Article Text

Download PDFPDF

Original research
Ischaemic stroke on anticoagulation therapy and early recurrence in acute cardioembolic stroke: the IAC study
  1. Shadi Yaghi1,
  2. Nils Henninger2,3,
  3. James A Giles4,
  4. Christopher Leon Guerrero5,
  5. Eva Mistry6,
  6. Ava L Liberman7,
  7. Daniyal Asad8,
  8. Angela Liu4,
  9. Muhammad Nagy2,
  10. Ashutosh Kaushal9,
  11. Idrees Azher6,9,
  12. Brian Mac Grory9,
  13. Hiba Fakhri6,
  14. Kiersten Brown Espaillat6,
  15. Hemanth Pasupuleti10,
  16. Heather Martin10,
  17. Jose Tan10,
  18. Manivannan Veerasamy10,
  19. Charles Esenwa7,
  20. Natalie Cheng7,
  21. Khadean Moncrieffe7,
  22. Iman Moeini-Naghani5,
  23. Mithilesh Siddu5,
  24. Erica Scher11,
  25. Tushar Trivedi11,
  26. Karen L Furie9,
  27. Salah G Keyrouz4,
  28. Amre Nouh8,
  29. Adam de Havenon12,
  30. Muhib Khan10,13,
  31. Eric E Smith14,
  32. M Edip Gurol15,16
  1. 1 Dpeartment of Neurology, Brown University, Providence, Rhode Island, USA
  2. 2 Neurology, University of Massachusetts Medical School, Worcester, Massachusetts, USA
  3. 3 Department of Psychiatry, University of Massachusetts, Worcester, Massachusetts, USA
  4. 4 Neurology, Washington University in Saint Louis, St Louis, Missouri, USA
  5. 5 The George Washington University School of Medicine and Health Sciences, Washington, District of Columbia, USA
  6. 6 Neurology, Vanderbilt University, Nashville, Tennessee, USA
  7. 7 Neurology, Montefiore Hospital and Medical Center, Bronx, New York, USA
  8. 8 Neurology, Hartford Hospital, Hartford, Connecticut, USA
  9. 9 Neurology, Brown University Warren Alpert Medical School, Providence, Rhode Island, USA
  10. 10 Neuroscience Institute, Spectrum Health, Grand Rapids, Michigan, USA
  11. 11 Neurology, NYU Langone Health, New York, New York, USA
  12. 12 Neurology, University of Utah Health Hospitals and Clinics, Salt Lake City, Utah, USA
  13. 13 Neurology, Michigan State University College of Human Medicine, Grand Rapids, Michigan, USA
  14. 14 Hotchkiss Brain Institute, University of Calgary, Calgary, Alberta, Canada
  15. 15 Neurology, Massachusetts General Hospital, Boston, Massachusetts, USA
  16. 16 Neurology, Harvard Medical School, Boston, Massachusetts, USA
  1. Correspondence to Dr Shadi Yaghi, Neurology, Columbia University Medical Center, New York, New York, USA; shadiyaghi{at}


Background and purpose A subset of ischaemic stroke patients with atrial fibrillation (AF) have ischaemic stroke despite anticoagulation. We sought to determine the association between prestroke anticoagulant therapy and recurrent ischaemic events and symptomatic intracranial haemorrhage (sICH).

Methods We included consecutive patients with acute ischaemic stroke and AF from the Initiation of Anticoagulation after Cardioembolic stroke (IAC) study from eight comprehensive stroke centres in the USA. We compared recurrent ischaemic events and delayed sICH risk using adjusted Cox regression analyses between patients who were prescribed anticoagulation (ACp) versus patients who were naïve to anticoagulation therapy prior to the ischaemic stroke (anticoagulation naïve).

Results Among 2084 patients in IAC, 1518 had prior anticoagulation status recorded and were followed for 90 days. In adjusted Cox hazard models, ACp was associated with some evidence of a higher risk higher risk of 90-day recurrent ischaemic events only in the fully adjusted model (adjusted HR 1.50, 95% CI 0.99 to 2.28, p=0.058) but not increased risk of 90-day sICH (adjusted HR 1.08, 95% CI 0.46 to 2.51, p=0.862). In addition, switching anticoagulation class was not associated with reduced risk of recurrent ischaemic events (adjusted HR 0.41, 95% CI 0.12 to 1.33, p=0.136) nor sICH (adjusted HR 1.47, 95% CI 0.29 to 7.50, p=0.641).

Conclusion AF patients with ischaemic stroke despite anticoagulation may have higher recurrent ischaemic event risk compared with anticoagulation-naïve patients. This suggests differing underlying pathomechanisms requiring different stroke prevention measures and identifying these mechanisms may improve secondary prevention strategies.

Data availability statement

Data are available on reasonable request.

Statistics from


Atrial fibrillation (AF) is associated with increased ischaemic stroke risk and studies have shown that oral anticoagulation therapy is associated with reduced risk of ischaemic stroke.1 The risk of ischaemic stroke in patients with AF, however, remains elevated despite anticoagulation therapy. A patient-level data meta-analysis of the SPAF trials showed that the overall risk of ischaemic stroke on warfarin was 2 per 100 patient-years. Importantly, this risk was almost double (nearly 4 per 100 patient-years) in patients with previous stroke or TIA.2 More recent studies comparing direct oral anticoagulants (DOACs) to warfarin3–5 and aspirin6 found a similar trend in that the rate of ischaemic stroke/systemic embolism was up to 1.5% per year in patients treated with DOACs and up to 2.5% per year in patients with prior stroke/TIA.

Several potential explanations for this phenomenon have been suggested, including interruption of anticoagulation therapy, inappropriate anticoagulation dosing, atrial enlargement, presence of additional vascular risk factors as assessed by the CHA2DS2Vasc score, non-paroxysmal (vs paroxysmal) AF, as well as non-cardioembolic (vs cardioembolic) stroke mechanisms.7–9 Notably, switching to a different anticoagulant class10 did not mitigate the risk of stroke recurrence, suggesting that pharmacokinetic and dosing considerations insufficiently explain the persistent risk increase.10 11 Yet, these studies did not adjust for important biomarkers of atrial disease as well as the presence/absence of a competing large artery atherosclerotic mechanism that portends a high risk of recurrence.12

In this study, we sought to assess the short-term risk of stroke recurrence to extend these findings in a large multicentre study by accounting for cardiac biomarkers as well as a competing large artery atherosclerotic stroke mechanism. Our primary hypothesis was that AF patients who were on anticoagulation prior to their ischaemic stroke would be at higher risk of recurrence even in short term when compared with anticoagulant naïve patients, regardless of poststroke preventive management.


Deidentified data may be shared on reasonable request to the corresponding author.

Patient population

The initiation of anticoagulation (IAC) study is a multicentre retrospective study that pooled ischaemic stroke registry data from consecutive patients with acute ischaemic stroke in the setting of AF treated at eight comprehensive stroke centres in the USA within the years 2015 and 2018. Methodological details have been previously described in detail.13–15 For the purpose of this study, we included all patients with information about anticoagulation treatment status at the time of index stroke as well those who survived and were followed during the first 90 days.

Primary predictor

Data on home medication use were collected by the sites. These included treatment with antiplatelets, statins, warfarin and DOACs. The primary predictor was home anticoagulation or anticoagulation prior to stroke (ACp) defined as being on a DOAC or warfarin at the time of the ischaemic stroke.

Primary outcomes

The coprimary outcomes were recurrent ischaemic events and symptomatic intracranial haemorrhage (sICH) as defined in prior studies.13–15 All outcomes were abstracted from medical records by the study site research assistant and confirmed by the site principal investigator.


Demographic factors: Age at the time of admission and sex.

Clinical variables: Stroke risk factors (history of hypertension, history of diabetes, history of hyperlipidaemia, history of prior stroke or TIA, active smoking), CHA2DS2Vasc score and NIHSS score.

Medications prior to admission: Antiplatelet agent and statin therapy.

Neuroimaging and vascular imaging variables: Largest ischaemic stroke lesion volume measured using the a*b*c/2 formula16 as well as the presence of intracranial or extracranial stenosis atherosclerosis with ≥50% luminal narrowing in the territory of the stroke abstracted from the vascular imaging report (ipsilateral atherosclerosis). The choice of brain imaging at baseline (CT vs MRI) was at the discretion of the treating physician.

Echocardiographic variables: Echocardiographic variables were abstracted by a transthoracic echocardiogram performed within 3 months from the ischaemic stroke. For this analysis, we included left ventricular ejection fraction and left atrial enlargement determined by left atrial diameter or volume measruements.17

In-hospital treatments: Intravenous alteplase, mechanical thrombectomy, bridging with heparin or low-molecular-weight heparin (LMWH) and initiation or resumption of anticoagulation.

Analytical plan

Patients were divided into two groups: Treatment with anticoagulation prior to the index stroke (ACp) and anticoagulation naïve (ACn). We compared clinical characteristics, in-hospital treatments and outcomes at 90 days between the two groups. To determine association between ACp and recurrent ischaemic events, we built a priori determined Cox regression models adjusting for the following factors based on prior studies: model 1: adjusting for the CHA2DS2-Vasc score,18 model 2 adjusting for the CHA2DS2-Vasc, presence of a small (<10 mL) vs large (≥10 mL) infarct18 and ipsilateral atherosclerosis,19 and model 3 adjusting for variables in model 2 in addition to anticoagulation initiated. To determine the possible association between ACp and sICH, we built a priori determined Cox regression models adjusting for the following factors, model 1 adjusted for age sex and small infarct size13 18; model 2 adjusted for model 1 and bridging with LMWH/heparin14; model 3 adjusted for model 2 plus anticoagulation initiated. Analyses were done using SPSS V.25 (IBM) and a two-sided p<0.05 was considered statistically significant.


Among 2084 patients included in the IAC study, 1518 subjects met the inclusion criteria. The reasons for exclusion were as follows: 195 patients were lost to follow-up at 90 days, 369 patients had a non-outcome related death within 90 days, and 2 patients had missing information on anticoagulation treatment prior to stroke. The baseline demographics, median CHA2DS2Vasc score and anticoagulation status prior to index stroke were not significantly different between those with vs without lost to follow-up within 90-days.

Among included patients, the mean age was 76.2±11.7 years and 49.9% (757/1518) were men; 546 patients (36.0%) were on anticoagulation at home (ACp) (278 (50.9%) were on a DOAC) and 972 patients (64.0%) were not on home anticoagulation (ACn). After the ischaemic stroke, 92 patients were switched to a different anticoagulant class: DOAC to warfarin (n=81 patients) and warfarin to DOAC (n=11 patients). In addition, 16.4% (245/1492) had evidence of atherosclerosis with 50% or more luminal narrowing in a vessel supplying the territory of the infarct.

Univariate analyses

Table 1 summarises the patient characteristics stratified according to prestroke anticoagulant status. In brief, when compared with patients in the ACn group, patients in the ACp group had a lower admission NIHSS score (median 6 vs 9, p=0.001), were more likely to have diabetes (39.7% vs 30.2%, p<0.001), hyperlipidaemia (59.9% vs 54.0%, p=0.027), prior stroke or TIA (39.6% vs 26.8%, p<0.001), congestive heart failure (29.7% vs 22.0%, p=0.001), coronary artery disease (38.1% vs 27.5%, p, 0.001), severe left atrial enlargement (43.9% vs 34.9%, p=0.002) and were more frequently on a statin at home (64.7% vs 49.5%, p<0.001), placed on anticoagulation after the ischaemic stroke (95.6% vs 80.2%, p<0.001) and less likely on antiplatelet medication at home (32.1% vs 58.9%, p<0.001).

Table 1

Baseline characteristics and outcomes of patients stratified by anticoagulation treatment status prior to the stroke

Association between anticoagulation prior to the index stroke and 90-day outcomes

Recurrent ischaemic events occurred in 104 patients (6.9%): recurrent stroke in 91 patients, recurrent TIA in 6 patients and recurrent symptomatic systemic embolism in 9 patients. Furthermore, 23 subjects had a sICH, with 91.3% (21/23) occurring after IAC. Predictors of ischaemic recurrence and haemorrhagic events in the IAC study were previously reported.13

Association between anticoagulation prior and recurrent ischaemic events

In minimally adjusted prespecified models (models 1 and 2), associations between ACp and recurrent ischaemic events were not statistically significant (table 2). In the fully adjusted model (model 3) adjusting for prespecified potential confounders including IAC and CHA2DS2-Vasc score, Cox regression analyses indicated a non-significantly higher risk for recurrent ischaemic events among subjects with ACp (adjusted HR 1.50, 95% CI 0.99 to 2.28, p=0.058) (table 2) (figure 1).

Figure 1

Upper figure shows Kaplan-Meier survival analysis showing the risk of recurrent ischaemic events within 90 days stratified by home anticoagulation status: anticoagulation prior (ACp) versus anticoagulation naïve (ACn). Lower figure shows Kaplan-Meier survival analysis showing the risk of symptomatic intracranial haemorrhage within 90 days stratified by home anticoagulation status: ACp versus ACn. sICH, symptomatic intracranial haemorrhage.

Table 2

Cox regression models showing associations between anticoagulation status before the stroke and recurrent ischaemic events and symptomatic intracranial haemorrhage

Association between anticoagulation prior and sICH

There was no association between ACp and sICH on unadjusted (HR 1.15, 95% CI 0.50 to 2.66, p=0.744) and adjusted Cox regression (adjusted HR 1.08, 95% CI 0.46 to 2.51, p=0.862) (table 2 and (figure 1).

Sensitivity analysis

We performed additional analyses restricted to patients with ACp, to determine whether switching anticoagulation had an impact on early recurrence and major bleeding in patients with ACp. In Cox regression analyses, there was no association between switching anticoagulant class and sICH unadjusted (HR 1.40, 95% CI 0.29 to 6.75, p=0.674) and after adjusting for potential confounders (HR 1.47, 95% CI 0.29 to 7.50, p=0.641). In addition, switching anticoagulation class was associated with non-significantly lower risk of early recurrent ischaemic events in unadjusted models (HR 0.35, 95% CI 0.11 to 1.13, p=0.079) but this association was not significant after adjusting for potential confounders (adjusted HR 0.41, 95% CI 0.12 to 1.33, p=0.136).

In addition, we performed a sensitivity analysis including patients with non-outcome related death and when the exact time of death was not recorded, the time of death was imputed as day 90. In these analyses and if the fully adjusted models ACp was associated with a non-significantly lower risk of recurrent ischaemic events persisted (adjusted HR 1.46, 95% CI 0.96 to 2.21, p=0.076) with a similar risk of sICH (adjusted HR 1.08, 95% CI 0.47 to 2.52, p=0.857).

Furthermore, to account for patient clustering by study site, we fit mixed-effects logistic regression models to our outcomes and our findings were not meaningfully different; in full adjusted models, ACp was associated with a non-significantly higher risk of recurrent ischaemic events (adjusted OR 1.42, 95% CI 0.93 to 2.19, p=0.106) but no increased risk of sICH (adjusted OR 1.05, 95% CI 0.49 to 2.25, p=0.809).


The results gained from our real-world study is in line with prior multicentre studies.10 11 We found that AF patients with ischaemic stroke on anticoagulation therapy had a heightened risk of recurrent ischaemic stroke even after adjustment for pertinent confounders. Switching anticoagulation class in AF patients who had an ischaemic stroke while using anticoagulation was not associated with reduced risk of recurrent ischaemic events, again a finding reported in a previous large multicentre study.11 Contrary to prior studies, our findings did not achieve statistical significance likely due to lower events rates, shorter follow-up duration, as well as a smaller sample size than that in prior studies.

Possible reasons for ischaemic stroke recurrence despite anticoagulation therapy might include suboptimal medication adherence, anticoagulation interruption and inappropriate dosing. Concomitant stroke mechanisms such as small vessel disease, large artery atherosclerosis and cancer-related hypercoagulability in patients with AF may augment recurrent stroke risk despite anticoagulation therapy. For many of these non-AF-related stroke mechanisms, anticoagulation is not superior to aspirin. For instance, the Warfarin Aspirin Recurrent Stroke Study showed no benefit of warfarin over aspirin in patients with non-cardioembolic stroke.20 This was also shown in the Warfarin Aspirin Symptomatic intracranial Disease trial where warfarin was not superior to aspirin in secondary stroke prevention in patients with intracranial atherosclerosis.21 In addition, symptomatic large artery atherosclerosis and cancer related hypercoagulability are associated with increased risk of early recurrence regardless of antithrombotic regimen.12 22 Furthermore, patients who have more vascular risk factors, more severe atrial disease and higher AF burden, remain at an increased risk of stroke despite anticoagulation therapy.7 Consistent with this notion, we found that in our study these patients with stroke recurrence despite oral anticoagulant therapy had a higher CHA2DS2Vasc scores and more frequently had a history of prior stroke, underlying coronary artery disease, and diabetes.

It is noteworthy that 19.6% of patients who were not on anticoagulation prior to the index event were not started on anticoagulation in the 90 days following the index event as compared with 4.4% of those who were on anticoagulation prior to the index event. There are several reasons that could account for this difference including more severe strokes in the ACn group or those who were not on anticoagulation prior to the index event having a contraindication to anticoagulation. In addition, it is also important to note that the percentage of patients on DOACs was higher than that reported in prior studies10 and may at least partially explain the relatively low bleeding complication events, in line with a prior study.23

Our findings have several therapeutic implications. It suggests that anticoagulation therapy alone may not be sufficient in reducing stroke risk in patients with AF thus highlighting the importance of improving stroke prevention beyond anticoagulation therapy.24 First, risk factor and lifestyle modification are the corner stone for prevention. Indeed, in patients with AF, risk factors such as hypertension and diabetes have been associated with increased stroke risk,25 thus controlling modifiable risk factors is an essential step in reducing the residual risk of ischaemic stroke in patients with AF started on anticoagulation. Second, AF frequently coexists with cerebrovascular atherosclerotic disease19 26 27 and small vessel disease28 and thus in patients with ischaemic stroke despite anticoagulation, it is essential to assess for a competing non-cardioembolic mechanism and ensure that stroke prevention strategies are targeted. For instance, finding severe carotid stenosis in a patients whose infarct is in its territory may lead to carotid revascularisation, which has been shown to reduce the risk of recurrent stroke.29 Third, left atrial appendage occlusion has recently emerged as a stroke prevention strategy in patients with AF and an alternative to long-term anticoagulation.30 The Novel Anticoagulation Agents in AF (PRAGUE-17) trial compared left atrial appendage occlusion to DOACs and showed that left atrial appendage occlusion is non-inferior to DOACs in reducing the risk of stroke and systemic embolism in patients with AF.31 The risk of ischaemic stroke however was 2% in each group suggesting that left atrial appendage occlusion may not superior to DOAC therapy in reducing the risk of ischaemic stroke31 and may not be effective by itself in reducing the residual risk of recurrence in patients with breakthrough strokes despite anticoagulation. Whether combining left atrial appendage occlusion with anticoagulation would reduce risk more than either alone is uncertain and has not yet been tested in trials.

Our study has several important limitations. First, our study is a multicentre retrospective study of comprehensive stroke centres which introduces bias and limits generalisability. Second, we lack data on anticoagulation adherence and risk factor control prior to presentation and therefore it is possible that some of the strokes occurred in the setting of poor risk factor control or non-compliance with anticoagulation. Third, nearly 9% of patients were lost to follow-up within 90 days. Fourth, while we collected data on infarct size and competing large artery atherosclerotic mechanisms, we lack data on other mechanisms such as dissection, small vessel disease and hypercoagulability from cancer. Fifth, we lack data on the imaging characteristics and stroke subtyping of recurrent events which in some cases could have been related to a competing non-cardioembolic stroke mechanism. Sixth, our study is an observational study and thus subject to treatment by indication bias. Seventh, we lack information on whether patients who were in the ACn whether they had new AF or known AF but not on anticoagulation although we suspect that most patients had known AF and were not prescribed anticoagulation (ACp) as as evidenced by (1) much higher antiplatelet use despite no difference in coronary artery disease and (2) 20% of these patient were still not anticoagulated at 90 days. Finally, the follow-up period in our study is only 90 days and thus it is unclear if the risk of recurrence remains elevated beyond the first 90 days as other studies suggested.

Notable strengths of our study include being a large multicentre study, collecting a wide range of variables and outcomes, and providing real-world data on the risk of early recurrence in patients with AF.


AF patients with ischaemic stroke on anticoagulation may be at higher risk for recurrent stroke compared with anticoagulation-naïve patients, even after adjusting for estimated risk and switching the anticoagulant type does not decrease such recurrent stroke risk. This raises the possibility that pathological mechanisms either not amenable to anticoagulation or not covered by anticoagulation alone are at play in these patients. Future research should seek to identify these mechanisms and test additional prevention strategies.

Data availability statement

Data are available on reasonable request.

Ethics statements

Patient consent for publication

Ethics approval

This study was approved by the institution review board of Rhode Island Hospital (IRB number 1347342-1) as well as individual study sites.



  • Twitter @neurogiles

  • Contributors SY, NH, AdH: Study concept design, drafting manuscript, and statistical analysis. JAG, CRLG, EM, ALL, DA, AL, MN, AK, IA, BMG, HF, KBE, HP, HM, JT, MV, CE, NC, KM, IM-N, MS, KLF, SGK, AN and MK: data collection and manuscript revisions. ESc and TT: data management and manuscript revision. ESm and MEG: study concept and design and manuscript revision.

  • Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.

  • Competing interests SY has a non-funded research collaboration with Medtronic. MEG reports grants from AVID (Eli Lilly), grants from Boston Scientific, and grants from Pfizer outside the submitted work. NH reports grants from NINDS during the conduct of the study; grants from NICHD of the NIH, grants from NINDS of the NIH, grants from CDMRP of the DoD, and personal fees from Astrocyte Pharmaceuticals outside the submitted work. EM reports grants from NIH/NINDS outside the submitted work. AdH reports research support from AMAG and Regeneron.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Linked Articles