Using controlled clinical trials to learn more about acute drug-induced liver injury


  • Potential conflict of interest: Although Dr. Pears is an employee of AstraZeneca, and Dr. Watkins is an academic consultant to several companies, the use of no particular drug was advocated or discouraged at this open, public meeting on general issues relating to drug-induced liver injury. Drs. Seligman, Avigan, and Senior are employees of the U.S. Food and Drug Administration and report no conflicts of interest.


Drug-induced liver injury (DILI) is of major interest to hepatologists and clinicians in general, patients, government regulators, and the pharmaceutical industry. Understanding why this form of injury occurs only in certain individuals has major implications for the development and availability of drug therapies and in the prevention of these events. A single controlled clinical trial may be unlikely to show cases of such rare events, but in the aggregate, clinical trials offer a unique resource for learning more about individual susceptibility and developing truly predictive new biomarkers for DILI. We pose the question as to whether clinical trials could be modified or improved to provide data that would better answer some of the outstanding issues. At a recent (March 2008) public meeting, experts from academia, industry, and regulatory bodies discussed several major issues regarding liver safety in clinical trials including: what signals of liver injury should justify stopping administration of study drug or allowing it to continue; if deliberate rechallenge should be done and under what circumstances; whether patients with liver disease should be included in clinical trials; and what kinds of new biomarkers will be needed to answer these questions more clearly. Past clinical trials have not provided data to settle those issues, and reliance has defaulted to consensus of expert opinions. Modified and better clinical trials with standardized collection of data and biospecimens are probably the best source of new and potentially valuable information to supplant current rules based on consensus of expert opinions and to understand by what mechanisms and how to distinguish those individuals who are susceptible to severe DILI. (HEPATOLOGY 2008.)

Drug-induced liver injury (DILI) is the adverse event that most frequently leads to regulatory action on drugs, including failure to approve, added label warnings, and withdrawal from the market1 and is now the leading cause of acute liver failure in the United States, exceeding all other causes combined.2 The major form of severe liver injury of concern is idiosyncratic hepatocellular injury with loss of liver function that occurs unpredictably in only a few of the many patients treated with the drug after its approval for clinical use and marketing. In addition to the threat DILI presents to the rare susceptible patient, withdrawal of drugs from the market or limitations in their availability because of DILI in a few denies beneficial treatment to the great majority of those who are not susceptible to DILI and who could take them safely.

The recent U.S. Food and Drug Administration (FDA) draft guidance document3 on “Drug-Induced Liver Injury: Premarketing Clinical Evaluation”, represents a formal recommendation regarding consistent approaches to detecting and investigating liver safety in clinical trials. At an open public meeting in March 2008, DILI experts from academia, the pharmaceutical industry, and the FDA met to discuss aspects of the draft guidance.4 Discussion was focused around four questions, the answers to which could improve the early detection and characterization of DILI: (1) when should the administration of an investigational drug to a participant be stopped during a trial? (2) should rechallenge be used? (3) should patients with stable underlying liver disease be included in controlled trials? and (4) how can clinical trials be used to identify better biomarkers for DILI? Complete slide presentations and discussion transcripts are available online.5 The following comments represent our interpretation of the significance and the state of understanding about those DILI issues arising from those discussions.

1. When Should an Investigational Drug Be Stopped?

A. What to Look for in a Patient or Subject During a Trial

Serum alanine aminotransferase (ALT) activity is the most frequently used biomarker to detect hepatocellular injury in clinical trials and is somewhat more liver-specific than aspartate aminotransferase (AST). However, both aminotransferases are widely distributed in other tissues,6 and ALT isoforms have recently been found.7 The absence of ALT elevations in a clinical trial is reassuring in terms of liver safety. Conversely, high incidence and magnitude of ALT elevations in a study population are clearly worrisome.

One problem with using serum ALT as a primary indicator of liver injury is that it is a volatile measurement that can rise or fall in hours or days. Single measurements of serum ALT activity on a given day do not establish peak values; multiple, close, serial measures are required. Another problem is that even fairly high elevations do not always indicate or predict serious DILI and hence the degree of risk to the drug recipient. A good example of this is shown by the clinical trials of tacrine, a drug proposed for treating Alzheimer's disease, in which about 50% of the patients developed ALT elevations,8 25% exceeding three times upper limit of normal (3× ULN) and 2% >20× ULN, despite close monitoring and stopping rules based on ALT levels. Nonetheless, no cases of symptomatic liver injury or jaundice were observed in the U.S. clinical trials of tacrine, and there have been very few postapproval reports of significant liver injury from the drug.

On the other hand, type-2 diabetics treated in clinical trials with troglitazone demonstrated much less frequent ALT elevations, compared to the tacrine trials.9 Fewer than 2% of subjects treated with troglitazone developed ALT increases >3× ULN and only 0.2% developed ALT elevations exceeding 20× ULN. However, two of 2502 patients in the troglitazone clinical trials developed hepatocellular injury with jaundice, indicating serious liver injury. Troglitazone was later withdrawn from the market after 89 cases of acute liver failure with 65 deaths and eight individuals who survived transplantation were reported in association with its marketed use.10 Both incidence and magnitude of observed ALT elevations were greater in tacrine trials than in the troglitazone trials, yet the potential for serious DILI appears negligible with tacrine. This discrepancy cannot be explained by differences in frequency or number of measurements, or by stopping rules based on ALT elevations. Currently, trials often use an ALT of 3× ULN as the criterion for stopping drug administration and removing a patient from the study, based on a 1978 expert consensus11 that this level be considered “markedly abnormal”.

B. Tolerance, Adaptability, and Susceptibility: Person-to-Person Differences

Serum ALT elevations caused by drugs often resolve despite continued treatment with an offending drug. Transient rises in serum ALT may reflect true liver injury by the drug, followed by a period in which the liver may “adapt” such that the injury subsides despite continuing treatment with the causative agent. The predictive value of isolated elevations of serum liver aminotransferases for clinically serious DILI is therefore poor. An observation documenting adaptation to DILI was made by Mitchell and colleagues12 involving the drug isoniazid. Following an outbreak of tuberculosis in a mental hospital, 201 men who were tuberculin-positive by skin test were begun on prophylaxis with isoniazid. Monthly serum samples were collected for a year from these patients, and from 107 patients who were negative for skin test and not treated with isoniazid but housed in the same facility. The sera were frozen and stored to be measured later after all men had been treated. Three men developed high serum AST elevations from 15× to 30× ULN, associated with serum total bilirubin levels (TBLs) above 3 mg/dL, during treatment (Fig. 1). In each case, serum abnormalities returned to baseline levels despite continuing therapy, suggesting adaptive tolerance to the drug. The authors concluded, “This raises the possibility that patients who progress to clinical hepatitis have livers that fail to adapt or to develop adequate repair mechanisms to cope with the insult.” Potential mechanisms of adaptation include changes in expression of hepatocyte enzymes and transporters,13, 14 and possibly immune mechanisms.15

Figure 1.

Graphic display of the time course of serum aspartate aminotransferase (red) and total bilirubin (green) values (measured after treatment was over) in three subjects treated with isoniazid (Data plotted from Table 4 in report from Mitchell et al.12). All three fulfill present definition by Temple as “Hy's Cases.” Despite this, all three adapted to the drug and became tolerant of it. Serum ALT was not measured in this study.

Patients treated with isoniazid or other drugs may be divided into three categories16 based on liver responses: “tolerators” who demonstrate no biochemical evidence of liver injury during treatment; “adaptors” who show biochemical evidence of liver injury that resolves despite continued drug treatment; and “susceptibles” who develop an initial injury that progresses to clinically significant liver injury if treatment is not stopped. We now believe that these three categories of patients probably exist for most and perhaps all drugs that are capable of causing hepatocellular injury progressing to liver failure. For example, the vast majority of patients in troglitazone clinical trials17 did not show ALT elevations of any magnitude and were therefore “tolerators.” There were also “adaptors” who experienced serum ALT elevations exceeding 10× ULN that resolved with continued drug treatment. Yet the experience unequivocally demonstrated that there were also “susceptibles” who developed significant liver injury in clinical trials (elevated ALT and TBL) and liver failure during marketed use. Tacrine and other drugs such as heparin would appear to be examples of drugs where the patient population consists of significant numbers of tolerators and adaptors, but few susceptibles.

C. Laboratory Variations and “Normal” Ranges

Stopping criteria based on asymptomatic ALT elevations are controversial, and need to be re-evaluated carefully based on systematically collected and analyzed data, supplanted by additional measurements over time and by other tests and clinical information. Using multiples of ULN can be problematic because, as for other clinical laboratory tests, individual laboratories usually establish reference ranges18 based on locally derived samples of apparently “normal” individuals, which may include variable numbers of those with unrecognized fatty liver, mild chronic hepatitis C, or other disorders. Hence, an identical blood sample can produce values at one laboratory that are within its normal range but above the ULN at the other. Prati et al.19 has noted that ULN values for ALT are probably too high at most laboratories due to inclusion of patients with occult liver disease, most notably fatty liver and mild chronic hepatitis C, in reference groups. With increasing standardization of automated analyzers for multiple serum variables, variations in control “normal” population samples have become more important than test method variations.18

Some discussants20 argue for using multiples of each person's baseline values for assessing elevations in test results. Establishing a reliable “baseline” value requires more than a single pretreatment measurement, which is not usually done in medical practice. Such a method might increase the sensitivity to detect early injury or changes, but might not increase the specificity of identifying serious DILI.

D. Severity of Liver Injury

A critical issue is determination of the severity of the liver injury, which cannot be done by measuring serum enzyme activities alone. As discussed above, even substantial increases in ALT to 10× to 30× ULN may not indicate serious liver damage, with some patients able to adapt and recover completely, even when exposure to the causative drug is continued. However, if there is sufficient liver injury to cause reduction in the overall capacity of the liver to carry out its functions, such as clearance of bilirubin from the plasma, more severe injury is present. It is useful21 to define five levels of liver injury (Fig. 2) ranging from mild and usually reversible elevations of serum enzyme activities only (Level 1); to more extensive injury with early impairment of liver function shown by increases in TBL or prothrombin time (Level 2); to serious clinical illness with obvious jaundice and disabling symptoms (Level 3); to acute liver failure (Level 4) with secondary loss of other organ functions such as brain (encephalopathy) or kidney (hepatorenal syndrome); to death of patient or liver requiring transplant (Level 5). Lower levels of DILI severity are observed more frequently and are therefore more likely to be detected in clinical trials. The relative distribution of the severity levels of DILI may vary by drug and the population sample exposed to it.

Figure 2.

We propose to define five levels of severity of DILI. The two lower levels are defined by serum tests of enzyme activities and bilirubin; the upper three levels by clinical assessments of patient status. Incidences of lower level cases tend to be higher, but the “pyramid” shape may vary for different drugs and populations samples exposed to them. Modified from slide #12 shown March 26, 2008 at:

It was recognized by Zimmerman22 in the late 1960s that a patient who presents with jaundice as a result of a drug-induced hepatocellular injury had at least a 10% chance of dying from liver failure (before liver transplantation was being done). The rise in TBL reflects a substantial risk because, with hepatocellular injury, this rise results from a major loss of functioning hepatocytes once other causes for increased bilirubin are excluded (hemolysis, decreased activity of bilirubin-conjugating enzymes [Gilbert syndrome] or transporters, or a significant element of cholestasis). Zimmerman's observation that hepatocellular DILI with jaundice indicates a serious reaction has been referred to by Temple1 as “Hy's Law.” Hy Zimmerman never defined “levels” of severity or quantitative elevations of serum TBL or aminotransferase activities that would denote jaundice or hepatocellular injury, nor how he determined the injury was drug-induced. But he continued to believe in the validity of his observation through two editions of his classic text,23, 24 although he was reluctant to have his name used eponymically.25, 26 The original Zimmerman observation included a mixed group of jaundiced patients he had seen in consultation, many of whom had more clinically serious DILI that we might now call Level 3. They roughly correspond to the levels of severity in patients reported to have an approximate 10% mortality or transplant rate in recent reports from Sweden27 and Spain28 that confirm his observation. It was notable in the Mitchell study12 (Fig. 1) that none of the three individuals treated with isoniazid who developed hepatocellular injury with elevations in serum TBL showed progressive liver injury, and each appeared to adapt despite continued treatment with isoniazid. This suggests that adaptation can occur even after hepatocellular injury has progressed to jaundice in some individuals, and will not necessarily be progressive in that person.

E. Hy's Law Case Definition: a Severity Issue

The Temple definition of “Hy's Law cases” as those patients showing ALT >3× ULN and TBL >2× ULN would include reactions with only Level 2 severity. The current draft guidance2 for clinical trials describes three criteria for a “Hy's Law case”: (1) the liver injury should be hepatocellular in nature and there should not be a prominent cholestatic component; (2) there should be no more likely alternative cause than DILI, such as acute viral hepatitis A or B, or other acute liver disease;29 and (3) there should be evidence that the drug causes more frequent but less severe hepatocellular injury (Level 1), as shown by more frequent ALT elevations greater than 3× ULN in the treated group relative to the group on control treatment. This last criterion in part reflects the Temple observation that Hy's Law cases have arisen generally from a larger subset of patients showing milder injury but who presumably are unable to adapt (termed by Senior16 as “Temple's Corollary” to Hy's Law). This criterion may not be fulfilled for a drug later proven to be hepatotoxic, as in clinical programs containing protocols for very short-term treatment of study subjects (e.g., many antibiotic protocols, or short-term bromfenac administration). Another is when there has been an absence of systematic serum ALT testing of clinical trial subjects or too-long intervals between tests, such that liver injury events are not reliably detected.

In an attempt to assess the sensitivity and specificity of the ALT and TBL combined test for the detection of diagnosable liver injury (such as acute viral hepatitis, common duct gallstones, or infiltrative lesions, among others), the Air Force Coronary Atherosclerosis Prevention Study trial done in the period 1991-1997 was reexamined.30 This was a Merck-sponsored randomized, double-blind, placebo-controlled, 5-year trial to investigate whether chronic lipid lowering with lovastatin would decrease the incidence of first major coronary events. The trial enrolled more than 6600 patients, of whom 85% were male, without clinical evidence of atherosclerotic cardiovascular disease, of whom 3248 received placebo for 5 years and had serum chemistries checked on 20 occasions (three tests occurring before randomization [baseline], eight in the first year at about 6-week intervals, and then nine more in years 2 through 5). Subjects were excluded at enrollment if they had any liver chemistry value exceeding 20% above the ULN. During the study, six of the 3248 placebo-treated subjects showed elevations of ALT >3× ULN and TBL >2× ULN. Clinical evaluation revealed a diagnosis of serious liver disease in each case: two had acute viral hepatitis, two had common duct gallstones, one had fatal amyloid infiltration of the liver, and one had fatal metastatic colon carcinoma. It was noted that when combined elevations in serum ALT and TBL occur in a clinical trial that are not related to the study medication (placebo), an alternate disease etiology generally is diagnosable. This supports the conclusion that in the treatment arm of a clinical trial, a Hy's Law case can be assumed to be drug-related if no other etiology can be proved.

When drug treatment is stopped at a relatively low level of liver injury, and the injury resolves without the patient qualifying as a Hy's Law case, it is not possible to conclude whether the drug has serious liver toxicity potential, or whether that patient may adapt to the drug. To determine this, it may be considered to continue treatment of asymptomatic patients experiencing ALT elevations as long as they are closely monitored and do not show symptoms of impaired liver function (elevated TBL or prothrombin time). The maximum level of ALT that should always prompt discontinuation of treatment is arguable; use of 10× ULN has been suggested,31 and >8× ULN was included in the draft guidance for further consideration.3 The risk of serious liver injury may be increased in a subject with even a low-level ALT elevation compared to others in the trial without such elevations. It is the duty of a clinical trial investigator to minimize the risks for enrolled subjects32 while ensuring that study design and data collection provide reliable results upon which public health decisions can be made.

There are few or no systematically collected data upon which to propose general stopping rules for asymptomatic elevations of serum enzymes only, but confirmed and rapidly rising ALT levels, especially if accompanied by symptoms or evidence of reduced liver function, are of concern.


AASLD, American Association for the Study of Liver Diseases; ALT, serum alanine aminotransferase activity; AST, serum aspartate aminotransferase activity; CDER, Center for Drug Evaluation and Research; DILI, drug-induced liver injury; eDISH, electronic tool for drug-induced serious hepatotoxicity; FDA, Food and Drug Administration; PhRMA, Pharmaceutical Research and Manufacturers of America; TBL, serum total bilirubin concentration; ×ULN, multiples of the upper limit of normal range.

2. Should Rechallenge Be Done? When? How?

Rechallenge is usually considered to be a deliberate readministration of a drug suspected to have caused an adverse effect, usually after interruption in treatment and recovery from the effect. Its purpose may be to prove more definitively that the suspected drug caused the observed reaction, or to ascertain if the person has adapted to and become tolerant so it could be restarted or continued if its benefits justify the possible risks. The obvious concern is that re-exposing the patient might lead to an accelerated and intensified reaction if immune mechanisms are involved.33 Rechallenges may also be inadvertent, and Seeff has proposed34 two other types: (1) continuing administration of the drug, or lowering the dose, despite evidence of an adverse effect or (2) when a patient resumes taking the drug after a prior adverse reaction to that drug. An observation of increased liver enzymes upon readministration of a drug (“positive rechallenge”), while posing clinical risks and ethical concerns, provides the strongest confirmation that the drug caused the injury. However, interpretation of “negative rechallenge” may be problematic because a rechallenge may be negative or equivocal even when the drug is the culprit. For example, it has been reported that hepatotoxicity due to tacrine8 and isoniazid35 does not always recur on rechallenge.

Recently, GlaxoSmithKline examined36 their database of clinical trials and postmarket adverse event reports for intentional or inadvertent rechallenge with medications suspected of causing hepatotoxicity. They noted that rechallenge can also represent a serious risk in patients with ALT elevations who did not qualify as Hy's Law cases at the initial reaction.

To administer a drug when it is clear that there is a high index of suspicion that the drug caused earlier hepatic injury should only be considered in the circumstance where there is a known important benefit to restarting therapy, there are no acceptable alternate therapies, and the patient gives informed consent.32 In the context of clinical trials, such circumstances are likely to be rare, because trials are conducted for the purpose of demonstrating efficacy, and it may not yet be known if the benefits of an experimental treatment might justify the risks. Several major pharmaceutical companies indicated that it is not their company policy to support rechallenge of study participants who have shown even Level 1 DILI.37 It is not yet known how common this policy is in companies conducting clinical trials internationally.38

Rechallenge should not generally be used to determine causality, especially for investigational agents whose benefits have not be established. It may be considered if there is known benefit, no acceptable alternatives, and the patient and institutional review board consent.

3. Should Patients with Liver Disease Be in Clinical Trials?

Competing considerations argue in favor of and against routine exclusion from clinical trials of subjects known to have underlying liver disease,39-41 but exclusion of patients with liver disease is perhaps the rule rather than the exception in most trial protocols. This approach reflects in part the central role of the liver in the metabolism and elimination of most drugs and the concern that inclusion of patients with liver disease will result in greater variation in systemic drug exposures that might affect efficacy, in addition to safety concerns. However, most approved drugs will be prescribed ultimately to patients with liver disease, in part because liver diseases (fatty liver, alcoholic liver disease, chronic hepatitis C or B) are prevalent and not always known by the patient or the prescribing physician. There is growing consensus that the patient populations participating in late-stage clinical trials should generally mirror the real-world target population that will actually be treated, which would include patients with the chronic liver diseases prevalent in our society. It is important to determine the cause and severity of any baseline liver disease in a patient before enrolling them into a controlled clinical trial.

There are, however, considerations that argue for excluding from clinical trials patients with liver disease.41 Chronic liver disease may create “noise” that may make it more difficult to identify true DILI risk. In chronic hepatitis C, the natural history includes waxing and waning ALT.42 Another consideration is the statistical power necessary to make conclusions concerning risk in patients with liver disease. Based on the prevalence estimates43-48 shown in Table 1, a study of 1000 patients in the United States would be unlikely to contain more than 20 with hepatitis C (unless intentionally enriched). Such a small sample might not detect a modest increase in incidence of DILI in patients with liver disease. Therefore, in order to determine a real increased propensity to DILI, consideration might be given to enriching trials with subjects with underlying disease, pooling data from subjects across trials, or carrying out specific trials in patients with underlying liver diseases.

Table 1. Estimated Prevalence of Liver Disease in the General U.S. Population
DiseasePrevalence Rate (Reference)
Nonalcoholic fatty liver disease13%-30%43, 44
Unexplained aminotransferase elevations5.4%45
Chronic hepatitis C1.8%46
Alcoholic liver disease0.7%47
Chronic hepatitis B0.5%48

It had been widely believed that acute DILI is not more common in patients with chronic liver disease23, 24 but this belief has been recently challenged by several studies. For example, antituberculous drugs show increasing DILI incidence in patients with hepatitis B,49 in combined infection with hepatitis C and human immunodeficiency virus,50 and highly active antiretroviral therapy has been associated with an increased DILI risk in patients with hepatitis C and human immunodeficiency virus coinfection.51, 52 There are also recognized examples where drugs can worsen certain underlying liver diseases, such as use of cytotoxic chemotherapy, immunosuppressants, or corticosteroids in patients with chronic hepatitis B,53 interferon in autoimmune hepatitis,54 and tamoxifen55 and methotrexate56 in those with fatty liver. Although these studies are relatively small and do not clearly distinguish DILI from recurrence of the underlying diseases, they do raise the possibility that in some circumstances there may be a modest increase in the incidence of DILI in patients with chronic liver disease, at least with some drugs.

A final argument for excluding patients is based on the assumption that patients with underlying chronic liver diseases will tend to have reduced hepatic reserve and therefore will tend to be at risk of a more severe manifestation of DILI than someone with a normal underlying liver, or be less able to recover. It could therefore be argued that it is unethical to include patients with chronic liver disease in clinical trials, especially if the study is not powered to determine whether patients with liver disease are at higher risk for developing liver injury.41 Benefits in understanding of the DILI potential of the drug and its utility in the general population must be balanced against what is known about the drug and how that impacts on the risk to the individual patient.

Debate continues between regulatory bodies and sponsors as to whether patients with underlying liver disease, even if mild and stable, should be included. Two major issues are that such people are likely to be treated if the drug were approved, but sponsors worry that including them might cause difficult interpretation of results. Even minor liver injury that is tolerated by people with normal, healthy livers may be less well tolerated by those with underlying liver disease.

4. Can Improved Biomarkers for DILI Potential Be Found?

The available means of detecting the potential of a new drug to cause significant DILI are currently suboptimal chiefly because the available biomarkers are less than ideal.57 Serum ALT appears to be sensitive but not entirely specific. The use of the combination of ALT and TBL elevations appears to be much more specific, but Hy's Law cases may be quite rare in clinical trials, especially if there are strict stopping rules based on ALT; stopping treatment early in the course of liver injury may prevent progression of the injury to a point sufficient to cause TBL elevation. The absence of Hy's Law cases, even in a very large prolonged clinical trial, may therefore not exclude a significant potential liver risk. Continuing to treat patients in the face of ALT elevations may “bring out” more Hy's Law cases, but involves increased risk to these patients.

The development of novel DILI biomarkers will address several needs (Table 2). There is need to identify and validate biomarkers to identify which drugs are capable of causing serious DILI. Ideally, the biomarkers would identify these drugs in relatively small clinical trials of short duration. Once drugs capable of causing serious DILI are identified, there is a need for reliable biomarkers that could be used to predict which individuals are especially susceptible to develop DILI from that drug either before drug exposure, or at a very early time after drug treatment has begun (e.g., assays of substances in blood or urine). Ideally, these biomarkers would discriminate between the “susceptibles,” “adaptors,” and “tolerators.” Upon their validation, such biomarkers could facilitate approval of drugs that would otherwise not be possible, by excluding identified susceptible people who should never be exposed to that particular drug.

Table 2. Potential New Uses of Biomarkers for DILI in Clinical Trials
  • To predict the likelihood that a particular drug will cause serious vs. mild vs. no hepatotoxicity in the target patient population.

  • To identify for a given drug the subsets of treated patients who are susceptible to progressive DILI vs those who are tolerators vs those who are adaptors to the potential DILI effect of the administered drug

  • To positively diagnose DILI in patients (rather than it being a diagnosis of exclusion), i.e., a pathognomonic test or procedure

Individual markers of DILI susceptibility may be specific only for subsets of hepatotoxic drugs because they are likely to incite hepatocellular injury through diverse pathways58, 59 Certain drugs may initiate liver injury by accumulation of reactive metabolites, reflecting the balance of specific toxification and detoxification enzyme activities, or drug transporters.60 Susceptibility to DILI in an individual may therefore reflect the aggregate effect of variation in the activities of multiple pathways. Activities of these pathways may also vary dramatically in different human populations. Failure to adapt to initial liver injury may reflect abnormalities in relatively few pathways possibly common to many drugs.

It is widely assumed that variation in susceptibility to DILI, at least in part, reflects genetic factors.61, 62 Genome-wide scanning or targeted gene analysis can be used to identify polymorphisms (e.g., single-nucleotide polymorphisms) associated with DILI susceptibility. Each approach has inherent advantages and disadvantages. Genome-wide screening does not depend on a preformed hypothesis,63–65 but requires large sample sizes. In contrast, targeted gene analysis screens for single-nucleotide polymorphisms (or other genetic variations) in a relatively small subset of genes that have been previously implicated in mechanisms relevant to DILI. Targeted gene analyses using case-control methods have been performed on genomic material available from patients who developed DILI and controls, and in some cases, statistically significant associations have been reported and replicated.61 To perform effectively as tools for risk management, biomarkers must have sufficient predictive power to enhance clinical treatment decisions. To date, the power of genetic tests to positively predict individual susceptibility to DILI has been low.

In addition to genetic studies, biomarkers, as tools in risk management, can also evolve from unbiased analysis of genomic or other “-omic” (transcriptome, proteome, or metabolome) data obtained in large clinical trials where individual patients have experienced various degrees of liver injury. Regardless of the methods used to identify potential biomarkers, it will be essential to validate markers in very large numbers of biospecimens obtained from well-phenotyped individuals including both healthy and diseased populations as well as in populations treated with many different drugs.66 The Drug Induced Liver Injury Network is building a well-annotated biobank (serum, lymphocytes, genomic DNA, and liver biopsies where available) obtained from patients who have experienced DILI,67 and the Severe Adverse Event Consortium has already begun genetic analysis on DNA banks obtained from patients who have experienced DILI.62

To be maximally useful for biomarker development, biospecimens would need to be obtained before treatment, as well as at intervals on treatment and before the onset of DILI. The clinical trial is the best available setting for such studies.68 To date, this has not been possible for cases that occur after a drug has been on the market.62, 67 The most cost-effective path would be to institute these protocols once a potential liver signal, such as ALT elevation, has been observed. One analytic method69 of managing a liver safety data base to effect this early identification is eDish (Fig. 3). This or a similar format could be directly linked to the biospecimens bank, and would allow immediate identification of individual subjects of interest and immediate access to all pertinent clinical and laboratory data for those patients for detailed evaluation.

Figure 3.

Graphic display of x-y plot of peak alanine aminotransferase (ALT) and total bilirubin (TBL) for 3922 patients in a single trial, half on experimental drug “X” (red triangles) and half on control drug “C” (green circles). The right lower quadrant (Temple's Corollary Range) shows the incidences of ALT elevations in patients on each drug. Each symbol represents one patient observed over a period from 1 to 3 years, and full information on any patient over his/her whole time of treatment can be obtained by pointing-clicking on the symbol, using the eDISH program.70

Another approach toward biomarker identification would be to conduct a clinical trial with a currently widely used drug known to be both clinically valuable and capable of causing severe DILI in some people. Isoniazid for the prophylaxis of latent tuberculosis is one such potential candidate.21 Enough people are treated with isoniazid yearly to make it possible to design a clinical trial that would enroll these patients as they begin isoniazid treatment. This trial would involve collecting biological specimens to identify tolerators, adaptors, and susceptibles, using the armamentarium of “-omics” technologies to identify potential biomarkers. Standardization of approaches to learning more about DILI from clinical trials fits well with the planned conference on reaching consensus on DILI terminology scheduled for December 2008.70

It is likely that improved biomarkers of DILI and susceptibility to DILI may be found and validated, but this will require prospective collection of data and biospecimens from a large number of clinical trials.

5. Conclusions

Despite prolonged discussion and debate, it has become evident that past clinical trials have not provided data upon which to base firm answers to the questions posed, resulting in default to expert opinions.. There is a real opportunity to enhance the understanding of DILI by making better use of controlled clinical trials. Kamath has suggested longer observations and using more important endpoints.71 We here propose standardized collection of biospecimens and data in clinical trials once a potential liver safety signal is detected. Careful analysis of these resources should lead to improved protocol guidelines to assess the liver safety of new medicines, and could uncover new biomarkers that will improve the safety of drugs in clinical trials and in clinical practice.