Potential conflict of interest: Nothing to report.
Identifying who is at risk of drug-induced liver injury: Is human leukocyte antigen specificity the key?†
Article first published online: 19 JAN 2011
Copyright © 2010 American Association for the Study of Liver Diseases
Volume 53, Issue 1, pages 358–362, January 2011
How to Cite
Chitturi, S. and Farrell, G. C. (2011), Identifying who is at risk of drug-induced liver injury: Is human leukocyte antigen specificity the key?. Hepatology, 53: 358–362. doi: 10.1002/hep.24094
- Issue published online: 12 JAN 2011
- Article first published online: 19 JAN 2011
Singer JB,LewitzkyS, Leroy E, Yang F, Zhao X,KlicksteinL, et al. A genome-wide study identifies HLA alleles associated with lumiracoxib-related liver injury. Nature 2010;42:711-714. Available at: www.nature.com (Reprinted with permission.)
Lumiracoxib is a selective cyclooxygenase-2 inhibitor developed for the symptomatic treatment of osteoarthritis and acute pain. Concerns over hepatotoxicity have contributed to the withdrawal or nonapproval of lumiracoxib in most major drug markets worldwide. We performed a case-control genome-wide association study on 41 lumiracoxib-treated patients with liver injury (cases) and 176 matched lumiracoxib-treated patients without liver injury (controls). Several SNPs from the MHC class II region showed strong evidence of association (the top SNP was rs9270986 with P = 2.8 × 10−10. These findings were replicated in an independent set of 98 lumiracoxib-treated cases and 405 matched lumiracoxib-treated controls (top SNP rs3129900, P = 4.4 × 10−12. Fine mapping identified a strong association to a common HLA haplotype (HLA-DRB1*1501-HLA-DQB1*0602-HLA-DRB5*0101-HLA-DQA1*0102, most significant allele P = 6.8 × 10−25, allelic odds ratio = 5.0, 95% CI 3.6-7.0). These results offer the potential to improve the safety profile of lumiracoxib by identifying individuals at elevated risk for liver injury and excluding them from lumiracoxib treatment.
Despite its relatively infrequent occurrence, drug-induced liver injury (DILI) is the leading cause of acute liver injury in the United States, an important cause of sporadic acute hepatitis in the community, a source of diagnostic and therapeutic challenges for treating clinicians and a common reason for premarketing and postmarketing drug withdrawals for pharmaceutical companies.
The selective cyclooxygenase-2 (COX-2) inhibitor, lumiracoxib, joins the long list of nonsteroidal anti-inflammatory drugs (NSAIDs) (benoxaprofen, bromfenac, ibufenac) withdrawn due to their association with DILI.1 When first introduced, lumiracoxib appeared to fulfill many of the desired attributes of an NSAID: three-fold to four-fold lower gastrointestinal (ulcer-related) complications than naproxen and ibuprofen and a slightly better cardiovascular disease track record than rofecoxib, another recently withdrawn COX-2 inhibitor.2 Ironically, although gastroenterologists should have welcomed the introduction of such an agent, it turns out that lumiracoxib has the potential for rare but serious hepatotoxicity. Worldwide, at least 20 cases of severe DILI associated with lumiracoxib have been reported, including 14 with acute liver failure, two deaths, and three liver transplants.3 Most cases occurred several months after starting lumiracoxib, but early presentations were also noted. Many cases involved daily doses exceeding 100 mg, but severe DILI was also reported in those patients who were prescribed 100 mg/day.
The U.S. Food and Drug Administration (FDA) issued a “nonapprovable” letter for lumiracoxib in 2007. Although the passing of one more NSAID is likely to be soon forgotten, there are two lessons to be learned for prescribers. Yet again, postmarketing surveillance has identified serious instances of DILI that were not foreseen in clinical trials. In the large TARGET (Therapeutic Arthritis Research and Gastrointestinal Event Trial) study, 2.6% had aminotransferase (AT) elevations greater than three times the upper limit of normal (3× ULN). There were six cases of probable or possible “clinical hepatitis”, but all resolved with cessation of the drug, and there were no reports of liver failure. Parallels can be drawn with troglitazone.4 However, whereas the relative rarity and unpredictability of many or now most causes of DILI has been recognized for more than 50 years, the genetic basis for such a host of susceptibility factors has been slow to document reliably since rare family clustering studies and indirect susceptibility tests were reported at least 25 years ago.5 The addition of lumiracoxib to the growing list of agents for which susceptibility to DILI has been linked to human leukocyte antigen (HLA) genotypes, as reviewed recently in Hepatology,6 provokes further consideration of the mechanistic significance and clinical utility of such associations.
The observations of Singer et al., who carried out a pharmacogenetic case-control analysis of participants enrolled into the two TARGET trials, are of particular interest.7 In the first phase of their study, 41 subjects with serum alanine aminotransferase (ALT) or aspartate aminotransferase (AST) > 5× ULN (“cases”) and 176 age-matched, sex-matched, race-matched, and clinical trial–matched individuals (who took lumiracoxib but had normal ALT/AST; “controls”) were recruited for a genome-wide association study (GWAS). This was performed using the Affymetrix assay 6.0, which can detect more than 900,000 single-nucleotide polymorphisms (SNPs). Several SNPs from the major histocompatibility complex (MHC) class II region on chromosome 6 were significantly represented in cases with DILI, the leading contender being rs9270986 (P = 2.8 × 10−10). Thirteen SNPs, including seven from phase 1 of the trial, were reevaluated in the second (validation) phase of the study, involving 98 cases and 405 lumiracoxib-exposed controls, respectively. Cases were defined here by ALT/AST > 3× ULN. The results of the replication phase confirmed the association of lumiracoxib-related DILI with the principal SNPs identified earlier, but did not find a similar relationship with cases of DILI drawn from small groups of controls receiving ibuprofen (n = 18) or naproxen (n = 9). Finally, fine mapping of the top SNPs showed strong association with a well-characterized MHC haplotype (HLA-DRB1*1501-HLA-DQB1*0602-HLA-DRB5*0101-HLA-DQA1*0102; most significant allele P = 6.8 × 10−25, allelic odds ratio = 5.0; 95% confidence interval [CI] = 3.6-7.0). Of these alleles, HLA-DQA1*0102 had the best negative predictive value (99%) and sensitivity (73.6%) in identifying cases at risk.
Before examining the implications of this study, it is worthwhile to look at the wider perspective of host/drug factors influencing susceptibility to DILI. Although the total dose of drug is critical in dose-dependent hepatotoxicity (e.g., acetaminophen), the relevance of this to idiosyncratic drug reactions is overshadowed by other host characteristics such as age, sex, comorbid illnesses, and coprescribed medications.8 A genetic predisposition to DILI is well recognized for drugs (phenytoin, sulfonamides) linked to hepatic injury as part of systemic hypersensitivity (“reactive metabolite syndrome”) and has been recognized for halothane.5 Other than these examples, the genetic contribution to DILI has only slowly been recognized, perhaps partly because of studies in the 1990s that showed a lack of association between HLA markers and DILI.9 Although some HLA markers were overrepresented in some cases (e.g., HLA A-11 in 75% of cases of diclofenac hepatitis), no overall association between specific HLA alleles and DILI could be discerned. Another limitation was the use of insensitive serological methods to determine HLA status instead of high-resolution genotyping on large case and control populations that is currently favored. These studies were also underpowered to detect meaningful associations with individual drugs. This poses a considerable challenge because cases of DILI are infrequent (typically between 1 and 10 per 100,000 persons exposed) and collating a case series requires considerable collaborative efforts. Furthermore, careful case definition is necessary; for DILI, this itself poses a considerable challenge.
Studies have usually used one of the causality scoring systems, such as the CIOMS (Council for International Organizations of Medical Sciences), which although laudable in many respects, lack sensitivity and specificity for several phenotypes of DILI, as reviewed elsewhere.10 Thus, in the lumiracoxib study, only 41 cases were included, and these were defined by ALT/AST changes and not by “clinical hepatitis”, which differs from another GWAS study that enrolled patients with flucloxacillin-associated DILI who had clinical features of liver disease.11 As a result of these logistic limitations, it is pertinent to consider whether the reported association between lumiracoxib-related AT elevations (DILI) and the HLA allele/extended haplotype is clinically meaningful. This cannot be conclusively determined from a study of this size, but supportive arguments have been put forward. Singer et al. noted the increasing sensitivity with increases in ALT rise; all patients with ALT > 20× ULN carried the specific HLA haplotype. Also, all three cases with substantial serum bilirubin increases that fulfilled “Hy's law” (ALT/AST > 3× ULN; serum bilirubin > 2 ULN), a reliable marker for high probability of significant hepatotoxicity,12 also carried the implicated HLA alleles. In other respects, the study by Singer and colleagues fulfills the necessary requisites for a GWAS: proper case definition (albeit by biochemical and not clinical presentation), matched controls in a ratio of cases:controls of 1:4, use of a replication cohort, and correction of P value for multiple comparisons.13
At the end of all this, what are the implications of this study in terms of pathogenesis of DILI and whether these observations can be used to prevent DILI in the future?
The physiological role of HLA class I (A, B, and C) and class II (DP, DQ, and DR) molecules on the cell surface is to present endogenous (class I) or exogenous material such as drugs (class II) to T lymphocytes through engagement with the T cell receptor. Recognition of small molecular weight drug/drug metabolites by T cells will occur either if presented in combination with a protein (“hapten” hypothesis) and MHC class II molecule (MHC peptide-complex), or by direct engagement with the MHC molecule (“pharmaceutical interaction” concept).14 In either scenario, it is conceivable that alterations in MHC alleles will disrupt proper drug–T cell engagement. The species differences in MHC restriction would account for the failure to predict human hepatotoxicity despite apparent safety in animal models.
In the study by Singer et al., there were no functional analyses that could shed light on the precise mechanisms of lumiracoxib-related DILI. It is, however, interesting that lumiracoxib is bioactivated to a reactive quinone imine,15 and possibly noteworthy that the structure of lumiracoxib closely resembles diclofenac. The latter is also associated with hepatotoxicity, and has metabolic pathways that can generate reactive metabolites capable of forming adducts with hepatic proteins and evoking an immune response.16 On the other hand, lumiracoxib shows no structural similarity to abacavir, which is associated with a severe cutaneous hypersensitivity reaction linked to one of the same HLA haplotypes (HLA-B*5701) as lumiracoxib. Interestingly, association with HLA-B*5701 is also shared with flucloxacillin, a synthetic penicillin associated with severe cases of drug-induced cholestasis.11
Unlike autoimmune hepatitis where specific HLA alleles can determine disease severity or treatment outcome, only limited genotype-phenotype correlations have been noted for instances of DILI. Interestingly, one of the same HLA haplotypes associated with lumiracoxib toxicity (HLA-DRB1*1501) is overrepresented among cases of liver injury resulting from amoxicillin-clavulanate.17 However, the latter causes early onset (<25 days) liver toxicity and has a completely different histologic pattern (mainly cholestatic injury), which differs from the usual late-onset hepatocellular reaction with lumiracoxib. Other recent associations of specific HLA alleles with DILI are listed in Table 1 and have been reviewed recently in Hepatology.6 It should be pointed out that not all HLA phenotypes are associated with increased susceptibility to DILI; HLA-DRB1*07 family of alleles conferred a reduced risk of DILI with amoxicillin-clavulanate as compared with population controls and treated nonaffected cases (odds ratio = 0.26 and 0.18, respectively).18 Overall, in most cases of DILI, the presence of a particular HLA allele is neither sufficient nor necessary for a particular adverse effect to occur. In addition to known and unknown host and environmental factors, the contributions of polymorphisms within drug-metabolizing systems, biliary transporters, and both innate and adaptive immune response pathways, as well as antioxidant, antiapoptosis, and other cell protective genes, need to be considered.6 It also remains possible that particular HLA alleles are in linkage disequilibrium with cardinal “susceptibility genes”, as turned out to be the explanation for the association between HLA A3 and C282Y, which led to the common form of genetic hemochromatosis.19
|Drug||Therapeutic Group||HLA Allele||Odds Ratio for Developing DILI (95% CI)|
|Ticlopidine||Antiplatelet agent||HLA A*3303||36.5 (7.3-184)|
|Flucloxacillin*||Antibiotic||HLA B*5701||80.6 (22.8-284.9)100 (20.6-485.8)|
|Ximelagatran*||Oral direct thrombin inhibitor||HLA-DRB1*0701||4.4|
|Lapatinib*||Tyrosine kinase inhibitor used in advanced breast cancer||HLA-DQA1*201||9 (3.2-27.4)|
Many consider the era of pharmacogenomic explanations for idiosyncratic adverse drug reactions to have begun with recognition of the association between hypersensitivity reactions to abacavir, a human immunodeficiency virus (HIV) protease inhibitor and HLA B*5701.20 Screening subjects for this HLA allele and withholding abacavir from those carrying it has almost completely abolished such reactions. However, unlike most cases of DILI, abacavir reactions are quite frequent (5%), and use of common agents like antimicrobials and NSAIDs is not usually subject to the same complex considerations as highly active antiretroviral therapy for HIV. A similar HLA-based screening strategy to exclude DILI is therefore unlikely to be logistically plausible or cost-effective unless screening costs become cheaper. In the case of lumiracoxib, excluding carriers of the HLA-DQA1*0102 allele would reduce the frequency of DILI to 1% but at the expense of excluding a considerable proportion (34%) of carriers, because less than 6% would actually develop hepatotoxicity. An alternative pharmacogenetic strategy is to restrict testing to those at increased risk of adverse drug reactions. For example, the FDA recommends screening Han Chinese patients for HLA-B*1502 before starting carbamazepine.21 Such screening is likely to be cost-effective because the allele in question is relatively common in that ethnic group (8%-12%) and, further, the odds ratio of developing a severe cutaneous reaction in persons carrying that allele is extremely high (>2500). This strategy would be useless in Caucasians who do not carry that specific HLA allele but also can develop similar reactions with carbamazepine. Likewise, a selective screening protocol cannot be applicable to lumiracoxib recipients because of the failure to identify specific characteristics that could be associated with a risk of DILI.
In the final analysis, routine pharmacogenetic testing would come down to costs, availability of alternative treatment options, and logistics (turn around times). Promising times lie ahead for the prospects of pharmacogenomic discovery to help unravel the multiple interactive mechanisms of DILI,6 but their impact on preventing DILI in the near future is still likely to be limited.
- 3Drug safety update: Lumiracoxib hepatotoxicity. Available from http://www.mhra.gov.uk/home/idcplg?IdcService=GET_FILE&dDocName=CON2033077&RevisionSelectionMethod=LatestReleased. Accessed December 2, 2010.
- 10Schiff's Diseases of the Liver, ed. 10. Philadelphia: Lippincott Williams & Wilkins; 2007., , , , eds.