BMJ Open Is Committed to Open Peer Review. As Part of This Commitment We Make the Peer Review History of Every Article We Publish Publicly Available

BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from

BMJ Open is committed to open peer review. As part of this commitment we make the peer review history of every article we publish publicly available.

When an article is published we post the peer reviewers’ comments and the authors’ responses online. We also post the versions of the paper that were used during peer review. These are the versions that the peer review comments apply to.

The versions of the paper that follow are the versions that were submitted during the peer review process. They are not the versions of record or the final published versions. They should not be cited or distributed as the published version of this manuscript.

BMJ Open is an open access journal and the full, final, typeset and author-corrected version of record of the manuscript is available on our site with no access controls, subscription charges or pay-per-view fees (http://bmjopen.bmj.com).

If you have any questions on BMJ Open’s open peer review process please email [email protected] http://bmjopen.bmj.com/ on September 27, 2021 by guest. Protected copyright. BMJ Open BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from

Patent foramen ovale closure, antiplatelet therapy or anticoagulation in patients with patent foramen ovale and cryptogenic stroke: a systematic review and network meta- analysis incorporating complementary external evidence

ForJournal: peerBMJ Open review only

Manuscript ID bmjopen-2018-023761

Article Type: Research

Date Submitted by the Author: 23-Apr-2018

Complete List of Authors: Mir, Hassan; McMaster University, Medicine; Siemieniuk, Reed; Mcmaster University, Clinical Epidemiology and Biostatistics; University of Toronto, Department of Medicine Ge, Long; The First Clinical Medical College of Lanzhou University, Foroutan, Farid; University Health Network, Medicine/Cardiology Fralick, Michael; University of Toronto, Eliot Phillison Clinician Scientist Training Program, Department of Medicine Syed, Talha; McMaster University Department of Medicine, Medicine Lopes, Luciane; Universidade de Sorocaba Kuijpers, Ton; Nederlands Huisartsen Genootschap, Guideline development and research Mas, Jean-Louis; Sainte-Anne Hospital, Neurology http://bmjopen.bmj.com/ Vandvik, Per; Norwegian Knowledge Centre for the Health Services, Agoritsas, Thomas; University Hospitals of Geneva, Division of General Internal Medicine & Division of Clinical Epidemiology; McMaster University Faculty of Health Sciences, Department of Health Research Methods, Evidence, and Impact Guyatt, Gordon; Mcmaster University, Clinical Epidemiology and Biostatistics

Cryptogenic stroke, Patent foramen ovale, Anticoagulation < on September 27, 2021 by guest. Protected copyright. Keywords: HAEMATOLOGY, Antiplatelet, PFO Closure

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 1 of 101 BMJ Open

Patent foramen ovale closure, antiplatelet therapy or anticoagulation in patients with patent BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 2 3 foramen ovale and cryptogenic stroke: a systematic review and network meta-analysis 4 5 incorporating complementary external evidence 6 7 Hassan Mir1, internal medicine and cardiology physician, Reed A.C. Siemieniuk1,2, PhD student and 8 3 4,5 6 9 internal medicine physician, Long Ge , PhD student, Farid Foroutan , PHD student, Michael Fralick , 10 1 11 research fellow and internal medicine physician, Talha Syed , internal medicine and cardiology 12 physician, Luciane Cruz Lopes7, clinical pharmacologist, Ton Kuijpers8, clinical epidemiologist, Jean- 13 14 Louis Mas9, professor, Per O. Vandvik10,11, associate professor, Thomas Agoritsas1,12, assistant 15 1 16 professor, Gordon H. GuyattFor, distinguished peer professorreview only 17 18 1. Department of Health Research Methods, Evidence, and Impact, McMaster University, 1280 Main St West, Hamilton, Ontario, Canada L8S 4L8 19 2. Department of Medicine, University of Toronto, Toronto, Ontario, Canada 20 3. Evidence-Based Medicine Center, School of Basic Medical Sciences, Lanzhou University, Lanzhou 730000, China 21 4. Department of Clinical Epidemiology and Biostatistics, McMaster University, 1280 Main St West, Hamilton, Ontario, Canada L8S 4L8 22 5. Heart Failure/Transplant Program, Toronto General Hospital, University Health Network, Toronto, Ontario, Canada 23 6. Eliot Phillipson Clinician Scientist Training Program, Department of Medicine, University of Toronto, Toronto, Canada, M5G 2C4 24 7. University of Sorocaba, UNISO, Sorocaba, Sao Paulo, Brazil 25 8. Department of guideline development and research, Dutch College of General Practitioners, Utrecht, The Netherlands 26 9. Sainte-Anne Hospital, Department of Neurology, 1 rue Cabanis, 75015 Paris. 27 10. Norwegian Institute of Public Health, Oslo, Norway 28 11. Department of Medicine, Innlandet Hospital Trust - division Gjøvik, Norway 29 12. Division General Internal Medicine & Division of Clinical Epidemiology, University Hospitals of Geneva, Rue Gabrielle-Perret-Gentil 4, CH-1211, Geneva, Switzerland 30 31

32 http://bmjopen.bmj.com/ 33 34 35 36 37 38 39

Abstract BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 2 3 Objective: To examine the relative impact of three management options in patients less than 60 4 5 years old with cryptogenic stroke and a patent foramen ovale (PFO): PFO closure plus antiplatelet 6 7 therapy, antiplatelet therapy alone, and anticoagulation alone. 8 9 Design: Systematic review and network meta-analysis (NMA) supported by complementary external 10 11 12 evidence 13 14 Data sources: Medline, EMBASE, and Cochrane CENTRAL. 15 16 Study selection: RandomisedFor controlled peer trials review (RCTs) addressing only PFO closure and/or medical 17 18 19 therapies in patients with PFO and cryptogenic stroke. 20 21 Review methods: We conducted an NMA complemented with external evidence and rated certainty 22 23 of evidence using the GRADE system. 24 25 26 Results: Ten RCTs in eight studies proved eligible (n=4416). Seven RCTs (n = 3913) addressed 27 28 PFO closure versus medical therapy. Of these, 3 (n=1257) addressed PFO closure versus antiplatelet 29 30 therapy, 3 (n=2303) addressed PFO closure versus mixed antiplatelet and anticoagulation therapies, 31 32 and 1 (n= 353) addressed PFO closure versus anticoagulation. The remaining three RCTs (n= 503) http://bmjopen.bmj.com/ 33 34 35 addressed anticoagulant versus antiplatelet therapy. PFO closure versus antiplatelet therapy 36 37 probably results in substantial reduction in ischaemic stroke recurrence (risk difference per 1000 38 39 patients over 5 years [RD]: -87, 95% credible interval [CrI] -100 to -33; moderate certainty).

40 on September 27, 2021 by guest. Protected copyright. 41 42 Compared with anticoagulation, PFO closure may confer little or no difference in ischaemic stroke 43 44 recurrence (low certainty) but probably has a lower risk of major bleeding (RD -20, 95% Crl -27 to -2, 45 46 moderate certainty). Relative to either medical therapy, PFO closure probably increases the risk of 47 48 49 persistent atrial fibrillation (RD 18, CI +5 to +56, moderate certainty) and device-related adverse 50 51 events (RD +36, 95% CI +23 to +50, high certainty). Anticoagulation, compared to antiplatelet 52 53 therapy, may reduce the risk of ischaemic stroke recurrence (RD -71, 95% CrI -100 to +17, low 54 55 certainty), but probably increases the risk of major bleeding (RD +12, CrI -5 to +65, moderate 56 57 58 certainty). 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 3 of 101 BMJ Open

32 http://bmjopen.bmj.com/ 33 34 35 36 37 38 39

INTRODUCTION: BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 2 3 Stroke is one of the leading causes of death worldwide and, for those who survive, often results in 4 5 substantial disability. Central to the management of patients with stroke is preventing additional 6 7 ischaemic events; the optimal management depends on the underlying aetiology. 8 9 In up to 40% of patients with an acute ischaemic stroke, investigations fail to identify a clear cause – 10 11 12 the Trial of Org 10172 in Acute Stroke Treatment (TOAST) criteria classify these strokes as “of 13 14 undetermined aetiology” or “cryptogenic” (1). Patients who have had a cryptogenic stroke are less 15 16 likely to have classical riskFor factors peer for stroke (e.g.,review hypertension, only hyperlipidaemia, diabetes mellitus) 17 18 19 and are more likely to have a patent foramen ovale (PFO) compared to patients with a stroke from a 20 21 known aetiology (2, 3). 22 23 A PFO can allow a systemic venous thrombus to travel directly into the systemic arterial circulation 24 25 26 instead of the pulmonary venous circulation – a phenomenon known as a paradoxical embolism (4, 27 28 5). The importance of this phenomenon as a cause of stroke is, however, controversial. Other 29 30 potential mechanisms of PFO-associated stroke include intracardiac thrombus formed in the tunnel of 31 32 the PFO, on the surface of the atrial septal aneurysm, or left atrial appendage (5). http://bmjopen.bmj.com/ 33 34 35 Until recently, randomised controlled trials (RCTs) had failed to definitively establish that closing a 36 37 PFO reduces the risk of subsequent stroke (6-8). Meta-analyses including earlier trials offered 38 39 conflicting conclusions (9-12). Current clinical practice guidelines recommend against PFO closure for

40 on September 27, 2021 by guest. Protected copyright. 41 42 most patients who have had a cryptogenic stroke (13). 43 44 In September 2017, two new trials (14, 15) and the long-term follow-up of a previously reported trial 45 46 (16) addressed the utility of PFO closure for patients less than 60 years old with cryptogenic stroke. 47 48 49 Evidence from recent meta-analyses that incorporate data from these RCTs suggests that PFO 50 51 closure may prevent stroke. However, despite evidence that anticoagulation is more effective than 52 53 antiplatelet therapy at preventing stasis-related venous or intracardiac thrombi, these meta-analyses 54 55 make the assumption that antiplatelet and anticoagulation therapies are similarly effective (17). 56 57 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 5 of 101 BMJ Open

This systematic review is part of the BMJ Rapid Recommendations project, a collaborative effort from BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 2 3 the MAGIC research and innovation programme (www.magicproject.org) and the BMJ (18). The aim 4 5 of the project is to respond to new potentially practice changing evidence and provide trustworthy 6 7 practice guidelines in a timely manner. In light of the new evidence for PFO closure and because 8 9 10 other reviews do not report clinically relevant comparisons (e.g., PFO closure vs. antiplatelet therapy), 11 12 we performed a systematic review to inform the parallel guideline published in a multi-layered 13 14 electronic format on bmj.com and MAGICapp 15 16 For peer review only 17 (https://www.magicapp.org/goto/guideline/JjXYAL/section/j79pvn) (19). As requested by the Rapid 18 19 Recommendations panel, in the absence of sufficient direct evidence to fully inform 20 21 recommendations, we conducted complementary analyses using indirect evidence. 22 23 24 25 METHODS 26 27 28 Protocol 29 30 31

32 The study protocol was registered with PROSPERO: CRD42017081567 (20). http://bmjopen.bmj.com/ 33 34 35 36 37 38 39

Guideline panel, patient involvement, and choice of outcomes BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 2 According to the BMJ Rapid Recommendations process (18), a multi-professional guideline panel 3 4 provided critical oversight to the systematic review and identified populations, subgroups, and 5 6 7 outcomes of interest. The BMJ Rapid Recommendation panel, and in particular three people who had 8 9 experienced a cryptogenic stroke (two of whom had undergone PFO closure), assessed the relative 10 11 importance of the outcomes (19). The guideline panel submitted a list of possible outcomes and then 12 13 14 rated the importance of each outcome on a scale from 1 (least important) to 9 (most important). For 15 16 items with scores of 4, theFor panel consideredpeer furtherreview and made onlya final decision by consensus, with 17 18 special consideration of the views of the patient panellists. The patients stressed the importance of 19 20 21 several outcomes that other panel members had identified and uniquely highlighted the importance of 22 23 detailed information about the device or procedure related adverse events. 24 25 26 Search strategy 27 28 29 30 A search from a previous systematic review that we judged as comprehensive included research 31

32 articles indexed before May 1, 2013 (12). We updated that review with a search of Medline, Medline http://bmjopen.bmj.com/ 33 34 in-process, EMBASE, and Cochrane CENTRAL from 1 January 2012 to 15 October 2017 using a 35 36 37 combination of keywords and MeSH terms for “patent foramen ovale” AND “stroke”, using the 38 39 sensitive search filters for therapeutic interventions developed by the Health Information Research

40 on September 27, 2021 by guest. Protected copyright. 41 Unit at McMaster University (Appendix 1). (21, 22) We also searched all references from included 42 43 44 studies and all of the studies citing the included studies on Google Scholar. 45 46 47 Study selection 48 49 50 51 We included RCTs addressing the relative impact of PFO closure versus antiplatelet therapy versus 52 53 anticoagulation in patients, and anticoagulation versus antiplatelet therapy with PFO and cryptogenic 54 55 stroke. We also included prior rigorous systematic reviews addressing antiplatelet versus 56 57 anticoagulant therapy in patients with deep venous thrombosis (DVT), atrial fibrillation, and heart 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 7 of 101 BMJ Open

failure. Reviewers screened titles and abstracts in duplicate and subsequently assessed eligibility BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 2 3 from the full text for all possible eligible articles. In comparing PFO closure to antiplatelet therapy for 4 5 the outcome of stroke, we used only RCTs in which at least 80% of medical therapy patients received 6 7 antiplatelet rather than anticoagulation. 8 9 10 11 Data collection 12 13 14 Two reviewers independently abstracted data from the included RCTs and resolved conflicts by 15 16 discussion. We searchedFor the clinical peer trial registration review page on clinicaltrials.govonly for additional outcome 17 18 19 data and emailed the corresponding authors of the included RCTs requesting outcome data when it 20 21 was missing or could not, as presented in the public domain, be included in meta-analytic estimates. 22 23 24 The patient-important outcomes we considered included ischaemic stroke, death, major bleeding, 25 26 27 persistent atrial fibrillation or flutter, transient or paroxysmal atrial fibrillation, device or procedure 28 29 related adverse events, pulmonary embolism, transient ischaemic attacks (TIAs), and systemic 30 31 embolism. For serious device or procedure-related complications, two independent clinician reviewers

32 http://bmjopen.bmj.com/ 33 34 categorized the specific serious adverse events in the primary studies as probably related or 35 36 unrelated to the device or implantation procedure; reviewers resolved disagreement with, if 37 38 necessary, input from a third reviewer. We used the definitions in the original studies for all outcomes 39

40 on September 27, 2021 by guest. Protected copyright. 41 except for persistent atrial fibrillation; our classification included the definition used in the original 42 43 study but also included patients who underwent a cardioversion. We also abstracted key outcomes 44 45 from the external evidence systematic reviews. 46 47 48 49 Risk of bias and certainty of evidence 50 51 We assessed risk of bias in duplicate using a modified Cochrane tool in which we used response 52 53 options of “definitely or probably yes” (assigned a low risk of bias) and “definitely or probably no” 54 55 56 (assigned a high risk of bias), an approach that has proved valid (23, 24). Specifically, we assessed 57 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 8 of 101

random sequence generation; concealment of treatment allocation; and blinding of participants, BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 2 3 caregivers, and outcome assessors. Reviewers resolved conflicts through discussion. 4 5 6 We followed the GRADE approach to rate the quality of evidence derived from pair-wise and network 7 8 meta-analyses (NMA) as well as from external evidence (25, 26). Direct evidence from RCTs starts at 9 10 11 high quality and may be rated down based on risk of bias, indirectness, imprecision, inconsistency, 12 13 and publication bias. For network estimates, we rated the quality of evidence in each of the direct, 14 15 indirect, and network estimates (26). The rating of indirect estimates starts at the lowest rating of the 16 For peer review only 17 18 two pairwise estimates that contribute to the indirect estimate and can be rated down further for 19 20 intransitivity. If direct and indirect estimates contributed similar power to the network estimate, then 21 22 we used the higher rating. The network estimates were further rated down for incoherence between 23 24 direct and indirect evidence. For analyses that incorporated external evidence, we rated down the 25 26 27 quality of evidence for indirectness (down one level for major bleeding and two levels for ischaemic 28 29 stroke) (27). Review authors and the linked Rapid Recommendation panel members came to 30 31 consensus regarding certainty of evidence ratings.

32 http://bmjopen.bmj.com/ 33 34 35 Statistical Analysis 36 37 38 We analysed patients in groups to which they were randomised. For both pairwise and NMA, we 39

40 on September 27, 2021 by guest. Protected copyright. 41 report pooled odds ratios (ORs); for estimates from NMA we report 95% credible intervals (CrI) from 42 43 the Bayesian analysis and for direct estimates 95% confidence intervals (CI). We present pooled risk 44 45 differences (RD) and their certainty/confidence intervals for all comparisons, applying relative risks to 46 47 baseline risk estimates when there were adequate numbers of events, and direct calculation of risk 48 49 50 differences when events were very infrequent. In reporting absolute events, we standardized absolute 51 52 estimates to a rate per 1,000 patients followed for 5 years assuming proportional event occurrence (in 53 54 other words, for a reported follow-up of 3 years we would multiply by 5/3). Atrial fibrillation and device 55 56 57 or procedure-related adverse events were, however, very likely to have occurred soon after the 58 59 procedure and we therefore report events per 1,000 patients as reported in the studies. 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 9 of 101 BMJ Open

The PFO closure arm was used for the baseline risk to allow consistency across comparisons BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 2 3 involving PFO closure versus antiplatelet agents and anticoagulants. Baseline risk estimates 4 5 represent the median risk of the outcome in the six RCTs of PFO closure. For comparisons of 6 7 antiplatelet agents and anticoagulants, the antiplatelet arm provided the baseline risk to estimate the 8 9 absolute effect estimate in the anticoagulation arm. The baseline risk for the antiplatelet arm 10 11 12 calculated in Table 2 considering the baseline risk in the PFO closure arm (10%) proved similar to the 13 14 median of three studies that included an antiplatelet arm (9%) and maintained consistency with the 15 16 PFO closure comparisons.For peer review only 17 18 19 20 Network meta-analysis 21 22 To compare effects of alternative medical therapies, for ischaemic stroke, recurrent TIA, death, major 23 24 bleeding, and systemic embolism, we conducted a NMA of RCTs within a Bayesian hierarchical fixed 25 26 27 effect framework with non-informative priors and adjusted for correlation between effects in the multi- 28 29 arm trial. We generated posterior samples using Markov Chain Monte-Carlo (MCMC) simulation 30 31 technique running the analysis in three parallel chains. We used 100,000 burn-in simulations to allow

32 http://bmjopen.bmj.com/ 33 34 convergence and then a further 500,000 simulations to produce the outputs. We assessed model 35 36 convergence using the Gelman and Rubin test, accepting a threshold of <1.05 and used the node- 37 38 splitting approach for the assessment of loop inconsistency (28, 29). We calculated direct estimates 39

40 on September 27, 2021 by guest. Protected copyright. 41 of absolute effects using the modified Dias model, incorporating lines into the Dias model (30). We 42 43 performed the network meta-analyses with R version 3.4.1 (R Core Team. 2016. Vienna, Austria: R 44 45 Foundation for Statistical Computing) using the gemtc library (31). 46 47 48 49 Pairwise meta-analysis 50 51 52 As there is no reason, for the outcomes of atrial fibrillation and device or procedure-related adverse 53 54 events, to expect any difference whether PFO closure is compared to anticoagulation or antiplatelet 55 56 57 therapy, for these outcomes we conducted a pairwise meta-analysis of PFO closure versus either 58 59 medical therapy. We used the random-effects Mantel-Haenszel (M-H) method to estimate relative risk 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 10 of 101

(RRs) and 95% confidence intervals (CIs), using the M-H risk difference approach for studies that did BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 2 3 not have any events in either arm. For outcomes with event rates less than 1%, we used the Peto 4 5 fixed effects method to estimate OR and 95% CI. For all outcomes, we examined statistical 6 7 heterogeneity among studies using the I2 statistic and visual inspection of the forest plots (27). 8 9 10 11 Meta-regression analysis 12 13 14 We hypothesised that anticoagulation may be more effective than antiplatelet therapy for preventing 15 16 stroke. Thus, we expectedFor that the peer effect of PFO review closure vs. medical only therapy would be largest in the 17 18 19 RCTs that had proportionally more patients in the medical therapy group that were treated with 20 21 antiplatelet therapy rather than anticoagulation. To explore this, we performed a random-effects meta- 22 23 regression of PFO closure vs. medical therapy, with the proportion of patients in the medical therapy 24 25 26 arm who received anticoagulation as the independent variable. We used random-effects meta- 27 28 regression with the studies were weighed by the inverse of their variance. 29 30 31 Modelling with external data from other indications

32 http://bmjopen.bmj.com/ 33 34 35 Because few patients were randomised to PFO closure versus anticoagulation (N=353) or 36 37 anticoagulation versus antiplatelet agents (N=405) and events were infrequent, the estimates of effect 38 39 of PFO closure versus anticoagulation were extremely imprecise. For instance, for ischaemic stroke

40 on September 27, 2021 by guest. Protected copyright. 41 42 the credible interval (Crl) around the relative effect included a relative reduction in events of 90% to a 43 44 more than quadrupling of events (NMA OR 0.51, CrI 0.09 to 4.46). We considered this information 45 46 essentially uninformative, as did the Rapid Recommendations panel. We nevertheless considered 47 48 49 the option of anticoagulation crucial to the decision, and therefore sought strategies for more 50 51 informative estimates of effect. 52 53 54 For ischaemic stroke, the most likely mechanism of cryptogenic stroke associated with PFO is stasis- 55 56 57 related thrombi, most notably paradoxical venous thromboembolism (VTE). The panel and systematic 58 59 review team therefore deduced that the relative effects of anticoagulation versus antiplatelet therapy 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 11 of 101 BMJ Open

in the secondary prevention of VTE could provide credible estimates of effect for patients with BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 2 3 cryptogenic stroke and PFO. For major bleeding, we deduced that the relative effects of VKA- 4 5 anticoagulation and low-dose ASA are similar to those seen in other common indications including 6 7 VTE, heart failure, and atrial fibrillation and could therefore provide credible estimates of relative 8 9 bleeding risk of antiplatelet agents (with or without PFO) and anticoagulants. 10 11 12 13 Thus, the panel and systematic review team ultimately decided to use this evidence to provide 14 15 indirect evidence of estimates of effect for three key outcomes (ischaemic stroke, major bleeding and 16 For peer review only 17 18 pulmonary embolism). We therefore performed a secondary analysis that also included RCT 19 20 evidence from other conditions (VTE, heart failure, and atrial fibrillation). 21 22 23 For ischaemic stroke, we digitised Kaplan-Meier curves and extracted patient-level time-to-event data 24 25 26 (32). For illustrative purposes we present the individual patient data by treatment arm (PFO closure 27 28 plus antiplatelet therapy, antiplatelet therapy alone, anticoagulation, and medical therapy left to the 29 30 discretion of the treating physician). Each enrolled patient contributed equally to the survival curves. 31

32 http://bmjopen.bmj.com/ 33 34 Subgroup analysis 35 36 37 We planned subgroup analyses based on the echocardiography-determined shunt size and the 38 39 presence of an atrial septal aneurysm. We hypothesised that patients with a larger shunt size will

40 on September 27, 2021 by guest. Protected copyright. 41 42 have a larger benefit with PFO closure than patients with a small shunt size and that patients with an 43 44 atrial septal aneurysm will have a larger benefit with PFO closure than patients without an atrial septal 45 46 aneurysm (33). 47 48 49 50 RESULTS 51 52 53 Our systematic literature search yielded 772 potentially relevant publications; ten trials presented in 54 55 56 eight reports ultimately proved eligible (Figure 1) (6, 8, 14-16, 34-36). Seven eligible multicenter 57 58 RCTs published in 2013 or later comparing PFO closure plus antiplatelet therapy versus medical 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 12 of 101

therapy enrolled 3913 patients (range per RCT: 120 to 980 patients) (Table 1). The CLOSE trial BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 2 3 included three separate randomized trials (14). In the first trial, patients eligible for all options were 4 5 randomised to PFO closure (N = 173), antiplatelet therapy (N=171), or anticoagulation with a vitamin 6 7 K antagonist or direct oral anticoagulant (N=180; 93% of which were a VKA). In the second trial, 8 9 patient’s ineligible for anticoagulation were randomised to PFO closure (N=65) or antiplatelet therapy 10 11 12 (N=64). In the third trial, patients with contraindications to PFO closure were randomised to 13 14 antiplatelet therapy (n=7) or anticoagulation (n=3). Of the remaining eligible RCTs, one specified that 15 16 all patients in the group whoFor were peer not randomised review to PFO closure only were treated with antiplatelet 17 18 19 therapy (N=664) (15). Four RCTs allowed the treating physician to choose between antiplatelet 20 21 and/or anticoagulation therapy for the patients randomised to medical therapy: a minority of these 22 23 patients (range 20% to 34%) were prescribed anticoagulation at the time of discharge from initial 24 25 26 hospital admission (6, 8, 16, 36). The last three studies compared antiplatelet therapy to 27 28 anticoagulation (N=503) (14, 34, 35). One study reported a composite endpoint for ischaemic stroke 29 30 and TIA; we therefore could not use the data in our meta-analyses of stroke (N=44) (34). However, 31 32 we were able to use data from this RCT in analyses of death and major bleeding. Another RCT http://bmjopen.bmj.com/ 33 34 35 reported only a composite outcome of stroke and death (N=98) (35), with non-significant results 36 37 suggesting a possible reduction with anticoagulation compared to antiplatelet therapy (HR 0.52, CI 38 39 0.16 to 1.67, p-value 0.28).

40 on September 27, 2021 by guest. Protected copyright. 41 42 43 We contacted the authors of three RCTs for additional outcome data (14, 34, 35); the lead author of 44 45 the CLOSE study (JLM) provided additional unpublished data (14). The authors of the two remaining 46 47 studies did not respond. 48 49 50 51 We identified a recent rigorous systematic review of 12 trials and 11 999 participants for secondary 52 53 prevention of VTE (17) and a recent rigorous systematic review of 15 trials and 4982 participants 54 55 56 across indications that provided a trustworthy estimate for an increase in major bleeds (37); both 57 58 reviews compared anticoagulant to antiplatelet therapy. 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 13 of 101 BMJ Open

Table 1 presents the characteristics of the eligible studies. In most, the mean age was approximately BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 2 3 45 years, approximately 50 to 60% were male, and approximately 50%, or more, had moderate or 4 5 larger shunt. The median follow-up was 3.9 years with a range of 1.2 to 5.9 years. 6 7 8 PFO closure plus antiplatelet therapy versus antiplatelet therapy alone 9 10 11 12 Table 2 provides estimates of effect and certainty of evidence for all patient-important outcomes, 13 14 including, as presented subsequently, complications of PFO-closure. PFO closure plus antiplatelet 15 16 therapy reduced the risk Forof ischaemic peer stroke comparedreview to antiplatelet only therapy alone (NMA OR 0.12, 17 18 19 0.04 to 0.27; risk difference per 1000 patient-years followed for 5 years [RD]: -87, 95% confidence 20 21 interval [CI] -100 to -33; moderate certainty) (Appendix 4, Figure 1). 22 23 24 Along with the three RCTs in which all medical patients received only antiplatelet therapy (14, 15, 36), 25 26 27 we included a meta-regression model data from an additional 5 RCTs that enrolled patients in which 28 29 more than 20% of medically treated patients received anticoagulation (Figure 2) (6, 8, 14, 16). The 30 31 reduction in stroke with PFO closure decreased as the proportion of patients receiving anticoagulation

32 http://bmjopen.bmj.com/ 33 34 in the medical therapy arm increased (p = 0.036). 35 36 37 Figure 3 presents a Kaplan-Meier curve of individual patient time-to-event data for each of the 38 39 randomised interventions, including PFO closure plus antiplatelet versus antiplatelet therapy alone

40 on September 27, 2021 by guest. Protected copyright. 41 42 and is again consistent with large relative reductions in hazard with PFO closure versus antiplatelet 43 44 therapy and smaller benefits of PFO closure when compared to mixed populations of patients 45 46 receiving anticoagulants and antiplatelet agents. 47 48 49 50 PFO closure did not appear to reduce the risk of TIA compared to antiplatelet therapy alone: NMA OR 51 52 0.82, CrI 0.32, 2.11; RD -6, -34 to +15; moderate certainty (Table 2; Appendix 4). 53 54 55 56 Systemic emboli were rare: there were 5 events in one study (15), and none in another (14); we 57 58 therefore calculated the risk difference directly. There was no important difference between PFO 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 14 of 101

closure and antiplatelet therapy: NMA OR 0.83, CrI 0.12, 7.25; RD -1, -6 to +4 per 1,000 patients; BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 2 3 high certainty (Table 2; Appendix 4, Figure 3). 4 5 6 Death occurred infrequently: in the PFO closure group a median of 9 per 1000 patients died over a 7 8 period of 5 years. The relative effect estimate for mortality was very imprecise, and the absolute 9 10 11 estimates included a small increase or decrease in deaths: NMA OR 3.28, CrI 0.2 to 174.22; RD +6, 12 13 CI -3 to +9; moderate certainty (Table 2; Appendix 4, Figure 4). 14 15 16 The risk of major bleedingFor did not peerdiffer significantly review between theonly PFO closure plus antiplatelet 17 18 19 therapy and antiplatelet therapy alone groups: NMA OR 0.48, CI 0.20 to 1.12; RD -7, CI -14 to +1, 20 21 moderate certainty (Table 2; Appendix 4, Figure 5). Pulmonary embolism was rare and no more 22 23 frequent in the PFO closure group (3/679, 0.4%) than the antiplatelet therapy alone group (1/458, 24 25 26 0.2%). No additional NMA evidence was available for pulmonary embolism because none of the 27 28 RCTs that included an anticoagulation arm reported pulmonary emboli; thus, we present a pair-wise 29 30 estimate for PFO closure vs. antiplatelet therapy alone: Peto OR 1.01, CI 0.09, 11.21; RD +0, CI -5 to 31

32 http://bmjopen.bmj.com/ 33 +48, high certainty (Table 2; Appendix 4, Figure 6). 34 35 36 PFO closure plus antiplatelet therapy versus anticoagulation 37 38 39 Table 3 provides estimates of effect and quality of evidence for all patient-important outcomes

40 on September 27, 2021 by guest. Protected copyright. 41 42 including, as presented below, complications of PFO-closure. One RCT randomised patients to PFO 43 44 closure plus antiplatelet therapy (n=173) or anticoagulation (n=180) (14). In the NMA, compared with 45 46 anticoagulation, PFO closure did not significantly reduce the risk of ischaemic stroke, but the CrI 47 48 49 around the relative effect was very wide (NMA OR 0.44, CrI 0.08 to 3.83; RD: -16, 95% CI -29 to +10; 50 51 low certainty). Figure 3 presents a Kaplan-Meier curve of individual patient time-to-event data for 52 53 each of the randomised interventions, including PFO closure plus antiplatelet versus anticoagulation 54 55 56 alone. 57 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 15 of 101 BMJ Open

Results from the meta-regression analysis described previously suggest that PFO closure was less BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 2 3 effective relative to medical therapy as the proportion of patients receiving anticoagulation increased 4 5 (Figure 2). 6 7 8 To complement the above analyses, we added the effect estimate from a systematic review of RCTs 9 10 11 comparing low-dose ASA to anticoagulation with a VKA for the secondary prevention of VTE to the 12 13 network meta-analysis (17). Similar to the primary analysis, the analysis with the external evidence 14 15 failed to show a difference between PFO closure and anticoagulation for ischaemic stroke, with a 16 For peer review only 17 18 point estimate substantially closer to 1.0: NMA OR 0.93, CI 0.31 to 2.76; RD: -2, -20 to +47; low 19 20 certainty (Table 3; Appendix 4, Figure 1). For this complementary analysis, we rated down our 21 22 certainty in the evidence to low because of very serious indirectness because this analysis relies on 23 24 the assumption that the relative effect of anticoagulation versus antiplatelet therapy is the same for 25 26 27 secondary prevention VTE as it is for secondary prevention of cryptogenic stroke in patients with a 28 29 PFO. 30 31

32 http://bmjopen.bmj.com/ 33 There was no important difference in TIA: NMA OR 1.27, CrI 0.40 to 4.52; RD +6, CI -22 to +22; 34 35 moderate certainty (Table 3; Appendix 4, Figure 2). No patient experienced a systemic embolism in 36 37 either arm of the RCT that included an anticoagulation arm (Appendix 4, Figure 3). 38 39

40 on September 27, 2021 by guest. Protected copyright. 41 Few patients randomised to PFO closure plus antiplatelet therapy experienced a pulmonary 42 43 embolism (median risk 5 per 1000 patients over 5 years). The RCT that randomised patients to PFO 44 45 closure versus anticoagulation did not measure pulmonary emboli (14). Using external evidence from 46 47 a systematic review of RCTs that compared ASA to anticoagulation with a VKA for secondary 48 49 50 prevention of VTE in the NMA, PFO closure plus antiplatelet therapy probably has a higher risk of 51 52 pulmonary embolism than anticoagulation: RD +4, CI +1 to +13; moderate certainty (Table 3; 53 54 Appendix 4, Figure 4). 55 56 57 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 16 of 101

Patients randomised to PFO closure plus antiplatelet therapy had a lower risk of major bleeding than BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 2 3 those randomised to anticoagulation: NMA OR 0.26, CI 0.07 to 0.82; RD -20, CI -27 to -2, moderate 4 5 certainty (Table 3; Appendix 4, Figure 5). Adding the effect estimate for major bleeding of low-dose 6 7 ASA compared to anticoagulation with a VKA from a systematic review of RCTs enrolling patients 8 9 with atrial fibrillation, VTE, or heart failure (37) to the NMA, resulted in an almost identical point 10 11 12 estimate, but a narrower confidence interval: OR 0.28, CrI 0.13 to 0.55; RD -17, CI -21 to -11; 13 14 moderate certainty (Table 3). 15 16 For peer review only 17 18 There was no difference in death between PFO closure and anticoagulation (RD -4, -13 to +9, 19 20 moderate certainty) (Table 3). 21 22 23 Complications of PFO closure 24 25 26 27 PFO closure, when compared to medical therapy, probably substantially increases the risk of atrial 28 29 fibrillation (including transient, persistent, and paroxysmal): six RCTs with 3560 patients; RR 4.50, 30 31 95% confidence interval (CI) 2.35 to 8.60; risk difference per 1000 patients (RD) +39, CI +15 to +84;

32 http://bmjopen.bmj.com/ 33 34 moderate certainty (Table 3; Appendix 6, Figure 1). All six studies assessed persistent atrial 35 36 fibrillation at a follow-up of 2.0 to 5.9 years. PFO closure probably increases the risk of persistent 37 38 atrial fibrillation: RR 4.84, CI 1.91 to 12.26; RD +18, CI +5 to +56; moderate certainty (Table 3; 39

40 on September 27, 2021 by guest. Protected copyright. 41 Appendix 6, Figure 2). PFO closure also probably increases the risk of transient atrial fibrillation: RR 42 43 3.76, CI 1.74 to 8.10; RD +12, CI +3 to +31; moderate certainty (Table 3; Appendix 6, Figure 3). The 44 45 studies were limited by risk of bias because they did not report sufficient detail to know how persistent 46 47 atrial fibrillation was detected (Tables 2 and 3). 48 49 50 51 Serious device or procedure-related adverse events occurred in approximately 36 per 1000 patients, 52 53 CI 23 to 50, high certainty (Table 2; Appendix 6, Figure 4). The most common device or procedure- 54 55 56 related complications were vascular complications (1%), conduction abnormalities (1%), device 57 58 dislocation (0.7%), and device thrombosis (0.5%). Although serious, air embolism (0.4%), cardiac 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 17 of 101 BMJ Open

tamponade (0.3%), and cardiac perforation (0.2%) were rare (Tables 2 and 3). No deaths were BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 2 3 attributed to the procedure or the device placement (Tables 2 and 3). 4 5 6 Anticoagulation versus antiplatelet therapy 7 8 9 10 Table 4 provides estimates of effect and certainty of evidence for all patient-important outcomes 11 12 comparing the two medical therapies. There may be fewer ischaemic strokes in patients randomised 13 14 to anticoagulation than to antiplatelet therapy, but the CI included no effect: NMA OR 0.27, CrI 0.03 to 15 16 1.21; RD: -71, CI -100 toFor +17; low peer certainty (Table review 4; Appendix only 7, Figure 1). Figure 3 presents a 17 18 19 Kaplan Meier curve of individual patient time-to-event data for each of the randomized interventions, 20 21 including anticoagulation versus antiplatelet therapy suggesting a lower rate of stroke in patients 22 23 receiving anticoagulant than antiplatelet therapy. 24 25 26 27 The complementary model including external evidence from a systematic of RCTs comparing low- 28 29 dose ASA to anticoagulation with a VKA for secondary prevention of VTE yielded a somewhat larger 30 31 reduction in strokes with anticoagulants and substantially narrowed the CI: OR 0.17, CrI 0.08 to 0.36;

32 http://bmjopen.bmj.com/ 33 34 RD -81, CI -91 to -62; low certainty (Table 4) (37). 35 36 37 There may be little or no difference in risk of TIA between anticoagulation and antiplatelet therapy 38 39 (NMA OR 0.65, CrI 0.19 to 1.98; RD -12, CI -34 to +24; low certainty (Table 4, Appendix 7, Figure 2).

40 on September 27, 2021 by guest. Protected copyright. 41 42 There were no reported systemic emboli, although it is not clear what the monitoring and 43 44 ascertainment processes were. 45 46 47 48 Our analysis suggested more major bleeding events among patients randomised to anticoagulation 49 50 than to antiplatelet therapy, but the CI included no effect: NMA OR 1.90, CI 0.68 to 5.53; RD +12, -5 51 52 to +65; moderate certainty (Table 4; Appendix 7, Figure 3). Including external evidence from a 53 54 systematic review of RCTs comparing low-dose ASA to anticoagulation with a VKA in patients with 55 56 57 atrial fibrillation, VTE, and heart failure yielded a very similar point estimate and substantially 58 59 narrowed the CI: OR 1.77, CI 1.36 to 2.31; RD +11, CI +5 to +18; moderate certainty (Table 4) (37). 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 18 of 101

The RCT that included an anticoagulation arm did not report pulmonary emboli; the risk of pulmonary BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 2 3 embolism was low in other RCTs that included an antiplatelet arm (median 5 per 1000 patient-years 4 5 over 5 years). External evidence from a systematic review of secondary prevention of VTE suggests 6 7 that the risk of pulmonary embolism is probably lower with anticoagulation than antiplatelet therapy: 8 9 RD -4, CI -5 to -3; moderate certainty (17). 10 11 12 13 Deaths were extremely rare and there was no appreciable difference between the anticoagulation 14 15 and antiplatelet therapy: RD +10, CI -3 to +357, low certainty (Table 4, Appendix 7, Figure 4). 16 For peer review only 17 18 19 Subgroup Analyses 20 21 22 Data proved insufficient to perform the prespecified subgroup analyses when anticoagulation and 23 24 antiplatelets were assumed to have different effects, as we had prespecified. We performed one post- 25 26 27 hoc meta-regression analysis to explore the effect of PFO shunt size on the relative effectiveness of 28 29 PFO closure compared to medical therapy. Five RCTs reported ischaemic stroke and the proportion 30 31 of patients with a moderate or large shunt vs. small shunt, all with slightly different definitions. PFO

32 http://bmjopen.bmj.com/ 33 34 closure was more effective in the RCTs that enrolled a higher proportion of patients with a moderate 35 36 or large shunt (interaction p=0.047) (Appendix 2, Figure 2). However, this is confounded by the fact 37 38 that the RCTs that enrolled a higher proportion of patients with moderate or large PFO shunts also 39

40 on September 27, 2021 by guest. Protected copyright. 41 had a larger proportion of patients with antiplatelet agents rather than anticoagulation (Appendix 2 42 43 Figure 3). When both possible effect modifiers were included in the model neither was statistically 44 45 significant: p=0.44 for proportion with moderate or large shunt and p=0.61 for proportion receiving 46 47 anticoagulation; however, this model is severely underpowered with two effect modifiers and only five 48 49 50 studies. 51 52 53 DISCUSSION 54 55 56 57 We found moderate certainty evidence that in patients less than 60 years old with a PFO and 58 59 cryptogenic stroke, PFO closure plus antiplatelet therapy results in a substantial reduction in the risk 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 19 of 101 BMJ Open

of recurrent ischaemic stroke compared to antiplatelet therapy alone by approximately 8.7% over 5 BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 2 3 years, but increases the risk of persistent atrial fibrillation (approximately 1.8%) and device-related 4 5 adverse events (approximately 3.6%) in the first year after procedure (Table 2). 6 7 8 Although the evidence regarding anticoagulation versus alternatives is of low quality and comes in 9 10 11 large part from indirect evidence, it is possible that anticoagulation and PFO closure may achieve a 12 13 similar reduction in the risk of recurrent ischaemic stroke – and thus appreciably fewer strokes than 14 15 antiplatelet agents alone. Anticoagulation, compared to either PFO closure plus antiplatelet therapy or 16 For peer review only 17 18 antiplatelet therapy alone probably increases the risk of major bleeding by approximately 2% over 5 19 20 years and probably reduces the risk of pulmonary embolism by approximately 0.4%. There does not 21 22 appear to be an important difference in the risk of death or in the risk of systemic emboli between any 23 24 of the interventions. 25 26 27 28 Our results provide further strong support for the hypothesis that stasis-related paradoxical venous 29 30 thromboemboli and/or intra-cardiac thrombi cause a large proportion of cryptogenic strokes in 31

32 http://bmjopen.bmj.com/ 33 younger patients with PFO. Given that is the case, it is plausible, though uncertain, that 34 35 anticoagulation has a similar relative benefit versus antiplatelet agents in preventing cryptogenic 36 37 strokes in patients with PFO as it does in preventing recurrent VTE. Our results, including the meta- 38 39 regression examining comparisons of PFO closure versus studies in which varying proportion of

40 on September 27, 2021 by guest. Protected copyright. 41 42 patients received anticoagulants (Figure 2) support this inference. 43 44 45 Only one of the RCTs included an arm in which all patients were offered anticoagulation or PFO 46 47 closure; as a result, the direct evidence includes relatively few patients and very few events. 48 49 50 Moreover, the evidence from studies of VTE is very indirect. Thus, we categorised the evidence 51 52 regarding anticoagulation versus PFO closure, and versus antiplatelet therapy, in preventing stroke 53 54 as low certainty (Tables 3 and 4). Resolving the uncertainty that currently bedevils decision-making 55 56 57 will require larger RCTs that compare anticoagulation to PFO closure. 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 20 of 101

A recent systematic review suggested that PFO closure may be more effective in patients with larger BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 2 3 shunt (38). We show that this subgroup analysis has low credibility because it is confounded by the 4 5 anticoagulation in the patients randomised to medical therapy in RCTs with a higher proportion of 6 7 patients with smaller shunts. PFO closure may be less effective relative to medical therapy when 8 9 patients receive anticoagulation and/or have smaller shunts. The results of the NMA, in addition to 10 11 12 indirect evidence from patients with a VTE, suggest that the subgroup effect by type of medical 13 14 therapy may be the most credible explanation. 15 16 For peer review only 17 18 PFO closure comes with infrequent but important risks: an approximately 3% to 4% chance of serious 19 20 device or procedure-related complications and an approximately 4% increased incidence of atrial 21 22 fibrillation, of which approximately half is persistent. Anticoagulation with VKAs carries a 1% to 2% 23 24 increase in major bleeding over 5 years in this patient population. The risk of major bleeding may be 25 26 27 marginally lower with some direct acting oral anticoagulants than with VKAs (17). Patients must also 28 29 consider the important practical implications of taking anticoagulants, including limitations to diet and 30 31 activities with an appreciable risk of trauma, possible drug interactions and, with VKAs, the need for

32 http://bmjopen.bmj.com/ 33 34 repeat laboratory testing. 35 36 37 Strengths and limitations: 38 39 Our study has several strengths. In contrast with recent systematic reviews summarising the recent

40 on September 27, 2021 by guest. Protected copyright. 41 42 RCTs on PFO closure (38-41); we separately addressed the three management alternatives of PFO 43 44 closure plus antiplatelet therapy, anticoagulation alone, and antiplatelet therapy alone. In doing so, 45 46 we applied several analytical approaches, including network meta-analysis, meta-regression, and 47 48 49 individual patient analyses from survival curves, as well as judicious use of external evidence, to bring 50 51 to bear all the relevant evidence for decision-making. In addition, we used the GRADE approach to 52 53 assess the certainty of evidence informing the estimates in the study. Our review also includes one 54 55 56 recently published study not included in prior reviews. We thus summarised all of the highest quality 57 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 21 of 101 BMJ Open

available evidence, providing optimal insight into the comparative effects of the alternative BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 2 3 management strategies for patients with PFO who have experienced a cryptogenic stroke. 4 5 The results are limited primarily by the available evidence. Of the ten RCTs that we included in this 6 7 review, three compared PFO closure to a medical therapy arm that was not predefined – the key 8 9 choice between antiplatelet and anticoagulation was left to the discretion of the physician and patient. 10 11 12 As corroborated in our evidence synthesis, there are probably important differences between 13 14 anticoagulation and antiplatelet therapy. As a result, these three RCTs cannot directly inform the 15 16 clinical decisions in whichFor antiplatelet peer therapy review and anticoagulation only are two distinct choices, and we 17 18 19 could not include them in many of the analyses bearing on the choice between PFO closure and 20 21 antiplatelet therapy. Our meta-regression analysis that included these studies suggested an inverse 22 23 relationship between stroke reduction and proportion of patients in the medical therapy arms 24 25 26 prescribed anticoagulation, thus providing support for the benefit of anticoagulants versus antiplatelet 27 28 therapy (Figure 2). 29 30 The NMA also has limitations: the network is sparsely populated, it includes patients with no 31 32 contraindications to any of three management strategies, those with patients with contraindications to http://bmjopen.bmj.com/ 33 34 35 anticoagulants, and those with contraindications to PFO. Nevertheless, we found no suggestion of 36 37 incoherence between direct and indirect estimates in the NMA and suggest that it provides the best 38 39 estimates available.

40 on September 27, 2021 by guest. Protected copyright. 41 42 The applicability of these findings to older patient populations, particularly those over 60 years of age, 43 44 and those with traditional cerebrovascular risk factors such as exposure to diabetes, hypertension, 45 46 and hyperlipidaemia, is uncertain. The proportion of patients with a cryptogenic stroke caused by 47 48 49 paradoxical emboli rather than large vessel atheroembolism or intracardiac thrombus almost certainly 50 51 decreases with age. Thus, we expect that the benefits of PFO closure would be smaller in older 52 53 patients and the harms (e.g. atrial fibrillation) greater. 54 55 56 57 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 22 of 101

CONCLUSIONS BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 2 3 In patients less than 60 years old with cryptogenic stroke and a PFO, PFO closure plus antiplatelet 4 5 therapy probably confers a substantial reduction in the risk of ischaemic stroke recurrence compared 6 7 to antiplatelet therapy alone, with a modest risk of persistent atrial fibrillation or flutter and serious 8 9 device or procedure-related complications. Based on low certainty evidence, anticoagulation may 10 11 12 confer a similar reduction in the risk of ischaemic stroke compared to PFO closure, but likely confers 13 14 a modest increased risk of major bleeding. Future RCTs comparing anticoagulation to PFO closure 15 16 would substantially reduceFor uncertainty peer for several review critical outcomes. only 17 18 19 20 21 Acknowledgement: We thank members of the BMJ Rapid Recommendations panel for critical 22 23 feedback on outcome selection, subgroup selection, GRADE judgments, and manuscript feedback. 24 25 26 We thank Rachel Couban for helping develop the search strategy. 27 28 Contributors: GHG and RAS conceived the study idea. HM performed the literature search and data 29 30 analysis. HM, RAS, TA, GHG, POV interpreted the data analysis. HM, RAS, and GHG wrote the first 31 32 draft of the manuscript. TS, LL, and MF acquired data and judged risk of bias in the studies. FF http://bmjopen.bmj.com/ 33 34 35 extracted patient level survival data from Kaplan-Meier curves. LG provided statistical advice. RAS, 36 37 GHG, TA, POV, TK, JLM, MF critically reviewed the manuscript. HM had full access to all of the data 38 39 in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

40 on September 27, 2021 by guest. Protected copyright. 41 42 HM is guarantor. 43 44 Funding: None 45 46 Competing interests: All authors have completed the BMJ Rapid Recommendations disclosure 47 48 49 form, which asks about any possible financial, intellectual, and professional conflicts of interest. JLM 50 51 has received consulting honoraria from Bayer, Bristol Meyer Squibb, Boehringer-Ingelheim, Daichii- 52 53 Sankyo, GECKO, and Servier. HM, RAS, TK, and GHG are also panel members of the linked Rapid 54 55 Recommendation guideline panel. 56 57 58 Ethical approval: Not required 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 23 of 101 BMJ Open

Data sharing: No additional data available BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 2 3 Transparency declaration: The lead author (HM) affirms that the manuscript is an honest, accurate, 4 5 and transparent account of the study being reported; that no important aspects of the study have 6 7 been omitted; and that any discrepancies from the study as planned have been explained. 8 9 10 11 12 13 14 15 16 For peer review only 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31

32 http://bmjopen.bmj.com/ 33 34 35 36 37 38 39

Legend: BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 2 3 Table 1: Characteristics of patients in eligible studies 4 5 Table 2 Summary of findings of PFO closure plus antiplatelet therapy vs antiplatelet therapy 6 7 Table 3: Summary of findings of PFO closure plus antiplatelet therapy vs anticoagulation therapy 8 9 Table 4: Summary of findings of anticoagulation vs antiplatelet therapy 10 11 12 13 14 Figure 1. PRISMA flow diagram of studies included in review treatment of patients with PFO and 15 16 cryptogenic stroke For peer review only 17 18 19 Figure 2. Meta-regression curve based on the proportion of anticoagulant in the medical therapy arm 20 21 for ischaemic stroke 22 23 Figure 3. Combined Kaplan-Meier curves with individualized patient data based on the type of 24 25 26 intervention for ischaemic stroke 27 28 29 30 31

32 http://bmjopen.bmj.com/ 33 34 35 36 37 38 39

Table 1: Characteristics of patients in eligible studies BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 2 3 Author n randomised Mean % Inclusion Moderate Atrial Most common 4 Age Male criteria or higher septal device used for 5 shunt (%) aneurysm closure 6 (1) > 10mm 7 (%) (7) 8 PFO closure plus antiplatelet vs antiplatelet therapy 9 Furlan (CLOSURE 1, 909 46.0 51.8 Cryptogenic 52.9 (2) 37.8 STARFlex 100% 10 2012) stroke, PFO, > 11 18yo and < 12 60yo 13 Mas (CLOSE, 2017) 473 43.4 59.0 Cryptogenic 92.5 (3) 31.8 Amplatzer 52%† 14 stroke, PFO, > 15 16yo and < 16 For peer review60yo only 17 Meier (PC Trial, 414 44.5 49.8 Cryptogenic 65.6 (4) 23.7 (8) Amplatzer 100% 18 2013) stroke, PFO, > 19 18yo and < 20 60yo 21 Saver (RESPECT, 980 45.9 54.7 Cryptogenic 48.8 (5) 35.7 (9) Amplatzer 100% 22 2017) stroke, PFO, > 23 18yo and < 24 60yo 25 Sondergaard 664 45.1 60.6 Cryptogenic 81.0 (2) NR for AP Cardioform 61%¥ 26 (REDUCE, 2017) stroke, PFO, > group 27 18yo and < 28 60yo 29 Lee (DEFENCE PFO, 120 51.5 55.8 Cryptogenic NA 10.8 Amplatzer 100% 30 2018) stroke, PFO, 31 no age limit

32 http://bmjopen.bmj.com/ 33 PFO closure plus antiplatelet vs anticoagulation 34 Mas (CLOSE, 2017) 353 NA NA Cryptogenic NA NA NA 35 stroke, PFO, > 36 16yo and < 37 60yo 38 Anticoagulation vs antiplatelet therapy 39 Homma (PICSS, 203 (98 with 57.9 59.1 PFO with or 41.4% (6) 11.5% NA

40 2002) cryptogenic without on September 27, 2021 by guest. Protected copyright. 41 stroke) cryptogenic 42 stroke 43 >30yo and < 44 85yo 45 Mas (CLOSE, 2017) 361 44.2 57.0 Cryptogenic NA NA NA 46 stroke, PFO, > 47 16yo and < 48 60yo 49 Shariat (2013) 44 61.4 63.6 Cryptogenic NA NA NA 50 stroke, PFO, > 51 18yo 52 1. Shunt size was measured based on the number of microbubbles in the left atrium within 3 cycles of 53 54 being seen in the right atrium on transthoracic or transoesophageal echocardiography 55 2. Greater than or equal to 25 Microbubbles 56 3. Greater than or equal to 30 Microbubbles 57 4. Greater than or equal to 20 Microbubbles 58 59 5. Size of shunt not clearly defined 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 26 of 101

6. At least 1 microbubble BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 7. Atrial septal aneurysm was assessed on transoesophageal echo and was defined as septal mobility or 2 3 protrusion 4 8. Atrial septal aneurysm greater than 15mm 5 9. Atrial septal aneurysm not clearly defined 6 † 13% Intrasept PFO occluder, 9% STARFlex septal closure System, 9% Premere, 6% Amplatzer cribriform 7 8 occluder, 6% Figulla flex II PFO occluder, 1% Atriasept II occluder, 1% Gore helex septal occluder, 1% 9 Amplatzer AS occluder, 1% Figulla flex II UNI occluder, 1% Figulla flex II ASD occluder 10 ¥ 39% Gore helex septal occluder 11 NA Not available 12 13 14 15 16 For peer review only 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31

32 http://bmjopen.bmj.com/ 33 34 35 36 37 38 39

Table 2: GRADE summary of findings of PFO closure plus antiplatelet therapy vs antiplatelet therapy alone in BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 2 3 patients with cryptogenic stroke 4 5 Absolute effect estimates per 1000 Certainty in effect 6 Outcome Study results and patient-years estimates Plain text summary 7 (Timeframe) measurements Antiplatelet PFO closure plus (Quality of evidence) 8 therapy antiplatelet therapy 9 10 Odds Ratio: 0.12 100 13 PFO closure plus (CI 95% 0.04 - 0.27) per 1000 per 1000 Moderate antiplatelet therapy 11 Ischaemic stroke Direct evidence in NMA from Due to serious probably results in a large (Standardized to 5 years) 12 1257 patients in 3 studies Difference: 87 fewer imprecision1 decrease in ischemic 13 Follow up 3.8 years (CI 95% 100 fewer* - 33 fewer) stroke 14 15 Odds Ratio: 3.28 3 9 16 (CI 95% 0.2 - 174.22) per 1000 per 1000 Moderate Death For peer review only There is probably little or 17 Direct evidence in NMA from Due to serious (Standardized to 5 years) 2 no difference in death 18 1257 patients in 3 studies Difference: 6 more imprecision Follow up 3.8 years 19 (CI 95% 3 fewer* - 9 more) 20 14 7 21 Odds Ratio: 0.48 (CI 95% 0.2 - 1.12) per 1000 per 1000 Moderate There is probably little or Major bleeding 22 Direct evidence in NMA from Due to serious no difference in major (Standardized to 5 years) 23 1257 patients in 3 studies Difference: 7 fewer imprecision3 bleeding 24 Follow up 3.8 years (CI 95% 14 fewer* - 1 more) 25 5 23 26 Relative risk: 4.84 PFO closure plus x per 1000 per 1000 27 Persistent atrial (CI 95% 1.91 - 12.26) antiplatelet therapy Moderate Based on data from 3560 probably increases 28 fibrillation or Flutter Difference: 18 more per 1000 Due to serious risk of bias4 29 (Standardized to 1 year) patients in 6 studies @ persistent atrial Follow up 3.9 years patients fibrillation 30 (CI 95% 5 more - 56 more) 31 5 17 32 Relative risk: 3.76 http://bmjopen.bmj.com/ per 1000 per 1000 PFO closure plus 33 Transient or paroxysmal (CI 95% 1.74 - 8.1) Moderate antiplatelet therapy atrial fibrillation or flutter Based on data from 3560 34 Difference: 12 more per 1000 Due to serious risk of bias5 probably increases (Standardized to 1 year) patients in 6 studies 35 @ transient atrial fibrillation Follow up 3.9 years patients 36 (CI 95% 3 more - 31 more) 37 0 36 38 PFO closure plus Risk difference: 0.04 per 1000 per 1000 Device or procedure antiplatelet therapy 39 (CI 95% 0.02 – 0.05) related adverse events High6 increase device or 40 Based on data from 3560 Difference: 36 more per 1000 on September 27, 2021 by guest. Protected copyright. (Standardized to 1 year) procedure related adverse 41 patients in 6 studies @ patients events 42 Follow up 3.9 years (CI 95% 23 more - 50 more) 43 44 Odds Ratio: 1.01 5 5 (CI 95% 0.09 - 11.21) per 1000 per 1000 PFO plus antiplatelet 45 Pulmonary embolism High Direct evidence in NMA from therapy has no effect on (Standardized to 5 years) 46 1137 patients in 2 studies Difference: 0 fewer pulmonary embolism 47 Follow up 4.3 years (CI 95% 5 fewer - 48 more) 48 49 Odds Ratio: 0.82 34 28 50 (CI 95% 0.32 - 2.11) per 1000 per 1000 Moderate There is probably little or Transient ischaemic attack Direct evidence in NMA from Due to serious no difference in transient 51 (Standardized to 5 years) 7 52 1257 patients in 3 studies Difference: 6 fewer imprecision ischaemic attack Follow up 3.8 years 53 (CI 95% 34 fewer* - 15 more) 54 Odds Ratio: 0.83 6 5 55 (CI 95% 0.13 - 7.25) per 1000 per 1000 There is little or no Systemic embolism High 56 Direct evidence in NMA from difference in systemic (Standardized to 5 years) 57 1257 patients in 3 studies Difference: 1 fewer embolism 58 Follow up 3.8 years (CI 95% 6 fewer* - 4 more) 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 28 of 101

Note: The baseline risk for PFO closure was used to estimate the absolute effect in the antiplatelet arm. The baseline risk in the PFO arm was calculated based on the BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 2 median risk of the outcome in the six RCTs included. This allowed for consistency in the absolute effect estimate in the PFO closure arms in table 2 and 3 thus allowing 3 ease in comparison. 4 5 1. Risk of bias: No serious. Despite inadequate/lack of blinding of participants and personnel, resulting in potential for performance bias, we decided not to 2 6 downgrade since we rated ischaemic stroke as an objective outcome (borderline decision). Inconsistency: No serious. Borderline decision I 54%, not rated down. Imprecision: Serious. Low number of events. 7 2. Imprecision: Serious. Wide confidence intervals, included appreciable harm. Low number of events. 8 3. Imprecision: Serious. Low number of events. 9 4. Risk of bias: Serious. Not clearly stated how this was measured or assessed with prolonged ECG monitoring. Also, it is not clear for all events whether it was 10 transient or persistent. 11 5. Risk of bias: Serious. Not clearly stated how this was measured or assessed with prolonged ECG monitoring. Also, it is not clear for all events whether it was transient or persistent. 12 6. Inconsistency: No serious. Inconsistency: 1 study as high 60/100 and 1 as low as 10/1000., Point estimates vary widely. Not rated down. 13 7. Imprecision: Serious. Wide confidence intervals, included both appreciable benefit and harm. Low number of events. x 14 Defined as persistent according to the study definition or requiring a cardioversion attempt. 15 *: The calculated confidence interval using risk difference, because of uncertainty in the point estimates, permits reductions greater than the point estimates in the 16 PFO group. To avoid confusion, we have truncated to present the maximum reduction as equal to the PFO event rate. @: In the first year after procedure ratherFor than 5 years. peer review only 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31

32 http://bmjopen.bmj.com/ 33 34 35 36 37 38 39

BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 2 3 Table 3: Summary of GRADE evidence profile of PFO closure plus antiplatelet therapy vs anticoagulation in 4 5 patients with cryptogenic stroke 6 7 Absolute effect estimates per 1000 patient- 8 Certainty in effect Outcome Study results and years 9 estimates Plain text summary Timeframe measurements PFO closure plus 10 Anticoagulation (Quality of evidence) 11 antiplatelet therapy 12 Odds Ratio: 0.44 29 13 13 (CI 95% 0.08 - 3.83) per 1000 per 1000 Low There may be little or no Ischaemic stroke 14 Direct evidence in NMA from Due to very serious difference in ischaemic (Standardized to 5 years) 1 15 353 patients in 1 studies Difference: 16 fewer imprecision stroke 16 FollowFor up 5.3 years peer (CIreview 95% 29 fewer* - 10 more) only 17 29 27 18 Odds Ratio: 0.93 Low Ischaemic stroke - modelling per 1000 per 1000 There may be little or no (CI 95% 0.31 - 2.76) Due to serious 19 data from VTE literature difference in ischaemic imprecision and 20 (Standardized to 5 years) stroke Difference: 2 fewer serious indirectness2 21 (CI 95% 20 fewer - 47 more) 22 23 Relative risk: 0.69 13 9 (CI 95% 0.02 - 32.36) per 1000 per 1000 Moderate 24 Death There is probably little or Direct evidence in NMA from Due to serious 25 (Standardized to 5 years) no difference in death 353 patients in 1 studies imprecision3 26 Difference: 4 fewer Follow up 5.3 years (CI 95% 13 fewer* - 9 more) 27 28 Odds Ratio: 0.26 27 7 PFO closure plus 29 (CI 95% 0.07 - 0.82) per 1000 per 1000 Moderate Major bleeding antiplatelet therapy 30 Direct evidence in NMA from Due to serious (Standardized to 5 years) probably decreases major 353 patients in 1 studies imprecision4 31 Difference: 20 fewer bleeding Follow up 5.3 years (CI 95% 27 fewer* - 2 fewer) 32 http://bmjopen.bmj.com/ 33 24 7 34 Odds Ratio: 0.28 PFO closure plus Major bleeding - modelling per 1000 per 1000 Moderate 35 (CI 95% 0.13 - 0.55) antiplatelet therapy data from VTE literature Due to serious probably decreases major 36 (Standardized to 5 years) indirectness5 Difference: 17 fewer bleeding 37 (CI 95% 21 fewer - 11 fewer) 38 39 Relative risk: 4.84 5 23 x PFO closure plus

Persistent atrial fibrillation or (CI 95% 1.91 - 12.26) per 1000 per 1000 Moderate on September 27, 2021 by guest. Protected copyright. 40 antiplatelet therapy Based on data from 3560 Due to serious risk of 41 flutter probably increases non- patients in 6 studies @ bias6 42 (Standardized to 1 year) Difference: 18 more per 1000 patients transient atrial fibrillation Follow up 3.9 years (CI 95% 5 more - 56 more) 43 44 Relative risk: 3.76 5 17 PFO closure plus 45 Transient or paroxysmal atrial (CI 95% 1.74 - 8.1) per 1000 per 1000 Moderate antiplatelet therapy 46 fibrillation or flutter Based on data from 3560 Due to serious risk of probably increases (Standardized to 1 year) patients in 6 studies @ bias7 47 Difference: 12 more per 1000 patients transient atrial fibrillation 48 Follow up 3.9 years (CI 95% 3 more - 31 more) 49

0 36 PFO closure plus 50 Risk difference: 0.04 Device or procedure related per 1000 per 1000 antiplatelet therapy 51 (CI 95% 0.02 – 0.05) High adverse event increases device or Based on data from 3560 52 (Standardized to 1 year) @ procedure related adverse patients in 6 studies Difference: 36 more per 1000 patients 53 (CI 95% 23 more - 50 more) events 54 Follow up 3.9 years 55 Odds Ratio: 1.27 22 28 56 (CI 95% 0.4 - 4.52) per 1000 per 1000 Moderate There is probably little or Transient ischaemic attack 57 Direct evidence in NMA from Due to serious no difference in transient (Standardized to 5 years) 8 58 353 patients in 1 studies Difference: 6 more per 1000 imprecision ischaemic attack Follow up 5.3 years 59 (CI 95% 22 fewer* - 22 more) 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 30 of 101 BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 5 1 Pulmonary embolism - Odds Ratio: 9.09 per 1000 per 1000 Moderate There is probably little or modelling data from VTE (CI 95% 3.7 - 25.0) 2 Due to serious no difference in literature 3 indirectness9 pulmonary embolism (Standardized to 5 years) Difference: 4 more per 1000 4 (CI 95% 1 more - 13 more) 5 6 Odds Ratio: 291.0 0 0 (CI 95% 0.0 - 999.0) per 1000 per 1000 Moderate There is probably little or 7 Systemic embolism Direct evidence in NMA from Due to serious no difference in systemic 8 (Standardized to 5 years) 353 patients in 1 studies imprecision10 embolism 9 Difference: 0 fewer per 1000 Follow up 5.3 years (CI 95% 11 fewer - 11 more) 10 11 Note: The baseline risk for PFO closure was used to estimate the absolute effect in the anticoagulation arm. The baseline risk in the PFO arm was calculated based on 12 the median risk of the outcome in the six RCTs included. This allowed for consistency in the absolute effect estimate in the PFO closure arms in table 2 and 3 and the 13 14 anticoagulation arms in table 3 and 4 allowing for ease in comparison. 15 16 1. Imprecision: Very Serious. Wide confidence interval. Low number of events. 2. Indirectness: Serious. In addition toFor the direct evidence peer from randomised review trials in patients with PFOonly and a cryptogenic ischaemic stroke, we additionally 17 considered external evidence from randomised trials that assessed the impact of anticoagulation vs. antiplatelet therapy for the secondary prevention of 18 venous thromboembolism. Imprecision: Serious. Wide confidence intervals, includes both appreciable benefit and harm. 19 3. Imprecision: Serious. Wide confidence intervals, includes both appreciable benefit and harm. Low number of events. 20 4. Imprecision: Serious. Wide confidence interval, included a not important benefit. Low number of events. 5. Indirectness: Serious. In addition to the direct evidence from randomised trials in patients with PFO and a cryptogenic ischaemic stroke, we additionally 21 considered external evidence from randomised trials that assessed the impact of anticoagulation vs. antiplatelet therapy for the secondary prevention of 22 venous thromboembolism. 23 6. Risk of bias: Serious. Not clearly stated how this was measured or assessed with prolonged ECG monitoring. Also, it is not clear for all events whether it was 24 transient or persistent. 25 7. Risk of bias: Serious. Not clearly stated how this was measured or assessed with prolonged ECG monitoring. Also, it is not clear for all events whether it was transient or persistent. 26 8. Imprecision: Serious. Wide confidence interval, includes both appreciable benefit and harm. Low number of events. x 27 Defined as persistent according to the study definition or requiring a cardioversion attempt *: The calculated confidence interval using risk difference, because of 28 uncertainty in the point estimates, permits reductions greater than the point estimates in the PFO group. To avoid confusion, we have truncated to present the 29 maximum reduction as equal to the PFO event rate. 30 @: In the first year after procedure rather than 5 years VTE, venous thromboembolism; AF, atrial fibrillation; CHF, congestive heart failure 31

32 http://bmjopen.bmj.com/ 33 34 35 36 37 38 39

BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 2 3 Table 4: Summary of GRADE evidence profile of anticoagulation vs antiplatelet therapy alone in patients with 4 5 cryptogenic stroke 6 7 Absolute effect estimates per 1000 8 Certainty in effect Outcome patient-years 9 Study results and measurements estimates Plain text summary Timeframe 10 Antiplatelet Anticoagulation (Quality of evidence) 11 y 12 Odds Ratio: 0.27 100 29 (CI 95% 0.03 - 1.21) per 1000 per 1000 Low 13 Ischaemic stroke Anticoagulation may decrease Direct evidence in NMA from 361 Due to very serious 14 (Standardized to 5 years) ischaemic stroke patients in 1 studies imprecision1 15 Difference: 71 fewer Follow up 5.3 years (CI 95% 100 fewer* - 17 more) 16 For peer review only 17 100 19 Odds Ratio: 0.17 18 Ischaemic stroke - modelling per 1000 per 1000 Low (CI 95% 0.08 - 0.36) Anticoagulation may decrease 19 data from VTE literature Due to very serious ischaemic stroke (Standardized to 5 years) indirectness2 20 Difference: 81 fewer 21 (CI 95% 91 fewer - 62 fewer) 22 23 Odds Ratio: 4.81 3 13 (CI 95% 0.31 - 224.43) per 1000 per 1000 Low 24 Death There may be little or no Direct evidence in NMA from 408 Due to very serious 25 (Standardized to 5 years) 3 difference in death patients in 2 studies Difference: 10 more imprecision 26 Follow up 3.2 years (CI 95% 3 fewer* - 357 more) 27 28 Odds Ratio: 1.9 14 26 29 (CI 95% 0.68 - 5.53) per 1000 per 1000 Moderate Major Bleeding Anticoagulation probably 30 Direct evidence in NMA from 408 Due to serious (Standardized to 5 years) 4 increases major bleeding 31 patients in 2 studies Difference: 12 more imprecision Follow up 3.2 years (CI 95% 5 fewer - 65 more)

32 http://bmjopen.bmj.com/ 33 14 25 Odds Ratio: 1.77 34 Major bleeding - modelling per 1000 per 1000 Moderate (CI 95% 1.36 - 2.31) Anticoagulation probably 35 data from VTE literature Due to serious increases major bleeding 36 (Standardized to 5 years) indirectness5 Difference: 11 more 37 (CI 95% 5 more - 18 more) 38 39 Odds Ratio: 0.65 34 22 (CI 95% 0.19 - 1.98) per 1000 per 1000 Low There may be little or no 40 Transient ischaemic attack on September 27, 2021 by guest. Protected copyright. Direct evidence in NMA from 361 Due to very serious difference in transient ischaemic 41 (Standardized to 5 years) 6 patients in 1 studies Difference: 12 fewer imprecision attack 42 Follow up 5.3 years (CI 95% 34 fewer* - 24 more) 43 44 5 1 Pulmonary embolism - Odds Ratio: 0.11 45 per 1000 per 1000 Moderate There is probably little or no modelling data from VTE (CI 95% 0.04 - 0.37) Due to serious difference in pulmonary 46 literature indirectness7 embolism. 47 (Standardized to 5 years) Difference: 4 fewer 48 (CI 95% 5 fewer - 3 fewer) 49 0 0 50 Not estimable per 1000 per 1000 Moderate Systemic embolism There is probably little or no 51 Direct evidence in NMA from 361 Due to serious (Standardized to 5 years) 8 difference in systemic embolism 52 patients in 1 studies Difference: 0 fewer imprecision 53 Follow up 5.3 years (CI 95% 11 fewer - 11 more) 54 Note: The baseline risk for antiplatelet was obtained using calculated absolute effect estimate. This was done to maintain consistency across the tables. The 55 56 calculated baseline risk (10%) was similar to the baseline risk calculated using the median of studies included (9%) 57 58 1. Imprecision: Very Serious. Wide confidence interval, includes appreciable harm. Low number of events. 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 32 of 101

2. Indirectness: Very Serious. In addition to the direct evidence from randomised trials in patients with PFO and a cryptogenic ischaemic stroke, we additionally BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 considered external evidence from randomised trials that assessed the impact of anticoagulation vs. antiplatelet therapy for the secondary prevention of 2 venous thromboembolism. 3. Imprecision: Very Serious. Wide confidence interval, includes both appreciable benefit and harm. Low number of events. 3 4. Imprecision: Serious. Wide confidence intervals, Low number of events. 4 5. Indirectness: Serious. In addition to the direct evidence from randomised trials in patients with PFO and a cryptogenic ischaemic stroke, we additionally 5 considered external evidence from randomised trials that assessed the impact of anticoagulation vs. antiplatelet therapy for the secondary prevention of 6 venous thromboembolism. We did not rate down with two levels because we felt the outcome is less indirect compared to VTE literature than Ischaemic stroke. 6. Imprecision: Very Serious. Wide confidence interval, includes both appreciable harm and benefit. Low number of events. 7 *: The calculated confidence interval using risk difference, because of uncertainty in the point estimates, permits reductions greater than the point estimates in the 8 PFO group. To avoid confusion, we have truncated to present the maximum reduction as equal to the PFO event rate. 9 10 11

12 13 14 15

16 For peer review only 17 18 19

20 21 22 23 24

25 26 27 28

29 30 31

32 http://bmjopen.bmj.com/ 33

34 35 36 37

38 39

40 on September 27, 2021 by guest. Protected copyright. 41

42 43 44 45 46

47 48 49 50

51 52 53 54

55 56 57 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 33 of 101 BMJ Open

BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 2

3 4 5

6 7 Records identified through 8 database searching 9 (total n = 997): 10 MEDLINE (n=326), 11 EMBASE (n=590), 12 CENTRAL (n=81) 13 14 15 Identification 16 For peer review only 17 18 Records after duplicates removed 19 (n = 772) 20 21

22 23 24 Records screened Records excluded 25 (n = 772) (n = 740) 26 27

Screening 28 29

30 Full-text articles assessed Full-text articles excluded, 31 for eligibility with reasons 32 http://bmjopen.bmj.com/

33 (n = 32) (n = 25) 34 35 36 37 Studies included in Study added (published

38 Eligibility qualitative synthesis after search completion) 39 (n = 7) (n = 1)

40 on September 27, 2021 by guest. Protected copyright. 41 42 43

44 Studies included in 45 quantitative synthesis 46 (meta-analysis)

47 (n = 8) 48 49 Included 50 51 52 Figure 1. PRISMA flow diagram of studies included in review treatment of patients with PFO and cryptogenic 53 54 stroke 55 56 57 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 34 of 101 BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 For peer review only 17 18 19 20 21 22 23 Figure 2. Meta-regression curve of the proportion of patients prescribed an anticoagulant in the medical 24 25 26 therapy arm for the outcome ischaemic stroke 27 28 29 30 31

32 http://bmjopen.bmj.com/ 33 34 35 36 37 38 39

40 on September 27, 2021 by guest. Protected copyright. 41 42 43 44 45 46 47 48 Figure 3. Combined Kaplan-Meier curves of individual participant data in the included trials, by type of 49 50 intervention for the outcome ischaemic stroke 51 52 Note: 53 1) Y-axis is truncated from 90% to 100% Event-free Survival 54 2) Mixed medical therapy includes studies where the proportion of antiplatelet agents was less than 80% and the proportion of anticoagulant was greater 55 than 25% 3) Proportion of patients: PFO Closure + antiplatelet therapy (N= 1829; 50%), Mixed medical therapy (N= 1153; 32%), Antiplatelet therapy (N= 458; 13%), and 56 Anticoagulation (N= 210; 6%) 57 4) K-M curves were available for the outcome of ischaemic stroke in the CLOSE study (14) (PFO closure, anticoagulation, and antiplatelet), the PC Trial (6) 58 (PFO closure and mixed medical therapy), the RESPECT trial (16) (PFO closure and mixed medical therapy) , and the REDUCE trial (15) (PFO closure and 59 antiplatelet). 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 35 of 101 BMJ Open

BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 2 APPENDICES 3 4 Appendix 1: Summary of search and strategies 5 6 7 Appendix 2: Tables 8 9 1) Table 1: Characteristics of eligible studies 10 11 12 2) Table 2: Direct and indirect estimates of effects for the network meta-analysis 13 14 Appendix 3: Figures 15 16 For peer review only 17 1) Figure 1: Risk of bias of included randomised controlled trials 18 19 2) Figure 2: Network of included RCTs with available direct comparisons for ischaemic stroke. 20 21 3) Figure 3: Meta-regression evaluating the effect of moderate or higher shunt size on the effectiveness 22 23 24 of PFO closure versus medical therapy 25 26 4) Figure 4: Meta-regression evaluating the proportion of patients on anticoagulation compared to those 27 28 29 with moderate or higher shunt size 30 31 Appendix 4: PFO closure vs antiplatelet forest plots (Direct evidence only) 32 Figure 1: Ischaemic stroke http://bmjopen.bmj.com/ 33 34 Figure 2: TIA 35 Figure 3: Systemic emboli 36 Figure 4: Death 37 Figure 5: Major bleeding 38 39 Figure 6: Pulmonary embolism

40 Appendix 5: PFO closure vs anticoagulation forest plots (Direct evidence only) on September 27, 2021 by guest. Protected copyright. 41 Figure 1: Ischaemic stroke 42 Figure 2: TIA 43 Figure 3: Systemic emboli 44 45 Figure 4: Pulmonary emboli 46 Figure 5: Major bleeding 47 Appendix 6: Complications of PFO closure vs medical therapy forest plots (Direct evidence only) 48 Figure 1: All atrial fibrillation 49 50 Figure 2: Persistent atrial fibrillation 51 Figure 3: Transient atrial fibrillation 52 Figure 4: Serious device or procedure related 53 Appendix 7: Anticoagulation vs antiplatelet forest plots (Direct evidence only) 54 Figure 1: Stroke 55 56 Figure 2: TIA 57 Figure 3: Major bleeding 58 Figure 4: Death 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 36 of 101

BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 2 3 4 5 Appendix 1: Summary of search and strategies 6 7 8 Database: OVID Medline Epub Ahead of Print, In-Process & Other Non-Indexed Citations, Ovid MEDLINE(R) 9 Daily and Ovid MEDLINE(R) 1946 to Present 10 Search Strategy: 11 ------12 1 cerebrovascular disorders/ or basal ganglia cerebrovascular disease/ or exp brain ischemia/ or carotid 13 14 artery diseases/ or carotid artery thrombosis/ or intracranial arterial diseases/ or cerebral arterial diseases/ or 15 exp "intracranial embolism and thrombosis"/ or exp stroke/ (237544) 16 2 (isch?emi$ adj6 (stroke$For or apoplex$ peer or cerebral review vasc$ or cerebrovasc$ only or cva or attack$)).tw. (62463) 17 3 ((brain or cerebr$ or cerebell$ or vertebrobasil$ or hemispher$ or intracran$ or intracerebral or 18 19 infratentorial or supratentorial or middle cerebr$ or mca$ or anterior circulation) adj5 (isch?emi$ or infarct$ 20 or thrombo$ or emboli$ or occlus$ or hypoxi$)).tw. (105788) 21 4 (TIA or TIAs).mp. (8376) 22 5 Embolism, Paradoxical/ (914) 23 6 heart atria/ and (embolism/ or thromboembolism/) (648) 24 25 7 ((paradoxic$ or crossed) adj5 embolism$).tw. (1585) 26 8 (cryptogenic adj5 stroke).tw. (1168) 27 9 or/1-8 (293081) 28 Annotation: population part 1: stroke or cryptogenic stroke 29 30 10 heart septal defects, atrial/ or foramen ovale, patent/ (14696) 31 11 heart septum/ or atrial septum/ or foramen ovale/ (10405)

32 12 (patent foramen ovale or PFO).tw. (5125) http://bmjopen.bmj.com/ 33 13 (atrial sept$ adj5 defect$).tw. (10211) 34 14 ((right to left or R-L or venous to arterial or venous-arterial or V-A) adj3 shunt).tw. (2212) 35 36 15 or/10-14 (30798) 37 Annotation: population part 2: PFO 38 16 9 and 15 (3491) 39 Annotation: population Stroke and PFO

40 on September 27, 2021 by guest. Protected copyright. 41 17 limit 16 to yr="2012 -Current" (1094) 42 Annotation: since last review 43 18 limit 17 to ("therapy (maximizes sensitivity)" or "therapy (maximizes specificity)" or "therapy (best 44 balance of sensitivity and specificity)") (429) 45 19 randomised controlled trial.pt. (496904) 46 47 20 controlled clinical trial.pt. (99253) 48 21 randomised.ab. (433409) 49 22 placebo.ab. (202740) 50 23 drug therapy.fs. (2114500) 51 52 24 randomly.ab. (298737) 53 25 trial.ab. (457112) 54 26 groups.ab. (1845391) 55 27 or/19-26 (4369043) 56 28 exp animals/ not humans.sh. (4677556) 57 58 29 27 not 28 (3778961) 59 30 17 and 29 (274) 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 37 of 101 BMJ Open

31 18 or 30 (485) BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 Annotation: Population stroke and PFO limit to since 2012 and RCTs 2 3 32 "prostheses and implants"/ or septal occluder device/ (47787) 4 33 Wound Closure Techniques/ (1141) 5 34 (close or closure or septal occluder).tw. (384980) 6 35 (cardioseal or gore helex or amplatzer or starflex or cardia or intrasept or premere).tw. (9507) 7 8 36 su.fs. (1918373) 9 37 or/32-36 (2271562) 10 Annotation: Septal occluder device as per Liu 2015 CDSR 11 38 exp Anticoagulants/ (212306) 12 39 anticoagulant$.tw. (55364) 13 14 40 (acenocoumarol$ or dicoumarol$ or ethyl biscoumacetate$ or phenprocoumon$ or warfarin$ or ancrod$ 15 or citric acid$ or coumarin$ or chromonar$ or coumestro$ or esculi$ or ochratoxin$ or umbelliferone$ or 16 dermatan?sul$ or dextran$For or edetic peer acid$ or enoxaparin$ review or gabexate$ only or heparin$ or lmwh$ or nadroparin$ 17 or pentosan sulfuric polyester$ or phenindione$ or protein c or protein s or tedelparin$).tw. (192083) 18 19 41 (argatroban or tinzaparin or parnaparin or reviparin or danaparoid or lomoparan or org 10172 or 20 mesoglycan or polysaccharide sulphate$ or sp54 or sp-54 or md805 or md-805 or cy222 or cy-222 or cy216 or 21 cy-216).tw. (2862) 22 42 (Marevan or Fragmin$ or Fraxiparin$ or Klexane).tw. (638) 23 43 exp Pipecolic acids/ae, tu (3633) 24 25 44 exp Vitamin K/ai (2564) 26 45 Vitamin K antagonist$.tw. (4641) 27 46 exp Antithrombins/ae, pd, de, tu (6778) 28 47 exp Blood coagulation factors/ai, de (16900) 29 30 48 exp Blood coagulation/de (19187) 31 49 (anticoagulat$ or antithromb$).tw. (67390)

32 50 or/38-49 (376368) http://bmjopen.bmj.com/ 33 Annotation: anticoagulants as per Berge 2002 CDSR 34 51 Factor Xa Inhibitors/ (3589) 35 36 52 Dabigatran/ (2442) 37 53 Rivaroxaban/ (2106) 38 54 (dabigatran or rivaroxaban or apixaban or edoxaban).mp. (7012) 39 55 anti-factor Xa.mp. (893)

40 on September 27, 2021 by guest. Protected copyright. 41 56 (factor Xa adj2 (antag* or inhibit*)).mp. (5256) 42 57 novel oral anticoagulant*.mp. (1068) 43 58 noac.mp. (829) 44 59 noacs.mp. (1129) 45 60 pradax.mp. (8) 46 47 61 pradaxa.mp. (127) 48 62 BIBR-953.mp. (8) 49 63 BIBR-953ZW.mp. (2) 50 64 xarelto.mp. (114) 51 52 65 BAY 59-7939.mp. (27) 53 66 BMS-562247.mp. (6) 54 67 eliquis.mp. (43) 55 68 lixiana.mp. (13) 56 69 DU-176.mp. (1) 57 58 70 DU-176b.mp. (27) 59 71 non-vitamin K.mp. (828) 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 38 of 101

72 or/51-71 (11701) BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 73 direct oral anticoagulant*.mp. (1161) 2 3 74 DOAC.mp. (449) 4 75 DOACs.mp. (570) 5 76 TSOAC.mp. (25) 6 77 TSOACs.mp. [mp=title, abstract, original title, name of substance word, subject heading word, keyword 7 8 heading word, protocol supplementary concept word, rare disease supplementary concept word, unique 9 identifier, synonyms] (40) 10 78 oral anticoagulant.mp. (4604) 11 79 (new or novel or direct or direct-acting or target-specific or targeted or non-vitamin K).mp. [mp=title, 12 abstract, original title, name of substance word, subject heading word, keyword heading word, protocol 13 14 supplementary concept word, rare disease supplementary concept word, unique identifier, synonyms] 15 (3920515) 16 80 78 and 79 (1801) For peer review only 17 81 72 or 73 or 74 or 75 or 80 (12741) 18 19 Annotation: NOACs or DOACs from 2016 Canadian Cardiac Society GL search RC 20 82 exp Platelet aggregation inhibitors/ (106651) 21 83 (antiplatelet$ or anti-platelet$ or antiaggreg$ or anti-aggreg$ or (platelet$ adj5 inhibit$) or (thrombocyt$ 22 adj5 inhibit$)).tw. (50701) 23 84 (alprostadil$ or aspirin$ or dipyridamol$ or disintegrin$ or epoprostenol$ or iloprost$ or ketanserin$ or 24 25 ketorolac tromethamine$ or milrinone$ or mopidamol$ or pentoxifyllin$ or procainamide$ or ticlopidine$ or 26 thiophen$ or trapidil$).tw. (85938) 27 85 (acetyl salicylic acid$ or acetyl?salicylic acid or clopidogrel$ or picotamide$ or ligustrazine$ or levamisol$ 28 or suloctidil$ or ozagrel$ or oky046 or oky-046 or defibrotide$ or cilostazol or satigrel or sarpolgrelate or 29 30 kbt3022 or kbt-3022 or isbogrel or cv4151 or cv-4151 or triflusal).tw. (27863) 31 86 (Dispril or Albyl$ or Ticlid$ or Persantin$ or Plavix).tw. (707)

32 87 exp Platelet glycoprotein gpiib-iiia complex/ai, de (3381) http://bmjopen.bmj.com/ 33 88 (((glycoprotein iib$ or gp iib$) adj5 (antagonist$ or inhibitor$)) or GR144053 or GR-144053 or abciximab$ 34 or tirofiban$ or eftifibatid$).tw. (5389) 35 36 89 (ReoPro or Integrilin$ or Aggrastat).tw. (468) 37 90 exp Platelet activation/de (26939) 38 91 exp Blood platelets/de (19928) 39 92 or/82-91 (202643)

40 on September 27, 2021 by guest. Protected copyright. 41 Annotation: antiplatelets as per Berge 2002 CDSR 42 93 37 or 50 or 81 or 92 (2776671) 43 Annotation: Intervention block 44 94 31 and 93 (366) 45 95 remove duplicates from 94 (326) 46 47 48 Central 49 Date Run: 16/10/17 21:33:56.134 50 Description: 51 52 53 ID Search Hits 54 #1 MeSH descriptor: [Stroke] explode all trees 7179 55 #2 MeSH descriptor: [Cerebrovascular Disorders] explode all trees 12133 56 #3 (isch*emi* near/6 (stroke* or apoplex* or cerebral next vasc* or cerebrovasc* or cva or attack*)) 57 58 10210 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 39 of 101 BMJ Open

#4 ((brain or cerebr* or cerebell* or vertebrobasil* or hemispher* or intracran* or intracerebral or BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 infratentorial or supratentorial or middle next cerebr* or mca* or "anterior circulation") near/5 (isch*emi* or 2 3 infarct* or thrombo* or emboli* or occlus* or hypoxi*)) 11401 4 #5 (tia or tias) 1201 5 #6 MeSH descriptor: [Embolism, Paradoxical] explode all trees 15 6 #7 ((paradoxic* or crossed) near/5 embolism*) 49 7 8 #8 (cryptogenic near/5 stroke) 143 9 #9 MeSH descriptor: [Heart Atria] explode all trees 565 10 #10 MeSH descriptor: [Embolism and Thrombosis] explode all trees 6758 11 #11 #9 and #10 49 12 #12 atria* near/3 (emboli* or thromboemboli*) 206 13 14 #13 #1 or #2 or #3 or #4 or #5 or #6 or #7 or #8 or #11 or #12 24313 15 #14 MeSH descriptor: [Heart Septal Defects, Atrial] explode all trees 187 16 #15 "patent foramen ovale"For or PFO peer 277 review only 17 #16 atrial sept* near/5 defect* 290 18 19 #17 (("right to left" or "R-L" or "venous to arterial" or "venous-arterial" or "V-A") near/3 shunt) 71 20 #18 #14 or #15 or #16 or #17 567 21 #19 #13 and #18 Publication Year from 2012 to 2017 81 22 23 24 25 26 27 28 29 30 31

32 http://bmjopen.bmj.com/ 33 34 35 36 37 38 39

Appendix 2: Tables BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 2 3 Table 1: Characteristics of eligible studies 4 5 6 7 8 PFO closure plus antiplatelet vs antiplatelet therapy 9 Percent PFO Percent Crossover from closure technical 10 med therapy to Crossover from performed success/PFO 11 PFO closure PFO closure to /patients in closure 12 Study, year Intervention Medical therapy Withdrawn or loss to follow-up (%) med therapy PFO cohort attempt 13 Medical 14 PFO (%) therapy (%) STARFlex Device. 15 Clopidogrel for 6 16 months and ASA for 2 ForASA 325mg peer daily review only 17 Furlan, 2012 years (56%) 1.8 0.65 1.73 8.72 90.6 89.4 18 19

20 PFO closure (Amplatzer 21 51.5%X) plus ASA 75mg 22 daily and clopidogrel 23 75mg daily x 3 months Aspirin or then aspirin or clopidogrel or 24 clopidogrel or aspirin aspirin plus ER 25 Mas, 2017 plus ER dipyridamole dipyridamole 8.8 5.1 4.25 8.82 99.6 88.6 26 27 28

29 30 Amplatzer occluder. Antiplatelet 31 ASA 5-6 months. (67%) or

32 clopidogrel or anticoagulation http://bmjopen.bmj.com/ 19.5 33 Meier, 2013 ticlopidine 1-6 months (31%) 15.2 13.3 6.4 96.1 97.4 ASA, warfarin, 34 Amplatzer occluder clopidogrel 35 100%. ASA 81-325mg (25%), or ASA 36 daily plus clopidogrel plus 37 75mg daily x 1 months dipyridamole 38 Saver, 2017 then ASA x 5 months (75%) 20.8 33.3 3.95 2.8 93.6 99.6 PFO Closure 39 (Cardoiform Septal

40 occluder 61%, Gore on September 27, 2021 by guest. Protected copyright. 41 helex Septal occluder 42 39%) plus antiplatelet (ASA 75-325mg daily, Antiplatelet (ASA 43 aggrenox, or 75-325mg daily, 44 clopidogrel. Clopidogrel ASA plus 45 300mg x 1 then 75 dipyridamole, or 46 Sondergaard, daily x 3 days for clopidogrel. 47 2017 everyone (100%) 8.3 14.7 6.3 7.3 96.1 97.4 PFO Closure (100% 48 Amplatzer PFO Antiplatelet with 49 occluder). Suggested ASA or ASA and 50 DAPT for 6 months Clopidogrel 75mg 51 with ASA 100mg and daily, or ASA and clopidogrel 75mg daily. Cilostazol 200mg 52 However, up to daily. Could 53 treating physician. anticoagulate 54 Could even use with warfarin 55 Lee, 2018 anticoagulation (INR 2.0-3.0) 0 0 0 11.7 88.3 100 56 PFO closure plus antiplatelet vs anticoagulation 57 Crossover from Crossover from Percent Percent 58 antiplatelet to anticoagulation technical technical Study, year Intervention Control Withdrawn or loss to follow-up anticoagulation to antiplatelet success/PFO success/PFO 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 41 of 101 BMJ Open

(%) (%) closure closure BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 attempts attempts 2 PFO Closure plus 3 antiplatelet Anticoagulation 4 5 6 PFO closure (Amplatzer 7 51.5%†) plus ASA 75mg Warfarin with

8 daily and clopidogrel target INR 2-3 75mg daily x 3 months (93%) or direct 9 then aspirin or oral 10 clopidogrel or aspirin anticoagulant 11 Mas 2017 plus ER dipyridamole (7%) NA NA NA NA NA NA 12 Anticoagulation vs antiplatelet therapy 13 Crossover from Crossover from 14 antiplatelet to anticoagulation 15 anticoagulation to antiplatelet Study, year Intervention Control Withdrawn or loss to follow-up (%) (%) 16 For peer review only 17 Anticoagulation 18 (%) Antiplatelet (%) 19 20 21

22 Aspirin or 23 Warfarin with target clopidogrel or 24 INR 2-3 (93%) or direct aspirin plus ER 25 Mas 2017 oral anticoagulant (7%) dipyridamole 23.5 5.7 5.17 20.3 26 Warfarin 2.5mg po Aspirin 80mg po Shariat, 2013 daily (INR 2-3) tid (100%) 8.7 4.2 0 0 27 Homma (PICSS, Warfarin (INR 1.4-2.8) Aspirin 325mg po 28 2002) F/U 24 months od 6.4 2.8 NA NA 29 † 13% Intrasept PFO occluder, 9% STARFlex Septal Closure System, 9% Premere, 6% Amplatzer cribriform occluder, 6% Figulla Flex II PFO occluder, 1% Atriasept 30 II occluder, 1% Gore Helex eptal occluder, 1% Amplatzer AS occluder, 1% Figulla Flex II UNI occluder, 1%Figulla Flex II ASD occluder 31 NA = Not available

32 http://bmjopen.bmj.com/ 33 34 35 36 37 38 39

Table 2: Direct and indirect estimates of effects BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 2 Indirect NMA 3 Outcomes Comparisons Direct [OR,95%CrI] 4 [OR,95%CrI]$ [OR,95%CrI] 5 PFO+antiplatelet vs. 0.14 (0.05, 0.34) Not estimable* 0.12 (0.04, 0.27) 6 antiplatelet 7 PFO+antiplatelet vs. 9.6e-16 (2.0e-47, 8 Ischemic stroke 0.45 (0.06, 3.18) 0.44 (0.08, 3.76) 9 anticoagulation 6.3e+16) 10 Antiplatelet vs. 4.67 (0.80, 26.95) Not estimable* 3.73 (0.84, 28.11) 11 anticogualation 12 PFO+antiplatelet vs. 13 3.1e+14 (4.32, 8.5e+62) Not estimable* 2.48 (0.16, 99.11) 14 antiplatelet 15 PFO+antiplatelet vs. 3.4e-06 (1.5e-14, Death 0.97 (0.02, 47.77) 0.59 (0.01, 22.76) 16 anticoagulationFor peer 728.3)review only 17 Antiplatelet vs. 18 0.72 (0.01, 21.93) Not estimable* 0.23 (0.01, 3.44) 19 anticogualation 20 PFO+antiplatelet vs. 0.54 (0.21, 1.40) Not estimable* 0.48 (0.20, 1.10) 21 antiplatelet 22 Major bleeding PFO+antiplatelet vs. 23 0.09 (0.00, 2.15) 0.30 (0.08, 1.12) 0.25 (0.07, 0.82) 24 anticoagulation 25 Antiplatelet vs. 0.60 (0.19, 1.80) Not estimable* 0.53 (0.18, 1.47) 26 anticogualation 27 PFO+antiplatelet vs. 28 0.75 (0.26, 2.13) Not estimable* 0.82 (0.31, 2.09) 29 antiplatelet 30 PFO+antiplatelet vs. 1.51 (0.34, 6.72) 0.88 (0.11, 6.85) 1.25 (0.39, 4.38) 31 Recurrent TIA anticoagulation

32 http://bmjopen.bmj.com/ Antiplatelet vs. 33 1.28 (0.26, 6.32) Not estimable* 1.53 (0.50, 5.21) 34 anticogualation 35 PFO+antiplatelet vs. 0.79 (0.12, 6.76) Not estimable* 0.73 (0.12, 5.52) 36 antiplatelet 37 PFO+antiplatelet vs. 0.06 (5.7e-09, 0.03 (4.0e-08, 38 Systemic embolism 0.01 (1.4e-10, 1.0e+06) anticoagulation 6.0e+05) 1280.0) 39

40 Antiplatelet vs. 0.04 (5.4e-08, on September 27, 2021 by guest. Protected copyright. 0.01 (1.4e-10, 4.0e+05) Not estimable* 41 anticogualation 1616.0) 42 43 $ Indirect results were from back-calculated 44 * Cannot be estimated because the intervention was not connected in a loop in the evidence network 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 43 of 101 BMJ Open

Appendix 3: Figures BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 For peer review only 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31

32 http://bmjopen.bmj.com/ 33 34 35 36 37 38 39

40 on September 27, 2021 by guest. Protected copyright. 41 Figure 1: Risk of bias of included randomised controlled trials 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 44 of 101 BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 Figure 2: Network of included RCTs with available direct comparisons for ischaemic stroke. 16 For peer review only 17 Note: The thickness of the lines (edges) is proportional to the number of RCTs evaluating each treatment and direct 18 19 comparison. 20 21 22 23 24 25 26 27 28 29 30 31

32 http://bmjopen.bmj.com/ 33 34 35 36 37 38 39

40 on September 27, 2021 by guest. Protected copyright. 41 42 43 44 45 46 47 48 49 50 Figure 3: Meta-regression evaluating the effect of the proportion of participants with a moderate or higher 51 52 53 shunt size on the effectiveness of PFO closure versus medical therapy in patients with a PFO and cryptogenic 54 55 stroke. PFO closure was more effective in RCTs that had a higher proportion of patients with moderate or 56 57 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 45 of 101 BMJ Open

higher PFO shunts, p=0.047. Note, the shunt size was not reported in the DEFENSE-PFO Trial and thus it was BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 2 3 not included in the analysis. 4 5 6 7 8 9 10 11 12 13 14 15 16 For peer review only 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31

32 http://bmjopen.bmj.com/ 33 34 35 Figure 4: Scatter plot of randomised trials of PFO closure in patients with cryptogenic stroke showing an 36 37 inverse relationship between the proportion of participants with a moderate or higher shunt and proportion 38 39

40 of patients prescribed anticoagulation in the medical therapy arm. Note, the shunt size was not reported in on September 27, 2021 by guest. Protected copyright. 41 42 the DEFENSE-PFO Trial and thus it was not included in the analysis. 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 46 of 101

Appendix 4: PFO closure vs antiplatelet therapy forest plots (Direct evidence only) BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 Figure 1: Ischaemic stroke 16 For peer review only 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 Figure 2: TIA

32 http://bmjopen.bmj.com/ 33 34 35 36 37 38 39

40 on September 27, 2021 by guest. Protected copyright. 41 42 43 44 45 46 47 Figure 3: Systemic embolism 48 49 50 51 52 53 54 55 56 57 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 47 of 101 BMJ Open BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 2 3 4 5 6 7 8 9 10 11 Figure 4: Death 12 13 14 15 16 For peer review only 17 18 19 20 21 22 23 24 25 26 27 Figure 5: Major bleeding 28 29 30 31

32 http://bmjopen.bmj.com/ 33 34 35 36 37 38 39

40 on September 27, 2021 by guest. Protected copyright. 41 42 43 Figure 6: Pulmonary embolism 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 48 of 101

Appendix 5: PFO closure vs Anticoagulation forest plots (Direct evidence only) BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 2 3 4 5 6 7 8 9 10 11 12 Figure 1: Ischaemic stroke 13 14 15 16 For peer review only 17 18 19 20 21 22 23 24 25 26 27 28 Figure 2: TIA 29 30 31

32 http://bmjopen.bmj.com/ 33 34 35 36 37 38 39

40 on September 27, 2021 by guest. Protected copyright. 41 Figure 3: Systemic embolism 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 Figure 4: Pulmonary embolism 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 49 of 101 BMJ Open

BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 2 3 4 5 6 7 8 9 10 11 Figure 5: Major bleeding 12 13 14 15 16 For peer review only 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31

32 http://bmjopen.bmj.com/ 33 34 35 36 37 38 39

Appendix 6: Complications of PFO closure vs medical therapy forest plots (Direct evidence only) BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 Figure 1: All atrial fibrillationFor peer review only 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31

32 http://bmjopen.bmj.com/ 33 34 Figure 2: Persistent atrial fibrillation 35 36 37 38 39

40 on September 27, 2021 by guest. Protected copyright. 41 42 43 44 45 46 47 48 49 50 51 52 53 54 Figure 3: Transient atrial fibrillation 55 56 57 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 51 of 101 BMJ Open

BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 Figure 4: Device or procedure related adverse events 16 For peer review only 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31

32 http://bmjopen.bmj.com/ 33 34 35 36 37 38 39

Appendix 7: Anticoagulation vs antiplatelet therapy forest plots (Direct evidence only) BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 2 3 4 5 6 7 8 9 10 11 12 13 Figure 1: Ischaemic stroke 14 15 16 For peer review only 17 18 19 20 21 22 23 24 25 26 27 28 29 Figure 2: TIA 30 31

32 http://bmjopen.bmj.com/ 33 34 35 36 37 38 39

References BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 2 3 1. Adams H, Bendixen, BH, Kappelle, LJ, et al. Classification of Subtype of Acute Ischemic Stroke: 4 Definitions for Use in a Multicenter Clinical Trial. Stroke. 1993;24(1):35-41. 5 2. Lechat P, Mas, JL, Lascault, G, et al. Prevalence of patent foramen ovale in patients with stroke. New 6 England Journal of Medicine. 1988;318(18):1148-52. 7 8 3. Lamy C, Giannesini, C, Zuber, M, et al. Clinical and Imaging Findings in Cryptogenic Stroke Patients With 9 and Without Patent Foramen Ovale The PFO-ASA Study. Stroke. 2002;33:706-11. 10 4. Alsheikh-Ali A, Thaler, DE, Kent, DM. Patent foramen ovale in cryptogenic stroke: incidental or 11 pathogenic? Stroke. 2009;40:2349-55. 12 5. Ning M, Lo, EH, Ning, PC, et al. . The Brain’s Heart – Therapeutic Opportunities for Patent Foramen 13 14 Ovale (PFO) and Neurovascular Disease. Pharmacology & Therapeutics. 2013;139(2):111-23. 15 6. Meier B, Kalesan, B, Mattle, HP, et al. Percutaneous closure of patent foramen ovale in cryptogenic 16 embolism. New England JournalFor of Medicine.peer 2013;368:1083-91. review only 17 7. Carroll J, Saver, JL, Thaler, DE, et al. Closure of patent foramen ovale versus medical therapy after 18 19 cryptogenic stroke. New England Journal of Medicine. 2013;368:1092-100. 20 8. Furlan A, Reisman, M, Massaro, J, et al. Closure or medical therapy for cryptogenic stroke with patent 21 foramen ovale. New England Journal of Medicine. 2012;366:991-9. 22 9. Khan A, Bin-Abdulhak, AA, Sheikh, MA, et al. Device closure of patent foramen ovale versus medical 23 therapy in cryptogenic stroke: a systematic review and meta-analysis. JACC Cardiovascular Interventions. 24 25 2013;6(12):1316-23. 26 10. Kent D, Dahabreh, IJ, Ruthazer, R, et al. Device Closure of Patent Foramen Ovale After Stroke: Pooled 27 Analysis of Completed Randomised Trials. Journal of the American College of Cardiology. 2016;67(8):907-17. 28 11. Li J, Liu, J, Liu, M, et al. Closure versus medical therapy for preventing recurrent stroke in patients with 29 30 patent foramen ovale and a history of cryptogenic stroke or transient ischaemic attack. Cochrane Database 31 Syst Rev. 2015(9):CD009938.

32 12. Spencer F, Lopes, LC, Kennedy, SA, et al. Systematic review of percutaneous closure versus medical http://bmjopen.bmj.com/ 33 therapy in patients with cryptogenic stroke and patent foramen ovale. BMJ Open. 2014;4:e004282. 34 13. Messé S, Gronseth, G, Kent, DM, et al. Practice advisory: Recurrent stroke with patent foramen ovale 35 36 (update of practice parameter): Report of the Guideline Development, Dissemination, and Implementation 37 Subcommittee of the American Academy of Neurology. Neurology. 2016;87(8):815-21. 38 14. Mas J-L, Derumeaux, G, Guillon, B, et al. Patent foramen ovale closure or anticoagulation vs. 39 antiplatelets after stroke. New England Journal of Medicine. 2017;377:1011-21.

40 on September 27, 2021 by guest. Protected copyright. 41 15. Søndergaard L, Kasner, SE, Rhodes, JF, et al. Patent foramen ovale closure or antiplatelet therapy for 42 cryptogenic stroke. New England Journal of Medicine. 2017;377:1033-42. 43 16. Saver J, Carroll, JD, Thaler, DE, et al. Long-term outcomes of patent foramen ovale closure or medical 44 therapy after stroke. New England Journal of Medicine. 2017;377:1022-32. 45 17. Castellucci L, Cameron, C, LeGal, G, et al. Efficacy and safety outcomes of oral anticoagulants and 46 47 antiplatelet drugs in the secondary prevention of venous thromboembolism: systematic review and network 48 meta-analysis. British Medical Journal. 2013;347:f5133. 49 18. Siemieniuk R, Agoritsas, T, Macdonald, H, et al. Introduction to BMJ Rapid Recommendations. BMJ. 50 2016;354:i5191. 51 52 19. Kuijpers T, Spencer, FA, Siemieniuk, RA, et al. Patent foramen ovale closure plus antiplatelet therapy 53 versus antiplatelet or anticoagulation therapy alone for cryptogenic stroke: a clinical practice guideline. British 54 Medical Journal. 2018. 55 20. Mir H, Siemieniuk, RA, Agoritsas, T, et al. Percutaneous closure versus antiplatelet therapy vs 56 anticoagulation therapy in patients with patent foramen ovale and cryptogenic stroke: a protocol for a 57 58 systematic review and network meta-analysis 2017 [Available from: 59 http://www.crd.york.ac.uk/PROSPERO/display_record.asp?ID=CRD42017081567. 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 54 of 101

21. Unit) HHIR. Search Strategies for EMBASE in Ovid Syntax [Website]. BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 http://hiru.mcmaster.ca/hiru/HIRU_Hedges_EMBASE_Strategies.aspx [updated February 9, 2016. Available 2 3 from: http://hiru.mcmaster.ca/hiru/HIRU_Hedges_EMBASE_Strategies.aspx. 4 22. Unit) HHIR. Search Filters for MEDLINE in Ovid Syntax and the PubMed translation [Website]. 5 http://hiru.mcmaster.ca/hiru/HIRU_Hedges_MEDLINE_Strategies.aspx2016 [updated February 9, 2016. 6 Available from: http://hiru.mcmaster.ca/hiru/HIRU_Hedges_MEDLINE_Strategies.aspx. 7 8 23. Akl E, Sun, X, Busse, JW, et al. Specific instructions for estimating unclearly reported blinding status in 9 randomised trials were reliable and valid. Journal of Clinical Epidemiology. 2012;65:262-7. 10 24. Higgins J, Altman, DG, Gotzsche, PC, et al. The Cochrane Collaboration's tool for assessing risk of bias in 11 randomised trials. British Medical Journal. 2011;343:d5928. 12 25. Guyatt G, Oxman, AD, Vist, GE, et al. GRADE: an emerging consensus on rating quality of evidence and 13 14 strength of recommendations. British Medical Journal. 2008;336:924-6. 15 26. Brignardello-Petersen R, Bonner, A, Alexander, PE, et al. Advances in the GRADE approach to rate the 16 certainty in estimates fromFor a network peer meta-analysis. review Journal of Clinical only Epidemiology. 2018;93:36-44. 17 27. Guyatt G, Oxman, AD, Kunz, R, et al. Rating the quality of evidence-indirectness. Journal of Clinical 18 19 Epidemiology. 2011;64(12):1303-10. 20 28. Gelman A, Rubin, DB. Inference from Iterative Simulation Using Multiple Sequences. Statistical Science. 21 1992;7:457-72. 22 29. vanValkenhoef G, Dias, S, Ades, AE, et al. Automated generation of node-splitting models for 23 assessment of inconsistency in network meta-analysis. Research Synthesis Methods. 2016;7:80-93. 24 25 30. Dias S, Sutton, AJ, Ades, AE, et al. Evidence synthesis for decision making 2: a generalized linear 26 modeling framework for pairwise and network meta-analysis of randomized controlled trials. Medical Decision 27 Making. 2013;33(5):607-17. 28 31. vanValkenhoef G, Lu, G, deBrock, B et al. Automating network meta-analysis. Research Synthesis 29 30 Methods. 2012;3:285-99. 31 32. Guyot P, Ades, AE, Ouwens, MJ, Welton, NJ. Enhanced secondary analysis of survival data:

32 reconstructing the data from published Kaplan-Meier survival curves. BMC Med Res Methodol. 2012;12(9). http://bmjopen.bmj.com/ 33 33. Mas J, Arquizan, C, Lamy, C, et al. Recurrent Cerebrovascular Events Associated with Patent Foramen 34 Ovale, Atrial Septal Aneurysm, or Both. New England Journal of Medicine. 2001;345:1740-6. 35 36 34. Shariat A, Yaghoubi, E, Farazdaghi, M, et al. Comparison of medical treatments in cryptogenic stroke 37 patients with patent foramen ovale: A randomized clinical trial. Journal of Research in Medical Sciences. 38 2013;18(2):94-8. 39 35. Homma S, Sacco, RL, DiTullio, MR, et al. Effect of Medical Treatment in Stroke Patients With Patent

40 on September 27, 2021 by guest. Protected copyright. 41 Foramen Ovale: Patent Foramen Ovale in Cryptogenic Stroke Study. Circulation. 2002;105:2625-31. 42 36. Lee P, Song, JK, Kim, JS, et al. Cryptogenic Stroke and High-Risk Patent Foramen Ovale: The DEFENSE- 43 PFO Trial. Journal of the American College of Cardiology. 2018;doi: 10.1016/j.jacc.2018.02.046. 44 37. Vazquez F, Gonzalez, JP, LeGal, G, et al. Risk of major bleeding in patients receiving vitamin K 45 antagonists or low doses of aspirin. A systematic review and meta-analysis. Thrombosis Research. 2016;138:1- 46 47 6. 48 38. De Rosa S, Sievert, H, Sabatino, J, et al. Percutaneous closure versus medical treatment in stroke 49 patients with patent foramen ovale. Annals of Internal Medicine. 2018:1-11. 50 39. Mojadidi M, Elgendy, AY, Elgendy, IY, et al. Transcatheter Patent Foramen Ovale Closure After 51 52 Cryptogenic Stroke. JACC Cardiovascular Interventions. 2017;10(21):2227-30. 53 40. Vaduganathan M, Qamar, A, Gupta, A, et al. Patent Foramen Ovale Closure for Secondary Prevention 54 of Cryptogenic Stroke: Updated Meta-Analysis of Randomized Clinical Trials. The American Journal of 55 Medicine. 2017;S0002-9343(17):31219-26. 56 41. Shah R, Nayyar, M, Jovin, IS, et al. Device closure versus medical therapy alone for patent foramen 57 58 ovale in patients with cryptogenic stroke. Annals of Internal Medicine. 2018:1-11. 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 55 of 101 BMJ Open

32 http://bmjopen.bmj.com/ 33 34 35 36 37 38 39

Legend: BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 2 3 Table 1: Characteristics of patients in eligible studies 4 5 Table 2 Summary of findings of PFO closure plus antiplatelet therapy vs antiplatelet therapy 6 7 Table 3: Summary of findings of PFO closure plus antiplatelet therapy vs anticoagulation therapy 8 9 Table 4: Summary of findings of anticoagulation vs antiplatelet therapy 10 11 12 13 14 15 16 Appendix 2: Tables For peer review only 17 18 19 1) Table 1: Characteristics of eligible studies 20 21 2) Table 2: Direct and indirect estimates of effects for the network meta-analysis 22 23 24 25 26 27 28 29 30 31

32 http://bmjopen.bmj.com/ 33 34 35 36 37 38 39

0 36 PFO closure plus 50 Risk difference: 0.04 Device or procedure related per 1000 per 1000 antiplatelet therapy 51 (CI 95% 0.02 – 0.05) High adverse event increases device or Based on data from 3560 52 (Standardized to 1 year) @ procedure related adverse patients in 6 studies Difference: 36 more per 1000 patients 53 (CI 95% 23 more - 50 more) events 54 Follow up 3.9 years 55 Odds Ratio: 1.27 22 28 56 (CI 95% 0.4 - 4.52) per 1000 per 1000 Moderate There is probably little or Transient ischaemic attack 57 Direct evidence in NMA from Due to serious no difference in transient (Standardized to 5 years) 8 58 353 patients in 1 studies Difference: 6 more per 1000 imprecision ischaemic attack Follow up 5.3 years 59 (CI 95% 22 fewer* - 22 more) 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 62 of 101 BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 5 1 Pulmonary embolism - Odds Ratio: 9.09 per 1000 per 1000 Moderate There is probably little or modelling data from VTE (CI 95% 3.7 - 25.0) 2 Due to serious no difference in literature 3 indirectness9 pulmonary embolism (Standardized to 5 years) Difference: 4 more per 1000 4 (CI 95% 1 more - 13 more) 5 6 Odds Ratio: 291.0 0 0 (CI 95% 0.0 - 999.0) per 1000 per 1000 Moderate There is probably little or 7 Systemic embolism Direct evidence in NMA from Due to serious no difference in systemic 8 (Standardized to 5 years) 353 patients in 1 studies imprecision10 embolism 9 Difference: 0 fewer per 1000 Follow up 5.3 years (CI 95% 11 fewer - 11 more) 10 11 Note: The baseline risk for PFO closure was used to estimate the absolute effect in the anticoagulation arm. The baseline risk in the PFO arm was calculated based on 12 the median risk of the outcome in the six RCTs included. This allowed for consistency in the absolute effect estimate in the PFO closure arms in table 2 and 3 and the 13 14 anticoagulation arms in table 3 and 4 allowing for ease in comparison. 15 16 1. Imprecision: Very Serious. Wide confidence interval. Low number of events. 2. Indirectness: Serious. In addition toFor the direct evidence peer from randomised review trials in patients with PFOonly and a cryptogenic ischaemic stroke, we additionally 17 considered external evidence from randomised trials that assessed the impact of anticoagulation vs. antiplatelet therapy for the secondary prevention of 18 venous thromboembolism. Imprecision: Serious. Wide confidence intervals, includes both appreciable benefit and harm. 19 3. Imprecision: Serious. Wide confidence intervals, includes both appreciable benefit and harm. Low number of events. 20 4. Imprecision: Serious. Wide confidence interval, included a not important benefit. Low number of events. 5. Indirectness: Serious. In addition to the direct evidence from randomised trials in patients with PFO and a cryptogenic ischaemic stroke, we additionally 21 considered external evidence from randomised trials that assessed the impact of anticoagulation vs. antiplatelet therapy for the secondary prevention of 22 venous thromboembolism. 23 6. Risk of bias: Serious. Not clearly stated how this was measured or assessed with prolonged ECG monitoring. Also, it is not clear for all events whether it was 24 transient or persistent. 25 7. Risk of bias: Serious. Not clearly stated how this was measured or assessed with prolonged ECG monitoring. Also, it is not clear for all events whether it was transient or persistent. 26 8. Imprecision: Serious. Wide confidence interval, includes both appreciable benefit and harm. Low number of events. x 27 Defined as persistent according to the study definition or requiring a cardioversion attempt *: The calculated confidence interval using risk difference, because of 28 uncertainty in the point estimates, permits reductions greater than the point estimates in the PFO group. To avoid confusion, we have truncated to present the 29 maximum reduction as equal to the PFO event rate. 30 @: In the first year after procedure rather than 5 years VTE, venous thromboembolism; AF, atrial fibrillation; CHF, congestive heart failure 31

32 http://bmjopen.bmj.com/ 33 34 35 36 37 38 39

29 30 Amplatzer occluder. Antiplatelet 31 ASA 5-6 months. (67%) or

32 http://bmjopen.bmj.com/ 33 34 35 36 37 38 39

Legend: BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 2 3 Figure 1. PRISMA flow diagram of studies included in review treatment of patients with PFO and 4 5 cryptogenic stroke 6 7 Figure 2. Meta-regression curve based on the proportion of anticoagulant in the medical therapy arm 8 9 for ischaemic stroke 10 11 12 Figure 3. Combined Kaplan-Meier curves with individualized patient data based on the type of 13 14 intervention for ischaemic stroke 15 16 For peer review only 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31

32 http://bmjopen.bmj.com/ 33 34 35 36 37 38 39

BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 2

3 4 5 Records identified through 6 database searching 7 (total n = 997): 8 MEDLINE (n=326), 9 EMBASE (n=590), 10 11 CENTRAL (n=81) 12 13 Identification 14 15 16 ForRecords peerafter duplicates review removed only 17 (n = 772) 18 19 20 21 22 23 Records screened Records excluded 24 (n = 772) (n = 740) 25

26 Screening 27 28 29 Full-text articles assessed Full-text articles excluded, for eligibility with reasons 30 31 (n = 32) (n = 25)

32 http://bmjopen.bmj.com/ 33 34 35 Studies included in Study added (published 36 Eligibility after search completion) 37 qualitative synthesis 38 (n = 7) (n = 1) 39

40 on September 27, 2021 by guest. Protected copyright. 41

42 Studies included in 43 quantitative synthesis 44 (meta-analysis) 45 46 (n = 8)

47 Included 48 49 50 Figure 1. PRISMA flow diagram of studies included in review treatment of patients with PFO and cryptogenic 51 52 stroke 53 54 55 56 57 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 71 of 101 BMJ Open BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 For peer review only 17 18 19 20 21 22 23 Figure 2. Meta-regression curve of the proportion of patients prescribed an anticoagulant in the medical 24 25 26 therapy arm for the outcome ischaemic stroke 27 28 29 30 31

32 http://bmjopen.bmj.com/ 33 34 35 36 37 38 39

40 on September 27, 2021 by guest. Protected copyright. 41 42 43 44 45 46 47 48 Figure 3. Combined Kaplan-Meier curves of individual participant data in the included trials, by type of 49 50 intervention for the outcome ischaemic stroke 51 52 Note: 53 1) Y-axis is truncated from 90% to 100% Event-free Survival 54 2) Mixed medical therapy includes studies where the proportion of antiplatelet agents was less than 80% and the proportion of anticoagulant was greater 55 than 25% 3) Proportion of patients: PFO Closure + antiplatelet therapy (N= 1829; 50%), Mixed medical therapy (N= 1153; 32%), Antiplatelet therapy (N= 458; 13%), and 56 Anticoagulation (N= 210; 6%) 57 4) K-M curves were available for the outcome of ischaemic stroke in the CLOSE study (1) (PFO closure, anticoagulation, and antiplatelet), the PC Trial (2) (PFO 58 closure and mixed medical therapy), the RESPECT trial (3) (PFO closure and mixed medical therapy) , and the REDUCE trial (4) (PFO closure and 59 antiplatelet). 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 72 of 101

BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 Appendices 2 3 Appendix 3: Figures 4 1) Figure 1: Risk of bias of included randomised controlled trials 5 2) Figure 2: Network of included RCTs with available direct comparisons for ischaemic stroke. 6 3) Figure 3: Meta-regression evaluating the effect of moderate or higher shunt size on the effectiveness 7 8 of PFO closure versus medical therapy 9 4) Figure 4: Meta-regression evaluating the proportion of patients on anticoagulation compared to those 10 with moderate or higher shunt size 11 Appendix 4: PFO closure vs antiplatelet forest plots (Direct evidence only) 12 Figure 1: Ischaemic stroke 13 14 Figure 2: TIA 15 Figure 3: Systemic emboli 16 Figure 4: Death For peer review only 17 Figure 5: Major bleeding 18 19 Figure 6: Pulmonary embolism 20 Appendix 5: PFO closure vs anticoagulation forest plots (Direct evidence only) 21 Figure 1: Ischaemic stroke 22 Figure 2: TIA 23 Figure 3: Systemic emboli 24 25 Figure 4: Pulmonary emboli 26 Figure 5: Major bleeding 27 Appendix 6: Complications of PFO closure vs medical therapy forest plots (Direct evidence only) 28 Figure 1: All atrial fibrillation 29 30 Figure 2: Persistent atrial fibrillation 31 Figure 3: Transient atrial fibrillation

32 Figure 4: Serious device or procedure related http://bmjopen.bmj.com/ 33 Appendix 7: Anticoagulation vs antiplatelet forest plots (Direct evidence only) 34 Figure 1: Stroke 35 36 Figure 2: TIA 37 Figure 3: Major bleeding 38 Figure 4: Death 39

32 http://bmjopen.bmj.com/ 33 34 35 36 37 38 39

40 on September 27, 2021 by guest. Protected copyright. 41 Figure 1: Risk of bias of included randomised controlled trials 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 74 of 101 BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 Figure 2: Network of included RCTs with available direct comparisons for ischaemic stroke. 16 For peer review only 17 Note: The thickness of the lines (edges) is proportional to the number of RCTs evaluating each treatment and direct 18 19 comparison. 20 21 22 23 24 25 26 27 28 29 30 31

32 http://bmjopen.bmj.com/ 33 34 35 36 37 38 39

40 on September 27, 2021 by guest. Protected copyright. 41 42 43 44 45 46 47 Figure 3: Systemic embolism 48 49 50 51 52 53 54 55 56 57 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 77 of 101 BMJ Open BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 2 3 4 5 6 7 8 9 10 11 Figure 4: Death 12 13 14 15 16 For peer review only 17 18 19 20 21 22 23 24 25 26 27 Figure 5: Major bleeding 28 29 30 31

32 http://bmjopen.bmj.com/ 33 34 35 36 37 38 39

32 http://bmjopen.bmj.com/ 33 34 Figure 2: Persistent atrial fibrillation 35 36 37 38 39

32 http://bmjopen.bmj.com/ 33 34 35 36 37 38 39

40 on September 27, 2021 by guest. Protected copyright. 41 42 43 Figure 3: Major bleeding 44 45 46 47 48 49 50 51 52 53 54 55 Figure 4: Death 56 57 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 83 of 101 BMJ Open BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 PRISMA NMA Checklist of Items to Include When Reporting A Systematic Review 2 Involving a Network Meta-analysis 3 4 Section/Topic Item Checklist Item Reported 5 # on Page # 6 TITLE 7 Title 1 Identify the report as a systematic review incorporating a 1 8 network meta-analysis (or related form of meta-analysis). 9 10 11 ABSTRACT 2-3 12 Structured 2 Provide a structured summary including, as applicable: 13 summary Background: main objectives 14 Methods: data sources; study eligibility criteria, participants, 15 and interventions; study appraisal; and synthesis methods, 16 Forsuch peer as network review meta-analysis. only 17 Results: number of studies and participants identified; 18 19 summary estimates with corresponding confidence/credible 20 intervals; treatment rankings may also be discussed. Authors 21 may choose to summarize pairwise comparisons against a 22 chosen treatment included in their analyses for brevity. 23 Discussion/Conclusions: limitations; conclusions and 24 implications of findings. 25 Other: primary source of funding; systematic review 26 registration number with registry name. 27 28 29 INTRODUCTION 30 Rationale 3 Describe the rationale for the review in the context of what is 4-5 31 already known, including mention of why a network meta-

32 http://bmjopen.bmj.com/ analysis has been conducted. 33 34 Objectives 4 Provide an explicit statement of questions being addressed, with 5 35 reference to participants, interventions, comparisons, outcomes, 36 and study design (PICOS). 37 38 METHODS 39

40 Protocol and 5 Indicate whether a review protocol exists and if and where it can 5 on September 27, 2021 by guest. Protected copyright. 41 registration be accessed (e.g., Web address); and, if available, provide 42 registration information, including registration number. 43 Eligibility criteria 6 Specify study characteristics (e.g., PICOS, length of follow-up) 6 44 and report characteristics (e.g., years considered, language, 45 publication status) used as criteria for eligibility, giving 46 rationale. Clearly describe eligible treatments included in the 47 treatment network, and note whether any have been clustered or 48 49 merged into the same node (with justification). 50 Information 7 Describe all information sources (e.g., databases with dates of 6 51 sources coverage, contact with study authors to identify additional 52 studies) in the search and date last searched. 53 Search 8 Present full electronic search strategy for at least one database, 6, 35-39 54 including any limits used, such that it could be repeated. 55 56 57 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 84 of 101 BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 Study selection 9 State the process for selecting studies (i.e., screening, eligibility, 6-7 2 included in systematic review, and, if applicable, included in the 3 meta-analysis). 4 Data collection 10 Describe method of data extraction from reports (e.g., piloted 6 5 process forms, independently, in duplicate) and any processes for 6 obtaining and confirming data from investigators. 7 Data items 11 List and define all variables for which data were sought (e.g., 6 8 PICOS, funding sources) and any assumptions and 9 simplifications made. 10 Geometry of the S1 Describe methods used to explore the geometry of the treatment 8-9 11 network network under study and potential biases related to it. This 12 should include how the evidence base has been graphically 13 summarized for presentation, and what characteristics were 14 compiled and used to describe the evidence base to readers. 15 Risk of bias within 12 Describe methods used for assessing risk of bias of individual 7-8 16 For peer review only 17 individual studies studies (including specification of whether this was done at the 18 study or outcome level), and how this information is to be used 19 in any data synthesis. 20 Summary 13 State the principal summary measures (e.g., risk ratio, difference 9-11 21 measures in means). Also describe the use of additional summary 22 measures assessed, such as treatment rankings and surface 23 under the cumulative ranking curve (SUCRA) values, as well as 24 modified approaches used to present summary findings from 25 meta-analyses. 26 Planned methods 14 Describe the methods of handling data and combining results of 9-11 27 of analysis studies for each network meta-analysis. This should include, but 28 not be limited to: 29 • Handling of multi-arm trials; 30 • Selection of variance structure; 31 • Selection of prior distributions in Bayesian analyses; 32 http://bmjopen.bmj.com/ 33 and 34 • Assessment of model fit. 35 Assessment of S2 Describe the statistical methods used to evaluate the agreement 9-11 36 Inconsistency of direct and indirect evidence in the treatment network(s) 37 studied. Describe efforts taken to address its presence when 38 found. 39 Risk of bias across 15 Specify any assessment of risk of bias that may affect the 7-8

40 studies cumulative evidence (e.g., publication bias, selective reporting on September 27, 2021 by guest. Protected copyright. 41 within studies). 42 Additional 16 Describe methods of additional analyses if done, indicating 9-11 43 analyses which were pre-specified. This may include, but not be limited 44 to, the following: 45 • Sensitivity or subgroup analyses; 46 • Meta-regression analyses; 47 48 • Alternative formulations of the treatment network; and 49 • Use of alternative prior distributions for Bayesian 50 analyses (if applicable). 51 52 53 54 55 56 57 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 85 of 101 BMJ Open BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 RESULTS† 2 3 Study selection 17 Give numbers of studies screened, assessed for eligibility, and 6,7, 30 4 included in the review, with reasons for exclusions at each stage, 5 ideally with a flow diagram. 6 Presentation of S3 Provide a network graph of the included studies to enable 44 7 network structure visualization of the geometry of the treatment network. 8 Summary of S4 Provide a brief overview of characteristics of the treatment 44 9 network geometry network. This may include commentary on the abundance of 10 11 trials and randomized patients for the different interventions and 12 pairwise comparisons in the network, gaps of evidence in the 13 treatment network, and potential biases reflected by the network 14 structure. 15 Study 18 For each study, present characteristics for which data were 26 16 characteristics Forextracted peer (e.g., study review size, PICOS, follow-up only period) and 17 provide the citations. 18 Risk of bias within 19 Present data on risk of bias of each study and, if available, any 43 19 studies outcome level assessment. 20 Results of 20 For all outcomes considered (benefits or harms), present, for 39-52 21 individual studies each study: 1) simple summary data for each intervention group, 22 and 2) effect estimates and confidence intervals. Modified 23 approaches may be needed to deal with information from larger 24 25 networks. 26 Synthesis of results 21 Present results of each meta-analysis done, including 27-32 27 confidence/credible intervals. In larger networks, authors may 28 focus on comparisons versus a particular comparator (e.g. 29 placebo or standard care), with full findings presented in an 30 appendix. League tables and forest plots may be considered to 31 summarize pairwise comparisons. If additional summary

32 measures were explored (such as treatment rankings), these http://bmjopen.bmj.com/ 33 should also be presented. 34 Exploration for S5 Describe results from investigations of inconsistency. This may 8 35 inconsistency include such information as measures of model fit to compare 36 consistency and inconsistency models, P values from statistical 37 tests, or summary of inconsistency estimates from different parts 38 of the treatment network. 39 Risk of bias across 22 Present results of any assessment of risk of bias across studies 43 40 studies for the evidence base being studied. on September 27, 2021 by guest. Protected copyright. 41 42 Results of 23 Give results of additional analyses, if done (e.g., sensitivity or 34 43 additional analyses subgroup analyses, meta-regression analyses, alternative 44 network geometries studied, alternative choice of prior 45 distributions for Bayesian analyses, and so forth). 46 47 DISCUSSION 48 49 Summary of 24 Summarize the main findings, including the strength of evidence 18-22 50 evidence for each main outcome; consider their relevance to key groups 51 (e.g., healthcare providers, users, and policy-makers). 52 Limitations 25 Discuss limitations at study and outcome level (e.g., risk of 21 53 bias), and at review level (e.g., incomplete retrieval of identified 54 research, reporting bias). Comment on the validity of the 55 assumptions, such as transitivity and consistency. Comment on 56 57 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 86 of 101 BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 any concerns regarding network geometry (e.g., avoidance of 2 certain comparisons). 3 Conclusions 26 Provide a general interpretation of the results in the context of 22 4 other evidence, and implications for future research. 5 6 7 FUNDING 8 Funding 27 Describe sources of funding for the systematic review and other 22 9 support (e.g., supply of data); role of funders for the systematic 10 review. This should also include information regarding whether 11 funding has been received from manufacturers of treatments in 12 the network and/or whether some of the authors are content 13 experts with professional conflicts of interest that could affect 14 use of treatments in the network. 15 16 PICOS = population, Forintervention, peer comparators, review outcomes, study design.only 17 * Text in italics indicates wording specific to reporting of network meta-analyses that has been added to 18 guidance from the PRISMA statement. 19 † Authors may wish to plan for use of appendices to present all relevant information in full detail for 20 items in this section. 21 22 23 24 25 26 27 28 29 30 31

32 http://bmjopen.bmj.com/ 33 34 35 36 37 38 39

40 on September 27, 2021 by guest. Protected copyright. 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 87 of 101 BMJ Open BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 2 3 5 9 4 5 Ç t 6 7 í 8 9 ww .aW # 10 # .aW h% .# ## 11 # % 12 13 t# ## ' # # % í 14 # % 15 16 í # For # peer review only % 17 18 {# 19 20 21 22 23 9 24 25 L 26 27 w 28 í # ## % 29 30 According to the BMJ Rapid Recommendations process (18), a multi-professional guideline 31 panel provided critical oversight to the systematic review and identified populations, subgroups, 32 http://bmjopen.bmj.com/ 33 and outcomes of interest. The BMJ Rapid Recommendation panel, and in particular three people 34 who had experienced a cryptogenic stroke (two of whom had undergone PFO closure), assessed 35 the relative importance of the outcomes (19). The guideline panel submitted a list of possible 36 outcomes and then rated the importance of each outcome suggested on a scale from 1 (least 37 important) to 9 (most important). Items with scores of 4 or more were considered, and a final 38 decision was made by consensus, with special consideration of the views of the patient panellists. 39 The patients stressed the importance of several outcomes that other panel members had 40 on September 27, 2021 by guest. Protected copyright. 41 identified and uniquely highlighted the importance of detailed information about the device or 42 procedure related adverse events. 43 44 45 2) Further explanation would be helpful to make clear why you included prior systematic 46 47 reviews addressing antiplatelet versus anticoagulant therapy in patients with deep venous 48 thrombosis (DVT), atrial fibrillation, and heart failure earlier in the paper. 49 50 w í # ## 51 52 . tCh b 53 b ! " tCh 54 # $ % C ! " & " ' 55 56 / " $ ! ) * " 57 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 88 of 101 BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 2 3 ba! hw %.! /L % ) %0% í " $ 4 ! " w w % í " " 5 " ! " " 6 7 % 8 9 C " &! " &$ " $ & " tCh 2 10 "' ! '$ # "'' ëÇ9% Ç" $ 11 " " " " $ 12 " $ ëÇ9 ' " $ 13 & tCh% C 6 ' ! " " ëY!2 14 2 !{! " " ëÇ9! " ! 15 ' " ' ' & 16 For peer review only 17 " " tCh % 18 19 Ç"! " $ $ " 20 " &$ " &! 6 ' $ ' % í 21 " $ $ " w/Ç " (ëÇ9! 22 " ! ' % 23 24 25 " L #í w/Ç 26 ( ) ") "* + /[h{9 (W[a 1 27 ( ) Ç /[h{9 w Y 28 5 4 í 29 4 í 4 30 31 w

32 http://bmjopen.bmj.com/ 33 í # 34 % h# 5 a /[h{9 35 % Ç # # 36 + % 37

38 39 ) í #

40 11 + on September 27, 2021 by guest. Protected copyright. 41 5 42 43 1 1 4 L 1 4 44 w 45 46 í # % 47 48 49 * Ç 5 1 tCh 50 51 tCh . 1 52 53 Ç 54 1 1 55 1 1 ! 9 56 57 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 89 of 101 BMJ Open BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 2 3 4 5 6 w 7 8 í % í # # # 9 # # # # # # 10 11 ## tCh # # # 12 % í # 13 +# - ## .% 14 15 í # 16 % b #For # peer review ## only tCh # 17 # # # # 18 # L ## # ## # # 19 20 # tCh 21 # # # # # 22 #% 23 24 a % L # # 25 tCh # # # # 26 # % Ç # ## 27 28 tCh # 1# 1 2%23% 29 í ' # +# # 30 31 %

32 http://bmjopen.bmj.com/ 33 b # # 34 # tCh # % í 35 5 ## 36 37 Although the evidence regarding anticoagulation versus alternatives is of low quality and comes 38 in large part from indirect evidence, it is possible that anticoagulation and PFO closure may 39 achieve a similar reduction in the risk of recurrent ischaemic stroke – and thus appreciably

40 fewer strokes than antiplatelet agents alone. on September 27, 2021 by guest. Protected copyright. 41 42 43 44 Further, in the next section addressing the issue, we have added the qualifier “though 45 admittedly uncertain”: 46 47 Our results provide further strong support for the hypothesis that stasis-related paradoxical 48 venous thromboemboli and/or intra-cardiac thrombi cause a large proportion of cryptogenic 49 strokes in younger patients with PFO. Given that is the case, it is plausible, though admittedly 50 uncertain, that anticoagulation has a similar relative benefit versus antiplatelet agents in 51 preventing cryptogenic strokes in patients with PFO as it does in preventing recurrent VTE. 52 53 Another passage from the discussion avoids, we think, overstatement: 54 55 Only one of the RCTs, however, included an arm in which all patients were offered 56 anticoagulation or PFO closure and as a result the direct evidence includes relatively few 57 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 90 of 101 BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 2 3 patients and very few events. Moreover, the evidence from studies of VTE is very indirect. 4 Thus, we categorised the evidence regarding anticoagulation versus PFO closure, and versus 5 antiplatelet therapy, in preventing stroke as low certainty (Tables 3 and 4). 6 7 8 : ò 1 9 10 1 11 12 w 13 í 1 14 15 % í 16 # # For peer # review only - 4.% 17 18 19 < L 5 1 1 20 21 22 í # 5 ' # tCh 23 # # # # # # # # # 24 # # #6 # # W!// 829: 25 26 92% Ç # 27 # # #% í 28 -92 #.% 29 30 31 = Ç 5 1

32 http://bmjopen.bmj.com/ 33 w 34 35 Ç # +# Ç# 8 ; 3 36 #% 37 38 39 > h 1 a! 40 #aI on September 27, 2021 by guest. Protected copyright. 41 42 w í % t - < #. 43 44 aI < # 2%4 ' 1## 45 # -% >9?. # -. 46 822@.% í aI # # . 47 . # 9? . 48 # # 8 9 . < 49 50 ## 51 #% 52 53 í # # # # # # 54 55 ### -D t ! !9 h aW í# 56 bW% 9 # # # 57 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 91 of 101 BMJ Open BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 2 3 Y#1a # % .a/ a w a #% 8298698-5..% a 4 # ## % í ## ## 5 6 # # 7 # < # -8 3.% 8 9 10 11 @ L /[h{9 1 1 12 1 13 14 AÇ ba! 1 15 16 For peer review only 17 w 18 19 !# ba! # # # # 20 # ba! 21 ##% í # # 22 ## 23 24 Ç" ba! " 9 " ba! $ ! " 25 $ " ! " " 26 ! " " tCh % b "! 27 28 " ' " ba! " " ' 29 '% C "! " 6 $ " " 30 ' $ " " % C "! " " " 31 $ " &%

32 http://bmjopen.bmj.com/ 33 34 35 36 / 37 38 39 w 9

40 on September 27, 2021 by guest. Protected copyright. 41 w 42 43 / 44 45 46 C h# 1 47 Ç 1# tCh # 48 # # # # % h## 49 # % 50 51 52 C L # 1 53 Ç # ## # 54 #% 55 56 C { ## 57 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 92 of 101 BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 2 3 Ç w v ## # % 4 C h## 5 6 9<## % 7 C t 1 + # E 8 ò # # ## % 9 10 11 w 12 13 í % 14 15 16 For peer review only 17 I !C 1 C w9{t9/Ç 18 1 1 19 !C !C Ç 1 1 ! 20 !C 21 22 23 24 We include the following definition for persistent atrial fibrillation in the Methods section: “We 25 used the definitions in the original studies for all outcomes except for persistent atrial 26 fibrillation; our classification included the definition used in the original study but also included 27 patients who underwent a cardioversion. We also abstracted key outcomes from the external 28 evidence systematic reviews.” 29 30 31

32 A a C http://bmjopen.bmj.com/ 33 ò 5 1 34 35 D I ëÇ9 L 1 36 b 1 1 37 38 39 w

40 on September 27, 2021 by guest. Protected copyright. 41 í # G 8% 42 43 44 45 " / tCh 1 1 a 46 1 ëÇ9 47 Ç1 ) 1 48 49 w Ç tCh # 50 % L ## tCh ## 51 # % h ## tCh 52 # # # % 53 54 Ç # # # # Ç# 3 # 55 56 # # # hw < # 57 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 93 of 101 BMJ Open BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 2 3 2%94 -/L 2%2@ H 2%;8. 2%8@ -/L 2%23 H 9%8. % Ç # # 4 # # <# % 5 6 7 C w# 1 + E /#E í## E 8 9 ò % 10 11 C L # 1 # I E 12 a #E 13 14 ò ## % 15 16 C w 1 For # E peer ! #review E only 17 w % 18 C! II 1 # # E 19 20 ò 21 í % 22 23 24

25 26 ) ! 9 27 28 L 1 tCh 29 (W[a 1 # 30 1 b 31 1 4

32 http://bmjopen.bmj.com/ 33 w 34 35 5 a /[h{9 # # 36 # -Ia. -[D.% h # 37 1 % Ç # 38 39 ## 1 # 5 a%

40 on September 27, 2021 by guest. Protected copyright.

41 42 ! # 43 44 45 9% Ç # 46 % t 9J 3 K # # # [hí9w # KL 47 48 w Ç PFO closure plus antiplatelet therapy probably has a higher 49 risk of pulmonary embolism than anticoagulation 50 51 52 % t 9: ; K Çh # w/ÇKL 53 54 w / % 55 56 57 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 94 of 101 BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 2 3 4 8% !< Ç# 9 2 tCh # # # 5 6 ## ' # % { 7 # L # 8 # % 9 10 w Ç# 9 % 11 12 13 14 15 16 w 8 For peer review only 17 18 w 19 20 21 / 22 a ## 1# 23 ## 24 # -tCh.% 25 26 h## # 5 ' # # 27 tCh # % # # tCh # % # 28 # # % # % 29 Ç # tCh # 30 # # 31 # % a tCh # 32 http://bmjopen.bmj.com/ 33 # ## ## 1# # # 34 #% 35 36 9. Ç ## 37 38 Ç 1 39 1 L L F 1

40 1 on September 27, 2021 by guest. Protected copyright. 41 tCh C > 42 1 1 ( 1 43 44 Ç 1 1 % 45 46 47 w 48 49 í ' # ## 50 51 h $ " "% L " $ " w/Ç 52 tCh :2 ; $ " " tCh 53 " $! ! " $ % L ! 54 ' $ "! & 2 $ ! 2 ! 55 $ ! 6 # ! ' ' 56 57 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 95 of 101 BMJ Open BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 2 3 " 2& % L ! " Dw!59 " " 4 $ " " $% h $ 5 6 ' " $ % í " " " " $ ' 7 ! " " " 8 " tCh " " # $ &% 9 10 11 12 D ! 1 1 13 ! 14 1 1 1 15 1 tCh ("))@ )D>: =@G ! 16 For peer review only 17 #tCh HIJ 18 Ü 19 tCh 20 21 w 22 23 í # G 8 3% 24 25 26 " L 1 27 1 Ç 28 t/ w9{t9/Ç w95Ü/9 29 Ç =@G ( 30 31 1

32 http://bmjopen.bmj.com/ 33 w 34 h # # # 1 % 35 36 . # % b # # 37 tCh # # # 38 # ## # H # # # 39 #%

40 on September 27, 2021 by guest. Protected copyright. 41 42 ) L /[h{9 D> @ 43 44 tCh Ç 1 tCh 45 D@G 46 47 w 48 í /[h{9 5 a # 49 50 % D 9 -bM483. D 8 -bM985. 51 # D ; -bM92. tCh # % Ç <' 52 ## #% 53 54 55 * ! C ! * < 56 57 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 96 of 101 BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 2 3 tCh 1 4 1 1 { 5 6 1 1 7 8 9 w 10 11 Ç # !< 31@ ## % 12 Ç 1 1#% Ç# 8 # 13 1# 14 % 15 16 ! # ## For 1# peer review only # 17 # 1 1# -%% 1# .% Ç 18 # 19 # # # % Ç # # 20 21 # ## # tCh # % L /[h{9 22 # ## # # tCh # 23 # % Ç 24 1#% 25 26 Ç # # a ##% Ç # 27 /[h{9 # 1 # w/Ç 9. 1 # # ## 28 8. 1 # # # % # ## 29 30 tCh # ;. 1 # # # % tCh # ## 31 # % Ç 1 # # tCh # # #

32 # % w 1 # # http://bmjopen.bmj.com/ 33 1# % b 1# ## # 1 34 # # # 1 # -[# 82286 I 35 8298.% 36 37 C # < - # 38 39 1#. tCh tCh 1# # tCh # # % Ç 40 on September 27, 2021 by guest. Protected copyright. 41 + # tCh 1 42 # a ##% 43 44 45 [# Ç% b 1# % { % 8228 ! 46 ;2689-9J. 8;9;183% 47 I Wt W 5 . WY [ D ! !9 í Lw% / 48 49 1# # # 1 % w % 8298 W 50 96;-8. 5:1992% 51 52 : Ç L 53 1 ( .aW D@ <+"*:9*<" C 54 < 55 56 57 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 97 of 101 BMJ Open BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 2 3 w 4 5 í # # # 1 % í 6 # # #% 7 8 9 10 11 < Ç Ya 12 1 1 1 (Ya 13 14 1 1 /[h{Üw9 L 15 w w # Y# a # 16 For peer review only 17 < # # ## # # 18 - .% Ç # ## 19 ## # #% C 20 ## - 1 21 22 . ## # % Ç ## # 23 % 24 25 26 = C Ya 1 Ya 27 28 1 29 Ç 30 1 31 Ç 1

32 http://bmjopen.bmj.com/ 33 w9{t9/Ç 5 1 34 1 35 36 w 37 í G % ! <# # 38 39 # # % Ç

40 % on September 27, 2021 by guest. Protected copyright. 41 42 43 > Ç 44 tCh Ç 45 1 1 " : I 46 47 Ç 48 1 Ç 1 49 1 50 51 w 52 53 Ç G # # 1 54 # % í # # .aW 55 w w % 56 57 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 98 of 101 BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 2 3 4 @ Ç 1 tCh 1 1 5 6 1 7 1 8 9 w 10 11 w ' G # ## % 12 t # ## < # ## 1## # ## 13 ## ## # % 14 15 16 ! ) Ç For (t/ peer w9{t9/Ç review 1 only C 17 ( 1 t 18 19 !< 3 tCh # # # % Ç 20 # :2? # # % Ç # # 21 22 J@%9? t/ # @3%:? w9{t9/Ç % b 23 # #% Ç t/ w9{t9/Ç # 24 C # % 25 26 27 D ! : Ç 1 { (D@ < t 28 1 (b 9 W a D@ "+":= @>D @@ 29 30 w 31

32 ! # ## # # {# { 829@ -Ç# {5.% http://bmjopen.bmj.com/ 33 í # # !< J 91;% 34 35 36 37 w ; 38 39 w

40 on September 27, 2021 by guest. Protected copyright. 41 42 / 43 {# 1# tCh # % # 44 % í N # 45 tCh # # 1# # # # 1# # # 46 47 # # 1# 1 # 48 # # # ## < # ## 49 # 1% 50 51 { 52 53 Ç # # % 54 9% b D% ë% t# 5% { Y% a Y% ë Ç% { t% a # -829:.% 55 P/# t C h# ë a# Ç t í / { 56 57 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 99 of 101 BMJ Open BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 2 3 Ç L ! Ü { w a 1!#%P { 35-8. 398139:% 4 8% b ó% D% h1ò t% C% ò {% [% I ù% Ç% ù ù% I% ù -829:.% P/# 5 6 # 1# 7 ' #%P W b #% 8 ;% ! Ç% !% !% I # a% t !% W !% . # Ç% Ç## I% Ç Ç% { [% ! /% 9 [% D {% .# -829:.% Pa 1!# / t C h# /# ë 10 a# Ç t w / { %P ! W / # 989-4. J351J44% 11 12 3% 5 w {% I% { W% { !% t # {% { /% L # -829:.% Pt 13 /# ë a# Ç { t í t C h# ! { w 14 a 1#%P ! L a% 15 4% ! 1{# 9% .% / Ç% I# 9% !% . I% !# a% [ -829:.% Pt 16 # # #For peer review only 1# ' 17 18 ## #%P W b #% 19 J% { w% a% b L% {% W !% w .% w% . Ç% a% C a% t% C# {% ë% w -829:.% 20 P5 /# ë a# Ç !# t C h# t í / 21 { ! { w a 1#%P ! L a% 22 23 w í # w 8G 24 % 25 26 27 28 w 3 29 30 31 w

32 http://bmjopen.bmj.com/ 33 / 34 L + L # # /## t 35 D# L # # # ## % 36 37 38 a # 39

40 Ç {w Ç 1 1 on September 27, 2021 by guest. Protected copyright. 41 42 43 w í # # ba! ## 44 45 Ç" ba! " 9 " & $ ! " 46 $ " ! " " " 47 48 ! " " tCh% 49

50 51 52 D í 1 1 53 I {w 54 1 1 55 56 57 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 100 of 101 BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 2 3 w í # # w 8G % 4 5 6 " L ba! L 1 C 1 7 8 9 Ç ## # -!< 8 Ç# 8. 10 5 L ba! 11 h / 12 Hhw>*G/ LJ Hhw>*G/ LJL Hhw>*G/ LJ 13 tChT # # % b 2%98 -2%23 2%93 -2%24 2%;3. 14 # # #C 2%8@. 15 tChT # # % 5%J19J -8%21 2%34 -2%2J 2%33 -2%2: L 16 For # peer review3@ J%;T9J. only ;%9:. ;%@J. 17 ! # # % 3%J@ -2%:2 b ;%@; -2%:3 18 19 # 8J%54. #C 8:%99. 20 tChT # # % ;%9T93 -3%;8 b 8%3: -2%9J 21 # # :%4TJ8. #C 55%99. 22 tChT # # % ;%312J -9%41 2%5@ -2%28 2%45 -2%29 5 23 # 93 @8:%;. 3@%@@. 88%@J. 24 ! # # % 2%@8 -2%29 b 2%8; -2%29 25 # 89%5;. #C ;%33. 26 tChT # # % b 2%3: -2%82 27 2%43 -2%89 9%32. 28 # # #C 9%92. a9 1 tChT # # % 2%;2 -2%2: 2%84 -2%2@ 29 2%25 -2%22 8%94. 30 # 9%98. 2%:8. 31 ! # # % b 2%4; -2%9: 2%J2 -2%95 9%:2. 32 # #C 9%3@. http://bmjopen.bmj.com/ 33 tChT # # % b 2%:8 -2%;9 34 2%@4 -2%8J 8%9;. 35 # # #C 8%25. tChT # # % 2%:: -2%99 9%84 -2%;5 36 9%49 -2%;3 J%@8. 37 w ÇL! # J%:4. 3%;:. 38 ! # # % b 9%4; -2%42 9%8: -2%8J J%;8. 39 # #C 4%89.

40 tChT # # % b 2%@; -2%98 on September 27, 2021 by guest. Protected copyright. 2%@5 -2%98 J%@J. 41 # # #C 4%48. 42 tChT # # % 2%29 -9%3192 2%2J -4%@125 2%2; -3%21 43 { 1 # 9%2T2J. J%2T24. 2: 98:2%2. 44 ! # # % 2%29 -9%3192 b 2%23 -4%31 45 46 # 3%2T24. #C 2: 9J9J%2. 47 U L # 1## 48 49 C / # 50 51 52 53 ) L 5 1 ( 1 Ya 54 a 55 C 56 57 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 101 of 101 BMJ Open BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 2 3 í "$ " " $ ' " " $ 4 &% Ç"! # " " tCh % " $ ' " w/Ç 5 6 " " $ " " $ " " 7 " $ " " % Ç # " ! 2 2 8 tCh % " $! " " " " $ " 9 " '% í 2 2 " " 10 " '$ " " % 11 12 C Y1a # 13 14 C " '$ tCh 15 " $! " $ ! ! " $ " 16 For peer review only 17 " "$ % 9" ' $ " % 18

19 20 21 22 23 24 25 26 27 28 29 30 31

32 http://bmjopen.bmj.com/ 33 34 35 36 37 38 39

ForJournal: peerBMJ Open review only

Manuscript ID bmjopen-2018-023761.R1

Article Type: Research

Date Submitted by the Author: 27-Jun-2018

Complete List of Authors: Mir, Hassan; McMaster University, Medicine; 2. Department of Medicine, University of Toronto, Toronto, Ontario, Canada Siemieniuk, Reed; Mcmaster University, Clinical Epidemiology and Biostatistics; University of Toronto, Department of Medicine Ge, Long; The First Clinical Medical College of Lanzhou University, Foroutan, Farid; University Health Network, Medicine/Cardiology Fralick, Michael; University of Toronto, Eliot Phillison Clinician Scientist Training Program, Department of Medicine Syed, Talha; McMaster University Department of Medicine, Medicine Lopes, Luciane; Universidade de Sorocaba Kuijpers, Ton; Nederlands Huisartsen Genootschap, Guideline development and research http://bmjopen.bmj.com/ Mas, Jean-Louis; Sainte-Anne Hospital, Neurology Vandvik, Per; Norwegian Knowledge Centre for the Health Services, Agoritsas, Thomas; University Hospitals of Geneva, Division of General Internal Medicine & Division of Clinical Epidemiology; McMaster University Faculty of Health Sciences, Department of Health Research Methods, Evidence, and Impact Guyatt, Gordon; Mcmaster University, Clinical Epidemiology and Biostatistics on September 27, 2021 by guest. Protected copyright. Primary Subject Cardiovascular medicine Heading:

Evidence based practice, Medical management, Neurology, Patient-centred Secondary Subject Heading: medicine, Research methods

Cryptogenic stroke, Patent foramen ovale, Anticoagulation < Keywords: HAEMATOLOGY, Antiplatelet, PFO Closure

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 1 of 56 BMJ Open

Patent foramen ovale closure, antiplatelet therapy or anticoagulation in patients with patent BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 2 3 foramen ovale and cryptogenic stroke: a systematic review and network meta-analysis 4 5 incorporating complementary external evidence 6 7 Hassan Mir1,2, internal medicine and cardiology physician, Reed A.C. Siemieniuk1,2, PhD student and 8 3 4,5 6 9 internal medicine physician, Long Ge , PhD student, Farid Foroutan , PHD student, Michael Fralick , 10 1 11 research fellow and internal medicine physician, Talha Syed , internal medicine and cardiology 12 physician, Luciane Cruz Lopes7, clinical pharmacologist, Ton Kuijpers8, clinical epidemiologist, Jean- 13 14 Louis Mas9, professor, Per O. Vandvik10,11, associate professor, Thomas Agoritsas1,12, assistant 15 1 16 professor, Gordon H. GuyattFor, distinguished peer professorreview only 17 18 1. Department of Health Research Methods, Evidence, and Impact, McMaster University, 1280 Main St West, Hamilton, Ontario, Canada L8S 4L8 19 2. Department of Medicine, University of Toronto, Toronto, Ontario, Canada 20 3. Evidence-Based Medicine Center, School of Basic Medical Sciences, Lanzhou University, Lanzhou 730000, China 21 4. Department of Clinical Epidemiology and Biostatistics, McMaster University, 1280 Main St West, Hamilton, Ontario, Canada L8S 4L8 22 5. Heart Failure/Transplant Program, Toronto General Hospital, University Health Network, Toronto, Ontario, Canada 23 6. Eliot Phillipson Clinician Scientist Training Program, Department of Medicine, University of Toronto, Toronto, Canada, M5G 2C4 24 7. University of Sorocaba, UNISO, Sorocaba, Sao Paulo, Brazil 25 8. Department of guideline development and research, Dutch College of General Practitioners, Utrecht, The Netherlands 26 9. Sainte-Anne Hospital, Department of Neurology, 1 rue Cabanis, 75015 Paris. 27 10. Norwegian Institute of Public Health, Oslo, Norway 28 11. Department of Medicine, Innlandet Hospital Trust - division Gjøvik, Norway 29 12. Division General Internal Medicine & Division of Clinical Epidemiology, University Hospitals of Geneva, Rue Gabrielle-Perret-Gentil 4, CH-1211, Geneva, Switzerland 30 31 Corresponding author:

32 http://bmjopen.bmj.com/ Dr Hassan Mir 33 1280 Main St West, Hamilton, Ontario, Canada L8S 4L8 34 [email protected] 35 36 37 38 39

Conclusions: In patients less than 60 years old, PFO closure probably confers an important BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 2 3 reduction in ischaemic stroke recurrence compared to antiplatelet therapy alone but may make no 4 5 difference compared to anticoagulation. PFO closure incurs a risk of persistent atrial fibrillation and 6 7 device-related adverse events. Compared to alternatives, anticoagulation probably increases major 8 9 bleeding. 10 11 12 13 14 Article Summary: 15 16 Strengths and limitationsFor of this study peer review only 17 18 - Summarized the highest quality available evidence by addressing three management 19 20 alternatives of PFO closure plus antiplatelet therapy, anticoagulation alone, and antiplatelet 21 22 therapy alone. 23 - Used network meta-analysis, meta-regression, external evidence, and individual patient 24 25 analyses from survival curves to assist in decision-making. 26 27 - Applied GRADE approach to thoroughly assess certainty of evidence. 28 - 29 Results were limited primarily by available evidence. 30 - Within the medical therapy arm, the decision to use antiplatelet or anticoagulation was left to 31

32 the discretion of the physician and patient. Due to this, three studies could not be used in the http://bmjopen.bmj.com/ 33 34 primary analysis. 35 36 37 38 39

32 The study protocol was registered with PROSPERO: CRD42017081567 (20). http://bmjopen.bmj.com/ 33 34 35 36 37 38 39

Patient and Public involvement BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 2 According to the BMJ Rapid Recommendations process (18), a multi-professional guideline panel 3 4 provided critical oversight to the systematic review and identified populations, subgroups, and 5 6 7 outcomes of interest. The BMJ Rapid Recommendation panel, and in particular three people who had 8 9 experienced a cryptogenic stroke (two of whom had undergone PFO closure), assessed the relative 10 11 importance of the outcomes (19). The guideline panel submitted a list of possible outcomes and then 12 13 14 rated the importance of each outcome on a scale from 1 (least important) to 9 (most important). For 15 16 items with scores of 4, theFor panel consideredpeer furtherreview and made onlya final decision by consensus, with 17 18 special consideration of the views of the patient panellists. The patients stressed the importance of 19 20 21 several outcomes that other panel members had identified and uniquely highlighted the importance of 22 23 detailed information about the device or procedure related adverse events. 24 25 26 Search strategy 27 28 29 30 A search from a previous systematic review that we judged as comprehensive included research 31

32 http://bmjopen.bmj.com/ 33 34 35 Statistical Analysis 36 37 38 We analysed patients in groups to which they were randomised. For both pairwise and NMA, we 39

The PFO closure arm was used for the baseline risk to allow consistency across comparisons BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 2 3 involving PFO closure versus antiplatelet agents and anticoagulants. Baseline risk estimates 4 5 represent the median risk of the outcome in the six RCTs of PFO closure. For comparisons of 6 7 antiplatelet agents and anticoagulants, the antiplatelet arm provided the baseline risk to estimate the 8 9 absolute effect estimate in the anticoagulation arm. The baseline risk for the antiplatelet arm (10%) 10 11 12 was calculated using the baseline risk in the PFO closure arm. This proved similar to the median of 13 14 three studies that included an antiplatelet arm (9%) and maintained consistency with the PFO closure 15 16 comparisons. For peer review only 17 18 19 20 Network meta-analysis 21 22 To compare effects of alternative medical therapies, for ischaemic stroke, recurrent TIA, death, major 23 24 bleeding, and systemic embolism, we conducted a NMA of RCTs within a Bayesian hierarchical fixed 25 26 27 effect framework with non-informative priors and adjusted for correlation between effects in the multi- 28 29 arm trial. We generated posterior samples using Markov Chain Monte-Carlo (MCMC) simulation 30 31 technique running the analysis in three parallel chains. We used 100,000 burn-in simulations to allow

40 on September 27, 2021 by guest. Protected copyright. 41 42 have a larger benefit with PFO closure than patients with a small shunt size and that patients with an 43 44 atrial septal aneurysm will have a larger benefit with PFO closure than patients without an atrial septal 45 46 aneurysm (33). 47 48 49 50 RESULTS 51 52 53 Our systematic literature search yielded 772 potentially relevant publications; ten trials presented in 54 55 56 eight reports ultimately proved eligible (Figure 1, Appendix 2: eTable1) (6, 8, 14-16, 34-36). Seven 57 58 eligible multicenter RCTs published in 2013 or later comparing PFO closure plus antiplatelet therapy 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 12 of 56

versus medical therapy enrolled 3913 patients (range per RCT: 120 to 980 patients) (Table 1). The BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 2 3 CLOSE trial included three separate randomized trials (14). In the first trial, patients eligible for all 4 5 options were randomised to PFO closure (N = 173), antiplatelet therapy (N=171), or anticoagulation 6 7 with a vitamin K antagonist or direct oral anticoagulant (N=180; 93% of which were a VKA). In the 8 9 second trial, patient’s ineligible for anticoagulation were randomised to PFO closure (N=65) or 10 11 12 antiplatelet therapy (N=64). In the third trial, patients with contraindications to PFO closure were 13 14 randomised to antiplatelet therapy (n=7) or anticoagulation (n=3). Of the remaining eligible RCTs, one 15 16 specified that all patientsFor in the group peer who were review not randomised only to PFO closure were treated with 17 18 19 antiplatelet therapy (N=664) (15). Four RCTs allowed the treating physician to choose between 20 21 antiplatelet and/or anticoagulation therapy for the patients randomised to medical therapy: a minority 22 23 of these patients (range 20% to 34%) were prescribed anticoagulation at the time of discharge from 24 25 26 initial hospital admission (6, 8, 16, 36). The last three studies compared antiplatelet therapy to 27 28 anticoagulation (N=503) (14, 34, 35). One study reported a composite endpoint for ischaemic stroke 29 30 and TIA; we therefore could not use the data in our meta-analyses of stroke (N=44) (34). However, 31 32 we were able to use data from this RCT in analyses of death and major bleeding. Another RCT http://bmjopen.bmj.com/ 33 34 35 reported only a composite outcome of stroke and death (N=98) (35), with non-significant results 36 37 suggesting a possible reduction with anticoagulation compared to antiplatelet therapy (HR 0.52, CI 38 39 0.16 to 1.67, p-value 0.28).

40 on September 27, 2021 by guest. Protected copyright. 41 42 43 There is risk of bias in most studies due to a lack of blinding of medical personnel or patients 44 45 regarding the placement of a PFO closure device (Appendix 3: eFigure 1). Also, half of the studies 46 47 had incomplete data; we contacted the authors of three RCTs for additional outcome data (14, 34, 48 49 50 35). The lead author of the CLOSE study (JLM) provided additional unpublished data however the 51 52 authors of the two remaining studies did not respond. 53 54 55 56 We identified a recent rigorous systematic review of 12 trials and 11 999 participants for secondary 57 58 prevention of VTE (17) and a recent rigorous systematic review of 15 trials and 4982 participants 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 13 of 56 BMJ Open

across indications that provided a trustworthy estimate for an increase in major bleeds (37); both BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 2 3 reviews compared anticoagulant to antiplatelet therapy. 4 5 6 Table 1 presents the characteristics of the eligible studies. In most, the mean age was approximately 7 8 45 years, approximately 50 to 60% were male, and approximately 50%, or more, had moderate or 9 10 11 larger shunt. The median follow-up was 3.9 years with a range of 1.2 to 5.9 years. 12 13 14 Table 1: Characteristics of patients in eligible studies 15 Author n randomised Mean % Inclusion Moderate Atrial Most common 16 For peerAge Male reviewcriteria or onlyhigher septal device used for 17 shunt (%) aneurysm closure 18 (1) > 10mm 19 (%) (7) 20 PFO closure plus antiplatelet vs antiplatelet therapy 21 Furlan (CLOSURE 1, 909 46.0 51.8 Cryptogenic 52.9 (2) 37.8 STARFlex 100% 22 2012) stroke, PFO, > 23 18yo and < 24 60yo 25 † 26 Mas (CLOSE, 2017) 473 43.4 59.0 Cryptogenic 92.5 (3) 31.8 Amplatzer 52% 27 stroke, PFO, > 28 16yo and < 29 60yo 30 Meier (PC Trial, 414 44.5 49.8 Cryptogenic 65.6 (4) 23.7 (8) Amplatzer 100% 31 2013) stroke, PFO, > 18yo and < 32 http://bmjopen.bmj.com/ 33 60yo 34 Saver (RESPECT, 980 45.9 54.7 Cryptogenic 48.8 (5) 35.7 (9) Amplatzer 100% 35 2017) stroke, PFO, > 36 18yo and < 37 60yo 38 Sondergaard 664 45.1 60.6 Cryptogenic 81.0 (2) NR for AP Cardioform 61%¥ 39 (REDUCE, 2017) stroke, PFO, > group

40 18yo and < on September 27, 2021 by guest. Protected copyright. 41 60yo 42 Lee (DEFENCE PFO, 120 51.5 55.8 Cryptogenic NA 10.8 Amplatzer 100% 43 2018) stroke, PFO, 44 no age limit 45 PFO closure plus antiplatelet vs anticoagulation 46 Mas (CLOSE, 2017) 353 NA NA Cryptogenic NA NA NA 47 stroke, PFO, > 48 16yo and < 49 60yo 50 Anticoagulation vs antiplatelet therapy 51 Homma (PICSS, 203 (98 with 57.9 59.1 PFO with or 41.4% (6) 11.5% NA 52 2002) cryptogenic without 53 54 stroke) cryptogenic 55 stroke 56 >30yo and < 57 85yo 58 Mas (CLOSE, 2017) 361 44.2 57.0 Cryptogenic NA NA NA 59 stroke, PFO, > 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 14 of 56

16yo and < BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 60yo 2 Shariat (2013) 44 61.4 63.6 Cryptogenic NA NA NA 3 stroke, PFO, > 4 18yo 5 1. Shunt size was measured based on the number of microbubbles in the left atrium within 3 cycles of 6 7 being seen in the right atrium on transthoracic or transoesophageal echocardiography 8 2. Greater than or equal to 25 Microbubbles 9 3. Greater than or equal to 30 Microbubbles 10 4. Greater than or equal to 20 Microbubbles 11 12 5. Size of shunt not clearly defined 13 6. At least 1 microbubble 14 7. Atrial septal aneurysm was assessed on transoesophageal echo and was defined as septal mobility or 15 protrusion 16 8. Atrial septal aneurysmFor greater peer than 15mm review only 17 18 9. Atrial septal aneurysm not clearly defined 19 † 13% Intrasept PFO occluder, 9% STARFlex septal closure System, 9% Premere, 6% Amplatzer cribriform 20 occluder, 6% Figulla flex II PFO occluder, 1% Atriasept II occluder, 1% Gore helex septal occluder, 1% 21 Amplatzer AS occluder, 1% Figulla flex II UNI occluder, 1% Figulla flex II ASD occluder 22 23 ¥ 39% Gore helex septal occluder 24 NA Not available 25 26 27 PFO closure plus antiplatelet therapy versus antiplatelet therapy alone 28 29 30 31 Table 2 provides estimates of effect and certainty of evidence for all patient-important outcomes,

32 http://bmjopen.bmj.com/ 33 including, as presented subsequently, complications of PFO-closure. PFO closure plus antiplatelet 34 35 therapy reduced the risk of ischaemic stroke compared to antiplatelet therapy alone (NMA OR 0.12, 36 37 38 0.04 to 0.27; risk difference per 1000 patient-years followed for 5 years [RD]: -87, 95% confidence 39

40 interval [CI] -100 to -33; moderate certainty) (Appendix 4: eFigure 1). on September 27, 2021 by guest. Protected copyright. 41 42 43 44 Along with the three RCTs in which all medical patients received only antiplatelet therapy (14, 15, 36), 45 46 we included a meta-regression model data from an additional 5 RCTs that enrolled patients in which 47 48 more than 20% of medically treated patients received anticoagulation (Figure 2) (6, 8, 14, 16). The 49 50 reduction in stroke with PFO closure decreased as the proportion of patients receiving anticoagulation 51 52 53 in the medical therapy arm increased (p = 0.036). 54 55 56 Figure 3 presents a Kaplan-Meier curve of individual patient time-to-event data for each of the 57 58 59 randomised interventions, including PFO closure plus antiplatelet versus antiplatelet therapy alone 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 15 of 56 BMJ Open

and is again consistent with large relative reductions in hazard with PFO closure versus antiplatelet BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 2 3 therapy and smaller benefits of PFO closure when compared to mixed populations of patients 4 5 receiving anticoagulants and antiplatelet agents. 6 7 8 PFO closure did not appear to reduce the risk of TIA compared to antiplatelet therapy alone: NMA OR 9 10 11 0.82, CrI 0.32, 2.11; RD -6, -34 to +15; moderate certainty (Table 2, Appendix 4: eFigure 2). 12 13 14 Systemic emboli were rare: there were 5 events in one study (15), and none in another (14); we 15 16 therefore calculated the riskFor difference peer directly. review There was no importantonly difference between PFO 17 18 19 closure and antiplatelet therapy: NMA OR 0.83, CrI 0.12, 7.25; RD -1, -6 to +4 per 1,000 patients; 20 21 high certainty (Table 2, Appendix 4: eFigure 3). 22 23 24 Death occurred infrequently: in the PFO closure group a median of 9 per 1000 patients died over a 25 26 27 period of 5 years. The relative effect estimate for mortality was very imprecise, and the absolute 28 29 estimates included a small increase or decrease in deaths: NMA OR 3.28, CrI 0.2 to 174.22; RD +6, 30 31 CI -3 to +9; moderate certainty (Table 2, Appendix 4: eFigure 4).

32 http://bmjopen.bmj.com/ 33 34 35 The risk of major bleeding did not differ significantly between the PFO closure plus antiplatelet 36 37 therapy and antiplatelet therapy alone groups: NMA OR 0.48, CI 0.20 to 1.12; RD -7, CI -14 to +1, 38 39 moderate certainty (Table 2, Appendix 4: eFigure 5). Pulmonary embolism was rare and no more

40 on September 27, 2021 by guest. Protected copyright. 41 42 frequent in the PFO closure group (3/679, 0.4%) than the antiplatelet therapy alone group (1/458, 43 44 0.2%). No additional NMA evidence was available for pulmonary embolism because none of the 45 46 RCTs that included an anticoagulation arm reported pulmonary emboli; thus, we present a pair-wise 47 48 49 estimate for PFO closure vs. antiplatelet therapy alone: Peto OR 1.01, CI 0.09, 11.21; RD +0, CI -5 to 50 51 +48, high certainty (Table 2, Appendix 4: eFigure 6). 52 53 54 55 56 57 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 16 of 56

Table 2: GRADE summary of findings of PFO closure plus antiplatelet therapy vs antiplatelet therapy alone in BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 patients with cryptogenic stroke 2 3 Absolute effect estimates per 1000 Certainty in effect Outcome Study results and patient-years 4 estimates Plain text summary (Timeframe) measurements 5 Antiplatelet PFO closure plus (Quality of evidence) 6 therapy antiplatelet therapy 7 Odds Ratio: 0.12 100 13 PFO closure plus 8 (CI 95% 0.04 - 0.27) per 1000 per 1000 Moderate antiplatelet therapy Ischaemic stroke 9 Direct evidence in NMA from Due to serious probably results in a large (Standardized to 5 years) 10 1257 patients in 3 studies Difference: 87 fewer imprecision1 decrease in ischemic 11 Follow up 3.8 years (CI 95% 100 fewer* - 33 fewer) stroke 12 13 Odds Ratio: 3.28 3 9 (CI 95% 0.2 - 174.22) per 1000 per 1000 Moderate 14 Death There is probably little or Direct evidence in NMA from Due to serious (Standardized to 5 years) no difference in death 15 1257 patients in 3 studies Difference: 6 more imprecision2 16 FollowFor up 3.8 years peer (CI 95%review 3 fewer* - 9 more) only 17 18 Odds Ratio: 0.48 14 7 (CI 95% 0.2 - 1.12) per 1000 per 1000 Moderate There is probably little or 19 Major bleeding Direct evidence in NMA from Due to serious no difference in major 20 (Standardized to 5 years) 3 21 1257 patients in 3 studies Difference: 7 fewer imprecision bleeding Follow up 3.8 years (CI 95% 14 fewer* - 1 more) 22 23 5 23 Relative risk: 4.84 PFO closure plus 24 x per 1000 per 1000 Persistent atrial (CI 95% 1.91 - 12.26) antiplatelet therapy 25 Moderate fibrillation or Flutter Based on data from 3560 4 probably increases 26 Difference: 18 more per 1000 Due to serious risk of bias (Standardized to 1 year) patients in 6 studies @ persistent atrial 27 Follow up 3.9 years patients fibrillation 28 (CI 95% 5 more - 56 more) 29 5 17 Relative risk: 3.76 30 per 1000 per 1000 PFO closure plus Transient or paroxysmal (CI 95% 1.74 - 8.1) 31 Moderate antiplatelet therapy atrial fibrillation or flutter Based on data from 3560 Difference: 12 more per 1000 Due to serious risk of bias5 probably increases 32 (Standardized to 1 year) patients in 6 studies http://bmjopen.bmj.com/ @ transient atrial fibrillation 33 Follow up 3.9 years patients 34 (CI 95% 3 more - 31 more) 35 0 36 PFO closure plus 36 Risk difference: 0.04 per 1000 per 1000 Device or procedure antiplatelet therapy (CI 95% 0.02 – 0.05) 37 related adverse events High6 increase device or Based on data from 3560 Difference: 36 more per 1000 38 (Standardized to 1 year) procedure related adverse patients in 6 studies @ 39 patients events Follow up 3.9 years (CI 95% 23 more - 50 more) 40 on September 27, 2021 by guest. Protected copyright. 41 Odds Ratio: 1.01 5 5 42 (CI 95% 0.09 - 11.21) per 1000 per 1000 PFO plus antiplatelet Pulmonary embolism High 43 Direct evidence in NMA from therapy has no effect on (Standardized to 5 years) 44 1137 patients in 2 studies Difference: 0 fewer pulmonary embolism 45 Follow up 4.3 years (CI 95% 5 fewer - 48 more) 46 47 Odds Ratio: 0.82 34 28 (CI 95% 0.32 - 2.11) per 1000 per 1000 Moderate There is probably little or Transient ischaemic attack 48 Direct evidence in NMA from Due to serious no difference in transient (Standardized to 5 years) 49 1257 patients in 3 studies Difference: 6 fewer imprecision7 ischaemic attack 50 Follow up 3.8 years (CI 95% 34 fewer* - 15 more) 51 52 Odds Ratio: 0.83 6 5 (CI 95% 0.13 - 7.25) per 1000 per 1000 There is little or no 53 Systemic embolism High Direct evidence in NMA from difference in systemic 54 (Standardized to 5 years) 1257 patients in 3 studies embolism 55 Difference: 1 fewer Follow up 3.8 years (CI 95% 6 fewer* - 4 more) 56 57 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 17 of 56 BMJ Open

32 http://bmjopen.bmj.com/ 33 34 low certainty). Figure 3 presents a Kaplan-Meier curve of individual patient time-to-event data for 35 36 each of the randomised interventions, including PFO closure plus antiplatelet versus anticoagulation 37 38 alone. 39

40 on September 27, 2021 by guest. Protected copyright. 41 42 Results from the meta-regression analysis described previously suggest that PFO closure was less 43 44 effective relative to medical therapy as the proportion of patients receiving anticoagulation increased 45 46 (Figure 2). 47 48 49 50 To complement the above analyses, we added the effect estimate from a systematic review of RCTs 51 52 comparing low-dose ASA to anticoagulation with a VKA for the secondary prevention of VTE to the 53 54 network meta-analysis (17). Similar to the primary analysis, the analysis with the external evidence 55 56 57 failed to show a difference between PFO closure and anticoagulation for ischaemic stroke, with a 58 59 point estimate substantially closer to 1.0: NMA OR 0.93, CI 0.31 to 2.76; RD: -2, -20 to +47; low 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 18 of 56

certainty (Table 3, Appendix 5: eFigure 1). For this complementary analysis, we rated down our BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 2 3 certainty in the evidence to low because of very serious indirectness because this analysis relies on 4 5 the assumption that the relative effect of anticoagulation versus antiplatelet therapy is the same for 6 7 secondary prevention VTE as it is for secondary prevention of cryptogenic stroke in patients with a 8 9 PFO. 10 11 12 13 There was no important difference in TIA: NMA OR 1.27, CrI 0.40 to 4.52; RD +6, CI -22 to +22; 14 15 moderate certainty (Table 3, Appendix 5: eFigure 2). No patient experienced a systemic embolism in 16 For peer review only 17 18 either arm of the RCT that included an anticoagulation arm (Appendix 5: eFigure 3). 19 20 21 Few patients randomised to PFO closure plus antiplatelet therapy experienced a pulmonary 22 23 embolism (median risk 5 per 1000 patients over 5 years). The RCT that randomised patients to PFO 24 25 26 closure versus anticoagulation did not measure pulmonary emboli (14). Using external evidence from 27 28 a systematic review of RCTs that compared ASA to anticoagulation with a VKA for secondary 29 30 prevention of VTE in the NMA, PFO closure plus antiplatelet therapy probably has a higher risk of 31

32 http://bmjopen.bmj.com/ 33 pulmonary embolism than anticoagulation: RD +4, CI +1 to +13; moderate certainty (Table 3, 34 35 Appendix 5: eFigure 4). 36 37 38 Patients randomised to PFO closure plus antiplatelet therapy had a lower risk of major bleeding than 39

40 on September 27, 2021 by guest. Protected copyright. 41 those randomised to anticoagulation: NMA OR 0.26, CI 0.07 to 0.82; RD -20, CI -27 to -2, moderate 42 43 certainty (Table 3, Appendix 5: eFigure 5). Adding the effect estimate for major bleeding of low-dose 44 45 ASA compared to anticoagulation with a VKA from a systematic review of RCTs enrolling patients 46 47 with atrial fibrillation, VTE, or heart failure (37) to the NMA, resulted in an almost identical point 48 49 50 estimate, but a narrower confidence interval: OR 0.28, CrI 0.13 to 0.55; RD -17, CI -21 to -11; 51 52 moderate certainty (Table 3). 53 54 55 56 There was no difference in death between PFO closure and anticoagulation (RD -4, -13 to +9, 57 58 moderate certainty) (Table 3). 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 19 of 56 BMJ Open

Table 3: Summary of GRADE evidence profile of PFO closure plus antiplatelet therapy vs anticoagulation in BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 patients with cryptogenic stroke 2 3 Absolute effect estimates per 1000 patient- Certainty in effect Outcome Study results and years 4 estimates Plain text summary Timeframe measurements PFO closure plus 5 Anticoagulation (Quality of evidence) 6 antiplatelet therapy 7 Odds Ratio: 0.44 29 13 8 (CI 95% 0.08 - 3.83) per 1000 per 1000 Low There may be little or no Ischaemic stroke 9 Direct evidence in NMA from Due to very serious difference in ischaemic (Standardized to 5 years) 1 10 353 patients in 1 studies Difference: 16 fewer imprecision stroke 11 Follow up 5.3 years (CI 95% 29 fewer* - 10 more) 12 29 27 13 Odds Ratio: 0.93 Low Ischaemic stroke - modelling per 1000 per 1000 There may be little or no 14 (CI 95% 0.31 - 2.76) Due to serious data from VTE literature difference in ischaemic imprecision and 15 (Standardized to 5 years) stroke Difference: 2 fewer serious indirectness2 16 For peer review(CI 95% 20 fewer - 47 more) only 17 18 Relative risk: 0.69 13 9 (CI 95% 0.02 - 32.36) per 1000 per 1000 Moderate 19 Death There is probably little or Direct evidence in NMA from Due to serious 20 (Standardized to 5 years) 3 no difference in death 21 353 patients in 1 studies Difference: 4 fewer imprecision Follow up 5.3 years (CI 95% 13 fewer* - 9 more) 22 23 Odds Ratio: 0.26 27 7 PFO closure plus 24 (CI 95% 0.07 - 0.82) per 1000 per 1000 Moderate Major bleeding antiplatelet therapy 25 Direct evidence in NMA from Due to serious (Standardized to 5 years) probably decreases major 353 patients in 1 studies imprecision4 26 Difference: 20 fewer bleeding 27 Follow up 5.3 years (CI 95% 27 fewer* - 2 fewer) 28 24 7 29 Odds Ratio: 0.28 PFO closure plus Major bleeding - modelling per 1000 per 1000 Moderate 30 (CI 95% 0.13 - 0.55) antiplatelet therapy data from VTE literature Due to serious probably decreases major 31 (Standardized to 5 years) indirectness5 Difference: 17 fewer bleeding 32 (CI 95% 21 fewer - 11 fewer) http://bmjopen.bmj.com/ 33 34 Relative risk: 4.84 5 23 x PFO closure plus 35 Persistent atrial fibrillation or (CI 95% 1.91 - 12.26) per 1000 per 1000 Moderate antiplatelet therapy Based on data from 3560 Due to serious risk of 36 flutter probably increases non- patients in 6 studies @ bias6 37 (Standardized to 1 year) Difference: 18 more per 1000 patients transient atrial fibrillation Follow up 3.9 years 38 (CI 95% 5 more - 56 more) 39 Relative risk: 3.76 5 17 40 PFO closure plus on September 27, 2021 by guest. Protected copyright. Transient or paroxysmal atrial (CI 95% 1.74 - 8.1) per 1000 per 1000 Moderate antiplatelet therapy 41 fibrillation or flutter Based on data from 3560 Due to serious risk of probably increases 42 (Standardized to 1 year) patients in 6 studies @ bias7 Difference: 12 more per 1000 patients transient atrial fibrillation 43 Follow up 3.9 years (CI 95% 3 more - 31 more) 44

45 0 36 PFO closure plus Risk difference: 0.04 Device or procedure related per 1000 per 1000 antiplatelet therapy 46 (CI 95% 0.02 – 0.05) High adverse event increases device or 47 Based on data from 3560 (Standardized to 1 year) @ procedure related adverse 48 patients in 6 studies Difference: 36 more per 1000 patients (CI 95% 23 more - 50 more) events 49 Follow up 3.9 years 50 Odds Ratio: 1.27 22 28 51 (CI 95% 0.4 - 4.52) per 1000 per 1000 Moderate There is probably little or Transient ischaemic attack 52 Direct evidence in NMA from Due to serious no difference in transient (Standardized to 5 years) 8 53 353 patients in 1 studies Difference: 6 more per 1000 imprecision ischaemic attack 54 Follow up 5.3 years (CI 95% 22 fewer* - 22 more) 55 56 Pulmonary embolism - Odds Ratio: 9.09 1 5 Moderate There is probably little or modelling data from VTE (CI 95% 3.7 - 25.0) per 1000 per 1000 Due to serious no difference in 57 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 20 of 56 BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from literature indirectness9 pulmonary embolism 1 Difference: 4 more per 1000 (Standardized to 5 years) (CI 95% 1 more - 13 more) 2 3 Odds Ratio: 291.0 0 0 4 (CI 95% 0.0 - 999.0) per 1000 per 1000 Moderate There is probably little or Systemic embolism 5 Direct evidence in NMA from Due to serious no difference in systemic (Standardized to 5 years) 10 6 353 patients in 1 studies Difference: 0 fewer per 1000 imprecision embolism 7 Follow up 5.3 years (CI 95% 11 fewer - 11 more) 8 Note: The baseline risk for PFO closure was used to estimate the absolute effect in the anticoagulation arm. The baseline risk in the PFO arm was calculated based on 9 10 the median risk of the outcome in the six RCTs included. This allowed for consistency in the absolute effect estimate in the PFO closure arms in table 2 and 3 and the 11 anticoagulation arms in table 3 and 4 allowing for ease in comparison. 12 13 1. Imprecision: Very Serious. Wide confidence interval. Low number of events. 14 2. Indirectness: Serious. In addition to the direct evidence from randomised trials in patients with PFO and a cryptogenic ischaemic stroke, we additionally 15 considered external evidence from randomised trials that assessed the impact of anticoagulation vs. antiplatelet therapy for the secondary prevention of venous thromboembolism. Imprecision: Serious. Wide confidence intervals, includes both appreciable benefit and harm. 16 3. Imprecision: Serious. Wide confidenceFor intervals, includespeer both appreciable review benefit and harm. Low only number of events. 17 4. Imprecision: Serious. Wide confidence interval, included a not important benefit. Low number of events. 18 5. Indirectness: Serious. In addition to the direct evidence from randomised trials in patients with PFO and a cryptogenic ischaemic stroke, we additionally 19 considered external evidence from randomised trials that assessed the impact of anticoagulation vs. antiplatelet therapy for the secondary prevention of venous thromboembolism. 20 6. Risk of bias: Serious. Not clearly stated how this was measured or assessed with prolonged ECG monitoring. Also, it is not clear for all events whether it was 21 transient or persistent. 22 7. Risk of bias: Serious. Not clearly stated how this was measured or assessed with prolonged ECG monitoring. Also, it is not clear for all events whether it was 23 transient or persistent. 8. Imprecision: Serious. Wide confidence interval, includes both appreciable benefit and harm. Low number of events. 24 x 25 Defined as persistent according to the study definition or requiring a cardioversion attempt *: The calculated confidence interval using risk difference, because of uncertainty in the point estimates, permits reductions greater than the point estimates in the PFO group. To avoid confusion, we have truncated to present the 26 maximum reduction as equal to the PFO event rate. 27 @: In the first year after procedure rather than 5 years 28 VTE, venous thromboembolism; AF, atrial fibrillation; CHF, congestive heart failure 29 30 31

32 http://bmjopen.bmj.com/ 33 Complications of PFO closure 34 35 36 PFO closure, when compared to medical therapy, probably substantially increases the risk of atrial 37 38 fibrillation (including transient, persistent, and paroxysmal): six RCTs with 3560 patients; RR 4.50, 39

40 on September 27, 2021 by guest. Protected copyright. 41 95% confidence interval (CI) 2.35 to 8.60; risk difference per 1000 patients (RD) +39, CI +15 to +84; 42 43 moderate certainty (Table 3, Appendix 6: eFigure 1). All six studies assessed persistent atrial 44 45 fibrillation at a follow-up of 2.0 to 5.9 years. PFO closure probably increases the risk of persistent 46 47 atrial fibrillation: RR 4.84, CI 1.91 to 12.26; RD +18, CI +5 to +56; moderate certainty (Table 3, 48 49 50 Appendix 6: eFigure 2). PFO closure also probably increases the risk of transient atrial fibrillation: RR 51 52 3.76, CI 1.74 to 8.10; RD +12, CI +3 to +31; moderate certainty (Table 3, Appendix 6: eFigure 3). 53 54 The studies were limited by risk of bias because they did not report sufficient detail to know how 55 56 57 persistent atrial fibrillation was detected (Tables 2 and 3). 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 21 of 56 BMJ Open

Serious device or procedure-related adverse events occurred in approximately 36 per 1000 patients, BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 2 3 CI 23 to 50, high certainty (Table 2, Appendix 6: eFigure 4). The most common device or procedure- 4 5 related complications were vascular complications (1%), conduction abnormalities (1%), device 6 7 dislocation (0.7%), and device thrombosis (0.5%). Although serious, air embolism (0.4%), cardiac 8 9 tamponade (0.3%), and cardiac perforation (0.2%) were rare (Tables 2 and 3). No deaths were 10 11 12 attributed to the procedure or the device placement (Tables 2 and 3). 13 14 15 Anticoagulation versus antiplatelet therapy 16 For peer review only 17 18 19 Table 4 provides estimates of effect and certainty of evidence for all patient-important outcomes 20 21 comparing the two medical therapies. There may be fewer ischaemic strokes in patients randomised 22 23 to anticoagulation than to antiplatelet therapy, but the CI included no effect: NMA OR 0.27, CrI 0.03 to 24 25 26 1.21; RD: -71, CI -100 to +17; low certainty (Table 4, Appendix 7: eFigure 1). Figure 3 presents a 27 28 Kaplan Meier curve of individual patient time-to-event data for each of the randomized interventions, 29 30 including anticoagulation versus antiplatelet therapy suggesting a lower rate of stroke in patients 31

32 http://bmjopen.bmj.com/ 33 receiving anticoagulant than antiplatelet therapy. 34 35 36 The complementary model including external evidence from a systematic of RCTs comparing low- 37 38 dose ASA to anticoagulation with a VKA for secondary prevention of VTE yielded a somewhat larger 39

40 on September 27, 2021 by guest. Protected copyright. 41 reduction in strokes with anticoagulants and substantially narrowed the CI: OR 0.17, CrI 0.08 to 0.36; 42 43 RD -81, CI -91 to -62; low certainty (Table 4) (37). 44 45 46 There may be little or no difference in risk of TIA between anticoagulation and antiplatelet therapy 47 48 49 (NMA OR 0.65, CrI 0.19 to 1.98; RD -12, CI -34 to +24; low certainty (Table 4, Appendix 7: eFigure 50 51 2). There were no reported systemic emboli, although it is not clear what the monitoring and 52 53 ascertainment processes were. 54 55 56 57 Our analysis suggested more major bleeding events among patients randomised to anticoagulation 58 59 than to antiplatelet therapy, but the CI included no effect: NMA OR 1.90, CI 0.68 to 5.53; RD +12, -5 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 22 of 56

to +65; moderate certainty (Table 4, Appendix 7: eFigure 3). Including external evidence from a BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 2 3 systematic review of RCTs comparing low-dose ASA to anticoagulation with a VKA in patients with 4 5 atrial fibrillation, VTE, and heart failure yielded a very similar point estimate and substantially 6 7 narrowed the CI: OR 1.77, CI 1.36 to 2.31; RD +11, CI +5 to +18; moderate certainty (Table 4) (37). 8 9 10 11 The RCT that included an anticoagulation arm did not report pulmonary emboli; the risk of pulmonary 12 13 embolism was low in other RCTs that included an antiplatelet arm (median 5 per 1000 patient-years 14 15 over 5 years). External evidence from a systematic review of secondary prevention of VTE suggests 16 For peer review only 17 18 that the risk of pulmonary embolism is probably lower with anticoagulation than antiplatelet therapy: 19 20 RD -4, CI -5 to -3; moderate certainty (17). 21 22 23 Deaths were extremely rare and there was no appreciable difference between the anticoagulation 24 25 26 and antiplatelet therapy: RD +10, CI -3 to +357, low certainty (Table 4, Appendix 7: eFigure 4). 27 28 29 Table 4: Summary of GRADE evidence profile of anticoagulation vs antiplatelet therapy alone in patients with 30 cryptogenic stroke 31 Absolute effect estimates per 1000 32 Certainty in effect http://bmjopen.bmj.com/ Outcome patient-years Study results and measurements estimates Plain text summary 33 Timeframe 34 Antiplatelet Anticoagulation (Quality of evidence) 35 y 36 Odds Ratio: 0.27 100 29 (CI 95% 0.03 - 1.21) per 1000 per 1000 Low 37 Ischaemic stroke Anticoagulation may decrease Direct evidence in NMA from 361 Due to very serious 38 (Standardized to 5 years) ischaemic stroke patients in 1 studies imprecision1 39 Difference: 71 fewer Follow up 5.3 years (CI 95% 100 fewer* - 17 more)

40 on September 27, 2021 by guest. Protected copyright. 41 100 19 Odds Ratio: 0.17 42 Ischaemic stroke - modelling per 1000 per 1000 Low (CI 95% 0.08 - 0.36) Anticoagulation may decrease 43 data from VTE literature Due to very serious ischaemic stroke (Standardized to 5 years) indirectness2 44 Difference: 81 fewer 45 (CI 95% 91 fewer - 62 fewer) 46 47 Odds Ratio: 4.81 3 13 (CI 95% 0.31 - 224.43) per 1000 per 1000 Low Death There may be little or no 48 Direct evidence in NMA from 408 Due to very serious (Standardized to 5 years) 3 difference in death 49 patients in 2 studies Difference: 10 more imprecision 50 Follow up 3.2 years (CI 95% 3 fewer* - 357 more) 51 52 Odds Ratio: 1.9 14 26 (CI 95% 0.68 - 5.53) per 1000 per 1000 Moderate 53 Major Bleeding Anticoagulation probably Direct evidence in NMA from 408 Due to serious 54 (Standardized to 5 years) 4 increases major bleeding 55 patients in 2 studies Difference: 12 more imprecision Follow up 3.2 years (CI 95% 5 fewer - 65 more) 56 57 Major bleeding - modelling Odds Ratio: 1.77 14 25 Moderate Anticoagulation probably 58 data from VTE literature (CI 95% 1.36 - 2.31) per 1000 per 1000 Due to serious increases major bleeding 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 23 of 56 BMJ Open BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from (Standardized to 5 years) indirectness5 1 Difference: 11 more (CI 95% 5 more - 18 more) 2 3 Odds Ratio: 0.65 34 22 4 (CI 95% 0.19 - 1.98) per 1000 per 1000 Low There may be little or no Transient ischaemic attack 5 Direct evidence in NMA from 361 Due to very serious difference in transient ischaemic (Standardized to 5 years) 6 6 patients in 1 studies Difference: 12 fewer imprecision attack 7 Follow up 5.3 years (CI 95% 34 fewer* - 24 more) 8 5 1 9 Pulmonary embolism - Odds Ratio: 0.11 per 1000 per 1000 Moderate There is probably little or no modelling data from VTE (CI 95% 0.04 - 0.37) 10 Due to serious difference in pulmonary literature 11 indirectness7 embolism. (Standardized to 5 years) Difference: 4 fewer 12 (CI 95% 5 fewer - 3 fewer) 13 14 0 0 Not estimable per 1000 per 1000 Moderate 15 Systemic embolism There is probably little or no Direct evidence in NMA from 361 Due to serious 16 (Standardized to 5 years) difference in systemic embolism Forpatients in 1 studiespeer review onlyimprecision8 17 Difference: 0 fewer Follow up 5.3 years (CI 95% 11 fewer - 11 more) 18 19 Note: The baseline risk for antiplatelet was obtained using calculated absolute effect estimate. This was done to maintain consistency across the tables. The 20 calculated baseline risk (10%) was similar to the baseline risk calculated using the median of studies included (9%) 21 22 1. Imprecision: Very Serious. Wide confidence interval, includes appreciable harm. Low number of events. 23 2. Indirectness: Very Serious. In addition to the direct evidence from randomised trials in patients with PFO and a cryptogenic ischaemic stroke, we additionally 24 considered external evidence from randomised trials that assessed the impact of anticoagulation vs. antiplatelet therapy for the secondary prevention of venous thromboembolism. 25 3. Imprecision: Very Serious. Wide confidence interval, includes both appreciable benefit and harm. Low number of events. 26 4. Imprecision: Serious. Wide confidence intervals, Low number of events. 27 5. Indirectness: Serious. In addition to the direct evidence from randomised trials in patients with PFO and a cryptogenic ischaemic stroke, we additionally 28 considered external evidence from randomised trials that assessed the impact of anticoagulation vs. antiplatelet therapy for the secondary prevention of 29 venous thromboembolism. We did not rate down with two levels because we felt the outcome is less indirect compared to VTE literature than Ischaemic stroke. 6. Imprecision: Very Serious. Wide confidence interval, includes both appreciable harm and benefit. Low number of events. 30 *: The calculated confidence interval using risk difference, because of uncertainty in the point estimates, permits reductions greater than the point estimates in the 31 PFO group. To avoid confusion, we have truncated to present the maximum reduction as equal to the PFO event rate.

32 http://bmjopen.bmj.com/ 33 34 35 36 37 38 Subgroup Analyses 39

40 on September 27, 2021 by guest. Protected copyright. 41 Data proved insufficient to perform the prespecified subgroup analyses when anticoagulation and 42 43 antiplatelets were assumed to have different effects, as we had prespecified. We performed one post- 44 45 46 hoc meta-regression analysis to explore the effect of PFO shunt size on the relative effectiveness of 47 48 PFO closure compared to medical therapy. Five RCTs reported ischaemic stroke and the proportion 49 50 of patients with a moderate or large shunt vs. small shunt, all with slightly different definitions. PFO 51 52 53 closure was more effective in the RCTs that enrolled a higher proportion of patients with a moderate 54 55 or large shunt (interaction p=0.047) (Appendix 8: eFigure 1). However, this is confounded by the fact 56 57 that the RCTs that enrolled a higher proportion of patients with moderate or large PFO shunts also 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 24 of 56

had a larger proportion of patients with antiplatelet agents rather than anticoagulation (Appendix 8: BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 2 3 eFigure 2). When both possible effect modifiers were included in the model neither was statistically 4 5 significant: p=0.44 for proportion with moderate or large shunt and p=0.61 for proportion receiving 6 7 anticoagulation; however, this model is severely underpowered with two effect modifiers and only five 8 9 studies. 10 11 12 13 DISCUSSION 14 15 16 We found moderate certaintyFor evidence peer that in review patients less than only 60 years old with a PFO and 17 18 19 cryptogenic stroke, PFO closure plus antiplatelet therapy results in a substantial reduction in the risk 20 21 of recurrent ischaemic stroke compared to antiplatelet therapy alone by approximately 8.7% over 5 22 23 years, but increases the risk of persistent atrial fibrillation (approximately 1.8%) and device-related 24 25 26 adverse events (approximately 3.6%) in the first year after procedure (Table 2). 27 28 29 Although the evidence regarding anticoagulation versus alternatives is of low quality and comes in 30 31 large part from indirect evidence, it is possible that anticoagulation and PFO closure may achieve a

32 http://bmjopen.bmj.com/ 33 34 similar reduction in the risk of recurrent ischaemic stroke – and thus appreciably fewer strokes than 35 36 antiplatelet agents alone. Anticoagulation, compared to either PFO closure plus antiplatelet therapy or 37 38 antiplatelet therapy alone probably increases the risk of major bleeding by approximately 2% over 5 39

40 on September 27, 2021 by guest. Protected copyright. 41 years and probably reduces the risk of pulmonary embolism by approximately 0.4%. There does not 42 43 appear to be an important difference in the risk of death or in the risk of systemic emboli between any 44 45 of the interventions. 46 47 48 49 Our results provide further strong support for the hypothesis that stasis-related paradoxical venous 50 51 thromboemboli and/or intra-cardiac thrombi cause a large proportion of cryptogenic strokes in 52 53 younger patients with PFO. Given that is the case, it is plausible, though uncertain, that 54 55 56 anticoagulation has a similar relative benefit versus antiplatelet agents in preventing cryptogenic 57 58 strokes in patients with PFO as it does in preventing recurrent VTE. Our results, including the meta- 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 25 of 56 BMJ Open

regression examining comparisons of PFO closure versus studies in which varying proportion of BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 2 3 patients received anticoagulants (Figure 2) support this inference. 4 5 6 Only one of the RCTs included an arm in which all patients were offered anticoagulation or PFO 7 8 closure; as a result, the direct evidence includes relatively few patients and very few events. 9 10 11 Moreover, the evidence from studies of VTE is very indirect. Thus, we categorised the evidence 12 13 regarding anticoagulation versus PFO closure, and versus antiplatelet therapy, in preventing stroke 14 15 as low certainty (Tables 3 and 4). Resolving the uncertainty that currently bedevils decision-making 16 For peer review only 17 18 will require larger RCTs that compare anticoagulation to PFO closure. 19 20 21 A recent systematic review suggested that PFO closure may be more effective in patients with larger 22 23 shunt (38). We show that this subgroup analysis has low credibility because it is confounded by the 24 25 26 anticoagulation in the patients randomised to medical therapy in RCTs with a higher proportion of 27 28 patients with smaller shunts. PFO closure may be less effective relative to medical therapy when 29 30 patients receive anticoagulation and/or have smaller shunts. The results of the NMA, in addition to 31

32 http://bmjopen.bmj.com/ 33 indirect evidence from patients with a VTE, suggest that the subgroup effect by type of medical 34 35 therapy may be the most credible explanation. 36 37 38 PFO closure comes with infrequent but important risks: an approximately 3% to 4% chance of serious 39

40 on September 27, 2021 by guest. Protected copyright. 41 device or procedure-related complications and an approximately 4% increased incidence of atrial 42 43 fibrillation, of which approximately half is persistent. Anticoagulation with VKAs carries a 1% to 2% 44 45 increase in major bleeding over 5 years in this patient population. The risk of major bleeding may be 46 47 marginally lower with some direct acting oral anticoagulants than with VKAs (17). Patients must also 48 49 50 consider the important practical implications of taking anticoagulants, including limitations to diet and 51 52 activities with an appreciable risk of trauma, possible drug interactions and, with VKAs, the need for 53 54 repeat laboratory testing. 55 56 57 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 26 of 56

Strengths and limitations: BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 2 3 Our study has several strengths. In contrast with recent systematic reviews summarising the recent 4 5 RCTs on PFO closure (38-41); we separately addressed the three management alternatives of PFO 6 7 closure plus antiplatelet therapy, anticoagulation alone, and antiplatelet therapy alone. In doing so, 8 9 we applied several analytical approaches, including network meta-analysis, meta-regression, and 10 11 12 individual patient analyses from survival curves, as well as judicious use of external evidence, to bring 13 14 to bear all the relevant evidence for decision-making. In addition, we used the GRADE approach to 15 16 assess the certainty of evidenceFor informingpeer the review estimates in the only study. Our review also includes one 17 18 19 recently published study not included in prior reviews. We thus summarised all of the highest quality 20 21 available evidence, providing optimal insight into the comparative effects of the alternative 22 23 management strategies for patients with PFO who have experienced a cryptogenic stroke. 24 25 26 The results are limited primarily by the available evidence. Of the ten RCTs that we included in this 27 28 review, three compared PFO closure to a medical therapy arm that was not predefined – the key 29 30 choice between antiplatelet and anticoagulation was left to the discretion of the physician and patient. 31 32 As corroborated in our evidence synthesis, there are probably important differences between http://bmjopen.bmj.com/ 33 34 35 anticoagulation and antiplatelet therapy. As a result, these three RCTs cannot directly inform the 36 37 clinical decisions in which antiplatelet therapy and anticoagulation are two distinct choices, and we 38 39 could not include them in many of the analyses bearing on the choice between PFO closure and

40 on September 27, 2021 by guest. Protected copyright. 41 42 antiplatelet therapy. Our meta-regression analysis that included these studies suggested an inverse 43 44 relationship between stroke reduction and proportion of patients in the medical therapy arms 45 46 prescribed anticoagulation, thus providing support for the benefit of anticoagulants versus antiplatelet 47 48 49 therapy (Figure 2). 50 51 The NMA also has limitations: the network is sparsely populated (Appendix 8: eFigure 3), it includes 52 53 patients with no contraindications to any of three management strategies, those with patients with 54 55 contraindications to anticoagulants, and those with contraindications to PFO. Nevertheless, we found 56 57 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 27 of 56 BMJ Open

no suggestion of incoherence between direct and indirect estimates in the NMA and suggest that it BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 2 3 provides the best estimates available (Appendix 9, eTable 1). 4 5 The applicability of these findings to older patient populations, particularly those over 60 years of age, 6 7 and those with traditional cerebrovascular risk factors such as exposure to diabetes, hypertension, 8 9 and hyperlipidaemia, is uncertain. The proportion of patients with a cryptogenic stroke caused by 10 11 12 paradoxical emboli rather than large vessel atheroembolism or intracardiac thrombus almost certainly 13 14 decreases with age. Thus, we expect that the benefits of PFO closure would be smaller in older 15 16 patients and the harms (e.g.For atrial peer fibrillation) greater.review only 17 18 19 20 21 CONCLUSIONS 22 23 In patients less than 60 years old with cryptogenic stroke and a PFO, PFO closure plus antiplatelet 24 25 26 therapy probably confers a substantial reduction in the risk of ischaemic stroke recurrence compared 27 28 to antiplatelet therapy alone, with a modest risk of persistent atrial fibrillation or flutter and serious 29 30 device or procedure-related complications. Based on low certainty evidence, anticoagulation may 31 32 confer a similar reduction in the risk of ischaemic stroke compared to PFO closure, but likely confers http://bmjopen.bmj.com/ 33 34 35 a modest increased risk of major bleeding. Future RCTs comparing anticoagulation to PFO closure 36 37 would substantially reduce uncertainty for several critical outcomes. 38 39

40 on September 27, 2021 by guest. Protected copyright. 41 42 Acknowledgement: We thank members of the BMJ Rapid Recommendations panel for critical 43 44 feedback on outcome selection, subgroup selection, GRADE judgments, and manuscript feedback. 45 46 We thank Rachel Couban for helping develop the search strategy. 47 48 49 Contributors: GHG and RAS conceived the study idea. HM performed the literature search and data 50 51 analysis. HM, RAS, TA, GHG, POV interpreted the data analysis. HM, RAS, and GHG wrote the first 52 53 draft of the manuscript. TS, LL, and MF acquired data and judged risk of bias in the studies. FF 54 55 extracted patient level survival data from Kaplan-Meier curves. LG provided statistical advice. RAS, 56 57 58 GHG, TA, POV, TK, JLM, MF critically reviewed the manuscript. HM had full access to all of the data 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 28 of 56

in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 2 3 HM is guarantor. 4 5 Funding: None 6 7 Competing interests: All authors have completed the BMJ Rapid Recommendations disclosure 8 9 form, which asks about any possible financial, intellectual, and professional conflicts of interest 10 11 12 (available upon request). To summarize, JLM has received consulting honoraria from Bayer, Bristol 13 14 Meyer Squibb, Boehringer-Ingelheim, Daichii-Sankyo, GECKO, and Servier. HM, RAS, TK, PV, TA, 15 16 and GHG are also panel Formembers peer of the linked review Rapid Recommendation only guideline panel. LG, FF, MF, 17 18 19 and TS do not have any competing interests to declare. 20 21 Ethical approval: Not required 22 23 Data sharing: Extra data can be accessed via the Dryad data repository at http://datadryad.org/ with 24 the doi: 10.5061/dryad.ng017rc 25 26 27 Transparency declaration: The lead author (HM) affirms that the manuscript is an honest, accurate, 28 29 and transparent account of the study being reported; that no important aspects of the study have 30 31 been omitted; and that any discrepancies from the study as planned have been explained.

32 http://bmjopen.bmj.com/ 33 34 35 36 37 38 39

Figure Legend: BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 2 3 Figure 1. PRISMA flow diagram of studies included in review treatment of patients with PFO and 4 5 cryptogenic stroke 6 7 Figure 2. Meta-regression curve based on the proportion of anticoagulant in the medical therapy arm 8 9 for ischaemic stroke 10 11 12 Figure 3. Combined Kaplan-Meier curves with individualized patient data based on the type of 13 14 intervention for ischaemic stroke 15 16 Foot note Figure 3. For peer review only 17 18 1) Y-axis is truncated from 90% to 100% Event-free Survival 19 2) Mixed medical therapy includes studies where the proportion of antiplatelet agents was less than 80% and the proportion of anticoagulant was greater 20 than 25% 21 3) Proportion of patients: PFO Closure + antiplatelet therapy (N= 1829; 50%), Mixed medical therapy (N= 1153; 32%), Antiplatelet therapy (N= 458; 13%), and 22 Anticoagulation (N= 210; 6%) 4) K-M curves were available for the outcome of ischaemic stroke in the CLOSE study (14) (PFO closure, anticoagulation, and antiplatelet), the PC Trial (6) 23 (PFO closure and mixed medical therapy), the RESPECT trial (16) (PFO closure and mixed medical therapy) , and the REDUCE trial (15) (PFO closure and 24 antiplatelet). 25 26

27 28 29 30 31

32 http://bmjopen.bmj.com/ 33 34 35

36 37 38 39

40 on September 27, 2021 by guest. Protected copyright.

41 42 43 44

45 46 47 48

49 50 51 52 53

54 55 56 57 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 30 of 56

32 http://bmjopen.bmj.com/ 33 34 35 36 37 38 39

40 on September 27, 2021 by guest. Protected copyright. 41 42 43 44 45 PRISMA flow diagram of studies included in review treatment of patients with PFO and cryptogenic stroke 46 47 279x361mm (300 x 300 DPI) 48 49 50 51 52 53 54 55 56 57 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 33 of 56 BMJ Open BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 For peer review only 17 18 19 20 21 22 23 24 25 26 27 28 29 30 Meta-regression curve based on the proportion of anticoagulant in the medical therapy arm for ischaemic 31 stroke 32 http://bmjopen.bmj.com/ 33 90x65mm (300 x 300 DPI) 34 35 36 37 38 39

40 on September 27, 2021 by guest. Protected copyright. 41 199x145mm (300 x 300 DPI) 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 35 of 56 BMJ Open

Appendices Legend: BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 2 3 Appendix 1: Summary of search and strategies 4 Appendix 2: Tables 5 • eTable 1: Characteristics of eligible studies 6 7 Appendix 3: Figures 8 • eFigure 1: Risk of bias of included randomised controlled trials 9 Appendix 4: PFO closure vs antiplatelet forest plots (Direct evidence only) 10 • eFigure 1: Ischaemic stroke 11 12 • eFigure 2: TIA 13 • eFigure 3: Systemic emboli 14 • eFigure 4: Death 15 • eFigure 5: Major bleeding 16 For peer review only 17 • eFigure 6: Pulmonary embolism 18 Appendix 5: PFO closure vs anticoagulation forest plots (Direct evidence only) 19 • eFigure 1: Ischaemic stroke 20 • eFigure 2: TIA 21 22 • eFigure 3: Systemic emboli 23 • eFigure 4: Pulmonary emboli 24 • eFigure 5: Major bleeding 25 Appendix 6: Complications of PFO closure vs medical therapy forest plots (Direct evidence only) 26 27 • eFigure 1: All atrial fibrillation 28 • eFigure 2: Persistent atrial fibrillation 29 • eFigure 3: Transient atrial fibrillation 30 31 • eFigure 4: Serious device or procedure related

32 Appendix 7: Anticoagulation vs antiplatelet forest plots (Direct evidence only) http://bmjopen.bmj.com/ 33 • eFigure 1: Stroke 34 • eFigure 2: TIA 35 36 • eFigure 3: Major bleeding 37 • eFigure 4: Death 38 Appendix 8: Meta-regression and NMA Figures 39 • eFigure 1: Meta-regression evaluating the effect of moderate or higher shunt size on the effectiveness

40 on September 27, 2021 by guest. Protected copyright. 41 of PFO closure versus medical therapy 42 • eFigure 2: Meta-regression evaluating the proportion of patients on anticoagulation compared to 43 those with moderate or higher shunt size 44 • eFigure 3: Network of included RCTs with available direct comparisons for ischaemic stroke. 45 46 Appendix 9: NMA Table 47 • eTable 1: Direct and indirect estimates of effects for the network meta-analysis 48 49 50 51 52 53 54 55 56 57 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 36 of 56

Appendix 1: Summary of search and strategies BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 2 3 Database: OVID Medline Epub Ahead of Print, In-Process & Other Non-Indexed Citations, Ovid MEDLINE(R) 4 Daily and Ovid MEDLINE(R) 1946 to Present 5 Search Strategy: 6 ------7 8 1 cerebrovascular disorders/ or basal ganglia cerebrovascular disease/ or exp brain ischemia/ or carotid 9 artery diseases/ or carotid artery thrombosis/ or intracranial arterial diseases/ or cerebral arterial diseases/ or 10 exp "intracranial embolism and thrombosis"/ or exp stroke/ (237544) 11 2 (isch?emi$ adj6 (stroke$ or apoplex$ or cerebral vasc$ or cerebrovasc$ or cva or attack$)).tw. (62463) 12 13 3 ((brain or cerebr$ or cerebell$ or vertebrobasil$ or hemispher$ or intracran$ or intracerebral or 14 infratentorial or supratentorial or middle cerebr$ or mca$ or anterior circulation) adj5 (isch?emi$ or infarct$ 15 or thrombo$ or emboli$ or occlus$ or hypoxi$)).tw. (105788) 16 4 (TIA or TIAs).mp. (8376)For peer review only 17 5 Embolism, Paradoxical/ (914) 18 19 6 heart atria/ and (embolism/ or thromboembolism/) (648) 20 7 ((paradoxic$ or crossed) adj5 embolism$).tw. (1585) 21 8 (cryptogenic adj5 stroke).tw. (1168) 22 9 or/1-8 (293081) 23 24 Annotation: population part 1: stroke or cryptogenic stroke 25 10 heart septal defects, atrial/ or foramen ovale, patent/ (14696) 26 11 heart septum/ or atrial septum/ or foramen ovale/ (10405) 27 12 (patent foramen ovale or PFO).tw. (5125) 28 13 (atrial sept$ adj5 defect$).tw. (10211) 29 30 14 ((right to left or R-L or venous to arterial or venous-arterial or V-A) adj3 shunt).tw. (2212) 31 15 or/10-14 (30798)

32 Annotation: population part 2: PFO http://bmjopen.bmj.com/ 33 16 9 and 15 (3491) 34 35 Annotation: population Stroke and PFO 36 17 limit 16 to yr="2012 -Current" (1094) 37 Annotation: since last review 38 18 limit 17 to ("therapy (maximizes sensitivity)" or "therapy (maximizes specificity)" or "therapy (best 39 balance of sensitivity and specificity)") (429)

40 on September 27, 2021 by guest. Protected copyright. 41 19 randomised controlled trial.pt. (496904) 42 20 controlled clinical trial.pt. (99253) 43 21 randomised.ab. (433409) 44 22 placebo.ab. (202740) 45 46 23 drug therapy.fs. (2114500) 47 24 randomly.ab. (298737) 48 25 trial.ab. (457112) 49 26 groups.ab. (1845391) 50 27 or/19-26 (4369043) 51 52 28 exp animals/ not humans.sh. (4677556) 53 29 27 not 28 (3778961) 54 30 17 and 29 (274) 55 31 18 or 30 (485) 56 57 Annotation: Population stroke and PFO limit to since 2012 and RCTs 58 32 "prostheses and implants"/ or septal occluder device/ (47787) 59 33 Wound Closure Techniques/ (1141) 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 37 of 56 BMJ Open

34 (close or closure or septal occluder).tw. (384980) BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 2 35 (cardioseal or gore helex or amplatzer or starflex or cardia or intrasept or premere).tw. (9507) 3 36 su.fs. (1918373) 4 37 or/32-36 (2271562) 5 Annotation: Septal occluder device as per Liu 2015 CDSR 6 38 exp Anticoagulants/ (212306) 7 8 39 anticoagulant$.tw. (55364) 9 40 (acenocoumarol$ or dicoumarol$ or ethyl biscoumacetate$ or phenprocoumon$ or warfarin$ or ancrod$ 10 or citric acid$ or coumarin$ or chromonar$ or coumestro$ or esculi$ or ochratoxin$ or umbelliferone$ or 11 dermatan?sul$ or dextran$ or edetic acid$ or enoxaparin$ or gabexate$ or heparin$ or lmwh$ or nadroparin$ 12 13 or pentosan sulfuric polyester$ or phenindione$ or protein c or protein s or tedelparin$).tw. (192083) 14 41 (argatroban or tinzaparin or parnaparin or reviparin or danaparoid or lomoparan or org 10172 or 15 mesoglycan or polysaccharide sulphate$ or sp54 or sp-54 or md805 or md-805 or cy222 or cy-222 or cy216 or 16 cy-216).tw. (2862) For peer review only 17 42 (Marevan or Fragmin$ or Fraxiparin$ or Klexane).tw. (638) 18 19 43 exp Pipecolic acids/ae, tu (3633) 20 44 exp Vitamin K/ai (2564) 21 45 Vitamin K antagonist$.tw. (4641) 22 46 exp Antithrombins/ae, pd, de, tu (6778) 23 24 47 exp Blood coagulation factors/ai, de (16900) 25 48 exp Blood coagulation/de (19187) 26 49 (anticoagulat$ or antithromb$).tw. (67390) 27 50 or/38-49 (376368) 28 Annotation: anticoagulants as per Berge 2002 CDSR 29 30 51 Factor Xa Inhibitors/ (3589) 31 52 Dabigatran/ (2442)

32 53 Rivaroxaban/ (2106) http://bmjopen.bmj.com/ 33 54 (dabigatran or rivaroxaban or apixaban or edoxaban).mp. (7012) 34 35 55 anti-factor Xa.mp. (893) 36 56 (factor Xa adj2 (antag* or inhibit*)).mp. (5256) 37 57 novel oral anticoagulant*.mp. (1068) 38 58 noac.mp. (829) 39 59 noacs.mp. (1129)

40 on September 27, 2021 by guest. Protected copyright. 41 60 pradax.mp. (8) 42 61 pradaxa.mp. (127) 43 62 BIBR-953.mp. (8) 44 63 BIBR-953ZW.mp. (2) 45 46 64 xarelto.mp. (114) 47 65 BAY 59-7939.mp. (27) 48 66 BMS-562247.mp. (6) 49 67 eliquis.mp. (43) 50 68 lixiana.mp. (13) 51 52 69 DU-176.mp. (1) 53 70 DU-176b.mp. (27) 54 71 non-vitamin K.mp. (828) 55 72 or/51-71 (11701) 56 57 73 direct oral anticoagulant*.mp. (1161) 58 74 DOAC.mp. (449) 59 75 DOACs.mp. (570) 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 38 of 56

76 TSOAC.mp. (25) BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 2 77 TSOACs.mp. [mp=title, abstract, original title, name of substance word, subject heading word, keyword 3 heading word, protocol supplementary concept word, rare disease supplementary concept word, unique 4 identifier, synonyms] (40) 5 78 oral anticoagulant.mp. (4604) 6 79 (new or novel or direct or direct-acting or target-specific or targeted or non-vitamin K).mp. [mp=title, 7 8 abstract, original title, name of substance word, subject heading word, keyword heading word, protocol 9 supplementary concept word, rare disease supplementary concept word, unique identifier, synonyms] 10 (3920515) 11 80 78 and 79 (1801) 12 13 81 72 or 73 or 74 or 75 or 80 (12741) 14 Annotation: NOACs or DOACs from 2016 Canadian Cardiac Society GL search RC 15 82 exp Platelet aggregation inhibitors/ (106651) 16 83 (antiplatelet$ or anti-platelet$ or antiaggreg$ or anti-aggreg$ or (platelet$ adj5 inhibit$) or (thrombocyt$ For peer review only 17 adj5 inhibit$)).tw. (50701) 18 19 84 (alprostadil$ or aspirin$ or dipyridamol$ or disintegrin$ or epoprostenol$ or iloprost$ or ketanserin$ or 20 ketorolac tromethamine$ or milrinone$ or mopidamol$ or pentoxifyllin$ or procainamide$ or ticlopidine$ or 21 thiophen$ or trapidil$).tw. (85938) 22 85 (acetyl salicylic acid$ or acetyl?salicylic acid or clopidogrel$ or picotamide$ or ligustrazine$ or levamisol$ 23 24 or suloctidil$ or ozagrel$ or oky046 or oky-046 or defibrotide$ or cilostazol or satigrel or sarpolgrelate or 25 kbt3022 or kbt-3022 or isbogrel or cv4151 or cv-4151 or triflusal).tw. (27863) 26 86 (Dispril or Albyl$ or Ticlid$ or Persantin$ or Plavix).tw. (707) 27 87 exp Platelet glycoprotein gpiib-iiia complex/ai, de (3381) 28 88 (((glycoprotein iib$ or gp iib$) adj5 (antagonist$ or inhibitor$)) or GR144053 or GR-144053 or abciximab$ 29 30 or tirofiban$ or eftifibatid$).tw. (5389) 31 89 (ReoPro or Integrilin$ or Aggrastat).tw. (468)

32 90 exp Platelet activation/de (26939) http://bmjopen.bmj.com/ 33 91 exp Blood platelets/de (19928) 34 35 92 or/82-91 (202643) 36 Annotation: antiplatelets as per Berge 2002 CDSR 37 93 37 or 50 or 81 or 92 (2776671) 38 Annotation: Intervention block 39 94 31 and 93 (366)

40 on September 27, 2021 by guest. Protected copyright. 41 95 remove duplicates from 94 (326) 42 43 Central 44 Date Run: 16/10/17 21:33:56.134 45 46 Description: 47 48 ID Search Hits 49 #1 MeSH descriptor: [Stroke] explode all trees 7179 50 #2 MeSH descriptor: [Cerebrovascular Disorders] explode all trees 12133 51 52 #3 (isch*emi* near/6 (stroke* or apoplex* or cerebral next vasc* or cerebrovasc* or cva or attack*)) 53 10210 54 #4 ((brain or cerebr* or cerebell* or vertebrobasil* or hemispher* or intracran* or intracerebral or 55 infratentorial or supratentorial or middle next cerebr* or mca* or "anterior circulation") near/5 (isch*emi* or 56 57 infarct* or thrombo* or emboli* or occlus* or hypoxi*)) 11401 58 #5 (tia or tias) 1201 59 #6 MeSH descriptor: [Embolism, Paradoxical] explode all trees 15 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 39 of 56 BMJ Open

#7 ((paradoxic* or crossed) near/5 embolism*) 49 BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 2 #8 (cryptogenic near/5 stroke) 143 3 #9 MeSH descriptor: [Heart Atria] explode all trees 565 4 #10 MeSH descriptor: [Embolism and Thrombosis] explode all trees 6758 5 #11 #9 and #10 49 6 #12 atria* near/3 (emboli* or thromboemboli*) 206 7 8 #13 #1 or #2 or #3 or #4 or #5 or #6 or #7 or #8 or #11 or #12 24313 9 #14 MeSH descriptor: [Heart Septal Defects, Atrial] explode all trees 187 10 #15 "patent foramen ovale" or PFO 277 11 #16 atrial sept* near/5 defect* 290 12 13 #17 (("right to left" or "R-L" or "venous to arterial" or "venous-arterial" or "V-A") near/3 shunt) 71 14 #18 #14 or #15 or #16 or #17 567 15 #19 #13 and #18 Publication Year from 2012 to 2017 81 16 For peer review only 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31

32 http://bmjopen.bmj.com/ 33 34 35 36 37 38 39

Appendix 2: Tables BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 2 3 eTable 1: Characteristics of eligible studies 4 5 6 7 8 PFO closure plus antiplatelet vs antiplatelet therapy 9 Percent PFO Percent 10 Crossover from closure technical med therapy to Crossover from performed success/PFO 11 PFO closure PFO closure to /patients in closure 12 Study, year Intervention Medical therapy Withdrawn or loss to follow-up (%) med therapy PFO cohort attempt 13 Medical 14 PFO (%) therapy (%) STARFlex Device. 15 Clopidogrel for 6 16 months and ASA for 2 ForASA 325mg daily peer review only 17 Furlan, 2012 years (56%) 1.8 0.65 1.73 8.72 90.6 89.4 18 19

29 30 Amplatzer occluder. Antiplatelet 31 ASA 5-6 months. (67%) or

32 clopidogrel or anticoagulation http://bmjopen.bmj.com/ 33 Meier, 2013 ticlopidine 1-6 months (31%) 15.2 19.5 13.3 6.4 96.1 97.4 34 ASA, warfarin, Amplatzer occluder clopidogrel 35 100%. ASA 81-325mg (25%), or ASA 36 daily plus clopidogrel plus 37 75mg daily x 1 months dipyridamole 38 Saver, 2017 then ASA x 5 months (75%) 20.8 33.3 3.95 2.8 93.6 99.6 PFO Closure 39 (Cardoiform Septal

40 occluder 61%, Gore on September 27, 2021 by guest. Protected copyright. 41 helex Septal occluder 42 39%) plus antiplatelet 43 (ASA 75-325mg daily, Antiplatelet (ASA aggrenox, or 75-325mg daily, 44 clopidogrel. Clopidogrel ASA plus 45 300mg x 1 then 75 dipyridamole, or 46 Sondergaard, daily x 3 days for clopidogrel. 47 2017 everyone (100%) 8.3 14.7 6.3 7.3 96.1 97.4 PFO Closure (100% 48 Amplatzer PFO Antiplatelet with 49 occluder). Suggested ASA or ASA and 50 DAPT for 6 months Clopidogrel 75mg 51 with ASA 100mg and daily, or ASA and 52 clopidogrel 75mg daily. Cilostazol 200mg However, up to daily. Could 53 treating physician. anticoagulate 54 Could even use with warfarin 55 Lee, 2018 anticoagulation (INR 2.0-3.0) 0 0 0 11.7 88.3 100 56 PFO closure plus antiplatelet vs anticoagulation 57 Percent Percent 58 Crossover from Crossover from technical technical Study, year Intervention Control Withdrawn or loss to follow-up antiplatelet to anticoagulation success/PFO success/PFO 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 41 of 56 BMJ Open

anticoagulation to antiplatelet closure closure BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 (%) (%) attempts attempts 2 PFO Closure 3 plus antiplatelet Anticoagulation 4 5 6 PFO closure (Amplatzer 7 51.5%†) plus ASA 75mg Warfarin with

8 daily and clopidogrel target INR 2-3 75mg daily x 3 months (93%) or direct 9 then aspirin or oral 10 clopidogrel or aspirin anticoagulant 11 Mas 2017 plus ER dipyridamole (7%) NA NA NA NA NA NA 12 Anticoagulation vs antiplatelet therapy 13 Crossover from Crossover from 14 antiplatelet to anticoagulation 15 anticoagulation to antiplatelet 16 Study, year Intervention ForControl peerWithdrawn or loss to follow review-up only(%) (%) 17 Anticoagulation 18 (%) Antiplatelet (%) 19 20 21

32 http://bmjopen.bmj.com/ 33 34 35 36 37 38 39

40 on September 27, 2021 by guest. Protected copyright. 41 eFigure 1: Risk of bias of included randomised controlled trials 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 43 of 56 BMJ Open

Appendix 4: PFO closure vs antiplatelet therapy forest plots (Direct evidence only) BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 eFigure 1: Ischaemic stroke 16 For peer review only 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 eFigure 2: TIA

32 http://bmjopen.bmj.com/ 33 34 35 36 37 38 39

40 on September 27, 2021 by guest. Protected copyright. 41 42 43 44 45 46 47 eFigure 3: Systemic embolism 48 49 50 51 52 53 54 55 56 57 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 44 of 56 BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 2 3 4 5 6 7 8 9 10 11 eFigure 4: Death 12 13 14 15 16 For peer review only 17 18 19 20 21 22 23 24 25 26 27 eFigure 5: Major bleeding 28 29 30 31

32 http://bmjopen.bmj.com/ 33 34 35 36 37 38 39

40 on September 27, 2021 by guest. Protected copyright. 41 42 43 eFigure 6: Pulmonary embolism 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 45 of 56 BMJ Open

Appendix 5: PFO closure vs Anticoagulation forest plots (Direct evidence only) BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 2 3 4 5 6 7 8 9 10 11 12 eFigure 1: Ischaemic stroke 13 14 15 16 For peer review only 17 18 19 20 21 22 23 24 25 26 27 28 eFigure 2: TIA 29 30 31

32 http://bmjopen.bmj.com/ 33 34 35 36 37 38 39

40 on September 27, 2021 by guest. Protected copyright. 41 eFigure 3: Systemic embolism 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 eFigure 4: Pulmonary embolism 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 46 of 56

BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 2 3 4 5 6 7 8 9 10 11 eFigure 5: Major bleeding 12 13 14 15 16 For peer review only 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31

32 http://bmjopen.bmj.com/ 33 34 35 36 37 38 39 40 on September 27, 2021 by guest. Protected copyright. 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 47 of 56 BMJ Open

Appendix 6: Complications of PFO closure vs medical therapy forest plots (Direct evidence only) BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 eFigure 1: All atrial fibrillation For peer review only 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31

32 http://bmjopen.bmj.com/ 33 34 eFigure 2: Persistent atrial fibrillation 35 36 37 38 39

40 on September 27, 2021 by guest. Protected copyright. 41 42 43 44 45 46 47 48 49 50 51 52 53 54 eFigure 3: Transient atrial fibrillation 55 56 57 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 48 of 56

BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 eFigure 4: Device or procedure related adverse events 16 For peer review only 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31

32 http://bmjopen.bmj.com/ 33 34 35 36 37 38 39

Appendix 7: Anticoagulation vs antiplatelet therapy forest plots (Direct evidence only) BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 2 3 4 5 6 7 8 9 10 11 12 13 eFigure 1: Ischaemic stroke 14 15 16 For peer review only 17 18 19 20 21 22 23 24 25 26 27 28 29 eFigure 2: TIA 30 31

32 http://bmjopen.bmj.com/ 33 34 35 36 37 38 39

40 on September 27, 2021 by guest. Protected copyright. 41 42 43 eFigure 3: Major bleeding 44 45 46 47 48 49 50 51 52 53 54 55 56 eFigure 4: Death 57 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 50 of 56

Appendix 8: Meta-regression and NMA Figures BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 For peer review only 17 18 19 20 21 22 23 24 25 26 27 28 eFigure 1: Meta-regression evaluating the effect of the proportion of participants with a moderate or higher 29 30 31 shunt size on the effectiveness of PFO closure versus medical therapy in patients with a PFO and cryptogenic

32 http://bmjopen.bmj.com/ 33 stroke. PFO closure was more effective in RCTs that had a higher proportion of patients with moderate or 34 35 higher PFO shunts, p=0.047. Note, the shunt size was not reported in the DEFENSE-PFO Trial and thus it was 36 37 38 not included in the analysis. 39

32 http://bmjopen.bmj.com/ 33 the DEFENSE-PFO Trial and thus it was not included in the analysis. 34 35 36 37 38 39

40 on September 27, 2021 by guest. Protected copyright. 41 42 43 44 45 46 47 48 49 50 51 52 eFigure 3: Network of included RCTs with available direct comparisons for ischaemic stroke. 53 54 Note: The thickness of the lines (edges) is proportional to the number of RCTs evaluating each treatment and direct 55 56 57 comparison. 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 52 of 56

Appendix 9: NMA Table BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 2 3 eTable 1: Direct and indirect estimates of effects 4 5 Indirect NMA 6 Outcomes Comparisons Direct [OR,95%CrI] 7 [OR,95%CrI]$ [OR,95%CrI] PFO+antiplatelet vs. 8 0.14 (0.05, 0.34) Not estimable* 0.12 (0.04, 0.27) 9 antiplatelet 10 PFO+antiplatelet vs. 9.6e-16 (2.0e-47, Ischemic stroke 0.45 (0.06, 3.18) 0.44 (0.08, 3.76) 11 anticoagulation 6.3e+16) 12 Antiplatelet vs. 13 4.67 (0.80, 26.95) Not estimable* 3.73 (0.84, 28.11) 14 anticogualation 15 PFO+antiplatelet vs. 3.1e+14 (4.32, 8.5e+62) Not estimable* 2.48 (0.16, 99.11) 16 antiplatelet For peer review only 17 PFO+antiplatelet vs. 3.4e-06 (1.5e-14, 18 Death 0.97 (0.02, 47.77) 0.59 (0.01, 22.76) 19 anticoagulation 728.3) 20 Antiplatelet vs. 0.72 (0.01, 21.93) Not estimable* 0.23 (0.01, 3.44) 21 anticogualation 22 PFO+antiplatelet vs. 23 0.54 (0.21, 1.40) Not estimable* 0.48 (0.20, 1.10) 24 antiplatelet 25 Major bleeding PFO+antiplatelet vs. 0.09 (0.00, 2.15) 0.30 (0.08, 1.12) 0.25 (0.07, 0.82) 26 anticoagulation 27 Antiplatelet vs. 28 0.60 (0.19, 1.80) Not estimable* 0.53 (0.18, 1.47) 29 anticogualation 30 PFO+antiplatelet vs. 0.75 (0.26, 2.13) Not estimable* 0.82 (0.31, 2.09) 31 antiplatelet

32 http://bmjopen.bmj.com/ PFO+antiplatelet vs. 33 1.51 (0.34, 6.72) 0.88 (0.11, 6.85) 1.25 (0.39, 4.38) 34 Recurrent TIA anticoagulation 35 Antiplatelet vs. 1.28 (0.26, 6.32) Not estimable* 1.53 (0.50, 5.21) 36 anticogualation 37 PFO+antiplatelet vs. 38 0.79 (0.12, 6.76) Not estimable* 0.73 (0.12, 5.52) antiplatelet 39 40 PFO+antiplatelet vs. 0.06 (5.7e-09, 0.03 (4.0e-08, on September 27, 2021 by guest. Protected copyright. Systemic embolism 0.01 (1.4e-10, 1.0e+06) 41 anticoagulation 6.0e+05) 1280.0) 42 Antiplatelet vs. 0.04 (5.4e-08, 43 0.01 (1.4e-10, 4.0e+05) Not estimable* 44 anticogualation 1616.0) 45 $ Indirect results were from back-calculated 46

47 * Cannot be estimated because the intervention was not connected in a loop in the evidence network 48 49 50 51 52 53 54 55 56 57 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 53 of 56 BMJ Open BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 PRISMA NMA Checklist of Items to Include When Reporting A Systematic Review 2 Involving a Network Meta-analysis 3 4 Section/Topic Item Checklist Item Reported 5 # on Page # 6 TITLE 7 Title 1 Identify the report as a systematic review incorporating a 1 8 network meta-analysis (or related form of meta-analysis). 9 10 11 ABSTRACT 2-3 12 Structured 2 Provide a structured summary including, as applicable: 13 summary Background: main objectives 14 Methods: data sources; study eligibility criteria, participants, 15 and interventions; study appraisal; and synthesis methods, 16 Forsuch peer as network review meta-analysis. only 17 Results: number of studies and participants identified; 18 19 summary estimates with corresponding confidence/credible 20 intervals; treatment rankings may also be discussed. Authors 21 may choose to summarize pairwise comparisons against a 22 chosen treatment included in their analyses for brevity. 23 Discussion/Conclusions: limitations; conclusions and 24 implications of findings. 25 Other: primary source of funding; systematic review 26 registration number with registry name. 27 28 29 INTRODUCTION 30 Rationale 3 Describe the rationale for the review in the context of what is 4-5 31 already known, including mention of why a network meta-

40 Protocol and 5 Indicate whether a review protocol exists and if and where it can 5 on September 27, 2021 by guest. Protected copyright. 41 registration be accessed (e.g., Web address); and, if available, provide 42 registration information, including registration number. 43 Eligibility criteria 6 Specify study characteristics (e.g., PICOS, length of follow-up) 6 44 and report characteristics (e.g., years considered, language, 45 publication status) used as criteria for eligibility, giving 46 rationale. Clearly describe eligible treatments included in the 47 treatment network, and note whether any have been clustered or 48 49 merged into the same node (with justification). 50 Information 7 Describe all information sources (e.g., databases with dates of 6 51 sources coverage, contact with study authors to identify additional 52 studies) in the search and date last searched. 53 Search 8 Present full electronic search strategy for at least one database, 6, 35-39 54 including any limits used, such that it could be repeated. 55 56 57 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 54 of 56 BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 Study selection 9 State the process for selecting studies (i.e., screening, eligibility, 6-7 2 included in systematic review, and, if applicable, included in the 3 meta-analysis). 4 Data collection 10 Describe method of data extraction from reports (e.g., piloted 6 5 process forms, independently, in duplicate) and any processes for 6 obtaining and confirming data from investigators. 7 Data items 11 List and define all variables for which data were sought (e.g., 6 8 PICOS, funding sources) and any assumptions and 9 simplifications made. 10 Geometry of the S1 Describe methods used to explore the geometry of the treatment 8-9 11 network network under study and potential biases related to it. This 12 should include how the evidence base has been graphically 13 summarized for presentation, and what characteristics were 14 compiled and used to describe the evidence base to readers. 15 Risk of bias within 12 Describe methods used for assessing risk of bias of individual 7-8 16 For peer review only 17 individual studies studies (including specification of whether this was done at the 18 study or outcome level), and how this information is to be used 19 in any data synthesis. 20 Summary 13 State the principal summary measures (e.g., risk ratio, difference 9-11 21 measures in means). Also describe the use of additional summary 22 measures assessed, such as treatment rankings and surface 23 under the cumulative ranking curve (SUCRA) values, as well as 24 modified approaches used to present summary findings from 25 meta-analyses. 26 Planned methods 14 Describe the methods of handling data and combining results of 9-11 27 of analysis studies for each network meta-analysis. This should include, but 28 not be limited to: 29 • Handling of multi-arm trials; 30 • Selection of variance structure; 31 • Selection of prior distributions in Bayesian analyses; 32 http://bmjopen.bmj.com/ 33 and 34 • Assessment of model fit. 35 Assessment of S2 Describe the statistical methods used to evaluate the agreement 9-11 36 Inconsistency of direct and indirect evidence in the treatment network(s) 37 studied. Describe efforts taken to address its presence when 38 found. 39 Risk of bias across 15 Specify any assessment of risk of bias that may affect the 7-8

40 studies cumulative evidence (e.g., publication bias, selective reporting on September 27, 2021 by guest. Protected copyright. 41 within studies). 42 Additional 16 Describe methods of additional analyses if done, indicating 9-11 43 analyses which were pre-specified. This may include, but not be limited 44 to, the following: 45 • Sensitivity or subgroup analyses; 46 • Meta-regression analyses; 47 48 • Alternative formulations of the treatment network; and 49 • Use of alternative prior distributions for Bayesian 50 analyses (if applicable). 51 52 53 54 55 56 57 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 55 of 56 BMJ Open BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 RESULTS† 2 3 Study selection 17 Give numbers of studies screened, assessed for eligibility, and 6,7, 30 4 included in the review, with reasons for exclusions at each stage, 5 ideally with a flow diagram. 6 Presentation of S3 Provide a network graph of the included studies to enable 44 7 network structure visualization of the geometry of the treatment network. 8 Summary of S4 Provide a brief overview of characteristics of the treatment 44 9 network geometry network. This may include commentary on the abundance of 10 11 trials and randomized patients for the different interventions and 12 pairwise comparisons in the network, gaps of evidence in the 13 treatment network, and potential biases reflected by the network 14 structure. 15 Study 18 For each study, present characteristics for which data were 26 16 characteristics Forextracted peer (e.g., study review size, PICOS, follow-up only period) and 17 provide the citations. 18 Risk of bias within 19 Present data on risk of bias of each study and, if available, any 43 19 studies outcome level assessment. 20 Results of 20 For all outcomes considered (benefits or harms), present, for 39-52 21 individual studies each study: 1) simple summary data for each intervention group, 22 and 2) effect estimates and confidence intervals. Modified 23 approaches may be needed to deal with information from larger 24 25 networks. 26 Synthesis of results 21 Present results of each meta-analysis done, including 27-32 27 confidence/credible intervals. In larger networks, authors may 28 focus on comparisons versus a particular comparator (e.g. 29 placebo or standard care), with full findings presented in an 30 appendix. League tables and forest plots may be considered to 31 summarize pairwise comparisons. If additional summary

32 measures were explored (such as treatment rankings), these http://bmjopen.bmj.com/ 33 should also be presented. 34 Exploration for S5 Describe results from investigations of inconsistency. This may 8 35 inconsistency include such information as measures of model fit to compare 36 consistency and inconsistency models, P values from statistical 37 tests, or summary of inconsistency estimates from different parts 38 of the treatment network. 39 Risk of bias across 22 Present results of any assessment of risk of bias across studies 43 40 studies for the evidence base being studied. on September 27, 2021 by guest. Protected copyright. 41 42 Results of 23 Give results of additional analyses, if done (e.g., sensitivity or 34 43 additional analyses subgroup analyses, meta-regression analyses, alternative 44 network geometries studied, alternative choice of prior 45 distributions for Bayesian analyses, and so forth). 46 47 DISCUSSION 48 49 Summary of 24 Summarize the main findings, including the strength of evidence 18-22 50 evidence for each main outcome; consider their relevance to key groups 51 (e.g., healthcare providers, users, and policy-makers). 52 Limitations 25 Discuss limitations at study and outcome level (e.g., risk of 21 53 bias), and at review level (e.g., incomplete retrieval of identified 54 research, reporting bias). Comment on the validity of the 55 assumptions, such as transitivity and consistency. Comment on 56 57 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 56 of 56 BMJ Open: first published as 10.1136/bmjopen-2018-023761 on 25 July 2018. Downloaded from 1 any concerns regarding network geometry (e.g., avoidance of 2 certain comparisons). 3 Conclusions 26 Provide a general interpretation of the results in the context of 22 4 other evidence, and implications for future research. 5 6 7 FUNDING 8 Funding 27 Describe sources of funding for the systematic review and other 22 9 support (e.g., supply of data); role of funders for the systematic 10 review. This should also include information regarding whether 11 funding has been received from manufacturers of treatments in 12 the network and/or whether some of the authors are content 13 experts with professional conflicts of interest that could affect 14 use of treatments in the network. 15 16 PICOS = population, Forintervention, peer comparators, review outcomes, study design.only 17 * Text in italics indicates wording specific to reporting of network meta-analyses that has been added to 18 guidance from the PRISMA statement. 19 † Authors may wish to plan for use of appendices to present all relevant information in full detail for 20 items in this section. 21 22 23 24 25 26 27 28 29 30 31

32 http://bmjopen.bmj.com/ 33 34 35 36 37 38 39

Correction: Patent foramen ovale closure, antiplatelet therapy or anticoagulation in patients with patent foramen ovale and cryptogenic stroke: a systematic review and network meta-analysis incorporating complementary external evidence

Mir H, Siemieniuk RAC, Ge LC, et al. Patent foramen ovale closure, antiplatelet therapy or anticoagulation in patients with patent foramen ovale and cryptogenic stroke: a systematic review and network meta-analysis incorporating complementary external evidence BMJ Open 2018;8:e023761. doi: 10.1136/bmjopen-2018-023761

The previous version of this manuscript contains an error in the author list. Author "Long Ge" was previously incorrectly listed as "Long Cruz Ge".

Open access This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/. © Author(s) (or their employer(s)) 2018. Re-use permitted under CC BY-NC. No commercial re-use. See rights and permissions. Published by BMJ.

BMJ Open 2018;8:e023761corr1. doi:10.1136/bmjopen-2018-023761corr1

BMJ Open 2018;8:e023761corr1. doi:10.1136/bmjopen-2018-023761corr1 1