How to assess pharmacogenomic tests for implementation in the NHS in England

Pharmacogenomic testing has the potential to target medicines more effectively towards those who will benefit and avoid use in individuals at risk of harm. Health economies are actively considering how pharmacogenomic tests can be integrated into health care systems to improve use of medicines. However, one of the barriers to effective implementation is evaluation of the evidence including clinical usefulness, cost‐effectiveness, and operational requirements. We sought to develop a framework that could aid the implementation of pharmacogenomic testing. We take the view from the National Health Service (NHS) in England.

K E Y W O R D S health service, implementation, pharmacogenomics 1 | INTRODUCTION

| Pharmacogenomics
Pharmacogenomics-the study of the interaction between drugs and the genome-could allow prescribers to target medicines more effectively at those who benefit and to avoid their use in those who could be harmed. Genetic variants can affect pharmacokinetic mechanisms (drug handling), such as drug transport or metabolism, or pharmacodynamic responses (the effect of drugs on their therapeutic target). This can lead either to drug toxicity or lack of efficacy, especially for drugs with a narrow therapeutic range and for prodrugs that rely on pharmacokinetic activation. 1 Commonly used drugs may bring benefits to only a minority of those who take them. For example, approximately 1 in 30 of those taking moderate doses of statins for primary prevention of atherosclerotic cardiovascular disease according to standard guidelines is likely to benefit over 10 years of treatment. 2 In terms of toxicity, adverse drug reactions are linked to 1 in 16 hospital admissions in the UK. 3 Understanding variability in drug response could allow drug treatment to be directed more accurately, and help curb unnecessary spending on medicines, which in the National Health Service (NHS) in England was £20.9 billion in the financial year 2019-20. 4

| United Kingdom approach
Genome UK: the Future of Healthcare outlines the UK government's ambitions to develop an evidence-based approach to implementing pharmacogenomics within mainstream healthcare. 5 More recently a joint report from the Royal College of Physicians and the British Pharmacological Society outlines how pharmacogenomics could be made more available in routine health care. 6

NHS England (NHSE) have established a National Genomic
Test Directory (NGTD), which outlines the genomic tests that are funded by the NHS in England. 7 The NGTD currently focuses on cancer and rare diseases, and includes four pharmacogenomic that test for four variants including the dihydropyrimidine dehydrogenase gene (DPYD) for fluoropyrimidines (Box 1), a mitochondrial RNR1 test for aminoglycoside antibiotics (Box 5) and TPMT and NUDT15 for purine analogue drugs. 7 This repertoire of tests may expand with NHSE plans to establish a pharmacogenomic test evaluation group.
NHSE have outlined a scoring framework and process by which test evaluation working group members can assess any genomic test for addition to the NGTD. 9 However, the information available to members to support this scoring process is unclear. Pharmacogenomic testing differs from many existing genomic tests in the What is already known about this subject • Pharmacogenomic testing has the potential to target medicines more effectively towards those who will benefit and avoid use in individuals at risk of harm.
• Health economies are actively considering how pharmacogenomic tests can be implemented into healthcare systems to make better use of medicines.
• One of the barriers for implementation is evaluation of the evidence including clinical usefulness, cost-effectiveness and operational requirements.

What this study adds
• We provide a framework that could aid the implementation of pharmacogenomic testing.
• We take the view from the National Health Service in England, although this framework could be applied more generally.
• We propose a centralized commissioning model that aims to reduce inequity and duplication but remain transparent and evidence-based. Here, we outline a framework that can be used by health care systems and commissioners to assess pharmacogenomic tests proposed for implementation. The current NGTD process is a positive first step but can be further developed by including these pharmacogenomic-specific considerations and adopting the gold standard evaluation methodology used by the National Institute for Health and Care Excellence (NICE).

BOX 1 Fluoropyrimidines and DPYD.
We further present a proposed model for national assessment of pharmacogenomic testing for the NHS.

| METHODS
We identified the themes of this framework after a literature search of EMBASE and Medline databases to identify prospective studies of pharmacogenomic testing focusing on clinical outcomes, and key studies on the mainstream implementation of pharmacogenomics.
We considered both efficacy (clinical trial evidence) and effectiveness (clinical usefulness in the real world). We use the term clinical usefulness to mean the extent to which a test reduces clinical uncertainty, influences clinical decisions, improves outcome, provides benefit greater than established measures and is generalizable. 11 We distinguish this from utility in the health economic sense, which is commonly measured in quality-adjusted life years (QALYs), "calculated by estimating the years of life remaining for a patient following a particular treatment or intervention and weighting each year with a quality-of-life score (on a 0 to 1 scale)." 12 By cost-effectiveness we mean the value for money, represented, for example, by the cost per QALY. We discussed these themes and developed this pragmatic framework of general principles. We formed a clinical expert advisory group by invitation of clinicians with expertise in pharmacology, pharmacogenomics, formulary evaluation and policy implementation who we knew through existing professional relationship. Our group held a series of online meetings to review the data from the literature search, prioritize themes and inform interpretation of the data, and developed this pragmatic framework of general principles.

| RESULTS
Themes that emerged from the discussions of the clinical advisory group are distilled as 10 questions. These provide a framework for evaluation of pharmacogenomic tests for clinicians and policy groups seeking to implement such tests. These are set out as below ( Figure 1).

| Is there robust evidence of a drug-gene pair association?
There are published, evidence-based, peer-reviewed guidelines of drug-gene pairs for which the pharmacokinetic or pharmacodynamic association is robustly established via biomedical research studies. 13,14 One set of guidelines advises how best to prescribe for patients of known genotype. 13 The other indicates which drugs suggest or mandate patient genotyping before prescribing-a more relevant consideration in the NHS at present. 14 The PharmGKB database lists 34 drugs whose European Medicines Agency (EMA) licences include actionable pharmacogenomic information, but pharmacogenomic testing has only been adopted into clinical practice for a few of these drugs. 15 Assessments of clinical usefulness and health economic utility are both necessary for effective implementation into practice.
F I G U R E 1 Checklist for assessing pharmacogenomic tests for implementation.
3.2 | Is there evidence in well-designed controlled trials that the pharmacogenomic test improves clinical outcomes?
Having established that variants in a specific gene affect a particular drug, the next step is to assess whether prospective pharmacogenomic information will improve its efficacy or safety. Ideally, this should be based on clinical outcomes rather than biomarkers alone.
Subsequently, the effectiveness of pharmacogenetic testing in clinical practice will need to be established.
Prospective randomized controlled trials in pharmacogenomics have proven challenging and costly, particularly for low prevalence variants that require large studies to reach adequate power. Ethnic differences in the prevalence of pharmacogenomic variants can also affect power calculations and the applicability of results. 16 Despite these challenges, robust randomized controlled trials in pharmacogenomics have emerged over the last decade, made more feasible by the falling cost of genotyping and multicentre collaborative approaches (see Box 2). [17][18][19][20][21][22][23][24][25] PREPARE, a randomized open-label multicentre European study published in 2023, examined the effect of screening for a panel of 12 genes in nearly 7000 patients who were to be prescribed one of the drugs whose adverse effects were likely to be related to one of the 12 genes.
The primary analysis showed a significant reduction in adverse reactions to the drug for which the test was performed (odds ratio 0Á70 [95% CI 0Á54-0Á91]; P = .0075). However, the odds ratio was unchanged by testing for the other 11 genes (odds ratio 0Á69 [95% CI 0Á61-0Á78]). The study concluded that, while feasible, challenges to wide-spread use of panel screening remained. 26 We agree that, even if testing for a relevant gene (or, rarely, genes) prior to treatment with a specific drug can be clinically useful, there is, up to now, no good evidence to support the clinical usefulness or cost-effectiveness of panel screening.
It will be important to implement tests that show demonstrably better patient outcomes or clinical pathways for clinicians, patients and policymakers to be confident in the adoption of pharmacogenetic testing, and ensure cost-effective use of NHS resources.

| Does the addition of the pharmacogenomic test improve clinical effectiveness (real-world outcomes)?
Pharmacogenomics is just one of many factors that determine an individual's drug response. Other important considerations include dose, age, sex, physiological factors such as pregnancy, exogenous factors such as diet or drugs, and diseases such as liver disease. 27 Up to now, few studies have evaluated whether implementation of genetic testing before starting treatment is both clinically useful and cost-effective. Our view is that such evaluation is necessary if a test is to be adopted widely in a financially constrained healthcare system. This might mean that the evidence threshold for implementation of pharmacogenomic testing is significantly higher than for other tests that guide therapeutic decisions. 28,29 For example, dosage adjustment in impaired kidney function has often been based on theoretical pharmacokinetic calculations rather than outcome data. However, studies demonstrate significant unwarranted variation and excessive use of non-genetic testing, which may contribute to patient harm and escalating costs, so this approach should not be regarded as a standard of best practice. 30 21,22 and reductions in clinical outcomes such as graft loss, acute rejection or adverse events were not observed. 22 13,14 These interventions need to be considered in the context of the clinical pathway and may need to be modified to allow for international differences in clinical practice and available treatments.

| Which patients should be tested and by whom?
Before pharmacogenetic tests are adopted, it is necessary to identify the criteria by which to judge whether a pharmacogenomic test will be clinically and cost-effective, and thereby define the patient cohort   Genotype-guided warfarin therapy is reported to improve the time in therapeutic range compared to standard dosing (mean difference 3.41%; 95% CI 0.71-6.10%; P = .01). 41 However, as more patients switch from warfarin to direct oral anticoagulant alternatives, with fixed dosing and less need for monitoring, testing for pharmacogenetic influences on warfarin will be a diminishing priority.
Tests targeted at specific genes or mutations have been the mainstay technology for assessing the clinical usefulness and decisions on clinical implementation of pharmacogenomic testing before instituting treatment. These tests are faster, cheaper and more widely available than next-generation sequencing. Turnaround times for tests depend on the technology used and can vary from 30 min for point of care testing, up to six weeks for whole genome sequencing. 6 Polymerase chain reaction, loop-mediated isothermal amplification and other gene-targeted technologies allow rapid testing for one or a small number of mutations. Gene panels that enable simultaneous testing for a range of pharmacogenomic targets have also been popular. 6 The choice of technology must also align with the clinical pathway so that test results are available when needed, to guide decisionmaking without causing harmful delays to treatment (Box 5).

| Is the pharmacogenomic test cost-effective and affordable?
The appraisal of a pharmacogenomic test should include measures of cost-effectiveness. Cost-effectiveness analyses of pharmacogenomic tests weigh up the value of testing according to health improvements against costs, and, like medicines, can be compared by using cost per QALY measures. These analyses must be assessed to ensure they are unbiased, relevant and consider key factors such as alternative treatments, rarity of outcome and cost of testing. 44 To date, cost-effectiveness studies have typically been limited to single drug-gene pairs associated with common but serious ADRs, such as HLA-testing for abacavir. 45 Testing a multigene pharmacogenomic panel has been proposed to be more cost-effective in the long-term.
Kimpton et al. 46 reported that, over a 5-year period, nearly half of patients aged 50-99 years would be expected to be prescribed 2 or more drugs associated with actionable pharmacogenetic variants. A panel approach to testing could inform prescribing decisions over a patient's lifetime but may be limited by lack of an interoperable patient record, the emergence of new treatments and evolving evidence on pharmacogenomic variants. While panel testing provides pharmacogenomic information for multiple treatments, it will be limited if the evidence for some included variants is weak. In some instances, genotyping will be less useful or less cost-effective than phenotyping. 47,48 The challenge of effectively interpreting results and communicating genetic risks to both clinician and patient is also amplified. 3.9 | Are any operational changes required to implement testing?
Pharmacogenomic testing can only be integrated into clinical pathways if scientific, laboratory and clinical teams collaborate to overcome operational barriers. For example, long laboratory turnaround times and inadequate resources to interpret and return the results can impede adoption. To achieve the benefits of personalized care, clinicians need to know when pharmacogenomic testing is appropriate, be confident in interpreting results and able to discuss the risks and benefits and the alternative treatment options with patients. 6 The large and rapidly increasing knowledge base for pharmacogenomics means that the introduction must be supported by evidence- Traditional genotyping assays take 3-4 days to return a result, but antibiotics for suspected sepsis must be delivered within 1 h. The PALOH study showed that a point of care test administered by clinical ward staff was feasible within the neonatal setting, with a turnaround time of less than 30 min. Genotype was used to guide antibiotic prescription without disrupting safe clinical practice. 43 3.10 | How will the impact of the pharmacogenomic test be evaluated?
Clear pharmacogenomic data standards will enable linking of laboratory and clinical datasets to measure the impact of pharmacogenomic testing over time, because clinical pathways and treatment decisionmaking are complex, variable and continuously evolving. Because the evidence on new variants is accumulating rapidly, evidence acquired at a single time-point may not achieve optimal benefit from pharmacogenomics and reporting to clinicians and patients needs to be revised as evidence evolves. Ongoing evaluation of prescribing behaviours and outcomes is also needed, to detect unanticipated consequences of pharmacogenomic testing (Box 6).

| DISCUSSION
Our view is that a structured approach should be taken to the assessment and governance of pharmacogenomic testing at a national, regional and local level, to ensure equitable access and reduce duplication (see Figure 2). The existing structure for medicines offers a good template for pharmacogenomic testing. For medicines, NICE provide independent expertise in health technology appraisal and costeffectiveness analysis to produce recommendations for adoption by the F I G U R E 2 Proposed model for the assessment, implementation and governance of pharmacogenomic tests in the NHS. GMS, genomic medicine service; NGTD, National Genomic Test Directory; NHSE, NHS England; NICE, National Institute for health and care excellence; RMOCs, regional medicines optimization committees.
NHS. Area Prescribing Committees and Drugs and Therapeutic Committees support local implementation of the guidelines.
Processes for central evaluation of genomic testing have started to be developed by NHSE via the NGTD. 7 These should further evolve to build on the gold standard evaluation methodology developed by NICE, with transparent methodology and incorporation of an independent analysis of the evidence, budget impact and pharmacogenomic-specific considerations highlighted within this review.

| CONCLUSION
Pharmacogenomic testing has the potential to improve treatment efficacy and safety. Implementation of testing into routine practice across a healthcare system will require careful consideration of the evidence, clinical usefulness, cost-effectiveness and operational requirements.
Our 10-point checklist outlines a standardized approach to evaluating applications to implement a pharmacogenomic test. A national approach to providing pharmacogenomic test recommendations and centralized commissioning will reduce inequity and duplication, but this process should be clearly set out, transparent, evidence-based, inclusive of the views of stakeholder and supported by appropriately skilled local teams to oversee implementation and monitor use of the tests in practice.  Reecha Sofat https://orcid.org/0000-0002-0242-6115