PharmGKB Summary: Very Important Pharmacogene Information for Human Leukocyte Antigen B (HLA-B)

Introduction

The human leukocyte antigen B (HLA-B) gene is a member of the major histocompatibility complex (MHC), a region of the human genome located on chromosome 6. The MHC (also known as the human leukocyte antigen (HLA) complex) includes three subregions, designated as class I, class II and class III. Each of these subregions contains a variety of genes that mainly code for proteins involved in the immune system. HLA-B is part of the class I group, along with HLA-A and HLA-C, all three of which code for their eponymous proteins [1]. Class II genes include HLA-DR, HLA-DP and HLA-DQ [2], and class III genes include complement components and cytokines such as complement factor B (CFB) and members of the tumor necrosis factor (TNF) family [1, 3]. The MHC is a large region of the genome, and contains many other genes besides the ones listed above; please see Horton et al. [1] for more details. The HLA genes are important within the field of pharmacogenetics: variations within these genes have been associated with severe drug reactions, as well as changes in how well a patient responds to a drug.

The HLA-B protein and the other class I group members are cell-surface molecules responsible for the presentation of endogenous peptides to CD8+ T-cells, and exist on almost all nucleated cells. This is in contrast to class II molecules, which display exogenous peptides to CD4+ T-cells, and are present only on antigen presenting cells (APCs) such as macrophages or dendritic cells [2, 4]. This presentation of peptides to T-cells assists in the recognition of pathogens [2]. As a class I molecule, most of the peptides that HLA-B presents come from the normal breakdown of host cellular proteins, and are recognized by the immune system as such (i.e. “self” peptides). However, when a cell becomes infected by a pathogen, the proteins presented will be from the pathogen and recognized as foreign or “non-self”. T cell antigen receptors (TCRs) on CD8+ cytotoxic T cells are responsible for this recognition, and will stimulate an immune reaction that destroys the cell [5].

Class I molecules are expressed in a codominant fashion, and humans inherit a set of HLA-A, B and C genes from each parent. Therefore, given allelic variations within these genes, up to six different class I molecules can be expressed on a cell surface. HLA-A, B and C are heterodimers consisting of an α chain, encoded by their respective genes, and a protein known as β2-microglobulin, which is encoded on chromosome 15. The α chain of HLA-B has four domains: one cytoplasmic, one transmembrane, one which binds to CD8+ cytotoxic T cells, and one which makes up a peptide-binding groove, where the presented peptide is nestled [5]. This peptide-binding region of the gene is highly polymorphic, and allelic differences between class I genes are often due to variations within this region [2, 5]. Indeed, allelic variants of class I genes can differ from one another by up to 20 amino acids. Peptides bind to the groove through interaction with specific amino acid residues, so any amino acid changes due to allelic variation may affect the peptide-binding specificity of a class I molecule [5] (class II molecules have more flexibility in peptide-binding; see Janeway et al. [5]). The type of extensive polymorphism seen in HLA genes allows a wide variety of peptides to be presented, and likely evolved in order to effectively combat pathogens [5]. In addition to affecting the peptides capable of being presented, allelic variants in the HLA-B gene have also been associated with susceptibility and resistance to numerous diseases, as well as adverse reactions to a wide range of pharmaceuticals. This makes HLA-B highly relevant to pharmacogenetic research. This Very Important Pharmacogene (VIP) summary on HLAB is available with interactive links to genetic variants and drugs on the PharmGKB website at http://www.pharmgkb.org/gene/PA35056.

HLA-B allele frequencies and nomenclature

Due to the highly polymorphic nature of class I genes, a large number of HLA-B alleles have been identified. Information on the frequencies of over 2800 HLA-B alleles in populations worldwide can be found at The Allele Frequency Net Database (http://www.allelefrequencies.net/) [6]; allele frequencies for specific polymorphisms will be discussed within relevant sections of this review. Systematic nomenclature for these alleles is invaluable given their quantity. The HLA nomenclature committee has provided a detailed nomenclature to this end, and comprehensive information on the allele naming process can be found at their website at http://hla.alleles.org [7]. Briefly, all HLA alleles receive at least a four-digit name consisting of two sets of two digits separated by a colon, such as HLA-B*57:01. The first set of digits before the colon describes the type, typically the antigen designation used to describe the HLA alleles prior to genetic sequencing. The second set of digits indicates the specific allele, numerically ordered based on when the DNA sequence was discovered; this set of digits describes nonsynonymous substitutions only [7, 8]. This paper will only refer to the first one or two sets of digits. However, longer names, up to four sets of digits separated by colons and possibly a letter suffix, can be assigned if more detail is necessary, such as type or location of nucleotide substitution (e.g. synonymous or intronic) or resultant protein expression (e.g. null protein or cytoplasmic protein); for more information on this process, please refer to the HLA nomenclature website (http://hla.alleles.org).

HLA-B and disease associations

A number of HLA-B alleles or allele groups have been associated with susceptibility or resistance to particular diseases. These include HLA-B*53 and resistance to malaria [9, 10], HLA-B*51 and susceptibility to the inflammatory condition Beçhet's disease [11, 12], and HLA-B*46 and increased risk of Graves’ disease, an autoimmune disorder [13]. Two particularly strong disease associations are HLA-B*57 and HIV long-term non-progression, and HLA-B*27 and ankylosing spondylitis.

HLA-B*57 and HIV long-term non-progression

Without treatment, almost all people infected with HIV will ultimately progress to AIDS. However, a small percentage of patients do not advance, even long after the median progression time. These patients are referred to as long-term non-progressors or “elite controllers”, and HLA-B*57 alleles, particularly *57:01 and *57:03, are highly enriched in this group of individuals [14-21]. Though this association is well known, the mechanism by which it occurs remains unclear. Košmrlj et al. used computer algorithms to predict that less than half the number of unique peptides (derived from the human proteome) bound to the HLA-B*57:01 protein as compared to HLA-B*07:01 (a non- HIV-protective form of the molecule). The authors suggested that this affected repertoire development, leading to T-cells that had been exposed to fewer self-peptides. This in turn may lead to a higher frequency of T-cells that recognize viral peptides, such as those from HIV, as well as T-cells that are more cross-reactive toward mutant epitopes. These qualities would enable the T-cells to better control the HIV infection, keeping the viral load in check and thereby making the development of AIDS unlikely [16].

HLA-B*27 and ankylosing spondylitis

Ankylosing spondylitis (AS) is a chronic inflammatory rheumatic disease, affecting mainly the axial skeleton and sacroiliac joints. It leads to inflammatory back pain, as well as other clinical features including enthesitis and anterior uveitis [22]. Presence of HLAB*27 leads to the greatest risk for AS, and this form of HLA-B is found in over 90% of AS patients with European ancestry. However, only 1-5% of HLA-B*27 individuals will go on to develop AS, and not all alleles of HLA-B*27 are associated with its development. While HLA-B*27:05, *27:02, *27:04 and *27:07 do confer risk, other types such as *27:06 and *27:09 do not appear to be associated with the disease [23]. As with HIV and HLA-B*57:01, the mechanism behind this association is unclear, though multiple theories have attempted to explain the relationship. Several of these theories are summarized in a review by McHugh and Bowness [24], including the arthritogenic peptide hypothesis, which suggests that HLA-B*27 binds particular peptides that give rise to a cytotoxic T-cell response, the misfolding and unfolded protein response (UPR) hypothesis, which suggests that the accumulation of abnormally folded HLA-B*27 molecules leads to an inflammatory response, and the free heavy chain and homodimer hypothesis, where AS results from the immune recognition of monomeric or dimeric β2-microglobulin-free and peptide-free HLA-B*27 molecules [24]. Recent GWAS analyses have identified a number of non-MHC genes associated with AS susceptibility; these genes may help explain the mechanism of AS pathogenesis. For example, multiple genes within the interleukin-23 (IL-23) proinflammatory cytokine pathway were associated with AS, indicating that this may be a core immunological pathway underlying disease development. Additionally, multiple aminopeptidase genes, such as ERAP1 and ERAP2, have been associated with AS. The protein products of these genes are involved in peptide trimming prior to HLA class I binding and presentation, suggesting that HLAB*27 may be involved in disease development through the aberrant trimming or presentation of peptides [25-28].

HLA-B pharmacogenetics

HLA-B alleles have been associated with reactions to a large number of different drugs. Some of these associations have been well studied, such as HLA-B*57:01 and abacavir hypersensitivity, HLA-B*58:01 and allopurinol-induced severe cutaneous adverse reactions (SCARs), and HLA-B*15:02 and carbamazepine-induced Stevens-Johnson syndrome (SJS) and toxic epidermal necrolysis (TEN). Other associations that are not as widely studied, but still show significant results include HLA-B*57:01 and flucloxacillin-induced liver injury and HLA-B*15:02 and phenytoin-induced SJS and TEN. A list of the HLA-B alleles and pharmacogenetic associations discussed in this review (along with their positive and negative predictive values, if available) can be seen in Table 1. For the alleles presented in this paper, there is no difference in phenotype depending on whether one or two HLA-B alleles are present, and therefore the pharmacogenetic studies discussed only consider whether an individual has the allele or not. Consequently, in the tables throughout this review, “prevalence” of an HLA-B allele refers to how many patients carry that allele, and not the frequency of the allele in the population. Some studies do use allele frequencies in their statistical analyses, and these cases are noted within the tables.

Table 1. List of HLA-B alleles and their associated drug phenotypes.

The phenotypes listed for each drug are more likely to occur in patients who carry the associated allele. For example, carriers of *57:01 who are given abacavir have an increased chance of a hypersensitivity reaction as compared to non-carriers. When available, positive and negative predictive values are also shown. An interactive version of this table is available online at http://www.pharmgkb.org/gene/PA35056.

HLA-B risk allele	Drug	Phenotype	Ref	PPV	NPV	Ref
*57:01	Abacavir	Hypersensitivity reaction	See Table 2	55%	100%	[41, 57]
	Flucloxacillin	Drug-induced liver injury	[57, 71]
*58:01	Allopurinol	SCARs, MPE	See Table 3	1.5%	100%	[91]
*15:02	Carbamazepine	SJS/TEN	See Table 4	1.8%	100%	[151]
	Phenytoin	SJS/TEN	See Table 5	33%1	100%1	[111]

PPV, positive predictive value; NPV, negative predictive value; SCARs, severe cutaneous adverse reactions; MPE, maculopapular eruption; SJS/TEN, Stevens-Johnson Syndrome/toxic epidermal necrolysis

¹For phenytoin-induced SJS only.

Many other HLA-B alleles besides the ones discussed in this paper have shown associations with various drug phenotypes. A table of all the HLA-B allele and drug phenotype associations currently annotated by PharmGKB can be found on the website at http://www.pharmgkb.org/gene/PA35056.

HLA-B Testing

Several options exist for determining whether a patient carries a particular HLA-B allele. One is by direct sequencing of the gene, and assignment of a star allele after checking the sequence against known HLA-B alleles. Though this method provides high-resolution genotyping and is the most accurate, it is both time-consuming and expensive and is not widely used [8]. An alternative and commonly used approach is genotyping, where the sequence variants known to define a particular HLA-B allele are detected using polymerase chain reaction (PCR) primers specific for each variant [8]. Quality assurance studies done on the accuracy of HLA-B*57:01 testing using sequence-specific primer PCR (SSP-PCR) across multiple laboratories have shown very high sensitivity and specificity, indicating that laboratories using this method appear to be offering effective screening for the allele [29]. Another method that offers cost-effective, rapid and sensitive screening for HLA-B*57:01 or HLA-B*58:01 is flow cytometry. HLA-B*57:01 and HLA-B*58:01 belong to a serological group known as HLA-B17. B17 monoclonal antibodies can be used to identify individuals who carry the HLA-B17 serotype, and these individuals can then undergo further DNA typing to determine whether they carry the *57:01 or *58:01 risk alleles. Since B17 is normally present in less than 10% of the population, assaying for presence of B17 first allows greater than 90% of a patient population to be eliminated from unnecessary HLA testing [30-32].

It is also possible to test for the presence of an HLA-B allele by genotyping for one or more single nucleotide polymorphisms (SNPs) nearby in linkage disequilibrium with that allele. However, linkage disequilibrium can vary across populations, and this method may have lower accuracy [8, 33]. The HCP5 SNP rs2395029 has been suggested as a potential marker for abacavir-induced hypersensitivity, since the variant allele has shown strong linkage disequilibrium with *57:01 [34-37]. However, it is not in complete concordance with *57:01 [38, 39], and individuals with the *57:01 allele but not the rs2395029 variant allele [38-40] as well as individuals with the rs2395029 variant allele but not the *57:01 allele [34, 35, 38, 39] have been noted. This type of incomplete concordance could result in the denial of abacavir to individuals who are not at risk for a hypersensitivity reaction, or administration of abacavir to individuals who are at risk for a hypersensitivity reaction [8, 40]. Additionally, the studies showing strong linkage between *57:01 and rs2395029 have been conducted in populations of mainly Caucasian or Hispanic descent; the strength of the linkage between the alleles has not been confirmed in large African or Asian populations [8]. Several studies have noted that caution should be used when using rs2395029 as a surrogate marker for HLA-B*57:01 [8, 41, 42]. However, due to the inexpensive and straightforward nature of this screening method, some laboratories do choose to perform SNP testing over allele-specific PCR [8].

It is important to note that currently, the high level of polymorphism within the HLA genes makes HLA genotyping at a high resolution challenging [43-45]. Present sequencing methods can result in ambiguous typing results with an inability to resolve phase [43, 44]. Additionally, different alleles may share similar sequences within the sequenced region [45], and defining polymorphisms may lie outside the amplified region [43, 44]. These issues may be resolved through next-generation sequencing (NGS), which allows for clonal amplification and massively parallel sequencing. These two properties provide phase information and the ability to sequence more and larger regions of genes, including intronic regions [43, 44].

A list of commercially available genetic tests for various HLA-B alleles can be found on PharmGKB at http://www.pharmgkb.org/views/viewGeneticTests.action; a more comprehensive list can be found at the Genetic Testing Registry (http://www.ncbi.nlm.nih.gov/gtr/). Since HLA-B expression is co-dominant, HLA-B genotyping results are either “positive”, with the * allele being present in one or both copies of the gene, or “negative”, where no copies of the allele are present; there is no intermediate phenotype [8].

Important variants

HLA-B*57:01

Abacavir

In addition to being enriched in HIV long-term non-progressors, the *57:01 allele is also strongly and independently associated with the development of an abacavir hypersensitivity reaction (HSR). Abacavir is a nucleoside reverse transcriptase inhibitor (NRTI) used for treatment of human immunodeficiency virus (HIV). It is generally well tolerated, and common side effects include nausea, headache and diarrhea [46]. However, approximately 5-8% of patients experience a HSR within the first 6 weeks of treatment [8]. Symptoms of a HSR include at least two of the following: fever, rash, cough, gastrointestinal symptoms (e.g., nausea, vomiting, abdominal pain), dyspnea and fatigue [8]. These symptoms worsen with continued treatment, but typically improve within 24 hours after discontinuation. However, drug rechallenge after discontinuation of abacavir due to a HSR can result in symptom recurrence within a matter of hours, and potentially life-threatening allergic reactions [47, 48]. An overview of abacavir and *57:01 pharmacogenetic studies can be found in Table 2.

Table 2. Summary of abacavir and HLA-B*57:01 pharmacogenetic studies.

p-values and odds ratios (ORs) listed pertain to the risk for an abacavir hypersensitivity reaction (HSR) in patients carrying the *57:01 allele as compared to non-carriers. p-values and ORs were calculated by comparing the frequency of *57:01 in patients who developed an HSR compared to the frequency in abacavir-tolerant controls. Prevalence of *57:01 is listed for patients who developed an HSR (case) and in those who were abacavir-tolerant (control). An interactive version of this table is available online at http:// http://www.pharmgkb.org/haplotype/PA165956565, and is updated as new results come to our attention.

Population		*57:01 prevalence	p-value	OR (95% CI)	Ref
American	White (74%)				[49]
	Black (14%)	Case: 37/84 (44%)
	Hispanic (11%)	Control: 1/113 (0.9%)
	Other (2%)1
	White (66%)	Case: 91/1892 (48%)	8.4 × 10−23	21.4 (9.5 – 48.1)	[50]
		Control: 7/171 (4%)
	Hispanic (19%)	Case: 11/512 (22%)	2.1 × 10−4	30.4 (1.7 – 531)
		Control: 0/53 (0%)
	Black (15%)	Case: 3/372 (8%)	0.27	3.5 (0.4 – 35.5)
		Control 1/41 (2%)
	White (54%)	Case: 42/423 (100%)			[55]
		Control: 8/202 (4%)		1945 (110 – 34352)
	Black (46%)	Case: 5/53 (100%)
		Control: 2/206 (1%)		900 (38 – 21045)
Australian	Caucasian (88%)				[48]
	African (3.5%)	Case: 14/18 (78%)
	Indigenous (5.5%)	Control: 4/167 (2%)	< 0.0001	117 (29 - 481)
	Asian (3%)
British	Caucasian (94%)	Case: 6/13 (46%)			[50]
	Other (6%)1	Control: 5/51 (10%)	0.006	7.9 (1.5 – 41.4)
Canadian	White (66%)				[152]
	Black (16%)	Case: 18/18 (100%)
	Aboriginal (6%)	Control: 2/470 (0.4%)	<0.0001	6934 (321 – 149035)
	Other (12%)4
Ugandan		Case: 0/6 (0%)5			[56]
		Control: 0/241 (0%)

¹Unspecified remaining races

²Data using a ‘standard’ case definition: cases of hypersensitivity were either “definite/probable” or “possible”. Results differed but were still significant for White and Hispanic patients (and still non-significant for Black patients) using a “restrictive” case definition - only cases that were “definite/probable”. For “restrictive” case definition data, please refer to the paper directly.

³Immunologically confirmed HSRs. For clinically confirmed HSR data, please refer to the paper directly.

⁴Including 4.7% Indo-Asian, 3.1% Hispanic, 2% Metis and 2.2% Oriental or Unknown.

⁵Clinically confirmed HSRs.

In 2002, two separate research groups published the first evidence that HLA-B*57:01 was present in a significantly higher percentage of patients showing an abacavir HSR compared to patients with no reaction. These studies were conducted in North American [49] and Australian [48] populations, and both included 200 participants. This association was subsequently confirmed by another study within a UK population of 64 patients [50]. However, these three studies were all conducted using predominantly Caucasian males, limiting their scope. Despite this limitation, several clinics began implementing prospective screening of these alleles with success [51-53]. A later study recognized the significance of the allele in White female and Hispanic populations, but found no significant associations for the Black population from this study [54]. This was likely due to the lower number of Black patients within the study, and the fact that Black populations tend to have a lower carriage rate of the allele [54, 55] – African populations often have *57:01 allele frequencies of less than 2.5%, in contrast to European populations, who often have *57:01 frequencies of 6 – 7% [8]. Indeed, a study in Ugandan patients failed to find the *57:01 allele in either patients with an abacavir HSR or tolerant controls [56].

In 2007, a study known as SHAPE (which included a similar number of White and Black participants) found that Black patients did have fewer cases of abacavir HSRs. However, 100% of both White and Black patients who had immunologically confirmed HSRs were positive for the HLA-B*57:01 allele [55]. This suggested that though immunologically confirmed HSRs are rare among Black populations due to the reduced frequency of *57:01, the allele has the same clinical implications in both populations [55]. A definitive association between *57:01 and abacavir HSRs came in 2008 with the results of the PREDICT-1 study, a double-blind, prospective, randomized study with 1956 patients from 19 countries. Patients were observed for six weeks and separated into two categories: those that underwent screening for the HLA-B*57:01 allele and were excluded from treatment with abacavir if they tested positive, and those that underwent standard care without any screening. Abacavir HSRs were immunologically confirmed with skin patch testing. The results of the study showed that screening eliminated immunologically confirmed HSRs - 0% of the patients screened had a HSR, while 2.7% of the control population did. This gave the screening a negative predictive value of 100% [57]. The positive predictive value of HLA-B*57:01 for abacavir-induced HSRs is typically cited as 55%, implying that around half of all patients who are HLA-B*57:01 positive will not develop an abacavir HSR [41, 58, 59]. This indicates that other genes and environmental factors are likely involved in the development of an abacavir-induced HSR. Research in this area has been scarce, but several studies have suggested a member of the 70 kilodalton heat shock protein (HSP70) family as a potential factor [60-62].

This body of evidence led the FDA to implement a boxed warning in 2008, detailing the risk of a HSR for abacavir-treated patients with the HLA-B*57:01 allele. The FDA also recommended that all patients be screened before being treated, and to not use abacavir in HLA-B*57:01-positive individuals [63]. The European Medicines Agency (EMA) [64], and therapeutic guidelines from the Clinical Pharmacogenetics Implementation Consortium (CPIC) [8] and the Dutch Pharmacogenetics Working Group (DPWG) [65] also recommend genotyping for this allele prior to beginning abacavir treatment.

The HLA-B protein has no direct effect on abacavir pharmacokinetics or pharmacodynamics, and it is still unclear how the HLA-B*57:01 allele affects susceptibility to drug hypersensitivity. Several hypotheses exist. One is the hapten concept, which suggests that small compounds such as drugs (“haptens”), bind to the peptides bound to immune receptors such as HLA-B, causing T cells to react and stimulate an immune reaction [66]. Another is the p-i concept (pharmacological interactions with immune receptors), which proposes that drugs bind directly and reversibly to immune receptors, stimulating an immune reaction [66]. Recent evidence seems to support an alternative hypothesis. Two studies, both published in 2012, found that abacavir can bind non-covalently and with specificity to the F pocket of the peptide-binding groove of HLA-B*57:01 [67, 68]. Due to the amino acid residues unique to the *57:01 protein, abacavir can bind only to this particular form of HLA-B. The binding of abacavir to HLA-B*57:01 is believed to change the shape and chemistry of the antigen-binding cleft, and consequently the repertoire of peptides which can bind the molecule. Indeed, both of these papers, as well as an additional paper by Norcross et al., all identified specific changes in the peptides presented by HLA-B*57:01 in the presence of abacavir, as compared to HLA-B*57:01 in the absence of the drug [67-69]. Since T cells are trained to be tolerant to a particular repertoire of peptides during their development in the thymus, the alteration in the peptides that can be presented may mean that these new peptides are perceived as foreign. This change would stimulate CD8+ T cell production and response, and would manifest as an abacavir HSR [67, 68]. Indeed, CD8+ T cells are abundant in skin biopsies of patients who present with a rash during an abacavir HSR [70].

Flucloxacillin

HLA-B*57:01 is also associated with flucloxacillin drug-induced liver injury (DILI). Flucloxacillin is a semi-synthetic penicillin, used widely across Europe and Australia for staphylococcal infection [71, 72]. However, it is also associated with the development of cholestatic hepatitis, with risk factors being female sex, age over 55 years old and treatment duration of longer than two weeks [72]. A genome-wide association study found that a marker in complete linkage disequilibrium with HLA-B*57:01 was also strongly associated with flucloxacillin DILI. Further analysis found that patients with this allele had a 80-fold increased risk of developing DILI, as compared to those without the allele [71]. However, flucloxacillin DILI is a very rare condition, with an estimated prevalence of only 8.5 out of every 100,000 patients. Additionally, the positive predictive value of *57:01 for DILI is only 0.12%, meaning that the majority of patients with *57:01 will not develop flucoloxacillin-induced DILI. Indeed, this positive predictive value indicates that almost 14,000 white patients would need to be tested for *57:01 and excluded from receiving the drug in order to prevent one case of DILI; in non-white or mixed populations, this number could be higher [73]. Given the low positive predictive value and the high number needed to test to prevent one case, pre-treatment screening implementation in the clinic is likely not feasible at this time. In their discussion of flucloxacillin HLA-B*57:01 screening, Phillips et al. suggest that a more practical approach is to carefully monitor patients receiving flucloxacillin, and to consider their HLA-B*57:01 genotype only if there is biochemical evidence for hepatitis, at which point the drug can be stopped [73]. It is uncertain how the presence of the *57:01 allele leads to an increased risk for flucloxacillin-induced DILI. However, it does not appear to be through the same mechanism as *57:01 and abacavir hypersensitivity: Wuillemin et al. found that in the presence of the HLA-B*57:01 molecule, flucloxacillin stimulated T cells (including CD8+ T cells) according to the p-i concept [74]. In a later study, Wuillemin et al. also showed that DILI might be caused by the infiltration of cytotoxic CD8+ and CD3+ T cells into the liver [75]. Consideration of the results from both Wuillemin et al. studies provides a possible explanation for the connection between *57:01 and DILI.

HLA-B*58:01

Allopurinol

The HLA-B*58:01 allele is strongly associated with allopurinol-induced SCARs, which include hypersensitivity syndrome (HSS) and SJS/TEN [76]. Allopurinol is mainly used for conditions associated with hyperuricemia, such as gout and tumor lysis syndrome [77]. The drug works by inhibiting the enzyme xanthine oxidase, which is responsible for the conversion of hypoxanthine and xanthine into uric acid. In this manner, the drug lowers the amount of uric acid created in the body [76, 78]. Most side effects from allopurinol are mild, with the most common complaint being gastrointestinal upset [76, 78]. However, allopurinol has also been associated with severe adverse reactions. SJS and TEN (or SJS/TEN when referencing both) are two forms of the same condition. Both involve severe mucocutaneous blistering and epidermal detachment, and usually present one to three weeks after treatment begins. SJS and TEN are distinguished by the amount of skin detachment: SJS is classified as 1-10% detachment of body surface area (BSA), SJS and TEN overlap as 10 – 30% of BSA, and TEN as greater than 30% of BSA. Though the occurrence of these symptoms is rare (only two patients per million per year), SJS and TEN can be permanently disabling or even fatal. The mortality rate correlates with the level of skin detachment, ranging from a 1 – 5% mortality rate for SJS, to a 25 – 35% mortality rate for TEN [79, 80]. In contrast to SJS/TEN, HSS (also known as drug-induced hypersensitivity syndrome (DIHS) or drug reaction with eosinophilia and systemic symptoms (DRESS)) typically has organ involvement, such as hepatitis or nephritis, in addition to fever and a severe rash [81-83] The risk of developing SCARs during allopurinol treatment is 0.1 – 0.4% [76]. In some association studies with allopurinol-induced SCARs, only the relationship with SJS/TEN is discussed, and HSS is not used as an associated phenotype.

The first association between HLA-B*58:01 and SCARs came from a 2005 Taiwanese study on Han Chinese patients, which found that 100% of the 51 patients that developed allopurinol-induced SCARs had the HLA-B*58:01 allele, while only 15% of the 135 allopurinol-tolerant patients carried the allele. This gave an odds ratio of 580 for development of SCARs with this particular allele [81]. Follow-up studies in Thai, Korean, Japanese, Han Chinese and European populations also found significant results [84-92]. However, differences exist between these ethnicities when considering the magnitude of risk for developing SCARs. While Han Chinese and Thai patients show exceptionally high odds ratios for developing SCARs, European and Japanese *58:01 carriers show comparatively much lower risks for SCAR development (Table 3). This disparity in odds ratios is likely due to variations in *58:01 frequencies between ethnicities, rather than a differing effect of the *58:01 allele depending on ethnicity. While the Han Chinese and Thais tend to have *58:01 allele frequencies of around 8%, Europeans and Japanese have allele frequencies of approximately 1.4% and 0.5%, respectively [6, 88, 90, 91]. The *58:01 allele may also be associated with a more mild cutaneous adverse drug reaction (cADR), maculopapular eruption (MPE), which presents as a rash consisting of macules and papules. Cao et al. found that 100% of their Han Chinese patients who developed a MPE while receiving allopurinol carried the *58:01 allele [84]. In contrast, Lee et al. found that none of their 12 patients who presented with MPE carried the *58:01 allele, while 6 out of 11 of their patients who developed SCARs did. However, this may be due to the large presence of Caucasians in their study, a population with a low *58:01 frequency. Eleven of the patients with MPE were Caucasian, as were the 5 patients who developed SCARs but did not carry *58:01 [93]. Table 3 provides an overview of allopurinol and *58:01 pharmacogenetic studies, and indicates whether a study has analyzed SCARs, SJS/TEN, HSS or MPE.

Table 3. Summary of allopurinol and *58:01 pharmacogenetic studies.

p-values and odds ratios (ORs) listed pertain to the risk for allopurinol-induced adverse reactions in patients carrying the *58:01 allele as compared to non-carriers. p-values and ORs were calculated by comparing the frequency of *58:01 in patients who developed adverse reactions to the frequency of *58:01 in allopurinol-tolerant controls or in healthy population controls. Prevalence of *58:01 is listed for patients who developed adverse reactions (case) and in those who were allopurinol-tolerant (tolerant control) or from a healthy population (population control). An interactive version of this table is available online at http://www.pharmgkb.org/haplotype/PA165956630, and is updated as new results come to our attention.

Population	*58:01 prevalence	p-value	Odds ratio (95% CI)	Ref
Australian	Case (SCARs)1: 6/11 (55%)			[93]
	Case (MPE)2: 0/12 (0%)
European	Case (SJS/TEN; Caucasians only): 15/27 (55%)			[90]
	Case (SJS/TEN; Mixed ethnicities)3: 19/31 (61%)
	Population control4: 28/1822 (1.5%)
	vs Caucasians only	<1 × 10−6	80 (34 – 187)
	vs Mixed ethnicities	<1 × 10−8	61 (32 – 118)
	Case (SCARs)5: 16/25 (64%)			[86]
	Tolerant control6: 1/23 (4.3%)	5.9 × 10=4	39.1 (4.5 – 340)
	Population control6: 63/3200 (2%)		88.5 (38 – 208)
Han Chinese	Case (SCARs): 51/51 (100%)			[81]
	Tolerant control: 20/135 (15%)	4.7 × 10−24	580 (34 – 9781)
	Population control: 19/93 (20%)	8.1 × 10−18	393 (23 – 6625)
	Case (SCARs): 19/19 (100%)			[85]
	Tolerant control: 4/30 (13%)		230 (11.7 – 4520)
	Case (SCARs): 16/16 (100%)			[84]
	Case (MPE): 22/22 (100%)
	Case (All cADRs7): 38/38 (100%)
	Tolerant control: 7/63 (11%)
	vs SCAR cases	7.4 × 10−12	248 (13.5 – 4585)
	vs MPE cases	9.2 × 10−14	339 (18.6 – 6186)
	vs cADR cases	7.0 × 10−18	580 (32.1 – 10457)
	Population control: 80/572 (14%)
	vs SCAR cases	1.8 × 10−18	202 (12 – 3398)
	vs MPE cases	3.7 × 10−24	275 (16.5 – 4584)
	vs cADR cases	3.2 × 10−38	471 (28.7 – 7744)
	Case (HSS): 1/1 (100%)			[153]
	Case (SJS): 1/1 (100%)			[154]
	Tolerant control: 0/1 (0%)
Japanese	Case (SJS/TEN): 4/208 (20%)			[89]
	Population control: 6/9868 (0.6%)	<0.0001	40.8 (10.5 – 159)
	Case (SJS/TEN): 10/368 (28%)			[92]
	Population control: 6/9868 (0.6%)	5.4 × 10−12	62.8 (21.2 – 186)
	Case (SCARs): 3/3 (100%)			[155]
Kenyan	Case (TEN)9: 1/1			[156]
Korean	Case (SCARs): 24/26 (92%)			[88]
	Tolerant control: 6/57 (11%)	2.5 × 10−11	97.8 (18.3 – 522)
	Population control: 59/485 (12%)	2.5 × 10−16	83.0 (19.0 – 361)
	Case (SCARs): 9/9 (100%)			[87]
	Tolerant control: 41/432 (9.5%)	<0.001	179 (10.2 – 3152)
	Population control: 59/485 (12%)	<0.001	136 (7.8 – 2381)
Thai	Case (SJS/TEN): 27/27 (100%)			[91]
	Tolerant control: 7/54 (13%)	1.6 × 10−13	348 (19.2 – 6337)

SCARs, severe cutaneous adverse drug reactions; SJS/TEN, Stevens-Johnson Syndrome/toxic epidermal necrolysis; MPE, maculopapular eruption; cADRs, cutaneous adverse drug reactions; HSS, hypersensitivity syndrome.

¹4 of the 6 patients with *58:01 were of Southeast Asian origin, and two were Caucasian. All patients with SCARs but without *58:01 (i.e. the 5 remaining patients) were Caucasian.

²One patient was of Southeast Asian origin, the remaining patients were Caucasian.

³Includes 27 Caucasian patients, and four non-Caucasian patients of Pakistani, Cuban, Indian and Senegalese background; all four non-Caucasian patients carried the *58:01 allele.

⁴Controls came from Western Europe but specific ethnic information was not provided; the majority of these controls were assumed to be Caucasian.

⁵Cases were exclusively Caucasian.

⁶Ethnicity of the controls was not specified.

⁷MPE and SCAR.

⁸Allele frequencies.

⁹Born in Germany to Kenyan parents

Due to the strong associations seen between *58:01 and allopurinol, the CPIC recommends genotyping prior to treatment with allopurinol, and suggests that the drug should be contraindicated in patients with one or more *58:01 alleles [76]. The American College of Rheumatology also recommends that *58:01 screening be considered when assessing the risks of the drug, especially in populations with high frequencies of the allele, such as the Han Chinese or Thais [94]. Unlike abacavir, no clinical trials have been published that test whether genotyping for the presence of HLA-B*58:01 can reduce the number of SCAR or SJS/TEN cases in patients treated with allopurinol. Based on data from Han Chinese and Thai populations, the negative predictive value of this allele for SJS/TEN development is 100%, but the positive predictive value is only about 1.5% [76]. This indicates that most patients who carry the allele will not develop SJS/TEN. Discovery of new genetic, or non-genetic, factors that lead to SJS/TEN or SCAR development may help increase the positive predictive value. Studies on the mechanism of SCAR development in HLA-B*58:01 carriers have been limited. A recent study suggested that it is a metabolite of allopurinol, oxypurinol, which causes the hypersensitivity reaction in individuals with the HLA-B*58:01 allele: oxypurinol was found to bind to the F pocket of HLA-B*58:01 through the p-i mechanism with a higher affinity compared to allopurinol [95].

HLA-B*15:02

Carbamazepine

HLA-B*15:02 is strongly associated with SJS/TEN in patients taking carbamazepine, an anticonvulsant and mood-stabilizing drug. Along with epilepsy and bipolar disorder, carbamazepine (CBZ) is also used to treat a variety of other conditions, such as schizophrenia, trigeminal neuralgia and carpal tunnel syndrome [96]. As with *58:01, the allele frequency of *15:02 varies worldwide. Han Chinese have an average allele frequency of almost 6%, but this value can range anywhere from 1.9% to 12.4% depending on the population [6]. Additionally, other Chinese populations, such as the Bulang, can have an allele frequency of close to 36% [97]. Thai and Malaysian populations also have high *15:02 allele frequencies, with average allele frequencies of close to 8%. In contrast, Koreans show a frequency of 0.3%, Japanese a frequency of 0.1%, and Caucasians a frequency of only 0.06% [6]. These percentages correlate with the number and strength of studies finding significant results linking *15:02 with the development of SJS/TEN: studies in Caucasian [98-100] and Japanese [89] patients have been limited and have shown exclusively non-significant results. One study in Koreans found a significant association when comparing allele frequencies for SJS/TEN patients against population controls, but no significant association when comparing against carbamazepine-tolerant controls [101]. In contrast, studies in Han Chinese are numerous and show very high odds ratios for CBZ-induced SJS/TEN [100, 102-107]. Studies in Indian [108], Thai [109-111], Malaysian [112, 113] and Singaporean [114, 115] populations have also seen significant associations. These pharmacogenetic studies are summarized in Table 4. Additionally, three meta-analyses (not included in Table 4) that combined studies with Chinese, Korean, Malaysian and Thai patients, all found odds ratios of approximately 80 for the development of CBZ-induced SJS/TEN in patients carrying the *15:02 allele [116-118].

Table 4. Summary of carbamazepine and *15:02 pharmacogenetic studies.

p-values and odds ratios (ORs) listed pertain to the risk for carbamazepine-induced Stevens-Johnson Syndrome or toxic epidermal necrolysis (SJS/TEN) in patients carrying the *15:02 allele as compared to non-carriers. p-values and ORs were calculated by comparing the frequency of *15:02 in patients who developed SJS/TEN to the frequency in carbamazepine-tolerant controls or in healthy population controls. Prevalence of *15:02 is listed for patients who developed SJS/TEN (case) and in those who were carbamazepine-tolerant (tolerant control) or from a healthy population (population control). An interactive version of this table is available online at http://www.pharmgkb.org/haplotype/PA165954769, and is updated as new results come to our attention.

Population	*15:02 prevalence	p-value	Odds ratio (95% CI)	Ref
British1	Case: 0/2 (0%)			[98]
	Tolerant control: 0/43 (0%)
Canadian2	Case: 3/9 (33%)3			[99]
	Tolerant control: 1/87 (1%)4	0.002	38.7 (2.7 – 2240)
European	Case: 4/12 (33%)5			[157]
	Cases: 0/20 (0%)			[100]
	Tolerant control: 0/43 (0%)	NS
	Population control: 4/8862 (0.1%)	NS
Han Chinese	Case: 9/9 (100%)			[106]
	Tolerant control: 11/80 (13.8%)	<0.001	115 (6.3 – 2111)
	Population control: 11/62 (17.7%)	<0.001	85.1 (4.6 – 1570)
	Case: 16/17 (94%)			[107]
	Tolerant control: 2/21 (9.5%)	<0.0001	152 (12 – 1835)
	Population control: 17/185 (9.2%)	<0.0001	158 (19 – 1266)
	Case: 44/44 (100%)			[102]
	Tolerant control: 3/101 (3%)	3.1 × 10−27	2504 (126 – 49522)
	Population control: 8/93 (8.6%)	1.4 × 10−21	895 (50 – 15869)
	Case: 59/60 (98.3%)			[123]
	Tolerant control: 6/144 (4.2%)	1.6 × 10−41	1357 (193 – 8838)
	Case: 13/18 (72.2%)			[105]
	Tolerant control: 12/93 (12.9%)	<0.001	17.6 (5.3 – 58.1)
	Population control: 10/93 (10.8%)	<0.001	21.6 (6.4 – 73.3)
	Case: 24/26 (92.3%)			[130]
	Tolerant control: 16/135 (11.9%)	3.5 × 10−18	89.3 (19 – 414)
	Case: 8/8 (100%)			[124]
	Tolerant control: 4/50 (8%)		184 (33.2 – 1021)
	Population control: 6/71 (8.5%)		173 (36 – 835)
	Case: 99/112 (88%)			[104]
	Tolerant control: 11/152 (7%)	5.8 × 10−43	97.6 (42 – 227)
	Case: 8/35 (22.9%)			[103]
	Tolerant control: 2/125 (1.6%)	0.000	18.2 (3.7 – 90.7)
	Case: 41/53 (77.4%)			[100]
	Tolerant control: 4/72 (5.6%)	< 0.001	58.1 (17.6 – 192)
	Population control: 60/710 (8.5%)	< 0.001	37.0 (18.5 – 74.2)
Indian	Case (SJS only): 6/8 (75%)			[108]
	Tolerant control: 0/10 (0%)	0.0014	71.4 (3.0 – 1698)
Japanese	Case: 0/7 (0%)			[89]
Korean	Case: 1/7 (14.3%)			[101]
	Tolerant control: 2/50 (4%)	NS	23.3 (0.9 – 634)
	Population control: 2/485 (0.4%)	0.042	40.3 (3.2 – 506)
Malaysian	Case: 12/16 (75%)			[112]
	Population control: 47/300 (15.7%)	7.9 × 10=6	16.2 (4.6 – 62.4)
	Case (SJS only): 6/6 (100%)6			[113]
	Tolerant control: 0/8 (0%)7	0.0003
Singaporean	Case: 5/5 (100%)8			[114]
	Tolerant control: 1/10 (10%)9		27.2 (2.7 – ∞)
	Case: 13/13 (100%)10			[115]
	Tolerant control: 3/26 (H.5%)11	6.9 × 10−8	181 (8.7 – 3785)
Thai	Case (SJS only): 6/6 (100%)			[110]
	Tolerant control: 8/42 (19%)	0.0005	25.5 (2.7 – 243)
	Case: 37/42 (88%)			[111]
	Tolerant control: 5/42 (12%)	2.9 × 10−12	54.8 (14.6 – 205)
	Case: 32/34 (94.1%)			[109]
	Tolerant control: 7/40 (17.5%)	<0.001	75.4 (13.0 – 719)

SJS, Stevens-Johnson Syndrome.

¹Caucasian only.

²Pediatrics. Multiple ethnicities included, please refer to the paper directly for more information.

³All three case patients with the allele were of Asian ancestry, countries unspecified.

⁴The one tolerant control with the allele was of Asian ancestry, country unspecified.

⁵Four patients with *15:02 were of Asian ancestry (Vietnam, China, Cambodia and Reunion Island). Remaining 8 patients without *15:02 were Caucasian (Germany, France)

⁶Four Malaysian, two Chinese.

⁷Six Malaysian, two Chinese.

⁸Pediatrics. Two Chinese, three Malaysian

⁹Pediatrics. Seven Chinese, two Malaysian, one Indian

¹⁰10 Chinese, three Malaysian

¹¹20 Chinese, six Malaysian

Due to the low frequency in Korean, Japanese and Caucasian populations, screening for exclusively *15:02 may not be sufficient from a global perspective. Other alleles have shown significant relationships with SJS/TEN within these ethnicities, such as HLAB*07:02 [98] and HLA-A*31:01 [100, 119-121]. Several studies in Japanese, Korean and Han Chinese patients suggest that carriers of the HLA-B*15:11 alleles [101, 105, 122] have an increased risk of developing SCARs. The average allele frequencies for *15:11 in Japanese and Korean populations are higher than for *15:02, with values of 0.4 - 0.8% for the Japanese, and 1.6% for Koreans [6]. Both *15:02 and *15:11 are part of the same HLA-B75 serotype [101, 122], and the two alleles share a 98.6% amino acid sequence homology [105]. Therefore, they may share similar structures that lead to the triggering of an immune reaction when carbamazepine is administered [105]. Studies on *07:02, *31:01 and *15:11 are limited compared to those on *15:02, and research into alleles beyond *15:02 may help improve predictive genetic testing for SJS/TEN.

Unlike other HLA-B alleles, the associations between HLA-B*15:02 and carbamazepine are phenotype specific. Many of the studies mentioned above ([98-100, 104, 106, 110, 113, 114, 116, 123, 124]) also looked for associations with HSS, as well as MPE. However, none of the studies found any associations between *15:02 and carbamazepine-induced HSS or MPE, indicating that the allele may be particular to the development of SJS or TEN. One study has assessed the link between *15:11 and carbamazepine-induced HSS, and found no significant association [101]. In contrast, HLA-A*31:01 has shown strong associations with the development of carbamazepine-induced HSS or MPE in European [99, 100, 120], Han Chinese [100, 123], Japanese [121] and Korean [101] populations, as well as in a meta-analysis which included patients of all four aforementioned ethnicities [116].

A 2011 study assessed the value of genotyping prior to carbamazepine treatment, with close to 4500 Taiwanese subjects of Han Chinese descent participating. The 367 patients who were found to be positive for *15:02 were told not to take the medication, while the remaining 4120 took the drug as normal. Due to ethical considerations, historical incidences of SJS/TEN were used as a control. While no cases of SJS/TEN occurred in the study, historical data estimations found that 10 cases of SJS/TEN would have likely appeared in the study cohort, a significant difference [125]. While those of Caucasian ethnicity do tend to have lower frequencies of the *15:02 allele, genotyping prior to treatment is still valuable for these individuals, as they may be unaware of Asian ancestry, or fail to alert their doctor to their heritage. At this time, the US Food and Drug Administration (FDA) recommends genotyping for *15:02 prior to treatment with carbamazepine in all Asian populations, though does not make recommendations for patients of other ethnicities [126]. The negative predictive value of this allele for patients in Taiwan is suggested to be 100%, and the positive predictive value 7.7% [126]. This low positive predictive value implies that additional genetic or non-genetic factors likely play a role in the development of SJS/TEN in patients taking carbamazepine. Despite the low positive predictive value, it may be advisable to avoid the drug in *15:02-positive patients, given that there are effective alternatives to carbamazepine [119, 126]. Indeed, both CPIC and the Canadian Pharmacogenomics Network for Drug Safety (CPNDS) recommend that a different agent be used if a patient is found to be a carrier of the *15:02 allele, due to a strong increased risk for SJS/TEN [33, 119].

The mechanism of CBZ-induced SJS/TEN is still unclear. In a recent study, Wei et al. found that peptide-loaded HLA-B*15:02 presented carbamazepine to cytotoxic T lymphocytes without any prior processing or drug metabolism. Only HLA-B*15:02 could bind CBZ, as opposed to HLA-B*15:01, *15:03, *40:01 or *51:01. Endogenous peptides already loaded onto the molecule were found to be required before CBZ could be presented [127]. The authors suggested that the binding of CBZ to HLA-B*15:02 activates and induces clonal expansion of cytotoxic (CD8+) T lymphocytes, eliciting a severe immune reaction that leads to SJS or TEN [127]. The skin reaction seen in cases of SJS or TEN are thought to be due to CD8+ T lymphocytes, which are found in abundance in skin blister cells of people with SJS/TEN, and are believed to release cytotoxic proteins that induce keratinocyte apoptosis [127-129]. One of the studies that looked at the mechanism of abacavir-induced HSR also found that CBZ binds to HLAB*15:02, but no mention was made regarding whether loaded endogenous peptides are necessary. However, the authors noted that there was a repertoire shift in the peptides bound to HLA-B*15:02 in the presence of CBZ, albeit at a smaller magnitude than that of peptides bound to *57:01 in the presence of abacavir. This change could lead to an immune reaction by the same mechanism suggested for an abacavir-induced HSR [67]. Further studies in this area should help elucidate the precise manner by which the *15:02 allele affects the development of SJS/TEN in patients taking carbamazepine.

Phenytoin and other antiepileptics

*15:02 is also associated with SJS/TEN in people taking phenytoin, another antiepileptic; pharmacogenetic results are presented in Table 5. Studies linking *15:02 with SJS/TEN include four in Han Chinese patients [130-133] and one in Thai patients limited to SJS cases only [110]. Additionally, a meta-analysis of four studies found a significant association between *15:02 and phenytoin-induced SJS/TEN [118]. Though these studies have shown statistically strong results, they have been limited in number and population size. One study in Thai children found no link between *15:02 and phenytoin-induced SCARs [134]. Variations within the CYP2C9 gene have also shown associations with phenytoin-related adverse reactions. Phenytoin is primarily metabolized to its inactive form by CYP2C9, and alleles that result in decreased enzymatic activity, specifically CYP2C9*2 and *3, have been linked with increased phenytoin concentrations [135-138] and an increased risk for neurological toxicities [139, 140] and cutaneous adverse drug reactions [133, 141]. Consideration of both HLA-B*15:02 and CYP2C9*2 and *3 may be important in any future clinical pre-treatment screening programs.

Table 5. Summary of phenytoin and *15:02 pharmacogenetic studies.

p-values and odds ratios (ORs) listed pertain to the risk for phenytoin-induced Stevens-Johnson Syndrome (SJS) or toxic epidermal necrolysis (TEN) in patients carrying the *15:02 allele as compared to non-carriers. p-values and ORs were calculated by comparing the frequency of *15:02 for patients who developed SJS/TEN to the frequency in phenytoin-tolerant controls. Prevalence of *15:02 is listed in patients who developed SJS/TEN (case) and in those who were phenytoin-tolerant (tolerant control). An interactive version of this table is available online at http://www.pharmgkb.org/haplotype/PA165954769, and is updated as new results come to our attention.

Population	*15:02 prevalence	p-value	Odds ratio (95% CI)	Ref
Han Chinese	Case (SJS/TEN): 7/15 (46.7%)			[130]
	Tolerant control: 15/75 (20%)	0.045	3.5 (1.1 – 11.2)
	Case (SJS/TEN): 8/26 (30.8%)			[131]
	Tolerant controls: 9/113 (8%)	0.0041	5.1 (1.8 – 15.1)
	Case (SJS): 1/2 (50%)			[132]
	Combined case (SJS/TEN)1: 10/29 (35%)
	Tolerant control: 9/113 (8%)	0.0012	6.1 (2.2 – 17.0)
	Case (SJS/TEN): 13/48 (27.1%)	0.0253	5.0 (2.0 – 13)	[133]
	Tolerant control: 9/130 (6.9%)
Thai	Case (SJS): 4/4 (100%)			[110]
	Tolerant controls: 0/7 (0%)	0.005	18.5 (1.8 – 188)
	Case (SCARs): 1/17 (5.88%)	0.35	0.35 (0.03 – 3.9)	[134]
	Tolerant control: 3/17 (17.6%)

¹Combined study results from [90,111,130].

²p-value for comparison in *15:02 frequencies between tolerant controls and combined cases.

³Bonferroni-corrected p-value.

Currently, the FDA recommends that physicians consider the risks associated with SJS/TEN in patients taking phenytoin who have Asian ancestry and carry the *15:02 allele, particularly in light of the evidence linking *15:02 with CBZ-induced SJS/TEN [142]. Phenytoin, carbamazepine, oxcarbazepine, phenobarbital and lamotrigine are all known as aromatic antiepileptic drugs (AEDs) due to the presence of an aromatic ring in their structure – symptoms of hypersensitivity were found to occur twice as often in patients given aromatic AEDs as opposed to non-aromatic AEDs (e.g. levetiracetam or topiramate) [143], suggesting that the presence of the ring may be involved in the higher risk for adverse reactions [144]. A recent study looking at children taking carbamazepine, oxcarbazepine or phenobarbital found that those who developed SJS had a higher frequency of the *15:02 allele as compared to tolerant controls or healthy population controls [145]. Wei et al., in their study on the mechanism by which *15:02 is associated with CBZ-induced SJS/TEN, noted that oxcarbazepine, which has a tricyclic ring structure similar to CBZ, was also capable of binding with HLA-B*15:02, though not as strongly as CBZ [127]. However, studies linking HLA-B*15:02 with oxcarbazepine-induced SJS or MPE have shown mixed results [131, 146-148]. Though no individual studies have found significant associations between *15:02 and lamotrigine-induced adverse reactions [130, 131, 149, 150], a meta-analysis of four studies did find a significantly increased risk for SJS/TEN for patients carrying *15:02 who receive lamotrigine [118].

Conclusions

The HLA-B gene has shown associations with a wide range of diseases and adverse drug reactions. Despite this, very little is understood about the mechanisms by which variations in an immune system gene can affect the propensity for certain pharmacological reactions or particular illnesses. However, some progress has been made in understanding the mechanisms behind abacavir HSRs and carbamazepine-induced SJS/TEN, particularly in the last couple years. Studies on abacavir HSRs have paved the way for pharmacogenetic implementation within the clinic: HLA-B*57:01 genotyping prior to abacavir treatment is one of the key examples of pharmacogenetics being used in routine medical practice. Given that HLA-B*58:01 and *15:02 have also shown strong pharmacogenetic associations, these alleles may also be good candidates for clinical integration. HLA-B has been shown to be an influential gene across many areas of medicine, and future studies should improve our understanding of its role in both disease and pharmacology.