity, Specificity, and Predictive Value of Mortality Depending on Number of Adverse Prognostic Factors Present
Number of Adverse Factors* PresentNDeath N (%)Sensitivity (%)Specificity (%)PPV (%)NPV (%)Diagnostic Accuracy (%)
  • *

    Serum bilirubin ≥10.8 mg/dL, Prothrombin time prolongation ≥26 seconds, Encephalopathy ≥ grade III.

  • Abbreviations: PPV, positive predictive value; NPV, negative predictive value.

At least one of three5637 (66.1%)10013.666.710067.7
At least two3630 (83.3%)81.172.783.369.677.9
All three1515 (100%)40.51001005062.7

Prognostic Power of King's College Hospital Criteria

At admission, 22 of (31.4%) of 70 patients with ATT-ALF had King's College Hospital (KCH)28 criteria for death (≥3 criteria); however, 16 of them died, and six survived. In a univariate analysis, KCH criteria did not discriminate survivors (06/23, 26.1%) from nonsurvivors (16/47, 34%; P = 0.59), and six of 23 survivors had three or more bad prognostic factors; among 47 who died, only 16 had three or more bad prognostic factors). Therefore, KCH criteria had a low sensitivity (34.04%) and a low negative predictive value (35.4%) for mortality. However, the specificity and positive predictive value was better (74% and 72.7%, respectively).

Prognostic Power of Model End-Stage Liver Disease Score.

From 1986 to 1996, the model of end-stage liver disease (MELD) score could not be calculated because international normalized ratios were not available. Therefore, prognostic evaluation of MELD could be done only in 21 consecutive patients with ATT-ALF who were admitted during 1997-2008. The median (range) MELD score among patients who died (n = 15) was 41 (22-52), and in those who survived (n = 6) was 30 (24-39); the difference was statistically significant (P = 0.009). With a MELD scores cutoff value of 33 or higher, the sensitivity, specificity, positive predictive value, and negative predictive value for predicting mortality were 73%, 66.6%, 86.6%, and 50%, respectively.


The current study is the largest prospective study on ATT-ALF reported to date. ATT was the cause in 5.7% (n = 70) of the ALF patients (n = 1223) admitted to our hospital during the past 23 years. Although the proportion of ALF caused by ATT is small, it is substantial in absolute number and is a cause of concern because tuberculosis is highly prevalent in South Asia and Africa, where a large population reside. The estimated prevalence of tuberculosis in India is 283 per 100,000 populations11 (approximating 3.3 million when applied to total population of India—1169 million). Approximately 0.33 million people in India die of tuberculosis each year.11 Patients consuming ATT have a 0.01% risk of developing ALF,9, 10 and if we presume that even half (1.65 million) of such patients are taking ATT at one point of time, the expected number of ATT-ALF patients will be 1650 within 30 days of taking the drugs. Considering a mortality rate of 67% as documented in the current study, 1105 of these 1650 ATT-ALF will die because of ATT. In association with hepatitis virus infection, we noticed implication of ATT in another 15 (1.2%) ALF patients. Therefore, the contribution of ATT in magnitude of ALF population may be even higher in countries in which tuberculosis as well as hepatitis virus(es) are endemic. In the current study, approximately 63% of the patients with ATT-ALF consumed ATT empirically without objective evidence of tuberculosis (Table 1), which could have been avoided. In the endemic region of tuberculosis, physicians often treat patients with pyrexia of unknown origin with ATT. The current study identified the risk associated with such therapy. Such information is important for health authorities and physicians dealing with the treatment of tuberculosis to prevent unwarranted deaths of ATT-ALF.

The presentation of patients with ATT-ALF was uniformly homogenous. Most had hyperacute presentation, the median IEI being 4.5 (0-30) days. Patients with ALF attributable to HEV and non-A non-E also had hyperacute presentation (Table 2).13 The clinical presentation was similar in all ALF regardless of causes. During hospitalization, the frequency of complications such as gastrointestinal bleed, seizures, and renal failure were also similar among patients with ATT, HEV, and non-A non-E induced ALF. The liver failure rather than its cause resulted in complications. ATT-ALF has been shown to occur within first 2 months to 14 months after initiation of therapy.3, 5 In this study, the median duration of ATT before onset of ALF was short (30 days). Therefore, the underlying cause of liver injury in these patients may be idiosyncratic reaction to ATT rather than dose-related cytotoxicity.

Compared with reported age at risk for ATT hepatotoxicity (>35 to <50 years),12, 29, 30 patients in the current study were younger, mean age (±SD) being 32.87 (15.8) years. Seventy percent (49 of 70) were younger than 35 years. Seventy percent of the study populations were females. The reason for female preponderance is unclear. An epidemiological study reported that women in reproductive age had rapid progression of tuberculosis from infection to active disease compared with age-matched men.31 Therefore, women may be receiving ATT more frequently than are men. Furthermore, women have been documented to have a higher incidence of ATT hepatotoxicity in various studies.30, 32 There is a sex-based inequality in developing countries that may affect the female's health in terms of poor health seeking practices leading to inappropriate use and monitoring of anti-tubercular treatment. Irrespective of cause, female preponderance (71%) among patients with ALF has been documented in a US ALF study as well.33 Ten percent of ATT-ALF patients were pregnant. It may be because of sex predisposition for hepatotoxicity to ATT per se rather than pregnancy predisposing to ATT-ALF. It also may be because of the high fertility rate (2.9%) in India.34 A substantial number (15%-60%) of pregnant women with HEV hepatitis have been shown to progress to ALF.35-37 It is not clear whether pregnant women with ATT hepatitis progress more often to ALF. However, the proportion of pregnancy in ATT-ALF was significantly less than in HEV and non-A non-E associated ALF (Table 2).

ATT-ALF was associated with a poor outcome. The mortality rate was 67.1%, significantly higher than that of HEV-ALF (46%) (P = 0.001). Only 32.9% patients survived with intensive medical treatment. The spontaneous survival of ALF from idiosyncratic drug reactions, including those from ATT, in the United States is 20% to 30%, and if we assume approximately 15% of patients who undergo liver transplantation would actually have survived with medical therapy alone, then the survival of ALF from ATT in the United States would be approximately 35% to 45%, which is quite close to the survival described here. This suggests that the intensive care unit improves outcomes of ATT-ALF only modestly. Three-fourths (76%) of ATT-ALF patients died within 5 days of hospitalization; hence, liver transplantation should be considered at the earliest. The reason for high mortality among ATT-ALF patients is unclear. The possible explanation could be decreased hepatocyte regeneration because of fatty infiltration and idiosyncratic damage associated with ATT. Another notable finding in our study was a very high (80%) rate of mortality among the ALF patients with combined ATT and viral cause. However, it was not significantly higher than that of ATT-ALF (P = 0.53), which may be attributable to smaller number of patients in the former group. Several studies have shown that hepatitis B or hepatitis C virus co-infection increases the risk of ATT-induced hepatotoxicity.38-40 Because ATT-ALF has high mortality, and the current study showed that two thirds of such patients had been treated with ATT without precise evidence of tuberculosis, increased physician awareness of ATT-ALF is necessary. It is largely preventable. Avoiding indiscriminate use of ATT, prescribing medications in appropriate doses and duration, monitoring of liver functions, and withdrawal of drug at the appropriate time may prevent ATT-ALF. Although there are several known risk factors for ATT hepatotoxicity, early predictors of ALF subsequent to documented transaminitis is unknown.

Our study also intended to identify early predictors of mortality. In univariate analysis, four clinical parameters (serum bilirubin, advanced encephalopathy, cerebral edema, and PT prolongation) discriminated survivors from nonsurvivors (Table 3). On multivariate analysis, three independent predictors of mortality were serum bilirubin (≥10.8 mg/dL), PT prolongation (≥26 seconds), and grade III/IV hepatic encephalopathy at presentation. The odds ratio of serum bilirubin of at least 10.8 mg/dL was 13. Therefore, onset of mild hyperbilirubinemia in patients receiving ATT should be of concern. We also found that the mortality rates were increasing with the increasing number of these adverse prognostic factors. The KCH criterion was found unhelpful in ATT-ALF because it yielded a low sensitivity. In a subgroup analysis, MELD score (at ≥33) discriminated survivors from nonsurvivors with a reasonable sensitivity (73%) and specificity (66.6%). However, the presence of any two of three poor prognostic factors of our study yielded the best prognostic power, with a sensitivity of 81% and specificity of 73% (Table 4). A relatively lowered prognostic power of MELD may be attributable to noninclusion of encephalopathy grade at presentation, a variable that was identified as an important predictor of outcome in these patients.

The strength of our study is that the data were from a single center, ensuring a homogeneous patient cohort. Ironically, the lack of a liver transplantation facility enabled us to study the comprehensive natural history of ATT-ALF. However, our study included mostly referral patients and did not detect patients who either self-resolved or died before reaching hospital. Therefore, the actual magnitude and outcome may be different from that described here. Another weakness of our study is the definition of cerebral edema based on clinical parameters. In the absence of intracranial pressure monitoring, the frequency of cerebral edema could have been underestimated. Furthermore, we could not evaluate the prognostic power of APACHE II, SOFA, and Clichy criteria in patients with ATT-ALF, because some of the parameters needed to accomplish them were not available.

In conclusion, ATT-ALF constituted 5.7% of ALF at our center. In 63% of the patients, ATT was consumed empirically. Most patients had a hyperacute presentation. The presenting symptoms and complications were similar with ALF because of other causes. Female preponderance was documented. The mortality rate was high. Serum bilirubin, PT prolongation, and grade of encephalopathy at presentation were independent predictors of mortality. However, considering a high mortality rate, determining which factors are predictors may be less important. Because use of ATT in our country is exceedingly prevalent, increased physician awareness about this complication is needed, and ATT should not be prescribed without objective evidence of tuberculosis.