Drug-Induced Liver Injury in GI Practice.

Although drug-induced liver injury (DILI) is a rare clinical event, it carries significant morbidity and mortality, leaving it as the leading cause of acute liver failure in the United States. It is one of the most challenging diagnoses encountered by gastroenterologists. The development of various drug injury networks has played a vital role in expanding our knowledge regarding drug-related and herbal and dietary supplement-related liver injury. In this review, we discuss what defines liver injury, epidemiology of DILI, its biochemical and pathologic patterns, and management.

autoimmune hepatitis

acute liver failure

alkaline phosphatase

alanine aminotransferase

acetaminophen

aspartate aminotransferase

drug‐induced liver injury

Drug‐Induced Liver Injury Network

U.S. Food and Drug Administration

green tea extract

genome‐wide association

herbal and dietary supplement

hepatitis B

hepatitis C

human leukocyte antigen

immune checkpoint inhibitor

idiosyncratic DILI

N‐acetylcysteine

nonalcoholic fatty liver disease

Roussel Uclaf Causality Assessment Method

upper limit of normal

vanishing bile duct syndrome

veno‐occlusive disease/sinusoidal obstruction syndrome

Although drug‐induced liver injury (DILI) is a rare clinical event, it carries significant morbidity and mortality, leaving it as the leading cause of acute liver failure (ALF) in the United States.( ) It is one of the most challenging diagnoses encountered by gastroenterologists. DILI is also the most common single adverse event that has led to withdrawal of drugs from the marketplace, drug attrition, and failure of implicated drugs to obtain U.S. Food and Drug Administration (FDA) approval.( )

Defining DILI

Liver injury is recognized by abnormal liver biochemistries with or without associated clinical symptoms. Using liver biochemistry criteria, which will increase the specificity of hepatotoxicity causality assessment and eliminate false positives, is key.( ) This aids in early detection, prediction, and risk stratification of suspected cases of DILI. The updated Roussel Uclaf Causality Assessment Method (RUCAM) uses an alanine aminotransferase (ALT) >5‐times the upper limit of normal (ULN) and/or alkaline phosphatase (ALP) >2‐times ULN to identify liver injury.( ) Conventionally, liver biochemistry elevations to this degree, lesser elevations that are sustained over time, rapidly rising tests, or any elevation combined with signs of liver dysfunction, such as increase in the international normalized ratio or encephalopathy, are clinically significant and worthy of investigation.

Burden of DILI in the United States and Abroad

In western countries, acetaminophen (APAP)‐related liver injury remains one of the leading causes of DILI.( ) Given the challenges in detection and reporting, the incidence of DILI is difficult to ascertain. Annual incidence of DILI ranges from 2.3‐13.9/100,000 inhabitants in population‐based studies from Europe.( , , ) The highest incidence of non‐APAP‐related DILI was reported at 19.1/100,000 inhabitants/year with a steady increase in age‐standardized incidence of DILI, in an Icelandic population‐based study.( ) Although most of the cases reported in the western countries are DILI secondary to prescription medications, traditional/complimentary and dietary supplements are the main causative agents of DILI in Asia.( , ) In the only U.S. population–based study, the yearly incidence of DILI was found to be approximately 3/100,000 residents.( )

U.S. DILI Network

The National Institute of Diabetes and Digestive and Kidney Diseases established the U.S. DILI Network (DILIN) in 2003 to identify, enroll, and characterize cases of non‐APAP DILI and herbal and dietary supplements (HDSs).( ) The U.S. DILIN has two registry studies at eight different academic centers across the United States. The network has expanded our understanding of DILI. Antimicrobials were noted to be the most common causative agents, accounting for 45% of cases in a 2004 study, followed by HDSs, cardiovascular drugs, central nervous system agents, antineoplastic agents, and analgesics.( ) Of the antimicrobials, amoxicillin‐clavulanate (22%), isoniazid (11.7%), and nitrofurantoin (10.2%) were the top three implicated agents.( ) The network has also noted an increasing proportion of HDS‐related liver injury, from 7% in 2004‐2005 to 20% in 2013‐2014.( , )

Patterns and Outcomes of DILI

Recognizing the pattern of liver injury at the initial presentation is vital. It provides a useful foundation to establish a differential diagnosis and guides the diagnostic evaluation accordingly. Assessing the pattern of liver injury as hepatocellular, cholestatic, or mixed is based on which liver enzyme elevation predominates. For example, hepatocellular injury suggests that an elevation of ALT and/or AST is more prominent than ALP. Conversely, cholestatic injury suggests a predominant elevation of ALP. The R‐ratio is a quantitative expression of the injury pattern; it is defined as the ratio of serum ALT to ALP values, both expressed as multiples of ULN, obtained at the onset of injury. An R‐ratio of >5 indicates hepatocellular injury, <2 indicates cholestatic injury, and 2‐5 indicates mixed injury.( , ) Table 1 lists the drugs and pattern of associated liver injury.

Table 1

Drugs and Pattern of Liver Injury

			Pattern of Liver Injury
			Hepatocellular		Cholestatic		Mixed
Drug		APAP Isoniazid Macrolides Minocycline Nitrofurantoin Inhaled anesthetics Phenytoin Carbamazepine Valproate Sulfonamides Amiodarone Allopurinol NSAIDs Fluroquinolones		Amoxicillin/clavulanate Trimethoprim/ sulfamethoxazole Anabolic/androgen containing steroids Chlorpromazine Azathioprine Phenytoin Fluroquinolones Carbamazepine Amiodarone Sulfasalazine		Azathioprine Flavacoxib Sulfasalazine Phenytoin Carbamazepine Allopurinol Amiodarone Fluoroquinolones

Abbreviation: NSAIDs, nonsteroidal anti‐inflammatory drugs.

Hepatocellular injury is the most common pattern of DILI across various networks, with 52%‐75% of cases reported from Spanish, Latin, and U.S. DILIN, and in the Swedish Adverse Drug Reaction Advisory Committee (SADRAC).( , , , ) Patients with hepatocellular injury tend to be younger, less likely to be clinically jaundiced, but have higher frequency of liver‐related deaths. Hepatocellular injury is 2‐3 times more likely to lead to liver transplantation. The clinical course tends to be more protracted in those with cholestatic injury.( )

Hepatocellular DILI with jaundice is an important pattern of biochemical injury to recognize, commonly referred to as “Hy’s Law.”( , ) It was first observed by Hyman Zimmerman in an analysis of 114 patients taking isoniazid, patients with an increased ALT >3‐times ULN, and total bilirubin >2‐times ULN without initial findings of cholestasis and after excluding other potential causes experienced a case fatality rate of 10%.( ) This observation has been used by the FDA to identify drugs with the potential of severe liver injury. Similar fatality rates from DILI with hepatocellular jaundice have been seen in the Spanish DILIN (11.7%) and SADRAC (12.7%).( , , )

Adaptation is a transient and modest rise of liver tests due to a drug that persists without progression, or regresses back to normal despite continued use of a drug thought to be the cause. The common examples of liver adaptation are statins and isoniazid, although both agents have also been implicated in cases of idiosyncratic DILI.( )

From a clinical standpoint, the development of low‐level liver enzyme elevation (ALT < 3‐times ULN or direct bilirubin <2‐times ULN) in an otherwise asymptomatic patient presents an important challenge. The clinician must also weigh the risks and benefits of a drug for the patient, as the benefit of a medication may outweigh the risk of hepatic injury. In this context, if low‐level enzymes are detected that, in the clinician’s assessment, are due to a drug, the patient has no liver‐related symptoms, and the drug is of evidence‐based clinical value to the patient, the enzymes should be followed monthly for at least 3 months to confirm their nonprogressive nature. A concurrent hepatic ultrasound to detect drug‐related morphological alterations such as steatosis is also wise. If the enzymes’ direct bilirubin rises, or new steatosis is detected, an alternative agent should be sought, even if within the same class of drugs. If no better alternative exists, such as may be the case with some chemotherapies, dose reduction or drug continuation with close enzyme and imaging follow‐up is recommended.

Most patients with DILI have both clinical and biochemical recovery. However, a small proportion of cases may develop chronic liver disease or chronic DILI, conventionally defined as persistence of liver‐enzyme elevations for more than 6 months after withdrawal of an offending drug. The incidence of chronic DILI varies from 5%‐20% in various DILIN registries and population‐based studies.( , , )

Types of DILI

Conventionally, DILI is classified as intrinsic and idiosyncratic. Intrinsic DILI is predictable, occurring in a dose‐dependent manner. The typical example of an intrinsic DILI is APAP‐related liver injury. The injury is usually within a short duration of time with a predictable latency period after drug exposure. APAP‐related liver injury alone is responsible for about 46% of the cases of ALF in the United States. Pharmacodynamic studies have demonstrated a dose‐response relationship between the extent of biochemical abnormalities and dose of APAP ingested. At higher doses of APAP, the conjugation pathways for its metabolization are saturated, thereby generating N‐acetyl‐p‐benzoquinone‐imine, a highly reactive toxic metabolite leading to hepatocellular necrosis.( )

Idiosyncratic DILI (IDILI) is unpredictable and, arguably, may be the most common type of drug‐associated hepatotoxicity. It has a variable latency period and is difficult to replicate in experimental models. Historically, IDILI has been considered to be dose‐independent. In the last decade, however, there have been studies showing an association of IDILI with drugs prescribed at a daily dose of more than 50 mg.( , ) Whether there is true association between dose and IDILI is not fully clear. The pathogenesis of idiosyncratic DILI can either be related to the pathway of metabolism of the drug and/or activation of the immune system, as will be discussed subsequently.

Mechanism of Action

Direct hepatotoxicity occurs in the setting of a known hepatotoxic agent, which leads to cell death through necrosis or apoptosis. However, the mechanism of IDILI involves a complicated process involving the drug, its metabolites, and the host immune system. Most drugs implicated in DILI are lipophilic and metabolized by the liver. They undergo phase 1 reaction, which is usually mediated by the hepatic cytochrome p450 system. The generation of intermediate bioactive or reactive metabolites is an important step, leading to both intrinsic and idiosyncratic DILI.( ) These toxic intermediate products are usually inactivated by phase 2 reactions, such as glutathione or sulfate conjugation. If conjugation pathways are overwhelmed by the excess production of the toxic reactive metabolites or caused by depletion of the conjugating factors, covalent binding of the reactive metabolites to mitochondrial proteins results. This leads to production of reactive oxygen species and ATP depletion, causing cellular organelle dysfunction from activation of stress kinase signaling pathways and disruption of membrane permeability pores. The result is hepatocyte dysfunction, necrosis, and/or cell death.( , )

Immune‐mediated injury is an important mechanism for IDILI. Both the innate and the adaptive immune system play a vital role. Over the years, there have been various hypotheses for the mechanism of immune‐mediated DILI. Under the hapten hypothesis, reactive metabolites irreversibly bind with cellular proteins to form neoantigens (also referred to as “haptens”), which then are presented to major histocompatibility complex molecules on antigen‐presenting cells, triggering an immune response against the hepatocyte by recruitment of cytotoxic T lymphocytes, natural killer cells, and B cells. In some instances, the haptens can induce development of autoantibodies against cytochrome p450 enzymes, leading to cellular injury and death, as in the case of halothane‐related liver injury.( , , ) The pharmacological interaction concept proposes that a drug or its metabolite can bind directly to the human leukocyte antigen (HLA) molecule to trigger a T cell–mediated injury, especially in genetically susceptible individuals.( )

The liver has a remarkable capacity for immune tolerance. This is a necessary adaptation that serves as a barrier to hepatocyte dysfunction and injury from an inflammatory state, which develops in response to the constant exposure to ingested antigens. This is managed by induction of peripheral immune tolerance to incoming antigens.( , ) The adaptive immune response can be up‐regulated in the setting of a sublethal stress from a hapten. HLA associations are indicative of the adaptive immune response in IDILI, and in genetically susceptible individuals, overt liver injury may occur as a result of “defective adaptation.”( ) Conceivably, this complex interplay between the immune and nonimmune pathways is responsible for the unpredictability of idiosyncratic DILI. Figure 1 details the mechanism of direct and immune‐mediated DILI.

Fig. 1

Mechanism of direct and immune mediated pathways of DILI. Abbreviations: RM, reactive metabolite; ROS, reactive oxygen species.

Risk Factors for DILI

Patient‐Related Factors

Susceptibility to DILI is influenced by modifiable and nonmodifiable host demographic, clinical, and environmental factors.

Age

Various large prospective DILI registries differ on advancing age as a risk factor for DILI. A population‐based Icelandic study showed increasing incidence of DILI with age: 5 times higher for patients over 70 years than those between 15 and 29 years. The former group was also noted to have a higher prescription rate of drugs,( ) which could be a potential reason for increased DILI in this group as opposed to advanced age itself. The U.S. and Spanish networks did not identify advancing age as a risk factor for all‐cause DILI.( , ) Increasing age does pose as a risk factor for liver injury from medications like isoniazid, amoxicillin/clavulanate, and nitrofurantoin.( ) Cholestatic DILI is more common in the elderly as compared with hepatocellular injury in younger individuals.( ) Thus, age may confer a susceptibility to DILI in a drug‐specific manner.

Gender

Although epidemiological data from Spain, the United States, and Iceland do not suggest a heightened risk of DILI in women,( , , ) others have found women to be more susceptible to DILI from some drugs. These agents include minocycline and nitrofurantoin, in which DILI is typically characterized by autoimmune features.( , ) Women with acute liver injury are more likely to progress to acute liver failure, as demonstrated in various registries.( , , )

Race

Race and ethnicity are potential factors influencing DILI frequency, liver injury patterns, and outcomes. In a U.S. DILIN cohort, African‐Americans with DILI tended to be younger with higher rates of chronic DILI. The most common agents in African‐Americans were trimethoprim/sulfamethoxazole, methyldopa and phenytoin, as compared with amoxicillin/clavulanate in Caucasians. The pattern of liver injury as well as the time of recovery of DILI does not appear to be have significant differences when comparing African‐Americans and Caucasians, the predominant pattern being that of a mixed injury. African‐Americans were twice as likely to develop severe liver injury, either leading to death or liver transplantation, despite absence of disparity in attaining health care.( ) The reasons for worse liver‐related outcomes and chronicity of DILI in African‐Americans in unclear. Agents associated with hypersensitivity reactions are known to be the common causes for DILI in African‐Americans, which can be as a result of genetic factors. African‐Americans have higher risk of severe cutaneous reactions to allopurinol due to higher frequency of HLA‐B*5801.( )

Pregnancy

Data on DILI during pregnancy is limited and limited to therapeutic agents used to treat gestational hypertension and hyperthyroidism. The most common drugs are methyldopa, hydralazine, propylthiouracil, and antimicrobial agents like tetracycline.( , ) It is imperative to differentiate DILI during pregnancy from other more common etiologies of abnormal liver tests including viral hepatitis, gallstone disease, or pregnancy‐related complications such as intrahepatic cholestasis of pregnancy.

Alcohol

The pathophysiology of APAP‐associated liver injury in the setting of alcohol use is complex and driven by the manner in which alcohol is consumed. Acute alcohol co‐ingestion with APAP may be protective, as they compete with each other for use of CYP2E1 substrate, in turn reducing the byproduct N‐acetyl‐p‐benzoquinone‐imine. In contrast, chronic alcohol use augments APAP hepatotoxicity by acting as a CYP2E1 inducer. Chronic malnutrition, as may occur in patients with alcoholism, may further augment risk for DILI due to glutathione depletion in the malnourished state. In IDILI, there are no data to suggest increased risk of all‐cause DILI in the setting of chronic alcohol use. Chronic alcohol consumption is a risk factor for DILI from isoniazid, methotrexate, and halothane.( )

Chronic Liver Disease and Comorbid Conditions

In a 2015 U.S. DILIN study, patients with pre‐existing liver disease were noted to have higher rates of severe DILI and 3‐times higher risk of mortality in comparison to those without prior liver disease.( ) Nonalcoholic fatty liver disease (NAFLD) is also a potential factor to increased risk of DILI( ); however, it should not preclude the use of statins in patients with NAFLD. In a 2010 post hoc analysis of the prospective Greek Atorvastain and Coronary Heart Disease Evaluation study, those receiving statin therapy had significantly lower rates of cardiovascular events and significant improvement in their liver chemistries.( ) In several prospective studies, similar results of improvement of liver enzymes in patients with NAFLD on statin therapy have been demonstrated.( , , ) The data from these studies provide evidence that statin use is safe for dyslipidemia with NAFLD.

Genetics

Single nucleotide polymorphisms in numerous genes and HLA regions have been shown to have an increased risk for DILI. A U.K. genome‐wide association (GWAS) identified HLA‐B*5701 genotype as a major determinant of flucloxacillin DILI.( ) Another European and U.S. GWAS showed evidence of HLA genotypes HLA DRB1*15:01 and DQB1*0602 playing a significant role in amoxicillin‐clavulanate DILI susceptibility.( ) HLA‐B*35:02 has been associated with increased risk of minocycline DILI, with a 16% carrier frequency in DILI cases in comparison to 0.6% in the population control in a GWAS study.( ) The rarity of occurrence of DILI in relation to a given drug confers a negative predictive value greater than 95% to these HLA alleles. HLA genotyping can potentially increase the accuracy and confidence in diagnosing DILI.( )

Drug‐Related Factors

Dose

There are data to suggest that drugs with daily oral dosing of ≥50 mg account for 70%‐80% of DILI cases in multiple DILI registries and population‐based studies.( , , )

Drug Metabolism

Ghabril in 2010 noted oral medications with more than 50% hepatic metabolism had significantly higher frequency of ALT >3‐times ULN, liver failure, and fatal DILI. In the same study, drugs with biliary excretion had significantly higher frequency of jaundice. Medications with more than 50% hepatic metabolism and dose more than 50 mg/day were noted to have an additive effect, leading to higher risk of hepatotoxicity.( )

Lipophilicity

Oral medication with high lipophilicity measured as LogP is known to influence drug pharmacokinetics and toxicity. Drugs with higher lipophilicity have an increased volume of distribution, higher chances of off‐target binding, and increased risk of toxicity with subsequent generation of increased reactive metabolites.( ) A drug’s hepatotoxic potential has been conceptualized as the “rule of two,” characterized by high lipophilicity (LogP > 3) and daily dosing (>100 mg). These parameters have been associated with an increased risk of DILI.( )

Making a Diagnosis

DILI is a challenging and complex diagnosis. Given the overlap between presentation of DILI with both acute and chronic liver diseases, it remains a diagnosis of exclusion. Obtaining a detailed history and recognition of the pattern of liver injury is key. Using the R‐ratio at presentation assists in a careful selection of diagnostic tests to evaluate for competing diagnoses: hepatic or systemic. In addition to a patient’s demographic variables, obtaining a full medication history is vital. Given the potential for prolonged latency periods, it is imperative to corroborate a patient’s medication list from their pharmacy and investigate over‐the‐counter medication/HDS use. This process can provide crucial information to establish a temporal relationship between drug exposure and development of signs/symptoms of liver disease. The spectrum of clinical presentation with DILI is broad, ranging from asymptomatic elevation in liver enzymes to nonspecific symptoms characterized by malaise, abdominal pain, and nausea to jaundice, pruritis, and encephalopathy. Establishing a timeline of onset of symptoms with respect to exposure of a culprit drug can be helpful, as the pattern of liver injury can change over the course of evolution of DILI. Figure 2 shows a stepwise approach to diagnosing DILI.

Fig. 2

Approach to diagnosis of DILI. Abbreviations: ANA, antinuclear antibody; ASMA, anti–smooth muscle actin; EtOH, ethanol; HBsAg, hepatitis B surface antigen; HCV, hepatitis C virus; HEV, hepatitis E virus; HSV, herpes simplex virus; IgG, immunoglobulin G; IgM, immunoglobulin M; LDH, lactate dehydrogenase.

Causality Assessment

A physician’s awareness of a drug’s hepatotoxic potential and associated phenotypic pattern is useful when considering a diagnosis of DIL. LiverTox (https://www.ncbi.nlm.nih.gov/books/NBK547852/) provides a comprehensive characterization of drugs and HDSs, describing their typical patterns and presentations.

A number of DILI‐specific causality assessment methods have been developed. These include general scales, algorithms, and expert opinions. The assessment provided by an algorithm or scale primarily depends on the weight given to each criterion. The validity of the method can vary as a result of differences in prioritizing the parameters. Table 2 details three liver‐specific causality assessment scales.( , , ) Causality scales are confounded by the absence of a true “gold standard” test for DILI. There is also varied interobserver reliability and reproducibility. Hence, these scales do not substitute a physician’s clinical judgement.

Table 2

Specific Causality Assessment Scales for DILI

Scale	Description	Comments
CIOMS‐RUCAM	‐ Initiated in 1989, RUCAM published in 1993 ‐ Numerical weight given to each key feature ‐ Features include chronology, risk factors, concomitant drug use, other etiologies, drug’s hepatotoxic potential, and response to rechallenge ‐ Overall score reflects causality probability ‐ Categorizes into five probability degrees: definite, probable, possible, unlikely, and excluded	‐ 86% sensitivity, 89% specificity ‐ Limited reliability; intrarater and interrater reliability of 0.54 and 0.45, respectively
Maria & Victorino	‐ Developed in 1997 ‐ Referred as CDS or M&V scale ‐ Validated using real and fictitious cases and compared with classification of three external experts ‐ Categorizes into five probability degrees: definite, probable, possible, unlikely, and excluded	‐ 84% agreement between the scale and expert opinions ‐ Definite score assigned only when “positive rechallenge” is present ‐ Drugs on the market for >5 years and without documented hepatotoxicity potential are given lower scores ‐ Poor performance with drug with long latency periods
DDW‐J	‐ Proposed in Japan ‐ Derived from CIOMS scale ‐ Modifications in chronologic criteria, concomitant drug use, and extrahepatic manifestations ‐ Uses an in vitro drug lymphocyte stimulation test ‐ Categorizes into three probability degrees: definite, probable, and unlikely	‐ Limited access and lack of standardization limit the use of this scale ‐ Shown to be superior to CIOMS and M&V scale

Abbreviations: CDS, Clinical Diagnostic Scale; CIOMS, Council for the International Organization of Medical Sciences; DDW‐J, Digestive Disease Week–Japan; M&V, Maria & Victorino.

The DILIN uses a structured expert opinion process for categorization of probability of DILI. They describe the likelihood of DILI based on percent probability of diagnosis of DILI: definite (>95%), highly likely (75%‐95%), probable (50%‐74%), possible (25%‐49%), or unlikely (<25%).( ) An assessment for the first 300 patients enrolled to U.S. DILIN in 2003 determined the RUCAM approach was more conservative in assigning a high level of causality than the DILIN strategy.( ) Although the two methods have been compared, the DILIN method has not been externally validated.

Liver Biopsy

Typically, a liver biopsy is not necessary to establish a diagnosis of DILI; however, it can prove useful in excluding other etiologies for liver injury and to assess the degree of inflammation and necrosis. The lack of resolution of liver injury serves as a strong rationale to obtain a liver biopsy.

Phenotypes

Clinicopathologic Phenotypes

Kleiner described 18 histopathologic patterns on liver biopsies of patients with suspected DILI. Most of the cases could be classified into one of five patterns: acute hepatitis, chronic hepatitis, acute cholestasis, chronic cholestasis, and cholestatic hepatitis.( )

Necroinflammatory

Inflammation, necrosis, and apoptosis are typical findings in hepatocellular DILI. The pattern of necrosis can be zonal versus nonzonal, and acute versus chronic hepatitis–like.( ) Acute hepatitis–like injury affects the hepatic parenchyma predominantly. Severe acute hepatitis is characterized by lobular disarray, from extensive sinusoidal architectural disruption, which correlates with a higher degree of hepatocyte apoptosis and confluent necrosis.( ) The main differential for acute hepatitis–like DILI includes acute viral hepatitis and fulminant presentation of autoimmune hepatitis (AIH). Confluent necrosis with acute hepatitis has been seen in some cases of diclofenac related DILI.( ) Chronic hepatitis–like DILI has a predominant portal infiltrate with interface hepatitis. Pure zonal necrosis, typically zone 3 necrosis, can mimic hypoxic‐ischemic liver injury, typical of APAP hepatotoxicity. This usually signifies cytotoxicity secondary to toxic drug metabolites, as opposed to necrosis with immune cell infiltration suggestive of involvement of the adaptive immune system.( )

Cholestatic

Acute cholestasis from DILI can be categorized as bland cholestasis or cholestatic hepatitis. Bland cholestasis is classically associated with injury caused by anabolic steroids and estrogens.( ) Both forms of acute cholestasis share common features of zone 3 hepatocellular and canalicular cholestasis. Concurrent granulomatous or eosinophilic inflammation in cholestatic hepatitis is suggestive of an immuno‐allergic or hypersensitivity‐type immune reaction.( ) Penicillins, especially with beta‐lactamase inhibitors (amoxicillin/clavulanate), sulfonylureas, methimazole and cephalosporins, have been implicated in causing cholestatic hepatitis.( , , , ) Azathioprine and mercaptopurine can cause liver injury, resembling both forms of acute cholestasis.( ) Vanishing bile duct syndrome (VBDS) is an ominous finding in cholestatic DILI, commonly identified by a paucity of bile ducts in over 50% of portal areas in a biopsy with at least 10 portal areas. A DILIN study confirmed the increased likelihood of development of chronic DILI in 94% of patients with VBDS in comparison to 47% in those without.( ) Amoxicillin/clavulanate, azithromycin, and fluroquinolones were major causes of VBDS in this study cohort.

Steatosis and Steatohepatitis

Historically, the risk of drug‐associated fatty liver disease related to tamoxifen or methotrexate use increased with the presence of obesity or diabetes.( ) Steatohepatitis has been associated with amiodarone, methotrexate, and tamoxifen. Microvesicular steatosis may indicate mitochondrial injury, which has a higher risk of severe liver injury, seen in DILI secondary to aspirin, valproate, amiodarone, and fialuridine.( , , , , , )

Vascular Injury

DILI from a vascular injury can present acutely with abdominal pain, hepatomegaly, abnormal liver enzymes, and acute onset portal hypertension. This usually occurs in the setting of Budd‐Chiari syndrome and veno‐occlusive disease/sinusoidal obstruction syndrome (VOD/SOS). VOD/SOS affects small veins and sinusoids, and typically presents as zone 3 necrosis. VOD/SOS is most commonly observed in the context of stem cell transplantation, exposure to toxins, pyrrolizidine alkaloids, onco‐therapeutic agents, and purine analogues. Thrombosis of the major hepatic veins in Budd‐Chiari can lead to massive congestion and ischemic necrosis. This is now an uncommon scenario, as it was associated with early formulations of contraceptive steroids.( , , , ) Nodular regenerative hyperplasia, typically identified with a reticulin stain, has been seen in cases of exposure to azathioprine (as a result of active metabolite 6‐thioguanine), oxaliplatin, and chlorambucil, in addition to various non–drug‐related disease states.( , , )

Immune Phenotypes

Drug‐Induced AIH

In multiple case cohorts, 2%‐9% of cases of AIH were considered to be induced by drugs, and drug‐induced AIH is responsible for 9% of all cases of DILI.( , , ) Differentiation between drug‐induced AIH and idiopathic AIH can be difficult for both the gastroenterologist and the pathologist, as patients often present with serologic and histologic markers of idiopathic AIH. As discussed previously, identification of DILI risk alleles can aide in the diagnosis of drug‐related AIH, such as HLA DRB1*15:01, which is positive in 57%‐67% of DILI cases from amoxicillin/clavulanate, in comparison with 15%‐20% cases in the general population. Other medications associated with drug‐induced AIH include minocycline, nitrofurantoin, hydralazine, infliximab, and methyldopa.( , )

Immunotherapy‐Related Liver Injury

Immune checkpoint inhibitors (ICIs) increase T‐cell responses and restore antitumor immune responses, which have been suppressed by cancer with the goal of inducing tumor rejection. The various targets for ICIs include cytotoxic T‐lymphocyte antigen 4 (CTLA‐4, the target for ipilimumab), programmed cell death 1 (PD‐1, the target for pembrolizumab and nivolumab), and programmed cell death ligand 1 (the target for avelumab, durvalumab, and atezoliumab).( ) However, the therapeutic reversal of immune tolerance following administration of these agents comes at the expense of immune‐related adverse events (irAEs), including hepatotoxicity.( ) A recent meta‐analysis showed a higher rate of all‐grade and high‐grade hepatotoxicity with CTLA‐4 inhibitors in comparison to PD‐1 inhibitors.( ) In other clinical trials with ipilimumab treatment, 11% of patients had early discontinuation of treatment due to hepatotoxicity, and this rose to 30% early discontinuation in combination therapy with ipilimumab and nivolumab.( , ) The mechanisms of the specific irAEs, which include rash, diarrhea, colitis, hepatitis and endocrinopathies, are not fully understood. It is postulated that hepatitis occurs from immune T‐cell activation, leading to secretion of CD4 T‐helper cell cytokines and cytolytic CD8 T‐cell tissue infiltration.

Hepatotoxicity from ICIs varies from an asymptomatic elevation of aminotransferases to acute hepatitis and fulminant liver failure. The pattern of liver injury can be nonspecific as well, ranging from hepatocellular to cholestatic. ICI‐related hepatitis is typically a seronegative hepatitis in comparison to cases of idiopathic AIH.( ) The 2018 American Society of Clinical Oncology Clinical Practice Guideline details the grading system for ICI‐related hepatotoxicity and its management.( ) Table 3 details the grades of toxicity from ICIs and proposed management.

Table 3

ICI Hepatotoxicity Grading and Management

Parameter	Grade 1	Grade 2	Grade 3	Grade 4
ALT, AST (×ULN)	1‐3	4‐5	6‐20	>20
Bilirubin (×ULN)	1‐1.5	1.6‐3	4‐10	>10
Liver disease/decompensation			X	X
ICI treatment	Continue	Hold	Stop	Stop
Lab monitoring	Twice weekly	Twice weekly	Daily	Daily
DILI treatment		Prednisone	IV Solumedrol	IV Solumedrol

Abbreviation: IV, intravenous.

Treatment

The most important step in management of suspected DILI is the discontinuation of the implicated agent and avoiding re‐exposure. In most cases (up to 90% or more), spontaneous recovery occurs as a result, without need for further treatment measures. Such improvement with discontinuation of a suspected offending agent is termed “dechallenge” and serves as good evidence of a causal association with injury.( , , )

Short‐term administration of a bile acid resin can be used in DILI by enhancing drug clearance by interrupting enterohepatic circulation of leflunomide and terbinafine.( , ) Carnitine is an antidote to valproate hepatotoxicity. It regulates mitochondrial acetyl‐CoA levels, leading to enhanced fatty acid uptake and beta oxidation in the mitochondria.( , ) The role of N‐acetylcysteine (NAC) in APAP‐related liver injury has been well established. NAC should be considered in patients with ALF from IDILI. The U.S. ALF study group randomized 45 patients with ALF from non‐APAP DILI to receive NAC versus placebo infusion. The NAC group had 2‐times higher transplant‐free survival rates.( ) The use of corticosteroids should be limited to DILI in the setting of drug‐induced AIH, immune therapy–related severe hepatitis, or in the presence of features of hypersensitivity. Corticosteroid treatment in one retrospective analysis was associated with lower survival in patients with more severe liver injury.( ) When corticosteroids are used, close monitoring is essential, as a failure of response may either suggest no benefit or an alternative diagnosis, and they should be stopped. With improvement of liver injury, withdrawal of steroids should be performed gradually with close follow‐up to detect resurgence of liver injury or uncover an AIH that warrants ongoing therapy.

Ursodiol can be used as pretreatment prophylaxis for patients undergoing myeloablative hematopoietic stem cell transplantation with good tolerability. Defibrotide is used as both a treatment for severe SOS and pretreatment prophylaxis in patients at high risk of developing SOS.( ) Liver transplantation remains the primary rescue treatment for ALF from DILI. Timely identification of ALF should prompt a referral to a liver transplant center. Table 4 details the targeted therapies for DILI.

Table 4

Specific Treatments for DILI

Agent	Treatment
APAP	NAC
Valproic acid	Carnitine
Leflunomide	Bile salt resin
Amanita (mushroom)	Silymarin
ICIs	Steroids
Sinusoidal obstruction syndrome	Ursodiol, defibrotide

HDS‐Induced Liver injury

More than half of U.S. households use some form of dietary supplement, most commonly multivitamins and minerals.( , , , ) The incidence of hepatotoxicity from HDSs is steadily increasing in the United States, up to 20% in 2013‐2014.( , ) The current regulatory framework established by the Dietary Supplement Health and Education Act of 1994 is suboptimal. The requirements for HDS products before being marketed are significantly less stringent than for pharmaceutical products; specifically, there is no requirement for assessments of product efficacy and safety. The FDA and other regulatory bodies take action only after a supplement has been shown to have potential toxicity.( )

Among HDSs, bodybuilding products are the most common cause of liver injury.( ) The typical clinical scenario is that of prolonged jaundice in young men, with nonfatal outcomes from use of anabolic androgenic steroids.( , ) Non‐bodybuilding HDS‐related liver injury is commonly seen in women with a hepatocellular pattern of injury. This group of patients with HDS‐related liver injury tend to have worse outcomes, with transplantation rates up to 11%.( ) Hillman found that in comparison to prescription medications, HDSs tend to have a higher transplantation rate and lower rate of ALF‐specific transplant‐free survival.( ) Non‐bodybuilding HDSs can be either single‐ingredient or multi‐ingredient products. One of the most common single‐ingredient HDSs is green tea extract (GTE). Results from a randomized Minnesota study showed that the health of postmenopausal women without chronic liver disease were 7 times more likely to have ALT elevations with GTE use. GTE dechallenge leads to a downtrend in ALT levels, and ALT elevation recurred following GTE rechallenge,( ) indicating a causal relationship between GTE and hepatotoxicity.

Multi‐ingredient products also pose a risk for HDS‐related liver injury.( , ) Given their unclear chemical descriptors, identification of the culprit agent is difficult. In addition, there can be contamination of HDSs with microbials, heavy metals, and mycotoxins. Large batch‐to‐batch variation in HDS product content is common.( , , , ) The U.S. DILIN using toxicologic analysis confirmed the suspicion of supplement mislabeling in multi‐ingredient products, with rates up to 50%‐80%.( ) The potential for more severe liver injury with these products remains largely unknown. However, the risk for toxicity may result from the complex interactions among individual ingredients, supplement overuse/misuse, or combination use with prescription medications.

Summary

DILI remains the leading cause for ALF in the U.S. adult population. It is a diagnosis of exclusion and requires a meticulous process of evaluation and a high index of clinical suspicion to attribute a potential agent as the cause of the liver injury. The accurate identification of clinical and biochemical patterns of liver injury at the onset enables prognostication for individual cases. HDSs are also a common cause for DILI, although the exact ingredient responsible for injury is difficult to confirm. The DILINs have played a vital role in expanding our understanding of liver injury from both drugs and HDSs. With continued advancements and expansion of drug injury registries, our hope is that the understanding of DILI epidemiology, mechanism of injury, and establishment of causality will continue to improve.