Development of a point-of-care genetic test for effective treatment of ischaemic stroke: an early model-based cost-effectiveness analysis

Background: People who have experienced a stroke are at high risk of recurrent strokes. Clopidogrel is prescribed to people who have had a non-cardioembolic stroke. There is evidence that clopidogrel is not effective for patients with CYP2C19 loss-of-function alleles. Pharmacogenetic testing is a potential strategy to identify such patients and guide prescription of appropriate antiplatelet treatment. This study aimed to provide an early estimate of the cost-effectiveness of using a point-of-care pharmacogenetic CYP2C19 test in the UK National Health System. Methods: A decision-analytic model comprising a linked decision tree and Markov model were created in R comparing pharmacogenetic testing with current prescribing practice. In the pharmacogenetic testing arm, patients identified to have one of three loss-of-function alleles were prescribed modified-release dipyridamole and aspirin or aspirin alone. Indicative data were sourced from reviews of the literature supported by expert consultation to select the most appropriate value for the input parameters. The healthcare costs (£;2021) and quality adjusted life years resulting from each strategy were estimated and the incremental cost-effectiveness of testing calculated. Deterministic threshold analysis and probabilistic sensitivity analysis (PSA) was conducted to account for uncertainty in the parameter estimates. Results: The pharmacogenetic testing strategy generated 0.107 additional QALYs per patient tested and saved £512. Pharmacogenetic testing dominated current prescribing practice. The results were robust to extreme changes in key input variables. The PSA suggested that there was a 77% chance that pharmacogenetic testing would be cost-effective with a 62% chance it is cost-saving. Conclusions: A point-of-care pharmacogenetic test to guide prescription of clopidogrel for people who have experienced a stroke has the potential to provide a significant health gain by preventing secondary strokes and may save resources in the health system. This early economic analysis has also informed the direction for future research.


Introduction
In the United Kingdom there are over 100,000 hospital admissions for strokes annually and over a million people alive who have previously suffered at least one stroke 1,2 .People who have suffered a stroke have a shorter life expectancy and experience a lower quality of life, than people who have not suffered strokes 3,4 .Stroke survivors also have a high risk of experiencing subsequent strokes contributing to negative impacts on mortality and morbidity 5 .
To reduce the risk of secondary stroke, people who have had an ischaemic, non-cardioembolic stroke (hereafter shortened to 'stroke'), caused by a thrombus in one of the cerebral arteries, are prescribed anti-platelet medications to prevent further ischaemic events 6 .Current guidelines by the National Institute for Health and Care Excellence (NICE) recommend the use of 300mg of aspirin daily for 14 days followed by 75mg clopidogrel daily as the first-line treatment.If clopidogrel cannot be tolerated then modified release-dipyridamole, aspirin, or a combination of these medicines can be used depending on which option is best tolerated 7 .
Clopidogrel is a prodrug and is metabolised to its active form by an enzyme encoded by the CYP2C19 gene.In the UK, 20-30% of individuals have one or two loss-of-function (LoF) alleles which result in the poor metabolism of medicines such as clopidogrel which, in turn, lowers the antiplatelet effect of the medicine.A potential consequence of ineffective clopidogrel is an increased risk of a secondary stroke.A systematic review conducted by Pan et al. (2017) suggested that people with LoF alleles (*2, *3, and *8) receiving clopidogrel had a risk of secondary stroke approximately double to individuals without the LoF alleles (Risk ratio: 1.92, 12.0% vs 5.8%) 8 .Patients with the LoF alleles also had a higher risk of adverse vascular events compared with non-carriers 8 .
If a patient's CYP2C19 genotype were known prior to prescription, those carrying LoF alleles could be prescribed an alternative anti-platelet medication if required, such as modified-release dipyridamole or aspirin.Using a genetic test to tailor the prescription of medicines to patients is known as pharmacogenetics 9 .The goal of pharmacogenetic testing is to improve outcomes and make the best use of healthcare budgets by providing additional information for clinicians to select medicines that are effective with a lower risk of side effects 10 .
A prototype point-of-care test to detect CYP2C19 LoF allleles has recently been developed by Genedrive PLC 11 .A pointof-care tests enable a result to be delivered in a short timescale and/or near to the patient 12 .The CYP2C19 pharmacogenetic point-of-care test (hereafter called CYP2C19-test) uses a small platform which can be located within stroke units and produces results in around 40 minutes 13 .This allows clinicians rapid access to genetic information to guide prescription of antiplatelet medicines in patients who have been admitted following a stroke.
Before implementing the CYP2C19-test in the NHS, it is important to determine the potential impact of changing the prescribing pathway on healthcare costs and consequences (benefits and harms) for patients.It is important to understand whether the total health benefits achieved for patients who have experienced a stroke outweigh the health benefits potentially lost by other patients in the health system when resources are allocated to the provision of the CYP2C19-test.The aim of this study was to conduct an early cost-effectiveness analysis of the use of the CYP2C19-test to guide prescription of clopidogrel, modified release dipyridamole, and aspirin.Early economic evaluations do not seek to provide definitive evidence as to whether an intervention is cost-effective or not but an indication as to whether it may be cost-effective and what areas of uncertainty may require further data collection to inform reimbursement decisions.

Methods
This study used a decision-analytic model-based cost-effectiveness analysis with deterministic (one-way) and probabilistic sensitivity analysis.Table 1 provides an overview of the key aspects of the design of the analysis.

Model conceptualisation
The first stage in structuring the decision-analytic model that formed the basis for the cost-effectiveness analysis was to establish how the proposed CYP2C19-test and treatment pathways with subsequent events compared with current prescribing should be represented.This model conceptualisation was completed in line with published recommendations 14 .Several important decisions and assumptions were made during the model conceptualisation process to focus the cost-effectiveness analysis of CYP2C19-test for secondary prevention of stroke informed by a systematic review and expert input.
The systematic review used an electronic search of two databases (Embase: from inception to December, 2019; MEDLINE: from inception to December, 2019).A search strategy was developed combining search terms for clopidogrel, genotyping, and stroke (taken from (7)) with a published electronic search strategy comprising terms to identify economic evaluations 15 .After de-duplication this electronic search yielded 119 results.None of the identified studies met the inclusion criteria of having a primary focus on stroke patients.This result matches the findings of a published systematic review of economic evaluations of pharmacogenomics technologies in cardiovascular disease which identified 16 studies investigating CYP2C19 genotyping in acute coronary syndrome, but none in patients experiencing a stroke 16 .
In the absence of any existing published studies on CYP2C19 pharmacogenetic testing, a bespoke model structure was conceptualised, informed using expert input (two clinical geneticists and two consultants specialising in the treatment of stroke).An existing decision-analytic model-based cost-effectiveness analysis produced as part of a NICE technology appraisal of clopidogrel was used to inform the selection of relevant health states for people experiencing a stroke 19 .The model conceptualisation process focused on defining: the relevant jurisdiction; study perspective; time horizon; relevant pathways.
-The Relevant Jurisdiction: The UK NHS.
-Study Perspective: the analysis assumed a relevant health care service perspective because the analysis aimed to inform how to spend the NHS budget.In accordance with this perspective only items of resource use and associated costs accruing to the health services were included.Costs to individual patients were excluded from the analysis.
-Time horizon: a life-time horizon was assumed because patients who have had an initial stroke remain at risk of subsequent strokes for the remainder of their lives.
-Relevant pathways: the model needed to capture the initial testing pathways (CYP2C19-testing or current prescribing) to identify which treatment pathway is relevant for a patient and then follow subsequent events that may occur including up to death from stroke or other causes (see Table 1).

Model Structure
The conceptualised model was structured using a decision tree to represent the testing or current prescribing pathways, linked with a state transition Markov model, for the treatment pathway and subsequent health events related to stroke.A bespoke decision tree was created (see Figure 1) to represent the CYP2C19-test process compared with no testing.The events associated with treatment type were captured using a stated transition Markov model (see Figure 2) with five health states: no additional strokes; one additional stroke; two or more

Decision problem
What are the incremental costs and consequences, and the degree of uncertainty in these estimates, of using a point-of-care CYP2C19-test to guide treatment of the secondary prevention of stroke?Intervention CYP2C19 LoF alleles 2*, 3*, and 8* testing using point-of-care strategy in the hospital setting The results of the test are then used to guide treatment for the prevention of secondary stroke: -Patients with no LoF alleles receive clopidogrel.If clopidogrel not tolerated they can be switched to modified release dipyridamole and aspirin.Patients can also be switched to aspirin alone if modified release dipyridamole is not tolerated -Patients with one or more of the three LoF receive modified release dypridamole and aspirin.Patients can then be switched to aspirin alone if modified release dipyridamole is not tolerated.If dipyridamole is not tolerated then aspirin can be prescribed.

Comparator
No testing and all patients receive clopidogrel.Populating the model A process of sequential reviews of the literature supported by expert consultation was used to obtain the most appropriate value for the parameters.The process of populating the model is now described for each parameter needed for the decision tree and Markov model.

Probabilities
The main route of action for the CYP2C19-test is to change the probability of having a subsequent stroke for patients receiving clopidogrel.In the general population estimates suggest approximately 75% of people of European ethnicity can metabolise clopidogrel into its active component while 25% cannot 8 .The probability of a patient receiving clopidogrel having a stroke was calculated as the mean (average) probability of patients from either metaboliser status group having a stroke.
The CYP2C19-test provides information to separate the population into two distinct groups of potential (i) responders and (ii) non-responders.Patients who receive clopidogrel and metabolise clopidogrel are classed as 'responders' are assumed to have a lower risk of stroke than the average patient in the population.Patients who do not metabolise clopidogrel are assumed to be non-responders and have a higher risk of stroke.These patients are instead prescribed modified release dipyridamole and/or aspirin.It is assumed that patients switched to modified release dypiridamole and/or aspirin have similar levels of secondary stroke to clopidogrel in the following the current prescribing pathway strategy.The assumed reduction in risk of patients with no LOF alleles having a stroke when receiving clopidogrel is taken from systematic review and meta-analysis by Pan et al., 2017 8 .The proportion of patients with CYP2C19 loss-of-function alleles was also taken from this study The transition probabilities used to populate the treatment component of the model (Markov model) were based on evidence synthesised from four trials in NICE technology appraisal 90 19 .Relevant transition probabilities included the likelihood of one or more additional strokes when patients were receiving each medicine and the likelihood of vascular death.All-cause mortality was included based on the Office of National Statistics (ONS) mortality data 21 .The probability of patients experiencing the adverse events of minor bleeding events, major bleeding events, and worsening cardiac heart function were included based on estimates in the NICE Technology Appraisal 90 19 .

Resource use and costs
Table 2 describes the relevant items of resource use and costs consistent with the selected study perspective.In each health state in the state transition Markov model patients were assigned a cost of treatment based on the severity of the health state.
Patients were also assigned a one off cost when they had an additional stroke, died due to a stroke, or died due to other causes.It was necessary to use summary data from published sources 19 .The primary data source was based on 2010 costs, these were inflated to 2021 prices using the hospital & community health service index (2010 to 2015) and its replacement, the NHS cost inflation index (2015 to 2021) 23 .

Health consequences
In each health state in the state transition Markov model patients were assigned a health utility value to represent the severity of the health state.Utility values are a preference-based measure of health-related quality of life 24 and are needed to convert the estimated years of survival into a quality adjusted value to estimate quality-adjusted life-years (QALYs) by simple multiplication.
The utility values used by the evidence review group appraising the published NICE Technology Appraisal 90 report for the stroke health states were not publicly available.Therefore, the utility values used by Bristol-Myers Squibb in their submission to the NICE appraisal of clopidogrel were used.A utility weight of 0.61 was used for the no further stroke state.A decrement of 0.174 was applied to provide values for the one additional stroke and two or more additional strokes states (utility = 0.436).Disutilities associated with experiencing adverse events associated with each medicine were available from the evidence review group report.These were 0.0033 for a minor bleed, 0.1426 for a major bleed, and 0.0163 for worsening cardiac heart function.

Data analysis
A base case analysis was used to calculate the total costs and QALYs for a sample of 60,424 patients over a lifetime from the pre-defined starting age of 67 years.An incremental analysis was performed.This process involves first calculating the total costs, life-years and QALYs for each pathway.The incremental cost per QALY gained was then calculated using the formula: Where QALY 2 is the total QALYs of CYP2C19-testing pathway, QALY 1 is the total QALYs for the current prescribing pathway, and C 2 and C 1 are the respective total costs for each pathway.
Net health and monetary benefit per patient and in total across the patient population were calculated 25 .The net health benefit (NHB) and net monetary benefit (NMB) of the intervention represents the difference between the change in health for patients receiving the intervention and the change in health for patients in the health system resulting from a reallocation of resources to the intervention.The values of NHB and NMB are calculated using the following formulae Where ΔQALY is the incremental QALYs per patient arising from the use of the intervention, ΔCost is the incremental cost per patient from the use of the intervention, and λ is the NICE cost-effectiveness threshold, which in this study was taken to be £20,000 per QALY.The total NHB and NMB of the intervention was calculated by multiplying these figures by the population size.If NHB or NMB are positive, then the introduction of the intervention will increase the total health produced by the health system.
Two types of sensitivity analysis were used to determine the impact of uncertainty on the results of the cost-effectiveness analysis: deterministic threshold analysis and probabilistic sensitivity analysis.In the threshold analysis, key parameters were varied one at a time until the intervention became costineffective defined by exceeding a threshold of £20,000 per QALY gained.The maximum or minimum values for the risk ratio for patients receiving stratified clopidogrel, the CYP2C19test cost, and the tolerance of clopidogrel and modified release dipyridamole were determined and reported.
To incorporate uncertainty in all of the input parameters simultaneously, a PSAs was conducted.Distributions were assigned to each parameter, representing the pattern of uncertainty around the point estimate (see appendix 1).The model was re-run across 10,000 iterations, with new parameter values being randomly drawn from the distributions in each iteration.For each iteration, the estimated incremental costs, benefits, ICERs, NHB, and NMB, were recorded.An incremental cost-effectiveness plane was created to illustrate the distribution of the incremental health costs and consequences.The probability that the intervention would be cost-effective given the parameter uncertainty was calculated by determining the proportion of iterations which exhibited a positive NHB and NMB.

Results
This section first presents the results of the base case analysis and then each sensitivity analyses.

Base case analysis
The deterministic base case analysis suggested that using the CYP2C19-test was £512 less expensive compared with no testing per patient and generated 0.107 additional QALYs (see Table 3).This indicates that CYP2C19 testing was the dominant strategy as it was cheaper and more effective when compared with no testing.The testing strategy generated a NHB of 0.133 QALYs per patient tested or a NMB of £2,652 per patient.Across the assumed patient population of 60,424, this equates to a total NHB of 8,012 QALYs or a total NMB of £160,244,448.

Probabilistic Sensitivity Analysis
Table 3 shows the results from the PSA.These closely match the estimates expected values from the base case analysis but slight differences are observed because the PSA returns the average (mean) values from a random sample of runs of the model.The probabilistic sensitivity analysis suggested that CYP2C19 testing was £124 less expensive compared with no testing per patient and generated 0.092 additional QALYs (Table 3).This indicates that CYP2C19 testing was the dominant strategy as it was cheaper and more effective when compared with no testing.The testing strategy generated a net health benefit of 0.098 QALYs per patient and a net monetary benefit of £1,964 per patient.This equates to a total net health benefit of 5,934 QALYs and a total net monetary benefit of £118,672,736 for a population of 60,424.
Figure 3 shows the plot of the probabilistic sensitivity analysis results.The analysis suggested that there is a 77% chance that the intervention is cost-effective, at a threshold of £20,000 per QALY, given the uncertainty in the parameter estimates.In 62% of the iterations in the probabilistic sensitivity analysis the intervention was cost saving to the health system.

Threshold Analysis
The results of the cost-effectiveness analysis were robust to substantial changes in the parameters included in the threshold analysis.The risk ratio for the effectiveness of clopidogrel for patients with no loss-of-function compared with the un-stratified population could rise from 0.702 to as high as 0.94 before the intervention became cost-ineffective at a threshold of £20,000 per QALY.The cost of the CYP2C19-test could  rise from £60 to £1,357 before it became cost-ineffective.Tolerance to taking clopidogrel could fall to as low as 4% and CYP2C19-testing remain cost-effective.Changing the assumed tolerance to taking modified-release dipyridamole did not impact the absolute cost-effectiveness of the intervention at any assigned level.

Discussion
This study used an early model-based cost-effectiveness analysis to estimate the incremental costs and QALYs associated with introducing a CYP2C19-test to guide prescription of clopidogrel in patients who had experienced a stroke.The use of the CYP2C19-test was found to dominate current prescribing practice of universal clopidogrel prescription, improving patients'health and reducing costs to the health system.Sensitivity analysis confirmed that the intervention was highly likely to be cost-effective and likely to be cost saving even when accounting for uncertainty in the currently available evidence.
The results of this study suggest that the introduction of a CYP2C19-test could significantly reduce the number of secondary and tertiary strokes in the UK.When clopidogrel is offered universally, its clinical effectiveness is, on average, very similar to modified-release dipyridamole or aspirin.However, this observed effectiveness is a product of the mixing of two groups of people within a population; those with LoF alleles and those without.In this early analysis, patients without LoF alleles receiving clopidogrel had an annual risk of stroke of 2.79% while those with LoF alleles had a stroke risk of nearly 9%.Testing patients and switching the medicine for people with LoF alleles to modified-release dipyridamole or aspirin approximately halved their risk of having a stroke in the next year.
Studies investigating the cost-effectiveness of CYP2C19 testing have been published but these are not relevant to the UK setting.The results of these international studies echo those found in this early economic analysis.Studies using similar modelling strategies in China and Canada have shown that testing for CYP2C19 variants is highly likely to be cost-effective in clinical practice, albeit it with less chance of saving money for the health system 30,31 .A study investigating the costeffectiveness of testing in people of Asian heritage, where LoF alleles are more prevalent, found testing to be cost-effective when patients with LoF alleles were switched from clopidogrel to ticagrelor 32 .
This study presented an early model-based cost-effectiveness analysis of CYP2C19 testing to guide prescription of clopidogrel in the UK.Much of the key data used in this analysis was taken from the 2010 NICE appraisal of clopidogrel compared to modified-release dipyridamole and aspirin.This study may now be out of date and a full economic evaluation using an up-to-date search of the literature should be conducted to confirm the findings presented in this early analysis.In particular, it is likely that the cost of treating patients with stroke will have changed in the 12 years since this study due to changes in treatment and management pathways.However, the aim of this study was to create an early economic evaluation that could be further developed.The public availability of the model will allow researchers to update the input parameters with new information from trials of CYP2C19 testing in order to address the sources of uncertainty in the model.
This study was limited in the focus on the consequences of CYP2C19 testing on the prevention of further ischaemic strokes.However, the appropriate prescription of anti-platelet medications will also help reduce the risk of other cardiovascular diseases.As such, the benefits of testing may be underestimated.Furthermore, the model developed in this study only includes patients who suffered an ischemic stroke and did not capture those who had an index transient ischemic attack (TIA).
The risk of developing an ischemic stroke following a TIA is significantly increased compared to control populations.As such effective secondary prevention is critical in this population and, as with ischemic stroke, clopidogrel is a commonly used agent.By not including patients with TIA in this model, the benefits of CYP2C19 guided genotyping may have been underestimated.In addition, the CYP2C19 gene is implicated in the metabolism of a range of other drugs including proton pump inhibitors and selective serotonin uptake inhibitors 33,34 .If the information from the test can be stored in a patient's records for use by other clinicians in their prescribing decisions, then there would likely be additional health benefits at little additional cost.
It should be noted that this model made a number of assumptions, including the fact that all index strokes were treated as atherosclerotic in origin and indicated antiplatelet therapy.In reality, a proportion of strokes will be cardioembolic in nature and require anticoagulant treatment.This granularity was not captured in this modelling exercise and should be considered when interpreting the findings.Finally, this model includes two main therapeutic strategies, clopidogrel and modified-release dipyridamole plus aspirin.Ticagrelor has been investigated in several recent studies for patients with ischaemic stroke, and particularly CYP2C19 LoF allele carriers.If ticagrelor becomes used as part of routine practice, then this agent should be included in any future economic modelling 35,36 .

Conclusion
This early decision-analytic model-based analysis has produced indicative estimates of the relative cost-effectiveness of using a point-of-care CYP2C19 test to guide the treatment of the secondary prevention of stroke.This early analysis suggested that using a point-of-care CYP2C19 test to guide clopidogrel prescription would be likely to be cost-saving to the NHS.
There was considerable uncertainty in this indicative analysis that used currently available evidence.There is a clear need for a definitive trial to evaluate the clinical and cost-effectiveness of using a point-of-care CYP2C19 test to guide the treatment of the secondary prevention of stroke.The model is also limited and rather overly simplistic in a number of ways such as using estimates from many years ago and only considering point of care testing and this reviewer is not convinced that the overall costs of point of care testing using the device under consideration have been fully accounted for.On the other hand as acknowledged by the authors the overall cost of recurrent stroke and indeed other events is underestimated in this model.
The discussion should include acknowledgement of other potential ways of testing beyond point of care testing.
The assumption of using all ischaemic strokes as opposed to removing the proportion with cardioembolism inflates the overall population and consequently all subsequent global saving estimates for pharmacogenetics testing.This should be corrected.
In this context there is also an important typographical error in the last sentence of the paragraph "Estimating the patient population"...."all strokes included in this model would be treated as atherothrombotic (and treated with anticoagulants)" -the authors mean to say "antiplatelets".

Is the study design appropriate and is the work technically sound? Yes
Are sufficient details of methods and analysis provided to allow replication by others?

Kaavya Narasimhalu
National Neuroscience Institute, Singapore, Singapore In this study, the authors examine the cost effectiveness of point of care testing for three LOF alleles prior to the prescription of clopidogrel, with aspirin and dipyridamole or aspirin as alternative agents.This is an interesting study, but it can be improved if the comments below were addressed.
Major comments: It would be better for the authors to look at the LOF rates in the UK population that uses the NHS based on its current ethnic breakdown rather than to focus on those of European ancestry alone.More so considering the LOF rates are higher in non-European populations. 1.
Part of the issue with dipryridamole as an alternative is that it is poorly tolerated compared to aspirin and clopidogrel.Is there data available about the rates of discontinuation of dipyridamole within stroke patients?Could that information be incorporated into your models? 2.
Minor comments: I think there is a mistake in the last sentence of the estimating the patient population section.Do the authors mean to say."An assumption was made that all strokes included in 1.
this model would be treated as atherothrombotic (and treated with antithrombotics) as opposed to cardioembolic strokes (which would be treated with anticoagulants)."?
First paragraph of discussion section, I believe there is a space missing in between these words: "improving patients'health".
2. treatment to prevent further strokes in patients who have had an ischaemic stroke.Overall, this a well-written and useful article, although more detail is needed on the model/inputs.The model makes a lot of assumptions, which is to be expected from an early economic evaluation, but in which case we would also expect there to be more exploration of the sensitivity of the results to those assumptions.There was insufficient detail to know what the model inputs were, which may in part be because Appendix 1 was missing.Specific comments are given below:

Is
The final sentence of the abstract is the first point it becomes clear in the abstract that this is an early economic analysis.This needs to be clearer earlier on in the abstract, as to the aim of the model (as you state in last sentence of the Intro). 1.
Introduction 3 rd para: Give confidence interval alongside estimate of RR. 2.
Intro 3 rd Para "dipyridamole or aspirin": Clarify that this is in combination or if dipyridamole not tolerated then aspirin monotherapy.Same comment elsewhere where this is used.

3.
Intro 5 th para: We recognise that the focus here is on the cost-effectiveness of GeneDrive, but Genomadix is also available as a point-of-care test, and that should at least be mentioned in the Intro and Discussion.Laboratory-based testing is also an option, and again should be mentioned in discussion.

4.
Methods: Many of the model inputs are quite old now, and searches only run until Dec 2019.We don't propose updating them at this stage, as this would be a large undertaking, but this is a limitation, which should be acknowledged more clearly in the methods section and discussion (it is mentioned briefly in discussion).

5.
Table 1: Intervention/Comparator. Can you be clearer when switching occurs between treatments in the model?Do a proportion of patients switch immediately due to intolerance, or is there a switching proportion over time?

6.
Table 1 footnote: Reference is to the old NICE methods manual.Update to the 2022 version.7.
Methods_Model Structure subsection: There is just one sentence to refer the reader to the decision tree, and no further explanation given.Please add in a bit of text explaining the paths through the decision tree, especially the treatment switching pathways.Why is there an MRD+ASP path on the Universal clopidogrel tree.Due to intolerance to clopidogrel?What are the timings between the different steps along the pathway in the decision tree?Was the decision tree run for the first cycle i.e. 1 year?Or shorter?The reader can't be expected to know all this from the tree and text provided.Model Assumptions: Some of these are listed, but many are not.For example we think it is assumed that the test is 100% sensitive and 100% specific.Also it's assumed that all those with a test result will receive the appropriate treatment.

11.
Markov Model: Again the description of the model is very brief.Explain why you chose those particular states for the model (e.g.choice of one further and > 1 further stroke).How were bleeding events /ICH captured in the model?12. P.6 sentence before Populating the Model section: Is this a typo and should read "… treated as atherothrombotic (and treated with antiplatelets) as opposed to …".

13.
Populating the Model: A table that lists all model parameters and assumed input values and distributions for PSA, and evidence sources that they are based upon is missing.Later you refer to "Appendix 1", so maybe you have prepared this table, but we couldn't find this appendix.

14.
Populating the model_Probabilities subsection, end of first para: "It is assumed that patients switched to MRD+ASA or ASA have similar levels of secondary stroke to clopidogrel in the following the current prescribing pathway strategy".Are you assuming that in LoF patients on MRD+ASA they have the same probability of stroke as for NoLoF patients on Clopidogrel?This seems like a strong assumption...if it were the case, then why not give everyone MRD+ASA in the first place?Also, it sounds like you are assuming equal efficacy for MRD+ASA and ASA mono, which again doesn't seem reasonable to me.Can you clarify the assumptions and justify them, providing the evidence sources that support the assumptions?In general the second half of this paragraph is unclear.

15.
Resource Use and Costs section/Table 2: Comment on the test costs -is the cost listed per patient?Does that depend on the population size and life-time of the device?16.
Resource Use and Costs: Highlight here the limitation that you are using old costs for health states that will likely be out of date.It would be useful in the table to have the original price year of all the costs in table 2, as not all costs are derived from TA90.Although the costs were inflated to 2021 £ there may some costs which have changed.It would be of particular interest to have the price year used for the cost of the PoC tests.17.
Health consequences: I don't think you include carer disutility, which is relevant for stroke patients.This should be acknowledged as a limitation in discussion.

18.
Data Analysis section: This is fine but quite lengthy.If short on words, it would be better to use them to describe the model in more detail.

19.
Given the large number of parameters and assumptions, threshold analysis is only reported for 3 parameters: RR for clopidogrel in No LoF vs general population; test cost; and clopidogrel intolerance.It would be good to see something like a tornado plot showing the ICER ranges for all parameters in the model.Also, any other assumptions that could be varied in scenario analyses e.g.diagnostic accuracy, uptake of targeted treatment.

20.
Results: It is reported that "changing the assumed tolerance to taking modified-release dipyridamole did not impact the absolute cost-effectiveness of the intervention at any assigned level".These results may be in the missing Appendix 1, but if not, it would be useful to be able to see these numbers reported.

21.
Discussion 2 nd para: It's unclear which figures here are assumptions, which are based on evidence sources, and which are results from the model.E.g. effectiveness of clopidogrel universally has similar effectiveness on average to Dipyridamole.Is there a reference supporting that?Or is it a model output?

22.
Discussion para 3: Ticagrelor -should note that this is not currently available in the UK NHS.23.
Discussion para 6: Assumptions made.Also assumed perfect diagnostic test accuracy, 100% uptake of targeted treatment, % with LoF, relative risk of stroke on different treatments/LoF status, etc.

24.
Figure 3: There appears to be a strong correlation between the incremental benefits and incremental costs in the cost-effectiveness plane.It would be interesting to hear the authors' interpretation of this result and what model mechanism(s) may be driving it in the discussion.

25.
Conclusion, last sentence: I'm not sure I agree with this conclusion, as there has been a large trial of a guided therapy (ticagrelor vs clopidogrel) for secondary prevention of stroke in LoF patients (CHANCE-2, Wang et al. 2021), plus a trial of clopidogrel vs aspirin with results by LoF status (CHANCE).Together with sensitivity and specificity of the test, and the model, that may be sufficient.The problem is that there isn't any evidence on dipyridamole vs clopidogrel in a LoF population, i.e. as in the UK context.I think you could reword this to be more specific to the evidence needs for the UK context.

26.
Although this is only an early economic model you may want to complete the Drummond et al.CHEERS checklist for health economic evaluations in the Appendix.Many of the checklist items seem to be included in the manuscript.

Is the work clearly and accurately presented and does it cite the current literature? Yes
Is the study design appropriate and is the work technically sound?Yes

Are sufficient details of methods and analysis provided to allow replication by others? Partly
If applicable, is the statistical analysis and its interpretation appropriate?Not applicable Are all the source data underlying the results available to ensure full reproducibility?Partly Are the conclusions drawn adequately supported by the results?Yes

Figure 3 .
Figure 3. Cost-effectiveness plane showing the results of the probabilistic sensitivity analysis.

Figure 2 :
Figure 2: The health states are labelled as 'No further stroke' in the figure but called 'No additional stroke' in the text.It would be good to use the same language for consistency here.10.

Methods guide for technology appraisal 2013 18 . Figure 1. Decision tree of the CYP2C19 testing component. Figure 2. State transition Markov model.
If clopidogrel is not tolerated then the patient is switched to modified release dypridamole and aspirin.Patients can then be switched to aspirin alone if modified release dipyridamole is not toleratedPopulationA cohort of patients in the UK representing an estimate of the number of people who have their first stroke each year (n=71,091).The mean age of the cohort at the start of the analysis was set at 67 years.Data relevant to a European cohort of patients were used to populate the model.
additional strokes; vascular death; non-vascular death.Patients were assumed to enter the state transition Markov model in the 'no additional stroke' state.Patients then moved through the health-states based on a series of transition probabilities.

Estimating the patient population. The
population used in this study was an estimate of the number of people who experience a first ischaemic stroke in the UK in 2019.A report produced by Public Health England estimated that there were

Cost Effectiveness of Genotype- Guided Antiplatelet Therapy in Asian Ischemic Stroke Patients: Ticagrelor as an Alternative to Clopidogrel in Patients with CYP2C19 Loss of Function Mutations
Publisher Full Text 33.Rouby N El, Lima JJ, Johnson JA: . Clin Drug Investig.2020; 40(11): 1063-1070.PubMed Abstract |

plus aspirin versus clopidogrel plus aspirin for platelet reactivity in patients with minor stroke or transient ischaemic attack: open label, blinded endpoint, randomised controlled phase II trial
. BMJ. 2019; 365: l2211.PubMed Abstract | Publisher Full Text | Free Full Text

Yes If applicable, is the statistical analysis and its interpretation appropriate? Yes Are all the source data underlying the results available to ensure full reproducibility? Yes Are the conclusions drawn adequately supported by the results? Yes Competing Interests:
No competing interests were disclosed.

have read this submission and believe that I have an appropriate level of expertise to confirm that it is of an acceptable scientific standard.
This is an open access peer review report distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

the work clearly and accurately presented and does it cite the current literature? Yes Is the study design appropriate and is the work technically sound? Yes Are sufficient details of methods and analysis provided to allow replication by others? Yes If applicable, is the statistical analysis and its interpretation appropriate? Yes Are all the source data underlying the results available to ensure full reproducibility? Yes Are the conclusions drawn adequately supported by the results? Yes Competing Interests:
No competing interests were disclosed.