Oral Anticoagulation Use and Left Atrial Appendage Occlusion in LAAOS III
LAAOS III (Left Atrial Appendage Occlusion Study III) showed that left atrial appendage (LAA) occlusion reduces the risk of ischemic stroke or systemic embolism in patients with atrial fibrillation undergoing cardiac surgery. This article examines the effect of LAA occlusion on stroke reduction according to variation in the use of oral anticoagulant (OAC) therapy.
Information regarding OAC use was collected at every follow-up visit. Adjusted proportional hazards modeling, including using landmarks of hospital discharge, 1 and 2 years after randomization, evaluated the effect of LAA occlusion on the risk of ischemic stroke or systemic embolism, according to OAC use. Adjusted proportional hazard modeling, with OAC use as a time-dependent covariate, was also performed to assess the effect of LAA occlusion, according to OAC use throughout the study.
At hospital discharge, 3027 patients (63.5%) were receiving a vitamin K antagonist, and 879 (18.5%) were receiving a non–vitamin K antagonist oral anticoagulant (direct OAC), with no difference in OAC use between treatment arms. There were 2887 (60.5%) patients who received OACs at all follow-up visits, 1401 (29.4%) who received OAC at some visits, and 472 (9.9%) who never received OACs. The effect of LAA occlusion on the risk of ischemic stroke or systemic embolism was consistent after discharge across all 3 groups: hazard ratios of 0.70 (95% CI, 0.51–0.96), 0.63 (95% CI, 0.43–0.94), and 0.76 (95% CI, 0.32–1.79), respectively. An adjusted proportional hazards model with OAC use as a time-dependent covariate showed that the reduction in stroke or systemic embolism with LAA occlusion was similar whether patients were receiving OACs or not.
The benefit of LAA occlusion was consistent whether patients were receiving OACs or not. LAA occlusion provides thromboembolism reduction in patients independent of OAC use.
What Is New?
A detailed analysis of the effect of left atrial appendage occlusion in patients receiving or not receiving oral anticoagulation during follow-up was performed.
This report provides extensive information about the use and nonuse of anticoagulants during LAAOS III (Left Atrial Appendage Occlusion Study III).
Data provide support for the concept that left atrial appendage occlusion provides protection against ischemic stroke and systemic embolism that is independent of the use of oral anticoagulant therapy.
What Are the Clinical Implications?
The addition of left atrial appendage occlusion to anticoagulation is a promising new paradigm for stroke prevention in high-risk patients with atrial fibrillation.
For surgical patients receiving left atrial appendage occlusion, there is a strong rationale for continuation of anticoagulation unless contraindicated.
Ischemic stroke in patients with atrial fibrillation is lessened with oral anticoagulants. On the basis of several randomized trials, the efficacy, safety, and net benefit of these agents have been conclusively established and they have become the mainstay for stroke prevention in patients with atrial fibrillation.1,2 An alternate approach to stroke reduction in atrial fibrillation is to mechanically prevent the left atrial thrombus from reaching the systemic circulation through excluding the left atrial appendage from systemic circulation. Randomized trials of an endovascular appendage occlusion device demonstrated that device closure of the left atrial appendage (LAA) was noninferior to oral anticoagulant (OAC) therapy for a composite outcome of ischemic and hemorrhagic events, but they failed to demonstrate a reduction in ischemic stroke itself.3 LAAOS III (Left Atrial Appendage Occlusion Study III) recently demonstrated that surgical LAA occlusion or removal reduced the risk of ischemic stroke or systemic embolism; the hazard ratio in favor of LAA occlusion was 0.67 (95% CI, 0.53–0.85; P=0.001).4
LAAOS III tested LAA occlusion in patients who were also receiving regular care, and this included OAC therapy for stroke prevention in most. Because LAA occlusion provides permanent stroke preventive therapy in case of discontinuation from, or poor adherence to, OAC therapy, we hypothesized that the LAA occlusion would be complementary to OAC therapy and would provide additional benefit to patients already treated with OACs.
We previously reported that a high proportion of patients in both treatment arms of LAAOS III received OAC therapy; however, details of OAC use and whether it modified the effects of LAA occlusion has not been reported. The purpose of this article is to provide more details on OAC use during the study and to describe whether the effect of LAA occlusion was consistent regardless of whether patients used OAC.
The study protocol has previously been published,5 as have the main results of LAAOS III.4 In brief, this was a randomized trial of LAA occlusion versus no LAA occlusion at the time of other scheduled cardiac surgery. Patients were eligible if they were undergoing cardiac surgery with the use of cardiopulmonary bypass, had atrial fibrillation, and had ≥2 risk factors for stroke. Neither patients nor the health professionals overseeing antithrombotic therapy were aware of treatment assignments. The primary outcome of the trial was time to ischemic stroke or systemic embolism; all primary events occurring during the trial were adjudicated by a blinded committee consisting of stroke experts.
Participants were expected to receive guideline-directed stroke prevention therapy including OAC agents, but the exact use, or nonuse, of such drugs and their control was left in the hands of local physicians. Physicians could start, stop, and restart OACs at their discretion. At every follow-up contact, we collected data on OAC use or specific reasons for nonuse. Follow-up visits occurred at the time of hospital discharge, at 1 month and 6 months after surgery, and every 6 months thereafter until a common end date. The data, analytic methods, and study materials will not be made available to other researchers for purposes of reproducing the results. The protocol received ethical approval at all centers, and all patients provided written informed consent.
To assess whether OAC use modified the efficacy of LAA occlusion, Cox proportional hazard models for the primary outcome of ischemic stroke or systemic embolism, using different landmarks for initiation of follow-up, were created. These analyses were adjusted for the following risk factors for stroke: age, previous stroke or transient ischemic attack, hypertension, diabetes, heart failure, and previous myocardial infarction. The population included all patients randomly assigned who also had surgery. The statistical evaluation was performed using the software package SAS, release 9.2 (SAS Institute Inc., Cary, NC). In these analyses, patients were stratified according to OAC use in several different ways. In the first analysis, patients were stratified into 3 groups according to their OAC use during the entire follow-up period: (1) patients receiving OACs at every visit; (2) patients receiving OACs at some visits (but not all); and (3) patients not receiving OACs at any visit. In 3 landmark analyses, we stratified patients according to OAC use reported at 3 different times: (1) hospital discharge; (2) the 1-year visit; and (3) the 2-year visit. In another landmark analysis, patients were stratified according to the prescription of a vitamin K antagonist (VKA), direct oral anticoagulant (DOAC), or neither at the 1-year visit. We also did an unadjusted Cox proportional hazard landmark analysis beginning at hospital discharge for the outcome of major bleeding. In all these analyses, patients were included if they had the visit that initiated the landmark analysis, and we only included events occurring after the visit. These analyses were performed based on time to first primary outcome event and were adjusted for baseline risk of stroke. We reformulated the data into a counting process format, and the hazard ratio of LAA occlusion compared with no occlusion was calculated, adding OAC use as a time-dependent covariate. We then stratified the follow-up time of the reformulated data as “on OAC” and “not on OAC.” We also created a proportional hazards model using OAC use or nonuse as a time-dependent variable to stratify follow-up time. Patients could contribute days of follow-up to both the on OAC and not on OAC follow-up time according to their use of OACs as recorded at each 6-month visit. We calculated the adjusted hazard ratio and 95% CI for LAA occlusion compared with no occlusion for patient follow-up on OAC and not on OAC. Pinteraction values were calculated for each of these analyses. We did not adjust P values for multiple analyses, and we did not specify a critical threshold P value, as these analyses are considered descriptive and hypothesis generating.
Use of OACs During Follow-Up
Table 1 shows the baseline characteristics of patients, according to their use of OACs both at hospital discharge and at the 1-year visit. Patients receiving and not receiving OACs had similar CHA2DS2-VASc scores (a calculated stroke risk score for patients with atrial fibrillation). However, those receiving OACs were more likely to have persistent or permanent atrial fibrillation and to have had valve surgery.
|Characteristics||At hospital discharge||At 1-year visit|
|On OAC||Off OAC||P value||On OAC||Off OAC||P value|
|Type of atrial fibrillation, n (%)|
|Paroxysmal||1721 (44.1)||558 (65.9)||<0.01||1442 (43.0)||614 (70.0)||<0.01|
|Persistent||931 (23.8)||148 (17.5)||782 (23.3)||166 (18.9)|
|Permanent||1254 (32.1)||141 (16.6)||1126 (33.6)||97 (11.1)|
|Male, n (%)||2595 (66.4)||610 (72.0)||<0.01||2239 (66.8)||626 (71.4)||0.01|
|Myocardial infarction, n (%)||837 (21.4)||310 (36.6)||<0.01||707 (21.1)||275 (31.4)||<0.01|
|Stroke, n (%)||352 (9.0)||79 (9.3)||0.77||312 (9.3)||60 6.8)||0.02|
|Transient ischemic attack, n (%)||225 (5.8)||45 (5.3)||0.61||208 (6.2)||30 (3.4)||<0.01|
|Rheumatic heart disease, n (%)||304 (7.8)||23 (2.7)||<0.01||234 (7.0)||53 (6.0)||0.32|
|Peripheral artery disease, n (%)||382 (9.8)||110 (13.0)||<0.01||324 (9.7)||90 (10.3)||0.60|
|Heart failure, n (%)||2257 (57.8)||451 (53.2)||0.02||1918 (57.2)||452 (51.5)||<0.01|
|Diabetes, n (%)||1227 (31.4)||305 (36.0)||<0.01||1046 (31.2)||278 (31.7)||0.79|
|Hypertension, n (%)||3166 (81.0)||721 (85.1)||<0.01||2742 (81.8)||717 (81.8)||0.95|
|Median (Q1–Q3)||4.0 (3.0–5.0)||4.0 (3.0–5.0)||4.0 (3.0–5.0)||4.0 (3.0–5.0)|
|Rhythm on baseline ECG, n (%)|
|Atrial fibrillation||2395 (61.3)||323 (38.1)||<0.01||2097 (62.6)||296 (33.8)||<0.01|
|Aflutter||137 (3.5)||31 (3.7)||0.82||121 (3.6)||28 (3.2)||0.55|
|Sinus||1346 (34.5)||487 (57.5)||<0.01||1111 (33.2)||546 (62.3)||<0.01|
|Other||20 (0.5)||5 (0.6)||0.78||15 (0.4)||6 (0.7)||0.38|
|Procedure, n (%)|
|Isolated coronary artery bypass graft||711 (18.2)||291 (34.4)||<0.01||654 (19.5)||258 (29.4)||<0.01|
|Isolated valve||989 (25.3)||132 (15.6)||<0.01||871 (26.0)||120 (13.7)||<0.01|
|Any valve procedure||2740 (70.2)||425 (50.2)||<0.01||2341 (69.9)||453 (51.6)||<0.01|
|Any aortic procedure||235 (6.0)||44 (5.2)||0.36||201 (6.0)||44 (5.0)||0.27|
|Other||2204 (56.4)||424 (50.1)||<0.01||1823 (54.4)||499 (56.9)||0.19|
|Atrial ablation procedure, n (%)||1314 (33.6)||243 (28.7)||<0.01||1031 (30.8)||375 (42.8)||<0.01|
At the time of hospital discharge after surgery, 3906 patients (82.2%) were receiving an OAC (83.4% of patients randomly assigned to receive LAA occlusion and 81.0% of those randomly assigned not to receive LAA occlusion). Figure 1 shows the use of OACs at all subsequent follow-up visits. There were no important differences in OAC use between the 2 randomized groups. The frequency of OAC use declined gradually to 75% at the 5-year follow-up visit. At hospital discharge, 3027 patients (63.5%) were receiving a VKA, and 879 (18.5%) were receiving a DOAC. VKA use was similar between the LAA occlusion and no LAA occlusion arms (64.5% versus 62.4%, respectively). Figure 2 shows that the frequency of DOAC prescription increased over time, partially replacing VKA, such that by the 5-year visit, the frequencies of VKA and DOAC use were nearly the same. There were no appreciable differences in frequencies of DOAC prescription between LAA occlusion and no LAA occlusion treatment arms at 5 years (37.9 % versus 39.4 %, respectively).
For patients receiving a VKA, Figure S1 shows mean percentages of patients within, below, and above the therapeutic international normalized ratio (INR) range of 2.0 to 3.0 based on the most recent INR reported at each follow-up contact. INR control improved during follow-up. At the 30-day visit, 1679 patients (57.6%) were in the therapeutic range: 57.1% of those in the LAA occlusion arm and 58.0% of those in the no LAA occlusion arm. By the 1-year visit, INR control had improved to 72.9% and 72.2% in the therapeutic range in the LAA occlusion and no occlusion arms, respectively.
Table S1 shows the reasons for not using OACs at the 30-day and 3-year visits. The reasons did not differ significantly between treatment arms. Patient or physician refusal was the reason for nonuse of OACs for about one-quarter of cases. Bleeding risk was less frequently given as a reason for nonuse of OACs over time, whereas low stroke risk was stated more commonly at the 3-year visit than at the 30-day visit.
Proportional Hazard Analyses According to OAC Use
Table 2 presents the results of adjusted landmark analyses, according to OAC use, for ischemic stroke or systemic embolism. The benefit of LAA occlusion was consistent across all 3 groups, with hazard ratios (95% CI) in favor of LAA occlusion of 0.70 (0.51–0.96), 64 of 1434 (1.2%) versus 92 of 1453 (1.7%); 0.63 (0.43–0.94), 41 of 722 (1.4%) versus 63 of 679 (2.4%); and 0.76 (0.32–1.79), 9 of 217 (1.6%) versus 13 of 255 (2.0%) for patients on OACs at all visits, at some visits, and at no visits. Results from landmark analyses demonstrate a significant interaction at hospital discharge (Pinteraction=0.02), but not at the 1-year visit (Pinteraction=0.80) or at the 2-year visit (Pinteraction=0.50). In this subgroup of patients not receiving OACs at hospital discharge, there appears to be no significant effect of LAA occlusion on the primary outcome; however, the 95% CI of the hazard ratio is wide (0.61–2.89). The adjusted landmark analysis starting at the 1-year visit that stratified patients receiving OACs, according to whether they were prescribed a VKA or DOAC or neither, failed to demonstrate a statistically significant interaction (Pinteraction=0.38; Figure 3).
|Landmarks||Left atrial appendage occlusion||No left atrial appendage occlusion||Hazard ratio* (95% CI)||Pinteraction|
|On OAC at all visits||64/1434||1.2||92/1453||1.7||0.70 (0.51–0.96)||0.88|
|On OAC at 1%–99% visits||41/722||1.4||63/679||2.4||0.63 (0.43–0.94)|
|On OAC at no visits†||9/217||1.6||13/255||2.0||0.76 (0.32–1.79)|
|On OAC||55/1923||0.7||95/1874||1.3||0.55 (0.40–0.77)||0.02|
|Not on OAC||15/324||1.2||12/354||0.8||1.33 (0.61–2.89)||.|
|At 1 y (365 days)||.|
|On OAC||28/1631||0.6||54/1600||1.1||0.51 (0.32–0.81)||0.80|
|Not on OAC||6/426||0.4||14/439||1.0||0.42 (0.16–1.11)||.|
|At 2 y (730 days)||.|
|On OAC||17/1517||0.5||35/1492||1.1||0.48 (0.27–0.87)||0.50|
|Not on OAC||4/457||0.4||12/438||1.1||0.31 (0.10–0.99)||.|
|Patients on DOAC at 1-y visit||16/718||0.8||22/718||1.1||0.74 (0.39–1.40)||0.38|
|Patients on VKA at 1-y visit||12/911||0.4||32/882||1.1||0.36 (0.18–0.69)||.|
|Patients on neither DOAC nor VKA at 1-y visit||6/426||0.4||13/438||0.9||0.45(0.17–1.20)||.|
Table 3 demonstrates that the rates of major bleeding were lower in patients receiving OACs at all visits compared with those receiving OACs at some visits, and especially compared with patients who never received OACs during follow-up. Table 3 also shows that LAA occlusion had no effect of bleeding risk.
|Landmark||Left atrial appendage occlusion||No left atrial appendage occlusion||Hazard ratio (95% CI)||Pinteraction|
|Events/patients||Rate/100 person-years||Events/patients||Rate/100 person-years|
|On oral anticoagulant at all visits||116/1434||2.2||129/1453||2.5||0.91 (0.70–1.16)||0.88|
|On oral anticoagulant at 1%–99% visits||92/722||3.5||88/679||3.4||1.00 (0.74–1.34)|
|On oral anticoagulant at no visits||39/217||7.5||48/255||7.7||0.95 (0.62–1.45)|
Time-Dependent Covariate Analysis
A Cox proportional hazards model was created for the primary outcome with status of on OAC and off OAC as a time-dependent covariate based on the anticoagulation prescription at each follow-up visit starting at hospital discharge. As shown in Table 4, the effect of LAA occlusion as measured by the hazard ratio and 95% CI is nearly identical for follow-up periods on OAC (0.63 [0.45–0.89]) and not on OAC (0.63 [0.32–0.123]).
|Left atrial appendage occlusion||No left atrial appendage occlusion||Hazard ratio (95% CI)*||Hazard ratio (95% CI)|
|Events||Incidence/100 person-years||Events||Incidence/100 person-years|
|On oral anticoagulant||56||0.8||85||1.3||0.64 (0.48–0.87)||0.63 (0.45–0.89)|
|Not on oral anticoagulant||14||0.8||22||1.2||0.63 (0.32–1.23)|
The results of these analyses indicate that LAA occlusion reduces ischemic stroke, regardless of whether patients are prescribed OACs. Because most patients in LAAOS III were prescribed OACs most of the time during follow-up, far more strokes were prevented in LAAOS III by LAA occlusion when patients were prescribed OACs than when they were not.
In designing this study, we hypothesized that LAA occlusion would be complementary to OAC therapy because LAA occlusion and OAC therapy each reduce ischemic stroke by very different mechanisms. OACs are a group of systemic therapies that inhibit or reduce coagulation proteins throughout the body. On the other hand, LAA occlusion is a mechanical barrier to thrombus propagation and embolism from its most common location of origin. Furthermore, the mechanism by which LAA occlusion reduces ischemic stroke mitigates a major limitation of OACs, namely their dependence on regular daily long-term administration. The effectiveness of prophylactic drug therapy depends fundamentally on patients being prescribed the correct dose, and on patients consistently taking the prescribed dose. OAC therapies, especially DOACs, have a relatively short half-life, and even small interruptions provide opportunities for thrombus formation and embolization, rendering therapy ineffective. Because OAC therapy increases the risk of bleeding, some drug interruptions are inevitable, especially in elderly patients with other medical issues, which may need to be addressed by surgery and procedures. One distinct advantage of LAA occlusion is that after it is performed, its effects are permanent.
Using LAA occlusion together with anticoagulation is a new approach to ischemic stroke prevention in atrial fibrillation. Until now, LAA occlusion has been mostly seen as an alternative to anticoagulation. In large part, this is because achieving LAA occlusion has been considered a high-risk option, except in patients already undergoing open heart procedures for other reasons. However, the techniques for LAA occlusion with minimally invasive surgical procedures or by endovascular device insertion have been refined and made safer.6 Thus, combining anticoagulation with LAA occlusion using these approaches is now more practical. Clinical trials to test whether LAA occlusion provides benefit on top of OAC therapy in a more general population of individuals with atrial fibrillation are now warranted.
Patients with atrial fibrillation and multiple risk factors have an ongoing risk of ischemic stroke despite OAC therapy. For those with a CHA2DS2-VASc score of ≥4, the risk of stroke per year is ≥2%.7–9 Patients with atrial fibrillation who have had a stroke despite receiving OACs are a very high-risk group for recurrent stroke; their 1-year risk of recurrent stroke is 6%.10,11 Thus, multiple groups of patients have a need for augmented stroke prevention that cannot be addressed through anticoagulation therapy alone. It would appear to be justified if one can reduce this risk by adding LAA occlusion to background OAC, if it could be done at low risk outside the context of cardiac surgery for another indication.
The results of the analyses presented in this article have some limitations. Detailed information about the use of OAC therapy was gathered only at 6-month visits. Some of the patients designated as on OAC at a visit may have stopped it before the next visit or vice versa. The use of an OAC is a postrandomization variable; therefore, the results may be confounded by indication bias regarding the use of OACs. This bias likely explains the finding that the patients on OAC at fewer visits had more bleeding; discontinuation was the response to adverse events. The statistical power to detect a difference in the effect of LAA occlusion on stroke between patients on OACs or not is limited. Only a much larger study would have had sufficient power to be certain that there is no true difference in the effect of LAA occlusion between patients on and not on OACs. However, the substantial similarity in hazard ratios in the 3 subgroups based on the use of an OAC argues against any major differences in the benefits of LAA occlusion irrespective of OAC use. The landmark analysis from hospital discharge demonstrated a significant interaction; however, the CI for the subgroup of those not on an OAC at discharge was wide, which gives us less confidence in this finding.In conclusion, a key result of LAAOS III is that LAA occlusion provides protection against ischemic stroke and systemic embolism that is independent of OAC therapy. The results from the analyses in this article provide additional information about reasons for nonuse of OAC, type of OAC use during the trial, and INR control.
Sources of Funding
Funding for this trial was received from the following sources: the Canadian Institutes of Health Research, the Canadian Stroke Prevention Intervention Network, Hamilton Health Sciences Research Institute through the Population Health Research Institute, the Heart and Stroke Foundation of Canada, the RFA Program–Research Strategic Initiatives of Hamilton Health Sciences, the Canadian Network and Centre for Trials Internationally (CANNeCTIN), and McMaster University Surgical Associates.
Tables S1 and S2
direct oral anticoagulant
international normalized ratio
left atrial appendage
oral anticoagulant therapy
vitamin K antagonist
Disclosures Drs Connolly, Healey, and Whitlock have received institutional research grant support from Atricure and Boston Scientific. Dr Whitlock has received honoraria from Atricure. The remaining authors reported no conflict.
Ruff CT, Giugliano RP, Braunwald E, Hoffman EB, Deenadayalu N, Ezekowitz MD, Camm AJ, Weitz JI, Lewis BS, Parkhomenko A,. Comparison of the efficacy and safety of new oral anticoagulants with warfarin in patients with atrial fibrillation: a meta-analysis of randomised trials.Lancet. 2014; 383:955–962. doi: 10.1016/S0140-6736(13)62343-0CrossrefMedlineGoogle Scholar
Hart RG, Pearce LA, Aguilar MI. Meta-analysis: antithrombotic therapy to prevent stroke in patients who have nonvalvular atrial fibrillation.Ann Intern Med. 2007; 146:857–867. doi: 10.7326/0003-4819-146-12-200706190-00007CrossrefMedlineGoogle Scholar
Holmes DR, Doshi SK, Kar S, Price MJ, Sanchez JM, Sievert H, Valderrabano M, Reddy VY.Left atrial appendage closure as an alternative to warfarin for stroke prevention in atrial fibrillation: a patient-level meta-analysis.J Am Coll Cardiol. 2015; 65:2614–2623. doi: 10.1016/j.jacc.2015.04.025CrossrefMedlineGoogle Scholar
Whitlock RP, Belley-Cote EP, Paparella D, Healey JS, Brady K, Sharma M, Reents W, Budera P, Baddour AJ, Fila P,; LAAOS III Investigators. Left atrial appendage occlusion during cardiac surgery to prevent stroke.N Engl J Med. 2021; 384:2081–2091. doi: 10.1056/NEJMoa2101897CrossrefMedlineGoogle Scholar
Whitlock R, Healey J, Vincent J, Brady K, Teoh K, Royse A, Shah P, Guo Y, Alings M, Folkeringa RJ,. Rationale and design of the Left Atrial Appendage Occlusion Study (LAAOS) III.Ann Cardiothorac Surg. 2014; 3:45–54. doi: 10.3978/j.issn.2225-319X.2013.12.06CrossrefMedlineGoogle Scholar
Reddy VY, Gibson DN, Kar S, O’Neill W, Doshi SK, Horton RP, Buchbinder M, Gordon NT, Holmes DR. Post-approval U.S. experience with left atrial appendage closure for stroke prevention in atrial fibrillation.J Am Coll Cardiol. 2017; 69:253–261. doi: 10.1016/j.jacc.2016.10.010CrossrefMedlineGoogle Scholar
Connolly SJ, Ezekowitz MD, Yusuf S, Eikelboom J, Oldgren J, Parekh A, Pogue J, Reilly PA, Themeles E, Varrone J,; RE-LY Steering Committee and Investigators. Dabigatran versus warfarin in patients with atrial fibrillation.N Engl J Med. 2009; 361:1139–1151. doi: 10.1056/NEJMoa0905561CrossrefMedlineGoogle Scholar
Granger CB, Alexander JH, McMurray JJ, Lopes RD, Hylek EM, Hanna M, Al-Khalidi HR, Ansell J, Atar D, Avezum A,; ARISTOTLE Committees and Investigators. Apixaban versus warfarin in patients with atrial fibrillation.N Engl J Med. 2011; 365:981–992. doi: 10.1056/NEJMoa1107039CrossrefMedlineGoogle Scholar
Giugliano RP, Ruff CT, Braunwald E, Murphy SA, Wiviott SD, Halperin JL, Waldo AL, Ezekowitz MD, Weitz JI, Špinar J,; ENGAGE AF-TIMI 48 Investigators. Edoxaban versus warfarin in patients with atrial fibrillation.N Engl J Med. 2013; 369:2093–2104. doi: 10.1056/NEJMoa1310907CrossrefMedlineGoogle Scholar
Seiffge DJ, De Marchis GM, Koga M, Paciaroni M, Wilson D, Cappellari M, Macha Md K, Tsivgoulis G, Ambler G, Arihiro S,; RAF, RAF-DOAC, CROMIS-2, SAMURAI, NOACISP, Erlangen, and Verona registry collaborators. Ischemic stroke despite oral anticoagulant therapy in patients with atrial fibrillation.Ann Neurol. 2020; 87:677–687. doi: 10.1002/ana.25700CrossrefMedlineGoogle Scholar
Tanaka K, Koga M, Lee KJ, Kim BJ, Park EL, Lee J, Mizoguchi T, Yoshimura S, Cha J-K, Lee B-C,; CRCS-K Investigators and the SAMURAI Study Investigators. Atrial fibrillation-associated ischemic stroke patients with prior anticoagulation have higher risk for recurrent stroke.Stroke. 2020; 51:1150–1157. doi: 10.1161/STROKEAHA.119.027275LinkGoogle Scholar