Introduction

Embolic stroke of undetermined source (ESUS) accounts for approximately one‑sixth of all ischemic strokes. It is defined as nonlacunar infarction in which, despite indicated diagnostic evaluation, no clear underlying cause, such as atrial fibrillation (AF), large artery stenosis, or another specific etiology, can be established.^1-4 The ESUS concept, introduced in 2014, provided standardized diagnostic criteria that allowed for harmonization of clinical research, and it was quickly adopted in practice.^1,5 More recently, the definition and diagnostic algorithm have been updated (Table 1).⁵ Importantly, ESUS differs from the broader category of “cryptogenic” stroke, which also includes patients with incomplete diagnostic work‑up or multiple competing causes.⁶

The long‑term cardiovascular risk in ESUS patients is high. Within 5 years, approximately 30% of the ESUS patients experience recurrent stroke, and more than one‑third suffer from other major cardiovascular events.³ This highlights the need for effective secondary prevention strategies. However, risk reduction in ESUS remains particularly complex,⁷ as potential embolic sources often coexist, and it is rarely possible to determine which mechanism is truly causal.^8,9

The current standard antithrombotic therapy for ESUS is antiplatelets.¹⁰ This approach has persisted despite high‑quality research on the hypothesis that a significant proportion of ESUS patients have occult AF that has eluded detection. The hypothesis that anticoagulation would prove superior to antiplatelet therapy was tested in large randomized controlled trials (RCTs). Both the NAVIGATE‑ESUS (New Approach Rivaroxaban Inhibition of Factor Xa in a Global Trial vs ASA to Prevent Embolism in Embolic Stroke of Undetermined Source)¹¹ and the RE‑SPECT ESUS (Randomized, Double‑Blind, Evaluation in Secondary Stroke Prevention Comparing the Efficacy and Safety of the Oral Thrombin inhibitor Dabigatran etexilate versus Acetylsalicylic acid in patients with Embolic Stroke of Undetermined Source)¹² trials failed to show superiority of oral anticoagulants (OACs) over aspirin in the unselected ESUS population. More recently, trials focusing on narrower ESUS groups, such as the ATTICUS (Apixaban for Treatment of Embolic Stroke of Undetermined Source) trial,¹³ which enrolled ESUS patients with risk factors for cardioembolism, and the ARCADIA (Atrial Cardiopathy and Antithrombotic Drugs in Prevention After Cryptogenic Stroke) study,¹⁴ which focused on ESUS patients with AC, did not demonstrate benefits of OACs over aspirin (Table 2).

Given that these antithrombotic agents act through different mechanisms in stroke prevention— antiplatelets targeting platelet aggregation in atherosclerosis and OACs mainly inhibiting the coagulation cascade (eg, in cardioembolism)—it is plausible that within the heterogeneous ESUS population, there are still subgroups more likely to benefit from OACs rather than antiplatelet therapy. Identifying these subgroups remains a key challenge. The purpose of this review was to discuss which ESUS patients may derive clinical benefits from OACs, based on the evidence from the most recent RCTs.

The unselected population of embolic stroke of undetermined source

The NAVIGATE ESUS trial was the first large‑scale RCT assessing anticoagulation in ESUS patients.^15,16 A total of 7213 participants with recent ESUS were randomized to rivaroxaban 15 mg daily or aspirin 100 mg daily. After median follow‑up of 11 months, the trial was terminated early due to futility and excess bleeding in the rivaroxaban arm. The primary end point (any recurrent stroke or systemic embolism) occurred at an annual rate of 5.1% in the rivaroxaban group vs 4.8% in the aspirin group (hazard ratio [HR], 1.07; 95% CI, 0.87–1.33). The rates of recurrent ischemic stroke were identical (4.7% per year in both groups), but major bleeding was significantly higher in the rivaroxaban group (1.8% vs 0.7%; HR, 2.72; 95% CI, 1.68–4.39).¹¹

Despite the neutral primary result, prespecified and exploratory subgroup analyses provided important insights. In particular, the patients with left atrial (LA) enlargement, defined as LA diameter above 4.6 cm, showed a considerably reduced risk of recurrent stroke on rivaroxaban, as compared with aspirin.¹⁷ This finding suggests that atrial structural disease, even in the absence of AF, may represent a pathophysiologic substrate responsive to anticoagulation. Another secondary analysis focused on the patients with left ventricular (LV) dysfunction, defined by impaired contractility or regional wall motion abnormalities. In this group, rivaroxaban was associated with lower rates of recurrent stroke or systemic embolism than aspirin (HR, 0.36; 95% CI, 0.14–0.93).¹⁸ This aligns with previous evidence linking LV systolic dysfunction to increased thromboembolic risk.¹⁹ These subgroup findings highlight that, although the NAVIGATE‑ESUS was neutral overall, anticoagulation may be effective in patients with structural cardiac abnormalities predisposing to embolism.

Another large‑scale trial, RE‑SPECT ESUS, enrolled 5390 patients and compared dabigatran (110 mg or 150 mg twice daily) with aspirin (100 mg daily). The eligible patients were 60 years or older with a qualifying ESUS event within the prior 3 months or within the prior 6 months, if they also had at least 1 vascular risk factor for stroke (mild‑to‑moderate symptomatic heart failure, diabetes mellitus, hypertension, patent foramen ovale [PFO], prior transient ischemic attack / stroke before the qualifying stroke, or a CHA₂DS₂-VASc score ≥3). Patients aged 18–59 years with at least 1 of the aforementioned risk factors were also eligible if their ESUS occurred within the previous 3 months. The primary end point was time to the first recurrent stroke. During mean follow‑up of 19 months, the annual rate of recurrent stroke was 4.1% on dabigatran vs 4.8% on aspirin (HR, 0.85; 95% CI, 0.69–1.03), a nonsignificant difference.¹² Major bleeding rates were comparable between the groups (HR, 1.19; 95% CI, 0.85–1.66).

However, subgroup analyses again suggested differential effects. The patients aged 75 years or older appeared to benefit from dabigatran, with a 37% relative risk reduction in recurrent stroke (HR, 0.63; 95% CI, 0.43–0.94). Given the high prevalence of covert AF and AC in the elderly, this finding is biologically plausible.²⁰ Additionally, the Kaplan–Meier survival curves began to diverge approximately 2 years after randomization, suggesting that longer follow‑up may have showed a greater benefit of anticoagulation, possibly due to delayed AF detection. Thus, while the RE‑SPECT ESUS trial did not show superiority of anticoagulation overall, its subgroup analyses reinforced the potential value of targeting elderly patients or those at a high risk of atrial cardiopathy.

The NAVIGATE ESUS and RE‑SPECT ESUS RCTs demonstrated no superiority of anticoagulation over aspirin in the unselected ESUS population, but highlighted specific subgroups that could benefit from it, particularly those with atrial structural abnormalities or at advanced age.

Embolic stroke of undetermined source with cardioembolic risk factors

Recognizing the heterogeneity of ESUS, the ATTICUS trial focused on patients with clinical or imaging features suggestive of a cardioembolic source. The eligible participants had at least 1 marker of potential atrial or cardiac pathology, including enlarged LA (>4.5 cm), spontaneous echo contrast in the LA appendage, reduced appendage flow velocity (≤0.2 m/s), atrial high‑rate episodes (AHREs), PFO, or a CHA₂DS₂-VASc score equal to or greater than 4. The patients were randomized to the apixaban (5 mg twice daily) or aspirin (100 mg daily) groups, and were followed with continuous rhythm monitoring using an implantable cardiac device. The primary end point was the occurrence of new ischemic lesions on brain magnetic resonance imaging within 12 months.^14,21,22

The trial demonstrated that apixaban was not superior to aspirin in preventing new ischemic brain lesions, even in this enriched ESUS population with clinical or imaging features suggestive of a cardioembolic source. The safety profile was also comparable between the groups, with consistent effects across prespecified subgroups. Importantly, early analyses showed that AHREs were the most powerful predictors of subsequent clinical AF, highlighting their value in risk stratification.²³ While the ATTICUS trial was not powered to evaluate differences in clinical outcomes, such as recurrent stroke, its findings support the rationale for targeting ESUS patients with atrial electrical or structural abnormalities in future anticoagulation trials.

Embolic stroke of undetermined source with atrial cardiopathy

The ARCADIA trial specifically enrolled ESUS patients with evidence of AC but without documented AF. AC was defined by at least 1 of 3 criteria: abnormal P‑wave terminal force in lead V1 (>5000 μV × ms), elevated N‑terminal pro–B‑type natriuretic peptide level (NT‑proBNP >250 pg/ml), or increased LA diameter index (≥3 cm/m²). A total of 1015 patients were randomized to receive apixaban (5 mg or 2.5 mg twice daily) or aspirin (81 mg daily), with recurrent stroke as the primary outcome.¹³

The trial results were neutral, showing no significant difference between apixaban and aspirin. A major limitation, however, was that nearly half of the participants were included based solely on the elevated NT‑proBNP level, a biomarker influenced by multiple noncardiac conditions, and not specific to atrial disease. This likely resulted in substantial misclassification, attenuating any potential treatment effect. Subsequent meta‑analyses employing broader definitions of AC, incorporating electrocardiography, imaging, and biomarker evidence, suggested that anticoagulation may in fact reduce recurrence in this subgroup.²⁴

Taken together, the ATTICUS and ARCADIA trials underscore the central challenge of ESUS management: not all patients share the same underlying mechanism. The failure of anticoagulation in unselected ESUS populations does not undermine the ESUS concept itself but highlights the need for refined risk stratification. Patients with convincing evidence of AC or markers of atrial dysfunction may represent the subgroup most likely to benefit from anticoagulation, while others may be better managed with antiplatelet‑based strategies.

Meta‑analyses

Although individual randomized trials in ESUS have produced neutral results, several pooled analyses have explored whether anticoagulation may be advantageous in specific subgroups. A study‑level meta‑analysis of the 4 major RCTs—NAVIGATE‑ESUS, RE‑SPECT ESUS, ATTICUS, and ARCADIA—examined over 13 000 patients. In the overall population, there was no marked reduction in recurrent ischemic stroke with anticoagulation, as compared with aspirin (recurrence rate [RR], 0.91; 95% CI, 0.8–1.05; P = 0.2; I² = 0%). Similarly, the risk of major bleeding was similar (RR, 1.2; 95% CI, 0.5–2.84; P = 0.68), although heterogeneity was substantial (I² = 78%). No mortality difference was observed between the treatment arms (RR, 1.11; 95% CI, 0.87–1.42; P = 0.39; I² = 0%). These pooled data confirm that, in unselected ESUS patients, empiric anticoagulation does not improve outcomes in comparison with aspirin.²⁵

Closer inspection of the subgroups, however, offers more nuance. The patients with PFO represented an important subgroup. A pooled analysis of 6 studies involving approximately 2000 ESUS patients with PFO found that anticoagulation was associated with a lower risk of recurrent ischemic stroke than antiplatelet therapy (RR, 0.59; 95% CI, 0.35–0.98; P = 0.04; I² = 0%). Interestingly, this effect was predominantly observed in the studies using vitamin K antagonists rather than direct OACs (DOACs), raising questions about potential differences in mechanisms or study populations.²⁵

LA enlargement has also been examined. Subgroup analyses from 3 RCTs, covering 461 patients with enlarged atria (defined as LA diameter >4.6 cm, except in the ATTICUS study, where it was defined as >4.5 cm), did not show a benefit of anticoagulation (RR, 1.77; 95% CI, 0.17–18.03; P = 0.63; I² = 79%) over antiplatelet therapy. A striking observation emerged when comparing the trials that prohibited postrandomization cardiac monitoring with those that allowed it. In the NAVIGATE‑ESUS trial, which prohibited monitoring, rivaroxaban was superior to aspirin among the patients with LA enlargement (RR, 0.25; 95% CI, 0.07–0.89). By contrast, in the ATTICUS and ARCADIA trials, both of which allowed continuous monitoring and therefore facilitated early initiation of anticoagulation in the aspirin arm once AF was detected, aspirin was associated with lower RR of 6.65 (95% CI, 1.26–35.08; P = 0.03). This divergence suggests that trial design, particularly regarding rhythm monitoring, may influence observed treatment effects.²⁵

Age has also emerged as a modifier of treatment response. When the patients were stratified by age, a meta‑analysis showed no benefit of anticoagulation in those younger than 75 years. In contrast, in the patients aged 75 years or older, there was a consistent trend toward lower recurrent stroke with anticoagulation.²⁵ This finding is in line with the RE‑SPECT ESUS trial, where dabigatran was superior to aspirin in this subgroup.^20,26 The biological plausibility is strong, given the high prevalence of AC and occult AF in elderly patients.

A more recent comprehensive meta‑analysis pooled 9 RCTs including 15 451 patients with either cryptogenic stroke or ESUS.²⁷ Although anticoagulation did not demonstrate overall superiority in comparison with antiplatelet therapy, 2 clinical factors appeared to differentiate the patients more likely to benefit from coagulation. The first was the presence of supracardiac atherosclerosis. The patients with low‑risk plaques had a significantly reduced risk of recurrent ischemic stroke when treated with OACs, as compared with antiplatelets (RR, 0.53; 95% CI, 0.35–0.8; I² = 0%). By contrast, the patients with high‑risk supracardiac atherosclerosis, defined as nonstenotic plaques lower than or equal to 50% in the aortic arch or the carotid, vertebral, or intracranial arteries, derived no additional benefit (RR, 0.91; 95% CI, 0.78–1.96; I² = 0%), and those with any supracardiac atherosclerosis showed a neutral effect (RR, 1.13; 95% CI, 0.84–1.53; I² = 0%). The second factor was AC. The patients with markers or risk factors for AC experienced greater benefit from anticoagulation (RR, 0.84; 95% CI, 0.7–0.99; I² = 0%), as compared with those without such factors (RR, 1.05; 95% CI, 0.85–1.3; I² = 0%). No interaction was observed in terms of PFO status (P = 0.28).²⁷ These results reinforce the concept that the underlying pathophysiology, rather than broad ESUS designation, determines which patients derive net clinical benefit from anticoagulation.

A third systematic review and meta‑analysis, including 9 studies with 14 582 ESUS patients, specifically compared DOACs with aspirin for secondary prevention.²⁸ This analysis confirmed that DOACs did not significantly reduce the risk of recurrent stroke, which is consistent with the findings from individual RCTs. Importantly, there was no clear evidence of an increased risk of major bleeding or symptomatic intracerebral hemorrhage. Subgroup analyses of the patients with features suggestive of cardioembolism or AC showed a trend toward reducing recurrent stroke with DOACs, although it was insignificant. These findings further emphasize the need for refined risk stratification to identify ESUS subgroups that may benefit from anticoagulation.²⁸

Real‑world evidence

Observational registries have provided complementary perspectives by examining outcomes in ESUS patients managed in routine clinical practice. The CASPR (Cardiac Abnormalities in Stroke Prevention and Risk of Recurrence) study²⁹ is the most comprehensive effort to date. This multicenter retrospective registry included data from 27 sites and 2328 consecutive adult patients admitted with acute ESUS between 2015 and 2024. Treatment decisions regarding anticoagulation vs antiplatelet therapy were left at the physician’s discretion, thus reflecting real‑world practice.

In the overall population, CASPR found no advantages of anticoagulation over antiplatelets for prevention of recurrent stroke, consistent with the neutral findings of the randomized trials. However, subgroup analyses provided important additional insights. In the patients with evidence of LV injury, defined as impaired contractility, regional wall motion abnormality, or both, anticoagulation was associated with a markedly lower risk of adverse outcomes. Specifically, the adjusted HR for the primary outcome in this group was 0.35 (95% CI, 0.16–0.77; interaction P <⁠0.01). A trend toward a lower recurrent ischemic stroke rate was also observed (HR, 0.22; 95% CI, 0.05–1.08; interaction P = 0.04). Although the number of patients with LV injury treated with OACs was relatively small, these findings corroborate secondary analyses from the NAVIGATE‑ESUS trial and strengthen the argument that LV dysfunction identifies a subgroup of ESUS patients in whom anticoagulation is beneficial.²⁹

Conclusions

Up to one‑sixth of ischemic strokes are classified as ESUS, and the annual recurrence risk is approximately 4%–5% despite optimal secondary prevention strategies.² Large RCTs have not demonstrated superiority of DOACs over aspirin in the general ESU S population^11-14; however, subgroup analyses have consistently indicated that certain patient groups may derive meaningful benefits from anticoagulation.^30,31 Individuals with LA enlargement or other markers of AC, with LV dysfunction or injury, and patients aged 75 years or older, who carry a greater burden of covert AF, have shown signals of improved outcomes with oral anticoagulation. Similarly, patients with medically managed PFO may represent populations in which anticoagulation offers protective benefit (Table 3).^25,32

AHREs and subclinical AF are increasingly recognized as possible mechanisms linking AC to embolic stroke.³³ While the benefit of OACs is well established in patients with clinically documented AF, its role in short, device‑detected arrhythmias remains uncertain. Continuous rhythm monitoring after stroke frequently identifies brief, asymptomatic AF episodes with a low overall burden, which carry a modestly increased risk of stroke in comparison with clinical AF. Recent randomized studies showed that anticoagulation in patients with AHRE or subclinical AF can lower the incidence of ischemic stroke by approximately one‑third, though this advantage is accompanied by a higher rate of major bleeding.^33,34 The duration and timing of AHREs appear to influence the stroke risk, and further research is required to determine which patients with AHREs or subclinical AF are most likely to benefit from anticoagulation.

By contrast, patients lacking evidence of atrial disease or those with atherosclerosis do not appear to benefit from OACs, and empiric anticoagulation in such groups may expose them to unnecessary bleeding risk. These observations underscore the heterogeneity of ESUS and highlight the limitations of a uniform treatment approach. Instead, secondary prevention in ESUS should move toward precision medicine strategies that integrate cardiovascular imaging and biomarker profiling to better identify patients with covert cardioembolic sources.³⁵

Another therapeutic approach to consider is dual‑pathway inhibition with low‑dose rivaroxaban (2.5 mg twice daily) plus aspirin, as evaluated in the COMPASS (Rivaroxaban for the Prevention of Major Cardiovascular Events in Coronary or Peripheral Artery Disease) trial.³⁶ It demonstrated significant reductions in major cardiovascular and ischemic events among high‑risk patients without AF. This strategy may represent an attractive option for selected ESUS patients, although further studies are needed to confirm its efficacy and safety in this population.³⁷

Factor XI inhibitors represent a novel class of anticoagulants that may reduce thromboembolic risk while minimizing bleeding complications. Early clinical studies have shown encouraging safety profiles, and ongoing phase III trials, including OCEANIC‑STROKE (A Study to Test Asundexian for Preventing a Stroke Caused by a Clot in Participants After an Acute Ischemic Stroke or After a High‑risk Transient Ischemic Attack) and LIBREXIA‑STROKE (A Study of Milvexian in Participants After an Acute Ischemic Stroke or High‑Risk Transient Ischemic Attack), are evaluating the efficacy of factor XI inhibitors, as compared with placebo, against the background of standard antiplatelet therapy for secondary prevention after noncardioembolic stroke or high‑risk transient ischemic attack.³⁸ Given their favorable balance between efficacy and a bleeding risk, factor XI inhibitors could in the future represent a potential therapeutic option for selected ESUS patients in whom traditional OACs offer limited benefit or carry an excessive bleeding risk.

While most ESUS patients require appropriate antiplatelet therapy,¹⁰ certain subsets may require oral anticoagulation. Individualized management guided by underlying pathophysiological clues and informed by shared decision‑making between patients and clinicians remains the optimal approach.³⁹ Future research should focus on refining the definition of AC, validating more precise risk stratification tools, and evaluating targeted therapeutic strategies in well‑defined ESUS subgroups to reduce the burden of recurrent stroke in this diverse patient population.