Outcome of liver transplantation for drug-induced acute liver failure in the United States: Analysis of the united network for organ sharing database

Abstract

Acute liver failure (ALF) is an uncommon but potentially lethal drug-related adverse effect that often leads to liver transplantation (LT) or death. A retrospective cohort study was performed with the United Network for Organ Sharing Standard Transplant Analysis and Research files. Recipients who underwent LT for drug-induced acute liver failure (DIALF) from 1987 through 2006 were analyzed. A total of 661 patients transplanted for DIALF were included in the analysis. The 4 leading implicated drug groups were acetaminophen (n = 265; 40%), antituberculosis drugs (n = 50; 8%), antiepileptics (n = 46; 7%), and antibiotics (n = 39; 6%). One-year estimated survival probabilities were 76%, 82%, 52%, 82%, and 79% for acetaminophen, antituberculosis drugs, antiepileptics, antibiotics, and others, respectively. The lower rate of survival among those exposed to antiepileptics was observed mainly in children. Of the 22 patients less than 18 years old who had ALF due to antiepileptics, 73% died within the first year. The difference in overall survival between acetaminophen-related and non–acetaminophen-related ALF was not statistically significant. Patients with acetaminophen-related ALF required dialysis prior to LT at a significantly higher rate than all other drug groups (27% versus 3%-10%, P < 0.0001). According to Cox proportional hazards regression analysis, the independent pretransplant predictors of death after LT were being on life support, DIALF due to antiepileptic drugs at age less than 18, and elevated serum creatinine. In conclusion, the leading drug groups causing LT due to DIALF in the United States were acetaminophen, antituberculosis drugs, antiepileptics, and antibiotics. Children who had ALF due to antiepileptics had a substantially higher risk of death after LT in comparison with other drugs. Patients transplanted for acetaminophen-related ALF required dialysis at a significantly higher rate. Being on life support, DIALF due to antiepileptics (at age less than 18), and elevated serum creatinine were independent pretransplant predictors of poor survival after LT for DIALF. Liver Transpl 15:719–729, 2009. © 2008 AASLD.

Drug-induced liver injury (DILI) is a potential complication of drugs that might progress to acute liver failure (ALF) and death. It is also the leading cause of regulatory action, including withdrawal of drugs from the market, restrictions in indications, and warnings to healthcare providers and patients, in the United States over the past 5 decades.1, 2

DILI is the leading cause of ALF in the adult population in the United States.3 It is also the most commonly identified cause of ALF in children between the ages of 3 and 18.3 According to the 1998-2007 data of the ALF Study Group, acetaminophen was the most common cause of ALF, accounting for 46% of ALF cases among US adults; it was followed by indeterminate causes of ALF (14%) and other drugs (11%).3 In the pediatric population, acetaminophen was the second most common cause of ALF between the ages of 3 and 18 years, and it was followed by autoimmune hepatitis, other, and metabolic causes, whereas other drugs were the sixth most common cause of ALF in that age range.3 LT is an ultimate treatment for patients with drug-induced acute liver failure (DIALF) in whom spontaneous recovery is unlikely. Between 1998 and 2007, 26% of the patients who suffered from acetaminophen-induced ALF and 31% of the patients who had ALF due to other drugs died without spontaneous recovery or LT.3 Among patients with acetaminophen-induced ALF, 9% underwent LT, whereas among those who had ALF due to other drugs, 41% underwent LT.3

Between 1990 and 2002; a total of 270 patients underwent LT in the United States for ALF due to identifiable drugs.4 Of these 270 patients, acetaminophen alone was the cause of ALF in 46%.4 The 1-year survival rate after LT was reported to be 77%, and there was no significant difference in patient survival rates between the acetaminophen and non-acetaminophen groups.4 Similar 1-year posttransplant survival rates were found in a study done in Sweden.5 According to the analysis of the Swedish Drug Reactions Advisory Committee database, 17 patients with DIALF underwent LT between 1985 and 2003.5 Forty-one percent of liver transplants were for acetaminophen-induced ALF. The 1-year survival rate for DIALF after LT was reported to be 77%.5

Although there are several models for predicting prognosis in ALF prior to LT, no prognostic model has been created to predict post-LT outcome in patients with DIALF.6 The United Network for Organ Sharing (UNOS) database contains data on outcomes of almost all liver transplants performed in the United States since 1987. In this retrospective cohort study, this large database was analyzed to identify all the drugs associated with DIALF and LT, determine the survival rates after LT, and develop a model that will predict the risk of death after LT for DIALF.

ALF, acute liver failure; ALP, alkaline phosphatase; ALT, alanine aminotransferase; CI, confidence interval; DIALF, drug-induced acute liver failure; DILI, drug-induced liver injury; HCV, hepatitis C virus; INR, international normalized ratio; LT, liver transplantation; OPTN, Organ Procurement and Transplantation Network; SD, standard deviation; STAR, Standard Transplant Analysis and Research; UNOS, United Network for Organ Sharing.

PATIENTS AND METHODS

Study Population

The analysis was based on Organ Procurement and Transplantation Network (OPTN) data of February 6, 2008. Standard Transplant Analysis and Research (STAR) dataset files on a total of 88,973 liver transplantations performed between October 01, 1987 and December 31, 2006 were analyzed. STAR files contain up-to-date pretransplant and posttransplant data of transplants performed in the United States and reported to OPTN since October 1, 1987. We analyzed data from patients who were transplanted for DIALF between October 1, 1987 and December 31, 2006. Patients whose diagnosis was other than DIALF were excluded.

Pretransplant Variables

UNOS collects data from transplant centers and organ procurement organizations by using liver transplant candidate and recipient registration forms and deceased donor and living donor registration forms.7 Twenty pretransplant recipient and 6 donor variables were obtained from the UNOS database. These included drug group, age, gender, ethnicity, status on the waiting list, recipient life support (ventilator, artificial liver, or other mechanism), dialysis 1 week prior to transplant, serum total bilirubin, serum creatinine, serum albumin, serum alanine aminotransferase (ALT), prothrombin time–international normalized ratio, cold ischemia time, warm ischemia time, days on the liver transplant waiting list, transplant region, ABO blood type, hepatitis B surface antigen, hepatitis C virus antibody (in patients < 18 years old), and hepatitis C virus status (in patients ≥ 18 years old). The donor variables were age, gender, donor type (cadaveric or living), deceased donor, 3 or more inotrop medications at the time of incision, non–heart-beating donor, and type of liver (split or whole).

Statistical Analysis

Statistical analyses were performed with SAS 9.1. We used statistical methods to estimate the association between patient characteristics at the time of their first transplant and survival time. Survival time was defined as the time from a patient's first transplant to death. Patients who were alive at the end of follow-up were considered right-censored in the analysis. Events that occurred after the first transplant (eg, retransplants) were not considered in the analysis. Thus, for example, if a patient had a retransplant 2 years after the first transplant and died 3 years after the second transplant, the variables considered in the analysis would be those at the time of the patient's first transplant, and the survival time would be 5 years. To reduce the influence of extreme laboratory values, natural logarithm transformation was applied to the serum total bilirubin, serum albumin, serum creatinine, ALT, and international normalized ratio.

Univariate Analysis

The difference in categorical and quantitative variables among the different drug categories were analyzed by a chi-square test and analysis of variance, respectively. Univariate Cox proportional hazards regression analysis was used to assess the association between the recipient/donor variables and survival. The Kaplan-Meier method was used to estimate the survival function. Survival functions in different groups were compared with the log-rank test.

Multivariate Analysis

Two different Cox proportional hazards regression models were developed with the PROC PHREG procedure of SAS 9.1. The first model was developed to assess the association between drug type and survival, controlling for potentially confounding variables. An interaction term was created between the variables “antiepileptic drug group” and “age less than 18” as age less than 18 was found to be an effect modifier for the antiepileptic group. A second model was developed to determine the risk score with stepwise Cox regression analysis (entry and stay cutoff levels were set at 0.05). The proportional hazard assumption was evaluated by the inclusion of time-dependent variables in the model and the construction of log–log survivor plots.

Validation of the Risk Score Model

To assess the validity of our approach to developing a risk score, the study cohort was randomly divided into 4 equal subsets. A model for risk score was developed on the basis of the data of 3 subsets (training sample; 75% of the study cohort) with stepwise Cox regression analysis. Then, based on this model, the risk score and predicted survival probabilities in the fourth subset were estimated. The entire procedure was repeated 4 times, and each time a different subset of the data was left out. This resulted in a risk score and survival probabilities for each subject that were estimated on the basis of a data set that did not include that subject. The means of all predicted survival probabilities were calculated in strata defined by the risk score group. Risk score groups were categorized according to 33rd and 66th percentiles of the risk score. These means were then compared to the survival probabilities estimated by the Kaplan-Meier approach. A graphical comparison was performed between the estimated Kaplan-Meier and predicted survival probabilities based on the stepwise Cox models.

RESULTS

Patient Characteristics

A total of 73,977 patients underwent LT between October 01, 1987 and December 31, 2006. Six hundred sixty-one patients whose diagnosis was DIALF were included in the analysis. The patients transplanted for DIALF accounted for 12% of all patients transplanted for ALF.

Among 661 patients, 567 were adult, and 94 were pediatric (age less than 18). The drugs that were associated with ALF and LT are shown in Table 1. Among the entire cohort, the 4 leading implicated drug groups were acetaminophen (n = 265; 40%), antituberculosis drugs (n = 50; 8%), antiepileptics (n = 46; 7%), and antibiotics (n = 39; 6%; Table 1).

Table 1. Drugs and Toxins Listed in the United Network for Organ Sharing Database for Drug-Induced Acute Liver Failure

Drug Group Name	Number of Patients
* Alcohol, hydrocodone, phenobarbital, isoniazid, rifampin, oxycodone, clarithromycin, acetone, amphetamine, fluconazole, ibuprofen, cocaine, and antifreeze. † Rifampin and pyrazinamide. ‡ Felbamate and unknown antiepileptic. § Terbinafine, ciprofloxacin, telithromycin, levofloxacin, itraconazole, and moxifloxacin HCl. ∥ Chaparral herbal tea, kava kava, and unknown herbal. ¶ Pravastatin, ezetimibe, fluvastatin, and unknown statin. # Etodolac, naproxen, and indomethacin. ** Sertraline, cyclobenzaprine, paroxetine, and bupropion. †† Tamoxifen, asparaginase, dactinomycin, and unknown antineoplastic. ‡‡ Iron, niacin, infliximab, pemoline, carbon tetrachloride, sulfasalazine, zafirlucast, loratadine, clonazepam, dapsone, diltiazem, isoflurane, labetolol, butorphanol, steroids, angiotensin converting enzyme inhibitor, efavirenz, nevirapine/lamivudine, valsartan, vitamin A, lisinopril, allopurinol, terbinafine, and unknown drug or toxin. §§ Patients transplanted for acute liver failure secondary to an unknown drug or toxin.
All (adult and pediatric)	661
Acetaminophen	265
Acetaminophen alone	253
Acetaminophen with another drug*	12
Antituberculosis	50
Isoniazid alone	48
Isoniazid with another antituberculosis drug†	2
Antiepileptics	46
Phenytoin	20
Valproic acid alone	20
Carbamazepine	3
Other antiepileptics‡	3
Antibiotics	39
Nitrofurantoin	12
Ketoconazole	8
Amoxicillin and clavulanate	5
Trimethoprim-sulfamethoxazole	5
Minocycline	2
Other antibiotics§	7
Herbals∥ and mushrooms	20
Herbals	6
Amanita phalloides	7
Unknown mushrooms	7
Propylthiouracil	19
Statins	12
Atorvastatin	3
Cerivastatin	2
Simvastatin	2
Other statins¶	5
Nonsteroidal anti-inflammatory drugs	10
Diclofenac	3
Bromfenac	2
Ibuprofen	2
Other nonsteroidal anti-inflammatory drugs#	3
Disulfiram	9
Halothane	8
Antidepressants	7
Amitriptyline	2
Nefazodone	2
Other antidepressants**	3
Methotrexate	5
Antidiabetics	5
Troglitazone	4
Unknown antidiabetic	1
Antineoplastics††	5
Methyldopa	5
6-Mercaptopurine/azathioprine	3
Fialuridine	3
Other drugs‡‡	29
Unknown drugs/drug category/toxins listed as 4100§§	121

Among the pediatric patients, the 4 leading drug groups were acetaminophen (n = 27, 29%), antiepileptics (n = 22, 23%), antituberculosis drugs (n = 8, 9%), and propylthiouracil (n = 6, 6%; Table 2). Among the pediatric patients transplanted for antiepileptic-induced ALF, valproic acid was the leading drug (16 of 22 patients, 73%; Table 2).

Table 2. Drugs and Toxins Listed in the United Network for Organ Sharing Database for Pediatric Patients Transplanted for Drug-Induced Acute Liver Failure

Drug Group Name	Number of Patients
* Carbamazepine and unknown antiepileptic. † Dactinomycin, asparaginase, and unknown antineoplastic. ‡ Amoxicillin and trimethoprim-sulfamethoxazole. § Disulfiram, halothane, iron, pemoline, isoflurane, and vitamin A. ∥ Patients transplanted for acute liver failure secondary to an unknown drug or toxin.
All	94
Acetaminophen	27
Antiepileptics	22
Valproic acid alone	16
Phenytoin	3
Other antiepileptics*	3
Antituberculosis: isoniazid	8
Propylthiouracil	6
Mushrooms	3
Antineoplastics†	3
Antibiotics‡	2
Other drugs§	8
Unknown drugs/drug category/toxins listed as 4100∥	15

The median age of the entire cohort was 36 (range: 1-73). The antiepileptic group had the lowest mean age (Table 3). Of 46 patients in the antiepileptic group, 18 were less than 10 years old.

Table 3. Recipient and Donor Characteristics in Patients Transplanted for Drug-Induced Acute Liver Failure

Recipient Characteristic	Acetaminophen	Antituberculosis	Antiepileptics	Antibiotics	Other Drugs	P Value
Abbreviations: ALT, alanine aminotransferase; HCV, hepatitis C virus; INR, international normalized ratio; SD, standard deviation.
Number of patients (%)	265 (40)	50 (8)	46 (7)	39 (6)	261 (39)
Age (SD)	31.66 (12.17)	38.36 (18.26)	21.67 (18.54)	43.67 (14.47)	39.49 (17.62)	<0.0001
Gender (%)
Female	199 (75)	32 (64)	31 (67)	35 (90)	179 (69)	0.0273
Male	66 (25)	18 (36)	15 (33)	4 (10)	82 (31)
Ethnicity (%)
White	213 (80)	13 (26)	29 (63)	27 (69)	172 (66)	<0.0001
Black	27 (10)	15 (30)	15 (33)	2 (5)	42 (16)
Hispanic	20 (8)	14 (28)	2 (4)	5 (13)	30 (11)
Other	5 (2)	8 (16)	0	5 (13)	17 (7)
Serum total bilirubin at transplant (mg/dL; SD)	9.70 (9.67)	24.71 (12.53)	18.05 (11.76)	26.87 (11.95)	19.74 (12.25)	<0.0001
Serum albumin at transplant (g/dL; SD)	3.05 (0.60)	2.87 (0.66)	2.90 (0.69)	2.74 (0.77)	2.86 (0.60)	0.002
Serum creatinine at transplant (mg/dL; SD)	3.21 (2.26)	1.31 (1.09)	1.38 (1.35)	1.86 (2.10)	1.74 (1.66)	<0.0001
INR at transplant (SD)	3.50 (2.46)	3.49 (2.30)	2.84 (1.67)	4.20 (3.41)	3.89 (6.36)	0.8888
ALT at transplant (IU/L; SD)	3777.25 (3663.69)	341.47 (416.51)	1027.36 (2089.97)	1235.21 (2564.67)	1138.54 (1853.63)	<0.0001
Days on the waiting list (SD)	2 (2)	6 (8)	21 (58)	3 (5)	28 (171)	0.0678
Region (%)
1	14 (5)	3 (6)	3 (7)	4 (10)	5 (2)	0.0116
2	40 (15)	4 (8)	8 (17)	1 (3)	25 (10)
3	28 (11)	2 (4)	5 (11)	3 (8)	47 (18)
4	18 (7)	2 (4)	8 (17)	4 (10)	13 (5)
5	64 (24)	12 (24)	4 (9)	11 (28)	55 (21)
6	6 (2)	1 (2)	1 (2)	1 (3)	10 (4)
7	29 (11)	8 (16)	4 (9)	6 (15)	22 (8)
8	19 (7)	2 (4)	3 (7)	2 (5)	17 (7)
9	19 (7)	10 (20)	1 (2)	3 (8)	26 (10)
10	16 (6)	3 (6)	4 (9)	2 (5)	22 (8)
11	12 (5)	3 (6)	5 (11)	2 (5)	19 (7)
Dialysis prior week to liver transplant: yes (%)	72 (27)	5 (10)	3 (7)	1 (3)	24 (9)	<0.0001
Simultaneous kidney transplantation: yes (%)	7 (3)	1 (2)	0	0	2 (1)	0.3277
Recipient life support pretransplant: yes (%)	217 (82)	25 (50)	32 (70)	17 (44)	134 (51)	<0.0001
Cold ischemia time (hours; SD)	7.54 (3.70)	8.85 (6.60)	9.04 (3.67)	7.33 (3.24)	8.32 (6.19)	0.1702
Warm ischemia time (minutes; SD)	49.47 (27.62)	46.92 (16.03)	61.24 (32.44)	49.41 (16.64)	51.26 (25.47)	0.1151
Status 1, 1A, 1B: yes (%)	241 (91)	37 (74)	33 (72)	33 (85)	191 (73)	<0.0001
HCV status in patients ≥ 18 years old (%)
Positive	7 (4)	0	0	1 (4)	5 (3)	0.5546
Negative	147 (93)	22 (96)	14 (100)	25 (93)	140 (95)
Unknown	0	0	0	1 (4)	2 (1)
Not done	5 (3)	1 (4)	0	0	1 (1)
Missing 196 (including all drug categories)
HCV antibody in patients < 18 years old (%)
Positive	0	0	0	0	0	0.2495
Negative	11 (100)	4 (80)	5 (83)	1 (100)	18 (95)
Unknown	0	0	1 (17)	0	1 (5)
Not done	0	1 (20)	0	0	0
Missing 52 (including all drug categories)
Hepatitis B surface antigen (%)
Positive	4 (2)	1 (2)	1 (3)	0	12 (5)	0.0142
Negative	205 (94)	40 (98)	35 (92)	33 (92)	214 (94)
Unknown	0	0	1 (3)	1 (3)	0
Not done	8 (4)	0	1 (3)	2 (6)	2 (1)
Missing 101 (including all drug categories)	101
ABO blood type (%)
A	106 (40)	13 (26)	21 (46)	12 (31)	108 (41)	0.2008
B	29 (11)	12 (24)	6 (13)	5 (13)	28 (11)
AB	8 (3)	0	1 (2)	3 (8)	11 (4)
O	122 (46)	25 (50)	18 (39)	19 (49)	114 (44)
Donor Characteristic
Age (SD)	35 (17)	33 (19)	29 (19)	37 (19)	35 (18)	0.1774
Gender (%)
Female	122 (46)	23 (46)	16 (35)	14 (36)	105 (40)	0.4096
Male	143 (54)	27 (54)	30 (65)	25 (64)	156 (60)
Donor type (%)
Cadaveric	260 (98)	50 (100)	45 (98)	38 (97)	256 (98)	0.8939
Living	5 (2)	0	1 (2)	1 (3)	5 (2)
Type of liver (%)
Whole	256 (97)	50 (100)	42 (91)	38 (97)	248 (95)	0.2218
Split	9 (3)	0	4 (9)	1 (3)	13 (5)
Deceased donor/3 or more inotrope medications at the time of incision: yes (%)	7 (3)	2 (4)	0	1 (3)	2 (1)	0.3001
Non–heart-beating donor: yes (%)	7 (3)	0	0	0	1 (1)	0.141

In all drug groups, the proportion of female patients was higher than that of male patients. The percentage of females was highest in the acetaminophen and antibiotic groups (75% and 90%, respectively; Table 3).

White patients constituted the majority in all drug groups except for the antituberculosis drug group, in which the proportion of black patients was the highest (30%; Table 3).

The mean serum creatinine was higher in the acetaminophen group compared to all other groups (3.21 versus 1.31-1.86 mg/dL, P < 0.0001; Table 3). In addition, the proportion of patients who required dialysis 1 week prior to transplant was significantly higher in the acetaminophen group compared to all other groups (27% versus 3%-10%, P < 0.0001; Table 3).

Patients in the acetaminophen group were also more likely to require life support in comparison with all other groups (82% versus 44%-70%, respectively, P < 0.0001; Table 3).

The mean serum total bilirubin was higher in the antibiotic group compared to all other groups (26.87 versus 9.70-24.71 mg/dL, P < 0.0001; Table 3).

Predictors of Survival Based on Univariate Analysis

Among the 5 leading drug groups; patients transplanted because of antiepileptic drug–induced ALF had the highest death rate after LT (Fig. 1). The median survival time for the entire cohort was 14.4 years. Among all patients, including adults and children, 1-year estimated survival probabilities were 76%, 82%, 52%, 82%, and 79% for acetaminophen, antituberculosis drugs, antiepileptics, antibiotics, and other drugs, respectively (Fig. 1). Patients who had ALF due to antiepileptics had a significantly higher death rate after LT than patients who had ALF due to other drugs (P = 0.005; Fig. 1). Among the patients who had ALF due to antiepileptics, 1-year survival was only 27% in patients less than 18 years old versus 75% in patients 18 years old or older.

Figure 1. Comparison of survival probabilities of patients transplanted for drug-induced acute liver failure (P = 0.005).

One-year survival for acetaminophen and non-acetaminophen groups was 76%, and the difference in overall survival after LT between acetaminophen and non-acetaminophen groups was not statistically significant.

There was no significant difference between the survival rates of pediatric and adult patients (P = 0.56). Among pediatric patients, 1- and 2-year survival probabilities were 0.68 and 0.67, respectively. Among adult patients, 1-year survival and 2-year survival were 0.78 and 0.74, respectively.

One-year survival was almost identical for patients listed as status 1 (1, 1A, or 1B) and non–status 1 (0.76 and 0.78, respectively).

Patients who were transplanted only once were 0.67 times less likely to die than patients transplanted more than once. The patients who had ALF due to antiepileptics had a higher frequency of retransplantation than other patients who had ALF due to all other groups (24% versus 4%-9%, respectively).

Table 4 shows the impact of each recipient and donor variable on the risk of death after LT. Gender, serum creatinine, being on life support, warm ischemia time, and transplant year significantly affected the risk of death after LT. Males were 1.4 times more likely to die after LT than females. A doubling in the level of serum creatinine increased the risk of death after LT by a factor of 1.19. Patient who were on life support were 2.08 times more likely to die after LT than patients who were not on life support. For every hour increase in warm ischemia time, the recipient risk of death increased by 40%. For every 1-year increase in the calendar year of transplant, the risk of death decreased by 5%.

Table 4. Univariate Analysis of Risk Factors for Mortality After Liver Transplantation for Drug-Induced Acute Liver Failure

Recipient Characteristic	Hazard Ratio	95% CI		P Value
Abbreviations: ALT, alanine aminotransferase; CI, confidence interval; INR, international normalized ratio.
Drug group
Acetaminophen	1.00
Antituberculosis	0.86	0.50	1.46	0.572
Antiepileptics	1.85	1.21	2.84	0.005
Antibiotics	0.64	0.32	1.28	0.208
Others	0.88	0.65	1.19	0.413
Age
≥18	1.00
<18	1.11	0.78	1.58	0.561
Gender
Female	1.00
Male	1.40	1.06	1.84	0.017
Ethnicity
White	1.00
Black	1.38	1.00	1.93	0.054
Hispanic	0.66	0.40	1.10	0.109
Other	0.59	0.29	1.21	0.151
Serum total bilirubin at transplant (Loge value)	1.08	0.93	1.24	0.321
Serum albumin at transplant (Loge value)	0.81	0.43	1.55	0.530
Serum creatinine at transplant (Loge value)	1.29	1.11	1.49	0.001
INR at transplant (Loge value)	1.36	0.94	1.98	0.102
ALT at transplant (Loge value)	1.00	0.91	1.09	0.976
Days on the liver waiting list	1.00	1.00	1.00	0.533
Region
1	1.00
2	0.92	0.48	1.75	0.791
3	0.68	0.35	1.32	0.251
4	0.40	0.17	0.94	0.036
5	0.57	0.31	1.08	0.084
6	0.50	0.18	1.41	0.189
7	0.85	0.44	1.64	0.621
8	1.03	0.51	2.08	0.945
9	0.65	0.32	1.33	0.235
10	0.79	0.39	1.62	0.523
11	0.82	0.40	1.69	0.585
Dialysis prior week to liver transplant
No	1.00
Yes	1.02	0.70	1.49	0.912
Simultaneous kidney transplantation (%)
No	1.00
Yes	1.26	0.52	3.06	0.610
Recipient life support pre-transplant
No	1.00
Yes	2.08	1.53	2.84	<0.0001
Cold ischemia time (hours)	1.01	0.98	1.03	0.690
Warm ischemia time (hours)	1.40	1.05	1.87	0.022
Status 1, 1A, 1B
No	1.00
Yes	1.08	0.78	1.51	0.643
Transplant year	0.95	0.93	0.98	0.001
ABO blood type
A	1.00
B	1.35	0.90	2.02	0.145
AB	1.09	0.53	2.26	0.811
O	1.06	0.79	1.42	0.693
Donor Characteristic
Age	1.01	1.00	1.01	0.114
Gender
Female	1.00
Male	1.06	0.81	1.38	0.690
Donor type
Living	1.00
Cadaveric	1.25	0.40	3.92	0.698
Type of liver
Split	1.00
Whole	0.90	0.46	1.75	0.747
Deceased donor/3 or more inotrop medications at the time of incision
No	1.00
Yes	1.10	0.41	2.96	0.853
Non–heart-beating donor
No	1.00
Yes	2.07	0.66	6.51	0.214

Joint Associations Between Drug Groups and Survival Based on a Multivariable Model

In comparison with the acetaminophen group (reference group), LT for ALF caused by antituberculosis drugs, antibiotics, and other drugs was associated with a lower risk of posttransplant death by factors of 0.86, 0.63, and 0.89, respectively (Table 5). On the other hand, in comparison with the acetaminophen group, LT due to ALF caused by antiepileptics increased the risk of posttransplant death by a factor of 4.13 among those less than 18 years old and by a factor of 1.03 among those 18 years old or older. Patients less than 18 years old were 2.67 times more likely to die after LT than patients 18 years old or older among those who had ALF due to antiepileptics (P = 0.02). These results did not change after adjustments for gender, type of the liver, donor type, days on the liver transplant waiting list, status 1, number of transplants, deceased donor being on 3 or more inotrop medications at the time of incision, cold ischemia time, warm ischemia time, blood type, non–heart-beating donor, ethnicity, and transplant region.

Table 5. Relative Effect of Drug Groups on Risk of Death After Liver Transplantation for DIALF

Recipient Variable	Hazard Ratio	95% CI		P Value
Abbreviations: CI, confidence interval; DIALF, drug-induced acute liver failure. * Acetaminophen is the reference group.
Drug group
Antituberculosis versus acetaminophen*	0.861	0.51	1.47	0.585
Antiepileptics versus acetaminophen
Among those ≥ 18 years old	1.028	0.52	2.05	0.938
Among those < 18 years old	4.127	2.11	8.09	<0.0001
Antibiotics versus acetaminophen	0.627	0.32	1.25	0.184
Other drugs versus acetaminophen	0.89	0.66	1.2	0.453
Age: <18 years versus ≥18 years	0.66	0.41	1.08	0.099
Among those with DIALF due to antiepileptics	2.666	1.201	5.919	0.0159
Among those with DIALF due to other drugs	0.66	0.41	1.08	0.099

Cox Model To Predict the Survival After Liver Transplantation for DIALF

The independent pretransplant predictors of death after LT identified by Cox stepwise regression analysis were being on life support, DIALF due to antiepileptic drugs at age less than 18, and elevated serum creatinine (Table 6). By using these variables, the risk score can be calculated as follows:

• 

For a hypothetical patient 17 years of age who was transplanted for antiepileptic-induced ALF and was on life support with serum creatinine of 2 mg/dL at the time of LT, the risk score would be 2.00 [risk score = 0.25 × (log_e 2) + 0.59 + 1.24 = 2.00].

Table 6. Independent Pretransplant Predictors of Survival After Liver Transplantation in Patients with Drug-Induced Acute Liver Failure

Prognostic Factor	Regression Coefficient	Standard Error	Hazard Ratio	95% CI		P Value
Abbreviation: CI, confidence interval.
Life support	0.59	0.17	1.81	1.31	2.51	0.0004
Antiepileptic at age < 18 years	1.24	0.26	3.46	2.07	5.76	<0.0001
Serum creatinine at transplant (Loge value)	0.25	0.08	1.29	1.10	1.50	0.0014

The baseline survival probabilities at 3 and 6 months and at 1, 2, 3, 4, and 5 years are shown in Table 7. Baseline status was considered for a patient who did not have ALF due to antiepileptics at age less than 18, was not on life support, and had a creatinine value of 1.0 mg/dL. The predicted survival probabilities at 3 and 6 months and at 1, 2, 3, 4, and 5 years can be calculated by the determination of the baseline survival probability at time t in Table 7 and then the application of the numbers to the survival equation:

• 

For example, for a hypothetical patient with antiepileptic-induced ALF who is 50 years old and is on life support with a serum creatinine of 2 mg/dL, the predicted survival at 3 months after LT would be 80% [survival (3 months) = (0.90)]. Figure 2 shows observed and predicted survival probabilities for mild, moderate, and high risk scores. There is good graphical agreement between the observed and predicted survival probabilities.

Table 7. Estimated Baseline Survival Probabilities

	Time
	3 Months	6 Months	1 Year	2 Years	3 Years	4 Years	5 Years
S0(t) (Estimated Baseline Survival Probability)	0.90	0.88	0.87	0.86	0.85	0.84	0.83

Figure 2. Observed and predicted survival probabilities after liver transplantation for drug-induced acute liver failure categorized according to risk score group.

DISCUSSION

In a previous report, Russo et al.4 described the survival rates among those with DIALF in the United States from 1990 to 2002 on the basis of the UNOS database. We extended that work, using a broader time period, exploring the relationship between survival and drug type, and developing a risk score.

Our analysis showed that the 4 leading drug groups causing LT due to DIALF in the United States were acetaminophen, antituberculosis drugs, antiepileptics, and antibiotics. The patients transplanted for antiepileptic-induced ALF had a higher risk of death after LT than the patients transplanted for acetaminophen induced ALF among all age groups. This risk was significantly higher for patients younger than 18. Among all the antiepileptic drugs, phenytoin and valproic acid were the most commonly associated with LT. Among children, valproic acid was the leading drug associated with LT. Interestingly, in comparison with all other drug groups, the frequency of listing as status 1 was the lowest in patients transplanted for antiepileptic-induced ALF. In addition, the mean liver transplant waiting time was longer for antiepileptic group compared to acetaminophen, antituberculosis, and antibiotic groups. It is difficult to determine if these differences had a causal relationship with the higher post-LT mortality of these patients. In addition, the mean cold and warm ischemia times in the antiepileptic group were the longest in comparison with all other drug groups. The reason for this difference is unclear; however, it is conceivable that the longer mean cold and warm ischemia times could at least partially explain the higher frequency of retransplantation in the antiepileptic group. Despite the lower frequency of listing as status 1, longer mean transplant waiting times, longer mean cold and warm ischemia times, and higher retransplantation rates, the relatively low survival probability among youth who had antiepileptic-induced ALF persisted after we controlled for these variables in multivariate analysis.

The significantly higher rate of post-LT death among pediatric patients who had ALF due to antiepileptic drugs might be related to valproic acid–induced hyperammonemic encephalopathy as 73% of children had ALF due to valproic acid. Carnitine depletion has been suggested by several investigators as the cause of valproic acid–induced hyperammonemic encephalopathy in children who were exposed to this antiepileptic in high doses or were under long-term treatment.8, 9 The degree of encephalopathy might have a significant impact on the post-LT survival. However, this information was not available in the database.

We found that the rate of dialysis was highest among patients transplanted for acetaminophen-induced ALF. This is not surprising, given that renal failure occurs in a high percentage of patients who have acetaminophen-induced ALF.10 As expected, the highest mean serum creatinine at transplant was found in the acetaminophen group. The impact of elevated pretransplant serum creatinine on survival after LT has been previously shown in clinical studies.11–15 Several investigators have reported that pretransplant abnormal renal function is significantly associated with posttransplant bacterial infections, sepsis, and renal failure and is a significant predictor of poor survival after LT.11–15

Our analysis showed that the highest and lowest mean serum total bilirubin levels were among the patients who were transplanted for antibiotic-induced ALF and acetaminophen-induced ALF, respectively. A higher mean serum total bilirubin level was an expected finding when the majority of the drugs that led to liver transplantation in our analysis were known to cause either cholestatic liver injury or a mixed type of liver injury.

Pediatric patients constituted 14% of our study cohort. In this group, acetaminophen, antiepileptics, antituberculosis drugs, and propylthiouracil were the leading associated drugs. We separately analyzed the survival rates of pediatric patients and did not find any significant difference in their posttransplant survival compared to survival rates of adults.

Another interesting finding of our analysis was the significant decrease in the risk of death after LT for every 1-year increase in the calendar year of transplant. This finding differs from our previous observation for a different cohort of LT patients and could be the result of improved surgical techniques, early listing for LT, and more potent immunosuppressive drugs in recent years.16 The reason for the previously reported slight increase in the risk of death after LT for every 1-year increase in the calendar year of transplant is unknown.16 It could be related to differences in study populations or the inclusion of sicker human immunodeficiency virus carriers in more recent cohorts.16

The strength of our study is the relatively large size of the cohort, which consisted of a total of 661 patients. We included all the patients who underwent LT for DIALF since 1987, which was the earliest date of UNOS data collection. This study also provided the most comprehensive list of the drugs and toxins that caused LT in the United States for DIALF since 1987.

Our study had some limitations, as expected from any retrospective database analysis. We reported the drugs that were listed in the UNOS database as the causes of LT for DIALF. However, we cannot estimate the number of patients who may have had DIALF but were not listed for LT, may have died before undergoing LT, or may have spontaneously recovered.

We were unable to assess the causality between drug exposure and ALF. Ideally, causality should be assessed by a hepatologist with or without the assistance of one of the causality assessment tools (eg, the Roussel Uclaf causality assessment method) to confirm DILI.17, 18 Although there have been some prospective studies done in Europe on the outcome of LT for DIALF, prospective studies with a causality assessment are needed in the United States. A preexisting chronic liver disease might have an impact on the severity of DILI. We were unable to verify this information because of incomplete data.

According to the International Consensus Meeting criteria on drug-induced hepatotoxicity, DILI can be subtyped to hepatocellular, cholestatic, or mixed on the basis of the ratio of serum (ALT/upper limit of normal ALT) to serum [alkaline phosphatase (ALP)/upper limit of normal ALP].1, 5, 18–21 However, we were unable to report the type of liver injury as data on serum ALP were not available in the database.

We developed a prognostic model based on the entire DIALF population, including both adult and pediatric patients. Therefore, one of the variables (antiepileptic- related DIALF at age < 18) in the model is related to the pediatric population. A separate prognostic modelcould have been developed for adult and pediatric patients; however, this would limit the predictability of the model by significantly reducing the sample size. The prediction of our model is good as the observed and predicted survival probabilities are almost identical. We validated our model by splitting the study cohort into 4 equal subsets. Validation of our risk score model in an independent cohort could be a better way to validate the model; however, as our cohort represents all the patients transplanted in the United States for DIALF, that form of model validation could be done only with the use of a cohort of patients transplanted outside the United States.

In conclusion, the leading drug groups causing LT due to DIALF in the United States were acetaminophen, antituberculosis drugs, antiepileptics, and antibiotics. Patients transplanted for antiepileptic-related ALF had a higher risk of retransplantation and death after LT in comparison with all other drug groups. In all drug groups, the proportion of female patients was higher. The mean serum creatinine and proportion of patients requiring dialysis prior to transplant and undergoing simultaneous kidney transplantation were higher in the acetaminophen group than in all other groups. Being on life support, DIALF due to antiepileptics (at age less than 18), and elevated serum creatinine were found to be independent pretransplant predictors of poor survival after LT for DIALF.