Abstract
Pharmacogenetics involves genetic testing of individual patients to guide drug treatment. Proponents argue that pharmacogenetics will achieve major gains in drug safety and efficacy, and revolutionise marketing. Pharmacogenetics also raises several policy concerns, including the need for sound information for clinical decision-making on drug-genetic test combinations. Currently, the pharmacogenetics science base and the rate of emergence of clinical applications are uncertain. Most commentary on pharmacogenetics focuses on new compounds, yet older drugs cause most adverse events. Test regulation in the USA appears fundamentally different from Europe, where evidence of safety or efficacy may not be required. Genetics research is needed as part of postmarketing surveillance systems. In routine clinical practice, computer-based health records with relevant decision support systems will also be needed. Without health policy action, pharmacogenetics could produce a new generation of poorly evaluated tests and drugs, with medicine becoming significantly less evidence-based, leading to rising costs, patient hazard and exclusions of drug-related ‘genetic minorities’ from evaluated treatments.
Introduction
Drug licensing and medical practice have traditionally gravitated towards a ‘one size fits all’ approach. Promising new drugs with serious side-effects, even in a small number of patients, have usually been dropped from development or withdrawn, while drugs that are ineffective in some patients have continued to be used because the non-responders could not be easily identified. As a result the majority of prescribed drugs are relatively safe, but only a third of patients are thought to benefit from them. 1
Occasionally, the drawback of this approach becomes obvious. One example is when patients with minor injuries are given codeine. In 9% of the Caucasian population the enzyme needed to metabolise codeine into its active form is absent, and codeine is ineffective. 2 Studies of such genetic influences on drug response are not new, but recent advances in genotyping technology have greatly reduced costs and constraints on such work. New techniques are being used to identify novel biological targets for medicines (so-called ‘pharmacogenomics’), but pharmacogenetics offers a different approach, in which individual patients are genetically tested and the results are used to guide clinical decisions. 3 Proponents have argued that pharmacogenetics will revolutionise drug marketing and prescribing.4, 5
Applications of pharmacogenetics
Genetic tests could be used in three different ways to individualise drug treatment: to adjust doses; to maximise effectiveness; or to avoid idiosyncratic side-effects. In the first category, tests might identify individual differences in absorption, metabolism, transportation or elimination of a drug. Individually adjusted doses could help avoid adverse events in those who were previously ‘overdosed’, or achieve greater effectiveness in those receiving too little. Secondly, testing could identify those who would respond particularly well to a specific compound. For example, there is some suggestion that polymorphisms play a role in the response to certain categories of asthma medications. 6 In practice, one consequence of this type of testing might be subdivision of existing disease categories, according to differing underlying mechanisms. Alternatively, some differences in effectiveness may be due to genetic co-factors modifying response within the same underlying disease. A third type of genetic testing would identify those at risk from unusual adverse events. For example, most drugs implicated in heart rhythm changes and sudden cardiac arrest are metabolised by a key liver enzyme (CYP2D6) of which there is a polymorphism conferring increased risk. 7
If genetic tests were to work well clinically, the gains would be substantial. A recent meta-analysis in the USA suggested that adverse drug reactions are the fourth most common cause of death. 8 Phillips et al 9 found that, of 27 drugs frequently cited in adverse drug reaction studies, 59% are metabolised by at least one enzyme that has a variant allele known to cause poor metabolism.
In contrast to traditional therapeutic drug monitoring, 10 pharmacogenetic testing could be undertaken before treatment begins, does not require the assumption of steady-state conditions (or patient compliance) for the interpretation of results, could be performed less invasively (e.g. with mouth swabs), could provide predictive value for multiple drug substrates rather than a single drug, and would be constant over an individual's lifetime.
Currently, pharmacogenetic testing is used only in a few special circumstances, such as in the management of children with acute lymphatic leukaemia. 10 Significant bodies of knowledge have accumulated for potential application of pharmacogenetics in several areas, including hypertension, 11 cancer, 12 oestrogen therapy 13 and psychiatry.14, 15
Overall, Ingelman-Sundberg 16 has speculated that 10-15% of today's drugs are affected by a limited number of high penetrance genes, for which genotyping is likely to be of benefit. A further 35-40% of drug therapy is affected by polygenic factors.
Areas of policy concern
Although the case for the pharmacogenetic approach is appealing, there are also several areas of concern. For example, in addition to genetic differences, there are numerous other factors that alter drug response, including other drugs, nutrition, ageing, liver function and kidney function. We also need to remember that poor quality of care and medication errors are important in adverse events, 17 and that many patients do not take the drugs prescribed for them in the ways intended. In addition, issues of equity and the potential for adverse social implications from genetic testing have caused concern. An examination of the full implications of this technology is therefore necessary.
In this paper we focus on the need for accurate, transparent and timely clinical information to support future evidence-based practice with pharmacogenetic test-drug combinations. Good evaluative information would provide the basis for dealing with subsequent implementation issues, including deciding on clinical appropriateness, funding, delivery logistics and information management. Three core questions are addressed. What information about pharmacogenetic tests and drugs will doctors and patients need in order to make informed decisions on their use? What form of evaluation will be needed to provide the necessary clinical information? And what policy initiatives will be needed to enable or ensure that the evaluation is completed? The authors’ overview of policy issues presented here has been influenced by literature reviews and a programme of interviews and focus groups with policymakers, regulators, academics, pharmaceutical company scientists and clinicians in the USA and UK, reported in detail elsewhere.18, 19 The policy issues raised currently provide challenges mainly to those involved in regulation and health technology assessment within health systems internationally, although, as products come into use, the issues will impact at many levels of health care.
While many experts are confident about the potential contribution of pharmacogenetics in the longer term, new pharmacogenetic drug licensing applications have been conspicuous by their paucity. Thus far, research in the public domain is dominated by genetic testing linked to existing compounds, including agents to reduce blood clotting (warfarin) and antidepressants. Many tests are likely to be developed independently of the pharmaceutical companies, with target drugs that are already out of patent (generics).
The main emergent technical themes for health policy include uncertainty about the timing and extent to which pharmacogenetics will produce clinical applications, issues of test and drug evaluation, and issues in adverse event monitoring. Each of these is discussed below.
How soon will pharmacogenetics deliver clinical applications?
While much of the debate about pharmacogenetics has thus far assumed that the science is sound, this has been questioned. Existing pharmacogenetic association studies have been criticised16, 20 for often being based on poorly characterised patient material, having too few participants, patients not being thoroughly phenotyped and a severe publication bias in favour of reporting positive results. There are also concerns about both the cost-effectiveness of patient testing and the commercial attractiveness of stratifying the target population for new drugs. 21
Even where pharmacogenetic differences are clear-cut, their clinical significance can be limited. For example, selective serotonin reuptake inhibitors (SSRIs) for depression show considerable pharmacokinetic variability, but this usually does not explain insufficient response to therapy because SSRIs have a wide therapeutic index. 22 Given the many competing influences on the safety and efficacy of drugs, it may be that genetic tests alone will seldom provide good enough markers for clinical use. Developing clinical applications may take longer and be more complex than has been claimed.
Test and device evaluation
Most evaluative information for new drugs is driven by regulatory or licensing requirements, or generated by publicly funded research. However, the regulatory requirements for new tests are currently very different from those for drugs. In the past, clinicians largely decided on the choice of biological targets for clinical tests on the basis of academic evidence. However, equipment marketed to carry out tests on patients’ samples is regulated under in vitro medical device regulations. Arrangements for laboratory services offering testing are also distinct.
In Europe, test device regulation is principally concerned with technical or analytical accuracy, identifying whether a test measures the chemical or gene it claims to measure. 23 Manufacturers are not formally required to make or justify claims about the clinical usefulness of tests. The European In Vitro Diagnostic Medical Devices Directive 24 does not include clinical utility or predictive value amongst the ‘essential requirements’ of a new test-kit, and tests provided as a laboratory service are not covered by device regulation at all. But, on the other hand, in the USA, since the passage of the Safe Medical Devices Act of 1990, the US Food and Drug Administration (FDA) does require information not only on analytical but also clinical performance of test kits, including clinical or diagnostic sensitivity and specificity and expected predictive value. 25 Because the safety of a device such as a pharmacogenetic test is based on the impact that information generated by the device has on patient management (including harm from false-positive or false-negative results), the terms safety and effectiveness are seen as linked in US regulation. 26 How important these differences will be in practice has yet to be seen, with some experts predicting that the implementation of the European arrangements will ensure that clinical claims are made and supported. The extent to which this occurs should clearly be monitored.
Much genetic testing is not carried out with test equipment marketed to health professionals or the public, but rather by laboratories offering testing services. At present, in the USA and Europe, a variety of systems for ensuring the quality of procedures in laboratories are in place, but testing services have not been regulated to ensure the evaluation of tests for clinical utility, safety or effectiveness. Some have suggested that the new European device directive will cover commercial receipt of specimens for testing from outside organisations, although the details of how this will operate are not yet known. In the USA, debate continues as to whether the FDA has the legal right to regulate laboratory testing.
International harmonisation of the evaluation of pharmacogenetic testing is currently under discussion. As a number of poorly evaluated genetic tests are already being marketed to the public and technology offers the prospect of cheaply testing thousands of genetic targets at once, the problems of having a limited evidence base for pharmacogenetic tests could worsen dramatically.
Licensing and use of drugs
Much clinical information on new pharmaceuticals is derived from licensing trials. Several existing regulatory guidelines - notably on drug-drug interactions and metabolism - already encourage pharmacogenetic analysis. 27 However, there is currently little experience of pharmacogenetic test-new drug combinations in Phase III (pre-marketing) trials. Such products might require joint or sequential evaluation between drug and device agencies, but it is currently unclear how this will be coordinated in Europe, or how standards for evaluation will be set.
Industry optimists believe that identification of potential responders will permit smaller clinical trials. However, smaller trials would also reduce the chances of detecting significant adverse events and would be less generalisable. A drug licence granted under these circumstances would need to reflect the narrow patient selection, but there are two major problems arising from this approach: off-label use and population diversity.
Regulatory agencies have long been concerned about ‘off-label use’ of prescribed drugs - that is, use for new indications or different patient groups not included in the licence or label. A substantial proportion of adverse events and non-response is thought to be due to off-label prescribing. Avoiding the potential hazards of unlicensed use of drugs in genetic risk groups will depend on the weakest link in the health care delivery chain. Pharmacogenetics is thus likely to lead to further questioning of the scope of clinical discretion to use compounds off-label, and health care providers will have to strengthen quality assurance systems to ensure that targeted use of drugs is reliably delivered in routine clinical settings.
There are also ethical concerns about genetics-based selection for drug trials, as it involves explicit exclusion of certain groups from access to evaluated medication. There are substantial genetic differences in drug response genes geographically and between ethnic groups, 28 and a licensing approach that was effectively targeted at the larger genetic groupings within, for example, the American population of European extraction, might result in unacceptable exclusion of ethnic minority groups from trials. Similarly, pharmacogenetic selection of patients entering trials will raise questions about the generalisability of trial results to other countries. There are already requirements to conduct ‘bridging trials’ when marketing a product in countries where the population's metabolism may be different to that of the trial subjects, but pharmacogenetic selection could exacerbate the situation.
Post-marketing surveillance
Trials for drug registration are rarely large enough to link exposure and adverse events. Unexpected side-effects often become evident only after marketing, when large numbers of patients are exposed. Identifying genotype-phenotype correlations for adverse events will clearly require larger populations than are included in licensing trials. 29 Post-marketing surveillance systems are, therefore, the main vehicle by which genetic information could be collected for pharmacogenetic analysis of idiosyncratic adverse events. It has long been recognised that an independent monitoring body and extensive post-licensing observational studies are needed to improve post-marketing surveillance. 30
It has been suggested that patient screening programmes for implicated markers could also be used to restore previously withdrawn drugs to the market, by excluding those patients at risk. Supporting such decisions could be challenging, especially as error could lead to severe hazard and litigation, and testing systems would have to be highly reliable.
Defining the policy options for ensuring a clinical evidence base for pharmacogenetics
It is clear that a range of policy challenges might need addressing to support the development of pharmacogenetics in clinical practice. Much of existing medical knowledge is poorly evaluated, and a case could be made that pharmacogenetics does not raise any fundamentally new issues and should not be singled out for special attention. On the other hand, it could be argued that pharmacogenetics will greatly intensify existing problems. Without policy action, many poorly evaluated genetic tests could result, with medicine becoming significantly less evidence-based, leading to rising health care costs, increased waste, patient hazard and exclusions of drug-related ‘genetic minorities’ from evaluated treatments.
A range of specific choices face health policy-makers. It is a characteristic of innovative health technologies that they are dealt with initially by existing regulatory structures, and only when these prove inappropriate is the system adapted. 27 If this approach is followed, then ensuring the adequate evaluation of pharmacogenetic tests relevant to already marketed drugs appears to be the most urgent policy challenge.
A US workshop convened by the Centers for Disease Control has set out the current consensus on the academic standards for evaluation of genetic tests. 31 Their report makes a series of critical distinctions, identifying the elements involved in the validity of a test. The report also points out that the benefit of a genetic test can be evaluated only in the context of specific health outcomes. Thus, the starting point for considering the use of a genetic test is a well-defined clinical problem, for which the test is expected to improve care. Evaluation of all four aspects of validity of a test is then needed in relation to the identified clinical problem: analytical validity; clinical validity; clinical utility; and the ethical, legal and social implications of taking the test. It is clear that current device regulatory arrangements in the USA and Europe do not require such a thorough level of evaluation. The current imbalance in requirements for tests as against the exhaustive requirements for new drugs will be difficult to defend, especially as a misleading test result from a pharmacogenetic test could produce the same consequences for patients as receiving a dangerous drug.
In terms of new drugs and test combinations, current licensing systems have already begun to respond, but international variations in regulatory approach present a challenge to drug makers. Uncertainty, in the context of the very long lead times for the development of new drugs, could be a major disincentive to industry in pursuing pharmacogenetics. 32 In post-marketing surveillance, the case for incorporating genetic data into the analysis of serious adverse events seems strong, but there are policy choices to be made over the extent and organisation of such a programme.
In making these policy choices, a number of potential technical and political hurdles exist. Political issues will no doubt be important in shaping responses, including pressure from sectional interests, public fears about genetics, short-term expediency preventing pursuit of long-term goals, and the lack of organisational capacity at the appropriate national, regional and global levels.
Some pharmacogenetic tests will raise similar ethical issues to other genetic testing, with the potential to reveal sensitive personal information. 33 Where tests relate to the development of diseases (rather than drug metabolism), results may also be predictive of future health status. Informed consent and privacy of pharmacogenetic test results may be essential in most clinical uses of pharmacogenetic drugs. 34 However, current confidentiality requirements could greatly impede the use of clinical databases for pharmacogenetic research, 35 especially in the crucial area of pharmacovigilance.
In reviewing these complex issues, policy-makers will have to remember that pharmacogenetics will not necessarily be profitable for drug manufacturers 36 and may not develop without a balance between incentives and regulation. Given the problems of population size and R&D economics, there is likely to be a major public sector role in developing this technology. 21 Enabling both industry and academic researchers to contribute will be very important to delivering the potential benefits of pharmacogenetics. 37 Ensuring the availability of independent public sector expertise and resources to evaluate pharmacogenetic claims is also crucial to maximising benefits and minimising risks with this new technology. Targeting the public sector effort to the cost-effective areas of this technology will also require complex judgements.
If the technology becomes a regular part of practice, educational and decision support systems will be needed to ensure ordinary prescribers use pharmacogenetic products appropriately. Computer-based systems will probably be needed to achieve this. 38
Conclusion
Pharmacogenetics offers the promise of moving some medicines from a ‘one size fits all’ approach to a more individualised style of drug prescribing. Achieving the consequent gains in safety and efficacy is likely to pose complex challenges for existing health technology assessment arrangements, and will require difficult policy judgements.
While there is significant uncertainty over the pace and ultimate contribution of genetic testing, it is already clear that current test regulation is unlikely to produce the full clinical evaluative data that are seen as needed to inform doctors and patients. The routine quality of the clinical evaluation of pharmacogenetic tests is a major concern, especially in relation to tests offered by laboratory services, and for tests marketed under the new European arrangements. Pharmacogenetic testing holds out the exciting prospect of reducing some serious adverse events, with genetic data and research being incorporated into pharmacovigilance systems. A good policy regime for pharmacogenetics could benefit all, leading to better understood products and greater certainty for long-term investment decisions in pharmacogenetics.