Abstract
Background:
Left ventricular (LV) dysfunction is a potential cardioembolic source of ischemic stroke, but its role in recurrent stroke risk and treatment response remains unclear. We explore whether LV injury associates with risk of recurrent stroke and modifies the association between anticoagulation and stroke recurrence using real-world data.
Methods:
We performed a multicenter, retrospective study of Cardiac Abnormalities in Stroke Prevention and Recurrence (CASPR) cohort across 27 U.S. sites. Patients with LV ejection fraction (LVEF) ≥20% were included. LV injury, defined as LVEF 20-40% and/or wall motion abnormality (WMA), was the primary exposure and treatment effect modifier. The treatment of interest was anticoagulant versus antiplatelet therapy. The composite outcome included recurrent stroke, major bleeding, or death. Outcomes were evaluated using unadjusted and inverse probability weighting (IPW)-adjusted Cox proportional hazards models, with treatment effect modification tested by LV injury status.
Results:
Among 2,685 patients enrolled, 2,328 with complete data were analyzed (median age, 65 years; 49.8% female; median follow-up 1.6 years). LV injury was present in 310 patients (13.3%). Overall, 535 events occurred: 258 recurrent ischemic strokes, 28 hemorrhagic strokes, 67 major hemorrhages, and 256 deaths. LV injury was associated with higher unadjusted risk of the primary outcome (HR 1.51; 95% CI, 1.21–1.87), though nonsignificant after IPW adjustment (adjusted HR 1.29; 95% CI, 0.97-1.70). In the LV injury subgroup, anticoagulation versus antiplatelet therapy was associated with a lower risk of the primary outcome (adjusted HR 0.24; 95% CI, 0.10–0.59), relative to the non-LV injury subgroup (adjusted HR 1.28; 95% CI, 0.83–1.95; p[LV-interaction]=0.001). Similar interactions were seen for EF 20-40% (vs. >40%; adjusted HR 0.19; 95% CI, 0.04–0.86; p[LV-interaction]=0.001) and WMA (vs. no WMA; adjusted HR 0.33; 95% CI, 0.15–0.73)).
Conclusion:
After cryptogenic stroke, anticoagulation in those with LV injury was associated with lower rates of recurrent stroke, major bleeding, and death. These findings warrant confirmation in a dedicated randomized controlled trial.
Trial Registration:
Keywords: ischemic stroke, left ventricular dysfunction, anticoagulants
Graphical Abstract
INTRODUCTION
Left ventricular (LV) injury, defined by either a left ventricular ejection fraction (LVEF) <40% or an LV wall motion abnormality (WMA), has been associated with a heightened risk of ischemic stroke.1-3 Severely reduced LVEF, though relatively uncommon, is a well-established cardioembolic etiology of ischemic stroke with an increased risk ranging from 1.9- to 8.8-fold.4,5 Emerging evidence suggests that moderately reduced LVEF and LV WMA also increase the risk of stroke.1,2 However, the association between these LV injury subtypes and stroke pathogenesis is less well-defined, which may lead clinicians to classify such cases as cryptogenic.
Additionally, there is no definitive anti-thrombotic treatment recommendation per secondary stroke prevention guidelines for patients with ischemic stroke, sinus rhythm, and LV injury in the absence of an intracardiac thrombus.6 As a result, there is currently significant variability in clinical practice of anticoagulation prescription after an ischemic stroke in this population.7 While multiple secondary, post-hoc analyses of primary prevention studies have demonstrated the benefits of anticoagulation in patients with reduced LVEF, its net clinical benefit has been perceived as limited.8-10 This is due to the use of primary composite outcomes that dilute the treatment effect on stroke, the increased risk of hemorrhage associated with warfarin (the predominant anticoagulant used), and relatively low stroke rates observed in primary prevention studies.11 A secondary analysis of the NAVIGATE ESUS trial demonstrated treatment effect in patients with embolic stroke of undetermined stroke (ESUS) and LV injury, however, the primary outcome evaluated did not encompass safety measures such as intracranial and major hemorrhage risk with anticoagulation.12 While a secondary analysis of the ARCADIA trial indicated an anticoagulation treatment effect in patients with ESUS, markers of left atrial cardiopathy, and LV injury, the LV injury sample size was only 165 patients.13
Given these gaps in evidence, this study aimed to 1) determine if there was an increased risk of the composite outcome of recurrent stroke (ischemic or hemorrhagic), major hemorrhage, or all-cause death among patients with cryptogenic stroke and LV injury, and 2) evaluate whether there was heterogeneity in treatment association by LV injury presence between anticoagulation and the composite outcome in a real-world registry.
METHODS
Data Availability and Reporting Guidelines:
Data from this study will be made available to a qualified investigator upon reasonable request to the principal investigator of the CASPR Registry (co-author J.S.) following completion of data transfer agreements. We followed the Strengthening the Reporting of Observational Studies in Epidemiology reporting guidelines for this study.
Study Design and Participants:
The Cardiac Abnormalities in Stroke Prevention and Recurrence or CASPR study14 was a multi-center retrospective cohort of consecutive patients ages 18 and older with an ischemic stroke deemed cryptogenic by the treating clinician after a comprehensive workup for the underlying stroke etiology. Twenty-seven sites in the U.S. were recruited by existing networks and academic societies that have conducted similar research studies. Participating sites complied with their local Institutional Review Board regarding patient consent for CASPR and were encouraged to apply for a waiver of informed consent, as this study involves minimal risk. Patients were eligible for inclusion if the index stroke event occurred between January 1, 2016, to June 30, 2022.
Patients at each site hospitalized with an acute ischemic stroke were identified and screened using the institutional Get-with-the-guidelines Stroke Registry if the site was a Primary or Comprehensive Stroke Center; this registry is a quality reporting system in which participating hospitals input clinical data of all patients hospitalized with a diagnosis of stroke.15 A collaborating site that was not certified as either retrospectively queried its local registry. Finally, sites which were not thus certified and also did not maintain a local registry queried their electronic medical records for patients ascribed an acute ischemic stroke diagnosis by either an ICD-9 or 10 code of 430-438 or I60-69.16,17 These patients then further underwent chart review for eligibility assessment by local investigators. All patients included underwent either computed tomography (CT) or magnetic resonance imaging (MRI) of the brain demonstrating an acute cerebral infarction. Only patients with stroke onset or last known well within two weeks of hospitalization were included unless the exact time of onset was unknown. Upon screening, the mechanism of infarction was then reviewed by more detailed manual review of the electronic medical record, laboratory results, demographic, clinical and diagnostic testing.
The inclusion criteria in CASPR was anchored around a cryptogenic stroke diagnosis, defined by an unrevealing workup including: 1) transthoracic echocardiogram, 2) electrocardiogram and a minimum of 24 hours of cardiac telemetry, 3) cervical and intracranial vessel imaging, 4) no known and established source of cerebral embolism after completion of the aforementioned testing.18 Exclusion criteria eliminated patients with known stroke etiologies which included: 1) LVEF < 20%, 2) prior or new diagnosis of atrial fibrillation, 3) cervical or intracranial atherosclerosis in a vessel supplying the infarcted brain region with ≥ 50% luminal stenosis per North American Symptomatic Carotid Endarterectomy Trial (NASCET) criteria,19 4) cervical or intracranial arterial dissection, 5) acute myocardial infarction or cardiac arrest at the time of stroke, 6) intracardiac thrombus, and 7) small vessel disease. Patients without follow-up information available at 90 days if not expired within this timeframe and those enrolled in a randomized clinical trial in which investigators are blinded to anti-thrombotic strategy were also excluded. Finally, patients with incomplete exposure and outcome data were excluded.
Study Outcomes:
The primary outcome of interest was a composite endpoint encompassing recurrent stroke, major hemorrhage, and all-cause death. Recurrent stroke due to either ischemia or hemorrhage was defined by a clinical neurologic event associated with a radiographic correlate. Major hemorrhage was determined on the basis of the International Society of Thrombosis and Hemostasis as a new or worsening intracranial or major systemic extracranial hemorrhage resulting in a drop of at least 2 grams per deciliter in hemoglobin level or necessitating a blood transfusion.20
Study Exposures:
The primary exposure of interest was the presence of LV injury noted on cardiac imaging that included transthoracic echocardiogram, transesophageal echocardiogram, cardiac computed tomography, or cardiac magnetic resonance imaging. LV injury was defined by either LVEF ≤ 40% or an LV wall motion abnormality characterized by an LV wall that was hypokinetic, akinetic, aneurysmal, or dyskinetic. The CASPR registry sought to include patients with a lower LVEF than defined by the formal ESUS construct due to the variable antithrombotic practices for reduced LVEF,7 the inter-rater variability in ventricular function assessments, potential improvement following goal-directed medical therapy,21 and variability in the method and modality for quantifying ejection fraction, which may differ by up to 20% between methods.22,23 Per TOAST classification,24 these LV subtypes are medium- or high-risk LV sources of cardioembolism, for which the guidelines are currently equivocal regarding optimal antithrombotic therapy for stroke prevention.6 LVEF was extracted by site investigators from cardiac imaging reports and hardcoded into the CASPR database. Information about the presence, degree, and location of LV WMA were not recorded in dedicated CASPR database fields. Thus, LV WMA presence and degree (hypokinesis, akinesis, aneurysmal, and dyskinesis) were extracted from echocardiography reports (n=2296, 98.6% of cohort) using the findw string search function in SAS 9.4 (Cary, NC) with the search terms and expressions listed in Table S1, followed by manual adjudication by co-author, R.S., while blinded to all other patient information. Apical, mid, and basal location of ventricular WMAs, if present, were then also manually extracted from the echocardiography reports of patients verified to have a WMA. Rates of new atrial fibrillation (AF) that developed after the index cryptogenic stroke captured during longitudinal follow-up were recorded.
The secondary exposure of interest was anti-thrombotic medication treatment categorized as two mutually exclusive, binary categories of either 1) antiplatelet(s) without anticoagulation or 2) anticoagulation. These were the anti-thrombotic treatments prescribed upon discharge post-stroke hospitalization.
Covariates:
We recorded baseline variables that may be associated with LV injury, anti-thrombotic prescription, and the primary outcome. These included baseline covariates of age at the time of index stroke, sex, race (White, Black, Asian, Other), pre-stroke functional status captured by the modified Rankin scale, health insurance status, and anticoagulation use prior to the index stroke. Comorbidities prevalent prior to stroke hospitalization discharge that were noted included hypertension, dyslipidemia, diabetes, tobacco use, coronary artery disease, cancer, and prior stroke. We accounted for baseline National Institutes of Health Stroke Severity score and cardiac imaging findings of severe left atrial enlargement and the presence of a patent foramen ovale were noted. Finally, we accounted for neuroimaging characteristics noted at the time of the index stroke including multifocality of lesions noted on the diffusion weighted imaging (DWI) sequence, cortical location of DWI positive lesions, and the presence of lesions on the T2 fluid attenuated inversion recovery or FLAIR sequence in a cortical location.
Statistical Analysis:
The characteristics of patients at baseline stratified by LV injury status were described using medians and interquartile ranges for continuous variables and proportions for categorical variables, and Mann-Whitney 2-sample tests or Chi-squared tests, as appropriate. Frequency of LV injury subtypes were calculated. Factors associated with anticoagulation versus antiplatelet prescription among patients with and without LV injury were also compared.
Annualized rates of the composite and individual outcomes of interest were computed by LV injury status. Cox proportional hazard models were built without and then with inverse probability weighting (IPWLV injury) adjustment. A fixed effect for site was included in all models. IPWLV injury was constructed using for all baseline covariates. We opted not to implement a variable selection criterion when developing IPWLV injury to avoid missing potential confounders associated with the outcomes of interest despite weak signals in descriptive analyses.25 IPWLV injury was utilized to mitigate the risk of confounding by the factors associated with both the exposure of LV injury and outcome of interest. An a priori defined threshold of standardized mean differences after weighting of ≤0.1 would indicate that the covariates are well-balanced between the treatment groups after IPWLV injury.26 Patients were censored at the time of the primary outcome, death, or loss to follow-up. The proportional hazards assumption was tested by visualization of survival curves and Schoenfeld residuals and confirmed using log-log plots.
Antithrombotic treatment effect and its modification by LV injury status was investigated under an intention-to-treat principle, according to initial antithrombotic used at the time of the index stroke event. IPW for anticoagulation versus antiplatelet use or IPWtreatment was constructed as a function of all baseline covariates. IPWtreatment was utilized to mitigate the bias by confounding on the relationship between antithrombotic treatment indication and the outcomes of interest. Heterogeneous treatment effects by LV injury presence were formally investigated via subgroup and interaction analyses in Cox proportional hazard models adjusted for IPWtreatment.
Further analyses were conducted across subgroups defined by LVEF (20-40% versus > 40%) and WMA degree (hypokinesis versus akinesis, dyskinesis, or aneurysmal) and WMA location (basal, mid, and apical regions). The correlation between inter-regional wall motion changes was assessed using Φ. Natural language processing was performed using SAS 9.4 (Cary, NC). All statistical analyses were performed with R version 4.3.2 (2023-10-31) using the survival, WeightIt, gtsummary, and survminer packages. No missing data were imputed. All p-values < 0.05 were considered statistically significant.
RESULTS
Sample Description
Of the 2,685 patients included in CASPR, a total of 2,328 (86.7%) patients met inclusion criteria of non-missing exposure and outcome values (Figure 1). A total of 310 patients (13.3%) had LV injury, while the remaining 2,018 (86.7%) did not (Table 1). The median time of follow-up was 480 days in the analytic cohort [IQR 126-1,040] (median time in the no LV injury group: 480 days [IQR 124-1,039]; median time in the LV injury group: 488 days [IQR 131-1,056]; p > 0.9). A comparison of baseline characteristics of patients with and without missingness of exposure and outcome variables is presented in Table S2. Excluded patients were likely to be older, identify as Other/unreported race, have a prior ischemic stroke, coronary artery disease, cancer, and higher baseline NIHSS.
Figure 1.
Flow diagram of patient inclusion.
Table 1.
Baseline demographic and clinical characteristics stratified by LV injury status.
| Characteristic | Overall (n=2,328) |
No LV injury (n=2,018) |
LV injury (n=310) |
p-value |
|---|---|---|---|---|
| Demographics | ||||
| Age, median y [IQR] | 65 [54, 75] | 65 [54, 74] | 65 [56, 75] | 0.2 |
| Female sex, no. (%) | 1,157 (50%) | 1,028 (51%) | 129 (42%) | 0.002 |
| Race, no. (%) | 0.006 | |||
| White | 1,351 (58%) | 1,186 (59%) | 165 (53%) | |
| Black | 660 (28%) | 547 (27%) | 113 (36%) | |
| Asian | 69 (3.0%) | 62 (3.1%) | 7 (2.3%) | |
| Other | 248 (11%) | 223 (11%) | 25 (8.1%) | |
| Hispanic ethnicity | 125 (5.4%) | 111 (5.5%) | 14 (4.5%) | 0.5 |
| Insurance status | ||||
| 1 | 194 (8.3%) | 166 (8.2%) | 28 (9.0%) | 0.6 |
| 2 | 272 (12%) | 231 (11%) | 41 (13%) | 0.4 |
| 3 | 997 (43%) | 865 (43%) | 132 (43%) | >0.9 |
| 4 | 744 (32%) | 658 (33%) | 86 (28%) | 0.087 |
| 5 | 164 (7.0%) | 150 (7.4%) | 14 (4.5%) | 0.062 |
| Baseline Functional Status and Medication Use | ||||
| Pre-stroke mRS, median (IQR) [range] | 0.00 (0.00, 1.00) | 0.00 (0.00, 1.00) | 0.00 (0.00, 1.00) | 0.007 |
| AC use immediately prior to stroke, no. (%) | 11 (0.5%) | 8 (0.4%) | 3 (1.0%) | 0.2 |
| Comorbidities | ||||
| Hypertension, no. (%) | 1,620 (70%) | 1,372 (68%) | 248 (80%) | <0.001 |
| Dyslipidemia, no. (%) | 1,091 (47%) | 914 (45%) | 177 (57%) | <0.001 |
| Tobacco use, no. (%) | 887 (38%) | 753 (37%) | 134 (43%) | 0.046 |
| Diabetes, no. (%) | 628 (27%) | 507 (25%) | 121 (39%) | <0.001 |
| Coronary artery disease, no. (%) | 303 (13%) | 191 (9.5%) | 112 (36%) | <0.001 |
| Prior stroke, no. (%) | 366 (16%) | 309 (15%) | 57 (18%) | 0.2 |
| Prior cancer, no. (%) | 255 (11%) | 231 (11%) | 24 (7.7%) | 0.052 |
| Cardiac Findings | ||||
| Severe left atrial enlargement, no. (%) | 261 (11%) | 205 (10%) | 56 (18%) | <0.001 |
| Patent foramen ovale, no. (%) | 389 (17%) | 362 (18%) | 27 (8.7%) | <0.001 |
| LVEF, median (IQR) | 60 (55, 65) | 60 (57, 65) | 40 (30, 50) | <0.001 |
| Wall motion abnormality present | 243 (11%) | 21 (1.1%) | 222 (73%) | <0.001 |
| MRI Brain Findings | ||||
| Multifocal DWI lesions | 452 (22%) | 394 (21%) | 58 (22%) | 0.8 |
| Cortical DWI lesions | 1,697 (81%) | 1,477 (80%) | 220 (84%) | 0.2 |
| Cortical T2 lesion | 541 (28%) | 457 (27%) | 84 (34%) | 0.021 |
| Stroke Severity and Treatment | ||||
| Treatment with antiplatelet(s)-no anticoagulation, no. (%) | 2,098 (90%) | 1,831 (91%) | 267 (86%) | 0.011 |
| Baseline NIHSS, median (IQR) | 4 (2, 11) | 4 (2, 11) | 6 (3, 15) | <0.001 |
| Follow-up | ||||
| Time to follow-up or outcome event, median d (IQR) | 480 (126, 1,040) | 480 (124, 1,039) | 488 (131, 1,056) | >0.9 |
Abbreviations: AC, anticoagulation; DWI, diffusion-weighted imaging; IQR, interquartile range; LV, left ventricular; LVEF, left ventricular ejection fraction; mRS, modified Rankin Scale; NIHSS, National Institutes of Health Stroke Scale; T2, T2-weighted (MRI image)
Rates and time detection of post-stroke AF are summarized in Table S3. New-onset AF after stroke occurred in 214 patients (9.2%) and there was similar incidence noted between those with and without LV injury (9.0% vs 9.2%, p > 0.9). Median time to AF detection was 142 days (IQR 31-497) overall, 146 days (IQR 30-432) in patients without LV injury, and 114 days (IQR 40-646) among those with LV injury.
Factors associated with LV injury
Patients with and without LV injury were similar in age, but those with LV injury were significantly less likely to be female (42% versus 51%). Patients with LV injury were significantly less likely to be White (53% versus 59%), and more likely to be Black (36% versus 27%). Patients with LV injury were significantly more likely to have hypertension (80% versus 68%), dyslipidemia (57% versus 45%), diabetes (39% versus 25%), and coronary artery disease (36% versus 10%) (p<0.001 for all) In terms of imaging there was no difference in MRI DWI topology by LV injury status, however, patients with LV injury were more likely to have a cortical T2 lesion (34% versus 27%, p=0.021). Among patients with LV injury, median LVEF was 40% [IQR 30-50], while those without LV injury had a median LVEF 60% [IQR 57-65]. Patients with LV injury were significantly more likely to have a severely enlarged left atrium (18% versus 10%) and less likely to have a PFO (8.7% versus 18%) (p<0.001 for both). Clinically, patients with LV injury had a significantly higher NIHSS at baseline (median 6 [3-15] versus 4 [2-11], p<0.001). Patients with LV injury were less likely to be treated with antiplatelet therapy alone within 7 days of the index stroke (86% versus 91%, p=0.011).
LV Injury Subtypes
There were various subtypes of LV injury observed in this cohort (Table 2). There were 48 patients (1.9%) with an LVEF 15-29% despite CASPR exclusion criteria LVEF < 20%, 71 (2.8%) with an LVEF 30–40%, and 148 (5.8%) with an LVEF 41–50%. There were 243 patients with an LV WMA present. There were 187 patients (7.4%) with a WMA involving the basal wall of the LV, 192 patients (7.6%) with mid-wall WMA, and 183 patients with apical WMA (7.2%). The degrees of WMA present included hypokinesis (n=217; 9.3%), akinesis (n=57; 2.5%), dyskinesis (n=7; 0.3%), and aneurysmal (n=12; 0.5%). There was a strong correlation between inter-regional abnormalities (Φapical/basal = 0.68, Φapical/mid = 0.76, Φbasal/mid = 0.86). Among the patients with LV injury, overlap between reduced LVEF and LV wall motion abnormality (WMA) was common (Table S4), with the highest overlap noted between WMA presence and LVEF 40-50% (37%).
Table 2.
LV injury subtype frequencies in the CASPR cohort.
| LVEF (%) | WMA | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| Hypokinesis | Akinesis | Dyskinesis | Aneurysmal | Apical WMA |
Mid WMA |
Basal WMA |
|||||
| Total LV injury patients |
15-30 | 30-40 | 40-50 | > 50 | |||||||
| 310 | 48 (1.9%) | 71 (2.8%) | 148 (5.8%) | 2,282 (90%) | 217 (87%) | 57 (23%) | 7 (2.8%) | 12 (4.8%) | 183 (7.2%) | 192 (7.6%) | 187 (7.4%) |
Abbreviations: LV, left ventricular; WMA, wall motion abnormality
Among patients without LV injury, there were 187 patients prescribed an anticoagulant (8.0%). Factors significantly associated with AC use upon discharge, as presented in Table 3, included AC use immediately prior to the index stroke (19% versus 1%), prior stroke (26% versus 14%), not having hypertension, multifocal DWI lesions (31% versus 20%), and higher presenting NIHSS at the time of the index stroke (median 6 versus 4).
Table 3.
Baseline characteristics stratified by LV injury status and anticoagulant (AC) versus antiplatelet (AP) treatment.
| No LV Injury | LV Injury | |||||
|---|---|---|---|---|---|---|
| AC (n = 187) | AP (n = 1,831) | p-value | AC (n = 43) | AP (n = 267) | p-value | |
| Demographics | ||||||
| Age, median y (IQR) | 63 (51, 73) | 65 (54, 75) | 0.071 | 62 (54, 71) | 66 (56, 76) | 0.3 |
| Female sex, no. (%) | 108 (58%) | 920 (50%) | 0.05 | 16 (37%) | 113 (42%) | 0.5 |
| Race, no. (%) | 0.9 | 0.4 | ||||
| White | 106 (57%) | 1,080 (59%) | 21 (49%) | 144 (54%) | ||
| Black | 51 (27%) | 496 (27%) | 15 (35%) | 98 (37%) | ||
| Asian | 6 (3.2%) | 56 (3.1%) | 2 (4.7%) | 5 (1.9%) | ||
| Other | 24 (13%) | 199 (11%) | 5 (12%) | 20 (7.5%) | ||
| Hispanic ethnicity | 12 (6.4%) | 99 (5.4%) | 0.6 | 3 (7.0%) | 11 (4.1%) | 0.4 |
| Insurance status | ||||||
| 1 | 13 (7.0%) | 153 (8.4%) | 0.5 | 4 (9.3%) | 24 (9.0%) | >0.9 |
| 2 | 27 (14%) | 204 (11%) | 0.2 | 7 (16%) | 34 (13%) | 0.5 |
| 3 | 74 (40%) | 791 (43%) | 0.3 | 14 (33%) | 118 (44%) | 0.2 |
| 4 | 69 (37%) | 589 (32%) | 0.2 | 13 (30%) | 73 (27%) | 0.7 |
| 5 | 14 (7.5%) | 136 (7.4%) | >0.9 | 3 (7.0%) | 11 (4.1%) | 0.4 |
| Baseline Functional Status and Medication Use | ||||||
| Pre-stroke mRS | 0.00 (0.00, 1.00) | 0.00 (0.00, 1.00) | 0.003 | 0.00 (0.00, 0.50) | 0.00 (0.00, 1.00) | 0.12 |
| AC use immediately prior to stroke | 35 (19%) | 19 (1.0%) | <0.001 | 6 (14%) | 0 (0%) | <0.001 |
| Comorbidities | ||||||
| Hypertension, no. (%) | 115 (61%) | 1,257 (69%) | 0.046 | 35 (81%) | 213 (80%) | 0.8 |
| Dyslipidemia, no. (%) | 78 (42%) | 836 (46%) | 0.3 | 28 (65%) | 149 (56%) | 0.3 |
| Tobacco use, no. (%) | 61 (33%) | 692 (38%) | 0.2 | 21 (49%) | 113 (42%) | 0.4 |
| Diabetes, no. (%) | 37 (20%) | 470 (26%) | 0.077 | 18 (42%) | 103 (39%) | 0.7 |
| Coronary artery disease, no. | 14 (7.5%) | 177 (9.7%) | 0.3 | 20 (47%) | 92 (34%) | 0.13 |
| Prior stroke, no. (%) | 49 (26%) | 260 (14%) | <0.001 | 9 (21%) | 48 (18%) | 0.6 |
| Prior cancer, no. (%) | 29 (16%) | 202 (11%) | 0.067 | 3 (7.0%) | 21 (7.9%) | >0.9 |
| Cardiac Findings | ||||||
| Severe left atrial enlargement | 15 (8.2%) | 190 (11%) | 0.3 | 8 (19%) | 48 (18%) | >0.9 |
| Patent foramen ovale, no. (%) | 39 (21%) | 323 (18%) | 0.3 | 3 (7.0%) | 24 (9.0%) | >0.9 |
| LVEF, median (IQR) | 60 (56, 66) | 60 (57, 65) | >0.9 | 40 (30, 50) | 40 (30, 50) | 0.7 |
| Wall motion abnormality | 2 (1.1%) | 19 (1.1%) | >0.9 | 28 (67%) | 194 (73%) | 0.4 |
| MRI Brain Findings | ||||||
| Multifocal DWI lesions | 52 (31%) | 342 (20%) | 0.001 | 10 (28%) | 48 (21%) | 0.4 |
| Cortical DWI lesions | 141 (85%) | 1,336 (80%) | 0.11 | 31 (86%) | 189 (84%) | 0.7 |
| Cortical T2 lesion | 50 (32%) | 407 (26%) | 0.091 | 13 (38%) | 71 (33%) | 0.6 |
| Stroke Severity and Treatment | ||||||
| Baseline NIHSS, median (IQR) | 6 (2, 14) | 4 (2, 10) | 0.005 | 5 (3, 16) | 6 (2, 14) | 0.7 |
Abbreviations: AC, anticoagulation; AP, antiplatelet; DWI, diffusion-weighted imaging; IQR, interquartile range; LV, left ventricular; LVEF, left ventricular ejection fraction; mRS, modified Rankin Scale; NIHSS, National Institutes of Health Stroke Scale; T2, T2-weighted (MRI image)
Among patients with LV injury, there were 43 patients (13.9%) of patients prescribed AC. Patients with LV injury prescribed AC versus AP were significantly more likely to have been taking an AC immediately prior to the stroke (14% versus 0%).
LV Injury and Clinical Outcomes Association
There were a total of 535 incident events comprising 286 recurrent strokes (258 ischemic), 67 major hemorrhages, and 256 deaths (Table 4). The annualized rate of recurrent ischemic stroke was 5.9% per year among patients with LV injury and 7.9% per year among patients without LV injury. Patients with LV injury had higher unadjusted risks for the composite (HR 1.51, 95% C.I. 1.21-1.87), recurrent ischemic stroke (HR 1.38, 95% C.I. 1.00-1.92), and all-cause mortality (HR 1.60, 95% C.I. 1.17-2.18) outcomes compared to those without LV injury. The mean standardized differences between baseline covariates of patients with versus without LV injury met the threshold of <0.1 after IPW was performed, indicating that the groups were then balanced with respect to these measured factors (Table S5a, Figure S1a). In the IPWLV injury adjusted models, there was no significant association between LV injury and the composite outcome, recurrent ischemic stroke, major bleeding, and all-cause death. Patients with LV injury had higher risk of recurrent hemorrhagic stroke (HR 2.67, 95% C.I. 1.11-6.46). Kaplan-Meier curves stratified by LV injury subtypes are presented in Figure S2 depicting that patients with WMA and EF 20-40% experienced consistently lower event-free survival for the composite outcome compared with those without WMA.
Table 4.
Annualized outcome incidence rates by LV injury status.
| Outcome | Overall | No LV Injury (absolute events, %) |
No LV injury (annualized event rate/year) |
LV Injury (absolute events) |
LV Injury (annualized event rate) |
Unadjusted HR (95% CI) |
IPW-Adjusted HR* (95% CI, p-value) |
|---|---|---|---|---|---|---|---|
| Composite Outcome | 535 / 2,328 (23%) | 436 / 2,018 (22%) | 17.80% | 99 / 310 (32%) | 12.00% | 1.51 (1.21-1.87) | 1.29 (0.97-1.70) |
| Recurrent stroke - ischemic | 258 / 2,155 (12%) | 214 / 1,860 (12%) | 7.90% | 44 / 295 (15%) | 5.90% | 1.38 (1.00-1.92) | 1.14 (0.83-1.57) |
| Recurrent stroke - hemorrhagic | 28 / 2,298 (1.2%) | 21 / 1,991 (1.1%) | 1.30% | 7 / 307 (2.3%) | 0.60% | 2.19 (0.93-5.15) | 2.67 (1.11-6.46 |
| Major bleeding | 67 / 2,328 (2.9%) | 54 / 2,018 (2.7%) | 2.30% | 13 / 310 (4.2%) | 1.50% | 1.58 (0.86-2.89) | 1.09 (0.59-2.03) |
| All-cause death | 256 / 2,328 (11%) | 206 / 2,018 (10%) | 9.00% | 50 / 310 (16%) | 5.70% | 1.60 (1.17-2.18) | 1.26 (0.88-1.80) |
Abbreviations: LV, left ventricular; HR, hazard ratio
IPW model accounts for age, sex, race, ethnicity, insurance status, baseline mRS, AC use prior to stroke, hypertension, dyslipidemia, tobacco use, diabetes, coronary artery disease, prior stroke, prior cancer, left atrial enlargement, presence of patent foramen ovale, multifocal DWI lesions, cortical DWI lesions, cortical T2 lesions, and baseline NIHSS.
Treatment associations by LV injury status subgroups
Upon constructing the IPW for anticoagulation use, there was no significant difference between baseline covariates of patients treated with AC versus AP (Table S5b, Figure S1b). In the overall cohort, there was no significant association between AC use and the primary outcome in the IPW adjusted model (HR 1.04, 95% 0.70–1.54); Table 5). In the LV injury subgroup, AC versus AP therapy was associated with a significantly lower risk of the primary outcome (HR 0.24, 95% 0.10–0.59), compared to a non-significant higher risk of AC versus AP in the non-LV injury subgroup (HR 1.28, 95% CI 0.83–1.95, an effect difference further confirmed by a negative interaction term p[interaction]<0.001). Overlaid Kaplan-Meier survival curves demonstrate that, compared with the AP-No LV injury reference group, patients in the AC-LV injury demonstrated consistently lower event-free survival for the composite outcome, recurrent ischemic stroke, and all-cause death over 24 months (Figure 2). Event rates among LV injury presence-anticoagulation use subgroups are provided in Table S6. Recurrent hemorrhagic stroke rates were 0.7% and 0.9%, respectively, in the LV injury-AC and non-LV injury-AC groups. Major bleeding rates were numerically lowest among patients in the LV injury-AC group (2.4%).
Table 5.
Risk of primary outcome overall and within LV injury subgroups by treatment status.
| Subgroup | N (%) | IPW adjusted HR of AC vs. AP (95% CI) |
p-value (interaction) |
|---|---|---|---|
| Overall treatment Effect | 2328 (100%) | 1.04 (0.70–1.54) | - |
| LV injury presence subgroup | <0.001 | ||
| LV Injury | 310 (13.3%) | 0.24 (0.10–0.59) | |
| No LV Injury | 2018 (86.7%) | 1.28 (0.83–1.95) | |
| LVEF subgroup | 0.002 | ||
| EF 20-40% | 159 (6.8%) | 0.19 (0.04–0.86) | |
| EF >40% | 2169 (93.2%) | 1.16 (0.78–1.72) | |
| WMA subgroup | 0.001 | ||
| No regional WMA | 2086 (89.6%) | 1.20 (0.78–1.84) | |
| Any regional WMA | 242 (10.4%) | 0.33 (0.15–0.73) | |
| WMA location (reference=no WMA) | |||
| Apical WMA | 178 (7.6%) | 0.24 (0.12–0.49) | 0.002 |
| Mid WMA | 190 (8.2%) | 0.28 (0.10–0.84) | 0.003 |
| Basal WMA | 185 (7.9%) | 0.42 (0.24–0.75) | 0.001 |
| WMA severity subgroup | <0.001 | ||
| No regional WMA (reference) | 1727 (89.5%) | 1.20 (0.78-1.84) | |
| Hypokinesis | 150 (7.8%) | 0.37 (0.18-0.74) | |
| Akinesis/Dyskinesis/Aneurysmal | 53 (2.7%) | 0 (0-0) |
Abbreviations: IPW, inverse probability weighting; AC, anticoagulant; AP, antiplatelet; LV, left ventricular; EF, ejection fraction; WMA, wall motion abnormality
Figure 2.
Kaplan-Meier Curves with unadjusted Hazard Ratios for each Outcome by LV Injury Status.
We explored whether the treatment modification occurred with specific LV injury subtypes. AC treatment was beneficial in patients with LVEF 20-40% versus > 40% (HR 0.19, 95% C.I. 0.04-0.86) but not in patients with LVEF >40% (HR 1.16; 95% C.I. 0.78-1.72) with a p-value of interaction of 0.002. Patients with LV WMA experienced a reduced risk of the primary outcome when treated with anticoagulation over antiplatelet therapy (HR 0.33; 95% C.I. 0.15-0.73), unlike those without WMA treated with anticoagulation over antiplatelet therapy (HR 1.20; 95% C.I. 0.78-1.84; p[interaction]=0.001). This pattern was consistent across WMA locations compared to those without WMA: apical WMA (HR 0.24, 95% CI 0.12-0.49, p-interaction=0.002), mid WMA (HR 0.28, 95% CI 0.10-0.84, p[interaction]=0.003), and basal WMA (HR 0.42, 95% CI 0.24-0.75, p-value of interaction=0.001). Finally, treatment associations were also present in patients with a mild degree of WMA (hypokinesis: HR 0.37, 95% C.I. 0.18-0.74) compared to those with no WMA. These results are visually presented as a forest plot in Figure 3.
Figure 3.
Forest Plot of Adjusted Hazard Ratios for Recurrent Stroke: Anticoagulation Therapy Relative to Antiplatelet Therapy Across Subgroups
DISCUSSION
In this real-world, multi-center cohort of patients with cryptogenic strokes, the presence of LV injury was associated with recurrent ischemic stroke risk, major hemorrhage, and all-cause mortality in unadjusted analyses, but these were no longer significant after covariate adjustment. The prescription of anticoagulation therapy within 7 days of the index stroke was associated with a significantly reduced risk of the composite outcome of recurrent stroke, major hemorrhage, and all-cause death among patients with LV injury and no risk difference among those with cryptogenic infarcts without LV injury.
Our study demonstrated that AC use in patients with LV injury appeared to offer protective benefits, with significant reduction in the risk of the primary outcome. While reduced LVEF is a well-established stroke risk factor, guidelines remain equivocal on the relative benefits of AC versus AP following ischemic stroke in patients with LV injury.6 The COMMANDER HF trial27 demonstrated that in patients with an LVEF ≤40%, rivaroxaban reduced the incidence of stroke compared to placebo. In the NAVIGATE ESUS subgroup analysis12 of patients by LV dysfunction status defined by moderate or severe impairment of LV global contractility or LV WMA, patients randomized to rivaroxaban had a lower recurrent stroke risk, a finding that is consistent with our results. Importantly, the subgroup with LV injury in the NAVIGATE ESUS analysis was limited to only 7.1% of the trial cohort, thus our results substantially increase the number of cases in the literature. In the NAVIGATE ESUS subgroup analysis, nearly 81% qualified for the LV dysfunction subgroup definition solely based on the presence of WMA and treatment effects without LV injury subgroups were not investigated. Our study indicates that other patterns of LV injury may benefit from anticoagulation. These results contrast those from another multicenter registry in which pattern LV WMA was an independent predictor of recurrent ischemic stroke (aHR 1.70, 95% CI, 1.22-2.38), while EF <30% was not (unadjusted HR 0.99, 95% CI, 0.32-3.09).1 Exploration of AC versus AP treatment effect in different LV injury subtypes revealed that patients with moderately reduced EF (20%-40%), any WMA, and WMA involving any of the LV segments experienced a significantly lower composite outcome when treated with AC compared to reference subgroups.
While these findings suggest that a broad degree of LV injury is responsive to anticoagulation as a secondary ischemic stroke prevention strategy, the competing risk of hemorrhage with anticoagulation merits consideration. In our study, the primary outcome included the individual outcome of major hemorrhage to ensure that a positive result indicated potential net clinical benefit of anticoagulation. While no head-to-head clinical trials have been performed comparing the various types of direct oral anticoagulants, comparative analyses indicate that there is a differential bleeding profile associated with each medication.28,29 The heterogeneity of oral anticoagulants prescribed in CASPR likely reflects real-world practice and thus the robustness of the point estimates of benefit in the current clinical landscape. Further randomized controlled trials are needed to definitively confirm the safety of AC in patients with LV injury and to identify the specific subgroups that may benefit the most.
In this study, in unadjusted models, patients with moderately reduced LVEF or LV WMA who were diagnosed with cryptogenic stroke experienced a significantly higher risk of the primary outcome, which included recurrent stroke. However, after adjustment for potential confounders such as comorbidities, this association was attenuated and no longer statistically significant. This suggests that a proportion of the unadjusted effect may be influenced by factors like hypertension, diabetes, and coronary artery disease, conditions that are more prevalent in patients with LV injury and could contribute independently to stroke risk.6 It is also likely that power to detect associations between LV injury and outcomes upon adjustment was limited given the sample sizes and event rates. The Kaplan-Meier survival curves visually illustrate that patients with LV injury exhibit a steeper decline in ischemic stroke-free survival over time, particularly after 15 months of the index stroke. A heightened recurrent ischemic risk among patients with LV injury defined by the presence of either reduced LVEF or LV WMA may justify a re-evaluation of the criteria for classification of cardioembolic stroke subtype as well as a search for other forms of LV injury such as peak circumferential strain in the LV apical wall discerned by cardiac MRI.30 Furthermore, the presence of LV injury may prompt further advanced cardiac evaluation such as a cardiac MRI with greater sensitivity to detect an LV thrombus as a source of stroke.31,32 In a study of patients with a prior myocardial infarction of LVEF < 50% with acute ischemic stroke, an LV thrombus was detected by CMR in 20% of patients rather than 1.7% by transthoracic echocardiography.33 Imaging modalities in the stroke diagnostic workflow such as the extended CTA to detect intracardiac thrombi evaluated in the DAYLIGHT trial continue to evolve.34 The risk estimates from this study and others can guide their implementation into clinical practice.
While it has been demonstrated in the literature that moderately reduced LVEF and LV WMA are associated with increased risk of stroke recurrence,4,5 the specific mechanisms by which these types of LV dysfunction subtypes contribute to stroke development remain unclear. As a result, strokes with these subclinical forms of cardiac injury are often classified as cryptogenic. As definitive therapies are identified, there may be a role for a more comprehensive cardiac evaluation in patients with suspected cryptogenic stroke, as traditional diagnostic algorithms may fail to detect these subclinical cardiac abnormalities. Nuanced cardiac evaluation protocols, possibly incorporating advanced cardiac imaging, and multidisciplinary care pathway optimization between neurology and cardiology for those with suspected cryptogenic stroke and LV injury, may be essential to identifying at-risk patients and guiding a personalized approach to intervention selection based on LV characteristics.
The differential benefit of AC by LV injury status is likely multifactorial. One potential driver may be the increased risk of embolic phenomena in those with LV injury due to blood flow aberrancy, dysfunction of the endothelium, and activation of platelets and thrombin.35-37 These targets may be amenable to modification by anticoagulation rather than antiplatelet therapy alone. Interestingly, patients in our cohort with LV injury were significantly more likely to have a T2 FLAIR cortical lesion, suggestive of a remote cortical stroke which tends to be due to an embolic source.18 This finding requires further validation particularly as patients with LV injury harbor other vascular risk factors that contribute to stroke risk, such as small vessel disease, that may also play a role in stroke etiopathogenesis.
Our study has several limitations. With this observational study, a causal relationship cannot be definitively established between anti-thrombotic therapies and recurrent stroke risk among patients with LV injury due to potential unmeasured confounders. Nonetheless, we attempted to offset confounding by indication bias by employing an IPW methodology, a causal inference approach.38 Nonetheless, we attempted to offset confounding by indication bias by employing an IPW methodology. Inclusion was restricted to those hospitalized within two weeks of stroke onset and excluded those with >50% cervical or intracranial stenosis and small vessel disease to isolate cryptogenic mechanisms. This may limit generalizability and underestimate the contribution of overlapping cardiac and atherosclerotic pathologies. The specific type of antiplatelet or anticoagulant prescribed, medication use adherence, and crossovers were not captured in the dataset. There is known interrater reliability of LV injury detection by echocardiography,23,39,40 thus there is a possibility for exposure misclassification. LV injury could be detected using a variety of cardiac imaging modalities, which vary in their sensitivity and specificity for LV injury subtypes. Some of the LV injury subtypes were infrequent, likely leading to underpowered subgroup analyses. There was no subsequent echocardiographic data available.
This multicenter real-world cohort study demonstrated that patients with an ischemic stroke of undetermined etiology and LV injury may benefit from anticoagulation rather than antiplatelet monotherapy. Dedicated randomized trials are needed to definitively determine the optimal antithrombotic therapy for this patient population.
Supplementary Material
Checklist
Tables S1-S6
Figures S1-S2
Funding Sources:
This work was supported by funding from Medtronic and Philips.
Disclosure Statement:
Dr. Sharma reports grants from NIH Clinical Center.
Dr. Kerrigan reports compensation from Bristol-Myers Squibb for consultant services.
Dr. Nguyen reports compensation from brainomix for consultant services; compensation from Route 92 Medical, Inc. for consultant services; compensation from Genentech for other services; compensation from Medtronic for consultant services; compensation from American Stroke Association for other services; compensation from Aruna for consultant services; and compensation from Kaneka for other services.
Dr. Rothstein reports grants from American Heart Association.
Dr. Liebeskind reports compensation from Genentech for consultant services; compensation from Cerenovus for consultant services; compensation from Medtronic for consultant services; compensation from Rapid Medical Ltd for consultant services; and compensation from Stryker for consultant services.
Dr. Stretz reports grants from American Heart Association to other; grants from Duke University Medical Center to other; and grants from Massachusetts General Hospital to other.
Dr. Eklund reports grants from University of Colorado, Denver/Anschutz Medical Campus and compensation from University of Colorado, Denver/Anschutz Medical Campus for other services.
Dr. de Havenon reports stock options in Certus; stock options in TitinKM; and compensation from Novo Nordisk for consultant services.
Dr. Siegler reports grants from Medtronic; compensation from Novartis for consultant services; grants from Viz.ai; compensation from AstraZeneca for other services; compensation from Bayer for consultant services; grants from Philips.
The remaining authors report no disclosures relevant to this manuscript.
NON-STANDARD ABBREVIATIONS AND ACRONYMS:
- AC
anticoagulation agents
- AP
anti-platelet agents
- CASPR
Cardiac Abnormalities in Stroke Prevention and Recurrence
- CI
confidence interval
- CT
computed tomography
- DWI
Diffusion weighted imaging
- ESUS
embolic stroke of undetermined source
- FLAIR
Fluid-Attenuated Inversion Recovery
- HR
hazard ratio
- IPW
Inverse probability weighting
- LVEF
Left ventricular ejection fraction
- MRI
magnetic resonance imaging
- OR
odds ratio
- TOAST
Trial of Org 10172 in Acute Stroke Treatment
- WMA
Wall motion abnormality
REFERENCES
- 1.Choi J-Y, Cha J, Jung J-M, Seo W-K, Oh K, Cho K-H, Yu S. Left ventricular wall motion abnormalities are associated with stroke recurrence. Neurology. 2017;88; 586–594. [DOI] [PubMed] [Google Scholar]
- 2.Hays AG, Sacco RL, Rundek T, Sciacca RR, Jin Z, Liu R, Homma S, Di Tullio MR. Left ventricular systolic dysfunction and the risk of ischemic stroke in a multiethnic population. Stroke. 2006;37; 1715–1719. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Pfeffer MA, Braunwald E, Moyé LA, Basta L, Brown EJ, Cuddy TE, Davis BR, Geltman EM, Goldman S, Flaker GC, et al. Effect of Captopril on Mortality and Morbidity in Patients with Left Ventricular Dysfunction after Myocardial Infarction. New England Journal of Medicine. 1992;327; 669–677. [DOI] [PubMed] [Google Scholar]
- 4.Loh E, Sutton MStJ, Wun C-CC, Rouleau JL, Flaker GC, Gottlieb SS, Lamas GA, Moyé LA, Goldhaber SZ, Pfeffer MA. Ventricular Dysfunction and the Risk of Stroke after Myocardial Infarction. New England Journal of Medicine. 1997;336; 251–257. [DOI] [PubMed] [Google Scholar]
- 5.Hosseini Farahabadi M, Milani-Nejad S, Liu S, Yu W, Shafie M. Left Atrial Dilatation and Reduced Left Ventricular Ejection Fraction Are Associated With Cardioembolic Stroke. Frontiers in Neurology. 2021;12:680651. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Kleindorfer DO, Towfighi A, Chaturvedi S, Cockroft KM, Gutierrez J, Lombardi-Hill D, Kamel H, Kernan WN, Kittner SJ, Leira EC, et al. 2021 Guideline for the Prevention of Stroke in Patients With Stroke and Transient Ischemic Attack: A Guideline From the American Heart Association/American Stroke Association. Stroke. 2021;52; e364–e467. [DOI] [PubMed] [Google Scholar]
- 7.Baker AD, Schwamm LH, Sanborn DY, Furie K, Stretz C, Mac Grory B, Yaghi S, Kleindorfer D, Sucharew H, Mackey J, et al. Acute Ischemic Stroke, Depressed Left Ventricular Ejection Fraction, and Sinus Rhythm: Prevalence and Practice Patterns. Stroke. 2022;53; 1883–1891. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Beggs SAS, Rørth R, Gardner RS, McMurray JJV. Anticoagulation therapy in heart failure and sinus rhythm: a systematic review and meta-analysis. Heart (British Cardiac Society). 2019;105; 1325–1334. [DOI] [PubMed] [Google Scholar]
- 9.Eikelboom JW, Connolly SJ, Bosch J, Dagenais GR, Hart RG, Shestakovska O, Diaz R, Alings M, Lonn EM, Anand SS, et al. Rivaroxaban with or without Aspirin in Stable Cardiovascular Disease. New England Journal of Medicine. 2017;377; 1319–1330. [DOI] [PubMed] [Google Scholar]
- 10.Homma S, Thompson JLP, Pullicino PM, Levin B, Freudenberger RS, Teerlink JR, Ammon SE, Graham S, Sacco RL, Mann DL, et al. Warfarin and Aspirin in Patients with Heart Failure and Sinus Rhythm. New England Journal of Medicine. 2012;366; 1859–1869. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Reddin C, Judge C, Loughlin E, Murphy R, Costello M, Alvarez A, Ferguson J, Smyth A, Canavan M, O’Donnell MJ. Association of Oral Anticoagulation With Stroke in Atrial Fibrillation or Heart Failure. Stroke. 2021;52; 3151–3162. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Merkler AE, Pearce LA, Kasner SE, Shoamanesh A, Birnbaum LA, Kamel H, Sheth KN, Sharma R. Left Ventricular Dysfunction Among Patients With Embolic Stroke of Undetermined Source and the Effect of Rivaroxaban vs Aspirin: A Subgroup Analysis of the NAVIGATE ESUS Randomized Clinical Trial. JAMA Neurology. 2021;78; 1454–1460. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Sharma R, Jillella D, Zhang C, Elkind MSV, Kamel H, Di Tullio MR, Kronmal RA, Krishnaiah B, Yaghi S, Longstreth WT, et al. Apixaban and Recurrent Stroke Risk With Left Ventricular Dysfunction: A Secondary Analysis of the ARCADIA Trial. Stroke. 2025;56; 3118–3126. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Siegler JE, Goicoechea EB, Penckofer M, Eklund K, Yaghi S, Stretz C, Lineback CM, Stamm B, Peter S, D’Souza M, et al. Outcomes of Patients With Embolic Stroke of Undetermined Source Treated With Antiplatelet Agents or Anticoagulation: A Multicenter Cohort Study. Neurology. 2025;105; e213876. [DOI] [PubMed] [Google Scholar]
- 15.Schwamm LH, Fonarow GC, Reeves MJ, Pan W, Frankel MR, Smith EE, Ellrodt G, Cannon CP, Liang L, Peterson E, et al. Get With the Guidelines-Stroke is associated with sustained improvement in care for patients hospitalized with acute stroke or transient ischemic attack. Circulation. 2009;119; 107–115. [DOI] [PubMed] [Google Scholar]
- 16.Goldstein LB. Accuracy of ICD-9-CM coding for the identification of patients with acute ischemic stroke: effect of modifier codes. Stroke. 1998;29; 1602–1604. [DOI] [PubMed] [Google Scholar]
- 17.Shirley AM, Morrisette KL, Choi SK, Reynolds K, Zhou H, Zhou MM, Wei R, Zhang Y, Cheng P, Wong E, et al. Validation of ICD-10 hospital discharge diagnosis codes to identify incident and recurrent ischemic stroke from a US integrated healthcare system. Pharmacoepidemiology and Drug Safety. 2023;32; 1439–1445. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Hart RG, Diener H-C, Coutts SB, Easton JD, Granger CB, O’Donnell MJ, Sacco RL, Connolly SJ, Cryptogenic Stroke/ESUS International Working Group. Embolic strokes of undetermined source: the case for a new clinical construct. The Lancet. Neurology. 2014;13; 429–438. [DOI] [PubMed] [Google Scholar]
- 19.Ferguson GG, Eliasziw M, Barr HWK, Clagett GP, Barnes RW, Wallace MC, Taylor DW, Haynes RB, Finan JW, Hachinski VC, et al. The North American Symptomatic Carotid Endarterectomy Trial. Stroke. 1999;30; 1751–1758. [DOI] [PubMed] [Google Scholar]
- 20.Schulman S, Angerås U, Bergqvist D, Eriksson B, Lassen MR, Fisher W, Subcommittee on Control of Anticoagulation of the Scientific and Standardization Committee of the International Society on Thrombosis and Haemostasis. Definition of major bleeding in clinical investigations of antihemostatic medicinal products in surgical patients. Journal of thrombosis and haemostasis: JTH. 2010;8; 202–204. [DOI] [PubMed] [Google Scholar]
- 21.Olsen SL, Gilbert EM, Renlund DG, Taylor DO, Yanowitz FD, Bristow MR. Carvedilol improves left ventricular function and symptoms in chronic heart failure: a double-blind randomized study. Journal of the American College of Cardiology. 1995;25; 1225–1231. [DOI] [PubMed] [Google Scholar]
- 22.Nahar T, Croft L, Shapiro R, Fruchtman S, Diamond J, Henzlova M, Machac J, Buckley S, Goldman ME. Comparison of four echocardiographic techniques for measuring left ventricular ejection fraction. The American Journal of Cardiology. 2000;86; 1358–1362. [DOI] [PubMed] [Google Scholar]
- 23.Pellikka PA, She L, Holly TA, Lin G, Varadarajan P, Pai RG, Bonow RO, Pohost GM, Panza JA, Berman DS, et al. Variability in Ejection Fraction Measured By Echocardiography, Gated Single-Photon Emission Computed Tomography, and Cardiac Magnetic Resonance in Patients With Coronary Artery Disease and Left Ventricular Dysfunction. JAMA Network Open. 2018;1; e181456. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Adams HP, Bendixen BH, Kappelle LJ, Biller J, Love BB, Gordon DL, Marsh EE. Classification of subtype of acute ischemic stroke. Definitions for use in a multicenter clinical trial. TOAST. Trial of Org 10172 in Acute Stroke Treatment. Stroke. 1993;24; 35–41. [DOI] [PubMed] [Google Scholar]
- 25.Brookhart MA, Schneeweiss S, Rothman KJ, Glynn RJ, Avorn J, Stürmer T. Variable Selection for Propensity Score Models. American Journal of Epidemiology. 2006;163; 1149–1156. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Austin PC. Using the Standardized Difference to Compare the Prevalence of a Binary Variable Between Two Groups in Observational Research. Communications in Statistics - Simulation and Computation. 2009;38; 1228–1234. [Google Scholar]
- 27.Zannad F, Anker SD, Byra WM, Cleland JGF, Fu M, Gheorghiade M, Lam CSP, Mehra MR, Neaton JD, Nessel CC, et al. Rivaroxaban in Patients with Heart Failure, Sinus Rhythm, and Coronary Disease. New England Journal of Medicine. 2018;379; 1332–1342. [DOI] [PubMed] [Google Scholar]
- 28.Ingason AB, Hreinsson JP, Agustsson AS, Lund SH, Rumba E, Palsson DA, Reynisson IE, Gudmundsdottir BR, Onundarson PT, Bjornsson ES. Comparison of the effectiveness and safety of direct oral anticoagulants: a nationwide propensity score–weighted study. Blood Advances. 2023;7; 2564–2572. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Lau WCY, Torre CO, Man KKC, Stewart HM, Seager S, Van Zandt M, Reich C, Li J, Brewster J, Lip GYH, et al. Comparative Effectiveness and Safety Between Apixaban, Dabigatran, Edoxaban, and Rivaroxaban Among Patients With Atrial Fibrillation. Annals of Internal Medicine. 2022;175; 1515–1524. [DOI] [PubMed] [Google Scholar]
- 30.Sharma R, Hansen S, Merkler AE, Lima JAC, Longstreth WT. Left Ventricular Injury Detected by Cardiac MRI and Incident Ischemic Stroke and Dementia Risk: The Multi-Ethnic Study of Atherosclerosis. Neurology. 2025;104; e213606. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Baher A, Mowla A, Kodali S, Polsani VR, Nabi F, Nagueh SF, Volpi JJ, Shah DJ. Cardiac MRI Improves Identification of Etiology of Acute Ischemic Stroke. Cerebrovascular Diseases. 2014;37; 277–284. [DOI] [PubMed] [Google Scholar]
- 32.Camen S, Haeusler KG, Schnabel RB. Cardiac Imaging After Ischemic Stroke or Transient Ischemic Attack. Current Neurology and Neuroscience Reports. 2020;20; 36. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Takasugi J, Yamagami H, Noguchi T, Morita Y, Tanaka T, Okuno Y, Yasuda S, Toyoda K, Gon Y, Todo K, et al. Detection of Left Ventricular Thrombus by Cardiac Magnetic Resonance in Embolic Stroke of Undetermined Source. Stroke. 2017;48; 2434–2440. [DOI] [PubMed] [Google Scholar]
- 34.Sposato LA, Ayan D, Ahmed M, Fridman S, Mandzia JL, Elrayes M, Lodol F, Khaw AV, Mai LM, Bogiatzi C, et al. Extended CT angiography versus standard CT angiography for the detection of cardioaortic thrombus in patients with ischaemic stroke and transient ischaemic attack (DAYLIGHT): a prospective, randomised, open-label, blinded end-point trial. The Lancet Neurology. 2025;24; 489–499. [DOI] [PubMed] [Google Scholar]
- 35.Merlini PA, Bauer KA, Oltrona L, Ardissino D, Cattaneo M, Belli C, Mannucci PM, Rosenberg RD. Persistent activation of coagulation mechanism in unstable angina and myocardial infarction. Circulation. 1994;90; 61–68. [DOI] [PubMed] [Google Scholar]
- 36.Cusick DA, Bonow RO, Chaudhry FA. Left Ventricular Apical Thrombus and Myocardial Viability: A Dobutamine Stress Echocardiographic Study. Echocardiography. 2000;17; 547–554. [DOI] [PubMed] [Google Scholar]
- 37.Delewi R, Zijlstra F, Piek JJ. Left ventricular thrombus formation after acute myocardial infarction. Heart. 2012;98; 1743–1749. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38.Chesnaye NC, Stel VS, Tripepi G, Dekker FW, Fu EL, Zoccali C, Jager KJ. An introduction to inverse probability of treatment weighting in observational research. Clinical Kidney Journal. 2022;15; 14–20. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39.Khush KK, Nguyen J, Goldstein BA, McGlothlin DP, Zaroff JG. Reliability of transthoracic echocardiogram interpretation in potential adult heart transplant donors. The Journal of Heart and Lung Transplantation: The Official Publication of the International Society for Heart Transplantation. 2015;34; 266–269. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 40.Kataoka A, Scherrer-Crosbie M, Senior R, Gosselin G, Phaneuf D, Guzman G, Perna G, Lara A, Kedev S, Mortara A, et al. The value of core lab stress echocardiography interpretations: observations from the ISCHEMIA Trial. Cardiovascular Ultrasound. 2015;13:47. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
Data from this study will be made available to a qualified investigator upon reasonable request to the principal investigator of the CASPR Registry (co-author J.S.) following completion of data transfer agreements. We followed the Strengthening the Reporting of Observational Studies in Epidemiology reporting guidelines for this study.