Epidemiology and Genetic Risk Factors of Drug Hepatotoxicity

Abstract

Idiosyncratic DILI from prescription medications and herbal and dietary supplements has an annual incidence rate of approximately 20 cases per 100,000 per year. However, the risk of DILI varies greatly according to the drug. In the US and Europe antimicrobials are the commonest implicated agents, with amoxicillin/clavulanate the most common, while in Asian countries herbal and dietary supplements predominate. Several demographic factors are associated with the risk of DILI as well as the severity and outcome. Genetic analysis of DILI is currently limited but multiple polymorphisms of HLA genes and genes involved in drug metabolism and transport have been identified as risk factors for DILI.

Summary:

The epidemiology of idiosyncratic DILI suggests that overall it is an uncommon event. In the US and Europe, prescription antimicrobials are the most common cause but HDS are increasingly associated with DILI, particularly in Korea and Japan. The pattern of injury differs according to the implicated drug. Several differences are apparent in the risk of DILI with regards to demographic factors and some of these factors influence the severity of the injury and its outcome. Genetic analysis of DILI is limited although several large studies are ongoing. Multiple polymorphisms of HLA genes and genes involved in drug metabolism and transport have been identified as risk factors for DILI.

Introduction:

Drug induced liver injury (DILI) is the leading reason why drugs are withdrawn from the marketplace in the United States (1). An accurate assessment of the frequency of DILI is difficult as it mainly relies on voluntary case reports to national registries without an appreciation of the population at risk leading to under-reporting. In addition, the vast majority of DILI cases are idiosyncratic and the presentation, pattern of injury, latency and severity differ widely between drugs and sometimes even with the same drug. Despite these limitations, the last two decades have seen several advances in our understanding of the epidemiology and genetic risk factors associated with DILI which will be discussed in this article.

Epidemiology

Incidence:

Several large studies have tried to determine the incidence of DILI but encounter several methodological issues leading to a wide range of figures. In addition, the data is limited to only a handful of countries (figure 1). Table 1 lists some of these studies based on their country or region.

Figure 1:

World map illustrating countries with DILI incidence data.

Most of the world lacks any reliable data on DILI incidence.

Table 1:

National Studies of DILI incidence.

Country	Iceland	France	Korea	UK	Spain	Sweden
Year of study	2010–11	1997–2000	2005–7	1994–9	2004–9	1995–2005
Study Type	Prospective	Prospective	Prospective	Retrospective	Retrospective	Retrospective
Number of DILI cases	96	34	371	128	57	77
Crude DILI Incidence Rate/100,000 per year	19.1	13.9	12	2.4	3.01	2.3

The incidence of DILI varies from 2–19 cases per 100,000 per year but appears to be an underestimate in retrospective studies.

Perhaps the most accurate study to date was performed in Iceland using outpatient and inpatient prescription databases allowing an estimate of the population at risk (2).

• The crude annual incidence rate of DILI was 19.1 cases per 100,000 patients per year but there was a very wide range depending on the implicated drug with:

  • 43 per 100,000 for amoxicillin-clavulanate
  • 11 per 100,000 for diclofenac
  • 752 per 100,000 for azathioprine
  • 675 per 100,000 for infliximab.

An earlier French population based study suggested a crude annual incidence rate of 13.9 per 100,000 per year and was some 16 fold higher than would be expected by spontaneous reporting to local authorities (3). A recent study from Korea found a crude rate of 12 per 100,000 per year but only looked at patients that were hospitalized meaning the true incidence of DILI should be higher (4).

The higher incidence rate of DILI when assessed prospectively is illustrated by examining retrospective studies. Using general practice databases in the United Kingdom and Spain, the crude incidence rate ranged from 1.35 to 3 cases per 100,000 per year (5, 6) which in part could be explained by only identifying DILI if the patient was hospitalized or referred to a specialist. Similar figures were noted in a hepatology outpatient setting in Sweden of 2.3 per 100,000 per year (7).

A crucial point to note in these studies is how DILI was diagnosed and which laboratory criteria were used to establish a diagnosis. Typically, an elevation in transaminase level (ALT or AST) or alkaline phosphatase or bilirubin is required with some multiple of the upper limit of normal for these tests. However, this is not uniform between studies. Attributing causality to the implicated drug is also heterogeneous, relying on the exclusion of other causes such as viral hepatitis and autoimmune disease, a temporal relationship to the drug, and occasionally using consensus criteria or validated scoring systems such as the Roussel Uclaf Causality Assessment Method (RUCAM).

DILI registries exist in several countries and have the advantage of collecting very detailed data on each DILI case with a formal causality adjudication process but they are limited by the lack of information on the number of patients exposed to each agent. Nevertheless, these registries provide information on the types of drugs that cause DILI, the pattern of injury and the risk of mortality and morbidity. Table 2 details some of these registries and gives some comparison of the types of drugs involved in each country.

Table 2:

DILI registries across the world.

Country	US	Spain	Korea
Year of study	2004–13	1994–2004	2005–7
Study Type	Prospective	Prospective	Prospective
Number of DILI cases	899	461	371
Antimicrobials (% of total)	45.3%	32%	-
Herbal and Dietary Supplements (% of total)	16.1%	-	73%

There are few large prospective DILI registries but there is a striking difference comparing the US and Spain with Korea in terms of the implicated drugs with almost all the cases in the latter related to herbal and dietary supplements while prescription antimicrobials are common in the former.

In the US the National Institutes of Health has funded the DILI Network (DILIN) since 2004 which is a consortium of several academic centers which collect data on patients with DILI in a prospective and a retrospective study. This group recently published their findings on almost 900 patients that were prospectively enrolled (8). Antimicrobials accounted for 45% of all the cases and a further 16% were due to herbal and dietary supplements. The nature of these registries with a bias towards hospitalized patients is reflected in that 10% of patients died or underwent liver transplant and 17% developed chronic DILI. Similar findings were noted in the Spanish DILI Registry with a death or transplant rate of 11.7% in jaundiced patients and 32% of all cases related to the use of anti-infectious drugs (9). The contrast with Asian studies is marked when considering the type of agents that cause DILI. In a prospective Korean cohort of 371 DILI patients, more than 70% were thought to be related to herbal and dietary supplements (HDS) (4). In Japan, an analysis of 1676 DILI cases demonstrated that 10% were related to dietary supplements and 7.1% were related to Chinese herbal drugs (10) and a smaller study in Singapore found more than half of DILI cases were associated with traditional Chinese medicines (11). New data should start emerging from Latin America with the formation of the Latin DILI Network (LATINDILIN) which encompasses most South American countries along with Mexico (12).

Pattern of Injury:

Based on the ratio of serum ALT to ALP (expressed as multiples of upper limit of normal), the R ratio is used to classify DILI into:

• Hepatocellular (R>5)

• Cholestatic (R<2)

• Mixed (R of 2–5).

Using the R ratio, it is apparent that the majority of DILI cases in the large national registries are hepatocellular (figure 2). The pattern of injury will depend on the type of drug as demonstrated when examining the breakdown of the same figures in Asia where DILI from HDS is the predominant injury seen. Over 75% of cases in the Korean series were hepatocellular with very few cholestatic cases (4).

Figure 2:

Pattern of liver injury based on national data

The higher percentage of hepatocellular injury in Korea reflects the implicated drugs as the vast majority of cases there are related to herbal and dietary supplements. (Data from references 2, 4, 8, 9)

The pattern of injury will depend on the type of drug but also influences the outcome as hepatocellular injury is more than twice as likely to lead to a worse outcome such as death of liver transplantation in both the US and Spanish Registries (8, 9).

Specific Drugs:

The risk of DILI with any drug is difficult to ascertain using registry data where there is no information on the population at risk but overall it is a rare event. However, the registry studies are similar to population based studies in identifying antimicrobials as the leading cause of DILI at least in the US and Europe.

In the US DILIN the largest category of drugs causing DILI was antibiotics and as shown in table 3, of the top 10 implicated drugs, 9 were antibiotics headed by amoxicillin-clavulanate (8) which was also the leading agent in the Spanish DILI Registry (9). In addition, the population based study using the UK General Practice Research Database identified amoxicillin-clavulanate as second only to chlorpromazine in terms of the odds ratio in developing acute DILI (5).

Table 3:

Top 10 individual agents causing DILI in the US DILI Network (total N=899).

Rank	Individual Drug	N=899
1	Amoxicillin-clavulanate	91
2	Isoniazid	48
3	Nitrofurantoin	42
4	Sulfamethoxazole/trimethoprim	31
5	Minocycline	28
6	Cefazolin	20
7	Azithromycin	18
8	Ciprofloxacin	16
9	Levofloxacin	13
10	Diclofenac	12

It is notable that of the 10 agents, 9 are antimicrobials.

Anti-tuberculosis medications are very commonly prescribed, particularly in the developing world. The risk of DILI is difficult to ascertain but 5.3% of all the cases in the US DILIN were due to isoniazid (second only to amoxicillin-clavulanate) and 7% of the cases in the Spanish DILI Registry were due to isoniazid alone or in combination with other drugs (8, 9).

Similarly, statins are one of the most commonly prescribed medications and liver enzyme monitoring is recommended. However, their actual risk of DILI is rare (13, 14). In the US DILIN there were 29 cases of statin related DILI out of a total of 899 cases, much less than seen with several antimicrobials (8).

Other than antimicrobials, non-steroidal anti-inflammatory agents are another group that is commonly associated with DILI, particularly ibuprofen and diclofenac, accounting for up to 6–7% of cases in Iceland and Spain (2, 9), and to a lesser extent in the US (8).

Herbal and Dietary Supplements:

The association between HDS use and DILI is being increasingly recognized but studying this area is complicated by several issues.

• Firstly, there is a complete lack of any population data.

• Secondly, assessing causality is problematic as patients often take a combination of HDS products which often consist of multiple ingredients and take them haphazardly.

• Thirdly, HDS are not subject to the same federal regulations as prescription medicines so that actual content of each product is not always certain.

Despite these limitations some interesting data has recently emerged. In the US, the DILIN found 130 of 839 (15.5%) DILI cases were caused by HDS and the incidence appeared to be increasing (15). The 130 cases were divided into 45 cases due to body-building supplements (predominantly anabolic steroids) and 85 non-body-building HDS. The striking differences were the prolonged jaundice in the body-building cases but they all resolved whereas the remaining non-body building cases were predominantly in middle aged women and were more likely to result in a severe outcome such as death or transplant, even compared to DILI related to medications. Spanish and Latin American investigators also noted the pronounced jaundice and favorable outcome in young men taking body-building anabolic steroids (16). The Korean experience with non-body building HDS demonstrated that 4 of 371 patients with DILI died or needed liver transplant and ranged in age from 40–48 years (4).

Demographic Factors:

Age:

The large national DILI registries indicate that the average age of patients presenting with DILI is between 49 and 53 years (8, 9). This is similar to the average age in Korea where HDS is much more prevalent (4). This is a reflection of increased prescription medication and HDS use in middle aged and older people. Whether older age is a risk factor for DILI has been examined in the Spanish Registry. There was no effect on age in the overall cohort of 603 DILI cases but when the cases were broken down into cholestatic and hepatocellular injury, older age was independently associated with the former and younger age with the latter (17), confirming an earlier finding (18).

Recent data from the World Health Organization Safety Report Database examining 236 drugs known to be associated with DILI using data-mining methods also suggested that elderly patients (65 years or older) were much more likely to develop cholestatic injury while acute liver injury was more common in children (19). The effect in children may have been related to mitochondrial dysfunction while the cholestatic DILI seen in the elderly may have reflected the higher lipophilicity and biliary excretion in the drugs they were more commonly taking.

In the US DILIN, children are included and there are some interesting differences with adults (20). Although antimicrobials were common causes, amoxicillin-clavulanate was not one of the agents. The majority of cases had hepatocellular injury, with some very severe or fatal, chronic DILI was less common (7%) but 64% of all the cases had autoantibodies with a much higher incidence of eosinophilia compared to adult counterparts.

Children also appear to be at increased risk of DILI from certain drugs, particularly valproate and other anti-epileptic medications (21).

Gender:

The majority of large registry studies suggest a female preponderance of idiosyncratic DILI (2, 3, 4, 8) with the exception of the Spain where there was a slight increase in male patients (9). Although female gender is not independently associated with DILI, it was associated with worse outcome such as acute liver failure or the need for liver transplantation (17) in the Spanish Registry but not seen in the US DILIN (22). In addition, females were more likely to develop chronic injury (defined as >3 months) in an older Spanish cohort (18).

Although gender does not appear to increase the risk for DILI overall, for individual drugs such as macrolide antibiotics and diclofenac, there may be an increased risk in females (5).

Race/Ethnicity:

The effect of race/ethnicity on DILI is unclear since the European Registries are almost universally comprised of Caucasians. There is limited information in the US DILIN with a more heterogeneous population. The natural history study from the US DILIN demonstrated that almost 10% of patients died within 6 months of DILI onset and 19% had evidence of chronic injury (persisting beyond 6 months) (22). Asian race was an independent predictor of reduced time to liver-related death or liver transplantation and African-American race was an independent risk factor for chronic DILI.

Daily dose and polypharmacy:

An interesting observation is that almost all drugs that have been withdrawn from the market or have black-box warnings due to hepatotoxicity were given in doses of greater than 50mg. This phenomenon was examined using pharmaceutical databases in the US and Sweden and it did appear that there was a statistically significant relationship between the daily dose of medication and the risk of a severe outcome due to hepatotoxicity such as liver failure, transplantation or death. In addition, 77% of DILI cases occurred in medications given at a dose of 50mg or more (23).

The US DILIN failed to show an association between daily dose and outcome in 383 DILI cases but did note that a daily dosage of ≥50mg was associated with a shorter latency and a different pattern of injury, with less cholestasis (24). If a drug has significant hepatic metabolism and is given in a daily dose >50mg the risk of DILI is also increased (25).

The UK General Practice Research Database found that when 2 or more hepatotoxic drugs were given concurrently the risk of DILI increased 6-fold (5).

Hospitalization:

The referral bias of the large DILI registries is reflected in the over-representation of severe cases meaning many are hospitalized, for instance more than half in the US DILIN (8). However, DILI defined as elevation of liver enzymes is typically seen in the outpatient setting but several studies indicate that the incidence of DILI in the inpatient setting is as high as 1.4% (26) and is even higher when examining patients with jaundice or non-acetaminophen related acute liver failure (27). Again the commonly associated agents are anti-microbials, anti-tuberculosis and some cancer drugs.

Underlying liver disease:

The suggestion that DILI superimposed on chronic underlying liver disease should be more severe or associated with a worse outcome is controversial. The long-term follow up from the US DILIN noted that 24% of patients who died or were transplanted had underlying liver disease compared to 11% of patients who survived (P<0.02) (22). However, underlying liver disease was self reported and the severity was unclear. Azithromycin DILI appears to be particularly more common in patients with underlying liver disease (8). Statins are frequently used in patients with underlying fatty liver disease without an increase in the risk of DILI (28, 29) and it is unclear whether chronic viral hepatitis increases the risk of DILI from anti-tuberculosis medications (30, 31) but may do so with anti-HIV medications (32–34).

Genetic Associations with DILI

Background:

Epidemiologic studies indicate that the prevalence of DILI varies geographically, which likely reflects differing environmental and genetic risk factors for DILI. It is currently unknown the extent to which this variation is environmental versus genetic. No studies of immigrants have been published to determine if their prevalence of DILI reflects that of their host or mother country. Genetic studies have clearly shown that a genetic component does exist at least for select drugs based on candidate gene studies and genome wide association studies. Certain HLA haplotypes have been proposed to increase the risk of DILI. Whether or not a genetic component exists for general susceptibility to DILI remains uncertain.

The liver plays a key role in metabolizing toxins and drugs via enzyme mediated oxidative metabolism and conjugation reactions followed by biliary transport and secretion (figure 3). The genes encoding the enzymes and transporters involved in this process are highly redundant and polymorphic (35), which may have protected early human populations from catastrophic exposure to environmental toxins as new areas were settled. Due to natural selection, the frequency of these polymorphisms varies widely amongst different populations. Unfortunately, the polymorphic nature of these genes may also lead to rare idiosyncratic reactions to toxin or drug exposure that actually cause liver injury or may in some individuals accentuate formation of dose dependent hepatotoxins, such as acetaldehyde from ethanol. Not every alcoholic develops cirrhosis. The immune environment of the liver is normally a highly tolerant one in part due to its constant exposure to foreign antigen. Some drugs, however, are well-known potentially to induce immune mediated liver damage infrequently (e.g. halothane) that may even mimic autoimmune hepatitis. Consequently, immunomodulatory gene polymorphisms and mutations may also contribute to induction of drug induced liver injury.

Figure 3:

Candidate Gene Approach to DILI Genetics Discovery.

The large number of hepatic proteins involved in drug metabolism can be placed in three broad categories: bioactivation and detoxification enzymes and transporters. Drug hepatotoxicity may be dose dependent and may be due to the parent drug or its metabolites. Therefore, the relative activity levels of proteins involved in drug metabolism are likely important determinants of susceptibility to DILI. The redox state of hepatocytes is also known to affect the activity of these proteins. The genes encoding hepatocyte bioactivation and detoxification enzymes as well as transporters involved in drug metabolism are highly polymorphic, which may lead to uncommon or rare idiosyncratic reactions that only cause liver injury in select individuals.

Not surprisingly, given these two potential mechanisms of drug induced liver injury, early candidate gene studies in DILI focused mainly on genes encoding hepatic enzymes and transporters involved in drug metabolism as well as immunomodulatory genes. More recently large collaborative efforts have spawned a few genome-wide association studies (GWAS) and whole-genome sequencing (WGS) studies to confirm prior observations and to identify genes that may unexpectedly be associated with DILI. Results are available from only a handful of GWAS currently. Over 90% of published DILI genetic studies are small case control candidate gene studies and only 1 whole genome sequencing study has been published to date (figure 4). More than half of the case control studies focus on anti-tuberculosis treatment (AT).

Figure 4:

DILI Genetic Study Alternatives.

Abundant candidate gene studies have been performed to uncover genes and polymorphisms responsible for DILI but the scope of these studies is necessarily limited by preconceptions as to the cause of DILI. Genome-wide association studies and whole genome sequencing represent unbiased approaches to improving our understanding of DILI. To date few of the latter approaches have been applied to those with DILI, but there have been some preliminary results that show promise.

Goals of Genetic Studies:

There are several potential benefits in pursuing genetic analysis of DILI including those listed below:

• Identify those most susceptible to DILI in order to make genetic pre-screening cost effective

• Determine genetic risk factors for specific drugs or populations

• Identify genetic factors that predict DILI susceptibility regardless of drug or ethnic group

• Use genetic information to prevent exposure to drugs to which one is susceptible (i.e. personalized medicine) or to indicate closer monitoring during treatment

• To increase our mechanistic understanding of DILI, which may suggest novel treatments for DILI

Candidate Gene Studies:

The drug association that has been most often studied is DILI during anti-tuberculosis treatment (AT) given the high prevalence of TB in certain regions and the high frequency of AT associated DILI, which lends itself to case-control studies. Isoniazid is believed to be responsible for most of the liver injury associated with AT. In some cases, contradictory results have been obtained within the same country such as India, which may reflect local ethnic variation or small case numbers (figure 5). The genetic polymorphisms most frequently associated with AT induced liver injury are found in N-acetyl transferase 2 (NAT2), glutathione S-transferase M1 (GSTM1), glutathione S-transferase T1 (GSTT1), Cytochrome p450 2E1 (Cyp2E1), and Human leukocyte antigen (HLA) genes (36–43). NAT2 polymorphisms causing slow acetylation are associated with AT-DILI and certain Cyp2E1 polymorphisms may increase the severity of AT-DILI (41). Null mutations in GSTM1 and GSSTT1 have often been associated with AT-DILI, but not always (38). Not all studies have shown an association between AT-DILI and NAT2 polymorphisms, either. However, a meta-analysis of results from AT-DILI studies confirmed this association across different ethnic groups (44).

Figure 5:

Geographic Variability in Candidate Gene Analysis Results for DILI.

The majority of studies focused on anti-tuberculosis treatment (AT) induced liver injury since it is very common. NAT2 polymorphisms, which lead to slow acetylation of AT drugs were the most commonly associated with AT. Most of the other polymorphisms or mutations associated with DILI were population specific. An x before a gene name indicates, no association was found. Some negative association results may have been due to the limited power of a smaller study to detect a weak or moderate association.

Some other drug metabolism related genes associated with AT-DILI by case control studies include SCLO1B1, NOS2A, BACH 1, MAFK, UGT1A1, and PXR (45–48). SCLO181 encodes the organic anion transporting polypeptide 1B1 (OATP1B1), which is responsible for the hepatic uptake of xenobiotics and conjugated bile acids (45). The inducible isoform of nitric oxide synthase (iNOS, coded by NOS2A) regulates the redox state of the cell, which may affect the activity of drug metabolizing enzymes. BTB and CNC homology 1 protein (Bach1; coded by BACH1) and Mafk (coded by MAFK) are transcription factors that regulates expression of anti-oxidant enzymes such as GST (46). UDP-glucuronosyl-transferase 1A1 (UGT1A1) is a detoxification enzyme responsible for glucuronidation (47). The pregnane X receptor (PXR) regulates expression of a number of detoxification enzymes (48). These associations AT-DILI were not very strong in most studies and have not led to pre-testing of individuals prior to AT to prevent DILI.

HLA associations have been identified in case control studies for DILI related to other drugs as well, including nevirapine (49–51), lapatinib (52–54), and ticlopidine (55). Meta-analysis has been useful in confirming the association of nevirapine with HLA haplotypes (56). Conversely, DILI due to two immunomodulatory drugs (methotrexate and 6-mercaptopurine) have been associated with genetic polymorphisms in multi-drug resistance gene (MDR-1), methylenetetrahydrofolate reductase (MTHFR), thiopurine methyltransferase (TMPT), and inosine triphosphate pyrophosphatase (ITPA) (57–61). Nonsteroidal anti-inflammatory agents (NSAIDS) also frequently induce hepatotoxicity, however checking candidate gene polymorphisms as risk factors for NSAID induced hepatotoxicity was unrevealing in a Spanish population (62). The former results suggested that immune responses play a general role in DILI regardless of the drug in question. Unfortunately, the associated HLA haplotypes or alleles were typically population specific and drug specific and not useful for general DILI prevention.

Two case control studies have identified associations which were not drug specific. In Spain, a mixed group of drug induced liver injury cases was associated with GSTM1/GSTT1 null mutations (63). A Chinese study found that UGT1A9 polymorphisms were associated with drug induced liver injury, regardless of etiology (64). These general associations were intriguing, but did not have a high positive predictive value and have not been replicated. In contrast, a Spanish DILI study of 141 individuals with DILI from varied etiologies suggested that ABCB1, ABCB4 and ABCC2 polymorphisms do not enhance the risk of drug-induced hepatotoxicity (65). Much larger populations likely need to be studied to identify genetic risk factors common to liver injury due to different drugs. To that end, a number of international collaborations have been formed to increase the power of DILI studies to identify genetic risk factors. A uniform definition of DILI across studies may also help diminish discrepancies between studies.

Genome-wide Association Studies and Whole Genome Sequencing Studies:

In order to identify genetic polymorphisms that would be clinically useful, assumptions about the mechanism of DILI may need to be disregarded. Hence, GWAS and WGS studies have been published by large consortiums and studies are ongoing. The quality and size of the GWAS have increased over time and have produced useful results, particularly for individual drugs with a high frequency of DILI (Table 4). The definition of DILI, the methodology employed, and statistical analysis in the different GWAS have been variable. Replication groups were not utilized in all studies to confirm results. A small GWAS study showed an association between rho GTPase, a signaling molecule, and methotrexate-induce liver injury (66). All of the highly significant genetic associations thus far are restricted to HLA alleles or haplotypes, emphasizing the importance perhaps of T cell responses in DILI in general (67–72). The HLA associations identified by GWAS do overlap between certain drugs, but none is universal for all the drugs. Some associations as with Lumiracoxib do cross ethnic groups (71). The HLA association with flucloxacillin induced DILI was surprisingly strong, but given the frequency of HLA-B*5701 only 0.1% of individuals with that genotype are predicted to develop liver injury if exposed to flucloxacillin (70). Only one WGS study has been reported and the results confirmed a prior GWAS result (68).

Table 4:

Genome-wide Association Studies for DILI.

Drug Name	Class	Gene Association	OR	P value	Cohort	N	Reference
Methotrexate	Immuno-modulator	ARHGAP24 (rhoGTPase)	NA	9.0×10−3	Japanese	8	66
Flupirtine	Analgesic	HLA DRB1*16:01-DQB1*05:02	18.7	6.7×10−5	German	10	67
Lapatinib	Tyrosine kinase inhibitor	HLA-DRB1*07:01		2.0×10−18	International	34	68
Amoxicillin-clavulanate	Antimicrobial	HLA-A*0201	2.2	2.0×10−6	NW European & Spanish	177	69
		HLA-DQB1*0602	3.3	1.4×10−6	NW European & Spanish	177	70
Flucloxacillin	Antimicrobial	HLA-B*5701	80.6	9.0×10−19	United Kingdom	51	71
Lumiracoxib	Cyclooxygenase inhibitor	HLA-DRB1*1501-DQA1*0102	5.0	6.8×10−25	International	98	72
Ximelagatran	Thrombin inhibitor	HLA-DRB1*07:01	4.4	4×10−5	European	74	73
		HLA-DQA1*02	4.4	2×10−6	European	74	74

All of the highly significant genetic associations thus far are associated with HLA alleles or haplotypes. Some HLA alleles (e.g. HLA-DRB1*07:01) associate with multiple drugs. Certain HLA associations as with Lumiracoxib do cross ethnic groups. A small number of cases appears to be sufficient to identify strong associations as with Flucloxacillin and HLA-B*5701. Odds ratios (OR) and p values are given for secondary fine gene mapping where available rather than single nucleotide polymorphism (SNP) screening.

A GWAS of 783 individuals of European ancestry who experienced DILI due to more than 200 different implicated drugs was published by international collaborators in 2012 (73). No significant associations were identified after accounting for the known HLA associations with DILI due to flucloxacillin and amoxicillin/clavulanate. A trend towards an association for hepatocellular DILI was seen for STAT4, an immunomodulatory gene. No general HLA associations were confirmed. The lack of reproducible findings for mixed cases of DILI supports the idea that only weak to moderate determinants of DILI may be present in mixed cases and/or that strong associations reflect rare genetic variations. By increasing the number of cases in GWAS or WGS studies, we may be able to identify such risk factors for DILI.

Key Genetic Findings:

Genetic analysis of DILI is limited to date and several large studies are ongoing. The key findings thus far are listed below:

• Polymorphisms of genes involved in drug metabolism and transport are risk factors for DILI

• HLA polymorphisms are also associated with DILI

• Most genetic risk factors for DILI identified so far are drug and population specific

• Genetic pre-testing of individuals is not yet possible

Future Directions for Genetic Analysis of DILI:

Of course, there are many directions in which to follow up on the above key findings. The following are some suggested avenues:

• Expanded GWAS and WGS studies-both drug specific and non-specific analyses

• Microbiome analysis- Are we studying the wrong genome? The microbiome clearly affects drug metabolism and immune responses

• Environment influence and epigenetic factors- Gene expression is modified by other factors besides one’s gene sequence.

• Proteomics and metabolomics- Likewise protein activity is not simply a function of gene expression. Genetic analysis should be integrated with proteomics and metabolomics analyses.

• DILI treatment trials- Do we know enough yet? Genetic study results to date suggest modulation of drug metabolism, anti-oxidant levels, and immune responses may ameliorate DILI.

Key Points:

• Idiosyncratic DILI is an uncommon event.

• Antimicrobials are the commonest class of drugs associated with DILI.

• Herbal and dietary supplements are increasingly recognized as causes of DILI, particularly in Asian countries.

• Several epidemiological factors influence the risk, severity and outcome of DILI.

• Genetic analysis of DILI is currently limited but multiple polymorphisms of HLA genes and genes involved in drug metabolism and transport have been identified as risk factors for DILI.

Abbreviations used:

DILI

Drug-induced liver disease

DILIN

Drug-induced liver disease Network

ALT

alanine aminotransferase

AST

aspartate aminotransferase

ALP

alkaline phosphatase

HDS

Herbal and dietary supplements

NAT2

N-acetyl transferase 2

GST

glutathione S-transferase

Cyp2E1

Cytochrome p450 2E1

HLA

Human leukocyte antigen

OATP1B1

organic anion transporting polypeptide 1B1

iNOS

inducible isoform of nitric oxide synthase

Bach1

BTB and CNC homology 1

MAFK

Maf basic leucine zipper protein K

UGT1A1

UDP-glucuronosyl-transferase 1A1

PXR

pregnane X receptor

STAT4

signal transducer and activator of transcription 4

MDR-1

multi-drug resistance gene 1

MTHFR

methylenetetrahydrofolate reductase

TMPT

thiopurine methyltransferase

ITPA

inosine triphosphate pyrophosphatase

AT

Anti-tuberculosis treatment

GWAS

Genome-wide association study

WGS

Whole-genome sequencing