Adverse Drug Reactions and Pharmacogenomics: Recent Advances

Ana Alfirevic; Munir Pirmohamed


Personalized Medicine. 2008;5(1):11-23. 

In This Article

Drug Hypersensitivity

An immune etiology has been suggested for a large number of ADRs. Many attempts have been made to look for associations with specific HLAs. The field has been hampered by the demonstration of weak associations (some of which are listed in Table 1 ), which, importantly, could not be replicated. More recent studies, however, have demonstrated very strong associations. This change in fortune can perhaps be traced to improvements in several areas.

First, in the past, studies utilized HLA serotypes or low resolution genotyping, which failed to discern the highly complex allelic structure of the HLA genes. The development of relatively high throughput and affordable high-resolution HLA typing has made it possible to genotype HLA alleles to four digits. This has led to striking associations, for example, with abacavir hypersensitivity and HLA-B*5701, first described in Australasia[6] and then replicated in patient groups from two other continents.[7,8] Similarly, a very strong association has been observed between carbamazepine-induced Stevens–Johnson syndrome (SJS) and HLA-B*1502 in Han Chinese.[9,10]

Second, these associations seem to be specific to ethnic groups. Thus, the association with carbamazepine-induced SJS in Han Chinese is not seen in Caucasians.[11,12] For abacavir, however, data are contradictory. The association observed in Caucasians was initially not seen in an African–American population in one study.[13] However, a more recent study has shown that while abacavir hypersensitivity is less common in black people, if they are positive for HLA-B*5701, they should be considered to be at risk of hypersensitivity.[14] An important issue here is how well the phenotype is defined. Clinical diagnosis is often imperfect as the skin rash may not be recognized (because of skin colour), or may be falsely attributed to the drug when it is in fact due to another concomitantly administered drug or a viral infection (both common in HIV infection). Thus, it is crucial to improve the phenotype through laboratory testing or more objective assessment. For abacavir, patch testing has been used in the definition of hypersensitivity, and when this is positive, the diagnostic test criteria improve markedly.[14]

Another important issue to consider in relation to ethnicity is that most studies define race on the basis of self reporting. Whether genomic control procedures that correct for unforeseen population stratification would further refine these associations is unclear, but would require the study of large numbers of patients and controls, which seems unlikely to happen in the near future given the rarity of these reactions. The critical issue here is whether the identified differences between different ethnic groups reflect different predisposing genes, and whether they reflect differences in the linkage disequilibrium structure is unclear. Identification of the causal variant(s) is going to be crucial to answer this.

Third, due to a low frequency of severe toxicity for the majority of drugs, it takes a long time to collect the large number of patients needed to achieve adequate power for genetic association studies. Therefore, many studies in the past have been designed to investigate common organ toxicity, regardless of drug etiology. One example is drug-induced hepatotoxicity ( Table 2 ). Given that immune-mediated mechanisms are important in the pathogenesis of drug-induced hepatitis, it is biologically plausible to expect an association with HLA alleles. However, several studies have shown that when HLA alleles are considered collectively, they do not contribute significantly to drug-induced hepatitis.[15] When you can identify adequate numbers of patients with hepatotoxicity caused by individual drugs, it is possible to discern significant associations with individual HLA alleles, for example, as shown with co-amoxiclav[16] and ticlopidine.[17] Interestingly, genetic association studies can produce important associations, which would not have been suspected from the clinical picture. This has recently been demonstrated with the thrombin inhibitor, ximelagatran, which had to be withdrawn because of liver toxicity. The clinical picture with ximelagatran does not suggest involvement of the immune system, yet a whole genome association study has shown a significant association with HLA DRB1*0701.[18]

Finally, and perhaps most importantly, we should not underestimate the enormous value in genetic association studies afforded by an accurate phenotype. Thus, we have shown that severe hypersensitivity reactions to carbamazepine, but not mild maculopapular eruptions, have a predisposing locus on the MHC ancestral haplotype 8.1.[19,20] Similarly, as mentioned earlier, the most serious reactions to carbamazepine, SJS and toxic epidermal necrolysis, are strongly associated with HLA-B*1502 in Han Chinese.[9] However, hypersensitivity syndrome in the same population shows an association with the motilin gene.[10] Clearly, it may not always be simple to phenotype patients as clinical manifestations often overlap, but using simple investigations, as has been done with patch tests and abacavir hypersensitivity[14,21] or intermediate phenotypes,[22,23] may be helpful. With respect to the latter, it is going to be important to define intermediate quantifiable outcomes in the future in order to prevent misclassification of patients. Although we have concentrated on drug hypersensitivity in this section, the arguments for an accurate phenotype clearly apply to any form of drug response, be it efficacy or toxicity.

A parallel can be drawn here between immune-mediated ADRs and immune diseases with regard to predisposing HLA haplotypes. Significant differences in the frequencies of the HLA alleles determining susceptibility to immune diseases such as coeliac disease, Type I diabetes mellitus and multiple sclerosis, have been observed in various racial groups; some examples are presented in Table 3 . HLA-allele frequencies vary in both ethnically diverse healthy and patient populations, and in addition, certain susceptibility alleles may be absent in some ethnic groups, but not in others.[24,25] Furthermore, non-HLA regions seem to play a stronger role in disease susceptibility in some ethnic groups.[26–28] Moreover, HLA alleles and haplotypes also seem to be associated with the age of disease onset and more rapid disease progression in patients with Type I diabetes and coeliac disease.[29–31]