Translating new knowledge on drug-induced liver injury into clinical practice

Abstract

Drug-induced liver injury (DILI) is one of the main reasons for drug withdrawal from the market, and a cause of worldwide morbidity. Although several issues on DILI are still unsolved, there have been significant advances in new definitions and diagnosis tools. DILI is the result of a complex interaction between genetic and environmental factors, and constitutes an expanding area of investigation. DILI can mimic virtually all known hepatopathies, including vascular disorders and liver tumours. As part of this broad spectrum of clinical presentations, DILI severity ranges from asymptomatic elevations of aminotransferases to acute liver failure. Although biomarkers are emerging as valuable diagnostic tools, they are not available in clinical practice. Accurate DILI diagnosis is a challenging issue, particularly the establishing of causal relationships with the culprit agent and the exclusion of competing causes of liver injury. Given that the understanding of the mechanisms inducing DILI is growing, and both DILI causality assessment scales and the performance of international DILI networks have been improved, hepatotoxicity may be recognised earlier in clinical practice. In this review, advances and results obtained by DILI registries around the world, case characterisations, particularly those relevant to newer definitions in DILI, and the behaviour of chronic liver disease induced by drugs will be updated. In addition, recently published data on herbal and dietary supplements and new predictive scores for acute liver failure assessment will also be discussed.

Significance of this study.

• Drug-induced liver injury (DILI) can mimic the majority of liver injury patterns, and as a consequence of the lack of specific biomarkers, it remains as an exclusion diagnosis.

• Clinicians often face a challenging decision: an underdiagnosis of DILI increases the risk of re-exposure to the drug, while a wrong diagnosis may contraindicate a vital safe treatment.

• Specific DILI registries are collaborative multicentre networks that have become an essential tool for the study of DILI epidemiology worldwide.

• Both the growing use of herbal and dietary supplements (HDS) and the increased rate of HDS hepatotoxicity in occidental countries are very concerning issues.

• There is no specific therapy available for idiosyncratic DILI except for stopping the suspected drug, and to strongly recommend avoiding future intake of the same agent.

Introduction

We have witnessed an increase in the knowledge on drug-induced liver injury (DILI) and herb-induced liver injury (HILI) over the last few decades. Although we still lack valid tools to be implemented in clinical practice in the diagnostic and therapeutic field, new definitions of DILI and an improvement in diagnostic scores have been achieved in recent years. In addition to these achievements, the creation of hepatotoxicity registries has also improved the knowledge on the epidemiological behaviour of drugs and herbs worldwide. Clinicians and hepatologists have now more awareness and tools to suspect DILI even in situations presenting as confusing in clinical practice. This review is aimed at describing the most important advances in DILI and HILI made in recent years.

The growing usefulness of DILI registries

The well-known Spanish DILI Registry,^{1 2} US Drug-Induced Liver Injury Network (DILIN)³ and the recently created Latin-American DILI Network (LATINDILIN)^{2 4} are collaborative multicentre networks, with large databases of prospectively recorded bona fide DILI cases. Additionally, there are three well-conducted, prospective DILI studies performed in Iceland,⁵ Korea⁶ and India.⁷ All of them represent an essential resource and provide information about the most frequent phenotypes and clinical scenarios (table 1).

Table 1.

Major ongoing DILI registries worldwide and published prospective DILI studies

	Time period	DILI cases (n)	Main provided information
DILIN,3 USA	2004–2013	899	Fatal outcome 10% and 17.5% had evidence of chronic DILI. HDS was represented in 16% of the cases. Amoxicillin-clavulanate ranked first.
Spanish Registry,1 2 Spain	1994–2018	878	Anti-infectives (38%), central nervous system drugs (13%), musculoskeletal agents (11%) and cardiovascular drugs (10%) were the most represented drug classes. Amoxicillin-clavulanate ranked first.
LATINDILIN,2 Latin-American countries*	2011–2018	311	Hepatocellular cases in 60%, fatal cases 4.9%. Anti-infectives (32%) followed by musculoskeletal agents (14%) were the two leading causes. Amoxicillin-clavulanate ranked first.
Population-based study in Iceland5	2010–2012	96	Antibiotics (37%) with amoxicillin-clavulanate as the first single agent (22%). Fatal cases 1%.
Prospective nationwide study from Korea6	2005–2007	371	Hepatocellular phenotype in 76.3% and the most common causative class HDS (70, 5%). Death or OLT: 1.8%.
Single-centre experience from India7	1997–2008	313	Leading cause was antituberculosis drugs. ALF in 19%, overall mortality 17.3%.

*Argentina, Uruguay, Brazil, Chile, Mexico, Peru, Paraguay, Venezuela, Ecuador and Dominican Republic.

ALF, acute liver failure; DILI, drug-induced liver injury; DILIN, Drug-Induced Liver Injury Network; HDS, herbal and dietary supplements; LATINDILIN, Latin-American DILI Network; OLT, Orthotopic Liver Transplantation.

Anti-infectives were the most commonly involved therapeutic groups, with amoxicillin-clavulanate at the top of the list in all three DILI registries, whereas antituberculosis drugs were the most prevalent culprit agents in the Indian study,⁷ and herbal and dietary supplements (HDS) in Korea.⁶ Some differences that could be explained, at least in part, by differences in patterns of drug prescriptions and local availability, have been noticed when comparing DILI frequency by other drugs. Thus, in the Spanish network, amoxicillin-clavulanate is followed by tuberculostatics, ibuprofen and atorvastatin, while in the LATINDILIN, amoxicillin-clavulanate is followed by diclofenac, nimesulide and nitrofurantoin. In the DILIN it is followed by isoniazid, nitrofurantoin and sulfamethoxazole/trimethoprim, with diclofenac ranked in the last place of the top 10.

Chalasani et al ³ analysed some interesting subgroups among 899 subjects who had developed a well-characterised DILI episode. They characterised acute DILI in patients with chronic liver disease, with unusual latency periods, and those who had persistently altered liver biochemistries after 6 months. Outcomes were more severe in 10% of patients who had pre-existing liver disease (mortality rates 16% vs 5.2%; p<0.001). Antimicrobial agents were the most frequent agent identified in atypical latency cases (≤7 or >365 days). Nitrofurantoin and minocycline were present in 25% and 17% of DILI cases with prolonged latency, respectively (n=60). Nearly 18% of patients with acute DILI maintained abnormal biochemistry values at 6 months, and three-quarters of them continued with altered liver laboratories for 1 year. Whether these patients will progress to more severe liver condition or they will experience slowly resolving liver injury is still unclear.

Current impact of HDS in DILI

There are some inconsistencies regarding the definition of HDS. Herbals or herbal-derived compounds are relatively easy to define, but the definition of dietary supplements in the context of DILI causative agents can be more challenging. Many compounds and non-standardised ingredients are involved, such as food supplements, vitamins, minerals and anabolic steroids, which do not actually belong to the ‘dietary supplement’ category. A stricter definition is needed to categorise this group. Different proposals have been provided recently to define anabolic steroids as ‘hormonal compounds’ or abuse agents.⁷

The increment in HDS sales and in publications reporting HDS hepatotoxicity has doubled from 2005 to 2015.⁸ The increase in HDS liver toxicity could be explained due to several reasons. One of the most important is that consumption of herbal products is promoted by popular magazines as healthy medication. Additional reasons include misleading historical safety, easy purchase over internet and switch to HDS medication after failure of medicine, as happens with obesity.⁹

The global prevalence of herbal toxicity varies according to different geographical areas. Perhaps, the best estimate for the incidence of HILI comes from a population-based survey carried out in Iceland by Bjornsson and coworkers.⁵ This study revealed that the overall incidence of DILI in 2011 and 2012 was estimated to be 19 cases per 100 000 individuals. In this study, 16% of the cases were attributed to HDS. The Spanish DILI Registry has also recently shown that from 856 DILI cases, 32 were due to HDS, and 20 to anabolic steroids. Camellia sinensis (traditionally known as ‘green tea’) was on the top of the list, followed by Herbalife as the two leading causes of herbal toxicity.¹⁰ Dietary supplements distributed by Herbalife, most commonly used for weight loss, were implied in 29 cases described from USA, Europe and South America. Hepatotoxicity, typically hepatocellular, occurred more commonly in women and had a variable latency.¹¹ Four cases (13.8%) had positive rechallenge. The remaining nine cases (31%) had insufficient data to determine scores. For the 16 cases determined to have at least possible causal association, the median latency from ingestion to injury was 117 days (range 12–729). Most (15, 94%) were symptomatic at presentation. Until this potential association between HDS or some of its ingredients is clarified awareness might be advised.

The Latin-American network started working in 2011, and they have been able to recruit and analyse 312 patients with DILI episodes, of which 22 were induced by HDS and 14 by anabolic steroids.²

In the prospective study recently published by Navarro et al,¹² from the DILIN network, HDS accounted for 16% of total cases of DILI, and the proportion from 7% in 2004 to 19% in 2012.

Prevalence of HILI varies among different countries, ranging from 8% to 72%.¹³ Undeclared ingredients and contaminants are usually the culprits of HILI, thus masking diagnosis and disturbing the investigation. Due to this variety, it is vital that physicians try to obtain the product package.¹⁴

Although HILI requires the same stepwise approach used as when conventional DILI is suspected, the lack of previous hepatotoxicity reports is a challenging issue when assessing HDS causality.

Case characterisation and new definitions in DILI

Idiosyncratic DILI, as a difference with intrinsic reaction (table 2), is unpredictable and exhibits a wide range of clinical patterns, including acute or chronic cholestasis, hepatitis or even liver failure. In the clinical setting, liver biochemistry is used to define liver injury, and DILI is not an exception. However, as discrete elevations of aminotransferases are seen in prevalent diseases like non-alcoholic fatty liver, and considering that some drugs induce transient and minor alterations in liver tests, a consensus group proposed the following criteria as a requisite to consider a case as DILI¹⁵: (A) alanine aminotransferase (ALT) value ≥5 times of upper limit of normal (ULN); (B) alkaline phosphatase (ALP) value ≥2 times ULN, or (C) ALT value ≥3 times ULN and total bilirubin (TB) >2 times ULN. Accordingly, the phenotypes of liver injury are also defined by biochemical criteria. The pattern of damage is established using the R value, where: R=(ALT patient/ULN)/(ALP patient/ULN).

Table 2.

Major differences between intrinsic and idiosyncratic liver toxicity

	Intrinsic	Idiosyncratic
Incidence	High	Low
Predictability	Yes	No
Dose relation	Yes	Mostly not
Latency	Hours-days	Days-weeks-months
Pathogeneses	Interaction with liver enzymes Increased production of reactive oxygen species Oncolytic necrosis	Adaptive immune response
Examples of drugs involved	Acetaminophen Amiodarone Anabolic steroids Antimetabolites Cyclosporine Cholestyramine Valproic acid	Allopurinol Amiodarone Amoxicillin-clavulanate Diclofenac Isoniazid Ketoconazole Methyldopa

The resultant pattern classifies cases as hepatocellular (R≥5), cholestatic (R≤2) and mixed (R>2 and <5).⁴ As the Spanish DILI Registry pointed out, the pattern of liver injury may change over time, and the R value tends to decrease during disease progression.¹⁶ Hence, to avoid misclassification, it is recommended that the first available test after DILI detection is the one to determine the pattern of injury.

Diagnostic scales are essential tools in clinical assessment and the use of causality assessment scales can assist the diagnostic process when DILI is suspected. The recommended and most used by formal registries is the Roussel Uclaf Causality Assessment Method (RUCAM). The RUCAM scale (table 3) provides points based on timing, dechallenge, DILI risk factors, competing medications, competing diagnoses and rechallenge information. Results can range from 1 to 14, allowing the probability of occurrence of hepatotoxicity to be classified into ‘excluded’ (score <1), ‘unlikely’ (1–2), ‘possible’ (3–5), ‘probable’ (6–8) and ‘highly probable’ (>8).¹⁷

Table 3.

RUCAM causality assessment¹⁶

	Hepatocellular type	Cholestatic or mixed type
Time to onset
From the beginning of the drug
Suggestive Compatible	5–90 days (+2) <5 to >90 days (+1)	5–90 days (+2) <5 or >90 days (+1)
From cessation of the drug
Suggestive	≤15 days (+1)	≥30 days (+1)
Course after stopping the drug
Highly suggestive Suggestive Compatible Inconclusive Against	Decrease ≥50% in 8 days (+3) Decrease ≥50% in 30 days (+2) Not applicable No information or decrease ≥50% after 30 days (0) Decrease ≤50% after 30 days or recurrent increase (−2)	Not applicable Decrease ≥50% in 180 days (+2) Decrease ≤50% in 180 days (+1) Persistence/increase/nor information (0) Not applicable
Risk factors
Alcohol or pregnancy    Age	Alcohol presence (+1) Alcohol absence (0) >55 years (+1) <55 years (0)	Either presence (+1) Either absence (0) >55 years (+1) <55 years (0)
Concomitant drugs
None, no information or incompatible (0) Concomitant drug with compatible time to onset (−1) Concomitant drug known to be hepatotoxic (−2) Concomitant drug with clear evidence to be hepatotoxic (−3)
Exclusion of other causes of liver injury
Group 1
Viral acute hepatitis Biliay obstruction Alcoholism Recent history of ischaemia	All causes (both groups) ruled out (+2) Six causes of group 1 ruled out (+1) Four or five causes of group 1 ruled out (0) <4 causes ruled out of group 1 (−2)
Group 2
Complications of underline disease CMV, EBV or HSV
Previous information about liver toxicity of the drug
Reaction labelled in the product characteristics (+2) Reaction published but unlabelled (+1) Reaction unknown (0)
Response to readministration
Positive Compatible Negative Not done or uninterpretable	Doubling of ALT values with suspected drug alone (+3) Doubling of ALT with all the drugs (+1) Increase of ALT but less than ULN with drug alone (−2) Other situations (0)	Doubling of ALP (or TB) values with drug alone (+3) Doubling of ALP (or TB) with all the drugs (+1) Increase of ALP but less than ULN with drug alone (−2) Other situations (0)
Interpretation according to final score
<1 Excluded; 1–2 Unlikely
3–5 Possible; 6–8 Probable; >8 Highly probable

ALP, alkaline phosphatase; ALT, alanine aminotransferase; CMV, Cytomegalovirus; EBV, Epstein Barr virus; HSV, Herpes simplex virus; RUCAM, Roussel Uclaf Causality Assessment Method; TB, total bilirubin; ULN, upper limit of normal.

Once the damaging drug is withdrawn, most liver abnormalities in DILI episodes resolve before 6 months. However, 10%–20% of patients have persisting evidence of liver injury after this period time.^{18 19} The true incidence, risk factors and, most importantly, liver consequences of persistent alterations are not elucidated. Fontana et al ²⁰ described the outcomes of 113 patients with an altered liver test at 6 months of follow-up. Based on the liver enzyme values at 12 months from DILI episode, they categorised those patients as persisters (n=74) or resolvers (n=25). The proportion of cholestatic liver injury at presentation was higher in persisters (54.1% vs 20%, p<0.01).

Chronic outcome following acute idiosyncratic DILI was also recently analysed in a prospective, multicentre study carried out by Medina-Caliz et al,¹⁹ on behalf of the Spanish DILI Registry. The authors described the outcome of 25/298 patients (8%) who did not resolve liver biochemistry after 1 year of follow-up. The main drug classes involved in chronicity were statins (24%) and anti-infectives (24%), with the latter including 16% cases with amoxicillin-clavulanate and 8% cases with sulfamethoxazole and trimethoprim. Values of ALP and TB >1.1 times ULN and 2.8 times ULN in the second month from DILI onset, respectively, were found to predict chronic DILI (p<0.001). Histological examination was available for 16 patients showing two cases with ductal lesions, one with low-grade fibrosis and seven with cirrhosis. Drugs implied in this evolution were atorvastatin, bentazepam, ebrotidine, clopidogrel/atorvastatin, amoxicillin-clavulanate/ibuprofen and ranitidine. They proposed that 1 year might be the best cut-off point to define chronic DILI.

The challenging scenario of acute liver failure induced by DILI

Deterioration in liver function as a consequence of DILI has an estimated incidence rate of 1.91 events per 1 000 000 person-years (95% CI 1.31 to 2.70) and a mortality rate of 21%, according to a population representative cohort from USA.²¹ Reuben et al,²² on behalf of the Acute Liver Failure (ALF) Study Group, reported idiosyncratic DILI in 11% of ALF cases enrolled in 10 years. Among the 61 agents implicated, antimicrobials were the most frequently involved, particularly isoniazid, nitrofurantoin, trimethoprim-sulfamethoxazole and antifungal drugs. Patients with ALF due to idiosyncratic DILI had poor liver transplantation-free survival; spontaneously recovery occurred just in 27%, while 42% of the cohort underwent liver transplantation. The Spanish DILI Registry showed that antituberculosis drugs and flutamide topped the list of drugs associated with severe or fatal outcome, while cyproterone, flutamide and nimesulide were on the top in the LATINDILI Registry.²³

As pointed out by Hyman Zimmerman in the 60s, a patient who presents jaundice as a consequence of hepatocellular injury from DILI has a significant risk of dying (‘Hy’s Law’).

As a result of a prospective validation, from the Spanish DILI Registry, a new definition that predicted ALF development with higher specificity (63% vs 43%) was proposed. These new criteria to predict DILI-associated ALF include a total serum bilirubin >2 times ULN, and an nR of 5 or greater, taking that the value of nR is the higher of the following two alternatives²⁴:

nR=(ALT/ULN for ALT)/(ALP/ULN for ALP) or nR=(AST/ULN for AST)/(ALP/ULN for ALP)

Although this proposal arises as a helpful prognostic score (area under the receiver operating characteristic curve 0.77) still needs validation.

Valuable tools for current DILI diagnosis

DILI diagnosis continues to be a challenge that clinicians must face without confirmatory tests and lies on clinical suspicion and exclusion of alternative causes (figure 1). Liver biochemistries remain as the hallmark to detect the damage and classify it. However, clinicians must keep in mind that drugs can also induce the typical laboratory and even pathological features of classic autoimmune hepatitis (AIH). Neither special laboratory tests nor specific histological characteristics have been shown to be useful for distinguishing classic AIH from drug-induced AIH. However, the absence of biochemical relapse after corticosteroid withdrawal is still the most important tool in clinical practice to support the diagnosis of drug-induced AIH.

Figure 1.

Diagnostic process in DILI suspicion. ALT, alanine aminotransferase; CIOMS-RUCAM, Council for Internatiinal Organizations of Medical Sciences-Roussel Uclaf Causality Assessment Method; CMV, cytomegalovirus; DILI, drug-induced liver injury; EBV, Epstein Barr virus; HAV, hepatitis A virus; HBV, hepatitis B virus; HCV, hepatitis C virus; HEV, hepatitis E virus; HDS, herbal and dietary supplement; OTC, over-the-counter; TB, total bilirubin; ULN, upper limit of normal.

Drugs recognisably associated with autoimmune features include diclofenac, α-methyldopa, hydralazine, nitrofurantoin, minocycline, statins and anti-tumour necrosis factor-alpha agents.²⁵

Even if several genome-wide association studies have revealed a number-specific alleles in the human leucocyte antigen (HLA) region associated with idiosyncratic hepatotoxicity attributed to specific drugs (amoxicillin-clavulanate, flucloxacillin, lumiracoxib, lapatinib and terbinafine),^26–30 the clinical application of these pharmacogenomic findings as a prior prescription screening is however not useful. This is due to the fact that the identified HLA risk alleles to date have a low positive predictive value (PPV) for DILI. However, they have a high negative predictive value (NPV), which could support or refute the diagnosis. This could be especially important when two or more agents with a compatible temporal relationship are involved.

Liver biopsy is not required for DILI diagnosis. However, it may offer some relevant information in selected patients to support the diagnosis or an alternative cause.³¹ It may also have a role in the clinical scenario of bone marrow transplantation by establishing the diagnosis of veno-occlusive disease/sinusoidal obstruction syndrome.³² Liver biopsy becomes crucial to provide prognostic information when DILI fails to resolve several months after stopping the drug.^{33 34}

On the National Institutes of Health (NIH) web page (https://livertox.nih.gov/), clinicians can easily access for free to updated information about liver toxicity by drugs, herbals and dietary supplements. It has become an efficient tool providing a useful guidance on the diagnosis and management of patients with DILI.

Promising data on serum biomarkers

The lack of specific biomarkers results in an underdiagnosis and therefore a risk of re-exposure, or maintenance of the treatment that could lead to a more severe liver injury progression. Indeed, many cases of acute liver damage and even ALF remain undetermined concerning the aetiology (25%–40%).³⁵

To date, several DILI biomarker candidates have been evaluated in experimental mice and in idiosyncratic DILI, providing a physiopathological overview of the kinetics of these biomarkers. These potential biomarkers include serum soluble biomarkers, microRNAs, exosomes and metabolomic profiles.

Several studies have confirmed increased sensitivity and specificity of miRNA-122, specifically for DILI due to acetaminophen and amoxicillin-clavulanate.^{36 37}

Serum miRNA-122 combined with albumin serum level accurately identified subjects who died within 6 months of DILI occurrence (PPV 38%, NPV 100%). If confirmed prospectively, these two markers may be useful to identify patients at high risk of either mortality or liver transplantation.³⁸

According to a recent report, glutamate dehydrogenase appears to be more useful than miRNA-122 to identify patients with DILI, and K18, osteopontin and macrophage colony-stimulating factor receptor are promising candidates for prediction of prognosis during acute DILI events.³⁹

Limited progress in treatment

There is no specific therapy available for idiosyncratic DILI. Key recommendations for treatment and outcome are to stop the administration of the suspected culprit drug immediately, and to avoid future intake. The immediate suspension of the involved drug is the most effective action. When no alternative drug is available, it is imperative to make a balance between the progression risk of the underlying disease after drug withdrawal and the aggravation risk of liver injury caused by its administration.⁴⁰

Appropriate anti-inflammatory and hepatoprotective agents have been tested without success.¹⁶ The antioxidant N-acetylcysteine can quench various free radical species, and the earlier it is used, the better. It is considered a true antidote for acetaminophen overdose, and it is recommended for other forms of DILI with ALF at early stages, where it can improve survival.⁴¹ This alternative should be implemented within 12 hours from drug intake, and the patients must be referred to a transplantation centre.

Since an immune mechanism plays an important role in the pathogenesis of many patients with ALF, glucocorticoids should be considered in those cases associated with marked signs of hypersensitivity or autoimmunity. The benefits are controversial, and there are no controlled trials either with pulse or with step-down therapy. This could be explained by the fact that it is used in very severe cases, where the beneficial response can be underestimated. On the other hand, some authors consider that early and short time of corticosteroid course could be beneficial for severe cases with hyperbilirubinaemia, since it is associated with a shorter recovery time and reduced mortality.⁴²

Although there are some reports of cholestatic cases successfully treated with ursodeoxycholic acid, lack of compelling evidence does not allow us to make a formal recommendation for this agent.^{43 44}

In many cases of ALF induced by DILI, liver transplantation should be considered. Timing for liver transplantation remains as a primary challenge and is based on early recognition of those cases with idiosyncratic drug-induced ALF, who will not survive with medical therapy. The wide variety of causes of ALF introduces enormous heterogeneity in the outcome.²¹

Conclusions

Considering that DILI can mimic the majority of patterns of liver injuries, and the lack of specific biomarkers, it remains as a diagnosis of exclusion. Available guidelines and recommendations on diagnostic cut-offs in liver enzyme levels, defined different clinical phenotypes, a newly proposed severity assessment, and a new definition of chronicity can assist the DILI diagnostic process.

DILI registries are long-standing collaborative multicentre networks that have become an essential tool; most of the new concepts and definitions have emerged as a result of them, promoting vital physician awareness and have a favourable impact on public health education.