It has long been the goal of clinical pharmacologists to be able to provide the right patient, with the right drug, and at the right dose. To date this aspiration has only, to a modest extent, been achieved. Physicians select antimicrobial agents on the basis of the sensitivity of particular micro-organisms. Warfarin therapy can be reasonably safely accomplished by monitoring patients’ anticoagulant responses. For drugs undergoing renal excretion such as digoxin, simple tests of renal function can permit the selection of dosages that avoid toxicity. Similarly, measuring plasma concentrations of drugs during treatment can, in a few instances, ensure that patients are not “overdosed”. But the remarkable advances in genomics, achieved over the past 15 years, presage the emergence of truly “personalised medicine”, and offer the promise of being able to prevent adverse drug reactions.
Many drugs are metabolised to inactive metabolites that then undergo renal excretion. There are more than 30 families of drug-metabolising enzymes, in humans, and essentially all have genetic variants that produce functional changes in the encoded proteins [1, 2].
Mutations in the gene encoding the enzyme thiopurine S-methyltransferase (TPMP) reduce its activity and render patients susceptible to toxicity as a result of excessive plasma and tissue concentrations (2). In this instance, physicians are advised to assess patients’ TPMP activity before embarking on treatment with thioguanine, mercaptopurine and azathioprine. Similarly, mutations in the gene encoding UDP glucuronosyltransferase 1A1 (UGT1A1) impede the rate of metabolism of some drugs undergoing glucuronidation. Consequently physicians are advised to assess patients’ UGTA1A1 activity before embarking on treatment with irinotecan and nilotinib.
There has been much interest in the application of pharmacogenomics in the control of warfarin therapy. Warfarin is metabolised, predominantly, by a cytochrome P-450 enzyme CYP2C9. Two common allozymes (CYP2C9*2 and CYP2C9*3) have only a fraction of the levels of enzyme activity compared to the wild type (CYP2C9*1). Patients with one or two of these variant alleles require lower doses of warfarin, and are at higher risk of bleeding, than patients homozygous for CYP2C9*1. However, the anticoagulant effects of warfarin are also controlled by the activity of vitamin K epoxide reductase complex subunit 1 (VKORC1) and taken together with CYP2C9 activity account for 30 to 40% of the variation in warfarin dose requirements.
Although some advocate the use of pharmacogenomic techniques in controlling warfarin therapy there are (at least) three problems. First, even when the activities of CYP2C9 and VKORC1 are taken into account, therapeutic dosages of warfarin still vary fourfold between individuals. Pharmaogenomics do not obviate the need for regular anticoagulant monitoring. Second, other enzymes are involved in controlling the activity, and hence safe therapeutic doses, of warfarin particularly CYP4F2. Thirdly, environmental effects also influence warfarin activity. These include, most notably, the intake of vitamin K and alcohol.
The problems surrounding the use of pharmacogenomic techniques, in ensuring the safe and effective use of warfarin, are compounded for many other drugs that are primarily eliminated by metabolism. In particular, like warfarin, numerous drugs are metabolised by two or more pathways so that attempting to control for them all is currently impratical.
Idiosyncratic adverse drug reactions
Many adverse drug reactions are idiosyncratic effects with no obvious relationship to the product’s pharmacological properties . Examples include venous thromboembolism (VTE) with oral contraceptives, anaphylactic reactions and skin rashes as well as most forms of dug-induced liver damage. Recent work has started to throw light on the underlying mechanisms and indicate potential approaches for prevention.
Around 5% of Caucasians have a mutation (the Leiden mutation) in the gene responsible for the production of factor V. Women with factor V Leiden are at increased risk of VTE during use of oral contraceptives (28.5/10,000 woman years) compared with women taking oral contraceptives who do not have the Leiden mutation (3.8/10,000 woman years). There have therefore been suggestions that women should be advised to undergo screening for the factor V Leiden mutation prior to starting oral contraceptives.
There are two problems with this approach. First, although the incidence of VTE is ten-fold higher, in factor V Leiden positive women using oral contraceptives compared to those who are factor V Leiden negative, the risk of VTE – to an individual – is still small (less than 1%). Second, although the development of VTE during oral contraceptive use of is unpleasant, the risk of death from pulmonary embolism is also under 1%. As a consequence, the cost per life year saved by such screening has been estimated to be around US$ 300 million .
Striking associations between different HLA alleles, and drug induced hypersensitivity reactions , have recently been described (Table 1). A common characteristic of these studies has been the substantial effect sizes for these associations. For example, in patients receiving carbamazepine , the odds ratio between expression of HLA-A*3102, and the development of Stevens-Johnson syndrome, is 25.9 (95%CI 4.9 to 116.2). The clinical utility of screening for HLA-B*5701to prevent hypersensitivities to abacivir, has been demonstrated in a prospective randomised controlled trial and is now used as a routine investigation before embarking on treatment.
Whether similar advice should be applied to the other associations shown in Table 1 is less clear, and depends on a variety of factors including the sensitivity and specificities of the test as well as the population frequencies of the mutations. McCormack and colleagues suggest that to prevent one case of carbamazepine hypersensitivity 83 individuals would need to be screened for the HLA-A*3102 allele. At a cost of €100 per test, the resources needed to prevent one case would thus be €8,300. Since the overall mortality from the various manifestations of carbamazepine hypersensitivity is under 1% it means that the cost per life saved would be likely to be over €1 million.
Clinical pharmacologists are only at the foothills of exploring the potential of pharmacogenomics. Many idiosyncratic reactions will, at some future date, be shown to have a polygenic basis that will explain, for example, why only a few women with the factor V Leiden mutation, taking oral contraceptives, develop VTE while many do not. Similarly, other genetic factors will be identified to allow reactions associated with HLA alleles to be predicted with near certainty.
Despite the pessimism advanced earlier, in this review, pharmacogenomics holds real promise for the future. By 2032, and hopefully much sooner, all developed countries will routinely use electronic health records. This will not only enable physicians and prescribers to have ready access to past and current medical records (including diagnostic information) but also incorporate patients’ whole genome scans together with decision-support tools. This will enable automated drug selection preferences to be made that is individualised for each patient. Similarly, individualised drug dosage regimes will be available depending on the mutations in the relevant drug metabolising genes at no additional cost because the relevant genetic information will already be encoded in the record.
Pharmacogenomics really does hold promise even if – at the moment – its place in routine care is postponed.
Table 1 - Some associations between HLA alleles and dug-induced hypersensitivity reactions
|HLA alleles||Drug||Hypersensitivity reaction|
|HLA-B*1502||Carbamazepine||Stevens-Johnson syndrome/toxic epidermal necrolysis|
- N Engl J Med 2003;348:538-49. CDI #nnr#
- N Engl J Med 2011;364:1144-53. CDI #nrr#
- Rawlins MD. Therapeutics, Evidence and Decision-making. London: Hodder Arnold, 2011.
- Thromb Haemost 2008;100:447-52. CDI #rrr#
- Pharmacogenomics 2010;11:497-9. CDI #rrr#
- N Engl J Med 2011;364:1134-43. CDI #nrr#