Progress in understanding the genomic basis for adverse drug reactions: a comprehensive review and focus on the role of ethnicity
Abstract
A major goal of the field of pharmacogenomics is to identify the genomic causes of serious adverse drug reactions (ADRs). Increasingly, genome-wide association studies (GWAS) have been used to achieve this goal. In this article, we review recent progress in the identification of genetic variants associated with ADRs using GWAS and discuss emerging themes from these studies. We also compare aspects of GWAS for ADRs to GWAS for common diseases. In the second part of the article, we review progress in performing pharmacogenomic research in multi-ethnic populations and discuss the challenges and opportunities of investigating genetic causes of ADRs in ethnically diverse patient populations.
ADR: Adverse drug reaction; GWAS: Genome-wide association studies.
Adverse drug reactions (ADRs) are a major, preventable cause of morbidity and mortality. Germline genomic variation contributes to interindividual differences in drug response and the risk of ADRs. Understanding the genomic basis for drug-response and ADR risk is one of the primary goals of the field of pharmacogenomics. Initially, the field employed primarily candidate-gene association studies, focusing on genes with known roles in pharmacokinetics or pharmacodynamics. With improvements in genotyping technology, and taking lead from the field of genetics of common diseases, genome-wide association studies (GWAS) have increasingly been used to provide a more comprehensive view of the genomic landscape of drug response.
Previous reviews of GWAS in pharmacogenomics were published at a time when only few GWAS of ADRs had been reported, and only small numbers of loci had reached genome-wide levels of statistical significance [1]. Since that time, there has been a rapid increase in the number of GWAS performed for ADRs that has followed a similar trajectory to the early days of GWAS for common diseases (Supplementary Figure 1). However, the number of GWAS performed on ADRs still represents a small fraction of the total number of reported GWAS (Supplementary Figure 1). The objective of this article is twofold. First, we systematically review GWAS of ADRs published since 2010 to identify trends and themes emerging from these studies. In the second part of this manuscript we consider the role of ethnicity in furthering our understanding of the genomic basis of ADRs.
Progress in understanding the genomic basis of ADRs
We performed a comprehensive literature review to identify GWAS that examined ADR phenotypes from 2010 to present, using both the GWAS catalog maintained by the National Human Genome Research Institute of the USA [2] and PubMed (Figure 1). We identified a total of 560 publications, which, after manual curation, yielded 55 primary research articles describing a GWAS of an ADR phenotype. Of these, 38 had a lead SNP at p < 5 × 10-7 (Table 1). The 17 studies that did not report a lead SNP at p < 5 × 10-7 are shown in Supplementary Table 1. Below, we discuss themes emerging from these studies.
Sample size of GWAS of ADRs
One early observation from the field of pharmacogenomics was that, in contrast to GWAS of common disease that generally require thousands of cases and controls, GWAS of drug-response were able to detect loci with genome-wide levels of significance using relatively small sample sizes, often only dozens of cases and hundreds of controls [41]. Based on the 38 studies that reported variants with a p value of < 5 × 10-7 from our literature review (Table 1), the median sample size of the discovery cohorts was 829 individuals (range 47–5609) with a median number of cases in discovery cohorts of 117 individuals (range 14–1675). The smallest sample size for discovery that reached genome-wide significance was a study of allopurinol-related Stevens-Johnson syndrome and toxic epidermal necrolysis (SJS/TEN) that included 14 cases and 991 untreated healthy controls and detected an association with the HLA-B*58:01 allele at p = 5.39 × 10-12 with an odds ratio (OR) of 63 [3]. These observations suggest that GWAS of ADRs continue to be characterized by substantially smaller sample sizes than are typically used for common disease studies. Indeed, the median sample size for ADR GWAS (including those with or without a lead SNP at p < 5 × 10-7) is 457 individuals compared with 2389 individuals for common disease GWAS (based on nonpharmacogenomic GWAS in the GWAS catalog, Mann–Whitney U p = 6.1 × 10-15).
While these observations create a high level of optimism about the ability of GWAS to discover markers for ADRs even with moderately sized cohorts, it is nonetheless noteworthy that many published GWAS of pharmacogenomic phenotypes have reported null findings, in other words, the absence of loci with genome-wide levels of statistical significance. Since 2010, 17 of 55 (31%) published GWAS of ADRs reported no loci with genome-wide significance (Supplementary Table 1). This is similar to the percentage of GWAS of common diseases that have reported null findings (˜27% since 2005 in the GWAS catalog). This indicates that, despite their significantly smaller sample sizes, published GWAS of ADRs have a similar likelihood of detecting statistically significant associations compared with GWAS of common diseases.
Independent replication
Because of the possibility of false-positive associations detected by genetic association studies, particularly when using small sample sizes, replication in independent cohorts is essential. One common approach is to use a two-stage study design, in which loci identified in the discovery stage are carried forward for replication in a second independent cohort. This poses a unique challenge for pharmacogenomics research because of the low incidence of many ADR phenotypes, the nonstandardized manner in which these phenotypes are captured by medical records, and the resulting challenges of assembling multiple, large cohorts of appropriately-phenotyped cases and controls. Indeed, 13 of 38 studies listed in Table 1 (34%) did not include replication data in the initial report [4,6,8–10,17,21–22,28–29,31,36,40]. In some cases, the lead SNPs discovered by GWAS were validated in smaller sets of replication samples [5,19,27]. Notably, some recent studies did include replication in a second GWAS [11,30,39] or were replicated by other groups using patients of either the same ethnicity as the discovery population [27,28] or in populations of different ethnicities [15–16,29–31], providing strong evidence for the robustness and generalizability of these findings.
One strategy to further augment the number of patients available to study is to combine untreated healthy population controls with drug-treated cohorts. This approach has been successfully used in studies of clozapine-induced agranulocytosis, drug-induced SJS/TEN and carbamazepine (CBZ) induced hypersensitivity reactions [15,22,40], and can increase the statistical power to detect an association given a limited number of ADR cases and drug-treated controls. This approach is most likely to be beneficial in instances in which the prevalence of the ADR is low, and the percentage of healthy controls who would have developed the ADR had they received the drug is correspondingly small. In Supplementary Figure 2 we show, using simulated data, the effect of including healthy population controls on statistical power for GWAS of ADRs. The greatest gains in statistical power from this approach occur with a small number of cases, moderate effect sizes and limited number of treated controls (Supplementary Figure 2).
Are most ADRs immunological phenomena?
Many ADRs represent immunological phenomena due to an interaction between drug compounds and major histocompatibility (MHC) molecules in the host [42]. Accordingly, many of the associations detected in early GWAS of ADRs were with HLA alleles in the MHC region on chromosome 6 [1,43]. A prominent example is the association of abacavir-induced hypersensitivity with HLA-B*57:01, which was detected in 2002 by two independent groups [44,45]. Subsequently, screening for HLA-B*57:01 in HIV positive patients was shown to reduce the risk of abacavir hypersensitivity reaction [46], and testing for HLA-B*57:01 prior to abacavir administration is now used routinely in many jurisdictions [47,48]. Another early example of an immunologically-mediated ADR is the association of HLA-B*15:02 with SJS/TEN due to CBZ. This association was first reported in Han Chinese from Taiwan with an OR 2,504 [49]. The association was subsequently validated in other Asian populations including the Thai and Malay populations [50–52].
In contrast to these early results, we note that only 10 of the 38 loci (26%) associated with ADRs from GWAS reported since 2010 identified association with HLA alleles. This includes association of HLA-B*13:01 with dapsone hypersensitivity syndrome [23], HLA-A*31:01 with CBZ-induced hypersensitivity reactions [15,16] and HLA-DQA1–HLA-DRB1 with thiopurine-induced pancreatitis [38]. Notably, all of these ADRs represent drug hypersensitivity type phenotypes. These findings suggest that it may be possible to predict, a priori, which ADRs are most likely to have associations with HLA molecules based on the nature of the clinical phenotype. One notable feature of these ADR-associated HLA loci is the high ORs that they confer. Indeed, the median OR for HLA loci associated with ADRs in Table 1 is 5.3, compared with 2.9 for non-HLA ADR loci (Mann Whitney Exact p = 0.03). The highly penetrant effect of these HLA loci suggests that these drug hypersensitivity phenotypes may be more similar to monogenic traits, compared with the more polygenic nature of non-hypersensitivity ADR phenotypes.
In contrast to the studies described above, studies of non-hypersensitivity type ADRs have typically reported associations with loci outside the MHC region. For example, association of NUDT15 with thiopurine-induced leukopenia [39], EPHA5 with paclitaxel-induced neuropathy [27,28], RBMS3 with bisphosphonate-induced osteonecrosis of the jaw [14] and TCL1A with aromatase inhibitor-associated musculoskeletal toxicity [10]. These findings suggest that, in contrast to the trend observed from early GWAS for ADRs, most ADRs for which the genomic basis has been identified are not immunological phenomena.
Most ADR markers are in noncoding regions of the genome
One important theme emerging from GWAS of common diseases is that the vast majority (>90%) of disease-associated variants are in noncoding regions of the genome, suggesting the possibility of regulatory mechanisms leading to disease phenotypes and posing substantial challenges for investigating the biological mechanisms underlying these disease associations [53]. However, whether the loci identified by GWAS for ADRs are similarly likely to lie in noncoding regions of the genome has not been previously explored. The early reliance on candidate-gene studies in pharmacogenomics, which frequently focus on known functional variants, may have biased the view of the genetic landscape of pharmacogenomic loci toward coding and functional variants. In reviewing GWAS of ADRs performed since 2010, we note that only nine of 51 (18%) of the genomic variants identified are coding (including eight nonsynonymous variants and one synonymous variant), while 42 of 51 (82%) are noncoding variants. These figures indicate that, when viewed from the relatively unbiased perspective of GWAS, the genetic landscape of ADRs is similar to that of common diseases with regard to the likelihood of the associated SNPs occurring in coding versus noncoding regions of the genome.
When an association is detected with a coding SNP, demonstration of functionality is relatively straightforward. For example, a missense variant in SLCO1B1 associated with statin-induced myopathy [54–56] has been shown to result in diminished activity of the encoded transporter, leading to impaired statin uptake by the liver and higher plasma levels of the drug that could lead to myotoxicity [54,56–57]. In contrast, when a noncoding variant is associated with an ADR, it can be very challenging to determine the mechanism by which these variants influence drug-response and the risk of an ADR, or even whether these associations represent causal relationships with the ADR phenotype. Translating the findings from GWAS into biological insight and clinical actionability will require the ability to understand the mechanisms by which variants detected in GWAS influence drug response and to confirm whether these relationships are causal (i.e., SNP directly affects drug response) or correlative (i.e., due to linkage disequilibrium [LD] between the SNP and another, unobserved functional variant).
One intuitive hypothesis is that these noncoding variants may influence gene expression through regulatory mechanisms. In some instances, experimental evidence supports such mechanisms. For example, aromatase inhibitor-associated musculoskeletal toxicity has been associated with variants downstream of the TCL1A gene that were shown experimentally to alter the expression of TCL1A [10]. Similarly, a GWAS for lithium response identified a SNP in intron 6 of the GADL1 gene that was strongly associated with response to lithium therapy in bipolar I disorder [58]. Resequencing of GADL1 revealed a one-base deletion in intron 8 of the gene (IVS8+48delG) that was in complete LD with this variant and affected mRNA splicing of the GADL1 gene.
Recent progress in mapping regulatory regions of the genome has expanded our understanding of the role of noncoding DNA in disease pathogenesis. For GWAS of common diseases, 36.3% of noncoding SNPs are reported to overlap with DNase hypersensitivity sites, regions of the genome thought to be important in transcription factor binding and gene regulation [59]. We performed a similar analysis for the noncoding GWAS SNPs associated with ADRs in Table 1 using RegulomeDB [60] and found that 46.2% of ADR GWAS SNPs overlap with DNase hypersensitivity sites (chi-squared p = 0.3 vs GWAS SNPs of common disease). This suggests that, as is the case of GWAS SNPs associated with common disease, a substantial fraction of the noncoding SNPs associated with ADRs are likely to exert effects via gene regulation.
Ethnic diversity in pharmacogenomic studies
Interethnic differences in drug response have long been recognized as a valuable clue to the presence of genetic differences contributing to variation in drug efficacy or toxicity. Recently, several notable population-specific pharmacogenomic associations have been described, and in many cases the interethnic difference in phenotype displays strong correspondence with the frequency of the risk allele between populations. Most pharmacogenomic studies, as well as most of the clinical trials of medications, have been performed in predominantly European populations [61]. Very recently, there has been a trend of extending pharmacogenomic studies to other world populations. Whereas only one of seven (14%) GWAS performed for ADRs prior to 2010 was in a non-European population [1], 20 of the 55 (36%) ADR GWAS since 2010 were conducted in non-European populations, including two studies in Chinese populations, three studies in Korean populations, 13 studies in Japanese populations, one study in a Thai population and one study in a population of mixed Asian ancestry (Table 1 and Supplementary Table 1). To capture the full scope of human genomic diversity and fulfill the promise to improve drug safety in global populations, it will be essential to describe the diversity of pharmacologically important gene variants across populations in order to infer the potential risk of ADRs, drug efficacy and utility of pharmacogenomic markers in different populations.
Early population diversity studies focused on specific candidate genes and variants, such as CYP enzymes and drug transporters in a limited number of population groups [62–66]. One of the first comprehensive studies of pharmacogenomic variants surveyed 165 SNPs using the Affymetrix DMET assay in 2,188 individuals from three major East Asian populations, Caucasians and Africans [67]. This was followed by a larger survey of 1,156 variants in 19 global populations [68], and recently approximately 4,500 variants in the three major Singaporean populations (Chinese, Malays and Indians), as well as Europeans and Africans [69,70]. Collectively, these studies have identified numerous differences between global populations at specific pharmacogenomic variants. One of the most differentiated loci on a global level is the VKORC1 locus that encodes the drug target for warfarin. The frequency of the crucial -1639G (rs9923231) variant, associated with lower warfarin dose requirement, varies widely in different population groups: 91% in Chinese, 76% in Malays, 40% in Europeans and 13% in Indians [68,69]. This frequency distribution closely mirrors the differences in mean daily warfarin dose requirement in these populations: 3.5 mg/day in Chinese, 3.6 mg/day in Malays, 4.5 mg/day in Europeans and 6 mg/day in Indians [71,72].
Studies of population diversity at pharmacologically important gene variants have provided clues toward specific drugs for which populations may differ in drug response, pointing to gaps in knowledge and direction for future pharmacogenomic research. For example, Africans are enriched for variants that could explain higher ADR risk and poorer response to certain chemotherapeutic, antiretroviral and antituberculosis drugs [70], and East Asians harbor a higher frequency of a variant in a statin transporter [69] that may contribute to the higher risk of statin myopathy reported in that population [73]. Similarly, the ITPA rs1127354 variant is more common in Chinese (8–21%) and Japanese (13%) populations compared with Europeans (8%) and Africans (2%) [74], which may contribute to the higher incidence of ribavirin-induced anemia in Asians compared with Europeans [75].
Comparing trends of reported ADR rates with frequencies of their associated risk alleles across populations can provide further insight into the relationship between ADR risk and allele frequency. For example, the frequency of ADRs related to clopidogrel collected by the Health Sciences Authority in Singapore is significantly higher in the Chinese compared with Malay and Indian populations, consistent with the higher frequency of the CYP2C19*3 (rs4986893) loss-of-function variant in the Chinese population [69]. A major challenge in performing this type of population-level analysis is the paucity of high-quality data on ADR rates in different populations and the inherent limitations of passive reporting mechanisms to capture a significant proportion of all ADRs.
Consistency of GWAS findings across different ethnicities
To gain additional insight into the role of ethnicity in published GWAS results, we examined the 27 drug-ADR pairs represented by the 38 ADR GWAS in Table 1, and asked in how many had the lead GWAS SNP been investigated in one or more populations of different ethnicity compared with the original discovery cohort. Of the 27 drug-ADR pairs, we identified seven (26%) in which the lead GWAS SNP was studied in at least one other population of different ancestry, using either a GWAS or a candidate gene approach, for a total of 37 published studies (Table 2). In 33 of these 37 replication studies (89%), the lead GWAS SNP was successfully replicated with the same direction of effect at p < 0.05. This suggests that, in many cases, ADR markers detected by GWAS are shared across populations.
Cases in which pharmacogenomic loci are not replicated in independent populations may suggest that the initial discovery represented a false-positive result, that the replication is falsely negative, or that the variant exerts a population-specific effect. Disentangling these possibilities can be challenging. A recent example is the previously described association of GADL1 variants with response to lithium treatment. This association was initially discovered in an ethnically Chinese population from Taiwan [58]. Three subsequent studies in the Japanese and Indian populations failed to detect association of the rs17026688 variant with lithium response [109–111]. Despite the lower frequency of this variant in the Indian (MAF = 4%) and Japanese (MAF = 13%) compared with the Chinese (MAF = 27%) population [74], these follow-up studies were seemingly adequately powered (>80%) to detect this association based on the reported OR of 74, suggesting that type II error is unlikely. The reasons for these discrepant findings remain uncertain, but highlight the challenges of performing appropriate replication studies and suggest that careful attention to the ethnicity of the population, as well as consistency in phenotypic definition and ascertainment are essential.
One particular challenge of performing appropriate replication studies is the lack of consistent phenotypic definitions used. The application of standardized definitions of ADRs would facilitate the identification of common mechanisms involved and aid in the assessment of replication studies.
In some cases, population-specificity has been observed in either the magnitude of association or the specific ADR phenotype. For example, the association of HLA-B*58:01 and allopurinol-induced SJS/TEN is consistent across all populations studied, but the effect size is considerably larger in Asians than in Europeans (Table 2). In addition, HLA-A*31:01 is associated with CBZ-induced SJS/TEN and hypersensitivity syndrome (HSS) in Europeans, while in Asian population this marker is associated with CBZ-induced HSS but not with CBZ-induced SJS/TEN. Another example of a population-specific association is HLA-B*15:02 and phenytoin-induced SJS/TEN, which to-date has only been reported in Asian populations [112].
Opportunities presented by ethnicity
Despite these challenges posed by working with populations of diverse ethnicity, the availability of different world populations also provides several opportunities for pharmacogenomics research. One clear benefit for pharmacogenomic studies in populations of different ancestry is the ability to identify more genetic loci associated with a certain phenotype. One of the most well-known examples is the association between HLA-B*15:02 and CBZ-induced SJS/TENS that is observed only in South and East Asians. The allele is prevalent in Central and South Asians but almost absent in Koreans, Japanese, African–Americans and Europeans [113]. In Japanese and European populations, another HLA allele, HLA-A*31:01, is associated with CBZ-induced hypersensitivity [15,16]. Another example is thiopurine-induced toxicity, associated with TPMT loss-of function alleles (primarily *2, *3 and *4) in Europeans [114]. These alleles are very rare or absent in Asian populations (combined frequencies of *2 and *3 alleles is 0.006 in Chinese compared with 0.051 in Europeans), but the ADR occurs at least as frequently [69,115]. This suggested that there must be a novel genomic basis for thiopurine-induced hematological toxicity in Asian populations. Indeed, very recently, a nonsynonymous SNP in NUDT15 was discovered that is strongly associated with thiopurine-induced leukopenia in Koreans [39]. This variant is relatively common in Asian populations (10–13%) but rare in Europeans (2%) [39].
Similarly, variants in the COMT and TPMT genes are associated with cisplatin-induced ototoxicity in two multi-ethnic Canadian populations [116,117], but are very rare in Asian populations (frequency of COMT rs9332377 and TPMT rs12201199 is 0–1% in Chinese [69]). Recently, a GWAS for cisplatin-induced ototoxicity was reported which identified ACYP2 rs1872328 as a risk variant in a mixed population of predominantly European and African–American patients [20]. Interestingly, this variant is also absent in the Chinese population [74], again suggesting that this marker is unlikely to explain cisplatin-induced ototoxicity in populations of Chinese ancestry and that there may be additional, as yet undiscovered risk markers in the Chinese population.
Even in cases where specific variants have been robustly associated with a phenotype in multiple populations, the clinical validity can differ in different populations. Warfarin is one such example, where dosing algorithms including the CYP2C9*2 & *3 and VKORC1 -1639G variants explain approximately 55% of warfarin dose variability in Europeans and Asians, but only approximately 20% in African–American populations. This may explain why clinical trials of genotype-guided dosing of warfarin have yielded less favorable results in African–American populations [118]. This has led to the hypothesis that other genetic variants, either in CYP2C9 and VKORC1, or in other genes, may explain some of the unaccounted for variability in warfarin dose requirement in African–Americans. Accordingly, GWAS for warfarin dose in African–Americans have identified a novel SNP, rs12777823, in CYP2C9 that is associated with warfarin dose [119] and which was not detected in GWAS in Japanese and Europeans [120–122]. Additional novel SNPs in VKORC1 (rs61162043) and CYP2C9 (rs7089580) associated with warfarin dose in African–Americans have been identified through sequencing [123]. Very recently, an exome sequencing study in African–Americans identified a variant in FPGS (rs7856096) associated with warfarin dose requirement [124]. This variant is common in Africans (11–37%) and South Asians (2–9%) but rare or absent in other populations [74] and is associated with lower warfarin dose requirement (5.8 mg/week lower dose in carriers of this variant).
Studying populations of diverse ethnicities can also allow fine-mapping of causal variants. Integrating association signals from different populations can significantly narrow down the list of potential causative variants by leveraging on differential LD patterns between populations [125]. Such a strategy has been used effectively in the study of common disease phenotypes such as blood lipids and BMI [126,127].
Challenges posed by ethnicity
The first challenge ethnicity poses to genomic and pharmacogenomic studies is the added complexity it brings to the collective interpretation of studies of a particular phenotype. As in common diseases, replication of genetic association studies is crucial for establishing the authenticity of a genetic marker. However, as discussed above, studies can fail to replicate for a variety of reasons such as insufficient power, inconsistent phenotype definition and different environmental background resulting in different gene-environment interactions. Failure to replicate genetic associations in a population of different ancestry to the original study adds to the complexity of its interpretation with additional explanations such as differential LD structures affecting tagging efficiency of the tested SNPs or allelic heterogeneity.
The second challenge is the difficulty in capturing population-specific genetic variation. The 1000 genomes project revealed substantial population differentiation in low-frequency variants, highlighting the inadequacy of SNP genotyping arrays in capturing rare genetic variants across multiple populations. In particular, African populations have shorter LD blocks, more private rare variants and thus shorter shared haplotypes with other populations [128]. Approaches to accurately capture these rare variants will be essential to fully explore the genetic architecture of ADRs in global populations.
Finally, population admixture compounds the difficulty of defining ethnicity even further. It is possible to study large admixed populations resulting from relatively well-defined ancestral populations, such as African–Americans, who are admixed between African and European ancestries. In fact, a study of genes involved in drug absorption, distribution, metabolism and excretion (ADME) in African–Americans revealed greater genetic complexity at functionally important ADME genes than their ancestral populations [129]. Admixed populations also exhibit greater genetic diversity at the individual level [70,130]. Increasingly, more diverse patterns of admixture are emerging, adding to the difficulty of assigning such individuals to a specific ancestral population by either self-reported ancestry or ancestry informative genetic markers.
Conclusion
The application of GWAS to the field of pharmacogenomics has resulted in a rapid advancement in our understanding of the genomic basis of specific ADRs. In particular, many more non-HLA loci for ADRs have been uncovered in recent years. While many of these loci have robust statistical significance regarding their association with an ADR phenotype, most are located in noncoding regions of the genome, and their biological significance remains to be determined. Approaches to understanding the molecular mechanisms underlying these associations, identifying causal variants and distinguishing true effects from false positive results will be essential to fully capitalize on the value of these findings and to enable clinical translation of these results. There has also been an increase in the proportion of GWAS of ADRs conducted in non-European populations. This has led to the recognition of population-specific signals in some cases, highlighting the complex nature of the genetics of ADRs. Studies of pharmacogenomic diversity across populations can facilitate our understanding of such differences and provide guidance on the direction forward. It is clear that ethnicity is an important factor in pharmacogenomics, and information on population diversity of relevant pharmacogenomic loci should contribute, at least in part, to the design of replication and discovery studies in different populations, with the ultimate goal of extending the benefits of pharmacogenomics research to all world populations to improve drug efficacy and reduce the burden of ADRs.
Future perspective
The current trend of extending pharmacogenomic studies to non-European populations appears set to continue. We anticipate additional developments in methodology for trans-ethnic fine mapping in the coming years and envisage the synergy between the growing number of studies in different ancestral populations and improved techniques to elucidate the genetic basis of more pharmacogenomic traits. Next-generation sequencing technologies are also rapidly maturing and have already made an entrance into the field of pharmacogenomics. We anticipate that whole-exome or whole-genome sequencing will become increasingly common, and will complement the GWAS approach to identify more rare variants associated with ADR phenotypes.
| Drug | ADR | Pair† | Sample size (case/control; [replicate case/control]) | Study population | Gene | Lead SNP (rsid, if available)‡ | OR (95% CI) | p value | Ref. |
|---|---|---|---|---|---|---|---|---|---|
| Allopurinol | SJS/TEN | 1 | 14/991; [141/65] | Japanese | HLA-B*5801 | 62.8 (21.2–185.8) | 5.39 × 10-12 | [3] | |
| 57/2248 | European | HCP5 | rs9469003 | 1.73 (1.44–2.08) | 1.60 × 10-9 | [4] | |||
| HLA-B*5801 | haplotype with 6 SNPs | 7.77 (4.66–12.98) | NA | ||||||
| Amoxicillin-clavulanate | Liver injury | 2 | 201/532; [177/219] | White European, US | HLA-DQB1 | rs9274407 (F41Y) | 3.1 (2.3–4.2) | 4.8 × 10-14 | [5] |
| Antiarrhythmic drugs | Torsades de Pointes | 3 | 216/771 | North-western European | C18orf21 | rs2276314 (T44A) | 2 (1.5–2.7) | 4.0 × 10-7 | [6] |
| A haplotype | rs2276314 and rs767531 | NA | 1.6 × 10-9 | ||||||
| Antidepressant | Emergent suicidal ideation | 4 | 32/329; [42/434] | European | rs1630535 | 10.535 (4.2–26.4) | 1.3 × 10-7 | [7] | |
| General side effect§ | 113/1326 | Mixed (Caucasian, African–American and others) | rs13432159 | NA | 9.0 × 10-8 | [8] | |||
| Sexual side effect# | 128/1311 | USP44 | rs7136572 | NA | 2.80 × 10-7 | ||||
| Sexual dysfunction¶ | 36/165 | Japanese | MDGA2 | rs1160351 | 2.92 (1.79–4.76) | 3.04 × 10-7 | [9] | ||
| Aromatase inhibitors | Musculoskeletal ADR | 5 | 293/585 | North American | TCL1A | rs7158782 | 2.13 (1.58–2.87) | 4.73 × 10-7 | [10] |
| TCL1A | rs7159713 | 2.13 (1.58–2.87) | 4.73 × 10-7 | ||||||
| Asparaginase | Allergy | 6 | 125/197; [63/100] | Mixed (White, Black, Hispanics, Asian, Others) | GRIA1 | rs4958351 | 1.75 (1.41–2.17)†† | 3.5 × 10-7 | [11] |
| Aspirin | Respiratory disease | 7 | 117/685; [142/996] | Korean | HLA-DPB1 | rs1042151 (M105V) | 2.4 (1.68–3.42) | 5.11 × 10-7 | [12] |
| Bevacizumab | Hypertension | 8 | 437/483; [35/150] | Mixed | SV2C | rs6453204 | 3.3 | 6.0 × 10-8 | [13] |
| Bisphosphonate | Jaw osteonecrosis | 9 | 30/17; [30/1830‡‡] | White | RBMS3 | rs17024608 | 5.8 (3.0–11.0) | 7.47 × 10-8 | [14] |
| Carbamazepine | Cutaneous ADR | 10 | 65/3987; [145/0] | European | HLA-A*3101 | rs1061235 | 12.41 (1.27–121.03) | 3.5 × 10-8 | [15] |
| 53/882; [61/376] | Japanese | HLA-A*3101 | rs1633021 | 10.8 (5.9–19.6) | 1.18 × 10–13 | [16] | |||
| Chemotherapeutic agents | Neutropenia/leukopenia | 11 | 805/4804 | Japanese | AGR2 | rs10253216 | 1.478 (1.155–1.891) | 1.68 × 10-7 | [17] |
| EBF1 | rs10040979 | 1.452 (1.120–1.883) | 4.60 × 10-7 | ||||||
| TNFRSF1A | rs4149639 | 4.443 (2.571–7.677) | 2.89 × 10-7 | ||||||
| RICH2 | rs11651483 | 1.357 (1.120–1.643) | 3.37 × 10-7 | ||||||
| FGD6 | rs12310399 | 1.852 (1.329–2.580) | 2.46 × 10-7 | ||||||
| Neutropenia/leukopenia§§ | 67/203; [48/0] | Japanese | MCPH1 | rs2916733 and rs1031309 | 2.88 (2.05–4.03) | 2.20 × 10-10 | [18] | ||
| Chemotherapy | Alopecia | 12 | 303/880; [23/0] | Japanese | CACNB4 | rs3820706 | 3.71 (2.24–6.15) | 8.13 × 10-9 | [19] |
| Cisplatin | Hearing loss | 13 | 145/93; [68##] | Mixed | ACYP2 | rs1872328 | 4.50 (2.63–7.69)†† | 3.9 × 10-8 | [20] |
| Citalopram | Vision/hearing side effects | 14 | 1675/87 | Mixed (Caucasian, African–American and others) | EMID2 | rs17135437 | NA | 3.27 × 10-8 | [21] |
| Clozapine | Agranulocytosis | 15 | 161/1196 | US, UK | HLA-DQB1 | 126Q (Q126H) | 0.19 (0.12–0.29) | 4.7 × 10-14 | [22] |
| HLA-B | 158T (A158T) | 3.3 (2.3–4.9) | 6.4 × 10-10 | ||||||
| Dapsone | Hypersensitivity | 16 | 39/833; [68/1290] | Chinese | HLA-B/MICA | rs2844573 | 6.18 | 3.84 × 10-13 | [23] |
| *13:01 | 21.67 (10.41–45.12) | 2.04 × 10-16 | |||||||
| Hydrochlorothiazide | Hyperuricemia | 17 | 36/112; [37/110] | African–American | LUC7L2 | rs6947309 | 6.95 (1.42–34.06) | 7.18 × 10-8 | [24] |
| Lumiracoxib | Liver injury | 18 | 41/217; [98/405] | Mixed (Caucasian, Black, Hispanic, others) | HLA-DRB1 | rs9270986 | 5.3 (3.0–9.2) | 2.8 × 10-10 | [25] |
| HLA-DRB1*1501 HLA-DQB1*0602 HLA-DRB5*0101 HLA-DQA1*0102 | NA | 5 (3.6–7.0) | 6.8 × 10-25 | ||||||
| Nevirapine | Rash | 19 | 72/77; [88/145] | Thai | CCHCR1 | rs1265112 and rs746647 | 4.36 (2.58–7.36) | 1.2 × 10-8 | [26] |
| Paclitaxel | Neuropathy | 20 | 206/649; [66/205] | European, African–American | EPHA5 | rs7349683 | 1.63 (1.34–1.98)†† | 9.6 × 10-7 | [27] |
| FZD3 | rs7001034 | 0.57 (0.48–0.69) | 3.1 × 10-9 | ||||||
| 36/108 | White European | EPHA5 | rs7349683 | 1.68 (1.42–1.99)†† | 1.4×10-9 | [28] | |||
| pegIFN/ribavirin | Haemolytic anemia | 21 | 118/1168 | Mixed (European–American, African–American, Hispanics) | ITPA | rs7270101 | NA | 8.5 × 10-76 | [29] |
| ITPA | rs1127354 (P15T; P32T) | NA | 1.7 × 10-58 | ||||||
| 59/606; [32/226] | Japanese | ITPA | rs1127354 (P15T; P32T) | 0.2 (0.092–0.46) | 3.5 × 10-44 | [30] | |||
| Thrombocytopenia | 1294/310 | Mixed (European–American, African–American, Hispanics) | ITPA | rs1127354 (P15T; P32T) | NA | 1.38 × 10-12 | [31] | ||
| ITPA | rs7270101 | NA | 3.39 × 10-7 | ||||||
| 107/196; [137/254] | Japanese | DDRGK1 | rs11697186 | 4.6 (2.7–7.8) | 8.17 × 10-9 | [32] | |||
| Neutropenia | 22 | 164/302; [50/354] | Japanese | PSMD3 | rs2305482 | 2.18 (1.61–2.96) | 3.05 × 10-7 | [33] | |
| Penicillin | Allergy | 23 | 387/1124; [299/362] | Spain, Italy | ZNF300 | rs4958427 | 1.88 (1.52–2.33) | 1.8 × 10-8 | [34] |
| Phenytoin | Severe cutaneous ADR | 24 | 60/412; [45/3373] | Taiwanese, Japanese, Malay | CYP2C | rs1057910 (I359L) | 12 (6.6–20.0) | 1.1 × 10-17 | [35] |
| Risperidone | Hip circumference | 25 | 265/473 | Mixed (White, Black, Spanish, Hispanic, Latino, others) | MEIS2 | rs1568679 | NA | 1.28 × 10-8 | [36] |
| Second-generation antipsychotic drugs | Weight gain | 25 | 139##; [205##] | Mixed (Caucasian, African–American, others) | MC4R | rs489693 | NA | 5.59 × 10-12 | [37] |
| Thiopurine | Pancreatitis | 26 | 172/2035; [78/472] | 168 sites around the world | HLA-DQA1*02:01-HLA-DRB1*07:01 | rs2647087 | 2.59 (2.07–3.26) | 2 × 10-16 | [38] |
| Thiopurine | Leukopenia | 27 | 33/307; [313/325] | Korean | NUDT15 | rs116855232 (R139C) | 35.63 (22.47–56.51) | 4.88 × 10-94 | [39] |
| Various drugs | SJS/TEN | – | 96/5053 | European | RPA3OS | rs17137412 | 4 (2.5–6.2) | 1.2 × 10-8 | [40] |
| Drug | ADR | Sample size (case/control) | Study population | Gene | Top SNPs (rsid, if available) | OR (95% CI) | p value | Rep | Ref. |
|---|---|---|---|---|---|---|---|---|---|
| Allopurinol | SJS/TEN | 18/493 | Japanese | HLA-B | *5801 | 62.8 (21.2–185.8) | 5.388 × 10-12 | NA | [3]† |
| 25/23 | Portuguese | 44 (3.18–608.19)‡ | NR but sig | Y | [76] | ||||
| 7/25 | Japanese | 65.6 (2.9–1497.0)§ | 9.73 × 10-4 | Y | [77] | ||||
| 38/63 | Han Chinese | 580.07# | <0.0001 | Y | [78] | ||||
| 57/2248 | Europeans | 7.77 (4.66–12.98) | NR but sig | Y | [4]† | ||||
| 7/128 | Europeans | 13.63 (2.77–69.45) | 0.248 | N | [79] | ||||
| 55/678 | Chinese, Thai, Koreans | 96.6 (24.49–381.0)¶ | <0.001 | Y | [80] | ||||
| Aspirin | Respiratory disease | 141/996 | Korean | HLA-DPB1 | rs1042151 | 2.4 (1.68–3.42) | 5.11 × 10-7 | NA | [12]† |
| 59/57 | Polish | *0301 | 5.3 (1.9–14.4) | 0.0001 | Y | [81] | |||
| *0401 | 0.48 (0.28–0.83) | 0.008 | N | ||||||
| 76/73 | Korean | *0301 | 5.2 (1.8–14.7) | 0.004 | Y | [82] | |||
| 19/21 21/18 | Caucasian (British), Caucasian (German) | *0401 | NR | 0.007 0.008 | Y | [83] | |||
| Carbamazepine | Severe cutaneous adverse reactions | 77/420 | Japanese | HLA-A | *3101 | 9.5 (5.6–16.3)# | 1.09 × 10-16 | NA | [16]† |
| 9.5 (4.6–19.5)†† | 2.06 × 10-9 | ||||||||
| 33.9 (3.9–295.6)‡ | 2.35 × 10-4 | ||||||||
| 145/257 | European | 12.41 (1.27–121.03)†† | 3.5 × 10-8 | Y | [15]† | ||||
| 25.93 (4.93–116.18)‡ | 8.0 × 10-5 | ||||||||
| 8.33 (3.59–19.36)‡‡ | 8.0 × 10-7 | ||||||||
| 15/33 | Japanese | 11.2 (2.7–47.1)‡,†† | 0.001 | Y | [84] | ||||
| 91/144 | Taiwanese | 17.5 (4.6–66.5)‡‡ | NR but sig | Y | [85] | ||||
| 18/93 | Southern Han Chinese | 6.14 (0.24–155.12)‡ | 0.33 | N | [86] | ||||
| 194/152 | Chinese | 6.86 (2.4–19.9)†† | 2.7 × 10-3 | Y | [87] | ||||
| NS‡, ‡‡ | N | ||||||||
| 93/329 | Europeans | 57.6 (11.0–341.0)†† | <0.001 | Y | [88] | ||||
| NS‡ | N | ||||||||
| Asians | 23 (4.2–125.0)†† | <0.001 | Y | ||||||
| NS‡ | N | ||||||||
| Clozapine | Agranulocytosis | 161/4300 | European, Ashkenazi Jews | HLA-DQB1 | 126Q | 0.19 (0.12–0.29) | 4.7 × 10-14 | NA | [22]† |
| HLA-B | 158T | 3.3 (2.3–4.9) | 6.4 × 10-10 | ||||||
| 6/25 | Jewish | HLA-B38, DR4, DQw3 | NR | Sig | Y | [89] | |||
| 31/21 | Ashkenazi Jewish | HLA-DRB1 | *0402 | 6.8 | 0.002 | Y | [90] | ||
| HLA-DQB1 | *0302 | 4.9 | 0.008 | ||||||
| HLA-DQA1 | *0301 | 3.1 | 0.03 | ||||||
| HLA-B | *38 | 50 | <.0001 | ||||||
| HLA-DR4 | 23.3 | 0.0004 | |||||||
| Non-Jewish (US) | HLA-DR | *02 | 5.95 | 0.01 | |||||
| HLA-DQB1 | *0502 | 9.68 | 0.02 | ||||||
| HLA-DQA1 | *0102 | 5.37 | 0.01 | ||||||
| 11/50 | Israeli | HLA-B | *38 | < 0.001 | Y | [91] | |||
| 82/132 | Caucasian | HLA-DQB1 | 6672G>C | 16.9 | Sig | Y | [92] | ||
| 5/13 | Israeli | HLA-DQB1 | *0201 | 100 vs 54% | 0.09 | N | [93] | ||
| Paclitaxel | Neuropathy | 206/649 | European | FGD4 | rs10771973 | 1.57 (1.30–1.91) | 2.6 × 10-6 | NA | [27]† |
| 33/84 | African–Americans | 1.93 (1.13–2.38) | 6.7 × 10-3 | Y | |||||
| pegIFN/ribavirin | Anemia | 1286§§ | European, African–American, Hispanics | ITPA | rs1127354 | A allele protective | 1.7 × 10-58 | NA | [29]† |
| 923§§ | Japanese | A allele protective | 3.5 × 10-44 | Y | [30]† | ||||
| 60§§ | Hong Kong (Chinese) | A allele protective | <0.001 | Y | [94] | ||||
| 67§§ | Japanese | A allele protective | 0.09 | N | [95] | ||||
| 292§§ | Japanese | A allele protective | <0.0001 | Y | [96] | ||||
| 354§§ | European | Deficiency protective¶¶ | <0.0001 | Y | [97] | ||||
| 175§§ | Taiwanese | A allele protective | 3.91 × 10-12 | Y | [98] | ||||
| 546§§ | Mixed (majority Caucasian) | Deficiency protective¶¶ | < 0.0001 | Y | [99] | ||||
| 193§§ | Italian | Deficiency protective¶¶ | <0.001 | Y | [100] | ||||
| 379§§ | Italian | A allele protective | 1.26 × 10-11 | Y | [101] | ||||
| 561§§ | Japanese | A allele protective | 5.44 × 10-9 | Y | [102] | ||||
| 167§§ | Italian | A allele protective | 4.09 × 10-8 | Y | [103] | ||||
| 161§§ | Spanish | A allele protective | <0.001 | Y | [104] | ||||
| 304§§ | White and Black | A allele protective | 3.12 × 10-13 | Y | [105] | ||||
| 61§§ | Japanese | A allele protective | 0.001 | Y | [106] | ||||
| 238§§ | Caucasian | A allele protective | 10-6 | Y | [107] | ||||
| Phenytoin | Severe cutaneous adverse reactions | 105/3655 | Taiwanese, Japanese, Malay | CYP2C9 | *3 | 11 (6.2–18.0) | <0.00001 | NA | [35]† |
| 10/39 | Koreans | 167 | 0.007 | Y | [108] |
Executive summary
Progress in understanding the genomic basis of ADRs using GWAS
Genome-wide association studies (GWAS) for adverse drug reactions (ADRs) continue to be characterized by smaller sample sizes than GWAS for common diseases.
Only two-thirds of reported ADR GWAS included independent replication; replication of these loci represents an ongoing challenge.
In contrast to early GWAS of ADRs, most studies since 2010 (74%) have reported associations with loci outside of the MHC region.
Similar to GWAS of common diseases, the majority of ADR-associated genomic variants (82%) are noncoding variants, and a similar proportion of these compared with those associated with common diseases overlap with potentially functional regions of the noncoding genome.
Pharmacogenomic studies in multi-ethnic populations
Several ADR GWAS loci have been successfully replicated in populations of different ancestry compared with the original discovery cohorts.
Population-specific ADR associations have been observed.
Ethnic diversity creates both challenges and opportunities in the search for genetic factors underlying drug response.
Financial & competing interests disclosure
This work was supported by the Biomedical Research Council of A*STAR and the National University of Singapore. The authors have no other relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript apart from those disclosed.
No writing assistance was utilized in the production of this manuscript.
References
- 1 . Genome-wide association studies in pharmacogenomics. Nat. Rev. Genet. 11(4), 241–246 (2010).
- 2 . www.genome.gov/gwastudies.
- 3 A whole-genome association study of major determinants for allopurinol-related Stevens-Johnson syndrome and toxic epidermal necrolysis in Japanese patients. Pharmacogenomics J. 13(1), 60–69 (2013).
- 4 Genome-wide association study of Stevens-Johnson Syndrome and Toxic Epidermal Necrolysis in Europe. Orphanet J. Rare Dis. 6(1), 52 (2011).
- 5 Susceptibility to amoxicillin-clavulanate-induced liver injury is influenced by multiple HLA class I and II alleles. Gastroenterology 141(1), 338–347 (2011).
- 6 Genome wide analysis of drug-induced torsades de pointes: Lack of common variants with large effect sizes. PLoS ONE 8(11), e78511 (2013).
- 7 Genome-wide association study of antidepressant treatment-emergent suicidal ideation. Neuropsychopharmacology 37(3), 797–807 (2012).
- 8 Pharmacogenomic study of side-effects for antidepressant treatment options in STAR*D. Psychol. Med. 42(06), 1151–1162 (2012).
- 9 Genome-wide association study of SSRI/SNRI-induced sexual dysfunction in a Japanese cohort with major depression. Psychiatry Res. 198(3), 424–429 (2012).
- 10 Genome-wide associations and functional genomic studies of musculoskeletal adverse events in women receiving aromatase inhibitors. J. Clin. Oncol. 28(31), 4674–4682 (2010).
- 11 Genetic variations in GRIA1 on chromosome 5q33 related to asparaginase hypersensitivity. Clin. Pharmacol. Ther. 88(2), 191–196 (2010).
- 12 Genome-wide association study of aspirin-exacerbated respiratory disease in a Korean population. Hum. Genet. 132(3), 313–321 (2013).
- 13 Genetic variant predicts bevacizumab-induced hypertension in ECOG-5103 and ECOG-2100. Br. J. Cancer. 111, 1241–1248 (2014).
- 14 . Genomewide pharmacogenetics of bisphosphonate-induced osteonecrosis of the jaw: the role of RBMS3. Oncologist 17(2), 279–287 (2012).
- 15 HLA-A*3101 and carbamazepine-induced hypersensitivity reactions in Europeans. N. Engl. J. Med. 364(12), 1134–1143 (2011).
- 16 Genome-wide association study identifies HLA-A*3101 allele as a genetic risk factor for carbamazepine-induced cutaneous adverse drug reactions in Japanese population. Hum. Mol. Genet. 20(5), 1034–41 (2011).
- 17 Genome-wide association study of chemotherapeutic agent-induced severe neutropenia/leucopenia for patients in Biobank Japan. Cancer Sci. 104(8), 1074–1082 (2013).
- 18 Genome-wide association study of epirubicin-induced leukopenia in Japanese patients. Pharmacogenet. Genomics 21(9), 552–558 (2011).
- 19 A genome-wide association study of chemotherapy-induced alopecia in breast cancer patients. Breast Cancer Res. 15(5), R81 (2013).
- 20 Common variants in ACYP2 influence susceptibility to cisplatin-induced hearing loss. Nat. Genet. 47(3), 263–266. (2015).
- 21 Genome-wide pharmacogenomic study of citalopram-induced side effects in STAR*D. Transl. Psychiatry 2, e129 (2012).
- 22 Clozapine-induced agranulocytosis is associated with rare HLA-DQB1 and HLA-B alleles. Nat. Commun. 5, 4757 (2014).
- 23 HLA-B*13:01 and the dapsone hypersensitivity syndrome. N. Engl. J. Med. 369(17), 1620–1628 (2013).
- 24 Hydrochlorothiazide-induced hyperuricaemia in the pharmacogenomic evaluation of antihypertensive responses study. J. Intern. Med. 276(5), 486–497 (2014).
- 25 A genome-wide study identifies HLA alleles associated with lumiracoxib-related liver injury. Nat. Genet. 42(8), 711–714 (2010).
- 26 Genome-wide association study identifies variations in 6p21.3 associated with nevirapine-induced rash. Clin. Infect. Dis. 53(4), 341–348 (2011).
- 27 A genome-wide association study identifies novel loci for paclitaxel-induced sensory peripheral neuropathy in CALGB 40101. Clin. Cancer Res. 18(18), 5099–5109 (2012).
- 28 Genome-wide association study identifies ephrin type A receptors implicated in paclitaxel induced peripheral sensory neuropathy. J. Med. Genet. 50(9), 599–605 (2013).
- 29 ITPA gene variants protect against anaemia in patients treated for chronic hepatitis C. Nature 464(7287), 405–408 (2010).
- 30 ITPA polymorphism affects ribavirin-induced anemia and outcomes of therapy–a genome-wide study of Japanese HCV virus patients. Gastroenterology 139(4), 1190–1197 (2010).
- 31 Genome-wide association study of interferon-related cytopenia in chronic hepatitis C patients. J. Hepatol. 56(2), 313–319 (2012).
- 32 Genome-wide association study identified ITPA/DDRGK1 variants reflecting thrombocytopenia in pegylated interferon and ribavirin therapy for chronic hepatitis C. Hum. Mol. Genet. 20(17), 3507–3516 (2011).
- 33 Genome-wide association study identifies a PSMD3 variant associated with neutropenia in interferon-based therapy for chronic hepatitis C. Hum. Genet. 134, 279–289 (2014).
- 34 HLA-DRA variants predict penicillin allergy in genome-wide fine-mapping genotyping. J. Allergy Clin. Immunol. 135(1), 253–259 (2015).
- 35 Genetic variants associated with phenytoin-related severe cutaneous adverse reactions. JAMA 312(5), 525–534 (2014).
- 36 Genomewide pharmacogenomic study of metabolic side effects to antipsychotic drugs. Mol. Psychiatry 16(3), 321–332 (2011).
- 37 Association between common variants near the melanocortin 4 receptor gene and severe antipsychotic drug-induced weight gain. Arch. Gen. Psychiatry 69(9), 904–912 (2012).
- 38 HLA-DQA1-HLA-DRB1 variants confer susceptibility to pancreatitis induced by thiopurine immunosuppressants. Nat. Genet. 46(10), 1131–1134 (2014).
- 39 A common missense variant in NUDT15 confers susceptibility to thiopurine-induced leukopenia. Nat. Genet. 46(9), 1017–1020 (2014).
- 40 Genome-wide association study of serious blistering skin rash caused by drugs. Pharmacogenomics J. 12(2), 96–104 (2012).
- 41 Genome-wide approaches to identify pharmacogenetic contributions to adverse drug reactions. Pharmacogenomics J. 9(1), 23–33 (2009).
- 42 . Delayed drug hypersensitivity reactions. Ann. Intern. Med. 139(8), 683–693 (2003).
- 43 . Using genome-wide association studies to identify genes important in serious adverse drug reactions. Annu. Rev. Pharmacol. Toxicol. 52(1), 21–35 (2012).
- 44 Genetic variations in HLA-B region and hypersensitivity reactions to abacavir. Lancet 359(9312), 1121–1122 (2002).
- 45 Association between presence of HLA-B*5701, HLA-DR7, and HLA-DQ3 and hypersensitivity to HIV-1 reverse-transcriptase inhibitor abacavir. Lancet 359(9308), 727–732 (2002).
- 46 HLA-B*5701 screening for hypersensitivity to abacavir. N. Engl. J. Med. 358(6), 568–579 (2008).
- 47 Refining abacavir hypersensitivity diagnoses using a structured clinical assessment and genetic testing in the Swiss HIV Cohort Study. Antivir. Ther. 13(8), 1019–1028 (2008).
- 48 First large, multicenter, open-label study utilizing HLA-B*5701 screening for abacavir hypersensitivity in North America. AIDS 22(13), 1673–1675 (2008).
- 49 A marker for Stevens-Johnson syndrome. Nature 428, 486 (2004).
- 50 Carbamazepine and phenytoin induced Stevens-Johnson syndrome is associated with HLA-B*1502 allele in Thai population. Epilepsia 49, 2087–2091 (2008).
- 51 Association between HLA-B*1502 and carbamazepine-induced severe cutaneous adverse drug reactions in a Thai population. Epilepsia 51, 926–930 (2010).
- 52 . Cutaneous adverse drug reactions seen in a tertiary hospital in Johor, Malaysia. Int. J. Dermatol. 49, 834–841 (2010).
- 53 Systematic localization of common disease-associated variation in regulatory DNA. Science 337(6099), 1190–1195 (2012).
- 54 SEARCH Collaborative Group, SLCO1B1 variants and statin-induced myopathy–a genomewide study. N. Engl. J. Med. 359(8), 789–799 (2008).
- 55 The SLCO1B1*5 genetic variant is associated with statin-induced side effects. J. Am. Coll. Cardiol. 54, 1609–1616 (2009).
- 56 Differential effect of the rs4149056 variant in SLCO1B1 on myopathy associated with simvastatin and atorvastatin. Pharmacogenomics J. 12(3), 233–237 (2012).
- 57 . SLCO1B1 polymorphism markedly affects the pharmacokinetics of simvastatin acid. Pharmacogenet. Genomics 16(12), 873–879 (2006).
- 58 Variant GADL1 and response to lithium therapy in bipolar I disorder. N. Engl. J. Med. 370(2), 119–128 (2014).
- 59 . Linking disease associations with regulatory information in the human genome. Genome Res. 22(9), 1748–1759 (2012).
- 60 Annotation of functional variation in personal genomes using RegulomeDB. Genome Res. 22(9), 1790–1797 (2012).
- 61 . Genome-wide approaches in pharmacogenomics: heritability estimation and pharmacoethnicity as primary challenges. Pharmacogenomics 13(10), 1101–1104 (2012).
- 62 . CYP2D6 allele frequency in European Caucasians, Asians, Africans and their descendants. Pharmacogenomics 3(2), 229–243 (2002).
- 63 . An interethnic comparison of polymorphisms of the genes encoding drug-metabolizing enzymes and drug transporters: experience in Singapore. Drug Metab. Rev. 37(2), 327–378 (2005).
- 64 . Genetic variation in drug transporters in ethnic populations. Clin. Pharmacol. Ther. 84(3), 412–416 (2008).
- 65 . CYP2D6 worldwide genetic variation shows high frequency of altered activity variants and no continental structure. Pharmacogenet. Genomics 17(2), 93–101 (2007).
- 66 Genetic variation in eleven phase I drug metabolism genes in an ethnically diverse population. Pharmacogenomics 5(7), 895–931 (2004).
- 67 Genetic variation in metabolizing enzyme and transporter genes: comprehensive assessment in 3 major east asian subpopulations with comparison to Caucasians and Africans. J. Clin. Pharmacol. 50(8), 929–940 (2010).
- 68 Pharmacogenomics, ancestry and clinical decision making for global populations. Pharmacogenomics J. 14(3), 217–222 (2014).
- 69 Pharmacogenomic diversity in Singaporean populations and Europeans. Pharmacogenomics J. 14(6), 555–563 (2014).
- 70 Higher frequency of genetic variants conferring increased risk for ADRs for commonly used drugs treating cancer, AIDS and tuberculosis in persons of African descent. Pharmacogenomics J. 14(2), 160–170 (2014).
- 71 Interethnic variability of warfarin maintenance requirement is explained by VKORC1 genotype in an Asian population. Clin. Pharmacol. Ther. 79(3), 197–205 (2006).
- 72 Warfarin pharmacogenetics: a single VKORC1 polymorphism is predictive of dose across 3 racial groups. Blood 115(18), 3827–3834 (2010).
- 73 HPS2-THRIVE randomized placebo-controlled trial in 25 673 high-risk patients of ER niacin/laropiprant: trial design, pre-specified muscle and liver outcomes, and reasons for stopping study treatment. Eur. Heart J. 34(17), 1279–1291 (2013).
- 74 1000 Genomes: A Deep Catalog of Human Genetic Variation. www.1000genomes.org.
- 75 . Impact of Hispanic or Asian ethnicity on the treatment outcomes of chronic hepatitis C: results from the WIN-R trial. J. Clin. Gastroenterol. 45(8), 720–726 (2011).
- 76 HLA-B*58:01 is a risk factor for allopurinol-induced DRESS and Stevens-Johnson syndrome/toxic epidermal necrolysis in a Portuguese population. Br. J. Dermatol. 169(3), 660–665 (2013).
- 77 HLA-B*58:01 strongly associates with allopurinol-induced adverse drug reactions in a Japanese sample population. J. Dermatol. Sci. 71(2), 150–152 (2013).
- 78 HLA-B*58:01 allele is associated with augmented risk for both mild and severe cutaneous adverse reactions induced by allopurinol in Han Chinese. Pharmacogenomics 13(10), 1193–1201 (2012).
- 79 A study of HLA class I and class II 4-digit allele level in Stevens-Johnson syndrome and toxic epidermal necrolysis. Int. J. Immunogenet. 38(4), 303–309 (2011).
- 80 . Association of HLA-B*5801 allele and allopurinol-induced Stevens Johnson syndrome and toxic epidermal necrolysis: a systematic review and meta-analysis. BMC Med. Genet. 12, 118 (2011).
- 81 Aspirin-induced asthma and HLA-DRB1 and HLA-DPB1 genotypes. Clin. Exp. Allergy 27(5), 574–577 (1997).
- 82 HLA association in aspirin-intolerant asthma: DPB1*0301 as a strong marker in a Korean population. J. Allergy Clin. Immunol. 113(3), 562–564 (2004).
- 83 . An analysis with sequence-specific oligonucleotide probes of the association between aspirin-induced asthma and antigens of the HLA system. J. Allergy Clin. Immunol. 92(1 Pt 1), 114–123 (1993).
- 84 . HLA-A31 strongly associates with carbamazepine-induced adverse drug reactions but not with carbamazepine-induced lymphocyte proliferation in a Japanese population. J. Dermatol. 39(7), 594–601 (2012).
- 85 Genetic susceptibility to carbamazepine-induced cutaneous adverse drug reactions. Pharmacogenet. Genomics 16(4), 297–306 (2006).
- 86 Association between HLA and Stevens-Johnson syndrome induced by carbamazepine in Southern Han Chinese: genetic markers besides B*1502? Basic Clin. Pharmacol. Toxicol. 111(1), 58–64 (2012).
- 87 Genotype-phenotype association between HLA and carbamazepine-induced hypersensitivity reactions: strength and clinical correlations. J. Dermatol. Sci. 73(2), 101–109 (2014).
- 88 HLA-A*31:01 and different types of carbamazepine-induced severe cutaneous adverse reactions: an international study and meta-analysis. Pharmacogenomics J. 14(3), 281–288 (2014).
- 89 . HLA-B38, DR4, DQw3 and clozapine-induced agranulocytosis in Jewish patients with schizophrenia. Arch. Gen. Psychiatry 47(10), 945–948 (1990).
- 90 . HLA associations in clozapine-induced agranulocytosis. Blood 86(3), 1177–1183 (1995).
- 91 HLA-B38 and clozapine-induced agranulocytosis in Israeli Jewish schizophrenic patients. Eur. J. Immunogenet. 25(1), 11–13 (1998).
- 92 Candidate gene analysis identifies a polymorphism in HLA-DQB1 associated with clozapine-induced agranulocytosis. J. Clin. Psychiatry 72(4), 458–463 (2011).
- 93 An association between clozapine-induced agranulocytosis in schizophrenics and HLA-DQB1*0201. Int. J. Neuropsychopharmacol. 1(1), 41–44 (1998).
- 94 Role of IL28B and inosine triphosphatase polymorphisms in the treatment of chronic hepatitis C virus genotype 6 infection. J. Viral Hepat. 20(7), 470–477 (2013).
- 95 Effect of laparoscopic splenectomy in patients with Hepatitis C and cirrhosis carrying IL28B minor genotype. BMC Gastroenterol. 12, 158 (2012).
- 96 Clinical milestones for the prediction of severe anemia by chronic hepatitis C patients receiving telaprevir-based triple therapy. J. Hepatol. 59(4), 667–674 (2013).
- 97 Variants of the inosine triphosphate pyrophosphatase gene are associated with reduced relapse risk following treatment for HCV genotype 2/3. Hepatology 59(6), 2131–2139 (2014).
- 98 Association between IPTA gene polymorphisms and hematological abnormalities in hepatitis C virus-infected patients receiving combination therapy. Gut Liver 8(6), 605–611 (2014).
- 99 ITPA genotype protects against anemia during peginterferon and ribavirin therapy but does not influence virological response. Hepatology 59(6), 2152–2160 (2014).
- 100 Inosine triphosphatase deficiency helps predict anaemia, anaemia management and response in chronic hepatitis C therapy. J. Viral Hepat. 20(12), 858–866 (2013).
- 101 Association of ITPA polymorphisms rs6051702/rs1127354 instead of rs7270101/rs1127354 as predictor of ribavirin-associated anemia in chronic hepatitis C treated patients. Antiviral Res. 100(1), 114–119 (2013).
- 102 Several factors including ITPA polymorphism influence ribavirin-induced anemia in chronic hepatitis C. World J. Gastroenterol. 18(41), 5879–5888 (2012).
- 103 Inosine triphosphatase polymorphisms and ribavirin pharmacokinetics as determinants of ribavirin-associate anemia in patients receiving standard anti-HCV treatment. Ther. Drug Monit. 34(2), 165–170 (2012).
- 104 Variants in the ITPA gene protect against ribavirin-induced hemolytic anemia in HIV/HCV-coinfected patients with all HCV genotypes. J. Infect. Dis. 205(3), 376–383 (2012).
- 105 Variants in the ITPA gene protect against ribavirin-induced hemolytic anemia and decrease the need for ribavirin dose reduction. Gastroenterology 139(4), 1181–1189 (2010).
- 106 Influence of ITPA polymorphisms on decreases of hemoglobin during treatment with pegylated interferon, ribavirin, and telaprevir. Hepatology 53(2), 415–421 (2011).
- 107 Inosine triphosphatase genetic variants are protective against anemia during antiviral therapy for HCV2/3 but do not decrease dose reductions of RBV or increase SVR. Hepatology 53(2), 389–395 (2011).
- 108 . Genetic polymorphism of cytochrome P450 2C9 in diphenylhydantoin-induced cutaneous adverse drug reactions. Eur. J. Clin. Pharmacol. 60(3), 155–159 (2004).
- 109 . Variant GADL1 and response to lithium in bipolar I disorder. N. Engl. J. Med. 370(19), 1856–1857 (2014).
- 110 Consortium on Lithium Genetics, Variant GADL1 and response to lithium in bipolar I disorder. N. Engl. J. Med. 370(19), 1857–1859 (2014).
- 111 . GADL1 gene polymorphisms and lithium response in bipolar I disorder: lack of association from an Indian population. Psychiatr. Genet. 25(1), 39–40 (2015).
- 112 Clinical pharmacogenetics implementation consortium guidelines for CYP2C9 and HLA-B genotypes and phenytoin dosing. Clin. Pharmacol. Ther. 96(5), 542–548 (2014).
- 113 Clinical Pharmacogenetics Implementation Consortium guidelines for HLA-B genotype and carbamazepine dosing. Clin. Pharmacol. Ther. 94(3), 324–328 (2013).
- 114 Clinical Pharmacogenetics Implementation Consortium guidelines for thiopurine methyltransferase genotype and thiopurine dosing. Clin. Pharmacol. Ther. 89(3), 387–391 (2011).
- 115 Adverse reactions to azathioprine cannot be predicted by thiopurine S-methyltransferase genotype in Japanese patients with inflammatory bowel disease. J. Gastroenterol. Hepatol. 24(7), 1258–1264 (2009).
- 116 Genetic variants in TPMT and COMT are associated with hearing loss in children receiving cisplatin chemotherapy. Nat. Genet. 41(12), 1345–1349 (2009).
- 117 Replication of TPMT and ABCC3 genetic variants highly associated with cisplatin-induced hearing loss in children. Clin. Pharmacol. Ther. 94(2), 243–251 (2013).
- 118 A pharmacogenetic versus a clinical algorithm for warfarin dosing. N. Engl. J. Med. 369(24), 2283–2293 (2013).
- 119 Genetic variants associated with warfarin dose in African–American individuals: a genome-wide association study. Lancet 382(9894), 790–796 (2013).
- 120 A genome-wide scan for common genetic variants with a large influence on warfarin maintenance dose. Blood 112(4), 1022–1027 (2008).
- 121 A genome-wide association study confirms VKORC1, CYP2C9, and CYP4F2 as principal genetic determinants of warfarin dose. PLoS Genet. 5(3), e1000433 (2009).
- 122 Genome-wide association study identifies genetic determinants of warfarin responsiveness for Japanese. Hum. Mol. Genet. 19(23), 4735–4744 (2010).
- 123 The missing association: sequencing-based discovery of novel SNPs in VKORC1 and CYP2C9 that affect warfarin dose in African Americans. Clin. Pharmacol. Ther. 89(3), 408–415 (2011).
- 124 Genetic variant in folate homeostasis is associated with lower warfarin dose in African Americans. Blood 124(14), 2298–2305 (2014).
- 125 . Identifying candidate causal variants via trans-population fine-mapping. Genet. Epidemiol. 34(7), 653–664 (2010).
- 126 Trans-ethnic fine-mapping of lipid loci identifies population-specific signals and allelic heterogeneity that increases the trait variance explained. PLoS Genet. 9(3), e1003379 (2013).
- 127 A systematic mapping approach of 16q12.2/FTO and BMI in more than 20,000 African–Americans narrows in on the underlying functional variation: results from the Population Architecture using Genomics and Epidemiology (PAGE) study. PLoS Genet. 9(1), e1003171 (2013).
- 128 1000 Genomes Project Consortium, An integrated map of genetic variation from 1,092 human genomes. Nature 491(7422), 56–65 (2012).
- 129 . Increased genetic diversity of ADME genes in African Americans compared with their putative ancestral source populations and implications for pharmacogenomics. BMC Genet. 15, 52 (2014).
- 130 . Pharmacogenetics: implications of race and ethnicity on defining genetic profiles for personalized medicine. J. Allergy Clin. Immunol. 133(1), 16–26 (2014).
