Abstract
Clinical presentations of drug-induced liver injury (DILI) cover essentially the entire spectrum of known liver diseases. However, in the last 8 years the form of liver injury that has most frequently resulted in labeling restrictions is idiosyncratic hepatocellular injury leading to acute liver failure. This rare form of DILI has a characteristic clinical presentation that includes an acute onset after uneventful treatment with drug for weeks to months. Serum alanine aminotransferase rises to very high levels and the appearance of jaundice indicates a high mortality even if the therapy is discontinued. Drugs that can cause this type of injury almost always are associated with frequent (2–15% of all treated patients) and minor serum aminotransferase elevations. These elevations are believed to reflect true liver injury, but often reverse even if drug therapy is continued. The basis for this “adaptation” is not known, as is why some patients do not adapt and develop progressive liver injury. Understanding how drugs cause severe idiosyncratic hepatocellular toxicity has been frustrated by the lack of good preclinical models. Indeed, because these events occur so rarely, the vast majority of humans are not good models. Studies of genomic DNA from affected individuals should provide important insight but not the complete answer because environmental factors almost certainly contribute to individual susceptibility. The most fruitful approach may therefore lie in focused and well-controlled phenotype/genotype studies of the rare patients who have survived this type of injury. The National Institute of Diabetes and Digestive and Kidney Diseases of The National Institutes of Health has recently sponsored a cooperative agreement (UO1) to create a Drug Induced Liver Injury Network (DILIN). DILIN consists of University of Michigan, Indiana University, University of Connecticut, University of California, San Francisco, University of North Carolina, and Duke University. This network should provide heretofore missing resources required to address the problem.
Introduction
Drug-induced liver injury (DILI) remains the single most common adverse effect that can result in failure to obtain regulatory approval to market a new drug, and postmarketing regulatory actions including labeling restrictions and withdrawal from the marketplace. Although this has been true for many years, interest in DILI has increased substantially recently. A number of factors have contributed to this increased interest in DILI, principal among them has been the experience with the drug troglitazone (Rezulin).
The Troglitazone Story
Troglitazone is an orally administered PPAR-γ agonist that was used to treat type II diabetes. It was the first in its class approved by the FDA for marketing in 1997, and was enthusiastically received by doctors and patients. Within less than 1 year of product launch however, the company began to receive reports of acute liver failure associated with troglitazone treatment. Diabetics have a high background incidence of liver disease relative to most other patient populations (El-Serag et al., 2004), and this is largely attributable to a high rate of nonalcoholic fatty liver disease (Clark and Diehl, 2002) and an increased rate of chronic hepatitis C infection (Bahtiyar et al., 2004). The diabetes-associated liver diseases, however, would not be expected to cause acute liver failure.
Liver injury produced by a drug often has a characteristic clinical presentation or “signature,” and this is also the case with troglitazone. The troglitazone “signature” was the relatively acute onset of hepatocellular injury generally occurring within 1 to 7 months after starting treatment with progression on to liver failure if treatment was not stopped in time (Graham et al., 2003). Figure 1 illustrates some of the important features of the Rezulin signature. Liver chemistries remained normal during the first few weeks or months of treatment, with the injury first heralded by relatively high elevations in serum ALT and AST. Elevation of serum bilirubin was delayed relative to the transaminase elevations, with jaundice typically appearing weeks after the onset of transaminase elevations. Elevations in serum alkaline phosphatase were generally mild and a late phenomenon. Progressive increases in ALT and AST often occurred for days or weeks after stopping the drug treatment. In the hypothetical case shown (Figure 1), the injury progressed to total liver failure and death as evidenced by the progressive rise in serum bilirubin as serum transaminases are falling. In this case, the fall in serum amino-transferases reflects a paucity of functioning liver cells and not resolution of injury.
It should be noted that all cases of liver injury attributed to any given drug need not always mirror the classic signature presentation of that drug, and this may also be the case with troglitazone (Menon et al., 2001; Bonkovsky et al., 2002). Nonetheless, the majority of cases of troglitazone liver injury appear to follow the signature pattern illustrated.
With the recognition that troglitazone could cause acute liver failure, routine monitoring of serum liver chemistries was recommended. When the next-in-class compounds entered the market and appeared to have improved liver safety (Isley, 2003), troglitazone was withdrawn from the market.
The troglitazone experience has had a major impact on the pharmaceutical industry. A series of news articles originating from the Los Angeles Times, but widely published in other U.S. newspapers, brought public attention to troglitazone, and cast in a negative light certain aspects of troglitazone’s development, FDA review, and FDA approval (Rezulin, 2000). This negative publicity has contributed to a heightened awareness of liver safety issues at the Food and Drug Administration, resulting in higher scrutiny of liver safety databases in NDA applications.
Another strong influence on the pharmaceutical industry has been ongoing and extensive litigation regarding patient claims of liver injury related to troglitazone (〈http://www.injuryboard.com/view.cfm/TOPIC-635〉). This has created a large cohort of plaintiff attorneys who are well-versed on issues regarding the liver safety of drugs. These attorneys have established relationships with expert witnesses, and have in hand extensive expert reports and source documentation largely applicable to DILI from any drug. As troglitazone litigation winds down, there has been concern in the pharmaceutical industry that liver injury due to other drugs will become a preferred target for costly litigation. It is understandable that those in decision-making capacities in pharmaceutical companies have placed special emphasis on attempting to screen out new molecular entities with potential liver safety issues.
One fortunate consequence of the troglitazone litigation is that virtually all of the data relevant to the liver safety of troglitazone has been “produced” and reviewed by multiple experts around the world. This ‘postmortem’ examination of troglitazone is largely unpublished, but some comments can be made. First, the standard preclinical toxicity screening performed, which included toxicity studies in rat, dog, and monkey, failed to detect the potential for troglitazone to produce irreversible liver injury. Second, a “liver signal” was not perceived in the clinical trials database, and there were several reasons for this. The incidence of ALT elevations greater than the upper limit of normal was actually higher in the patients receiving placebo than in those receiving troglitazone (unpublished data). This apparent “hepatoprotective” effect of troglitazone probably relates to the ability of PPAR-γ agonists to improve fatty liver disease (Liangpunsakul and Chalasani, 2003). ALT elevations greater than 3 times the upper limit of normal were observed more frequently in troglitazone-treated (approximately 2%) versus placebo (0.5%) patients (Watkins and Whitcomb, 1998). However, since there were no stopping rules based on serum ALT in the clinical trials, some patients experiencing ALT elevations greater than 10 times the upper limit of normal were continued on troglitazone treatment, and serum ALT returned to baseline in spite of continued treatment (Watkins and Whitcomb, 1998). This observation was interpreted as indicating either that the elevations were not due to the drug, or that the elevations were not the result of true liver injury.
Two patients in the NDA database did experience hepato-cellular jaundice and in retrospect, these events are consistent with the troglitazone “signature” (Watkins and Whitcomb, 1998). However, in both of these patients, serum ALT continued to demonstrate rises days after discontinuing therapy. This was interpreted as evidence that the liver injury was not related to troglitazone treatment, although we now know that this is quite consistent with troglitazone liver injury.
The late Hyman Zimmerman first noted that patients coming to medical attention with drug induced hepatocellular jaundice have at least a 10% mortality rate (Black et al., 1975). Since the 2 patients with hepatocellular jaundice in the troglitazone trials represent approximately 0.1% of the treated patients, the Zimmerman observation (which is now dubbed “Hy’s Rule” (Reuben, 2004)) would predict that fatal liver injury could occur in approximately 1 in 10,000 treated patients. This estimate may be near the actual rate observed postmarketing (Graham et al., 2003). The troglitazone experience has therefore been viewed as validating “Hy’s Rule” in the clinical trial setting. Medical examiners at the FDA are now instructed to search safety databases for individuals experiencing bilirubin elevations in the setting of hepatocellular injury (Temple, 2001). The finding of even a single such patient (without an alternate explanation for the abnormalities) can mean serious difficulty with the approval process.
As a result of the troglitazone experience and the effect it has had in intensifying review of liver safety data at the FDA, there has been intense interest in improving understanding DILI with the goal of improving preclinical and clinical means of assessing liver safety.
Other Liver-Related Regulatory Actions
Table 1 lists major regulatory actions that have resulted from DILI over the 8-year period beginning in 1996. It is interesting to note that with the exception of terbenafine (mixed hepatocellular/cholestatic injury (Ajit et al., 2003) and valproic acid (microvesicular steatosis (Eadie et al., 1988)), all other drugs listed typically produce a predominantly hepatocellular injury capable of causing acute liver failure (〈http://www.fda.gov/medwatch/safety.htm〉). In other words, although DILI can mimic every known histopathologic and clinical presentation of liver disease, hepatocellular injury is the predominant “signature” presentation for all but 2 of the 16 drugs undergoing major regulatory actions during this period. Furthermore, with the exception of acetaminophen, the drugs causing hepatocellular injury would also be considered to cause “idiosyncratic” injury. In other words, liver injury caused by the drug is not clearly dose-related and occurs in a very small subset of all treated patients. Such unusual patients are presumed to have a rare blend of genetic and/or environmental factors that render them susceptible.
In reviewing the data available in the literature on, or prescribing information for, these 13 drugs, several general statements can be made. First, with the possible exception of felbamate, each of these drugs was associated with an increased incidence of serum ALT elevations greater than 3 times the upper limit of normal relative to placebo treatment. With felbamate, the control patients were receiving other antiseizure drugs known to be associated with ALT elevations, and this may account for the apparent lack of difference between treatment and controls. A second general statement is that the majority of patients who experience ALT elevations are not at risk of developing significant liver injury. That is, the majority of patients who experience ALT elevations due to these drugs will have resolution of the liver injury despite continued exposure to the drug. Indeed, the ability to adapt to DILI appears to be a general phenomenon for drugs capable of causing severe idiosyncratic injury. For example, approximately 15% of patients treated with isoniazid will experience ALT elevations greater than 3 times the upper limits of normal. However, less than 1:100 will develop symptomatic hepatitis with continued treatment (Black et al., 1975).
The explanation for these reversible ALT elevations observed during continued drug therapy include the possibility that the mechanisms underlying them have no relationship at all to liver injury capable of progressing to liver failure. The alternate and generally favored explanation is that patients capable of developing progressive liver injury represent a subset of those patients with ALT elevations. Support for this conclusion is that the ALT elevations observed in the troglitazone clinical trials (Watkins and Whitcomb, 1998) were similar in timing to the signature pattern of acute liver failure (Figure 1). The concept that patients capable of developing irreversible liver injury represent a subset of those developing ALT elevations is illustrated in Figure 2.
The mechanisms potentially underlying this “adaptation phenomena” are not known. Figure 3 shows multiple factors that could underlie interindividual differences in susceptibility to liver injury from a given medication. The hepatocellular injury indicated by rising serum ALT should have no long-term health consequences to the individual if adaptation occurs. The most important variable in understanding susceptibility to significant DILI may therefore be the individual patient’s ability to adapt to the initial injury. Indeed, an intriguing possibility is that a single inherited defect in adaptation is a prerequisite for progression of DILI independent of the causative drug. This would be analogous to hemolytic anemia produced by multiple different drugs in patients with G-6-PD deficiency. Discovery of the mechanism’s underlying adaptation may therefore provide a critical insight into idiosyncrasy.
Unfortunately, there are currently few animal models that could be used to address this adaptation to DILI. Development of such models is one of the defined research goals of the recently released “Action Plan for Liver Disease Research” by the National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK) (〈http://www.niddk.nih.gov/fund/divisions/ddn/ldrb/topics.htm〉). Adaptation generally cannot be studied in clinical trials because treatment is usually stopped when the serum ALT rises to greater than 3 times the upper limit of normal. It is therefore not possible to distinguish patients who were capable of progressive liver injury versus those who would adapt had therapy been continued. Identifying such individuals will be facilitated by the recent creation of the Drug-Induced Liver Injury Network (DILIN) (Figure 4).
Drug-Induced Liver Injury Network (DILIN)
This network has initiated 2 studies in the summer of 2004 (〈http://dilin.dcri.duke.edu〉). The first study undertaken by this network, termed the Retrospective Study, is creating a registry of patients who sustained severe and potentially irreversible liver injury due to isoniazid, phenytoin, valproic acid, and combination amoxycillin/clavulanic acid at any time since 1994. These drugs were chosen because they have characteristic signatures in terms of clinical presentation; the drugs are often administered to healthy individuals not likely to be receiving other hepatotoxic medications; and, because some information concerning mechanisms of liver injury have been proposed. The second study undertaken by DILIN is prospectively enrolling patients who have clinically significant DILI from any drug. In both studies, plasma and genomic DNA will be prepared and lymphocytes immortalized for future studies. Liver tissue will also be obtained where possible. In addition, the subjects agree to be contacted yearly for up to 20 years and may be offered enrollment in phenotype/genotype studies.
Summary
The great current focus on DILI research is, in large part, due to the ramifications of the troglitazone experience. Current preclinical and clinical liver safety testing does an excellent job in keeping from the public drugs capable of causing many different forms of liver injury. However, these practices have not been completely successful in avoiding drugs capable of causing acute idiosyncratic hepatocellular injury. Defining the mechanisms that initiate this type of injury, and the adaptation that usually follows, will be extremely challenging. However, the scientific payoff is likely to be great and go far beyond the area of DILI. Indeed, idiosyncratic hepato-cellular injury due to drugs is an ideal model for studying susceptibility to all environmental diseases. This is because this research involves large populations with a known exposure to a chemically defined xenobiotic (i.e., the drug). Biomarkers for liver injury (serum liver chemistries) are cheap, safe, and sensitive and can therefore be repetitively administered. It seems likely that the environmental and genetic factors found to underlie idiosyncratic liver injury will also be found to underlie environmental illnesses affecting many other organs.