Current perspectives on the clinical management of cryptogenic stroke

Abstract

Introduction:

Cryptogenic stroke is a heterogenous entity defined as an ischemic stroke for which no probable cause is identified despite thorough diagnostic evaluation. Since about a quarter of all ischemic strokes are classified as cryptogenic, it is a commonly encountered problem for providers as secondary stroke prevention is guided by stroke etiology.

Areas Covered:

In this review, the authors provide an overview of stroke subtype classification schemes and diagnostic evaluation in cryptogenic stroke. They then detail putative cryptogenic stroke mechanisms, their therapeutic implications, and ongoing research. This review synthesizes the available evidence on PubMed up to December 2022.

Expert opinion:

Cryptogenic stroke is an evolving concept that changes with ongoing research. Investigations are focused on improving our diagnostic capabilities and solidifying useful constructs within cryptogenic stroke that could become therapeutically targetable subgroups within an otherwise nonspecific entity. Advances in technology may help move specific proposed cryptogenic stroke mechanisms from undetermined to known source of ischemic stroke.

1. Introduction

Cryptogenic stroke is a heterogeneous entity defined as an ischemic stroke for which no probable cause is identified despite thorough diagnostic evaluation [1]. First conceptualized in the 1980s for research purposes in the National Institute of Neurological Disorders and Stroke (NINDS) Stroke Data Bank [2], cryptogenic stroke has since been increasingly addressed in clinical practice. While the exact definition may vary by stroke subtype classification system, the categorization from the Trial of ORG 10172 in Acute Stroke Treatment (TOAST) is commonly referenced, which qualifies cryptogenic stroke as those not attributable to a definite source of cardioembolism, large artery atherosclerosis, small artery disease, or other determined uncommon etiology despite standard evaluation, and may include strokes with incomplete evaluation or multiple plausible etiologies [3]. Based on this classification, a significant proportion (roughly 25–40%) of all ischemic strokes are considered cryptogenic [4–6]. Though over time providers have classified fewer ischemic strokes as of undetermined etiology due to increasingly advanced and extensive diagnostic evaluations [6], cryptogenic stroke remains a common challenge for neurologists as treatment is informed by identified stroke mechanisms.

Various etiologies for cryptogenic stroke have been proposed, which broadly fall into embolic or non-embolic causes (Figure 1). Given that many cryptogenic stroke patients from the NINDS Stroke Data Bank were later found to have potential cardioembolic sources and cryptogenic strokes often correspond to cortical infarcts on neuroimaging, which is suggestive of embolism, most putative mechanisms for cryptogenic stroke are embolic [7]. Of embolic mechanisms, occult cardiac sources represent the majority, which is indirectly supported by histopathologic evidence suggesting thrombi from cryptogenic strokes are similar in composition to those of known cardioembolic source [8]. Other embolic sources may be from undefined thrombophilia or substenotic cerebrovascular disease. Non-embolic causes may be related to in situ thrombosis, vasospasm, or impaired cerebral hemodynamics. In this review, we will discuss clinically relevant classification schemes, distinct mechanistic entities, and therapeutic implications in the current management of cryptogenic stroke.

Figure 1.

Schematic for proposed embolic and non-embolic mechanisms of cryptogenic stroke.

2. Evolution of Classification Schemes

Though TOAST was first widely adopted as a stroke classification system in the 1990s for its ease of use, additional schemes have developed along with our understanding of cryptogenic stroke (Table 1). These include modified TOAST criteria (SSS-TOAST), the Causative Classification System (CCS), and ASCO phenotyping [9–11]. These systems sought to address limitations of TOAST: modest interrater reliability, considering strokes with insufficient investigation as cryptogenic, and limiting strokes to only one assigned etiology when frequently patients have overlapping risk factors [12].

Table 1.

Ischemic stroke classification systems

Features	TOAST [3]	SSS-TOAST [9]	CCS [10]	ASCOD [11,13]
Publication year	1993	2005	2007	2009, updated 2013
Stroke etiologies	• Large-artery atherosclerosis • Cardioembolism • Small-vessel occlusion • Other determined etiology • Undetermined etiology	• Large artery atherosclerosis • Cardioaortic embolism • Small-artery occlusion • Other causes • Undetermined causes	• Large artery atherosclerosis • Cardioaortic embolism • Small-artery occlusion • Other causes • Undetermined causes	• Atherothrombosis (A) • Small vessel disease (S) • Cardioembolism (C) • Other causes (O) • Dissection (D)
Undetermined subgroups	• Two or more causes identified • Negative evaluation • Incomplete evaluation	• Cryptogenic embolism • Other cryptogenic • Incomplete evaluation • Unclassified (more than one probable mechanism or no probable evidence to establish a single cause)	• Cryptogenic embolism • Other cryptogenic • Incomplete evaluation • Unclassified (more than one probable mechanism or no probable evidence to establish a single cause)	• None
Estimated percent of ischemic strokes classified as cryptogenic	40–50 [3,14,122]	25–35 [9,122]	25–35 [14,122]	40* [14,122]
Interrater reliability (kappa)	0.54–0.75 [12]	0.76–0.96 [9]	0.81–0.91 [10]	0.48–0.88 [14]
Comments	• Easy to use • Widely used in clinical practice in some form • Does not clearly specify minimum diagnostic evaluation	• Structured algorithm to weight evidence as “evident”, “probable”, or “possible” • Identifies one causal mechanism but ignores interactions between multiple competing mechanisms	• Refined from SSS-TOAST • Automated web-based algorithm • Focuses on one causative mechanism • May best discriminate 90-day stroke recurrence	• Phenotype-based scheme • Retains all diagnostic information to assign likelihood to all potential causes • Only deemed undetermined etiology if all graded evidence is 0 (disease is absent) • More complex

^*estimation derived after converting phenotypic to causative classification with only grade 1 (highest) evidence considered causal

2.1. Modified TOAST and CCS

The SSS-TOAST criteria use the same five major stroke subtypes of TOAST, but each category is subdivided based on weight of evidence as “evident”, “probable”, or “possible” [9]. The classification system uses a three-step decision algorithm to determine stroke etiology: 1) a mechanism is deemed “evident” only if it is the sole potential mechanism confirming one of the causative categories, 2) if more than one “evident” stroke mechanism, the system uses specific stroke characteristics to assign one mechanism more “probable” than others, and 3) in absence of any “evident” cause, investigation is made for a “possible” cause before determining stroke of undetermined cause. The creators of SSS-TOAST demonstrated a similar, if not improved, interrater reliability compared to the original version [9]. Building upon SSS-TOAST, the same group developed CCS, which is further refined and automated to facilitate consistency across centers [10]. It is publicly available online at https://ccs.mgh.harvard.edu/ccs_title.php.

2.2. ASCO phenotyping

To better account for multiple concurrent possible stroke mechanisms and underlying risk factors, the ASCO system introduced a “complete stroke phenotyping” classification, in which every category of stroke etiology (A for atherosclerosis; S for small-vessel disease; C for cardiac pathology; and O for other causes) is graded based on level of evidence [11]. This phenotype-based scheme retains all diagnostic information and seeks to better describe overlapping diseases. It was later modified to ASCOD, with D representing dissection [13]. A stroke is only deemed completely unknown in etiology if all graded evidence is 0 (disease is absent). SSS-TOAST, CCS, and ASCO all reduce the amount of stroke classified as cryptogenic [9,14], but at the cost of complexity.

2.3. Embolic stroke of undetermined source

In 2014, as interest in cryptogenic stroke as a therapeutic target grew, Hart and colleagues introduced the term “embolic stroke of undetermined source” (ESUS), given supporting evidence that most cryptogenic strokes appear embolic and therefore may respond to anticoagulation [15]. They defined ESUS as a non-lacunar stroke detected on neuroimaging without significant (≥50%) proximal arterial stenosis, major-risk cardioembolic sources, or other specific causes of stroke identified. This construct has been the subject of recently completed and ongoing randomized controlled trials (RCT) comparing anticoagulant versus antiplatelet therapy for secondary stroke prevention after ESUS [16–18].

3. Overview of Diagnostic Evaluation and Therapeutics

3.1. Standard Diagnostic Evaluation

Cryptogenic stroke is a diagnosis of exclusion and, therefore, should only be diagnosed after ruling out major causative mechanisms. The American Heart Association (AHA) considers ischemic stroke to be cryptogenic after a minimum standard stroke evaluation, which involves basic laboratory tests, brain imaging, neurovascular imaging, and cardiac evaluation (Table 2) [19]. Routine laboratory tests will screen for myocardial ischemia, hematologic disorders, and traditional vascular risk factors (hyperlipidemia and diabetes mellitus). Structural brain imaging may be acquired by computed tomography (CT) or magnetic resonance imaging (MRI). While CT is more readily available, brain MRI has added sensitivity for detecting acute infarction, especially small or brainstem infarcts, and may provide helpful information on infarct topography [20]. For instance, acute infarct in multiple vascular territories may suggest a proximal embolic source, while multiple infarcts in the same vascular territory may point towards a culprit large artery lesion (Figure 2). Though small subcortical infarcts on neuroimaging with a corresponding lacunar stroke syndrome are generally considered to be etiologically small vessel disease, small deep infarcts may be cryptogenic in some patients; as a rule of thumb, small deep infarcts may be considered cryptogenic in patients aged <50 years without any traditional vascular risk factors and no neuroimaging features of small vessel disease [1].

Table 2.

Standard and further diagnostic evaluation of ischemic stroke

Standard diagnostic evaluation of ischemic stroke
Diagnostics	First-line tests
Laboratory tests	• Complete metabolic panel • Hemoglobin A1c, lipid profile • Complete blood count, INR, partial-thromboplastin time • Troponin T • Urine toxicology screen
Brain imaging	• Computed tomography (CT) of the brain • Magnetic resonance (MR) imaging of the brain
Neurovascular imaging	• CT angiography of the head and neck • MR angiography of the head and neck • Carotid ultrasound and/or transcranial Doppler if unable to obtain CT or MR angiography
Cardiac evaluation	• Electrocardiogram • Inpatient cardiac telemetry • Echocardiogram, often transthoracic
Further diagnostic evaluation of ischemic stroke (if cause still undetermined)
Diagnostics	Possible tests (guided by suspected cause(s))
Advanced vessel imaging	• Catheter angiography • High-resolution MR with vessel wall imaging
Advanced cardiac evaluation	• Prolonged outpatient cardiac telemetry or loop recording • Transesophageal echocardiogram • Cardiac CT or MRI
Biofluid testing	• Arterial hypercoagulability testing • Venous hypercoagulability testing • Serum inflammatory and/or autoimmunity markers • Cerebrospinal fluid evaluation • Occult cancer serum biomarkers • Genetic testing
Other	• Imaging for cancer screening • Brain biopsy

Figure 2. Neuroimaging clues in standard evaluation of cryptogenic stroke.

A-C) Axial diffusion-weighted images from brain magnetic resonance imaging of an older patient with embolic stroke of undetermined source. Subsequently, the patient was newly diagnosed with paroxysmal atrial fibrillation on prolonged cardiac monitoring. From superior (A) to inferior (C), the images show the main symptomatic left middle cerebral artery territory infarct (solid arrows) and multiple infarcts in different vascular territories (dashed arrows). D-F) Neuroimaging from a middle-aged patient with recurrent dominant hemisphere cryptogenic strokes, which were later attributed to substenotic carotid disease. Panel D shows acute left basal ganglia infarct (solid arrow) on diffusion weighted imaging. Panel E shows a chronic left frontal infarct (broken dashed arrow) on fluid attenuated inversion recovery sequence. Panel F shows a coronal view of a computed tomography angiogram of the neck demonstrating mild luminal stenosis and irregularity in the left internal carotid artery (circle).

First-line neurovascular imaging modalities used to detect extra- and intracranial vascular lesions include non-invasive techniques, unless urgent endovascular therapy is planned: CT angiography (CTA) of the head and neck, MR angiography (MRA) of the head and neck, or carotid ultrasound and transcranial Doppler when CT or MR is contraindicated or not available. Both CTA and MRA have generally good sensitivity and specificity for detection of significant extra- and intracranial artery stenoses but may have lower performance in substenotic lesions [21–23]. Routine cardiac evaluation after ischemic stroke involves structural assessment and cardiac monitoring. Often transthoracic echocardiography (TTE) is preferred as a first-line structural assessment as it is non-invasive, inexpensive, and readily available [24]. A 12-lead electrocardiogram (ECG), inpatient cardiac telemetry, and short-term ambulatory cardiac monitoring comprise routine cardiac monitoring after stroke [19].

3.2. Further Diagnostic Evaluation

If standard stroke evaluation does not reveal a clear stroke etiology, then the stroke may be considered cryptogenic and further advanced diagnostic investigation could be warranted. It should be noted many ischemic strokes of less common causes, especially among young patients without traditional vascular risk factors, may be provisionally deemed cryptogenic based on initial evaluation and then later reclassified as other determined cause after further diagnostic testing elucidates a cause. Because a wide variety of advanced testing exists, clinicians should carefully consider the patient’s demographics, comorbidities, clinical syndrome, and data from routine stroke evaluation in deciding relevance of additional testing that will change clinical management [19]. For instance, cortical infarcts in multiple vascular territories may suggest proximal embolic sources such as occult cardioembolism or hypercoagulable state to be more likely causes. Diagnostic algorithms devised by SSS-TOAST, CCS, or ASCO may help guide further testing, which will be led by the suspected stroke mechanism. These algorithms may also help determine which etiology is more likely if multiple possible causes exist. Since many cryptogenic strokes are thought to be embolic [15], advanced testing will often center on searching for occult cardiac sources and other embolic etiologies. This may include prolonged cardiac monitoring, advanced cardiac imaging like cardiac CT or MRI, advanced vascular imaging such as catheter angiography or vessel wall imaging, and hematologic testing for hypercoagulable states (Table 2) [1].

3.3. Overview of Therapeutics

Acute therapy for cryptogenic stroke is the same as for other types of ischemic stroke, but secondary stroke prevention can be challenging without an exact treatment target [19]. Besides optimizing lifestyle and modifiable vascular risk factors, antiplatelet therapy is generally recommended for most cases of cryptogenic stroke [19]. This is informed by multiple RCTs which compared anticoagulation to aspirin. Among cryptogenic stroke patients in the Warfarin-Aspirin Recurrent Stroke Study (WARSS), warfarin was not superior to aspirin in recurrent stroke or death rate, apart from small subgroups defined by absence of baseline hypertension and posterior circulation ischemia without brainstem infarction [25,26]. NAVIGATE ESUS (Rivaroxaban vs Aspirin in Secondary Prevention of Stroke and Prevention of Systematic Embolism in Patients with Recent ESUS) and RE-SPECT ESUS (Dabigatran Etexilate for Secondary Stroke Prevention in Patients with ESUS) compared rivaroxaban and dabigatran, respectively, to aspirin in patients with ESUS. Both found no advantage of direct oral anticoagulants (DOAC) over aspirin in recurrent stroke prevention, with potentially more bleeding events in the DOAC arms [16,17]. In a secondary analysis of NAVIGATE ESUS, rivaroxaban modestly reduced recurrent stroke in a small subset of ESUS patients with moderate to severe left atrial enlargement, but further confirmation is needed before being adopted in standard clinical practice [27]. In cases of minor stroke defined by National Institutes of Health Stroke Scale (NIHSS) ≤5 or high-risk transient ischemic attack (TIA) (ABCD² score ≥4), a 21-day course of dual antiplatelets followed by long-term single antiplatelet therapy should be considered [28,29]. Otherwise, single antiplatelet therapy is the current mainstay treatment for most cases of cryptogenic stroke until an etiology is found.

4. Occult atrial fibrillation

The paroxysmal nature of atrial fibrillation (AF) frequently allows it to evade detection on initial evaluation, thereby classifying many strokes as cryptogenic until a diagnosis of AF is made. A meta-analysis of 50 studies found that nearly a quarter of patients with ischemic stroke or TIA will have newly detected AF using sequential cardiac monitoring methods [30]. More broadly, atrial high-rate episodes (AHRE), which refers to implanted device detected asymptomatic atrial tachyarrhythmias in individuals with no prior AF, is about 20% at 1-year among those with cardiac devices. AHRE includes subclinical AF and both have a relationship with clinical AF and stroke [31].

Among older adults with hypertension, no prior history of AF, and recent pacemaker or defibrillator implantation, the prospective ASSERT (Asymptomatic Atrial Fibrillation and Stroke Evaluation in Pacemaker Patients and the Atrial Fibrillation Reduction Atrial Pacing Trial) study linked subclinical atrial tachyarrhythmia (atrial rate >190 beats per minute for more than 6 minutes) to ischemic stroke (hazard ratio (HR) 2.5 when compared to no detected atrial tachyarrhythmia, 95% confidence interval (CI): 1.25–5.08) [32]. Subsequent randomized trials confirmed that longer monitoring would yield more AF detection in cryptogenic stroke than standard monitoring for 24 hours (9% at 6 months and 30% at 3 years) [33,34]. Further, older age, a higher CHA₂DS₂-VASc score, cortical infarct topography, and evidence of structural atrial disease such as dilatation, appendage size and morphology, p-wave dispersion on ECG, and elevated N-terminal pro-brain natriuretic peptide (NT-proBNP) serum levels, may be patient selection factors that increase the likelihood of detecting occult AF with prolonged cardiac monitoring after cryptogenic stroke [34–39].

However, the overall benefit of prolonged cardiac monitoring in cryptogenic stroke remains unclear due to questions about the causal nature between occult AF and cryptogenic stroke, uncertainty when detection of occult AF would warrant anticoagulation over antiplatelet therapy, and the unknown ideal duration of prolonged cardiac monitoring. Among ASSERT study participants who had least three months of continuous cardiac monitoring and suffered an incident ischemic stroke or systemic embolic event (n=51), only 8% had occult AF detected within 30 days preceding the event, suggesting a lack of temporal association and thus no definitive evidence of causality [32]. Further, even if found, there is limited data to suggest that occult AF would be best treated with anticoagulation or antiplatelet therapy. In the LOOP trial, which included participants aged ≥70 years with no known AF and ≥1 stroke risk factor, an implantable loop recorder compared to no monitoring did not demonstrate a significant reduction in future stroke risk [38]. Participants did not necessarily have cryptogenic stroke at baseline, but the prolonged monitoring group had triple the amount of AF detection and anticoagulation initiation than usual care, implying not all occult AF may merit anticoagulation [38]. Moreover, the optimal duration of prolonged monitoring, programmable device threshold to detect AHREs, and the temporal burden of occult AF detected that would benefit from treatment remain unknown. Other device detected rhythms such as premature atrial contractions may also be related to AF [27], but their exact role in stroke is not clear.

Nonetheless, even if there is no established lower limit of stroke risk from dysrhythmia burden, it is likely that occult AF will progress in duration to a point of similar stroke risk as overt AF, especially as patients age [39]. Over 80% of North American providers will initiate anticoagulation after discovering occult AF, according to two randomized trials [33,34]. Both the AHA and the American Academy of Neurology (AAN) have recommended considering prolonged cardiac monitoring in cryptogenic stroke patients and offering anticoagulation if occult AF is found regardless of the duration [19,40]. The ongoing ARTESiA (Apixaban for the Reduction of Thrombo-Embolism in Patients with Device-Detected Sub-Clinical Atrial Fibrillation; NCT01938248) and NOAH (Non-vitamin K Antagonist Oral Anticoagulants in Patients with Atrial High Rate Episodes; NCT02618577) trials may shed light on these therapeutic uncertainties [41,42].

5. Atrial Cardiopathy

Atrial cardiopathy refers to structural and functional abnormalities in the atria that may precede any dysrhythmia and can occur with or without AF. The biological reasoning that AF causes embolic stroke lies in the nineteenth century paradigm of Virchow’s triad, in which a fibrillating atrium leads to blood stasis that begets clot formation. However, a growing body of evidence suggests the relationship is more complex [43]. First is the issue of temporality; the ASSERT study, for instance, showed only a small minority of participants with incident embolic events had preceding atrial tachyarrhythmia [32]. Second, if dysrhythmia is the cause of stroke through blood stasis, then one would expect rhythm control to mitigate stroke risk. However, a meta-analysis of 8 RCTs has shown that successful rhythm control had no such benefit [44]. Third, biomarkers of atrial dysfunction are increasingly associated with increased risk of stroke, even in the absence of AF [45–50]. Together, these lines of evidence have been used to argue that the relationship between AF and stroke does not clearly satisfy Bradford Hill’s epidemiologic criteria for causation [51]. Rather, AF may be another sign of atrial cardiopathy, which is itself the underlying cause for stroke.

The evidence linking biomarkers of atrial dysfunction and stroke highlight the dissociation between AF and stroke and serve as a basis for conceptualizing atrial cardiopathy. Multiple cohorts, including the Framingham Heart Study and the Northern Manhattan Study (NOMAS), have demonstrated left atrial enlargement on echocardiogram is associated with ischemic stroke risk independent of AF [45,46]. Studies using transesophageal echocardiogram (TEE) or advanced cardiac imaging have also implicated certain left atrial appendage (LAA) morphologies in AF to have variable relationships with stroke, suggesting dysrhythmia may not be the sole pathway to stroke [52]. Electrocardiographically, ECG markers like P-wave terminal force in lead V1, frequent premature atrial contractions, and paroxysmal supraventricular tachycardia have also been linked to increased non-lacunar stroke risk in the absence of overt AF [46–48]. Similarly, independent of AF, serum biomarkers of cardiac disease like NT-proBNP may also be related to cardioembolic stroke [50].

Though the relatively new concept of atrial cardiopathy is evolving, it has provided an initial framework to study a subset of cryptogenic stroke patients who otherwise may not have a clear diagnostic or therapeutic direction, especially given the difficulty in identifying AF. Preliminary studies have shown some atrial cardiopathy biomarkers may be common among the cryptogenic stroke population [53], and that some combination of them might be used to identify a subset of stroke patients [54]. Since many of these biomarkers are easier to obtain than prolonged cardiac monitoring, it may simplify the diagnostic evaluation in this population after further validation studies.

Lastly, if atrial cardiopathy is closely related to AF, a major indication for anticoagulation, then atrial cardiopathy would be a plausible intervenable target for anticoagulation as well. Though the WARSS trial did not show a benefit of warfarin over aspirin in secondary stroke prevention among cryptogenic strokes, post-hoc analysis has shown patients with high NT-proBNP may have reduced risk of stroke or death at 2 years with warfarin when compared to aspirin. Both NAVIGATE ESUS and RE-SPECT ESUS were negative trials comparing DOACs to aspirin in cryptogenic stroke, with a subgroup analysis in NAVIGATE ESUS showing potential benefit of rivaroxaban among ESUS participants with left atrial enlargement, but they did not specifically select based on biomarkers for atrial cardiopathy [16,17].

The ongoing ARCADIA (Atrial Cardiopathy and Antithrombotic Drugs in Prevention after Cryptogenic Stroke; NCT0392215) trial is comparing apixaban to aspirin in secondary stroke prevention in cryptogenic stroke and specifically atrial cardiopathy, based on thresholds of P-wave terminal force, NT-proBNP, or left atrial diameter [18]. Similarly, ATTICUS (Apixaban for Treatment of Embolic Stroke of Undetermined Source; NCT0242716) compared apixaban to aspirin in the prevention of new ischemic lesions on neuroimaging among ESUS patients with at least one risk factor for AF/cardiac thromboembolism (increased left atrial size, spontaneous echo contrast in left atrial appendage, low appendage flow velocities, atrial high-rate episodes, high CHA₂DS₂-VASc score, or patent foramen ovale). The ATTICUS trial’s enrollment was recently stopped after interim analysis showed futility but final analyses are forthcoming [55].

6. Atrial Septal Abnormalities

Atrial septal abnormalities, such as patent foramen ovale (PFO), atrial septal defect (ASD), and atrial septal aneurysm (ASA), are associated with ischemic stroke [56]. PFO and ASD are speculated to be a source of paradoxical embolism in which venous and right sided thromboemboli enter the systemic arterial circulation through a right-to-left shunt [57]. Shunting can occur transiently in individuals without any right heart pathology, both during Valsava maneuvers and at rest [56]. ASA, defined as redundant, localized outpouching of the fossa ovalis region of the atrial septum that protrudes at least 10–15 mm beyond the plane of the atrium, is thought to be related to paradoxical embolism given its association with PFO and ASD or itself a cardioembolic source due to in situ platelet aggregation [58].

Cryptogenic stroke patients have increased prevalence of atrial septal abnormalities [59]. However, they are also common in the general population. Notably, PFO has an estimated prevalence of 25% in the general adult community [56]; therefore, other potential high-risk sources must be properly investigated before attributing stroke to atrial septal abnormalities. The Risk of Paradoxical Embolism (RoPE) score is a 10-point system that uses clinical and imaging characteristics to estimate the probability a PFO is the likely cause of cryptogenic stroke. High RoPE scores found in younger patients with no vascular risk factors and a cortical infarct convey a higher risk PFO, while a lower RoPE score among older patients with traditional risk factors suggests an incidental PFO [60]. Other anatomic features of the atrial septal abnormality such as shunt size, presence or absence of ASA, and venous thromboembolism are included in the PASCAL classification system to categorize the likelihood a stroke is caused by PFO [61].

To identify an atrial septal abnormality, TTE, TEE, and transcranial Doppler (TCD) are most used in clinical practice. TCD with agitated saline is the most sensitive diagnostic evaluation for right-to-left shunting, but it is unable to differentiate intracardiac shunting from other sources of shunting, such as pulmonary arteriovenous malformations [62], and cannot rule out other cardioembolic causes [1,63]. TTE with agitated saline is often used as the first-line modality given its accessibility and tolerability, but TEE is more sensitive in the detection of PFO and allows adequate visualization of relevant anatomic features [64], as well as other potential causes of embolism such as fibroelastoma or myxoma Thus, in patients with high suspicion for PFO-associated ischemic stroke, TEE is often performed as the initial study or ultimately in follow-up from TTE or TCD.

Unlike other cryptogenic stroke mechanisms which lack conclusive data for therapeutics beyond antiplatelet therapy, PFO-associated stroke may be additionally treated with percutaneous PFO closure [19]. In the PFO in Cryptogenic Stroke Study (PICSS), which was conducted within WARSS, investigators compared warfarin to aspirin in the medical treatment of PFO-associated stroke, finding no difference in recurrent ischemic stroke [65]. In the early 2010s, the first three RCTs comparing recurrent stroke outcomes with PFO closure via percutaneous device against antiplatelet therapy alone for PFO-associated stroke did not show a significant benefit of closure by intention-to-treat analyses [66–68], but three subsequent trials at the end of the decade found a clear benefit of PFO closure [69–71]. The three later trials (RESPECT extended follow-up, REDUCE, and CLOSE) had more stringent patient selection factors such as requiring neuroimaging confirmation of infarct and high-risk PFO anatomic features, which may have led to the demonstrated benefit of PFO closure. In 2021, a meta-analysis of individual level data (n=3740) from 6 of these trials concluded PFO closure was associated with a relatively lower risk of recurrent stroke when compared to medical therapy alone (0.47 vs 1.09%, adjusted HR 0.41, 95% CI: 0.28–0.60). Further, the PASCAL classification system may provide superior patient selection for PFO closure than the RoPE score alone [72].

Accordingly, both the AHA and AAN recommend interdisciplinary neurology and cardiology consideration of PFO closure with antiplatelet therapy over antiplatelet therapy alone in patients 18–60 years of age with ESUS and PFO with high-risk anatomic features [19,73]. Though there is a risk of periprocedural complications and future AF, the data would suggest the overall benefits of PFO closure outweigh the risks in that subpopulation [72]. In those with PFO without high-risk anatomic features or patients older than 60 years, the benefit of PFO closure is not certain [19]. In other isolated atrial septal abnormalities, there continues to be uncertainty regarding optimal secondary stroke prevention beyond antiplatelet therapy.

7. Heart Failure

In the absence of comorbid high-risk stroke mechanisms, such as AF or recent myocardial infarction, congestive heart failure (HF) with reduced left ventricular ejection fraction (LVEF) (<30%) is considered by stroke classification systems to be of low or uncertain risk of stroke, and therefore a possible cryptogenic stroke mechanism [9–11]. Though HF is prevalent among the general population and an estimated 25% of all ischemic stroke patients have some degree of left ventricular systolic dysfunction [74,75], data on stroke risk specific to HF is limited due to heterogeneous study designs and difficulty in separating out commonly concurrent high-risk mechanisms such as AF [76]. In the Framingham Heart Study, stroke risk was 3–4% per year for adults with HF, although many participants also had atrial dysrhythmia [77]. Post-hoc analyses from two RCTs have estimated the stroke risk among HF patients without AF to be 1.1% per year, compared to 1.7% among HF patients with AF [78]. HF with preserved ejection fraction is even less certain as a stroke etiology, but thought to be of similar stroke risk to HF with reduced LVEF [79]. Therefore, the stroke risk in HF is likely elevated, but the precise extent is not known.

Further, stroke prevention in HF with sinus rhythm is also an area of uncertainty. Like atrial dysrhythmia, HF is thought to confer embolic risk due to relative stasis [80]. In reduced LVEF, this may be rationalized through a low flow state, but the pathophysiologic mechanism is less clear in HF with preserved ejection fraction [81]. Accordingly, anticoagulants may prevent thromboembolic events in some forms of HF. Two large RCTs, the Warfarin and Antiplatelet Therapy in Chronic Heart failure (WATCH) trial and Warfarin versus Aspirin Treatment in the Reduced Cardiac Ejection Fraction (WARCEF) study, compared warfarin vs antiplatelet therapy in prevention of composite events, including ischemic stroke, among HF patients with reduced LVEF (≤35%) in sinus rhythm. Both found the warfarin arms had less incident ischemic stroke, but at the cost of significantly higher major hemorrhagic complications [82,83]. With the rise of DOACs, which have more favorable bleeding risks, the COMMANDER HF (A Study to Assess the Effectiveness and Safety of Rivaroxaban in Reducing the Risk of Death, Myocardial Infarction or Stroke in Participants with HF and Coronary Artery Disease Following an Episode of Decompensated HF) trial compared rivaroxaban 2.5 mg twice daily against placebo in safety and composite efficacy outcome of death, myocardial infarction, and ischemic stroke among participants with worsening of chronic HF with LVEF ≤40%, coronary artery disease, and no AF. The trial found no difference in the composite outcome, but a slight benefit in ischemic stroke within the DOAC arm compared to placebo (2 vs 3%; HR 0.66, 95% CI: 0.47–0.96) without significantly elevated bleeding risk [84,85]. Further study is needed to confirm these findings. In summary, for secondary prevention after cryptogenic stroke with chronic HF and no other strong indication for anticoagulation, there is still clinical equipoise in optimal management between antiplatelet and anticoagulant therapy.

8. Substenotic Atherosclerosis

Current criteria for large artery atherosclerosis to qualify as a major causative stroke mechanism require luminal stenosis of 50% or greater [3,9,10]. However, numerous studies have shown substenotic (<50%) lesions to be more common ipsilateral to cerebral infarct of undetermined source than the contralateral unaffected side, implicating both extracranial and intracranial large artery substenotic atherosclerosis in ischemic stroke [86–88]. Speculatively, in the extracranial carotids, vulnerable plaques may rupture and cause artery-to-artery embolism even in the absence of significant luminal stenosis. Plaque ulceration, intraplaque hemorrhage, fibrous cap rupture, and lipid-rich cores may indicate vulnerable carotid plaques at high-risk for ischemic stroke [89]. Many of these features can be identified on standard stroke neuroimaging [87]. About a quarter of ischemic strokes harbor ipsilateral substenotic carotid atherosclerosis and, in some populations, high-risk substenotic carotid plaques may be more prevalent than those with ≥50% stenosis [90]. Therefore, substenotic carotid atherosclerosis is likely an underrecognized potential cause of ischemic stroke.

Similarly, some proportion of strokes from large-artery intracranial atherosclerosis may not be appreciated because of substenotic plaque [88]. In NOMAS and the Atherosclerosis Risk in Communities study, about 30–40% of the general older population have intracranial artery stenosis, most of which would be classified as substenotic [91,92]. Autopsy-based studies have shown some cryptogenic ischemic infarcts to be associated with substenotic intracranial atherosclerotic plaques with superimposed thrombi [92]. Putative mechanisms for stroke in substenotic intracranial atherosclerosis include artery-to-artery embolism, but also branch artery disease when plaque extends into the ostium of small perforators [94]. An emerging area of research leverages recent advances in neuroimaging modalities like high-resolution MRI with vessel wall imaging to better identify high-risk plaque characteristics beyond luminal stenosis [95].

Therapeutically, surgical revascularization of substenotic carotid stenosis has not been demonstrated to be of benefit [96]. In the Warfarin-Aspirin Symptomatic Intracranial Disease (WASID) trial, which did include substenotic intracranial atherosclerosis, anticoagulation was not superior to aspirin in secondary stroke prevention in patients with intracranial atherosclerosis [97]. Therefore, the contemporary mainstay of treatment in both substenotic extra- and intracranial atherosclerosis remains antiplatelet agents and optimization of vascular risk factors [19].

9. Aortic Atheroma

Aortic atherosclerotic plaques or atheroma are a putative source of systemic embolism and ischemic stroke. Plausibly, a proportion of ischemic stroke may be caused by aortic atheroma given its estimated prevalence of about 30% among older community-dwelling adults [98]. It is thought that complex aortic atheromas may become unstable and form superimposed lipid-rich thrombi prone to embolization [99].

Longitudinal population-based studies have had conflicting results regarding the exact relationship between aortic atheroma and cryptogenic stroke due to methodological differences [100]. In the Stroke Prevention: Assessment of Risk in the Community (SPARC) study of over 1100 participants who had TEE, complex aortic atheroma (>4mm in thickness with or without mobile debris) was not associated with cryptogenic stroke, but rather thought to be a marker generalized atherosclerosis [101]. Conversely, in a prospective case-control study by Amarenco and colleagues, those with aortic atheroma ≥4 mm identified by TEE had a 9-fold increase in the odds for ischemic stroke overall (95% CI: 3.3–25.2) and about 5-times increased odds for cryptogenic infarcts (95% CI: 2.2–10.1), independent of atherosclerotic risk factors [102]. Other case-control studies have also reported thick or protruding aortic plaque, ulceration, or mobile components to be associated with ischemic stroke [103,104]. Together, it is difficult to tease out if aortic atheroma is a bystander index of atherosclerotic disease or the culprit lesion for stroke. There is likely a combination of both in the community.

In a subgroup analysis of those with complex aortic atheroma in the Stroke Prevention in Atrial Fibrillation-III (SPAF-III) trial, which was designed for stroke prevention in AF with at least one thromboembolic risk factor, patients treated with therapeutic warfarin (median INR 2.3) still had relatively higher risk of thromboembolic events than those without aortic atheroma (16 vs 4% per year, p=0.02), suggesting other approaches are needed [105]. The randomized, open-label Aortic Arch Related Cerebral Hazard (ARCH) trial compared dual antiplatelet therapy to warfarin in adults with non-disabling ischemic stroke with aortic atheromas ≥4 mm, finding the dual antiplatelet group to have non-significantly less future vascular events (HR 0.76, 95% CI: 0.36–1.61). However, the trial was stopped early due to prolonged recruitment time, so the results were inconclusive [106]. Surgical or endovascular procedures such as aortic atherectomy should be considered experimental approaches given lack of data [107]. Therefore, secondary stroke prevention among those with aortic atheromas remains the same as general guidelines for atherosclerotic stroke [19].

10. Occult Malignancy

Active, diagnosed cancer is an established risk factor for ischemic stroke, and may be due to both arterial and venous malignancy-associated hypercoagulability [108–110]. Conversely, it is becoming increasingly appreciated that ESUS can be the initial presentation of undiagnosed cancer [110]. Using Medicare-linked data from a large American cancer registry, Navi and coauthors found the risk of an arterial thromboembolic event was increased by 60–70% within 1 year before cancer diagnosis and became progressively greater within 5 months preceding diagnosis [111]. Therefore, a proportion of cryptogenic strokes is likely due to occult malignancy, though the exact rate remains unknown.

Further, it is unclear if and how cryptogenic stroke patients should be screened for occult malignancy. Retrospective observational studies have shown historical clues such as weight loss, infarcts in multiple vascular territories or the so called “three-territory sign” (i.e., presence of infarcts in posterior and both anterior circulation territories), TCD microemboli, and elevated serum markers (D-dimer, fibrinogen, inflammatory markers) are all associated with occult malignancy in cryptogenic stroke [112–114]. Often clinicians obtain whole body CT scans to screen for occult malignancy during cryptogenic stroke evaluation [1]. However, the SOME (Screening for Occult Malignancy in Patients with Idiopathic Venous Thromboembolism) trial demonstrated less than 4% of patients with an unprovoked deep venous thrombosis were later newly diagnosed with cancer and comprehensive CT scans did not significantly improve diagnostic yield compared to routine evaluation and age-appropriate cancer screening [115]. Extrapolating from these data on venous thromboembolism, the yield of extensive imaging to search for malignancy in cryptogenic stroke patients is likely low. The yield of TEE in visualizing nonbacterial thrombotic endocarditis, which is a potential mechanism for malignancy-associated hypercoagulability to cause stroke, may be higher than TTE but currently there is no consensus on optimal diagnostic imaging in this subpopulation of cryptogenic stroke patients [116].

For secondary stroke prevention in cancer patients, there have been no completed studies directly comparing treatment options, and society guidelines do not make any firm treatment recommendations in this subgroup of patients who would otherwise have no strong indication for anticoagulation [19]. Data for secondary stroke prevention in cancer patients is limited to post-hoc subgroup analyses of DOACs in AF trials, which showed comparable efficacy to warfarin in rates of stroke and systemic embolism [117,118]. Frequently, anticoagulation may be empirically started for suspected cancer-associated hypercoagulability, but this is based on extrapolation from trials on venous thromboembolism in malignancy [119]. Traditionally, low-molecular weight heparin is used due its longstanding history of efficacy, but more recently DOACs have been shown to be equally or more efficacious in preventing recurrent venous thromboembolism, with similar or slightly increased major bleeding rates when compared with low-molecular weight heparin, such that oncology guidelines now consider DOACs as first-line treatment options as an alternative to low-molecular weight heparin [119,120]. In the case of ESUS with possible occult malignancy, a particular area of concern is possible increased intracranial hemorrhage risk associated with cancer. The recently completed TEACH pilot trial tested enoxaparin vs aspirin in patients with cancer and ischemic stroke, but it had difficulty with enrollment as 40% of participants randomized to enoxaparin crossed over to aspirin given discomfort with enoxaparin injections [121]. Larger blinded clinical trials are being planned.

11. Prognosis and recovery after cryptogenic stroke

Cryptogenic strokes generally have better prognosis than other defined stroke subtypes. Using TOAST, CCS, and ASCO criteria, Arsava and co-investigators found ischemic strokes without determined etiology had a lower risk of 90-day stroke recurrence and mortality than large-artery atherosclerosis and cardioembolism, but not small vessel disease, in all three stroke classifications schemes [122]. Longer term recurrent stroke risk may be about 20% at 2 years in cryptogenic stroke, though this estimate is based on older data before increasingly advanced diagnostic evaluation and improvements in medical therapy [123]. Similarly, cryptogenic stroke has equal or less stroke severity in terms of NIHSS and infarct volume than strokes due to large-artery atherosclerosis or cardioembolism [122]. Logically, this follows the reasoning that upfront stroke evaluation would rule out known high-risk mechanisms before deeming a stroke cryptogenic. Since cryptogenic stroke is heterogeneous, the recurrent event risk likely also differs by suspected mechanism. For example, ipsilateral substenotic carotid atherosclerosis may increase the risk of recurrent event in cryptogenic stroke, consistent with the known high-risk nature of symptomatic carotid stenosis [124].

An estimated half of cryptogenic stroke patients have favorable long-term functional outcomes of modified Rankin scale of 2 or better [122,124]. Like other subtypes of ischemic stroke, rehabilitation will be an important component of recovery that often is underappreciated. Unlike acute stroke treatment, the exact subtype of ischemic stroke seems to have little bearing on the effect of motor recovery or outcome with rehabilitation [125].

12. Conclusion

Cryptogenic stroke, a diagnosis of exclusion, is a frequently encountered clinical conundrum for healthcare providers. The current mainstay treatment for secondary stroke prevention in the majority of cryptogenic stroke is optimization of modifiable vascular risk factors and antiplatelet therapy, with a notable exception for potential endovascular PFO closure in PFO-associated stroke [19]. Beyond this, the absence of definite etiology in cryptogenic stroke creates uncertainty in optimal treatment because stroke prevention is informed by mechanism. Therefore, the clinical impetus rests in further diagnostic evaluation that would uncover potentially treatable causes of ischemic stroke. With a widely variable set of advanced diagnostic tools in an otherwise nonspecific clinical entity, appropriateness of further testing must be carefully tailored to the patient and clinical scenario. Newer constructs within cryptogenic stroke such as ESUS and atrial cardiopathy may solidify with further study into clear therapeutic targets. To continue significant progress since conceptualization in the 1980s NINDS Stroke Data Bank, much work is still needed to shed light on the underlying causes of cryptogenic stroke.

13. Expert Opinion

Cryptogenic stroke is an evolving concept that changes with research. As we have discussed, the common endpoint in the multiple avenues of research in cryptogenic stroke is to reduce the number of ischemic strokes labeled as cryptogenic such that an optimal treatment modality may ultimately be identified. In part, this will require developing targetable constructs, such as atrial cardiopathy, within the nonspecific boundaries of cryptogenic stroke. Beyond ongoing interventional trials (Table 3), new applications of technology to advance diagnostic capabilities in stroke is an exciting facet of research that may help better define subsets of cryptogenic stroke.

Table 3.

Summary of ongoing interventional therapeutic trials in cryptogenic stroke.

Mechanism	Trial Name	Inclusion Criteria	Treatment Arms	Primary Outcome(s)	Estimated Study Completion Date
Occult atrial fibrillation	ARTESiA41	Adults aged ≥55 years with prior stroke/systemic embolism and implanted device detected sub-clinical atrial fibrillation ≥6 minutes in duration but no episode >24 hours	Apixaban 5mg (or 2.5mg) twice daily vs aspirin 81mg daily	• Composite efficacy of ischemic stroke and systemic embolism • Major bleeding	December 2023
	NOAH42	Stroke free adults aged ≥65 years with implanted device detected atrial high rate episodes (≥170 bpm atrial rate and ≥6 minutes in duration)	Edoxaban 60mg (or 30mg) daily vs aspiring 100mg daily	• Time to first ischemic stroke, systemic embolism or cardiovascular death	December 2023
Atrial cardiopathy	ARCADIA18	Adults aged ≥45 years with recent ESUS and evidence of atrial cardiopathy (≥1 of P-wave terminal force >5000 μV × ms in ECG lead V1, serum NT-proBNP > 250 pg/mL, and left atrial diameter index ≥3 cm/m2 on echocardiogram)	Apixaban 5mg (or 2.5mg) twice daily vs aspirin 81mg daily	• Incidence of recurrent stroke	June 2024
	ATTICUS55	Adults aged ≥18 years with recent ESUS, implanted cardiac monitoring prior to study, and at least one non-major risk factor for cardiac embolism (left atrial size >45mm, spontaneous echo contrast in left atrial appendage (LAA), LAA flow velocity ≤0.2 m/s, atrial high rate episodes, patent foramen ovale, CHA2DS2-VASc score ≥4)	Apixaban 5mg (or 2.5mg) twice daily vs aspirin 100mg daily	• Occurrence of at least one new ischemic lesion on brain magnetic resonance imaging at 12 months when compared to baseline imaging	Enrollment ended September 2021, final analysis forthcoming

Technology as a means to advance understanding of cryptogenic stroke is evident in the case of occult cardioembolic sources, which arguably have had a large share of attention. Improvements in prolonged cardiac monitoring over the last decade have made it possible to capture occult arrhythmias long suspected of being related to cryptogenic stroke in a more feasible and unobtrusive manner to patients [1]. Advanced cardiac imaging techniques such as cardiac MRI with flow dynamics have helped identify left atrial fibrosis and flow patterns as predictive of ischemic stroke even before AF, pushing the envelope with atrial cardiopathy as a construct within cryptogenic stroke [43]. Improvements in older technologies also indispensably contribute to our understanding of cryptogenic stroke. For instance, though practical use of TEE was first described in the 1970s, improvements in a more flexible endoscopic probe, higher resolution images, and 3-dimensional visualization have allowed for more robust study of factors potentially relevant in cryptogenic stroke, such as left atrial appendage morphology [54,57,126].

Looking ahead, molecular cell biology and specialized imaging may be two areas that particularly hold promise. Several studies have examined cellular and histological clot composition of large vessel occlusions extracted during endovascular thrombectomy. Quantifying the relative proportion of red blood cells, fibrin, platelets, and leukocytes may provide clues on the source of the thromboembolism in cryptogenic stroke [127,128]. In a similar thread, emerging data have suggested RNA may serve as a precision biomarker in the determination of stroke etiology [129–131]. Various circulating cells that may contribute to thromboembolism formation in the peripheral blood can express different RNA signatures and these differences have been preliminarily applied to classify probable etiology of cryptogenic stroke in conjunction with neuroimaging [132]. On the imaging front, high-resolution MRI with vessel wall imaging has become increasingly popular as an adjunctive tool in the assessment of extra- and intracranial arterial stenosis as it offers information beyond degree of luminal stenosis like markers of plaque vulnerability [133]. Advanced ultrasound techniques such as elastography may provide an alternative and cheaper avenue to assessing plaque vulnerability [134]. Together, these promising diagnostic modalities would need validation of their findings with pathology and demonstration of feasibility in routine clinical practice.

As the field gains further insight into cryptogenic stroke, a fundamental question is raised: at what point does a putative cause of cryptogenic stroke stop being cryptogenic? Since its definition hinges on the absence of findings, data, and therapeutic clarity, then one approach would be to consider when a cause goes from characterization marked by absence to presence. As Allan Ropper writes on PFO closure, within a decade, the flip from negative to positive results on endovascular PFO-closure for PFO-associated stroke shifted the viewpoint of PFO-associated stroke from exclusionary features to positive characteristics of patient selection factors such that one could arguably consider PFO-associated stroke to no longer be cryptogenic [135]. With continued research, we would hope the same evolution can be applied to other groups within cryptogenic stroke.

Article Highlights:

• Cryptogenic stroke is a heterogeneous entity defined as an ischemic stroke for which no probable cause is identified despite thorough diagnostic evaluation; it accounts for about 25% of all ischemic stroke.

• Cryptogenic stroke is a diagnosis of exclusion after standard stroke evaluation fails to yield a probable cause.

• Advanced diagnostic testing should be carefully geared towards the patient and clinical scenario. If a cause is discovered, then the stroke ceases to be cryptogenic.

• The current mainstay secondary prevention for cryptogenic stroke is optimization of modifiable vascular risk factors and antiplatelet therapy, with the notable exception of possible endovascular patent foramen ovale (PFO) closure in PFO-associated stroke.

• Most proposed cryptogenic stroke mechanisms are embolic, thus many interventional trials have sought to test the efficacy of anticoagulation over antiplatelet therapy in cryptogenic stroke, though conclusive benefit has yet to be seen.