Do Preclinical Testing Strategies Help Predict Human Hepatotoxic Potentials?

Abstract

Overt hepatotoxicity due to drug administration is a real and present issue in drug development and regulatory circles. Preclinical drug development is intended to identify potential risks and target tissues prior to introduction of new molecular entities into the human population. The standard regimen is testing at various multiples of the intended human therapeutic dose in at least 2 species of animals, one rodent (rats or mice), one non-rodent (dogs, nonhuman primates, minipigs, and rabbits, as examples) for at least two weeks of repeated dosing. Experience has shown that this regimen “works” most of the time. However, preclinical models are not infallible and are not always predictive. Whether the lack of predictivity is due to individual human genetic sensitivities, immunologically mediated phenomena, disease mediation or idiosyncratic reactions, the animal models are limited in detecting these characteristics and other low incidence phenomena. While it is uncommon for drug developers to continue development with products that elicit overt hepatic toxicity early in the animal testing, some products have made it through the approval process and then shown significant adverse effects. Some of the drugs (acetaminophen, isoniazid, trovafloxacin, troglitazone, bromfenac, clarithromycin, telithromycin) that have shown this propensity will be discussed in detail from early preclinical development to marketing and, in some instances, to limitations to usage or removal from the U.S. marketplace.

Drugs for which development or marketing discontinued due to hepatotoxicity include: Troglitazone, Bromfenac, Tienilic acid, Temafloxacin, Nomifensin, Perhexilin, Ibufenac, and Benoxaprofen.

Drugs for which usage has been limited or that have been withdrawn in certain countries due to hepatotoxicity include: Zileutin, Trovafloxacin, Tolcapone, and Felbamate.

Individual drugs may elicit single or multiple morphologic changes in the same liver-amiodarone elicits phospholipidosis (with lamellar inclusions within hepatocytes and other cells visualized by EM) and steatohepatitis but by different mechanisms. Phenylbutazone may elicit necrosis, cholestasis, granulomata, and/or a combination of these findings.

Depending on the level of the liver enzymes, drug injuries can be classified as hepatocyte injuries (necrosis and/or steatosis), cholestatic injuries (bile secretion and/or bile flow disturbances) and mixed disorders. Liver vessels are rarely involved in drug-induced lesions. Antibiotics and non-steroidal anti-inflammatory drugs account for many of the drug-induced hepatotoxic events. Thus widespread use of drugs with weak hepatotoxicity signals for “routine therapies” may be unwise due to cumulative effects.

Warning Signs of Idiosyncratic Liver Injury (Alden et al., 2003)

1. Structural alerts

2. Glutathione depletion

3. Glutathione drug/metabolite conjugates

4. Covalent binding

5. Bioaccumulation in the liver

6. Drug interactions

7. Toxicity gene induction responses

8. Ames or micronucleus assay positive response

9. Liver injury in the rat and/or higher species without a safety margin at steady state concentration of drug in the liver

10. P450 enzyme induction and liver hypertrophy without a safety margin

What is idiosyncratic toxicity? Various scientific disciplines define the term differently. From a preclinical toxicologist’s perspective, it means that the toxicity is rare and unpredictable from animal studies. The erratic temporal and dose relationships that characterize idiosyncratic drug responses suggest the possibility that some event during the course of therapy renders tissues particularly susceptible to toxic effects of the drug. Latency between initiation of therapy and onset of liver disease may give some idea as to pathogenesis. Early onset is strong evidence for direct toxicity of the drug or a metabolite which is more often predictable (e.g., acetaminophen overdose). In contrast, delayed onset of symptoms may occur weeks to months after exposure. Intermediate latency is often characteristic of hypersensitivity reactions including eosinophilia, rash, and rapid response to rechallenge (e.g., amoxicillin-clavulinate reaction). The long latency type is not usually associated with signs of hypersensitivity and response to rechallenge (e.g., isoniazid and troglitazone) is variable. These delayed reactions may occur as a result of metabolism of the drug or response to injury. In man, unpredictable reactions may occur against a background of mild, asymptomatic liver injury characterized by increased ALT >3× the upper limits of normal. The majority of subjects with increased ALT have some sort of adaptive change and idiosyncratic reactions may occur in those subjects who do not possess this adaptive ability. Thus, the rare occurrence of immune-mediated liver disease is often superimposed on preexisting increased ALT levels, suggesting that the drug has a toxic potential but, in rare subjects, the potential leads to metabolism-dependent hypersensitivity.

In the Unites States, in all age groups, 10% of liver disease is caused by drugs (Lee, 2003). Acetaminophen is responsible for ~40% of all acute liver failure cases in adults (Chitture, 2002). Mortality for acute icteric/toxic hepatitis in man is ~10% irrespective of the causative drug. This reaction is characterized by jaundice, markedly elevated serum enzyme levels, coagulopathy and encephalopathy, indicative of fulminant liver failure. Well-demarcated zone 3 necrosis with less inflammation than expected and minimal portal reaction is the usual scenario. Eosinophils within an infiltrate may suggest a hypersensitivity reaction but their presence is neither proof nor necessary for the diagnosis. Intrinsic hepatotoxins include drugs that predictably induce liver damage when taken in sufficient quantities. The type of damage is usually characteristic of a particular drug and often zonal necrosis follows a short latency period. This phenomenon is usually reproducible in experimental animals. Preclinical testing effectively detects this type of lesion and the drugs submitted for approval today rarely elicit these types of damage.

Drugs with known intrinsic hepatotoxic effect can be divided into 2 subgroups: those with a direct hepatotoxic effect and those with an indirect effect. Both types can cause cytotoxic and cholestatic types of liver injury. A direct cytotoxic effect refers to the direct development of liver cell injury at the subcellular level, affecting various organelles with subsequent steatosis, necrosis or both. Examples of this type of injury include carbon tetrachloride (zonal necrosis without inflammation), cocaine (zonal necrosis sharply demarcated from the normal liver) and phosphorus (Kupffer cells proliferate and are filled with ceroid). Endothelial cells can show proliferative changes. Portal tracts usually are unaffected. Paraquat elicits a direct type of injury with the biliary epithelium being the primary target. Indirect injury means that the cellular injury is caused by alterations of specific metabolic pathways or selective effects on membrane cell receptors and DNA or RNA, either within the nuclei or in the cytosol. Acetaminophen is an example of a drug that causes this type of injury. Drugs that elicit indirect cholestatic toxic injury by selective interference with bile excretion and uptake from blood include contraceptives and anabolic steroids.

Unlike toxic responses to xenobiotics that are dose-related and have a characteristic temporal pattern, idiosyncratic drug reactions have the following characteristics: (1) usually occur in < 5% of the subjects exposed to the drug, (2) are unrelated to the drug’s pharmacologic effect, (3) demonstrate no clear dose relationship, and (4) occur with inconsistent temporal patterns (Zimmerman 2000). Due to the small numbers of subjects and, potentially, animals affected, predictability of the preclinical testing is not very accurate. If animal models existed that could predict idiosyncratic reactions, they could be used to detect potential problems early in drug development and could be used to help design appropriate human clinical trials. For detection of hepatotoxicity, monitoring for additional liver parameters (i.e., leucine aminopeptidase, 5′-nucleotidase, ornithine carbamyl transferase, glutamate dehydrogenase, cytochrome c, proteonomics) may clarify mechanisms to minimize injury or to determine if alternative drug candidates that do not have similar potential for toxicity would be better utilized. When idiosyncratic drug toxicity takes place, it requires that multiple critical and discrete events occur: the probability of the event is a product of the probabilities of each discrete event. Under this hypothesis, idiosyncratic drug reactions can be evaluated experimentally by studying key determinants (chemical properties, exposure, environmental factors, and genetic factors). Chemical properties of importance include: formation of reactive metabolites, metabolism by cytochrome P450 isoforms, P450 induction, and significant pharmacokinetic interactions with concurrently administered drugs or agents.

Idiosyncratic reactions are qualitatively different from the drug’s pharmacologic activity, are often immunologically mediated and have a host-dependent component. There are distinctions based upon whether the reaction is considered a hypersensitivity reaction or a metabolic idiosyncrasy. True examples of hypersensitivity are unusual but when they happen, symptoms usually occur weeks following exposure and may re-occur upon re-exposure with a much shorter latency period. Most drug-related hepatotoxicity is related to aberrant and unique enzyme systems or interference with normal enzymes (i.e., CYP3A4 inhibitor or substrate). This may lead to the production of toxic metabolites from drugs (as with isoniazid, para-aminosalicylic acid, valproic acid). CYP3A4 is the most common isoform involved in troglitazone toxicity, whether directly or by formation of metabolites leading to toxicity and covalent protein adduct formation.

In the “multiple determinant hypothesis” for idiosyncratic drug toxicity, it has been proposed that the toxicity is the result of “multiple, discrete but necessary processes.” Using this hypothesis, it would be possible to study idiosyncratic reactions by evaluating the chemical properties of the test agent, the exposure level, the environmental factors and the genetic factors. A modest inflammatory response can enhance tissue sensitivity to a variety of chemicals. These observations have lead to the hypothesis that inflammation during therapy may decrease the threshold for toxicity and render an individual susceptible to a reaction that might not otherwise occur.

Clinical/epidemiological studies of the relationships between inflammation and drug idiosyncrasy in man may provide insight into the applicability of findings in animals to those in man. Relatively small differences in the temporal relationship between an inflammatory episode and drug exposure could determine whether an idiosyncratic reaction occurs. Genes controlling such entities as hepatocellular glutathione and other antioxidants, proliferative repair and signal transduction may be important in defining idiosyncratic responses. If this premise is correct, genetic differences in animals/man may render some more likely to develop idiosyncratic responses to the same exposure of drug. In the future, genotyping might identify at-risk subjects prior to drug exposure.

Idiosyncratic toxicity may be an extreme sensitivity to a very low dose of compound or individual factors that increase the expression of the intrinsic toxicity of a drug and/or its metabolite. The consequences include hepatocellular necrosis, apoptosis or sensitization to cytokines or inflammatory mediators produced by nonparenchymal cells. This implies that little is usually known or understood about the underlying mechanisms of the toxicity. It is likely that many other extrinsic factors influence the severity of the reaction including disease state, nutritional status, concomitant medications, inflammatory mediators, etc.

Why can’t we reliably predict idiosyncratic drug-related hepatotoxicity? By definition, idiosyncratic reactions cannot be predicted by preclinical studies. This may be due to (1) unavailable models (as with community acquired pneumonia) and/or (2) use of an inappropriate model (as with rodents and nitric oxide synthetase inhibitors). Differences between species in absorption, distribution, metabolism and/or excretion may limit the potential of the model to demonstrate the selective toxicity in man. There have been several compounds that have progressed into human clinical trials and even to approval that have caused severe hepatotoxicities. Were there preclinical signals that were missed or were the toxicities completely unexpected and due to “idiosyncratic” reactions? There are many possible reasons but no definitive explanations for these cases. There may be significant differences in metabolism between the animal species tested and man. There may be disease-related or genetic predispositions that are not tested in animal models. There may indeed be an immune-mediated idiosyncratic response. This idiosyncratic reaction may be manifested by fever, rash, eosinophilia and multi-organ failure/involvement.

For example, development of the anxiolytic agent, panadiplon, was halted after human exposures elicited hepatotoxicity. A similar toxicity was demonstrated in the rabbit, but not rats, dogs or monkeys. Additionally, the confounders such as age, disease state, concurrent medications, genetic susceptibility, etc. are not addressed in the preclinical models as these models generally utilize young, healthy animals as specified in the Good Laboratory Practice regulations (21CFR part 58).

Drugs with high potential for toxicity in man need to be very carefully evaluated but one must consider the intended population—the risk for normal healthy volunteers may not be acceptable but the risk for subjects with HIV, end stage renal or hepatic disease or cancer may be reasonable. Quantitative safety limits need to be determined and/or dose limiting toxicity clearly identified. Selection of the starting dose is critical as too low a starting dose may provide no valuable information and prolong the clinical trial period. However, this implies that the appropriate preclinical animal models are being used. Secondary effects due to exaggerated pharmacology must be well defined. For example, while the test animals may show ataxia and convulsions, they are not able to express headache, visual abnormalities or sweating. Thus there would appear to be only a modest correlation for neurologic signs and symptoms between animals and man. Serious adverse effects on the brain and spinal cord are usually visualized on histologic examination so detection via the conventional models would appear to be useful in finding “the bad actors.” Preclinical toxicity testing using animal models screens out the most potent hepatotoxins. Unfortunately, hepatotoxicity at extremely high doses in animals does not necessarily predict human response and the choice of animal model may be imperfect. Compounds that do not elicit significant hepatic injury in multiple animal species usually are not inducers of serious hepatotoxicity in man. Whether or not the animal studies show toxic properties at high doses, cautious monitoring during clinical trials and post-marketing needs to be a universal truth.

In a multinational study of adverse effects noted in pre-clinical models and later toxicities shown in humans, data showed that effects in humans were better predicted by dogs than by rats, mice, or rabbits in most instances (Kaplowitz, 2001). Although many fewer nonhuman primate (NHP) studies are conducted than in other models, it does not appear that the NHP is better than dog or mini pig models at predicting human hepatic toxicoses for most xenobiotics. As a rule, repeat dose testing for a minimum of two weeks is appropriate to permit the development of and progression of organ pathology when developing a new molecular entity. These studies should include primary and secondary pharmacology, as well as toxicity across at least 2 species.

In the FDA White Paper on Nonclinical Assessment of Hepatotoxicity in Man, the following paradigms are discussed: Tier 1 testing is the ‘standard’ testing of 2 mammalian species (1 rodent, 1 nonrodent) for at least the intended duration of treatment and preferably at multiples of the human exposure. Ideally, in these studies one can identify a dose at which no adverse effects are found (NOAEL) and it is possible, using risk:benefit paradigms, to assign a safety margin and a potentially safe dose to start with in man. These Tier 1 studies are conducted in young, healthy animals, under the assumption that extrapolations can be made to both healthy and diseased humans. In the case of FIAU, the potential for hepatotoxicity was not appreciated during preclinical evaluation for safety. Use of these models does not enable drug developers to detect hepatotoxicities occurring at low incidence or potential drug-drug interactions that may occur in humans being treated with multiple drugs concomitantly. Immune-mediated hepatotoxicity can only rarely be predicted in preclinical studies. While the rat and the dog are the most commonly used species, it is evident that more appropriate models can be utilized (e.g., woodchucks for nucleoside analogs, minipigs for sepsis products). In the case of fialuridine (FIAU), a thymidine analog, the healthy rodent, dog and nonhuman primates did not provide evidence that the drug might be hepatotoxic. When the agent was tested in woodchucks infected with hepatitis, the syndrome of delayed toxicity (hepatic steatosis, increased bilirubinemia, increased prothrombin time, and lactic acidosis) was very comparable to the adverse effects in the subjects with Hepatitis B (the intended treatment population). Subjects that were treated with FIAU for 9–13 weeks developed signs of severe and sometimes fatal hepatotoxicity associated with lactic acidosis. Subsequently, it has been demonstrated that the major target is the mitochondrial DNA in hepatocytes with a high affinity for thymidine kinase 2, constitutively expressed in mitochondria. This lead to mutations in mitochondrial enzymes and defective oxidative phosphorylation. This, in turn, lead to leakage of electrons and production of superoxide anion and increased oxidative stress.

Agents that bioaccumulate offer a real challenge to safety evaluation. Covalent binding of reactive drug metabolites typically demonstrates bioaccumulation. Plasma levels may not reflect tissue levels.

In the event of inconsistencies between the results from the various animal models, it is incumbent upon the developer to try to understand the factors that might demonstrate species specificity for the adverse effect. This would enable a more realistic risk:benefit evaluation. If the only adverse effect is an increase in ALT, with no histologic correlates, then it is incumbent on the developer to determine the source (e.g., body weight effects, muscle necrosis, trauma, anemia, P450 induction).

As mentioned earlier, several human hepatotoxicants did not show significant hepatotoxic potential in preclinical testing. The following drugs are examples:

Acetaminophen

An improved testing strategy may improve our ability to detect potential hepatotoxicants (idiosyncratic or classical) but probably not improve our ability to detect immune-mediated injury unrelated to reactive metabolites. Acetaminophen is not commonly associated with idiosyncratic drug reactions. The underlying mechanism of its toxicity is bioactivation of acetaminophen by CYP2E1, CYP1A2, and CYP3A4 pathways to N-acetyl-p-quinonoimine (NAPQI), which as an electrophilic species is thiol-reactive and arylates glutathione and other nonprotein thiols (Cohen et al., 1998). Once the hepatocellular glutathione is sufficiently depleted, NAPQI binds in a covalent manner to cysteine residues, producing an oxidative stress. This may link with precipitation of apoptosis and/or necrosis. Regular use of alcohol or anorexia may be causative in acetaminophen toxicosis, even with therapeutic doses. One proposed mechanism is induction of CYP2E1, depletion of glutathione, and inadequate glucuronidation. In this case, host factors may make a normally “safe” drug elicit significant toxicities via environmental factors.

Clarithromycin

In preclinical models the following effects were described:

Monkeys

Increased ALT, AST, and LDH were reported. Vacuolated cells were found in many organs. Increased lipid droplets in hepatocytes with necrosis were reported. On electron microscopic (EM) evaluation, electron dense material in liver, kidney, cornea, and pancreas corresponding to vacuoles on light microscopy were diagnosed. Cytoplasmic rarefaction of centrilobular hepatocytes was noted but considered reversible after 1 month.

Rats

Increased P450 enzymes and cytochrome b5 were noted at all time points. No increase in LFTs in some studies but increased ALT, AST, LDH, leucine aminopeptidase, and bilirubin (~5× increase all values) were seen and were not considered reversible. Increased liver weights and diffuse microvacuolar steatosis with hepatocellular necrosis and increased lipid in Kupffer cells were reported in most animals. Increased multinucleated hepatocytes with focal necrosis were noted in long term studies. EM demonstrated membranous cytoplasmic bodies, autophagic vacuoles and dense bodies in hepatocytes and bile duct epithelial cells. Neonatal rats dosed for 2 weeks showed multifocal degeneration of bile duct.

Dogs

Increased ALT (2–10×), ALP (5–20×), GGT (3–5×) were noted with repeated dosing. Diffuse pericholangitis with necrotic changes in bile duct epithelium, biliary hyperplasia, intrahepatic cholestasis, hepatocellular necrosis, and Kupffer cell hyperplasia were also seen. EM descriptions included cytoplasmic organelle changes with membranous cytoplasmic bodies, hyperplasia of SER in hepatocytes, and Kupffer cells. Juvenile dogs treated for 3 weeks developed fatty deposition in centrilobular hepatocytes, with inflammatory cells in portal areas.

In vitro assays using liver slices showed increased ALT, AST, and LDH at highest concentrations—erythromycin stearate M< clarithromycin < M1 < erythromycin etiolate.

Human experience with clarithromycin has a confusing picture with regard to hepatic toxicity. In 1993, Yew et al. published results from subjects with Mycobacterium spp. in which they reported elevated liver enzymes in elderly subjects with high doses of clarithromycin. Additional cases have been reported where the pattern was primarily cholestatic with minimal elevations of ALT and AST but significant elevations of alkaline phosphatase and/or GGT. It appears that these subjects had dose-related toxicity, not a hypersensitivity reaction.

Telithromycin

This compound is a ketolide. It differs from other macrolides by the lack of a α-L-cladinose on the erythronolide A ring. The drug is metabolized by the CYP3A4 isoform of cytochrome P450, a potential pathway for the generation of hepatotoxic metabolites, as well as by CYP4501A. It is primarily excreted via the liver and the kidney in all species evaluated. In the various preclinical models the following findings were reported: In monkeys, increased LFTs and total bilirubin were reported. In rats, AST, ALT, and leucine aminopeptidase were increased 2–15×. Hepatic phospholipidosis and moderate to severe hepatic necrosis were noted at doses ~1.8× the human dose.

Dogs

Demonstrated ALT and AST levels increased 4–6×. Phospholipidosis in many tissues was discussed at doses ~0.9× the human dose. In dogs, rats and monkeys, electron microscopic evaluation showed telithromycin stored in lysosomes of hepatocytes, bile duct epithelia, and renal tubular epithelia.

In the clinical trials prior to approval, one subject was reported to have a serious adverse hepatic event with centrilobular necrosis and eosinophilic infiltration in the liver. The sequelae included chronic hepatitis with marked activity and extensive bridging fibrosis. This diagnosis was considered consistent with drug-induced immunologic hepatic injury. Another subject was shown to have markedly increased ALT (13× ULN) and AST (9× ULN) with resolution after several weeks off therapy. Additionally, increased transaminases (>8× ULN) and adverse hepatic events were reported in 43 subjects and more frequently with 7–10 days of dosing compared to 5 days of dosing. More females (24) than males (16) were affected. There appears to be a possible trend towards more reports consistent with a cholestatic pattern of liver injury in males vs. a cytolytic (hepatocellular) pattern in females but the paucity of “hard” data makes this conclusion tentative. Of the 43 subjects, only 14 had information regarding results from an imaging study (ultrasound or CT scan) and hepatitis serologies.

Troglitazone

A more completely characterized hepatotoxin is the thiazolidinedione, troglitazone. Thiazolidinediones, more commonly called glitazones, are the first drugs to specifically target muscular insulin resistance (one of the major underlying metabolic defects in type 2 subjects), associated with increased risk of atherosclerosis and cardiovascular complications in subjects with type 2 diabetes mellitus. Drugs from this class act as ligands for the (subunit of the peroxisome proliferator-activated receptor (PPAR-(), a transcription factor which is directly involved in the genetic regulation of glucose homeostasis and lipid metabolism in fat cells, endothelial cells, macrophages, and smooth muscle cells. Thiazolidinediones have been shown to interfere with expression mediators of insulin resistance originating in adipose tissue (e.g., increased free fatty acids, decreased adiponectin) resulting in improved insulin sensitivity.

Monkeys

In animal studies, it appeared that the cholestatic potential was a possible factor in the hepatotoxicity shown. Chronic monkey studies showed bile duct hyperplasia at doses approximating those in humans with an AUC ~3.8× the human AUC. There was no NOEL assigned to this finding.

Dogs

An increase in ALT was reported from a dog metabolism study at clinically relevant doses. Troglitazone also elicited plasma volume expansion and cardiomegaly and increased fatty marrow in all species tested. The high concentration of troglitazone sulfate, a conjugated metabolite, elicited an interaction with hepatobiliary transport of bile acids at the level of the canalicular bile salt export pump. Activation of the PPARγ with troglitazone has been shown to modulate the profibrinogenic and proinflammatory actions in hepatic stellate cells. This may indicate a possibility that interference with these receptors may cause hepatic toxicity. At high levels, troglitazone induces mitochondrial damage (dissipation of the membrane potential).

Troglitazone elicited life-threatening, idiosyncratic reactions in man. The hepatotoxicity may be due to a reactive quinone, quinone epoxide, or quinone methide and/or reactive isocyanate (from thiazolidine sulfoxide). This drug elicited idiosyncratic injury to the liver of patients via metabolism to potentially reactive intermediates. Troglita-zone is an equal mixture of 4 stereoisomers as there are 2 asymmetric chiral sites. It undergoes extensive metabolism to metabolites: (1) sulfate conjugate at levels 7–10× parent concentration, (2) quinone metabolite approximately equal to parent concentration and (3) glucuronide metabolite with negligible concentrations. These include p-benzoquinone, an o-quinomethide, a quinone epoxide and a reactive α-ketoisocyanate and sulfenic acid intermediate from the thiazolidinedione ring bioactivation. Troglitazone induces CYP3A, the isoform that bioactivates the drug. It is uncertain which of the metabolites is important in eliciting the hepatotoxicity. In subjects with hepatic impairment, the capacity to eliminate the metabolites was impaired but the formation of metabolites was not appreciably affected.

A mild, reversible hepatotoxicity appeared in <2% of the subjects in the clinical trials. In the controlled clinical trials, 1.3% of subjects given troglitazone developed ALT values >3× the ULN. 1% developed ALT values >8× the ULN or more. After drug approval, several cases of severe and sometimes fatal liver failure were reported. Subjects with liver failure showed elevated serum enzymes and bilirubin and other signs of hepatocellular necrosis and cholestasis. Signs appeared at inconsistent times after initiation of therapy and with no consistent dose relationship. There have been speculations about immunological hypersensitivity and metabolic polymorphisms as origins for the severe toxicity but the mechanism remains unknown. Monthly monitoring was not sufficient to protect against adverse events. While the incidence of adverse events appeared to decrease after increased monitoring was instituted, rare cases of severe toxicity (“rapid risers”) occurred within weeks of a normal/minimally elevated ALT. Unlike the other drugs in the class, troglitazone induces cytochrome P450 3A4 which is partially responsible for its metabolism and may predispose to drug interactions. Histologically, troglitazone elicits a pattern of hepatocellular necrosis with bridging necrosis and fibrosis and/or collapse. However, the mechanism of troglitazone-induced fulminant hepatitis is uncertain.

With troglitazone, the unique therapeutic opportunity provided by the product complicated the evaluation of risk:benefit. It was necessary to weigh the long term effects of control of blood sugar in subjects with diabetes mellitus versus the complication of acute liver failure at an incidence of 1:20,000. This strategy may have been effective but compliance with the monitoring and lack of identification of “rapid ALT risers” made it very difficult to protect the population. Additionally, several other products in the class have been approved and provide a more favorable risk:benefit ratio after many months or years of post-marketing experience. In the treated population, the ALT levels with rosiglitazone and pioglitazone are comparable to controls while with troglitazone, ALT levels were ~3× greater than controls.

Kennedy et al. synthesized analogues of N-(3,5-dichlorophenyl)succinimide (NDPS) to test the nephrotoxic potential via metabolism on the succinimide ring. Male F344 rats were given one of the analogues in corn oil, ip, and evaluated diuresis, proteinuria, increased BUN, increased kidney weights and proximal tubular damage. While NDPS produced severe nephrotoxicity, none of the analogues were nephrotoxicants. None elicited increased ALT or liver weights. However, 3-(3,5-dichlorophenyl)-2,4-thiazolidinedione (DCPT) elicited centrilobular necrosis. Thus replacement of the succinimide ring with a thiazolidinedione ring produced a more significant effect on the liver. Since troglitazone has the thiaxolidinedione ring, this analogue (DCPT) may be useful for investigating the toxicity.

Troglitazone might have been better tested in the Type II (KKA) diabetic mouse. The upregulated hepatic PPAR(is the suspected sensitizer to hepatotoxicity and the mechanisms may include increased transcription of genes involved in lipid metabolism, hepatic steatosis, increased sensitivity to oxidative stress, macrophage dysfunction and/or upregulation of mitochondria functionalities. The PPARγ receptor is highly upregulated in the liver of obese and diabetic rodents. Thus, some drugs which are both activators and ligands of this receptor may have little effect in normal, healthy individuals of many species but would elicit increased transcriptional activity in diabetics.

Trovafloxacin

This is a synthetic broad-spectrum antibacterial agent, a fluoronaphthyridone related to the fluoroquinolone antibacterials that differs from other quinolone derivatives by having a 1,8-naphthyridine nucleus.

Trovafloxacin is metabolized by conjugation (the role of cytochrome P 450 oxidative metabolism of trovafloxacin is minimal). Thirteen percent of the administered dose appears in the urine in the form of the ester glucuronide and 9% appears in the feces as the N-acetyl metabolite with 2.5% of the dose found in the serum as the active N-acetyl metabolite. Other minor metabolites (diacid, sulfamate, hydroxycarboxylic acid) have been identified in both urine and feces in small amounts (<4% of the administered dose). Protein binding is moderate (~70%) in most species and the fecal route of elimination is primary in rats, dogs, and humans. Biliary excretion is extensive with glucuronidation being the major metabolic pathway. The plasma t 1/2 was much shorter in animals than in humans.

In a 4-week oral study in rats, a minimal to mild “fatty change” was reported in male rat livers. At a dose of trovafloxacin 10 times the highest human dose based on mg/kg or approximately 5 times based on mg/m², elevated liver enzyme levels that correlated with centrilobular hepatocellular vacuolar degeneration and necrosis were observed in dogs in a 6-month study. A subsequent study demonstrated reversibility of these effects when trovafloxacin was discontinued. In beagles treated for 6 months, the major hepatic finding was vacuolar degeneration and necrosis with mild bridging fibrosis and inflammation from a couple of animals from the high-dose group (~10× human dose). These animals had significantly elevated liver function tests (ALT, ALP, and GGT) during the study. Sequential liver biopsies were taken in a 6-month study where increased liver enzymes ($3× ULN) were reported. In 3/16 animals on study, necrotizing perivenular hepatocellular inflammation was noted in animals with increased serum enzymes but no hepatic effects were noted in animals without increased enzymes.

Human clinical trials showed diverse hepatic events: elevated transaminases alone, cholestasis alone, one case of hepatic necrosis. The increased enzyme levels were dose and time related. In the first quarter after approval, 3× increases in adverse events reported (29 total).

Bromfenac

This cyclo-oxygenase inhibitor is a nonsteroidal anti-inflammatory agent with no narcotic activity. No significant bioaccumulation occurs and it is >99% protein bound. Elimination is primarily through the kidney in monkeys and man. In the mouse carcinogenicity study, cytologic alterations to hepatocytes and ulcers in the GI tract were found. The carcinogenicity study in the rat showed dose-dependent vacuolar change to hepatocytes, cytoplasmic alterations, inflammation and necrosis were noted as well as GI ulcers and papillary necrosis of kidney. As a result of the 2 carcinogenicity studies (1 rat, 1 mouse), it was concluded that the liver, in addition to the previously identified kidney and GI tract, were the target organs of toxicity. Rats: In acute toxicity studies, the kidney and GI tract were the target tissues, with females more susceptible than males to the adverse effects. No hepatic lesions were found in rats treated for <13 months. Vacuolar change, generally mild, was noted in hepatocytes of mice and rats at subtherapeutic doses. Glutathione depletion was found in the liver and kidney of rats.

Monkeys

The 3 month rhesus study revealed enteritis and gastritis, but in the 12-month rhesus study, GI ulcers with mild liver enzyme elevations (first time seen) but no histologic correlates were found. In the preclinical models, the potential for GI effects (ulceration throughout the tract) were approximately the same as for indomethacin. It was not possible to achieve exaggerated dosing in animals making it difficult to determine other toxic effects.

In general, NSAIDs cause ALT elevations in >15% of subjects as an overall generalization.

During clinical trials, bromfenac was associated with reversible, minor elevations in AST. In clinical use patients described, with recommended therapeutic doses of 25–100 mg/d for at least 90 days (labeled recommended duration >10 days), severe symptomatic hepatocellular disease with associated hypoprothrombinemia was reported. There was no evidence of a hypersensitivity reaction. Extensive confluent parenchymal necrosis that originated in central zones was associated with lymphocytic infiltrate with nodular regeneration in 2 subjects with more prolonged disease prior to transplantation.

Histologic lesions with bromfenac in man included acute lobular hepatitis with bridging necrosis and fibrous expansion of portal tracts. Prognosis is usually good with resolution, but fulminant liver failure requiring liver transplant has been recorded after inadvertant rechallenge.

Risk benefit

With bromfenac, continued use was not justified after the delayed idiosyncratic severe hepatotoxicity was identified. This was due to the availability of many alternative treatments.

Isoniazid

Isoniazid is the hydrazide of isonicotinic acid used to treat M. tuberculosis infections.

Rats

In studies with LPS (lipopolysaccharide) doses that elicit a mild inflammation (increased cytokines and COX2 expression), no tissue injury ensues. However, when these doses are coadministered with aflatoxin B or a potentially hepatotoxic drug, the threshold for toxicity is lowered more than 10-fold. Both biliary injury (increased GGT) and hepatocellular necrosis (increased ALT) are demonstrated. In some studies, animals not only demonstrated increased sensitivity to toxins with LPS but also showed a change in tissue target for toxicity. This appears to depend on the agent and the exposure paradigm, thus mimicking the human idiosyncratic reactions.

Isonicotinic acid hydrazine (isoniazid) at 150 mg/kg ip to rats elicited an increased in total plasma lipids, triglycerides, cholesterol, phospholipids and free fatty acids for <30 hours. postdosing. This response was followed by an increase in these same parameters in liver with a decrease in adipose tissue. This implies increased mobilization of depot fat into the liver. Total phospholipids decreased after a decrease in phosphatidyl choline and an increase in phosphatidyl ethanolamine fractions of phospholipids in the liver. These changes were not shown in rabbits given 100 mg/kg ip and these animals were comparable to controls. This experiment demonstrate that isoniazid at higher doses can induce hepatic steatosis in rats as shown by changes in plasma lipid parameters.

Carcinogenesis

Isoniazid has been reported to induce pulmonary tumors in a number of strains of mice.

Drug-drug interactions and drug–disease interactions are sometimes the cause of significant human pathologies. Most of the interactions are due to effects on the cytochrome P450 system that is responsible for the metabolic transformation within hepatocytes by the mixed function oxidase system contained within the smooth endoplasmic reticulum. While a common practice, polypharmacy may affect the rate and extent of reactive metabolite formation. For example, when isoniazid is used concurrently with rifampin, an enzyme inducer, hepatotoxicity is more common. When isoniazid is used concurrently with para-aminosalicylic acid for tuberculosis, hepatic reactions to isoniazid were uncommon as para-aminosalicylic acid is an enzyme inhibitor. Rifampicin is often co-administered with isoniazid. Rifampicin is a high-affinity ligand for the PXR nuclear receptor and a CYP3A4 inducer. Thus, the drugs are more rapidly cleared by CYP-mediated metabolism but there is also an increased formation of reactive metabolites. In most clinical cases, drug–drug interaction is a potential confounder in the interpretation of events but there is often no way to discern the contribution of each drug.

Isoniazid has also had clinically significant interactions with acetaminophen by increasing acetaminophen toxicity Isoniazid has been shown to cause chronic hepatitis with or without cirrhosis, usually with acinar necrosis and inflammation. However the histologic appearance is indistinguishable from viral or auto-immune hepatitides. Isoniazid causes hepatotoxicity in <10% of subjects exposed to the drug. Human polymorphisms in INH acetylation were recognized many years ago and showed that rapid acetylators were more prone to adverse effects. Studies in animals, which identified acetyl-hydrazine as an hepatotoxic metabolite, seemed to support polymorphism in acetylation as a determinant to sensitivity. Large epidemiologic studies failed to confirm the link between rapid acetylation status and hepatotoxicity in man (Gurumurthy) but isoniazid toxicity is usually attributed to genetic polymorphisms eliciting bioactivation via reduced capacity for N-acetylation of the hydrazine component.

Risk factors for isoniazid toxicity include Hepatitis B, Hepatitis C, HIV, alcohol abuse, age, female, slow acetylator phenotype, concurrent use of rifampicin or pyrazinamide. With isoniazid, latency may be shortened when used concurrently with rifampicin. Isoniazid-induced ALT elevation occurs in >20% of subjects exposed to the drug. LPS are commonly detected in human blood with spikes associated with GI distress. LPS is acknowledged as a hepatotoxic factor via the mobilization and activation of macrophages and thus eliciting hepatocyte apoptosis via TNFα mediation. LPS may be an example of an environmental confounder sensitizing the liver to an exaggerated response to xenobiotics.

From the label:

Severe and sometimes fatal hepatitis associated with isoniazid therapy may occur and may develop even after many months of treatment. The risk of developing hepatitis is age related. Approximate case rates by age are: 0 per 1,000 for persons under 20 years of age, 3 per 1,000 for persons in the 20–34 year age group, 12 per 1,000 for persons in the 35–49 year age group, 23 per 1,000 for persons in the 50–64-year age group, and 8 per 1,000 for persons over 65 years of age. The risk of hepatitis is increased with daily consumption of alcohol. Precise data to provide a fatality rate for isoniazid-related hepatitis is not available; however, in a U.S. Public Health Service Surveillance Study of 13,838 persons taking isoniazid, there were 8 deaths among 174 cases of hepatitis.

Therefore, patients given isoniazid should be carefully monitored and interviewed at monthly intervals. Serum transaminase concentration becomes elevated in about 10–20% of patients, usually during the first few months of therapy, but it can occur at any time. Usually enzyme levels return to normal despite continuance of drug, but in some cases progressive liver dysfunction occurs. Patients should be instructed to report immediately any of the prodromal symptoms of hepatitis, such as fatigue, weakness, malaise, anorexia, nausea, or vomiting. If these symptoms appear or if signs suggestive of hepatic damage are detected, isoniazid should be discontinued promptly, since continued use of the drug in these cases has been reported to cause a more severe form of liver damage.

Given the preceding examples and the technologic advances that are occurring in the drug development community, it behooves all of us to help develop a more predictive preclinical schema for hepatotoxic potential in humans. However, it is important to remember that the complex physiologic responses in animals and man will never be completely duplicated in a test tube or on a computer screen. Thus, animal preclinical testing strategies need to maximize the information gathered, especially with respect to hepatotoxic potential, to provide more predictive insights to enable safe and effective drug products to reach the marketplace. Cardiovascular evaluations are rarely done in rodent or lapine models. Increasingly minipigs or dogs are evaluated by telemetry. Equally important is the need to evaluate at multiple time points to include peak plasma concentrations and sequential samplings after that point. Recently it was determined by the ILSI Working Group on QT Evaluation that rabbit Purkinje fibers did not provide very useful information on cardiovascular effects of drugs and recommended that the hERG assay be performed early in the development process for all potential drug candidates. Can we determine or develop similarly sensitive predictors for hepatotoxicity?