MJ - Decision on Manuscript ID BMJ

Transcription

1 MJ - Decision on Manuscript ID BMJ

2 Body: 02-Jun-2017 Dear Dr. López-López Manuscript ID BMJ entitled "Oral anticoagulants for prevention of stroke in atrial fibrillation: systematic review, network meta-analysis and cost-effectiveness analysis" Thank you for sending us your paper. We sent it for external peer review and discussed it at our manuscript committee meeting. We recognise its potential importance and relevance to general medical readers, but I am afraid that we have not yet been able to reach a final decision on it because several important aspects of the work still need clarifying. We hope very much that you will be willing and able to revise your paper as explained below in the report from the manuscript meeting, so that we will be in a better position to understand your study and decide whether the BMJ is the right journal for it. We are looking forward to reading the revised version and, we hope, reaching a decision. Kristina Fister kfister@bmj.com *** PLEASE NOTE: This is a two-step process. After clicking on the link, you will be directed to a webpage to confirm. *** 37f3a **Report from The BMJ s manuscript committee meeting** These comments are an attempt to summarise the discussions at the manuscript meeting. They are not an exact transcript. Members of the committee were: xelizabeth Loder (chair), Angela Wade (statistician), John Fletcher, Wim Weber, Sophie Cook, Tiago Villanueva, Rubin Minhas, Kristina Fišter Decision: Put points Detailed comments from the meeting: First, please revise your paper to respond to all of the comments by the reviewers. Their reports are available at the end of this letter, below. Please also respond to these additional comments by the committee: * A bibliographic database search of this topic returns many relevant and recent results. What is it about your paper that would make readers choose it over other meta analyses? In other words, what does your paper add over what is already available in recent literature? This does not seem to be the first network meta analysis or the first one with ranking.

3 * Much of the material is currently in the appendices, which makes the paper very readable but we felt you might have overdone it slightly. We agreed with one of the reviewers that more of the methods descriptions should be included in the main paper. * The economic analysis is likely to be out of date in a year as dabigatran comes off patent in We thought you might want to insulate the paper a bit by discussing how much price would need to drop to make an important difference to your conclusions and how much the makers of apaxiban would need to drop their price in reaction. * Are the messages in your paper entirely in tune with what was published in the HTA report? In your response please provide, point by point, your replies to the comments made by the reviewers and the editors, explaining how you have dealt with them in the paper. Comments from Reviewers Reviewer: 1 Recommendation: Comments: This is a very helpful paper for both patients and carers, not least as it clearly sets out that DOACs are cost effective as first line treatments - and in which order. The results are very clear and relevant. Although there is mention of the increased convenience for patients and carers more could be made of this. An additional couple of sentences recognising that the DOACs have significant benefits for patients and carers. The article mentions at line 25 the inconvenience and risk but it does read as if that is for docs rather than patients and carers. As a carer for someone with dementia who is on Apixaban the benefit of that over warfarin is significant, both in time taken for both of us if warfarin levels had to be monitored and the added distress for the patient. As such it could be pointed out that given the uncertainty around the cost estimates the additional benefits and convenience for older patients and their carers is a significant factor to be considered and one that should be factored into any head to head future trial. Additional Questions: Please enter your name: Jeni Bremner Job Title: Independent Consultant (Patient Reviewer) Institution: Phillips Kay Partnership Reimbursement for attending a symposium?: No A fee for speaking?: No A fee for organising education?: No Funds for research?: No Funds for a member of staff?: No

4 Fees for consulting?: No Have you in the past five years been employed by an organisation that may in any way gain or lose financially from the publication of this paper?: No Do you hold any stocks or shares in an organisation that may in any way gain or lose financially from the publication of this paper?: No If you have any competing interests <A HREF=' lists/declaration-competing-interests'target='_new'> (please see BMJ policy) </a>please declare them here: Reviewer: 2 Recommendation: Comments: Summary This is a systematic review, network meta-analysis (NMA) and cost-effectiveness analysis of all currently available and approved orally administered direct acting oral anticoagulants (DOACs), warfarin and antiplatelet agents for the prevention of thromboembolism in patients with atrial fibrillation (AF). The objective of this research was to compare, both directly and indirectly, the results of multiple randomised controlled trials which assessed the safety and efficacy of these antithrombotic agents in an attempt to identify those that are most clinically effective and cost-effective but don t provoke adverse events. The authors also aimed to rank the antithrombotic agents in order of most to least safe, efficacious and economical. The authors found that many of the DOACs reduced the risk of stroke or systemic embolism compared with warfarin but that some DOACs increased the risk of stroke when compared to others. Furthermore, the risk of all-cause mortality was lower with all DOACs than with warfarin and that most DOACs reduced the risk of major and intracranial bleeding compared with warfarin. The authors concluded that apixaban 5mg bd was ranked as the best intervention for most outcomes and was cost-effective when compared with warfarin. Review This was an ambitious undertaking by the authors and one that is highly relevant and currently topical. They produced a good manuscript that has clear clinical importance. It reads well, makes sense and has a clear message. An essential consideration in the management of AF is described that is the optimal (and most cost-effective) treatment for patients to prevent thromboembolism whilst minimising the risk of bleeding. Much research is currently being conducted in the post-approval era on the real-world benefits and risks of using direct oral anticoagulants (DOACs) for thromboprophylaxis and the current study adds to this growing body of knowledge. The findings of this work are important and increasingly so due to the exponential increase in the global prevalence of AF currently being observed. The importance of these data relate to patients, clinicians and healthcare systems. If further elucidated, clinical practice (including informing clinical guidelines which is a particular strength of network meta-analyses) could be altered accordingly. Originality This work adds to work that has been conducted thus far on the use of DOACs versus warfarin for the prevention of thromboembolism in patients with AF. A number of NMAs have been conducted to date (mostly in the last few years) to

5 understand the differences (and similarities) of all antithrombotic therapies and the risk of morbidity and mortality related to their use (or not). A quick PubMed search allowed me to identify: o Lin L et al. Cinical and safety outcomes of oral antithrombotics for stroke prevention in atrial fibrillation: a systematic review and network meta-analysis. JAMDA 2015;1103:e1-19. o Lip GYH et al. Relative efficacy and sagety of non-vitamin K oral anticoagulants for non-valvular atrial fibrillation: network meta-analysis comparing apixaban, dabigatran, rivaroxaban and edoxaban in three patient subgroups. In J Cardiol 2016;204: o Tawfik A et al. Systematic review and network meta-analysis of stroke prevention treatments in patients with atrial fibrillation. Clin Pharmacol 2016;8: o Tereshchenko LG et al. Comparative effectiveness of interventions for stroke prevention in atrial fibrillation: a network meta-analysis. J Am Heart Assoc 2016;5:e o Guo L et al. Comparative efficacy of clinical events prevention of five anticoagulants in patients with atrial fibrillation (a network meta-analysis). Am J Cardiol 2017;119: The authors state that this is the first study of DOACs for stroke prevention in AF that has provided a rank order for their use, in terms of both individual efficacy and safety outcomes and overall cost-effectiveness. However, a similar study was performed in 2012 looking at dabigatran, apixaban and rivaroxaban but not edoxaban and obviously not with data from RCTs conducted since then (Wells G et al. Safety, effectiveness, and cost-effectiveness of new oral anticoagulants compared with warfarin in preventing stroke and other cardiovascular events in patients with atrial fibrillation. CADTH Therapeutic Review 2012). Importance of work This work is important for all readers (clinicians, researchers, patients, educators and policymakers) in the thrombotic management of patients with AF and could definitely contribute to decision-making. Information regarding safety, efficacy and cost-effectiveness is provided and, therefore, there is importance for many. The content of medical advice given could change for clinicians and educators and patient quality-of-life may be positively impacted if one DOAC is prescribed over another due to a better risk:benefit ratio and safety profile. Furthermore, costs could be significantly reduced if one was recommended that was of lower cost than another. Therefore, the findings of this research have wide-reaching influence and significance. Scientific reliability Research Question The research question is clear when taking all relevant and available evidence into consideration, what is the most effective (in terms of action, safety and cost) antithrombotic therapy available for the management of patients with AF? The authors answered their research question appropriately, although I have some comments and queries to be addressed below. Overall design of study The overall design of the study presented is adequate to answer the research question posed. Participants studied Minimal data is provided about the participants studied. Only mean age, proportion of males and AF type was available to understand who was included in the cohorts of each individual RCT (Appendix 4). Therefore, readers are not provided with a clear picture as to who was actually studied. There are no details included on

6 important cohort features such as risk factor profiles, comorbidities or concomitant pharmacotherapies. Furthermore, the inclusion and exclusion criteria of all trials included in the analyses were not provided. The proportion of male patients included in each RCT varies quite significantly (i.e. ARISTOTLE-J had a cohort that was 83% male). Taking into consideration that some of the included RCTs were majority male, do the authors believe that the results of their work would be generalisable to females (particularly considering factors such as females being at greater risk of thromboembolism)? Background There are a couple of points to be added/clarified for improved reading: o Page 3, line 29: the authors should state what the therapeutic window of warfarin is where they discuss the optimal therapeutic range just so the reader has a reference when INR is discussed further on. o Page 3, line 31: the definition of DOACs that the authors give should read direct-acting (non-vitamin K antagonist) oral anticoagulants not directly acting. o Page 3, lines 33-34: factor X inhibitors should be factor Xa inhibitors. o Page 3, lines 55-56: the authors discuss that as no trials have directly compared different DOACs against each other, so it is difficult to determine which drug should be recommended as a first choice. Although this is true, I believe that the authors should be more circumspect when suggesting that it should be known which one drug is the best for all AF patients. There will always be physician and patient preferences and specific clinical reasons why one anticoagulant is chosen over another. In addition, there is now the added safety consideration with the advent of reversal agents for some but not all DOACs. o Page 4, lines 6-9: We conducted a systematic review, network meta-analysis (NMA) and cost-effectiveness analysis to compare DOACs for stroke prevention in AF against each other and against warfarin but they were actually comparing DOACs against each other and against warfarin and antiplatelet agents. I feel that there should be more in the Background about the current guidance and recommendations for DOAC use in the management of AF. Ultimately, clinical guidelines are what will be influenced by work such as the current study. Methods I don t feel that the methods have been described adequately enough. The authors refer the reader to a detailed report of the methods and results which is available elsewhere (page 4, lines 21-22). Furthermore, no reference to Appendix 1 (which gives some additional information of the methods) is provided within the body of the manuscript. The authors looked at Medline, Premedline, Embase and the Cochrane Library. Did the authors search all 6 databases in the Cochrane Library or just some of them? This should be made clearer for clarity of the methodology employed. Did the authors allocate the outcomes of interest according to the intention-to-treat principle? The authors excluded trials of warfarin with a target INR<2 (subtherapeutic) which is appropriate due to patients being at increased risk of thromboembolism. However, they did include trials of warfarin with a target INR of 3-4 (supratherapeutic) which puts patients at increased risk of bleeding. Why were these trials not also excluded? Both subtherapeutic and supratherapeutic INR ranges are outside of clinical practice guidelines for the management of AF. Given that there is significant evidence for some DOACs to increase the risk of gastrointestinal bleeding, the authors should have included this as a separate outcome of interest for analysis. Page 4, lines 34-35: we screened the studies included in previously published NMAs of DOACs against out eligibility criteria. What studies are these studies? References should be provided.

7 Page 4, line 38: we applied no restrictions on language. How many papers were in languages other than English? How were these translated? Due to the significant heterogeneity that exists between the cohorts, treatment duration and length of follow-up of each and every study included in these analyses, it is perhaps more appropriate to use a random-effects model for the NMA conducted. The authors could conduct both fixed-effects and random-effects NMAs and then assessment of model fit and choice of model could be made based on assessment of the deviance information criterion (DIC) and comparison of residual deviance to number of unconstrained data points (Cooper NJ et al. Addressing between-study heterogeneity and inconsistency in mixed treatment comparisons: application to stroke prevention treatments in individuals with non-rheumatic atrial fibrillation. Stat Med 2009;28: ). It was appropriate for the authors to conduct a meta-regression analysis to assess the dependence of effect estimates on the average time spent in the therapeutic range for those on warfarin therapy due to the large variability that exists between the cohorts of the included studies. Regarding the patient involvement in this study It is good that the authors engaged two patient group representatives but am not sure that including these representatives as co-applicants on the grant application for this work and as co-authors of this paper makes them entirely independent or detached from the findings of this work. Figure 1 this flow chart needs to be more explicit for complete understanding of the methods. o Records excluded (n = 2918) what are the reasons for the exclusions? These should be included for transparency. o Full-text articles excluded, n = 95 the numbers provided only add up to 92. What about the other 3? o Full-text articles assessed as eligible but not used in data extraction why were these not used in data extraction? o Reasons for exclusion at the point of full-text assessment are only provided for n = 178 (actually 175) but only 23 studies are included in the NMA which means that 200 full-text articles were excluded. What are the reasons for the other articles (n = 22) being excluded? This needs to be included for transparency. I have some concerns with the main assumptions and structure of the economic model used (details in Appendix 3) although I am by no means an expert in this field. o Having no evidence available for the effect of prior bleeds or intracranial haemorrhage on mortality, the authors assumed that having a history of bleeds or intracranial haemorrhage would have the same effect on future risk of death as a history of stroke. This has the potential to over- or underestimate the risk of bleeds on mortality. o The authors assumed that systemic embolism, myocardial infarction, transient ischaemic attack and clinically relevant bleeding had no long term management costs. This is not correct for all patients and is dependent on the severity of the event, patient characteristics and comorbidities. o Patients may switch from DOAC to warfarin or warfarin to no treatment after ischaemic stroke, bleed, systemic embolism or transient ischaemic attack. Patients who have had a thromboembolic event will most likely never have antithrombotic therapy discontinued. o Event rate and relative treatment effects are assumed not to vary with age and relative mortality rate in AF patients relative to the general population does not vary with age these two assumptions do not take into consideration patient characteristics or comorbidities (e.g. chronic kidney disease) and ignores the fact that AF has an independent effect on mortality (Stewart et al. A population-based study of the long-term risks associated with atrial fibrillation: 20-year follow-up of the Renfrew/Paisley study. Am J Med 2002;113:359-64).

8 Results The analyses presented certainly answer the research questions posed by the authors. However, I have a few questions and comments. I feel that some additional analyses should have been conducted to comprehensively understand the findings of this work and to identify potential differences in sub-groups, for example: o Males versus females o Individuals at different CHA2DS2-VASc scores o Those with chronic kidney disease versus those without o Older versus younger individuals More detail is required for the risk of bias section of the results particularly with the number of outcomes assessed and for this to be reinforced within the Limitations section of the Discussion. There is no comment about selective reporting. This is an essential element of any meta-analysis and additional details here will provide readers with information regarding the appropriateness of the comparison of the different RCTs. Regarding the risk of bias assessments presented in Appendix 7 what do the numbers refer to under the Interventions compared column? A legend needs to be included. Some of the results that the authors present as weak evidence are not statistically significant and, therefore, should not be presented as evidence for any associations. For example, on page 7, lines 7-20 all of the ORs and credible intervals presented are not significant. Therefore, the claims that there is weak evidence that the risk of MI is higher with dabigatran etc, that the risk of all-cause mortality was lower with all the DOAC interventions compared with warfarin INR 2-3 and that the risk of intracranial bleeding is higher with rivaroxaban etc (page 7, lines 35-37) are unfounded. It would be perhaps easier for readers to interpret the results of this work with the presentation of forest plots rather than tables of data. It was reassuring that the meta-regression analysis revealed no evidence of effect modification due to the mean time within therapeutic range for those on warfarin therapy. However, the fact that there were not enough data to analyse the influence of other effect modifiers (page 7, lines 56-57) is a disadvantage. With the cost-effectiveness results, the authors need to provide some context and clarity around the results for readers who are not familiar with cost-effectiveness analysis. For example, the timeframe for which total costs apply to need to be included for complete appreciation; explain what most cost-effective against willingness to pay per QALY threshold means. Table 1 some of the numbers do not add up within the table. I understand that this is a table of selected outcomes but I feel that readers should have the complete picture of events arising from each trial and know where the numbers for the all columns originate from. Furthermore, some of the all columns had less events than in one or two of the specific event columns. For example, all stroke for AFASAK II, ARISTOTLE, AVERROES, ENGAGE AF-TIMI 48, J-ROCKET AF and ROCKET AF has a lower number of events than the number of events within the stroke or systemic embolism column. A similar case exists for the all bleeding column. What is the reason for these discrepancies? Table 1 I suggest that the stroke or systemic embolism column be split into stroke and systemic embolism columns to allow the identification of differences with these two thromboembolic outcomes. Table 1 perhaps the authors should include proportions in parentheses after the event numbers to provide some context due to the differences in study size. Tables 2 & 3 these results could be presented as forest plots. Interpretation and conclusions

9 The interpretation of the findings are warranted and sufficiently derived from the data presented. However, I believe that only based on an analysis of this type, that a single answer on what therapy should be prescribed can t be conclusively derived (i.e. all AF patients should be prescribed apixaban). Other factors including individual patient risk factors, costs, lifestyle modifications, burden of monitoring and patient values and preferences should also be considered when anticoagulant treatment recommendations are made by physicians. For example, apixaban is taken BD this may be challenging for patients with AF and comorbid cognitive impairment. It may be more appropriate to prescribe a once daily therapy. Discussion on the potential impact of trial-specific biases on the effect estimates should also be included. More discussion should be included on how the results of this work would influence clinical practice guidelines for the management of AF. Furthermore, the influence of physician and patients preferences on decision-making in AF is an important consideration when making clinical decisions and needs discussion within the context of the findings of this work. References All references are up-to-date. However, the format of the citations is not consistent in the reference list (for example Thrombosis and haemostasis for reference 26 but Thrombosis & Haemostasis for reference 27). The authors need to ensure that all citations are complete and in BMJ format. References 11, 18 and 35 don t have complete citations. Reference 32 the senior author should be Lip GY not GY L. Page 4, lines 3-5: in addition, the effects of DOACs may have been overestimated in clinical trials because some individuals randomised to warfarin were not maintained within the therapeutic INR target (INR 2-3). (12-14). Reference 12 is for an actual clinical trial of dabigatran and a search takes you to a page on Clinicaltrials.gov. I don t think this is an appropriate reference the publication reference should be provided. In addition, references are all Japanese trials and, therefore, I am not convinced that the statement made by the authors is properly supported by the references provided given they are of studies only conducted in one population. Abstract/summary/key messages/what this paper adds The abstract is clear and well written and reflects accurately the findings of the analyses and the body of the manuscript. The Data sources paragraph states that the authors looked at Medline, Premedline, Embase and the Cochrane Library. Did the authors search all 6 databases in the Cochrane Library or just some of them? This should be made clearer for clarity of the methodology employed. Research designs should have the relevant checklist The authors provided a PRISMA checklist in the supplementary files and this accurately reflected the contents of the manuscript. However, the PRISMA for Network Meta-Analyses (PRISMA-NMA) extension (published in 2015) should have been used instead (Hutton B, Salanti G, Caldwell DM, et al. The PRISMA Extension Statement for Reporting of Systematic Reviews Incorporating Network Meta-analyses of Health Care Interventions: Checklist and Explanations. Ann Intern Med 2015;162:777-84). Additional administrative/minor considerations for the authors: - Some sentences need adjustment for clarity: o Page 2, line 41: Our NMA informs choice DOACs should be Our NMA informs choice of DOACs o Page 3, line 57:.costs of DOACs are offset efficacy benefits should be costs of DOACS are offset by efficacy benefits

10 o Page 4, line 3: reduced need for therapeutic monitoring or. The or needs to be deleted or the sentence completed. o Page 4, lines 20-21: the Centre for Reviews and Dissemination should be the University of York Centre for Reviews and Dissemination and Cochrane should be the Cochrane Collaboration. o Page 8, line 53: cost effectiveness analysis should be cost-effectiveness analysis. o Page 9, line12: for individuals with specific or contraindications or idiosyncratic reactions. The or needs to be removed after specific or the sentence completed. o Page 10, line 42: directly acting anticoagulants (DOAC) should be direct acting anticoagulants (DOACs). o Page 10, line 54: DOASs should be DOACs. - Ensure that all acronyms are defined at first mention (NHS EED). I would advise the Editors to request a major revision before considering this manuscript for publication within The BMJ. 17 April, 2017 Dr Jocasta Ball BBiomedSci (Hons), PhD National Health & Medical Research Council of Australia/National Heart Foundation of Australia Early Career Fellow Senior Research Fellow, Pre-Clinical Disease and Prevention Baker Heart and Diabetes Institute, Melbourne, Australia Additional Questions: Please enter your name: Dr Jocasta Ball Job Title: NHMRC/NHFA Early Career Fellow Senior Research Officer Institution: Baker Heart and Diabetes Institute Reimbursement for attending a symposium?: No A fee for speaking?: No A fee for organising education?: No Funds for research?: No Funds for a member of staff?: No Fees for consulting?: No Have you in the past five years been employed by an organisation that may in any way gain or lose financially from the publication of this paper?: No Do you hold any stocks or shares in an organisation that may in any way gain or lose financially from the publication of this paper?: No If you have any competing interests <A HREF=' lists/declaration-competing-interests'target='_new'> (please see BMJ policy) </a>please declare them here:

11 Reviewer: 3 Recommendation: Comments: The authors present a digest of a larger health technology assessment (HTA) report published recently by NHS, National Institute for Health Research. The paper is focused on efficacy and safety of oral anticoagulants in AF evaluated on published trial data identified through a systematic review. The authors represents results based on network meta analyses along with cost-effectiveness analyses based on a Markov multistate model and reported in terms of costs, QALY, net benefit etc. Overall the authors present a comprehensive analysis using appropriate methods, and the methods seem being used in an appropriate and competent way. It is reported that the HTA gives the full details on the analyses, especially the Markov model relies on wide range of assumptions and the specification of a large number of parameters. The paper does give a summary on these assumptions along with an overall description of the model, which is fine. The paper is well written and structured. My main concern goes to the extent on the conclusion of the indirect comparisons which also provides the information to the cost-effectiveness analyses. The authors have judged risk of bias in each trial involved, which, as far as I can see, focus on issues in each trial internally. A crucial assumption underlying the (indirect) meta-analyses is homogeneity and similarity between trials. I might have missed it, but I do not think this is considered in the discussion of the study limitations. It is well known that the pivotal trials differ on important points on conduct of trial, patient comorbidity, primary/secondary prevention, new users/switchers. These differences may be very critical towards assuming the study populations are homogeneius and that similarity can be assumed. A discussion of these issues can be found in Song F, Loke YK, Walsh T, Glenny A-M, Eastwood AJ, Altman DG. Methodological problems in the use of indirect comparisons for evaluating healthcare interventions: survey of published systematic reviews. BMJ. 2009;338: I suggest that these issues should be discussed and it should be evaluated how these may impact the level of conclusion that may be drawn on this type of analysis. Especially the advice for policy makers in the second paragraph in the discussion seems very firm to me. Minor points. Figure 1, please check numbers, they do not seem to count up correctly. First sentence in Background, for information a very recent paper on the prevalence of AF in UK has just appeared Lane DA, Skjøth F, Lip GYH, Larsen TB, Kotecha D. Temporal Trends in Incidence, Prevalence, and Mortality of Atrial Fibrillation in Primary Care. J Am Heart Assoc. 2017;6(5):1-10. doi: /jaha Last sentence on the same page (3) "Also, it remains..." it seems as some words are missing. Additional Questions: Please enter your name: Flemming Skjøth Job Title: Senior Biostatistician Institution: Aalborg Thrombosis Research Unit, Aalborg University Reimbursement for attending a symposium?: No

12 A fee for speaking?: No A fee for organising education?: No Funds for research?: No Funds for a member of staff?: No Fees for consulting?: Yes Have you in the past five years been employed by an organisation that may in any way gain or lose financially from the publication of this paper?: No Do you hold any stocks or shares in an organisation that may in any way gain or lose financially from the publication of this paper?: No If you have any competing interests <A HREF=' lists/declaration-competing-interests'target='_new'> (please see BMJ policy) </a>please declare them here: I have received consultancy fees from Bayer. Reviewer: 4 Recommendation: Comments: Overall, this is an interesting article on an important topic. The overall approach seems appropriate, and the conclusions seem appropriate given the results shown. There are some limitations that prevent me from recommending the manuscript for immediate publication. My area of expertise lies more in the area of cost-effectiveness analysis (CEA), so my comments will be focused on that aspect of this article. I would like to acknowledge the challenge to cover both a network met-analysis _and_ CEA in the same article. I think the authors did a pretty good job of keeping the CEA methods in the appendix to save space in the main paper. With that said, I would have liked to have seen some additional detail on the methods (probably added to the appendix). This would help me as a reviewer (and readers) to evaluate the quality of exactly what the authors did. And, I think some additional sensitivity analysis would have been valuable. Overall, it appears the authors did a good job on the CEA. A Markov modeling approach is a good approach for this. I like the Markov model diagram (Figure A3.2). The model structure is slightly different from many other published CEA's on stroke prevention for atrial fibrillation (e.g. including combinations of events, not including stroke severity). But, it appears as though it should be good to answer the questions posed. And, for example, although the model does NOT include stroke severity, presumably, the studies of stroke costs and utility already could have been an appropriately weighted average of those (although without further documentation of all parameters, I'm not sure)? Anyway, the article could benefit from additional documentation of all parameters used in the model, along with sources used for each parameter (all probabilities/rates/relative risks, costs, and utilities) and ranges used in sensitivity

13 analysis (and distributions for the probabilistic sensitivity analysis). This probably should go in a table (or tables) in the appendix. I like the ranges of results shown in Table 4 (so presumably, the authors have ranges for all the inputs). I like the probabilistic sensitivity analysis (cost-effectiveness acceptability curves) shown in Figure 5. But, I think additional sensitivity analyses could be instructive. Typically, one-way sensitivity analyses of all parameters (tornado diagram or similar table) would be useful to readers to understand what parameters have the largest impact, and also to give readers confidence in the validity of the model (e.g. if the price of apixaban goes up, the cost-effectiveness ratio becomes less favorable). I also think this sensitivity analysis could give important policy insights (e.g. when dabigatran goes off patent next year, its price could drop substantially, altering its cost-effectiveness relative to the others). I just have a couple of other minor comments: I think some of the numbers in the results on page 8 might have been reversed (or the columns in Table 4 are mis-labeled). I think dabigatran should be 6365 pounds of incremental net benefit and rivaroxaban should be 5279 pounds. Is the assumption that a myocardial infarction does not have any long-term management costs? For example, it appears the Bayer analysis (Single Technology Appraisal (STA) of Rivaroxaban (Xarelto )) did have an annual post-mi cost (admittedly, a relatively small amount of 140 pounds in comparison to the 5278 pounds for the event itself). Thank you. Additional Questions: Please enter your name: David William Hutton Job Title: Associate Professor Institution: University of Michigan Reimbursement for attending a symposium?: No A fee for speaking?: No A fee for organising education?: No Funds for research?: No Funds for a member of staff?: No Fees for consulting?: No Have you in the past five years been employed by an organisation that may in any way gain or lose financially from the publication of this paper?: No Do you hold any stocks or shares in an organisation that may in any way gain or lose financially from the publication of this paper?: No If you have any competing interests <A HREF='

14 lists/declaration-competing-interests'target='_new'> (please see BMJ policy) </a>please declare them here: