CYP2C19 genotype testing to guide clopidogrel use after ischaemic stroke or transient ischaemic attack

The committee said there was strong evidence that people with loss-of-function CYP2C19 alleles had worse outcomes when taking clopidogrel than people without loss-of-function alleles. But the evidence was less clear on the benefits of treatment with other antiplatelets. The committee recalled that clinical experts had said during scoping that dipyridamole plus aspirin was the most likely alternative antiplatelet that would be used in the NHS. But no data was found on the impact on people with loss-of-function alleles who had clopidogrel compared with dipyridamole plus aspirin.

A clinical expert noted that the latest Royal College of Physicians national clinical guideline for stroke recommends ticagrelor as an alternative antiplatelet for people who have had a TIA or minor stroke, but not a major stroke. Ticagrelor does not have a marketing authorisation for TIA or stroke in the UK. The committee recalled that the EAG reported evidence that, compared with clopidogrel, ticagrelor decreased the risk of secondary vascular events in people with loss-of-function CYP2C19 alleles. An alternative network meta‑analysis done by the EAG for the second committee meeting, which included further studies assessing ticagrelor (such as the THALES study), showed a higher risk of major bleed for ticagrelor compared with clopidogrel.

The committee considered that tests that only detect the most common loss-of-function alleles may be more likely to introduce inequalities. This is because less-common loss-of-function alleles are more prevalent in certain ethnic groups (see section 2.7). The EAG estimated that the combined prevalence of the *4, *8 and *35 alleles in the UK stroke population would be around 0.6%. But it noted that the *35 allele has a prevalence of up to 3% in people with sub‑Saharan African ethnic background, and that the *4 allele is more common in the Ashkenazi Jewish population. So, tests that detect a smaller range of alleles would likely disproportionately fail to identify people with loss-of-function CYP2C19 alleles in certain ethnic groups. Some clinical experts suggested that commissioners could consider the demographics in their local area when deciding how to do CYP2C19 genotype testing. Other committee members felt that a wide range of alleles should be tested for to minimise potential inequalities. The committee noted that the significance of some CYP2C19 variants, particularly if they are very rare, may be uncertain. Experts highlighted that there is information that could be used to guide decisions on the alleles tested for, such as the Association of Molecular Pathology Pharmacogenetics Working Group's recommendations on minimum and optional sets of alleles, or the Clinical Pharmacogenetics Implementation Consortium's guideline for clopidogrel and CYP2C19. The Association for Molecular Pathology's minimum set of alleles for testing includes the *2 and *3 loss-of-function alleles, with more alleles specified as optional for inclusion in an extended panel. In a survey of genomic laboratories done by the EAG, when asked which alleles would be tested for in a request for a CYP2C19 test, 2* and 3* were the alleles with the highest response. These are also the only loss-of-function alleles tested for when CYP2C19 genotype testing is done in Scotland (see section 3.19). A clinical expert commented that laboratory-based testing is adaptable and can change which alleles are tested for over time, whereas point-of-care tests can only detect certain alleles. Laboratory-based tests can test for a broader range of loss-of-function alleles than point-of-care tests, although this depends on the specific technologies used (see sections 2.10 to 2.16).

The committee agreed that CYP2C19 genotype testing would be appropriate for children and young people after an ischaemic stroke, if treatment with clopidogrel was being considered. Clinical experts noted that stroke in children and young people is very rare and normally has a different cause to stroke in adults. They also acknowledged that clopidogrel is not indicated for use in children. Clopidogrel is not normally prescribed except where there are other risk factors for cerebrovascular disease, such as cardiovascular conditions. The EAG did not identify any evidence for children or young people in their review of clinical effectiveness. But, clinical experts said that there is no biological reason why the interaction between drug and genotype would be different in children and young people compared with adults. The Clinical Pharmacogenetics Implementation Consortium states that it is reasonable to extrapolate its recommendations to children and young people if needed (the guidance was also based on data from studies in adults). Experts also said that the benefits from successfully preventing further clotting events by prescribing appropriate antiplatelet therapy would likely be larger for children and young people because of the longer expected remaining lifetime.

For some people who have a stroke or TIA it will not be their first event (that is, it is a recurrent stroke or TIA) and they will already be taking clopidogrel. The 2023 update to the Royal College of Physicians national clinical guideline for stroke recommends that if a person has a recurrent cardiovascular event while on clopidogrel, then clopidogrel resistance may be considered. The committee noted that this group was within the scope of the recommendations for using testing, provided continued treatment with clopidogrel is being considered. Clinical experts commented that further investigations are likely to be done for people who have a stroke or TIA while on antiplatelet therapy, for example to check for atrial fibrillation. An alternative antiplatelet treatment is also likely to be considered. Clinical experts said that there is variation in practice in terms of treatment decisions for people who have a recurrent stroke while taking clopidogrel. Treatment may include short-term (3 to 6 weeks) dual antiplatelet therapy (for example, aspirin with clopidogrel) before continuing on a single or dual antiplatelet treatment in the long term. Or, people may be switched to long-term aspirin or ticagrelor. A clinical expert said that without CYP2C19 genotype testing, people may be returned to long-term clopidogrel monotherapy. A clinical expert commented that, in centres where CYP2C19 genotype testing is done (see section 3.19), people with a loss-of-function allele will be switched to an alternative treatment such as aspirin with dipyridamole. A stakeholder highlighted that non‑adherence with prescribed antiplatelet treatments may be a cause of apparent clopidogrel resistance. Studies have reported high levels of non‑adherence with antiplatelet treatment. But, a patient expert commented that in their experience people prescribed clopidogrel tend to take it. People having a recurrent stroke or TIA while taking clopidogrel are potentially more likely to have a loss-of-function allele, particularly if other causes, such as non‑adherence to prescribed treatment, are ruled out.

Clinical experts felt that the full cost of laboratory-based testing was likely to be lower than the cost used in the EAG's model (£139 per test). A scenario run by the EAG that assumed tests would be done in batches, estimated a cost per test of £44. An expert commented that laboratory-based CYP2C19 genotype testing in their region was about £20 to £40 per test (when testing for 3 alleles). The EAG explained that the responses to its survey of genomic laboratory hubs did not express a clear preference for the method of CYP2C19 genotype testing, and there was a lack of agreement on the staff time needed. So, the true cost of laboratory-based testing is uncertain. The cost used for laboratory-based testing in the model was for the Agena Bioscience MassARRAY, with an assumed 1‑year lifespan. This was amended to 5 years in an updated base case, in response to consultation comments. The EAG's estimated cost per laboratory test included the cost of the iPlex testing platform, which experts indicated would also be used for tests other than CYP2C19 genotype testing. Some committee members said that the cost of laboratory testing could come down over time because of economies of scale or new technologies. But, the committee noted that costs for laboratory testing may have been underestimated if higher implementation costs were incurred but not captured in the EAG's estimates (see section 3.19). The EAG acknowledged the uncertainty about test costs. But it highlighted that a threshold analysis it had done showed that laboratory-based testing was cost effective compared with no testing if the cost per test was less than £1,920.

A clinical expert suggested that the costs used for the point-of-care tests may have been underestimated. This is because it is likely that multiple machines would be needed in each centre to handle the volume of testing or as backup in case of failure. The EAG clarified that its cost estimates were based on an average cost per test, which could account for multiple devices. A committee member highlighted that introducing point-of-care tests to stroke services would add new processes such as taking cheek swabs and running tests, which are not currently part of practice. The committee concluded that there is considerable uncertainty about the cost per test. But it acknowledged that when compared with no testing, costs would have to be much higher (around £1,800 higher for the Genomadix Cube and £1,900 for Genedrive) than those used in the EAG's model, for testing not to be cost effective.

Several stakeholders questioned how the care pathway had been modelled, stating that it may not reflect current UK practice. The EAG explained that it had provided scenario analyses that varied the modelled care pathway. These included which alternative antiplatelet is used instead of clopidogrel and when antiplatelet treatment is started. The tests were still cost effective when the choice of antiplatelet and timing of treatment were varied.

Testing done for people with minor stroke or TIA generated less net monetary benefit (NMB) than testing done for people with non‑minor stroke in the EAG's model. The EAG explained this was because people with non‑minor stroke had a higher rate of recurrent stroke in the long term, and so experience greater benefit from appropriate treatment. Experts noted that while recurrence rates were similar in the first 90 days, after that rates in the minor stroke and TIA population were much lower. The EAG explained that the parameter values were from a recent retrospective cohort study using the Framingham Heart Study data. Experts considered that the long-term recurrence rates for non‑minor stroke appeared to be in line with their experience, but that those for minor stroke and TIA were too low. Following the second committee meeting, clinical experts suggested alternative evidence sources (4 studies) that they considered better reflected the long-term recurrent stroke rates after a minor stroke or TIA. The EAG did further analyses in the minor stroke and TIA population using these higher rates of recurrent stroke, presented in an addendum to the main report. This increased the number of incremental quality-adjusted life years (QALYs) for the testing strategies (compared with no testing). This resulted in a higher NMB (ranging from £753 to £1,131) compared with the base-case analysis. But, the NMB remained lower than for the non‑minor stroke population.

The committee agreed that the 2 point-of-care tests considered in the assessment are different. So it was not appropriate to use data for the Genomadix Cube to model performance of the Genedrive CYP2C19 ID Kit. The committee recalled that several studies were identified for the Genomadix Cube test, but at the first committee meeting no data on accuracy or failure rate was available for the Genedrive test. By the second committee meeting, the Genedrive test had received its UKCA mark and the company provided data on test performance. The committee acknowledged that this showed that the test performed well. But it noted that further data on performance and failure rates would be beneficial. The company commented that further evidence is being generated. The committee welcomed this and encouraged centres already using this test to take part in data collection (see section 3.23). The committee also noted that several features of the Genedrive test could offer advantages over the Genomadix Cube (see sections 2.12 to 2.15). For example, its reagents do not need to be stored in a freezer and it can detect several additional alleles including those that occur in greater frequency in some ethnic groups (see section 3.8). The committee also noted that the estimated cost per test for Genedrive is less than for Genomadix. So the committee concluded that Genedrive was its preferred point-of-care test.

The committee stated a preference for laboratory-based tests over point-of-care tests. It noted that there was very little difference in the QALYs generated in the EAG's model by the different methods of testing. The committee had previously concluded that tests that detected fewer loss-of-function alleles would likely disproportionately affect certain ethnic groups (see section 3.8). While Genedrive detects more alleles than Genomadix, laboratory-based testing has the potential to detect an even broader range. Also, if needed, laboratory testing can be adapted more easily to assess other alleles in the future. Several stakeholders and experts also commented that centralised testing would reduce variability in testing offered across the NHS. Experts raised concerns that if left to local centres to implement testing with point-of-care tests, this would likely lead to considerable variation and could worsen health inequalities. A stakeholder also commented that centres usings point-of-care tests would need to have quality-assurance processes in place and consider who is responsible for ensuring these. Some committee members stated that existing infrastructure should be preferentially used over investing in new single-purpose technologies. Experts also highlighted that, in the future, pharmacogenomic testing may be reactive when clopidogrel is needed, but pre‑emptive pharmacogenomic tests for other treatments could be done at the same time. This would require a panel of tests that would be more easily done in a laboratory. There was also relatively little evidence for Genedrive (see section 3.15). So, despite Genedrive having the highest net benefit of the 3 tests assessed in modelling, the committee concluded that laboratory-based testing was its preferred method. But the committee also acknowledged that there are barriers to implementing laboratory-based testing (see section 3.19). If laboratory-based testing is not possible or feasible, or it will take a long time to develop capacity to provide it, then point-of-care tests could be used.

The committee agreed that CYP2C19 genotype testing was likely to be cost effective. It recalled that in the EAG's updated base-case analysis, provided for the second committee meeting, testing dominated no testing in probabilistic analysis (it cost less and produced more QALYs). Testing had positive NMB for all scenario analyses for the non‑minor stroke population. This was including all alternative antiplatelet agents modelled (dipyridamole plus aspirin, ticagrelor plus aspirin or aspirin alone), based on potential variation in UK practice. NMB was lower for the minor stroke and TIA population, and not positive in all scenario analyses. But the committee recalled that NMB estimates for this population were increased (although remained lower than for non‑minor stroke) in the EAG's additional analysis that used higher long-term recurrent stroke rates (see section 3.14).

The committee also considered that CYP2C19 genotype testing was likely to be cost effective for children or young people. The committee recognised that there was no data for children or young people, and that clopidogrel is rarely used in this population. But, clinical experts advised that, if clopidogrel was being considered, information on CYP2C19 genotype would still be useful (see section 3.9). The committee noted that CYP2C19 genotype testing was more cost effective in the EAG's scenario analysis, which used a younger cohort of adults (average age 40 years) than in the base case (average age 71 years). Clinical experts suggested that this would be more pronounced in children and young people because of their longer expected remaining lifetime.

Several stakeholders highlighted that there would be considerable challenges to implementing a new laboratory-based test for all people who have a stroke or TIA. This included the need to expand testing capacity. The committee noted that CYP2C19 genotype testing was being done by NHS Tayside in Scotland. An expert commented that this was now being considered for roll out nationally in Scotland. But they also highlighted that rolling out testing over a much larger number of people would have considerable additional challenges. Experts further highlighted a shortage of clinical scientists, which would impact on the possibility of increasing laboratory testing capacity. The committee acknowledged that implementing testing for everyone who has a stroke or TIA could be done in a stepwise process. This would involve a gradual increase in numbers tested while capacity was established. The committee suggested that the NHS may need to identify subgroups of the overall population to start offering testing to initially. If a group was to be prioritised for earlier roll out of testing, there would need to be capacity to test everyone in this group. The 2023 update to the Royal College of Physicians national clinical guideline for stroke recommends considering testing for clopidogrel resistance only for a subset of people who have had a stroke or TIA (a recurrent TIA or stroke while taking clopidogrel; see section 3.10).

The committee recalled that, compared with no testing, CYP2C19 testing after a non‑minor stroke generated more NMB than testing after a minor stroke or TIA. This was the case even when higher recurrent stroke rates for the minor stroke and TIA population were considered (see section 3.14). Estimates provided by the NICE resource impact team indicate that the non‑minor stroke population is around 35,900 people per year (rather than about 110,000 people who have any stroke or a TIA). The committee concluded that to facilitate implementation, testing could potentially be started in the smaller non‑minor stroke population before being expanded to the minor stroke and TIA population.

The committee also recalled that it considered that point-of-care tests could be recommended for use if laboratory-based testing is not available (see section 3.16). It noted that point-of-care tests may have less potential to detect a wider array of loss-of-function alleles, although this will depend on how many alleles laboratory-based testing includes; for example, if it goes beyond *2 and *3 alleles (see section 3.8). This could disproportionately affect certain ethnic groups. But, if the alternative is no testing, all groups will benefit from some form of testing; the loss-of-function allele that is most common across all groups is the *2 allele, which is included in both point-of-care tests assessed. The committee considered the implementation challenges and delays associated with laboratory-based testing. It agreed that point-of-care testing should be considered as an alternative if laboratory-based testing is not feasible at this scale, or while capacity for laboratory-based testing is increased.

Stakeholders highlighted concern that waiting for test results may mean that starting antiplatelet treatment is delayed inappropriately. The EAG's model included starting initial clopidogrel treatment and changing to an alternative antiplatelet treatment later if genotype testing indicated a loss-of-function allele was present. The committee agreed that this was appropriate, and that starting antiplatelet treatment should not be delayed while waiting for test results. Stakeholders also raised concern that if testing was recommended but not available, alternatives to clopidogrel may be used more often, including ticagrelor, which is more expensive. The committee noted that many publications already highlight the prevalence of CYP2C19 loss-of-function alleles, and that this is higher in some groups. For example, a recently published study reported that 57% of a cohort of people in the UK from Bangladeshi and Pakistani backgrounds had at least 1 loss-of-function allele (Magavern et al. 2023). Recommendations to use CYP2C19 genotype testing to inform choices about antiplatelet treatment already exist, for example from the Clinical Pharmacogenetics Implementation Consortium. Testing that is already being used in the NHS in Tayside (see section 3.19) has also been publicised. So, the incidence of loss-of-function alleles and their impact would already be widely known. The committee noted that this guidance does not replace any existing guidance on using antiplatelet therapy if genotype testing is not available. Experts also strongly emphasised the importance of shared decision making with people when deciding which treatment to use, and the important role stroke pharmacists can play here.

The committee encouraged centres using the Genedrive CYP2C19 ID Kit test to take part in data collection to further determine the accuracy and failure rate.