Clinical Gastroenterology and Hepatology 2014;-:-–-

Relationship Between Characteristics of Medications andDrug-Induced Liver Disease Phenotype and Outcome

Raj Vuppalanchi,* Raghavender Gotur,* K. Rajender Reddy,‡ Robert J. Fontana,§

Marwan Ghabril,* Andrzej S. Kosinski,k Jiezhun Gu,k Jose Serrano,¶ and Naga Chalasani*

*Department of Medicine, Indiana University School of Medicine, Indianapolis, Indiana; ‡Department of Medicine, University ofPennsylvania, Philadelphia, Pennsylvania; §Department of Internal Medicine, University of Michigan Medical School, Ann Arbor,Michigan; kDuke Clinical Research Institute, Durham, North Carolina; and ¶Liver Disease Research Branch, National Institute ofDiabetes and Digestive and Kidney Diseases, National Institutes of Health, Bethesda, Maryland

BACKGROUND & AIMS:

It is not known whether specific characteristics of medication are associated with type of drug-induced liver injury (DILI) or outcome. We examined the relationships among select charac-teristics of medications and DILI phenotype and outcome.

METHODS:

We analyzed 383 cases of DILI caused by a single orally administered prescription agent fromthe DILI Network Prospective Study with causalities of definite, highly likely, or probable.Relationship of daily dosage (‡50 mg vs £49 mg), preponderance of hepatic metabolism (‡50%vs <50%), or Biopharmaceutics Drug Disposition Classification System (BDDCS) class (1–4,based on solubility and metabolism of the drug) were compared with clinical characteristicsand outcomes.

RESULTS:

Compared with cases of DILI in the £49 mg/day group, those associated with daily dosages‡50 mg had shorter latency (median, 38 days vs 56 days; P [ .03) and a different biochemicalpattern of liver injury (P[ .04); no differences in recovery, severity, or outcome were observed.Patients with DILI caused by medications with or without preponderant hepatic metabolismdid not differ in clinical characteristics or outcomes. Compared with other classes of BDDCS,DILI caused by BDDCS class 1 medications had significantly longer latency (P < .001) andgreater proportion of hepatocellular injury (P [ .001). However, peak liver biochemical valuesand patients’ time to recovery, disease severity, and outcomes did not differ among the 4 BDDCSclasses.

CONCLUSIONS:

Characteristics of medications (dosage, hepatic metabolism, and solubility) are associated withfeatures of DILI such as latency and pattern of liver injury, but not with recovery, severity, oroutcome.

Keywords: DILIN; Liver Toxicity; Overdose; Side Effect; ALT.

Abbreviations used in this paper: ALT, alanine aminotransferase; BDDCS,Biopharmaceutics Drug Disposition Classification System; DILI, drug-induced liver injury; DILIN, Drug-Induced Liver Injury Network; ULN, upperlimit of normal.

© 2014 by the AGA Institute1542-3565/$36.00

http://dx.doi.org/10.1016/j.cgh.2013.12.016

Idiosyncratic drug-induced liver injury (DILI) likelyoccurs because of complex interplay among host

factors, environmental variables, and the nature of theoffending agent.1 Research efforts during the last decadehave therefore focused primarily on the understanding ofthe genetic factors that increase an individual’s suscepti-bility to DILI.2,3 However, medication characteristicssuch as daily dose administered and metabolism profileare starting to emerge as potential risk factors forDILI.4–7

It was initially pointed out by Uetrecht8 that drugsgiven at a daily dose of 10 mg or less are rarely if everassociated with a high incidence of DILI, and majority ofthe drugs that were either withdrawn from the market orreceived a black box warning due to hepatotoxicity wereprescribed at daily doses greater than 50 mg. Lammert

et al6 reported a statistically significant relationship be-tween daily dose of oral medicines and reports of liverfailure, liver transplantation, and death that were due toDILI. Lammert et al also reported that compounds with>50% hepatic metabolism had significantly higher fre-quency of alanine aminotransferase (ALT) >3� upperlimit of normal (ULN), liver failure, and fatal DILI.4

Recently, Chen et al5 suggested a combined effect of

2 Vuppalanchi et al Clinical Gastroenterology and Hepatology Vol. -, No. -

daily dosage of >100 mg and high lipophilicity on therisk of DILI among medications approved by the Foodand Drug Administration.

Drugs with higher lipophilicity resulting in increasedpermeability and uptake by hepatocytes may undergogreater biotransformation leading to toxic metabolitesand heightened susceptibility to DILI.7,9 Conversely,drugs with low lipophilicity have low permeability, un-dergo less hepatic metabolism, and may be at lower riskfor causing DILI because they are primarily eliminatedunchanged into the urine and bile.7,9 However, somedrugs (cefadroxil, cephradine, levofloxacin, loracerbef,ofloxacin, sotalol, and pregabalin) classified as havinglow permeability exhibit high absorption �90% becauseof presence of intestinal drug transporters.10 By evalu-ating metabolism, not permeability or lipophilicity, drugscan be categorized into 4 classes according to their sol-ubility and metabolism by using the BiopharmaceuticsDrug Disposition Classification System (BDDCS) classifi-cation: class 1 (high solubility, extensive metabolism),class 2 (low solubility, extensive metabolism), class 3(high solubility, poor metabolism), and class 4 (low sol-ubility, poor metabolism).11–13 The current study inves-tigated the relationship between DILI phenotypiccharacteristics and select drug properties such as (1)daily dosage �50 mg, (2) predominant hepatic meta-bolism (>50%), and (3) BDDCS classes 1–4.

Figure 1. Flowchart describing selection of study population.A total of 1083 subjects with DILI were enrolled up to March2012. Among these, 631 subjects had DILI from a single drug.After causality adjudication, there were 412 subjects withadjudication status of definite, highly likely, and probable(1, 2, or 3). An additional 29 cases were excluded because ofunknown dosage, drug start/stop date, or mode of adminis-tration. The final study cohort consisted of 383 DILI casesresulting from 107 different drugs that satisfied the study in-clusion criteria: (1) DILI occurring from a single agent, (2)administered orally, (3) known daily dosage, and (4) causalityof 1, 2, or 3.

Methods

We extracted data from the U.S. Drug-Induced LiverInjury Network (DILIN) prospective study database withthe following eligibility criteria: (1) cases that have beencausality adjudicated with a DILIN causality scoreprobable or higher as described previously14,15 and (2)DILI caused by single oral prescription agent taken daily.A DILIN severity score was assigned according to apreviously published scale from 1 (mild with serum totalbilirubin <2.5 mg/dL) to 5 (death or liver trans-plantation).16,17 Information regarding the drug dosageadministered to the patient was extracted from the DILINdatabase (DILIMED). Drugs were categorized on the ba-sis of daily dosage into those with and without appre-ciable hepatic metabolism that was based on informationwithin the drug’s approved package insert, DrugBank(www.drugbank.ca), or DailyMed (dailymed.nlm.nih.gov). BDDCS classification of the drugs was obtainedfrom previously published literature.11 By evaluatingmetabolism, not permeability, BDDCS is not subject tovariability as a result of transporter effects.11,12 De-mographic, biochemical, and clinical outcomes wereextracted, and analyses were performed on the followingvariables: (1) latency (time from drug exposure to DILIonset), (2) time to peak values for total bilirubin, (3) timeto recovery (peak to 50% reduction in total bilirubin),(4) DILIN severity score, and (5) outcomes (hospitali-zation, liver-related death, and transplant).

The data analyses were performed at the Duke Clin-ical Research Institute, the data coordinating center ofthe DILIN. SAS v 9.2 (SAS Institute Inc, Cary, NC) wasused to aid all analyses. Continuous variables weresummarized with mean and standard deviation or me-dian and interquartile range. Categorical data wereexpressed as counts and proportions. The Fisher exacttest was used to analyze differences between groups forthe categorical data, and differences between groups forcontinuous variables were assessed with a nonpara-metric test (Wilcoxon test for 2 groups or Kruskal–Wallisone-way analysis of variance for more than 2 groups).P values less than .05 were considered statistically sig-nificant. There was no adjustment made for multiplicityof comparisons. The authors had access to the study dataand have reviewed and approved the final manuscript.

Results

A total of 1083 DILI subjects enrolled up to March2012 were screened to identify 383 cases that met theeligibility criteria (Figure 1). A total of 107 drugs wereresponsible for these DILI episodes. We could not obtainthe details of hepatic metabolism on 3 (febuxostat,investigational drug, ranitidine) and the specifics ofBDDCS in 8 drugs (benzonatate, deferasirox, flavocoxid,investigational drug, phentermine, regorafenib, salsalate,and thiamazole). The mean age of the patients in thestudy cohort was 50 � 18 years, with 63% women and79% white.

- 2014 Medication Properties and the Characteristics of DILI 3

Relationship Between Drug Dose and Drug-Induced Liver Injury Phenotype and Outcomes

Episodes of DILI due to drugs with daily dosage�50 mg had significantly shorter latency period ascompared with those who received �49 mg per day (38[20–96] vs 56 [30–208] days, P¼ .03). The pattern of liverinjury (cholestatic/mixed/hepatocellular) at onset wasalso significantly different between 2 dosage groups(P ¼ .04), with a greater proportion of cholestatic liverinjury in �49 mg per day. However, no other significantdifferences were noted between the groupswith regard topeak values of liver biochemistries, time to peak, and timeto recovery (Table 1). The DILIN severity scores were notsignificantly different on the basis of dose classification(P ¼ .3). Similarly, clinical outcomes such as hospitaliza-tion rate, liver-related death, and liver transplantationrate were not different between 2 groups (Table 1).

Relationship Between Drug Metabolismand Drug-Induced Liver Injury Phenotypeand Outcomes

In the current cohort, a higher number of cases hadDILI from drugs that underwent significant hepaticmetabolism (n ¼ 305) compared with those withouthepatic metabolism (n ¼ 71). The 2 groups did not differwith regard to variables such as latency, pattern of liverinjury, time to peak, and time to recovery. The DILINseverity scores were not significantly different on thebasis of hepatic metabolism classification (P ¼ .3). Out-comes such as hospitalization rate (55% vs 44%, P ¼ .3),liver-related death (2% vs 5.6%, P ¼ .7), and livertransplant (3.3% vs 1.4%, P ¼ .7) were not differentbetween 2 groups (Table 1).

Table 1. Relationship Between Drug Characteristics and DILI P

Latency(d)

Pattern of liverinjury at onset(C/M/HC, %)

Time to peak(d)

Ti

Daily dose�50 mg (n ¼ 324) 38 (20–96) 21/25/54 8 (2–15)�49 mg (n ¼ 53) 56 (30–208) 32/11/57 10 (4–22)

(P ¼ .03) (P ¼ .040) (P ¼ .1)Hepatic metabolism

Yes (n ¼ 305) 42 (25–100) 23/24/53 8 (2–16)No (n ¼ 71) 34 (16–101) 23/21/56 7 (2–17)

(P ¼ .3) (P ¼ .9) (P ¼ .9)BDDCS

Class 1 (n ¼ 118) 81 (43–214) 12/12/76 7 (2–14)Class 2 (n ¼ 96) 35 (43–214) 25/29/46 10 (3–23)Class 3 (n ¼ 112) 28 (16–44) 32/29/39 7 (1–13)Class 4 (n ¼ 38) 38 (15–568) 26/24/50 11 (3–22)

(P < .001) (P < .001) (P ¼ .1)

NOTE. All values are expressed as median (interquartile range) unless otherwise einitial blood test by using multiples of ULN for both values. R � 2 cholestatic (C) ppattern of DILI.

Relationship of Biopharmaceutics DrugDisposition Classification System ClassificationWith Drug-Induced Liver Injury Phenotypeand Outcomes

A total of 99 drugs belonging to BDDCS classes 1–4were responsible for these DILI episodes (Table 2). Thedistribution of cases belonging to each BDDCS group wasdifferent, with underrepresentation of DILI cases causedby BDDCS class 4 agents. Four compounds (ciprofloxacin,nitrofurantoin, cefdinir, and levonorgestrel/ethinyles-tradiol) accounted for all cases of DILI belonging toBDDCS class 4 (Table 2). DILI cases from drugs belongingto class 4 differed from other groups in having a sig-nificantly lower percentage of jaundiced patients(P < .001), absolute eosinophilia >500/mL (P ¼ .008)with lower peak ALT (P ¼ .001) and aspartate amino-transferase (P < .001).The groups differed significantlywith regard to latency period, with class 1 having thelongest latency period (P < .001) (Table 1). The patternof liver injury at onset was significantly different be-tween the groups (P < .001), with class 1 having thehighest proportion of hepatocellular pattern (76%).However, no significant differences were noted betweenthe groups with regard to time to peak total bilirubin andtime to recovery. Hospitalization rate was significantlydifferent among the groups (P ¼ .01), with class 4 havingthe lowest rates at 45% (Table 1). However, outcomessuch as liver-related death (P ¼ .1) and transplantationrate (P ¼ .3) were not different (Table 1).

Discussion

Although prior studies have suggested a relationshipbetween drug characteristics and their propensity to

henotype and Outcomes (N ¼ 383)

me to recovery(d)

Hospitalization(%)

Liver-related

death (%) Transplant (%)

31 (13–51) 52 2.8 3.436 (12–70) 62 1.9 0(P ¼ .3) (P ¼ .2) (P ¼ 1.0) (P ¼ .4)

33 (13–56) 55 2.0 3.331 (12–45) 47 5.6 1.4(P ¼ .3) (P ¼ .2) (P ¼ .1) (P ¼ .7)

27 (12–44) 53 2.5 5.131 (10–55) 68 2.1 1.038 (17–53) 46 0.9 1.825 (7–49) 45 7.9 2.6(P ¼ .1) (P ¼ .01) (P ¼ .1) (P ¼ .3)

xpressed. R ratio was calculated by dividing ALT by alkaline phosphatase fromattern of DILI, 2 < R < 5 mixed (M) pattern of DILI, or R >5: hepatocellular (HC)

Table 2. List of Drugs Implicated in the DILI Episode

Class 1 (high solubility,extensive metabolism)

(n ¼ 118)

Class 2 (low solubility,extensive metabolism)

(n ¼ 99)

Class 3 (high solubility,poor metabolism)

(n ¼ 115)

Class 4 (low solubility,poor metabolism)

(n ¼ 40)

Alfuzosin (n ¼ 1)Aliskiren (n ¼ 1)Atomoxetine (n ¼ 3)Azathioprine (n ¼ 4)Bupropion (n ¼ 3)Clindamycin (n ¼ 1)Cyclophosphamide (n ¼ 2)Dantrolene (n ¼ 1)Diclofenac (n ¼ 9)a

Duloxetine (n ¼ 6)Escitalopram (n ¼ 1)Estradiol (n ¼ 1)Ethosuximide (n ¼ 1)Fluoxetine (n ¼ 1)Fluvastatin (n ¼ 1)Hydralazine (n ¼ 4)a

Isoniazid (n ¼ 32)Labetalol (n ¼ 1)Letrozole (n ¼ 1)Methylphenidate (n ¼ 1)Minocycline (n ¼ 22)a

Nicotinic acid (n ¼ 2)Promethazine (n ¼ 1)Propylthiouracil (n ¼ 2)Pyrazinamide (n ¼ 1)Quetiapine (n ¼ 1)Sertraline (n ¼ 1)Tamoxifen (n ¼ 4)Valaciclovir (n ¼ 1)Valproic acid (n ¼ 5)Verapamil (n ¼ 3)

Acitretin (n ¼ 3)Allopurinol (n ¼ 2)Amiodarone (n ¼ 4)Atorvastatin (n ¼ 4)a

Bosentan (n ¼ 1)Carbamazepine (n ¼ 2)Celecoxib (n ¼ 2)Disulfiram (n ¼ 2)Dronedarone (n ¼ 2)Drospirenone w/ethinylestradiol (n¼ 1)Etodolac (n ¼ 2)Ezetimibe w/simvastatin (n ¼ 2)Febuxostat (n ¼ 1)Fenofibrate (n ¼ 2)Imatinib (n ¼ 3)Itraconazole (n ¼ 1)Lamotrigine (n ¼ 5)Lansoprazole (n ¼ 1)Lapatinib (n ¼ 1)Leflunomide (n ¼ 2)Meloxicam (n ¼ 2)Mercaptopurine (n ¼ 5)Mesalazine (n ¼ 1)Metaxalone (n ¼ 1)Montelukast (n ¼ 2)Nevirapine (n ¼ 2)Olanzapine (n ¼ 1)Oxaprozin (n ¼ 1)Phenytoin (n ¼ 8)Prasugrel (n ¼ 1)Quinapril (n ¼ 1)Simvastatin (n ¼ 4)Sulfasalazine (n ¼ 1)Sulfamethoxazole w/trimethoprim

(n ¼ 12)Tacrolimus (n ¼ 1)Telithromycin (n ¼ 5)Temozolomide (n ¼ 1)Terbinafine (n ¼ 5)Thalidomide (n ¼ 1)Voriconazole (n ¼ 1)

Amoxicillin (n ¼ 6)Amoxicillin w/clavulanic acid (n¼ 61)Atenolol (n ¼ 1)Azithromycin (n ¼ 4)Benazepril w/HCTZ (n ¼ 1)Cefaclor (n ¼ 1)Cefadroxil (n ¼ 1)Cefalexin (n ¼ 1)Cefuroxime (n ¼ 1)Doxycycline (n ¼ 4)Erythromycin (n ¼ 2)Fluconazole (n ¼ 2)Gabapentin (n ¼ 1)Levofloxacin (n ¼ 4)Lisinopril (n ¼ 2)Methotrexate (n ¼ 1)Methyldopa (n ¼ 8)a

Metoclopramide (n ¼ 1)Moxifloxacin (n ¼ 2)Penicillamine (n ¼ 1)Pravastatin (n ¼ 1)Ranitidine (n ¼ 2)Rosuvastatin (n ¼ 4)Topiramate (n ¼ 3)

Cefdinir (n ¼ 1)Ciprofloxacin (n ¼ 11)Levonorgestrel w/ethinylestradiol

(n ¼ 1)Nitrofurantoin (n ¼ 27)a

NOTE. Drugs arranged as per BDDCS class. A total of 99 drugs with BDDCS class of 1, 2, 3, or 4 were responsible for 372 cases of DILI.Solubility refers to solubility in water. High solubility compound at highest marketed dose strength would be soluble in 250 mL water over the pH range of 1–7.5at 37�C.aHas been associated with autoimmune DILI phenotype.

4 Vuppalanchi et al Clinical Gastroenterology and Hepatology Vol. -, No. -

cause DILI, data regarding drug characteristics and DILIphenotype are lacking. In the current study, we lever-aged the unique features of DILIN database to examinethe relationships among drug characteristics and DILIphenotype and outcomes. The current observations ofshorter latency with daily dosage �50 mg and a higherpeak ALT with drugs that undergo hepatic metabolismdemonstrate the importance of the intrinsic drugproperties as a potential factor for the development of aDILI event. These findings also suggest that a thresholdlevel of parent drug, metabolite, and/or adducts is a

prerequisite for the DILI event. However, lack ofany significant differences in the clinical outcomessupports the current understanding that the mechanismof injury and subsequent recovery may be predomi-nantly mediated by the host immune response or hostsusceptibility due to genetic or acquired inability todetoxify hepatotoxic drug intermediates.1,14,15 Furtherstudies are needed to examine how these propertiesinfluence mechanisms involved in DILI such as CYPbioactivation reactions, covalent binding, or mitochon-drial toxicity.18,19

- 2014 Medication Properties and the Characteristics of DILI 5

Our examination of the relationship among drugproperties such as aqueous solubility/gastrointestinalpermeability by using BDDCS classification20,21 and DILIphenotype or outcomes led to some novel observations.First, drugs belonging to class 4 (cefdinir, ciprofloxacin,nitrofurantoin, and levonorgestrel with ethinylestradiol)were incriminated in lower percentage of DILI patients,possibly because BDDCS class 4 drugs are underrepre-sented in orally approved drugs. These BDDCS class 4drugs were associated with significantly lower peakserum aminotransferase levels presumably because oftheir low permeability and solubility resulting in reduceduptake and lower hepatic biotransformation. A detailedreview of these cases showed instances where prolongedexposure of the drug occurred (in some cases of nitro-furantoin, exposure occurred for several years) beforethe DILI episode occurred. Second, the current studyidentified 81% of the drugs causing DILI episodes asundergoing extensive hepatic metabolism, suggestingthat a reactive metabolite intermediate rather than thenative drug may be responsible for the development of aDILI event, possibly through the formation of adductcomplexes acting as neoantigens.

Autoimmune DILI is often associated with presence ofserum autoantibodies directed against hepatic cyto-chrome P450s or other hepatic proteins.22 The covalentbinding of the drug or reactive metabolite to livermicrosomal proteins results in formation of a neoantigenwith subsequent injury from immune response.23,24 It isthus plausible that drugs that belong to BDCSS classes1 and 2 (extensive metabolism) may be at increased riskof causing autoimmune DILI. In the current study, weobserved that several drugs that have previously beenimplicated to cause autoimmune DILI22 belonged toBDCSS classes 1 (hydralazine, minocycline, propylth-iouracil, methylphenidate, and diclofenac) and 2 (ator-vastatin, imatinib, fenofibrate, and terbinafine) (Table 2).However, 2 drugs with definite association with auto-immune DILI, ie, methyldopa (BDCSS class 3) andnitrofurantoin (BDCSS class 4), diverge from this pattern,suggesting additional role of host-related factors as beingcritical for a DILI event.

Last, the unique strengths of the DILIN cohort suchas the ability to examine DILI cases with strong cau-sality (1, 2, or 3) and availability of prospectivelycaptured DILI phenotype and outcomes allowed us toperform the current study. The pitfalls common in theliterature such as DILI case definition and causalityadjudication are minimized in the DILIN prospectivestudy because of its study design.16,17,25 The currentstudy failed to find any significant differences in clinicoutcomes based on daily dosage in contrast to theearlier study by Lammert et al,6 who evaluated theSwedish registry. It is likely that these differences couldbe related to the DILIN registry prospective protocoland the current study design that included only DILIcases from single agent and with causality limited to thescores of 1, 2, or 3. It is also possible that lack of

relationship between drug metabolism as well as BDDCSclassification with clinical outcomes might be due tolack of adequate sample size.

In summary, drug characteristics such as daily dosage�50 mg and BDDCS class 1 are related to select aspectsof DILI phenotype. However, the severity of liver diseaseand outcomes are not associated with drug characteris-tics. We speculate that DILI outcomes may be morerelated to host response rather than drug characteristics.

References

1. Tujios S, Fontana RJ. Mechanisms of drug-induced liver injury:

from bedside to bench. Nat Rev Gastroenterol Hepatol 2011;8:202–211.

2. Monshi MM, Faulkner L, Gibson A, et al. Human leukocyte an-tigen (HLA)-B*57:01-restricted activation of drug-specific T cellsprovides the immunological basis for flucloxacillin-induced liverinjury. Hepatology 2013;57:727–739.

3. Urban TJ, Shen Y, Stolz A, et al. Limited contribution of commongenetic variants to risk for liver injury due to a variety of drugs.Pharmacogenet Genomics 2012;22:784–795.

4. Lammert C, Bjornsson E, Niklasson A, et al. Oral medicationswith significant hepatic metabolism at higher risk for hepaticadverse events. Hepatology 2010;51:615–620.

5. Chen M, Borlak J, Tong W. High lipophilicity and high daily doseof oral medications are associated with significant risk for drug-induced liver injury. Hepatology 2013;58:388–396.

6. Lammert C, Einarsson S, Saha C, et al. Relationship betweendaily dose of oral medications and idiosyncratic drug-inducedliver injury: search for signals. Hepatology 2008;47:2003–2009.

7. Hughes JD, Blagg J, Price DA, et al. Physiochemical drugproperties associated with in vivo toxicological outcomes. Bio-org Med Chem Lett 2008;18:4872–4875.

8. Uetrecht J. Idiosyncratic drug reactions: current understanding.Annu Rev Pharmacol Toxicol 2007;47:513–539.

9. Kaplowitz N. Avoiding idiosyncratic DILI: two is better than one.Hepatology 2013;58:15–17.

10. Chen ML, Yu L. The use of drug metabolism for prediction ofintestinal permeability (dagger). Mol Pharmacol 2009;6:74–81.

11. Benet LZ, Broccatelli F, Oprea TI. BDDCS applied to over 900drugs. AAPS J 2011;13:519–547.

12. Benet LZ. The role of BCS (biopharmaceutics classificationsystem) and BDDCS (biopharmaceutics drug disposition clas-sification system) in drug development. J Pharm Sci 2013;102:34–42.

13. Benet LZ, Amidon GL, Barends DM, et al. The use of BDDCS inclassifying the permeability of marketed drugs. Pharm Res 2008;25:483–488.

14. Lucena MI, Molokhia M, Shen Y, et al. Susceptibility toamoxicillin-clavulanate-induced liver injury is influenced bymultiple HLA class I and II alleles. Gastroenterology 2011;141:338–347.

15. Daly AK, Donaldson PT, Bhatnagar P, et al. HLA-B*5701 ge-notype is a major determinant of drug-induced liver injury due toflucloxacillin. Nat Genet 2009;41:816–819.

16. Fontana RJ, Watkins PB, Bonkovsky HL, et al. Drug-InducedLiver Injury Network (DILIN) prospective study: rationale, designand conduct. Drug Saf 2009;32:55–68.

17. Chalasani N, Fontana RJ, Bonkovsky HL, et al. Causes, clinicalfeatures, and outcomes from a prospective study of drug-

6 Vuppalanchi et al Clinical Gastroenterology and Hepatology Vol. -, No. -

induced liver injury in the United States. Gastroenterology 2008;135:1924–1934.

18. Metushi IG, Cai P, Zhu X, et al. A fresh look at the mechanism ofisoniazid-induced hepatotoxicity. Clin Pharmacol Ther 2011;89:911–914.

19. Alfirevic A, Pirmohamed M. Predictive genetic testing for drug-induced liver injury: considerations of clinical utility. Clin Phar-macol Ther 2012;92:376–380.

20. Benet LZ. Predicting drug disposition via application of a Bio-pharmaceutics Drug Disposition Classification System. BasicClin Pharmacol Toxicol 2010;106:162–167.

21. Wu CY, Benet LZ. Predicting drug disposition via application ofBCS: transport/absorption/elimination interplay and develop-ment of a biopharmaceutics drug disposition classificationsystem. Pharm Res 2005;22:11–23.

22. Czaja AJ. Drug-induced autoimmune-like hepatitis. Dig Dis Sci2011;56:958–976.

23. Beaune PH, Bourdi M. Autoantibodies against cytochromes P-450 in drug-induced autoimmune hepatitis. Ann N Y Acad Sci1993;685:641–645.

24. Beaune PH, Lecoeur S, Bourdi M, et al. Anti-cytochrome P450autoantibodies in drug-induced disease. Eur J Haematol Suppl1996;60:89–92.

25. Aithal GP, Watkins PB, Andrade RJ, et al. Case definition andphenotype standardization in drug-induced liver injury. ClinPharmacol Ther 2011;89:806–815.

Reprint requestsAddress requests for reprints to: Naga Chalasani, MD, FACG, Division ofGastroenterology and Hepatology, Indiana University School of Medicine, 1050Wishard Boulevard, RG 4100, Indianapolis, Indiana 46202. e-mail: nchalasa@iu.edu; fax: (317) 278-1949.

Conflicts of interestThese authors disclose the following: Dr Vuppalanchi received compensationfor consulting related to drug hepatotoxicity from BMS. Dr Reddy served on theadvisory boards for Merck, Genentech-Roche, Gilead, BMS, Idenix, Vertex,Janssen, and Abbvie and has conducted research with Merck, Genentech-Roche, Gilead, BMS, Vertex, Janssen, and Abbvie. Dr Fontana receivedgrant support from Gilead and Vertex and served as a paid consultant toTibotec, Merck, and GSK. Dr Chalasani served as a paid consultant to Abbvie,Salix, BMS, Aegerion, Lilly, and Merck in the past 12 months and receivedgrant support from Intercept, Cumberland, Gilead, Takeda, and Enterome. Theremaining authors disclose no conflicts.

FundingThe U.S. Drug-Induced Liver Injury Network (DILIN) is supported by the NationalInstitute of Diabetes and Digestive and Kidney Diseases under the followingcooperative agreements: 1U01DK065021, U01DK065193, 1U01DK065201,1U01DK065193, 1U01DK065184, 1U01DK065211, 1U01DK065238, and1U01DK065176.