Troglitazone and liver injury

Troglitazone (TGZ) (Rezulin; Pfizer, Madison, NJ) was withdrawn from the market by the U.S. Food and Drug Administration (FDA) in March 2000, because it was associated with the development of acute liver failure.1 Approximately 35,000 TGZ injury claims are pending within the United States.2 However, even 4 years later, the mechanisms of TGZ-induced liver injury remain highly controversial and poorly understood. A review of the topic proposed that TGZ hepatotoxicity was mediated through the induction of steatosis and apoptosis and that the injury could be silent (without increased serum aminotransferases, bilirubin, or alkaline phosphatase).3 In addition, it has been suggested that TGZ may induce liver injury as a consequence of mitochondrial damage, oxidative stress, or by inhibiting the bile salt excretory protein (Bsep).3 Furthermore, these effects have been attributed, at least in part, to the accumulation of TGZ in the liver, particularly among patients with liver disease, as well as to the peculiar metabolism of the drug.3 Finally, it has been postulated that liver fibrosis and cancer may result from TGZ hepatotoxicity.3 These presumed chronic complications would constitute an extraordinary medical burden to the approximately 1.92 million patients treated with TGZ. Because these alleged long-term complications of TGZ-induced liver injury have significant medical, social, legal, and financial repercussions, this review evaluates the scientific merits of the proposed mechanisms and suggest alternative explanations. TGZ, troglitazone; FDA, Food and Drug Administration; Bsep, bile salt excretory protein; PPAR, peroxisomal proliferator-activated receptor; ALT, alanine aminotransferase; NASH, nonalcoholic steatohepatitis; Cmax, maximal concentration; TNF, tumor necrosis factor. Insulin resistance can result from an inherited defect in mitochondrial oxidative phosphorylation in children of patients with type 2 diabetes4 or from acquired defects in mitochondrial biogenesis,5 which leads to reductions in mitochondrial function, in elderly individuals.6 TGZ, a peroxisomal proliferator-activated receptor (PPAR)-γ agonist that improved insulin resistance, was used for the treatment of type 2 diabetes7, 8 in the United States by approximately 1.92 million patients from March 1997 to March 2000.9 As a consequence of the improved insulin sensitivity,8 TGZ had beneficial cardiovascular effects,10-13 protected pancreatic β-cell function,14 and corrected metabolic abnormalities associated with polycystic ovarian syndrome15 and lipodystrophy.16 In addition, TGZ antagonized the corticosteroid induction of insulin resistance and abnormal glucose tolerance.17 In clinical trials, 1.9% of the subjects who received TGZ and 0.6% of the subjects who received placebo had elevations (≥3 times the upper limit of normal) of serum alanine aminotransferase (ALT) concentrations.18 In this cohort of 2,510 subjects, TGZ induced overt liver injury and jaundice in two individuals.18 A retrospective analysis also concluded that approximately 2% of the patients developed TGZ-induced liver injury judging by the serum ALT elevations.19 By comparison, serum ALT elevations (≥3 times the upper limit of normal) occurred in approximately 25% of subjects taking tacrine,20 a drug that has not been associated with acute liver failure; however, probably fewer than 100,000 patients have been treated with this drug. Unfortunately, 94 of the 1.92 million patients developed liver failure while taking TGZ.1 Although only 49 of these liver failure cases were considered to be possibly or probably related to TGZ,9 the numbers may be higher because of incompleteness of reporting.1 TGZ treatment was associated with a characteristic hepatocellular injury,18, 21-35 with rare instances of either a mixed hepatocellular/cholestatic injury or a predominant cholestatic reaction.36-38 The liver injury associated with TGZ is idiosyncratic39; it is unpredictable, neither time- nor dose-dependent,40 and cannot be reproduced in animals.41-43 Although the mechanism of liver injury induced by TGZ is believed to be a metabolic idiosyncrasy (nonimmunological),39 some of the case reports had histological evidences suggestive of an immunological reaction16, 25, 27, 36 or responded favorably to corticosteroids.28, 38 One of these patients promptly developed a similar destructive, granulomatous cholangitis when treated subsequently with rosiglitazone, a related PPAR-γ agonist,38 supporting the diagnosis of an immunological reaction and suggesting a class effect. According to Zimmerman and Ishak44 and Klatskin and Conn,45 the occurrence of numerous eosinophils or granulomatous inflammatory infiltrates in the liver is sufficiently compelling, albeit circumstantial, evidence that hypersensitivity may play an important role in the injury provoked by TGZ in some patients. Although valuable to the medical and scientific community, neither the MedWatch Reports to the FDA nor the published case reports provide statistical support for the causal association between TGZ and acute liver failure. In addition, and except in rare instances (e.g., rechallenge38), presumed idiosyncratic liver injuries cannot be attributed unequivocally to TGZ or any other drug.40, 46, 47 As evidence of the diagnostic uncertainties, in the United States, the cases of acute liver failure of undetermined causes (17%) exceed those attributed as probable to all medications combined (13%), excluding acetaminophen.48 Nonetheless, in a prospective study of acute liver failure between 1998 and 2001 at 17 tertiary care centers in the United States, TGZ was prominently implicated, contributing 10% to 15% of the idiosyncratic drug reactions.48 In a small cohort of type 2 diabetic patients and based on a single case of acute liver failure, the incident rate of acute liver failure attributed to TGZ was 240/million person-years,49 compared with the estimated idiopathic acute liver failure of 1 to 2/million person-years.49, 50 Although a study suggested that type 2 diabetes may increase the risk of acute liver failure independently of underlying chronic liver disease or viral hepatitis,51 the study did not eliminate the possible contribution of decompensated chronic liver disease to liver failure. However, in a cohort of diabetic patients treated with hypoglycemic drugs, the incidence of serious liver injury was similar among those treated with sulfonylureas, insulin, metformin. or TGZ (approximately 100/million person-years).52 The estimated risk of unexplained acute liver failure (with coagulopathy and encephalopathy) in type 2 diabetes was approximately 40/million person-years,53 substantially higher than that of 1 to 2/million person-years estimated in the general population.49, 50 It is possible that the retrospective analysis could have introduced some biases,52 because rosiglitazone and pioglitazone have rarely been associated with serious liver injury.49, 50 Another confounding epidemiological variable is the development of subacute liver failure, in the absence of any known cause or TGZ treatment, in approximately 3% of patients with nonalcoholic steatohepatitis (NASH) and unrecognized cirrhosis.54 Thus, TGZ induced serious liver injury in a large number of patients, but the background incidence rate of acute liver failure among type 2 diabetic patients remains to be determined in prospective studies. TGZ is metabolized predominantly by sulfation to metabolite 1 (M1) and, to a lesser extent, by glucuronidation to M2 and by oxidation through cytochrome P450 (CYP)3A4 to a quinone M3.42, 55, 56 TGZ concentration in the liver is essentially the same as that in plasma because these pools are in equilibrium.57-61 In rats, the maximal concentration (CMax) of TGZ in plasma and in the liver was approximately 2.6 μmol/L, whereas the concentration of M1+M2+M3 in the liver was roughly 52 μmol/L.58 Thus, TGZ does not accumulate in the liver, as has been suggested,3 and the actual liver TGZ concentration should be used as a reference for cell culture toxicity studies rather than the combined CMax in the liver for TGZ and its metabolites (approximately 18.5 μg/g, approximately 55 μmol/La).57 Virtually all of the excreted dose in bile was represented by M1, M2, and M3. Indeed, TGZ was rapidly metabolized in the rat to M1 and M2. Thirty minutes after the intravenous administration of TGZ (at a dose approximately 3-fold greater than that given to patients), which is expected to result in much higher plasma TGZ concentrations than when administered orally,62 the concentration of TGZ in the liver was undetectable in male and was approximately 10 μmol/L in female rats.60 The diabetic KK mice had a faster metabolism of TGZ than normal mice, with approximately 66% of the dose being metabolized after 1 hour.57 Experiments in diabetic KK mice showed that the combined CMax for TGZ and its metabolites in the liver was lower (approximately 23 μmol/L) than in normal animals.57 Studies of up to 52 weeks in monkeys, the closest animal model to humans for the study of TGZ,42, 43 showed that at doses 160-fold higher than that given to patients, TGZ was metabolized efficiently but the sulfate M1 accumulated in plasma.42 In humans, after reaching a steady state, the plasma CMax for TGZ (approximately 3-6 μmol/L) was approximately 12% of the CMax for TGZ + M1 + M3.63 This indicates a very efficient metabolism of TGZ by the liver, supporting the notion that hepatic TGZ in humans is in equilibrium with plasma TGZ.59 The peak level occurs approximately 4 hours after taking TGZ, and the plasma concentration drops progressively from 4 to 24 hours, which would also decrease the liver concentration concomitantly.64, 65 Moreover, because TGZ is tightly bound to albumin in plasma (95%-99.8%), only 0.2% to 5% is available as free plasma TGZ capable of a steady-state equilibrium with hepatocytes.59, 63 It has been speculated that TGZ levels in the livers of patients with alcohol-induced liver disease and NASH should be higher than in normal individuals, leading to greater rates of liver injury in these patients.3 However, this hypothesis is not supported by the available experimental data. In patients with type 2 diabetes, the pharmacokinetics of TGZ were normal, and the plasma CMax (and consequently, the liver CMax) for TGZ, M1, and M3 were not different from those of healthy normal individuals.66 The CMax for TGZ was 16% and 39% lower in cirrhotic patients with moderate and severe liver impairment, respectively.64 The CMax of the quinone M3 was normal, whereas the CMax for the sulfate M1 was increased in cirrhotic patients.64 However, the M1 metabolite showed no toxicity in cultured hepatocytes,55 in HepG2 cells,56 or in monkeys, even at much higher plasma concentrations.42, 43 Indeed, the total integrated plasma concentration for M1 among cirrhotic patients was less than 10% of that in monkeys receiving an extremely high TGZ dose for 52 weeks.43 Likewise, the quinone M3 was not toxic to cultured rat hepatocytes even at concentrations of 100 μmol/L (approximately 50-fold the liver concentration).67 AlthoughTGZ has been proposed to exacerbate fatty liver by producing oxidative stress and inflammation in the liver,3 TGZ actually has potent antioxidant and anti-inflammatory properties. Increased oxidative stress can lead to mitochondrial damage within the cell, and, in turn, decreased mitochondrial oxidation of fatty acids results in the accumulation of triglycerides. Therefore, increased triglyceride concentration in the liver is an indicator of mitochondrial damage.68, 69 TGZ treatment induced fatty liver in both diabetic KK mice, and in hyperlipidemic (NZO × NON) F1 mice,70, 71 but the hepatic triglyceride concentration remained unchanged or was not reported,70, 71 reflecting the accumulation of unidentified lipids, apparently, unrelated to mitochondrial injury. Moreover, similar fatty changes in the liver occurred after the administration of rosiglitazone or pioglitazone.70-72 Therefore, these mouse models are poor predictors of the liver abnormalities associated with TGZ given that they do not reflect the pathological conditions observed in humans and that the current glitazones, which have not been associated with acute liver failure, induce similar abnormalities. TGZ, at concentrations relevant to human pharmacokinetics, did not induce changes in mitochondrial oxidation in normal cells.73 Although TGZ inhibited fatty acid oxidation and esterification in isolated hepatocytes from starved rats,74 this effect occurred at a concentration approximately 200-fold greater than that achieved therapeutically in the liver.58 Futhermore, TGZ prevented mitochondrial abnormalities and apoptosis, as well as decreased liver and pancreatic islet triglyceride concentration in Zucker diabetic rats and Long-Evans fatty rats.75-77 More importantly, TGZ decreased the liver size and fat accumulation in the livers of patients with insulin resistance and type 2 diabetes or lipodystrophy syndrome, as determined by magnetic resonance imaging and computed tomography scans.16, 78 In addition, TGZ improved mitochondrial oxidation in insulin-resistant patients,16 presumably through PPAR-γ signaling, which is consistent with the stimulation of mitochondrial synthesis and function by PPAR-γ coactivator-1α.5 The activation and induction of PPAR-γ caused by TGZ is also observed with the antioxidant vitamin E79 and anti-inflammatory drugs such as ibuprofen,80, 81 in the absence of cytotoxicity. TGZ is a potent anti-inflammatory drug because it blocks cytokine- and lipopolysaccharide-induced cytotoxicity, activation of macrophages and nitric oxide synthase and chemokine expression.81-87 Like vitamin E, TGZ acts as an antioxidant by inhibiting the oxidation of low-density lipoprotein cholesterol and by blocking the reactive products of oxidative stress.88, 89 These antioxidant and anti-inflammatory effects81, 84, 88, 89 may explain the improved patency of both carotid arteries and coronary artery stents in type 2 diabetic patients who are taking TGZ.90, 91 These beneficial effects of TGZ in patients and animals with insulin resistance are inconsistent with the hypothesis that TGZ induces oxidative stress and mitochondrial abnormalities,3 except possibly in rare, susceptible individuals. The proposal that TGZ and the sulfate M1 may induce intrahepatic cholestasis, contributing indirectly to the development of liver injury in patients taking TGZ,3 is plausible because in isolated rat liver canalicular preparations, TGZ inhibited Bsep.60 Also, impaired biliary excretion of taurocholic acid was observed in isolated perfused livers under albumin-free conditions.60, 92 However, this cholestatic effect of TGZ was prevented by adding albumin to the perfusate.92 This emphasizes the critical role of protein binding in the analysis of TGZ toxicity.59, 65 Although the intravenous administration of TGZ to rats induced acute cholestasis,60 there is no evidence that these effects of TGZ occur after oral administration in chronic studies in animals or in patients. Comparable experimental results have been observed with other drugs, including cyclosporin A, rifampicin, and glyburide, but there is rarely clinical correlation.93 For example, rifampicin inhibits Bsep in experimental studies94 but ameliorates pruritus (induced by bile acids) in patients with cholestasis.93, 95 This effect is achieved through the stimulation of CYP3A4 activity by rifampicin.96 In turn, 6-α hydroxylation of bile acids by CYP3A4 and subsequent conjugation at C6 by UDP-glucuronosyltransferase increases the renal clearance of bile acids.97, 98 The beneficial effects of rifampicin on bile acid excretion in patients, and the apparent discrepancy between animal and human studies, can be readily explained by the fact that rifampicin is a potent activator of human, but not mouse, pregnane X receptor, which induces CYP3A4.96 In other instances, the correlation between experimental inhibition of Bsep and clinical cholestasis is higher. Because bosentan, an endothelin-1 receptor antagonist that inhibits Bsep in experimental models, can induce cholestasis and liver injury in humans,99 it is used with strict monitoring requirements in patients with pulmonary arterial hypertension.100 The cholestatic effects of TGZ have not been confirmed with chronic studies in animals and humans. Increased serum alkaline phosphatase is a sensitive indicator of cholestasis,101 because bile acids induce the expression of alkaline phosphatase in the liver by enhancing mRNA translation.102 TGZ treatment in monkeys (60- to 120-fold the therapeutic dose) for up to 52 weeks did not increase serum alkaline phosphatase levels.42, 43 In addition, administration of TGZ to monkeys (approximately 20-fold the therapeutic dose) for 4 weeks did not affect sulfobromophthalein plasma clearance, suggesting that TGZ did not impair the hepatic uptake, transport, and biliary excretion of this organic anion.42 Similarly, the serum alkaline phosphatase levels were not increased in the cohort of subjects given TGZ in the clinical trials.103 Alternatively, compensatory mechanisms for the inhibition of Bsep activity by TGZ could exist. These may include the induction of CYP3A4 by TGZ and bile acids,104, 105 which would result in an increased renal excretion of C6-glucuronidated bile acids95-98 and the prevention of hepatocellular damage and inflammation, such as that associated with cholestasis in bile duct–ligated animals.106 PPAR-γ activation is important because it inhibits tumor necrosis factor alpha (TNF-α) production, a contributor to liver inflammation,107 and blocks lipopolysaccharide signaling, a mediator of hepatic injury in cholestasis.108 In brief, if TGZ induced liver injury in humans by affecting Bsep-dependent bile acid excretion or by any other cholestatic mechanism, the event has been rare because cholestasis was unusual among the case reports.36-38 Cell studies are important to understand the effects of drugs on molecular pathways, but the FDA does not require cell studies and considers animal studies sufficient for preclinical assessment of drug toxicity http://www.fda.gov/cder/guidance/index.htm. It has been argued that apoptosis may be a mechanism responsible for TGZ-induced liver injury.3 Although TGZ has been shown to induce cell death in various cancer cell lines and tumors in animals, it did not affect the corresponding normal primary cells or tissues.109-114 Moreover, these studies with tumoral cells cannot be reliably extrapolated to humans because cancer cells have abnormal mitochondrial morphology, protein content, DNA, and metabolism.115 In addition, TGZ-induced mitochondrial abnormalities in cancer or transformed cell lines were not associated with either apoptosis or cytochrome c release from mitochondria,116, 117 an activator of caspase 9 (a mitochondrial apoptotic pathway).118 Rather, TGZ facilitated apoptosis initiated by TNF-related apoptosis-inducing ligand, at least in some tumor cell lines, by increasing degradation of FLICE-like inhibitory protein long form (FLIPL),109 an inhibitor of caspase 8 (a non-mitochondrial apoptotic pathway).119 Induction of toxicity and cell death by TGZ in normal cells also has been demonstrated, but either with concentrations higher than those achieved in animals and patients given therapeutic doses of TGZ55, 120-122 or in cells that were cultured in the absence or with very low concentrations of albumin (or fetal calf serum).120, 122-124 (Table 1). By contrast, when normal cells, including primary hepatocytes, endothelial cells, peripheral blood monocytes, bone and cells, were cultured with the albumin and TGZ concentrations of TGZ pharmacokinetics, there was no even in the of TNF-related apoptosis-inducing a death receptor agonist that the apoptotic 1). Furthermore, TGZ toxicity to hepatocytes cultured which the TGZ by up to approximately was when albumin was to the TGZ was to human hepatocytes, judging by the decrease in but the of these results is because of the is associated with a of TGZ induced toxicity at concentrations approximately 20-fold greater than that achieved therapeutically in the the absence of albumin in the culture increases TGZ to and rosiglitazone also induced cytotoxicity, albeit with a higher concentration model for TGZ-induced liver toxicity could include highly human hepatocytes, with other liver cells or given role on injury and as well as mice Although apoptosis occurs rarely in normal at least apoptotic was in the of of the hepatic in normal human and rat in all including the liver, can be by because of the characteristic pathological such as and and apoptosis was only in case it was increased from background and with normal by of the apoptotic have been in the liver or other after TGZ treatment, in the other case or in animal The of in the absence of any pathological evidence of apoptosis, is in a case reports in association with severe necrosis and 27, The that TGZ-induced liver injury increased serum aminotransferases, bilirubin, or alkaline is not supported by the TGZ did not induce liver apoptosis in and stimulation of apoptosis by bile or A in mice is not silent because it in increases in serum TGZ was given to patients with NASH, a significant in serum ALT and a in the of necrosis and inflammation was observed in of the liver The other PPAR-γ rosiglitazone and also hepatic inflammation in patients with the TGZ treatment, patients did not have evidences of liver apoptosis by or changes and which are only in approximately of liver among NASH were not by TGZ However, the small number of patients could have the statistical The of these mitochondrial is may reflect changes because they also occurred after treatment or in rats, associated with decreased mitochondrial Thus, pathological liver apoptosis in patients treated with TGZ does not to be a mechanism for TGZ-induced liver injury. The of chronic liver injury is which apparent increase in fibrosis is expected in from patients with liver necrosis caused by Therefore, the diagnosis of in patients with liver necrosis cannot be In a only changes occurred in the liver fibrosis in NASH patients treated with reflecting in the hepatic cells are responsible for the of Thus, of the or of a liver activation of cells cells is for the development of liver Indeed, this is the of therapeutic to liver fibrosis by blocking cell activation by PPAR-γ It has been suggested that apoptosis, associated with increased serum ALT and inflammation, may be a to liver Moreover, apoptotic by a human cell induced expression of type apoptotic would be by cells in the liver and in the of macrophages and in the absence of an inflammatory remains to be is in and of the after cell death would the For example, the of TGZ with acid induced apoptosis and fibrosis in tumors with TGZ (at a dose approximately 200-fold greater than the therapeutic dose) was not associated with either apoptosis or fibrosis of these of the critical for the induction of type 1 expression and cell activation can be with TGZ and related PPAR-γ also the activation of hepatic cells, inhibiting the of the that the of the The that TGZ could have induced cancer in is not supported by the experimental data. Increased incidence of liver and hepatocellular occurred with TGZ in mice at and times the human therapeutic tumors of any type were induced by TGZ in rats at times or in monkeys at times the human therapeutic Furthermore, TGZ was neither in nor in bone of mice or rats, no In addition, of the tumors in mice had and fewer than 5% had of and Thus, there is no evidence that TGZ the development of liver fibrosis and or the induction of liver tumors in The metabolism of TGZ by the CYP3A4 in hepatocytes, is a for approximately of all as well as and bile and some and cardiovascular for example, the induction of CYP3A4 with which would increase TGZ metabolism M3. to the of the variable to TGZ have been in critical for the activity of This is important because the hepatic expression of CYP3A4 approximately and the in CYP3A4 function at least 20-fold among on the drug to of the in function is caused by of of the affect various and could have to the to which would TGZ metabolism can be determined in by a based on of The quinone which has a similar to that of vitamin has not been to be toxic after treatment of monkeys with up to the therapeutic TGZ 43 However, in susceptible individuals, M3 could with or to induce liver injury. Also, reactive are bound to protein and but the of these in the of is In a cohort of patients, combined were associated with increased for TGZ-induced liver single for drug metabolic risk for type 2 insulin and In a small of patients, a correlation was also observed between TGZ-induced liver injury and the combined The mechanisms by which TGZ caused severe liver injury remain but they may include or acquired Bsep, or evidence that TGZ induced pathological apoptosis of hepatocytes in culture experimental in animals or in patients. In addition, no scientific evidence the that it in the that it oxidative stress in the that it the in hepatocytes in transformed or cancer that it induces cholestasis in patients in rare that it steatosis in the that it causes silent liver and that patients are at risk of and liver The liver injury induced by TGZ was hepatocellular and to a metabolic Although in some cases evidence for an immunological this mechanism cannot explain of the of the mechanisms responsible for TGZ-induced liver from to the risk of similar with other The of for critical review of the and to of of of of and and Drug for valuable The for the of the prevented the of important