Drug-induced liver injury: Summary of a single topic clinical research conference


  • Potential conflict of interest: Dr. Watkins is on the advisory board for GlaxoSmithKline and is a consultant for Actelion, Critical Therapeutics, CSL, Encysive, Hoffmann-LaRoche, Valiant, VIA.


Idiosyncratic drug induced liver injury (DILI) remains poorly understood. It is assumed that the affected individuals possess a rare combination of genetic and non genetic factors that, if identified, would greatly improve understanding of the underlying mechanisms. This single topic conference brought together basic scientists, translational investigators, and clinicians with an interest in DILI. The goal was to define high priority areas of investigation that will soon be made possible by The Drug-Induced Liver Injury Network (DILIN). Since 2004 DILIN has been collecting clinical data, genomic DNA and some tissues from patients who have experienced bone fide DILI. The presentations spanned many different areas of DILI, and included novel data concerning mechanisms of hepatotoxicity, new “omics” approaches, and the challenges of improving causation assessment. (HEPATOLOGY 2006;43:618–631.)

Drug-induced liver injury (DILI) is a major problem for affected patients as well as for their health care providers. Not only is it the primary cause for acute liver failure in the United States, it is also a strong consideration in differential diagnosis when abnormal liver-related chemistries are identified in persons with minimal or no symptoms. There is presently no specific diagnostic test for DILI, or a means of confidently singling out an implicated drug among many being received, so that the physician must empirically decide which, if any, treatments to discontinue or substitute when DILI is suspected. DILI is also a serious problem for the pharmaceutical industry. Indeed, safety and toxicology issues are now major reasons for failure of drugs in clinical trials. Moreover, DILI is also the most common factor that compels regulatory actions due to drugs, such as failure of approval or restriction of indications of the drug, as well as the mailing of Dear Doctor letters.1 As drugs are increasingly designed to potently interact with known molecular targets, it will become possible to genetically determine ahead of time who among a patient population will likely respond to the drug. If likely responders can be identified, it should be possible to show efficacy in smaller and less costly clinical trials. Nevertheless, demonstration of drug safety will continue to require large clinical trials until the propensity to adverse events is eliminated from the molecule, or those susceptible to the toxicity can be identified before they are treated.

Among the several manifestations of drug toxicity, the most problematic is the idiosyncratic form. This term derives from Greek words meaning “mixture of characteristics” and in the context of drug toxicity, refers to the combination of genetic and non-genetic factors that make a rare patient susceptible to drug injury.2 There are currently no validated animal models for studying idiosyncratic DILI. Indeed, humans as a group are not appropriate subjects for study since they are only occasionally susceptible to the toxicity. Therefore, the best models to study idiosyncratic DILI are the rare patients who have already experienced the phenomenon.

Accordingly, the Drug-Induced Liver Injury Network (DILIN) was established in 2003 with the aim of studying the problem of hepatotoxicity3 (http://dilin.dcri.duke.edu). Supported by the National Institute of Diabetes and Digestive and Kidney Diseases, the National Institutes of Health, the goals of the network are to: (1) establish a registry of well characterized patients who have experienced clinically significant DILI and agree to be contacted and offered participation in future studies, (2) develop improved tools to aid in the diagnosis of DILI, and (3) provide to the scientific community genomic DNA, serum and immortalized lymphocytes obtained from the patients in the registry for future research purposes.

Inspired by this network study, this single topic conference was designed to bring together basic scientists, translational investigators, and clinicians with an interest in DILI. The presentations and lively discussions helped define important issues for future investigation utilizing the resources of DILIN.

Geoffrey C. Farrell: State-of-the-Art Lecture: Problems of Drug-Induced Liver Injury.

The diagnosis of liver injury can be considered when abnormalities of liver-related biochemical tests are identified, but liver disease requires knowledge of pathology. DILI can be classified as hepatocellular, cholestatic or mixed injury, using criteria established by the Council for International Organisation of Medical Sciences (CIOMS).4 This classification is aided by calculating an R ratio (alanine aminotransferase [ALT] divided by the alkaline phosphatase [AP]). A major diagnostic dilemma is that elevations in biochemical tests, such as the ALT or γ-glutamyl transpeptidase (GGT), can represent an adaptive response to drugs and may not be indicative of true injury. The incidence of DILI varies depending on the drug, ranging from 5–20/1,000 exposed to INH, chlorpromazine and PTU to 1–10/1 million who receive minocycline, statins and psychotropic agents. It is therefore not surprising that DILI is often not considered as a potential diagnosis when liver dysfunction is identified. One study of hospitalized patients showed that DILI events were not reported in the discharge summary in over 50% of the cases.5 The classes of drugs most frequently causing severe DILI appear to be anti-infectives and non-steroidal anti-inflammatory drugs (NSAIDS).6–9 Potential risk factors for DILI include a previous history of drug reactions, the very old and very young, female sex (especially for acute liver failure [ALF]), multiple drug therapy, immunological disorders (HIV/AIDS, SLE), preexisting liver disease and poor nutrition.10, 11 The diagnosis of DILI can be challenging because some drugs may occasionally lead to more than one presentation of DILI. Liver biopsy is rarely diagnostic, but can be supportive (based on the presence of eosinophils, granulomas, microvesicular steatosis, or zonal necrosis). Herbal medications are a common cause of ALT elevations and should always be sought as a potential cause for DILI.11, 12 The prognosis of DILI is generally good if the drug treatment is stopped11; chronic liver disease resulting from DILI is rare. Two recent studies have confirmed the original observation of Hyman J. Zimmerman (now termed Hy's rule) that patients with hepatocellular jaundice have ≈10% mortality even if the offending drug is discontinued.7, 8

William M. Lee: DILI and Acute Liver Failure.

More than half of all instances of ALF identified in the multi-center Acute Liver Failure Study Group (ALFSG) registry from 25 U.S. sites are due to DILI; 3/4 of the DILI cases are accounted for by acetaminophen.13–15 Disturbingly, the proportion of acetaminophen cases among all ALF cases has increased steadily since 1998, with acetaminophen accounting for 51% of all cases over the last 2 years.16 More than half of the cases of acetaminophen-induced ALF involved persons who had not intended to harm themselves (“unintentional” cases), among whom the use of narcotic/acetaminophen combinations was over-represented.17 Acetaminophen adducts have recently been identified as a highly specific and sensitive assay for acetaminophen hepatotoxicity and remain positive up to nine days after ingestion (unpublished observations). Importantly, these adducts have been detected in the serum of 20% of persons who were categorized as “indeterminate”, indicating that acetaminophen is even more commonly a cause for ALF than had previously been recognized.18 Among the non- acetaminophen cases of ALF due to DILI, antibiotics, particularly anti-tuberculous drugs, are the major responsible class. Four cases of ALF attributed to statins are in the registry. However, closer examination of the original medical records revealed that other drugs were more likely the cause of ALF in 3 of these cases. The ALFSG is presently conducting a randomized, double blind, controlled trial of N-acetylcysteine given IV in persons with non-acetaminophen ALF. The group plans to partner with DILIN to help increase enrollment.

Eve A. Roberts: DILI and Children

DILI is relatively uncommon in children, but it accounts for ≈25% of childhood ALF. Drugs which are prominent among those causing childhood DILI include acetaminophen; the group of anti-epileptics including phenytoin, carbamazepine, and phenobarbital; valproic acid; sulfonamides; pemoline (recently withdrawn from the market); minocycline; and most-anti-neoplastic drugs Where drug therapy plays a key role—epilepsy and childhood neoplasia—drug hepatotoxicity is rather frequent. Children can also develop DILI from herbals and from recreational drugs. As with adults, almost every major type of hepatic pathology can occur due to DILI.

In children, the diagnosis of DILI requires extreme vigilance and meticulous clinical investigation. In a recent study, the most commonly implicated drugs in fatal DILI were anticonvulsants and anti-neoplastics.19 Phenytoin has been commonly associated with DILI in children.20 Young children differ from adults in the activities of certain hepatic drug-metabolizing enzymes. For example, children are deficient in caffeine and theophylline metabolism during the first year of life, and this appears to reflect reduced expression of CYP1A2.21 Since children tend to receive fewer drugs than adults and have less exposure to relevant environmental factors (smoking, ethanol, hormonal factors), it should be easier to study DILI and identify genetic factors conferring susceptibility. This has been attempted with phenytoin, carbamazepine, phenobarbital, and sulfonamides relating to the multisystemic “hypersensitivity” syndrome in which hepatitis is prominent.22, 23 Congenital mitochondrial abnormalities may predispose to valproic acid hepatotoxicity.24 Although some data suggest that children may be more resistant than adults to DILI from acetaminophen, the drug remains a major cause of liver injury and liver failure in children. There are many different formulations of children's acetaminophen, often confusing to parents. Clinical findings may be nondescript in small children who develop severe acetaminophen hepatotoxicity from short-term inappropriate repetitive dosing with this drug, and the Rumack normogram must be interpreted cautiously.25 The recreational drug “Ecstasy” can cause liver failure and should be a consideration in teenagers presenting with hepatocellular injury.

Robert J. Fontana: Idiosyncratic Hepatocellular injury

DILI can be characterized biochemically and clinically as predominantly hepatocellular when the R ratio at presentation exceeds 5, cholestatic when R <2, or mixed when 5> R >2.4.With many drugs, any or all of these biochemical injury patterns may be seen but most have a “signature” profile with a characteristic pattern of liver injury and range of time to onset. Nevertheless, since DILI is a diagnosis of exclusion, all other conditions that present as hepatocellular injury must first be excluded. A review of drugs that have undergone regulatory actions in the last decade reveals that the majority have been associated with severe hepatocellular injury.2 Most cases of hepatocellular DILI have their onset within 1 year of starting the suspect medication. Clinically significant hepatocellular DILI is usually defined by a serum ALT value >3 times the upper limits of normal. Studies of patients with jaundice from severe acute hepatocellular DILI demonstrate a short-term mortality as high as ≈10% (Hy's Law).7, 26 In addition, patients with fulminant hepatitis from DILI have only a 20% likelihood of survival with supportive care.14 Therefore, any patient with hepatocellular DILI and jaundice who develops encephalopathy or coagulopathy should be referred for liver transplantation evaluation. The mechanisms responsible for hepatocellular DILI are largely unknown but may be a result of host metabolic idiosyncracy.27 Implicated but yet uncertain mechanisms include mitochondrial toxicity, oxidative stress, hepatic steatosis and hepatocyte necrosis or apoptosis.

F. Peter Guengerich: The Role of Bioactivation in Drug-Induced Liver Disease

The concept of bioactivation goes back to the pioneering work of J. A. and E. C. Miller in the 1940s.28 The focus on protein adducts of drugs as an explanation for drug toxicity developed in 1973 with the classic studies on acetaminophen by Brodie and Gillette.29 The cytochromes P450 are the major enzymes involved in the activation of drugs to reactive species, although other drug-metabolizing enzymes can contribute to the activation of drugs. Another factor is the balance between enzymatic activation and detoxication.30 Exactly what fraction of drug problems are related to bioactivation and metabolism is difficult to establish, but it was recently reported that thirteen of 22 (62%) of drugs withdrawn from the market for hepatotoxicity, or with black box warnings for hepatotoxicity, have been shown to produce reactive metabolites.31

One strategy for reducing risk of toxicity in drug development is to avoid certain chemical moieties (such as furans and thiophenes) that have the potential to generate reactive species in the context of medicinal chemistry.32 Another focus of activity is in deployment of methods to screen for potential toxicity early in the drug development process.32, 33 Approaches in use include in silico strategies, transcriptonomics, proteomics, metabolomics, and higher-throughput screens for covalent binding.33 The overall goal is to define better predictive markers, both in vitro and in vivo.

Michael H. Court: The Role of Detoxification

The extent of detoxification of the parent drug and/or its reactive metabolite is likely to represent an important determinant of individual risk for drug-induced hepatotoxicity. In contrast with the cytochrome P450 (CYP) enzymes, which are most commonly responsible for reactive metabolite formation, the phase 2 (or conjugation) enzymes are more frequently involved in detoxification reactions.34 Acetaminophen (APAP) is detoxified primarily through glucuronidation (by UDP-glucuronosyltransferases; UGTs) and secondarily through sulfation (by sulfotransferases; SULTs). APAP glucuronidation is mediated by 3 UGTs with differing kinetic properties, including high affinity (UGT1A6), intermediate affinity (UGT1A1) and low affinity (UGT1A9) isoforms [35]. Sulfation is primarily via the high affinity SULT1A1.36N-acetyl-p-benzoquinone imine (NAPQI), the main reactive metabolite of APAP, is detoxified through conjugation with glutathione.34 Although glutathione S-transferases (GSTs) are involved to some extent, the majority of NAPQI conjugation with glutathione may occur spontaneously.34 Both conjugation enzymes and CYPs are coordinately regulated by a variety of transcription factors including the xenobiotic responsive nuclear receptors PXR, CAR and AhR.37 However, a unique characteristic of conjugation enzymes is induction via the anti-oxidant response pathway mediated by the transcription factor NRF-2. NRF-2 knockout mice show significantly enhanced sensitivity to APAP hepatotoxicity associated with reduced expression of UGTs, SULTs, GSTs, as well as the cofactor synthesis enzymes.38, 39 Although minimally studied, polymorphism of the genes encoding each of these proteins could explain interindividual differences in susceptibility for toxicity. Some studies suggest that people with Gilbert's syndrome (1%-7% of population) caused by a TATA box mutation of the UGT1A1 gene may be at increased risk for APAP toxicity.40 Preliminary studies using DNA samples from people that developed acute liver failure as a result of APAP usage show a significant association with 3 non-synonymous polymorphisms with established functional relevance located in the UGT1A6 gene (unpublished observations).

Robert A. Roth: The Role of Inflammation

Evidence from animal studies indicates that an inflammatory stress can render the liver sensitive to otherwise nontoxic doses of many xenobiotic agents, including numerous drugs. For example, trovafloxacin (TVX) causes idiosyncratic adverse drug reactions (ADRs) in people, and rats given TVX experienced hepatotoxicity when co-exposed to a small dose of bacterial lipopolysaccharide (LPS) that causes a modest inflammatory reaction (unpublished observation). In contrast, no hepatotoxicity is seen upon co-treatment with LPS and levofloxacin, a drug which does not cause human idiosyncratic ADRs. Similarly, ranitidine causes liver toxicity in rats when it is co-administered with LPS, but famotidine does not.41, 42 Studies with ranitidine and LPS have suggested that inflammation can enhance susceptibility to DILI by several mechanisms, including increasing fibrin deposition in liver sinusoids causing tissue hypoxia, increasing plasminogen activator inhibitor-1 expression and by causing expression of chemokines that recruit and activate neutrophils.42–44 Other drugs whose toxicity is potentiated by LPS include halothane, chlorpromazine, acetaminophen and diclofenac.45 These observations raise the possibility that idiosyncratic DILI may be triggered by an inflammatory episode in the host.

Neil Kaplowitz: The Role of Innate Immunity in Acetaminophen Hepatotoxicity

Glutathione depletion and covalent binding are prerequisites for acetaminophen hepatotoxicity. However, the threshold for cell death can be modulated by intrahepatocyte signal transduction and transcription factors for protective or injurious pathways such as NRF-2 and JNK.46 In addition, extrahepatocyte factors also determine whether injured hepatocytes die or recover, by a mechanism that involves the participation of the innate immune response. It is presumed that the initiation of cellular injury triggers an innate immune response. The activation and infiltration of NK/NKT cells plays a pivotal role as depletion of these cells protects against acetaminophen.47 One key feature of these cells is the production of IFNγ, a pro-inflammatory cytokine; indeed, NK/NKT cells are the predominant source of IFNγ.47, 48 Furthermore, IFNγ null-mice are protected against acetaminophen.47, 48 Downstream of IFNγ, a variety of cytokines and chemokines are expressed which promote neutrophil infiltration. Although an area of ongoing controversy, immunodepletion of neutrophils can protect against acetaminophen; furthermore ICAM-1 null mice are protected against acetaminophen. Interestingly, Fas and FasL deficient mice are also protected against acetaminophen47 and silencing of Fas has also been shown to decrease toxicity.49 Although apoptosis is not a prominent feature, it is not clear if the Fas system is involved in acetaminophen-induced necrosis (upstream of mitochondria) or is promoting inflammation. At the same time as a pro-inflammatory cascade is activated by the initial “stress” of acetaminophen, a counter-regulatory anti-inflammatory cascade is simultaneously activated, so that the interplay modulates the threshold for injury. Thus, IL-10 and IL-6 null mice are more susceptible to acetaminophen toxicity.50, 51 The murine model of acetaminophen hepatotoxicity provides important clues to increase the understanding of predictable and idiosyncratic drug-induced liver injury. Both in the context of the early events in individual hepatocytes and subsequent events in the intact organ, competing pro- and anti-injurious mechanisms are battling to determine the outcome.

Herbert L. Bonkovsky; Autoimmune DILI: Clinical

Although the etiology and pathogenesis of autoimmune liver disease (AILD) remain uncertain, many clinical observations suggest that, in some subjects, exposure to drugs (chemicals) may trigger immune responses that target especially the liver.52 If not recognized promptly so that the responsible agent can be withdrawn, such reactions can give rise to chronic hepatitis (resembling viral hepatitis) or chronic non-suppurative cholangitis (resembling primary biliary cirrhosis [PBC]). Examples of the former include chronic hepatitis with autoimmune features, triggered by alpha-methyldopa, halothane, hydralazine and other hydrazine-containing drugs, minocycline, nitrofurantoin, and oxyphenisatin.26 Examples of the latter include chronic cholestatic syndromes, sometimes with biochemical, serological, and histological features of PBC, following exposure to chlorpromazine. Several antibiotics, notably penicillins, cephalosporins, and macrolides, may cause severe cholestatic hepatitis, but rarely, if ever, cause self-perpetuating AILD. Discussed in this presentation were clinical vignettes of three patients who had recently developed AILD triggered by statins, including one woman in whom re-challenge led to a second and severe exacerbation.53 These three patients double the reported instances of this association.54–56

Classical features of AILD include strong female preponderance (true for most auto-immune diseases), chronicity for >6 months (albeit arbitrary), the presence of hyperglobulinemia, circulating autoantibodies, extrahepatic clinical autoimmune syndromes, overrepresentation of HLA DR3 or DR4, and a low rate of spontaneous remission.57

Philippe H. Beaune: Autoimmune DILI: Mechanisms

Tienilic acid and dihydralazine are known to trigger, although rarely, a severe form of liver disease with clinical features that are consistent with immune mediated hepatitis; these include the development of certain autoantibodies (anti-LKM2 and anti-LM).58 Studies have demonstrated that each drug is metabolized by a cytochrome P450 (CYP) (CYP2C9 and CYP1A2 for the two drugs, respectively) into a reactive metabolite which binds to the CYP, behaving as a neoantigen. In unpublished studies using dihydralazine, two rat models have been developed that express the neoantigens in the liver, one of which resulted in transient elevation of aminotransferase and liver inflammation, while in the other model, autoantibodies developed in 2/4 rats. However, there was no sign of liver disease in the rats with the autoantibodies, suggesting that the antibodies may have been the result but not a cause of liver injury. It is nevertheless possible that the antibodies may cause injury to the liver in the presence of inflammation.

Eugene R. Schiff: Drug induced-cholestasis: Clinical

Drug-induced cholestasis is characterized by jaundice, pruritis and a disproportionately elevated serum alkaline phosphatase value.59 It can occur as an acute disorder manifesting as canalicular (bland jaundice), hepatocanalicular (cholestatic hepatitis), or ductular (cholangiolar hepatitis) disease, or it may exist as a chronic condition, exemplified by cholangio-destructive (vanishing bile duct syndrome) or a septal sclerosing cholangitis (cholangiosclerotic disease).60, 61 Canalicular cholestasis can rarely progress to the chronic vanishing bile duct syndrome which can be distinguished from primary biliary cirrhosis by the absence of both antimitochondrial antibody (AMA) and hepatic granuloma, and the rarity of its association with the Sicca syndrome.62 In addition, the vanishing bile duct syndrome can exist as a mild form which generally reverses, although progression to cirrhosis is expected with the major form. Numerous antibiotics (e.g., amoxicillin-clavulanic acid, erythromycin estolate, trimethoprim-sulfamethoxazole), psychotropic drugs (e.g., chlorpromazine, carbamazepine, haloperidol), and other drugs (e.g., oral contraceptives, androgens) have been reported to cause one or other form of chronic cholestasis. Treatment usually includes ursodeoxycholic acid.

James L. Boyer: Drug-Induced Cholestasis: Mechanisms

Biliary excretion of drugs and their conjugates is mediated by several members of the ATP-binding cassette superfamily of transporters that are located at the canalicular domain of the hepatocyte. These ATP dependent transport pumps include the multi-drug resistance P-glycoprotein-170 (Mdr1 in animals/MDR1 in man) and the multi-drug resistance polypeptide, (Mrp2 in animals/MRP2 in man). Mdr1/MDR1 transports a number of lipophilic organic cations, many of which are drugs, whereas Mrp2/MRP2 functions as a conjugate export pump to transport many different drug conjugates.63, 64 A number of drugs that are excreted into bile by the bile export pumps have also been implicated in cholestatic liver injury in animals and man.65

The mechanism of drug induced cholestasis is poorly understood. However, in vitro studies utilizing isolated canalicular membrane vesicles from rat and human liver suggest that some of these drugs (cyclosporin A and FK506) inhibit ATP dependent bile salt and leukotriene C4 (LTC4) transport. Other studies using in-vitro systems expressing canalicular transport proteins suggest that cyclosporin A, rifamycin and glibenclamide are capable of directly inhibiting the bile salt export polypeptide, Bsep, the major determinant of bile salt dependent bile formation.66 Bosanten, sulindac and triglitazone may also inhibit hepatic bile salt excretion. Drugs like cyclosporin A, rifampin and glibenclamide, cis-inhibit Bsep mediated bile salt transport whereas other drugs such as estrogen-17β-glucuronide and diclofinac need to be excreted into bile via Mrp2 prior to inhibiting canalicular membrane transporter proteins (trans inhibition).67, 68 Clinical observations suggest that some instances of drug-induced cholestasis occur because of drug/drug interactions, perhaps by binding to a transporter and altering the ability of the polypeptide to excrete the second compound.

Dominique Pessayre: Drug-Induced Steatosis

Drugs can cause deposition of small or large fat droplets in the liver. Microvesicular steatosis generally indicates acute disease. During chronic conditions, small fat vesicles can partly coalesce to give macrovacuolar steatosis, or the association of both microvesicular steatosis and macrovacuolar steatosis. Magnetic resonance imaging (MRI) is a generally reliable means of identifying and quantifying fat in the liver.69

There are two major pathways for the elimination of fat from the liver. Microsomal triglyceride transfer protein (MTP) forms triglyceride-rich VLDL particles, which are secreted, and mitochondria oxidize fatty acids. Drugs can cause steatosis by inhibiting MTP activity and thus hepatic VLDL secretion, and/or by impairing fatty acid oxidation through diverse mechanisms.70, 71 Drugs can sequester coenzyme A (e.g., salicylate, valproate), and/or they can inhibit mitochondrial β-oxidation enzymes (e.g., amineptine, amiodarone, 2-arylpropionate non-steroidal anti-inflammatory drugs, glucocorticoids, perhexiline, tamoxifen, tetracyclines and tianeptine).70 Drugs can also impair mitochondrial structure and function (female sex hormones), or they can decrease the synthesis and stability of mitochondrial transcripts (interferon-alpha).70 Finally, drugs can inhibit the replication of mitochondrial DNA (mtDNA), thus causing progressive mtDNA depletion (e.g., dideoxynucleosides, fialuridine, tamoxifen),72 or they can cause oxidative damage to, and rapid depletion of mtDNA (e.g., alcohol).73

Drugs and several other conditions (e.g., inborn beta-oxidation defects, inborn mitochondrial cytopathies, infections and cytokines, obesity and diabetes, or alcohol abuse) can combine to cause steatosis.70

Steatosis due to drugs and/or other etiologies can be associated with lipid peroxidation, oxidative stress, cytokine induction, and apoptosis.74 Steatosis can thus evolve into steatohepatitis, which can cause liver fibrosis in some patients.74

Laurie D. DeLeve: Sinusoidal Obstruction Syndrome (SOS): Clinical Features and Mechanisms

In North America and Western Europe, SOS (formerly known as VOD) is most commonly diagnosed in patients who have received a myeloablative chemotherapeutic regimen prior to stem cell transplantation for malignancy. SOS is also seen after exposure to various chemotherapeutic agents (gemtuzumab ozogamicin, 5FU/oxaliplatin, cytosine arabinoside, actinomycin, dacarbazine, mithramycin, urethane) and immunosuppressive drugs (azathioprine, 6-thioguanine).75 The diagnosis is usually based on clinical manifestations (hepatomegaly, weight gain, hyperbilirubinemia), but may be confirmed in difficult cases by transvenous liver biopsy.75 Management remains largely supportive and death occurs in 15% to 30%. The challenge for the future will be to use the improved understanding of the pathophysiology to devise novel prophylactic and/or therapeutic strategies.

SOS is initiated by damage to sinusoidal endothelial cells (SEC). Damage to the microcirculation causes ischemic necrosis of hepatocytes followed by sinusoidal fibrosis leading to occlusion of central veins. The best-studied experimental model for SOS is the monocrotaline model. Monocrotaline forms a reactive pyrrole, which binds covalently to F-actin.76 This leads to F-actin depolymerization with rounding up of SEC and to synthesis and activation of matrix metalloproteinase 9 (MMP-9) by SEC.77 F-actin depolymerization promotes rounding up of SEC and the MMP-9 activity on the contralumenal side of the SEC which releases SEC from their tethering in the space of Disse; these changes lead to rounding up of SEC with formation of gaps in the SEC barrier. The swollen SEC partially obstruct the sinusoidal lumen.78 Red blood cells follow the pathway of least resistance and enter the space of Disse through the gaps in the SEC barrier. Blood flowing in the space of Disse dissects off the sinusoidal lining cells, which embolize downstream and obstruct flow.78 Loss of SEC and Kupffer cells leads to a decline in nitrous oxide (NO) levels. Endogenous NO tonically suppresses MMP-9,79, 80 so that loss of producing cells permits further upregulation of MMP-9.81 Inhibition of MMP-977 or treatment with a liver specific NO donor (VPYRRO/NO)81 completely prevents SOS, thereby demonstrating the primacy of the sinusoidal damage to subsequent changes.

Paul B. Watkins: Adaptation to Injury

Drugs capable of inducing severe idiosyncratic DILI often cause more minor and transient ALT elevations in a larger proportion of treated patients. For example, up to 15% of patients treated with isoniazid develop ALT elevations greater than 3 × ULN, but these usually normalize with continued treatment. With the drug tacrine, ALT elevations up to 20 times the ULN have been observed to normalize with continued treatment.82 Some data suggest that these transient ALT elevations represent true liver injury and not just “adaptive changes”. Mice treated with certain halogenated hydrocarbons develop hepatic necrosis after one week of treatment, but this largely resolves with continued exposure.83 There are case reports of patients gradually increasing daily acetaminophen dosing (up to 65 grams/day) without signs of significant liver injury.84 This phenomenon of acetaminophen tolerance has been reproduced in rats and attributed to down-regulation in the cytochromes P450 involved in bioactivation of acetaminophen, and to increased levels of glutathione in the hepatocytes.84 Recent studies have shown that during recovery from acetaminophen toxicity, there are changes in the levels of certain basolateral and canalicular transporters that have been implicated in transport of bile acids and acetaminophen metabolites.85 These changes appear to be due, in part, to the antioxidant response (Nrf-2), the hepatic acute phase reaction (IL-6)86 and gene expression changes previously described in the regenerating liver.87 However, some evidence in man suggests that this adaptation can last for months after an episode of mild DILI, which is unlikely to be explained by these mechanisms.

Timothy J. Davern: Challenges in HIV Patients

The diagnosis of DILI in patients with HIV can be unusually challenging, in part because multiple potentially toxic drugs are used as a cocktail and it is not appropriate to discontinue a single agent empirically due to the rapid development of HIV resistance. The diagnostic challenge also results from the high incidence of underlying liver disease, including chronic viral hepatitis B or C, liver injury due to ethanol and illicit drug abuse, and steatohepatitis due to insulin resistance and dyslipidemia, a consequence of some HIV medications.88 In addition, HIV patients are prone to biliary tract infections and certain cancers (lymphomas and Kaposi's Sarcoma) that can involve the liver. Drugs from all three commonly used classes of antiretroviral agents —nucleoside/nucleotide analog reverse transcriptase inhibitors, protease inhibitors, and non-nucleoside analog reverse transcriptase inhibitors —have been associated with liver injury, particularly in persons with co-existing hepatitis B or C.89 Even when the diagnosis of DILI is likely, it is often difficult to identify which is the culprit among the many drugs taken by these patients. Furthermore, underlying liver disease, particularly viral hepatitis, can flare as HIV drugs effectively restore immune function. This “immune reconstitution” typically occurs in patients with low baseline CD4 counts and high HIV baseline viral titers.90 Studies have suggested that the risk of hepatotoxicity is significantly higher in patients with viral hepatitis, and this increased risk has been observed with multiple different anti-HIV regimens, potentially suggesting that the liver injury observed is due to flare of hepatitis and not direct hepatotoxicity.91 Mitochondrial toxicity is a particularly ominous form of liver injury in HIV patients due to nucleoside reverse transcriptase inhibitors that can inhibit mitchondrial γ-polymerase resulting in depletion of mitochondrial DNA, lactic acidosis and microvesicular steatosis.

Jeremy K. Nicholson: State-of-the-Art Lecture: Metabonomic Approaches in Mechanistic Hepatotoxicology

There are now over 150 words that end in “omics”, and many refer to various levels of biomolecular organization, such as “transcriptomics” or “proteomics”. The goal is to intergrate all of these bits of information into “global systems biology”, defined as “the statistical integration of multi-omic and bio-data to model and visualize systemic gene-environment interactions”.92 One problem is that a perturbation in a single biochemical pathway produces a “ripple effect” within a cell that over time can alter many pathways in that cell, and potentially in remote cells in the body. The metabolome is the collection of all endogenous metabolites produced by a cell or more complex living organisms. It can be argued that changes in gene expression are only relevant if they produce changes in the metabolome.

The term “metabolomics” refers to the measurement of all metabolites and pathways in a given sample or biological system. Several technologies are available that are capable of simultaneously measuring literally thousands of metabolites in biofluids, that includes urine as well as serum. This makes metabolomics very amenable to clinical research applications. An advantage of metabolomic analyses is that the measurements represent aggregate changes over the interval of the biofluid collection, unlike transcriptomics and proteomics approaches that measure concentrations at a specific point in time. Research in rats has identified specific metabolites in urine that can distinguish injury to various regions of the nephron, and can distinguish cholestatic from hepatocellular liver injury,93 and detect hydrazine induced hepatic steatosis94 (Fig. 1). A consortium of industries is collaborating to look at the value of metabolomics in preclinical screening for hepatotoxicity with the hope that improved urine and serum biomarkers of liver injury will be discovered.95 One problem with metabolomics is that many important biochemical pathways have not yet been characterized, particularly in man. Another major challenge is to be able to analyze and visualize the massive amounts of data generated in an unbiased fashion. Techniques to associate metabolite changes and to link these changes to specific biochemical pathways are being developed. An additional challenge is the relatively recent discovery that many metabolites in serum and particularly urine are actually products of colonic bacteria.96, 97 This may not be surprising since the aggregate mass of bacteria in the body can exceed 1 kilogram.

Figure 1.

Metabolomic signature of hydrazine-induced hepatic steatosis. Urine was collected from rats treated with toxic doses of hydrazine and vehicle (control). Shown are representative NMR spectra (600-MHz) obtained from urine an identical time point after treatment. Obtained with permission from J.K. Nicholson.

In summary, metabonomics is a powerful in vivo systems-based tool for investigating events and disease processes. In drug development, metabolomics offers real prospects for toxicity and efficacy prediction in vivo as both genetic and environmental contributions to the metabolic state can be derived non-invasively. Understanding gut microbial influences on host response to drugs may be crucial.

Jack W. Snyder: Proposal from the National Library of Medicine (NLM) to Develop a Web of Knowledge for Hepatotoxicity

It is proposed to develop LIVERTOX, an electronic, open-access “web of knowledge” for hepatotoxicology, designed to provide practicing physicians with comprehensive, “one-stop shop,” web-based information supporting and complementing the growing database of the Drug-Induced Liver Injury Network (DILIN). Utilizing modern search, metasearch, indexing, and interface design tools,98 the NLM proposes to integrate various resources in “LIVERTOX,” including those found in: (a) the DILIN database; (b) a comprehensive bibliography of hepatotoxicity literature; (c) animal data from the National Toxicology Program; (d) human data from FDA's NDA submissions and Adverse Event Reporting System; (e) international perspectives through NLM's World Library of Toxicology & Chemical Safety; (f) peer-reviewed pharm/tox/chem information from NLM's TOXNET; (g) digitized clinical and histopathological images; and (h) public-access “omics” databases. Information regarding proposed collection building strategies; resource and information types, editorial policies, peer review and quality control processes, and knowledge management challenges in hepatotoxicology is now available for consideration and comment (website:http://sis4.nlm.nih.gov/livertox/).

Jay H. Hoofnagle: The DILIN

Based on recommendations that derived from a workshop on drug-induced liver disease99 held in October 2000, sponsored by the National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK), National Institutes of Health (NIH), a Request for Applications was issued soon thereafter requesting applications to participate in a Clinical Research Network on Hepatotoxicity. The ensuing study, functioning as a cooperative agreement, was termed the Drug-Induced Liver Injury Network (DILIN).3 DILIN is composed of five interactive Clinical Centers and a Data Coordinating Center. The Clinical Centers are based at the University of Connecticut (Dr. Herbert Bonkovsky); the University of California at San Francisco (Dr. Timothy Davern,); the University of Indiana (Dr. Naga Chalasani); the University of Michigan, Ann Arbor (Dr. Robert Fontana) and the University of North Carolina at Chapel Hill (Dr. Paul Watkins). The Data Coordinating Center is located at Duke University (Dr. James Rochon). It is estimated that these centers, and their immediate referral networks, care for approximately 12 million patients. The objectives of DILIN are to: (1) provide a prospective, clinical database on unselected cases of hepatotoxicity, (2) develop standardized definitions, grading systems and clinical instruments to identify and assign causality to cases of suspected DILI, and (3) obtain biological samples (serum, urine, PBMCs and DNA) for studies on the pathogenesis of hepatotoxicity using biochemical, molecular, immunologic and genetic techniques.

The two components of the network are a Retrospective and a Prospective Study. The goal of the Retrospective Study is to establish a nationwide registry of patients who had suffered in the past severe idiosyncratic liver injury associated with the administration of isoniazid, phenytoin, clavulanic acid / amoxicillin and valproic acid. The Prospective Study is a longitudinal, epidemiological effort to collect all cases of drug-, toxin- and Complementary and Alternative Medicine (CAM)-induced liver toxicity. Both component studies are presently underway. All enrolled patients will become part of a registry, and have agreed to be contacted in the future and offered enrollment in additional studies. Program announcements will in the future solicit R21 and R01 grants to use the resources developed by DILIN. Industry partners are also welcome as both scientific collaborators and as a source of support to expand the network.

Naga P. Chalasani: Data Obtained From the DILIN Network

At the time of holding of the DILI Single Topic Conference, 40 patients had been enrolled into the retrospective protocol and 96 patients enrolled into the prospective protocol. The implicated medications for DILI cases enrolled into the retrospective study were amoxicillin/clavulinic acid,18 isoniazid,10 valproate7 and phenytoin.5 For the prospective study, 33 individuals (34%) had DILI presumably due to antimicrobials representing the single largest class of implicated agents (amoxicillin/clavulinic acid — 6, anti-TB — 6, nitrofurantoin — 5, other antimicrobials — 16). Other common classes of agents included anti-convulsants, herbal preparations, anesthetics, and NSAIDs. Polypharmacy was common among the enrolled subjects so that it was often not possible to isolate a single agent as the culprit for DILI. In fact, two or more agents were potentially responsible for DILI in 19% of our prospective cases. The cumulative enrollment of patients into retrospective and prospective studies is shown in Fig. 2. Our prospective study has the ability to serve as a surveillance system to capture early signals of DILI caused by recently marketed medications. For example, cases of DILI potentially caused by atomoxetine, etanercept, telethromycin and research agents were enrolled into the prospective study. A similar phenomenon was observed with the Spanish DILI Registry.8

Figure 2.

Enrollment in both the retrospective and prospective protocols conducted by the DILIN network. Cumulative enrollment by month is shown.

In the prospective protocol, three-quarters of the enrollees were Caucasian and nearly 60% were women. Their mean age ± SD was 47 ± 17 years with 63% requiring hospitalization and ≈60% having been been ill for longer than 4 weeks. Jaundice was present in approximately 60% of the subjects. The mean peak serum ALT was 19 times the upper limits of normal. Forty three percent of the patients underwent liver biopsies, although not all liver biopsies were available for review by the study pathologist, Dr. David Kleiner. Fewer than 20% of patients had known preexisting liver disease. The study enrollment continues.

Leonard B. Seeff: Improving Causality Assessment

Drug-induced adverse reactions involving the liver can mimic virtually all forms of liver disease, and therefore it has been difficult to develop a specific and accurate means of establishing a diagnosis of hepatotoxicity. Because there is no diagnostic biomarker, hepatotoxicity is essentially a diagnosis of exclusion. There have been efforts, however, to create causality instruments, the first of which was developed in Paris in 1989 at a meeting sponsored by the Roussel Uclaf Pharmaceutical Company, and was hence named the Roussel Uclaf Causality Assessment Method (RUCAM).4 Cases are rated based on numerical scoring, but the process is somewhat unwieldy and the scoring criteria poorly defined. A second instrument was developed in 1997 in Portugal by Maria and Victorino, termed the Clinical Diagnostic Scale (CDS),100 but in comparison with RUCAM, proved to be less effective,101 although others have found it a useful instrument.102 Most emphasis, however, has been placed on adjudication by an expert panel.103 Because there is no consensus on the preferred diagnostic approach, the DILIN investigators are presently working to develop an improved approach to causality assessment. A complex computer-based process is utilized in gathering and distributing relevant information. Specific information is recorded using specially designed clinical research forms that are then submitted to a 3-member Causality Assessment Committee for evaluation as an expert panel. Cases are adjudicated by the 3 members and differences reconciled during teleconferencing. If not fully agreed upon, the number of reviewers expands to 5 and the majority rules. The cases are assessed for validity using a 5-point scale for certainty of diagnosis. Thus, the effort is to refine the process of expert opinion. Cases are also scored using the RUCAM instrument and the results compared with those of the expert opinion approach. The work to improve causality assessment continues.

David Kleiner: Histopathology of DILI

When evaluating a patient for possible DILI, a liver biopsy can be helpful in narrowing the diagnostic possibilities or by demonstrating pathology that is characteristic of the suspect drug.104 For example, because acetaminophen is known to cause zone 3 necrosis, it can be eliminated from consideration if the biopsy shows a pattern of diffuse spotty lobular inflammation. This analysis is possible because particular drugs are generally only associated with certain kinds of injury.105 Pathology texts generally group hepatotoxins into their known patterns of injury, allowing easy reference. Some pathologic features, such as prominent eosinophilia, granulomas, zonal or massive necrosis or cholestasis with hepatitis may increase the index of suspicion for DILI. As is true for other liver diseases, liver biopsy allows better assessment of the degree of injury or progression to chronic liver disease. When confronted by DILI from novel agents, the pathology can shed light on potential mechanisms of toxicity: intrinsic vs. idiosyncratic, metabolic vs. hypersensitivity. In the case of fialuridine hepatotoxicity, the light microscopic and ultrastructural examination of the injured livers was the first clue that the drug was causing a severe metabolic type of mitochondrial injury.106 The usefulness of liver biopsy is limited by the existing knowledge about DILI, particularly information on the patterns of injury caused by various agents. A Web-based, public database of DILI linked to histopathologic images would be of great value to pathologists and clinicians seeking to sort out events in complicated clinical situations. To this end, all biopsies reviewed as part of DILIN are being photographed to create an image database of well-characterized DILI cases.

Mark W. Russo: Transcriptomics

There have now been many studies that have analyzed changes in rodent liver transcriptome during hepatotoxicity.107, 108 Such studies are difficult to perform in man due to the risks of liver biopsy. Studies performed in rats at the National Institutes of Environmental Health have shown that during acetaminophen hepatotoxicity, some changes in liver transcripts are mirrored by changes in the whole blood transcriptome (personal communication, Richard Paules, Ph.D.). The reason why changes in blood transcriptome may parallel changes in liver transcriptome during acetaminophen toxicity is not clear. Lymphocytes may express CYP2E1, the major enzyme responsible for the bioactivation of acetaminophen to its toxic metabolite, NAPQI109; hence the initial events in acetaminophen hepatotoxicity may occur in lymphocytes. In addition, components of the innate immune system, such as NKT cells, are involved in determining progression or resolution of acetaminophen injury,46 and this may be reflected in changes in blood transcriptome. To determine whether acetaminophen treatment caused consistent changes in blood transcriptome, a preliminary study in 9 healthy volunteers was conducted in the General Clinical Research Center at the University of North Carolina (UNC). Each subject was housed in the center for 7 days and received a constant soy shake diet. A 4 gram single oral dose of acetaminophen was administered on day 4, and the time dependent changes in blood transcriptome were measured by hybridization to Agilent chips (≈20,000 genes). At 48 hours, down-regulation of multiple mRNAs was observed in peripheral blood, including several coding for proteins involved in oxidative phosphorylation. The significance of these changes was supported by high resolution NMR analysis of urine that revealed changes in metabolites consistent with impaired oxidative phosphorylation. Studies are underway to examine the effects of recurrent acetaminophen treatment on the whole blood transcriptome and urinary metabolome.

David Cox: State of the Art Lecture: Genomic Approaches to Understanding DILI

Despite great strides in technology, it remains difficult to identify genes determining drug response, including hepatotoxicity. In part, this is because genetic factors may account for no more than 50% of variation in response to a drug. In addition, most human traits are not caused by a single genetic variation, but probably by 20 or more genetic changes across the genome (Fig. 3). Hence, any single genetic change may be responsible for only a small contribution to the trait. It will therefore be necessary to identify a set of genetic markers across the human genome that determines variation in drug response. This will require identification of single nucleotide polymorphisms that are sufficient to do association studies, and continued development of high-throughput, low-cost genotyping assays. Most importantly, the successful hunt for genetic bases of susceptibility to drug toxicity will probably require at least 100 individuals who very clearly have the adverse event of interest. National networks such as DILIN are essential to accomplish this.

Figure 3.

Susceptibility to DILI determined by multiple alleles. In this hypothetical example, DILI susceptibility is determined by five 5 distinct genes that each have common functional (green) and dysfunctional (red) alleles. The number of dysfunctional alleles possessed by an individual patient correlates with susceptibility. The most highly susceptible patients are those homozygous for dysfunctional alleles at all five loci (far right). Those with lowest susceptibility have no homozygosity for dysfunctional alleles (far left). This is the property of Perlegen Sciences, Inc. and has been provided by David R. Cox, M.D., Ph.D., Chief Scientific Officer.


It seems reasonable to assume that a combination of genetic and environmental factors account for why a rare patient will develop idiosyncratic hepatotoxicity. Elucidation of these factors should lead to clinical tests that can identify the susceptible patient so that the toxicity can be avoided. The identity of these factors should also shed light on underlying mechanisms and thereby allow development of safer drugs.

To date, most clues to relevant susceptibility factors have come from rodent studies of acetaminophen hepatotoxicity. It is hoped that susceptibility factors for acetaminophen hepatotoxicity identified in rodents will be relevant to understanding idiosyncratic liver injury, but this may not necessarily be the case. Acetaminophen is a classic predictable hepatotoxin and there is little evidence that it can cause a typical idiosyncratic injury. There exists substantial interindividual differences in the acetaminophen dose that will cause hepatotoxicity, but the liver injury observed remains pericentral necrosis that occurs within 24 to 48 hours of ingesting a toxic dose. In contrast, idiosyncratic hepatotoxicity characteristically occurs months into therapy and pericentral necrosis is unusual.

Invariably, validation of susceptibility factors will require study of large numbers of patients who have experienced idiosyncratic DILI and recovered. This is the primary rationale for two studies currently being conducted by the DILIN network. In these studies, patients with bone fide DILI are providing genomic DNA, immortalized lymphocytes, and serum. In the process of undertaking these studies, DILIN is wrestling with the large problems of causation assessment and hopes to create improved diagnostic tools that will be useful to the practicing clinician as well as the clinical researcher. A very important component of the DILIN studies is that all subjects have given permission to be recontacted and offered enrollment in additional studies, such as genotype:phenotype correlations or pedigree studies.

The Single Topic Conference provided an opportunity for basic, translational and clinical researchers to “sit at the same table” in discussions about the promise and challenges of DILI research. There appeared to be great enthusiasm for the resources being created by the DILIN network and a general optimism regarding the ability to use these resources to identify the susceptibility factors underlying idiosyncratic DILI. This is an exciting time for DILI research.