Adverse Drug Reactions and Pharmacogenomics: Recent Advances

Ana Alfirevic; Munir Pirmohamed


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

In This Article

Clinical Application of Pharmacogenetics

The clinical uptake of pharmacogenetic testing has been rather poor. Part of this may be related to the poor evidence that has been presented to date regarding the clinical utility of testing. This seems to be particularly true for CYP2D6, perhaps the most widely studied polymorphic enzyme, as pointed out by a recent systematic review of the use of CYP2D6 testing in patients prescribed selective serotonin re-uptake inhibitors for depression.[49] Even when there is relatively good evidence for the risk of severe toxicity and an enzyme deficiency, the uptake of testing is patchy. For instance, with thiopurine methyl transferase testing to prevent azathioprine toxicity, a survey in Europe showed that uptake was generally low, while a more recent study in the UK showed that testing varied enormously between different specialties.[50–52] Part of the reason for this may be the unresolved debate as to whether phenotyping or genotyping is superior and the relative lack of availability of phenotyping procedures in accredited laboratories.[53–55]

By contrast, pharmacogenetic testing for abacavir hypersensitivity has really taken off in some countries, including Australia and the UK ( Table 3 ).[6–8] Easy access to commercial laboratories providing relatively cost–effective genotyping and the availability of phenotyping methodologies may have helped the translation of this test into clinical practice.[56,57] However, since the test has such a high negative-predictive value, the association has been replicated in different populations, and the fact that it is cost–effective may have had more of an impact. Also important has been the relative willingness of HIV physicians to rapidly adapt their clinical practice to new evidence. The success of this testing has now been shown in three countries (Australia, UK and France), all of which have reported a drop in the frequency of abacavir hypersensitivity following implementation of preprescription genotyping.[58,59]

A test which has been available for many years is for glucose-6-phosphate dehydrogenase (G6PD), an essential red cell enzyme that protects erythrocytes from hemolysis caused by oxidative stress. G6PD is, in fact, the most common enzyme deficiency in the world and 57 mutations that account for more than 400 named G6PD variants have been described in the encoding gene located on chromosome Xq28. The list of drugs that should be avoided in G6PD deficiency has grown longer since the discovery of G6PD deficiency in the 1950s. However, a more recent analysis has shown that the concurrent infection may have been an important confounding factor for many drugs.[60] Commonly used drugs likely to induce hemolytic anemia in G6PD deficiency include primaquine, sulphamethoxazole, dapsone, nalidixic acid and nitrofurantoin.[61] G6PD is X-linked and therefore females who are heterozygotes for the deficiency can have two populations of erythrocytes (G6PD deficient and non-deficient). Screening for G6PD deficiency can be performed using the fluorescent spot test, which measures conversion of NADP to its reduced form (NADPH) in erythrocytes.[62] However, in heterozygous females, accurate detection is not possible because of their two erythrocyte populations and, therefore, PCR-based genetic testing is recommended. Most people with a polymorphic G6PD gene are asymptomatic. In newborns, however, jaundice, kernicterus and severe hemolytic anemia are potentially life-threatening conditions. Newborn screening has been proposed in populations with a high prevalence of G6PD deficiency. However, since a clear understanding of the relationship between kernicterus, severe hemolytic crises and G6PD mutations is still lacking, the American College of Medical Genetics did not recommend the inclusion of G6PD screening into the newborn screening panel in the USA.[63]

A test that has recently been approved by the US FDA is for UDP-glucuronyltransferase 1A1 polymorphisms and their association with irinotecan-induced diarrhea and leukopenia.[202] In vivo, irinotecan is hydrolyzed by carboxylesterase to its active metabolite 7-ethyl-10-hydroxycamptothecin (SN-38), which is 1000-fold more cytotoxic than the parent drug.[64] The major pathway for detoxification is glucuronidation catalyzed by the microsomal enzymes uridine disphosphate-glucuronosyltransferases (UGT). A reduction in glucuronidation rate has been associated with irinotecan toxicity and the interindividual variability in the glucuronidation rate is at least partly genetically determined. The most extensively studied UGT isoform is UGT1A1, which has a primary role in bilirubin detoxification. The presence of a dinucleotide (TA) insertion in the TATA box of the UGT1A1 promoter results in a 70% reduction in enzyme expression and is associated with Gilbert's syndrome.[65] The same genetic variant (UGT1A1*28) is associated with both neutropenia and diarrhea in patients treated with irinotecan.[66–68] Based on a prospective study, grade 4 neutropenia occurred in 50% of homozygous UGT1A1*28 patients on irinotecan and 12.5% of heterozygotes, while no wild-type homozygotes developed neutropenia.[69] Although the data are promising in populations of white origin, in Asian populations, different alleles, and in particular UGT1A1*6, may be more important for severe toxicity.[70,71] The FDA recommendations have not yet been extended to Europe and Japan; part of the reason for this may be related to differing predisposing variants in different populations. It is also important to note that there are other variants within the whole metabolic pathway for irinotecan, and their role in the overall contribution to irinotecan toxicity needs to be more clearly defined.

Advances in Genotyping Technology

While there has been slow progress in demonstrating the clinical validity and utility of genotyping tests, there has been marked progress ingenotyping technologies. This, together with increased competition, has driven down the costs of genotyping, now down to as low as a few pence per genotype. High-throughput genotyping technologies have been firmly established in many research laboratories around the world and whole genome scanning is now becoming available, the latest platforms able to genotype around 1 million SNPs. Recently, whole genome association studies have identified novel genetic associations in a number of diseases. In addition, sequencing technologies are also becoming cheaper and more highthroughput in nature. Indeed, the possibility exists that in the near future it will be possible to sequence the whole human genome in less than one day for about US$1000. However, one has to insert a note of caution here; while it may be possible to do this, our ability to interpret such huge amounts of data, and more importantly to link it with health-related data, is likely to be limited unless we make significant strides in bioinformatics and link it to health informatics.

As our evidence base for genetic associations improves, there will also be a need to develop near-patient testing technologies. These should be cheap and reproducible and importantly have a short turn-around time, so that drug prescribing decisions can be made without introducing unnecessary delays. The first example for the application of such technology may be for dose determination in patients being prescribed warfarin – indeed, there is increasing interest in developing point-of-care testing for the VKORC1 and CYP2C9 genotypes, which are important determinants of responsiveness to warfarin.[72,203]

Future of Pharmacogenetic Research in Adverse Drug Reactions

Although the advancement of pharmacogenetics into clinical practice has been slower than predicted, pharmacogenetics has a definite place in new drug development and safety prediction for some, but not all, drugs. Pretreatment genetic testing has been proven to be useful in preventing adverse reactions in cancer[73,74] and HIV therapy.[8,75,76] It is possible that genetic testing will soon be introduced in a larger number of centers to predict response to tamoxifen therapy in breast cancer patients.[77] Even if genetic testing is far from being realized for most drug response phenotypes because of low specificity and sensitivity, a low incidence of an ADR or the high cost of genotyping, pharmacogenetic research can sometimes lead to a change in clinical patient management and hospital policies. An example is the recent report of morphine poisoning in a breastfed neonate of a mother who was prescribed codeine for pain following an episiotomy.[78] Codeine is metabolized to morphine by CYP2D6 and neonates are known to have impaired capacity to eliminate morphine. The mother was genotyped for CYP2D6 and found to be an ultra-rapid metabolizer. The authors postulated that the concentration of morphine was higher than would be expected, and taken together with the fact that the baby had a polymorphism in UGT2B7[79] (an allele which increases conversion to the active morphine-6-glucuronide), the baby was diagnosed as having morphine poisoning.[78] The baby's death has triggered a series of new clinical strategies and recommendations for mothers who breastfeed while on codeine for postpartum pain relief.[80] It has also led to a change in the information provided in the British National Formulary.[204]

In order to perform high-quality research in ADR pharmacogenetics, we need large cohorts of patients with well-defined clinical phenotypes. Owing to the low frequency of most severe ADRs (if they were less rare they would have been discovered in the early stages of the clinical efficacy and toxicity studies and drugs would not have been licensed), it is necessary to conduct multicenter and multinational studies to be able to recruit a sufficient number of patients to achieve adequate statistical power ( Table 4 ). We strongly endorsed this approach in 2001[78] and this has also recently been espoused in a commentary in Nature.[81] Indeed, various global initiatives and networks have recently been formed, which encourage clinicians and basic scientists interested in pharmacogenomics to collaborate. One of the major research networks with multiple activities is the Pharmacogenetics Research Network organised by the NIH.[205] Central to the network activities is the PharmGKB database,[206] which aims to link pharmacogenetic knowlege with phenotypes and different drugs and diseases. In the UK, the Department of Health-based Pharmacogenetic Research Programme[207] provides a basis for future networking. In the area of drug-induced liver injury, there are multicenter collections ongoing in both the USA[208] and UK,[209] while EUDRAGENE and EURO-SCAR are collecting DNA samples from patients with various idiosyncratic and serious cutaneous ADRs, respectively. Various biobanks, including the UK Biobank,[210] are currently being established all over the world. Although their primary focus is on the genetics of complex diseases, they may also be of use in pharmacogenomics; however, this will be very much dependent upon the availability of good prescribing data linked to the samples.