The purpose of this brief review is to focus on a mechanistic and practical approach to the problem of drug hepatotoxicity, in the context of clinical liver disease. For extensive and detailed discussions of individual drugs and the types of hepatotoxicity they may cause, recent textbooks should be consulted .^1-3 

Hepatic drug metabolism
The liver plays an important role in the metabolism of drugs. How the liver acts upon drugs can have important implications for both drug-induced hepatic damage and liver cancers. Hepatic drug metabolism is categorized as activation (Phase I processes) and detoxification (Phase II processes). For most hepatotoxicity, the balance between activation and detoxification is critical. Numerous factors influence this balance including age or developmental stage, fasting or undernutrition, other drugs and chemical inducers, immunomodulators resulting from viral infection or inflammation. Chemical inducers may have different effects on Phase I and Phase II processes. The rate of absorption of the toxic drug, how it is distributed in the body, and how long it and its metabolites linger in the body, and whether it is toxic to other organ systems may all influence the nature and extent of drug hepatotoxicity. Whether the drug is taken as a large single dose (acute ingestion) or many smaller doses repeatedly (chronic ingestion) may also influence or modify its metabolism in the liver. Genetic features, namely, polymorphisms of cytochromes P450 and various Phase II enzymes also affect this critical balance. Hepatic drug metabolism also shows developmental variability. Different drug-metabolizing enzymes may be expressed differently in infants compared to adults, and in general clearance of many drugs is more rapid in pre-pubertal children than in adults.

Phase I processes generally involve the cytochromes P450.⁴  These processes include hydroxylation, dealkylation, and dehalogenation. Cytochromes P450 are diverse hemoproteins and, compared to most enzymes, they have less specificity for substrate. Thus the same cytochrome P450 may act on numerous drugs. Many cytochromes P450 are inducible: they become more abundant in response to treatment with various chemicals. The cellular mechanisms by which induction occurs are different for different cytochromes P450. Cytochrome P450 are grouped within families and subfamilies depending on the degree of similarity in gene sequence of the apoprotein. The cytochrome P450 3A (CYP3A) subfamily includes cytochromes induced by pregnenolone and by glucocorticoids, and CYP3A4 is the most abundant cytochrome P450 in human liver. Other important human hepatic P450s include CYP1A2 (induced by polycyclic aromatic hydrocarbons anvironmental contaminants), CYP2B6 (induced by phenobarbital), CYP2D6 (polymorphic), CYP2C19, CYP2E1 (induced by ethanol), and CYP4A subfamily (induced by drugs which cause peroxisome proliferation).

Phase I processes typically make the drug into a more polar chemical. Sometimes a Phase I process converts a drug to its active form (for example, prednisone is converted to prednisolone), but the usual effect is to expedite excretion by adding a chemical moiety ready for a conjugation reaction via a Phase II process. Phase II processes are performed by numerous different classes of enzymes, including glutathione S-transferases, glucuronosyl transferases, epoxide hydrolase, sulfotransferases, and N-acetyltransfer­ases. Typically these reactions complete the transformation of a hydrophobic chemical to a hydrophilic one which can be excreted easily in urine or bile. A few Phase II enzymes, such as some glucuronosyl transferases, can be induced. Some are polymorphic, such as N-acetyltransferase (either rapid or slow acetylators). In some metabolic diseases the activity of Phase II enzymes may be abnormal.^{5, 6} 

The product of a Phase I reaction may be an unstable or reactive metabolite, and Phase II reactions may inactivate such chemicals before cell damage occurs. Whether reactive metabolites actually damage a liver cell also depends on how much reactive metabolite binds to cellular components, whether these components are critical to cell integrity, and whether repair is possible. A reactive metabolite binding to vital intracellular proteins or membranes may lead to necrosis. Reactive metabolites may initiate apoptosis by directly damaging mitochondria or by initiating immune mechanisms. Binding to DNA may lead to mutagenesis or carcinogenesis. Besides lipid peroxidation, membranes can be altered by alkylation (addition of an aliphatic radical such as methyl or ethyl groups), arylation (addition of an aromatic group such as a phenyl group) or acylation (adding a radical derived from a carboxy acid).

Reactive metabolites are electrochemically unstable. Electrophilic intermediates (or electrophiles) are formed when electrons are lost from the original chemical; they carry a net positive charge. Examples include hydroxylamines, quinoneimines, and arene oxides. Nucleophiles are negatively-charged species, formed through activation of oxygen, such as halocarbon and nitroso radicals. They tend to bind to intracellular lipids, leading to lipid peroxidation. Glutathione reacts with electrophiles via conjugation reactions catalyzed by glutathione S-transferases. It also reacts with hydrogen peroxide and activated oxygen species via glutathione peroxidase. As a general rule, when toxic metabolites are the important cause of cell damage, high tissue concentrations of the parent drug are not found. Metabolite(s) covalently bound to cellular constituents may be detected.

The anatomical complexity of the liver accounts in part for the diversity of patterns of hepatotoxicity. Drug-associated injury may involve hepatocytes or non-parenchymal cells in the liver besides hepatocytes. Cytotoxic damage to bile duct cells, hepatic stellate cells or endothelial cells accounts for some of the clinical diversity of drug-induced liver disease. Within the hepatocyte, a drug or its reactive metabolites may interfere with biliary excretion or damage proteins within the biliary excretion apparatus, thus leading to cholestasis.⁷  The cellular specialization of hepatocytes is also important. For example, most drug metabolism occurs in zone 3 of the Rappaport acinus: necrosis of hepatocytes due to generation of a toxic metabolite may be most prominent in a zonal pattern.

Classification of drug hepatotoxicity
Drug hepatotoxicity can be acute or chronic. Acute hepatotoxic injuries develop over a relatively short time and show no histopathological features of chronicity. Subacute hepatotoxicity includes damage developing over weeks to months, with fibrosis and possibly cellular regeneration histologically. Chronic hepatotoxic lesions include those with fibrosis or cirrhosis, predominantly vascular lesions, and neoplasia. Some drugs can cause clinical liver disease indistinguishable from autoimmune hepatitis. These drugs include oxyphenisatin (generally no longer in use), a-methyldopa (rarely used), isoniazid, nitrofurantoin, and minocycline.

Our knowledge of mechanisms of hepatotoxicity is evolving. For a long time, hepatotoxicity has been categorized on the basis of predictability. Intrinsic hepatotoxins are differentiated from idiosyncratic heptatotoxins. The intrinsic hepatotoxin causes predict­able hepatic damage in almost any individual. The toxicity is dose-related in that higher doses cause worse damage, and animal models can be developed which exhibit the same type of hepatotoxicity. This classification has limited clinical applicability and delineates mechanisms of hepatotoxicity only superficially. In practice, most instances of hepatotoxicity, mainly those associated with medica­tions, are unpredictable, infrequent, and thus apparently sporadic. If such a reaction is accom­panied by systemic features including fever, rash, eosinophilia, atypical lymphocytosis and possibly other major organ involvement, then classically it has been regarded as an idiosyncratic hypersensitivity reaction, where "hypersensitivity" with its connotation of allergy is left undefined.

An alternate view is that idiosyncratic hepatotoxicity has a biochemical basis and is due to defective hepatic drug metabolism (often termed "metabolic idiosyncrasy"). Individu­als who have specific abnormalities in drug metabolism at risk for adverse drug reactions. If this abnormal metabolism is expressed in liver cells, then these rare individuals will develop hepatotoxicity—if they are exposed to the appropriate drug. In most instances a metabolite, not the drug itself, is responsible for hepatotoxicity. Typically, the problem is a defect in detoxification of the reactive metabolite because the detoxification system is itself focally defective and cannot meet the normal demands of metabolite production. Sometimes these individuals show systemic features interpreted as hypersensitivity: it is likely that interaction of the reactive metabolite with cellular components, such as the cell membrane, elicits an immune response. In such cases hypersensitivity is itself the conse­quence of metabolic idiosyncrasy, not a separate mechanism of drug hepatotoxicity. There may be strictly allergic drug hepatotoxicity, but investigations of the mechanism of drug-induced hepatotoxicity suggest that metabolic idiosyncrasy is much more common than previously supposed. It seems likely to account for hepatotoxicity with drugs which show two main patterns of toxicity: mild reversible toxicity in a comparatively large segment of patients and severe hepatotoxicity in a few individuals. Toxic metabolites are probably involved in both patterns of toxicity. Severe reactions occur in rare persons with abnormal generation of toxic metabolites or detoxification, irrespective of the appearance of drug allergy.

The major implication of the metabolic idiosyncrasy thesis is that most drug hepatotoxicity is predictable if one understands the pathways of hepatic biotransformation and detoxifica­tion for each drug. Given the plethora of drugs and hepatic biotrans­formation pathways, it is no wonder that most clinically-important drug hepato­toxicity appears sporadic and fortuitous. However, there is enough experimental data available now to warrant rethinking the intrinsic/idiosyncratic-allergic classification of drug hepatotoxicity. These definable metabolic defects in hepatic drug metabolism are particularly common in the types of drug hepatotoxicity which occur in children.

Recent research has focused on the role of the immune system in drug hepatotoxicity.⁸  With some drugs the connection between immune-mediated mechanisms and hepatic damage may be very direct: autoantibodies are elaborated against hepatic cytochromes P450. The P450 involves varies with different drugs: CYP2C9 for tienilic acid, CYP1A2 for dihydralazine, CYP2E1 for halothane. Reactive metabolites may alter other components of hepatocytes to form neo-antigens. Hepatocyte damage mediated through immune mechanisms may involve apoptosis or necrosis. Bile acid associated hepatocyte injury involves Fas activation leading to apoptosis. When toxic metabolites or reactive oxygen species or cytokines stimulate Kupffer cells specific mechanisms of cell damage are set into motion involving tumor necrosis factor-a (TNF-a) or nitric oxide (NO) produced by Kupffer cells. NO elaborated by Kupffer cells and hepatocytes appears to play a role in acetaminophen hepatotoxicity. Other cytokines, including CXC chemokines regulating leukocyte action, may modulate these effects. The vigor of the immune response in general, an individual polygenic trait, may also determine the importance of immune mechanisms in drug hepatotoxicity.

The following classification of types of drug hepatotoxicity is designed to avoid the drawbacks of other systems currently in use (Table 1). Drugs are categorized as being intrinsic hepatotoxins or contingent hepatotoxins. The intrinsic hepatotoxin is a true poison: it is causes predict­able hepatic damage in almost any individual in a dose-dependent fashion. The contingent hepatotoxin causes liver damage only in certain individuals either because of pre-existing genetic or environmental factors. Instead of being a universal poison, it is a "personal" poison. An example of a genetic factor potentiating a contingent hepatotoxin is carbamazepine hepatotoxicity which occurs in an individual who pharmacogenetically has inadequate Phase II processes to detoxify the reactive metabolite formed. An example of an environmental factor leading to a drug being a contingent hepatotoxin is moderate or appropriate doses of acetaminophen causing liver damage in an individual who uses alcohol chronically and therefore has induced hepatic CYP2E1. To specify the role of immune mechanisms in drug hepatotoxicity, drugs can be further categorized as "eliciting and immuno-allergic response." This may be a special type of contingent hepatotoxicity since genetic aspects of immune responsiveness may play an important role in whether an individual is likely to have an immune component with drug hepatotoxicity. Examples of possible immuno-allergic responses include (1) fever, rash, atypical lymphocytosis, eosinophilia, and multi-systemic involvement—the so-called systemic "hypersensitivity syndrome," (2) hepatic granulomatosis, and (3) clinical and histological features of "chronic active hepatitis," that is, a process resembling autoimmune hepatitis.

Drug hepatotoxicity produces a broad spectrum of clinical liver disease, summarized in Table 2. Most drug-associated liver disease involves cytotoxicity: direct severe damage to hepatocytes. Serum aminotransferases are elevated, and liver failure may occur. The exact mechanism of cell death is highly variable and probably differs with each drug or toxins. Necrosis is the predominant pattern, but apoptosis may play a role. Hepatocyte damage may be zonal, reflecting metabolic specialization in various parts of the hepatic lobule. Specifically, hepatocytes in have the highest concentration of drug-metaboliz­ing enzymes and thus the greatest potential for producing toxic intermediates. Zonal hepatocellular necrosis mainly in zone 3 of the Rappaport acinus suggests that metabolic activation of toxic metabolites has an important role in the pathogenesis of the toxicity; however, spotty necrosis scattered throughout the lobule does not exclude a mechanism involving toxic metabolites. Whenever hepatocellular damage is sufficiently severe, some degree of cholestasis will develop. Some drug-associated liver damage is mainly cholestatic. Clinically findings include jaundice, pruritus, markedly elevated serum alkaline phosphatase and mild elevations of serum aminotransferases. This type of drug-associated liver disease is often due to direct damage to the bile secretory apparatus by toxic metabolites.⁷  In particular, the bile salt excretory pump (BSEP, abnormal in Progressive Familial Intrahepatic Cholestasis type 2) appears to be an important target in forms of drug hepatotoxicity with prominent cholestatic features.

Principles of clinical management
In the individual patient any drug suspected of causing hepatotoxicity should be stopped. Most drug-induced liver disease resolves spontaneously when the hepatotoxic drug is discontinued. Severe chronic changes may not resolve. Certain hepatotoxins require timely treatment with specific antidotes, such as N-acetylcysteine in acetaminophen hepatotoxicity. Steroid treatment has appeared beneficial when severe acute hepatitis dominates a multisystemic hypersensitivity reaction as with phenytoin, carbamazepine, or phenobarbital,^{9, 10}  but large prospective studies are not available. In general, the use of steroids in drug-induced liver disease remains controversial. When drug hepatotoxicity leads to acute liver failure, the usual supportive measures are required and liver transplantation may be life-saving.

Making the diagnosis of drug-associated liver disease is critically important and not necessarily easy. A high index of clinical suspicion is essential. A meticulous history of the illness and a detailed history of all drugs taken, including over-the-counter preparations, with specific questions relating to potential exposure to environmental or industrial toxins are absolutely necessary. It is important to determine that the appropriate dosage was actually given. This is especially important with children and may involve having the parent or caregiver bring in the (possibly empty) bottles of the drug(s) used. Chemical analysis of herbal remedies may be required.

Liver biopsy, including electron microscopic examination if possible, is often helpful and can be definitive. Algorithms for determining the likelihood of an adverse drug reaction ^11-13 may be helpful. Clinical rechallenge is rarely possible because it poses too much risk to the patient. In vitro rechallenge of the patient's lymphocytes with generated toxic metabolites usually provides important corroborative evidence,¹⁴  but this remains a research investigation which is not generally available. Other rechallenge assays based on purely immunological mechanisms have proven difficult to interpret. Determination of a pharmacogenetic defect in drug metabolism or drug detoxification may have important implications for other primary relatives.

Resources for documenting drug hepatotoxicity include Meyler's Side Effects of Drugs, which is a published book series, a compendium plus regular updates. Databases of the reported experience of adverse reactions with drugs, including drug hepatotoxicity, are also available on the Internet (Table 3). Most algorithms to test likelihood of an adverse drug reaction attach importance to whether the adverse event has ever been reported before. It is sometimes useful to approach the pharmaceutical company which has developed the drug or the FDA to find out about reports not in the public domain. These resources however will provide little direct assistance with the greatest challenge in drug hepatotoxicity, which is to identify the first occurrence of significant drug hepatotoxicity, except where the experience with chemically similar compounds may provide some guidance or clues to the mechanism of hepatotoxicity.

Table 2. Examples of patterns of drug hepatotoxicity

Asymptomatic ­AST, ALT pemoline Acute hepatitis isoniazid, halothane Hepatitis-cholestasis erythromycin Zonal hepatocellular necrosis acetaminophen Bland cholestasis cyclosporine, estrogens Steatohepatitis (= NASH) perhexiline, methotrexate Phospholipidosis amiodarone Microvesicular steatosis valproic acid, fialuridine Granulomatosis sulfonamides Liver cell adenoma estrogens Malignant tumours estrogens, anabolic steroids Peliosis estrogens, androgens Hepatic vein thrombosis estrogens (N.B. oral contraceptive pill) Veno-occlusive disease thioguanine, busulfan Non-cirrhotic portal hypertension azathioprine

Table 3. Databases relating to drug hepatotoxicity available on the Internet

Medwatch	www.fda.gov/medwatch/
Pharmalert	www.space.tin.it/salute/pgallet.adv
Australian Adverse Drug Reaction Bulletin	www.health.gov.au/tga/adv/aadrb
Pharmacovigilance	www.pharmacovigilance.org
Drug Info	www.druginfonet.com

References

1. Zimmerman HJ. Hepatotoxicity: the Adverse Effects of Drugs and Other Chemicals on the Liver. 1st ed. New York: Appleton-Century-Crofts; 1978.
2. Farrell GC. Drug induced liver disease. New York: Churchill Livingstone; 1994.
3. Kaplowitz N, DeLeve L. Drug Hepatotoxicity. New York: Marcel Dekker; 2002.
4. Gonzalez FJ. The molecular biology of cytochrome P450s. Pharmacol Rev 1989;40:243-288.
5. Spielberg SP, Gordon GB. Glutathione synthetase-deficient lymphocytes and acetaminophen toxicity. Clin Pharmacol Ther 1981;29:51-55.
6. Stoner E, Starkman H, Wellner D, et al. Biochemical studies of a patient with hereditary hepatorenal tyrosinemia: Evidence of a glutathione deficiency. Pediatr Res 1984;18:1332-1336.
7. Bohan A, Boyer JL. Mechanisms of hepatic transport of drugs: implications for cholestatic drug reactions. Semin Liver Dis 2002;22:123-136.
8. Jaeschke H, Gores GJ, Cederbaum AI, Hinson JA, Pessayre D, Lemasters JJ. Mechanisms of hepatotoxicity. Toxicol Sci 2002;65:166-176.
9. Shear NH, Spielberg SP. Anticonvulsant hypersensitivity syndrome. In vitro assessment of risk. J Clin Invest 1988;82:1826-1832.
10. Roberts EA, Spielberg SP, Goldbach M, Phillips MJ. Phenobarbital hepatotoxicity in an 8-month-old infant. J Hepatol 1990;10:235-239.
11. Naranjo CA, Busto U, Sellers EM, et al. A method for estimating the probability of adverse drug reactions. Clin Pharmacol Ther 1981;30:239-245.
12. Danan G, Benichou C. Causality assessment of adverse reactions to drugs--I. A novel method based on the conclusions of international consensus meetings: application to drug-induced liver injuries. J Clin Epidemiol 1993;46:1323-1330.
13. Benichou C, Danan G, Flahault A. Causality assessment of adverse reactions to drugs--II. An original model for validation of drug causality assessment methods: case reports with positive rechallenge. J Clin Epidemiol 1993;46:1331-1336.
14. Spielberg SP. In vitro assessment of pharmacogenetic susceptibility to toxic drug metabolites in humans. Federation Proc 1984;43:2308-2313.