Detection of Atrial Fibrillation in Cryptogenic Stroke
November 30, 2025
By Alan Z. Segal, MD
Synopsis: The ANTARCTICA study is a pooled meta-analysis from multiple prospective studies of patients with ischemic stroke of varying etiologies, where loop monitoring was performed for atrial fibrillation detection. The groups were divided into cryptogenic stroke/transient ischemic attack or non-cryptogenic stroke/non-stroke. Both groups demonstrated an unadjusted rate of atrial fibrillation of about 30%.
Source: Yaghi S, Sposato LA, Shu L, et al. Device-detected atrial fibrillation in patients with and without cryptogenic ischemia: The ANTARCTICA pooled analysis. Stroke. 2025;56(10):2895-2903.
Atrial fibrillation (Afib) is a major risk factor for stroke. Based on the commonly accepted CHA2DS2-VASc score, patients with Afib and multiple risk factors may have a 15% yearly incidence of stroke if not treated with anticoagulation. Because Afib often can be asymptomatic, detection of Afib is a crucial tool in the primary prevention of stroke.
Among patients who already have had a cerebrovascular event (stroke or transient ischemic attack [TIA]), monitoring for Afib is a mainstay of the workup to determine stroke etiology and optimize secondary prevention. Supplementing in-hospital telemetry and Holter monitors, most patients receive mobile cardiac outpatient telemetry (MCOT) devices as standard of care. Depending on patient compliance, MCOT monitoring typically is 30 days, but longer-term monitoring, such as with an implantable loop recorder (ILR), may increase the yield for Afib detection.
The 2014 Cryptogenic Stroke and Underlying Atrial Fibrillation (Crystal AF) study established that ILRs can significantly increase rates of Afib detection.1 In Crystal AF, Afib in the ILR group was found in 8.9%, 12.4%, and 30% of subjects at six, 12, and 30 months, respectively, compared to 1.4%, 3%, and 3% among controls receiving standard follow-up. The implantable loop record detection of atrial fibrillation to prevent stroke (LOOP) study attempted to translate these promising observational results into a therapeutic trial.2 Over a five-year period, ILRs in LOOP detected Afib in 32% of strokes as compared to 12% receiving usual care.
After being diagnosed with Afib, the majority of patients in LOOP were initiated on direct oral anticoagulants (DOACs), but the study failed to achieve its primary outcome of stroke prevention. There were recurrent strokes in 4.5% of patients monitored with ILRs compared to 6.5% without them, a weak trend that was not statistically significant (P = 0.11). These low rates with or without an ILR are typical of stroke therapeutic trials, which historically have been hampered by unexpectedly infrequent rates of index events.
ILRs are used widely among stroke patients regardless of mechanism, since a diagnosis of Afib is not exclusive to patients with cryptogenic strokes. Afib also is found frequently in strokes of non-cardioembolic etiology. The Stroke of Known Cause and Underlying Atrial Fibrillation (STROKE-AF) study showed that Afib could be found in patients with strokes from large or small vessel atherosclerotic disease in 12% at one year and 21.7% at three years, compared to only 2.4% after three years without an ILR.
Because Afib in these patients is not thought to be the stroke mechanism, questions have been raised about what proportion of ILR-detected Afib is pathogenic rather than merely incidental, unrelated to the stroke. The ANTARCTICA study aimed to address this question by dichotomizing subjects into two groups, one with cryptogenic stroke or TIA (C-IS/TIA) and the other with either non-cryptogenic stroke or no prior history of stroke.
ANTARCTICA was a meta-analysis, pooling individual-participant data from two randomized studies and 12 prospective studies. There were 1,562 subjects with C-IS/TIA and 474 non-cryptogenic/non-stroke subjects. The unadjusted proportion of Afib detection in C-IS/TIA vs. non-cryptogenic/non-stroke subjects was 30% in both groups. Using a multivariable prediction model, C-IS/TIA subjects were more likely to have Afib compared to controls (odds ratio [OR], 1.90). The authors applied Bayes theorem to calculate the percentage of ILR-detected Afib that may be pathogenic in C-IS/TIA: (1 − 1/1.9) × 100 = 47%. Conversely, according to ANTARTICA, 53% of ILR-detected Afib occurs incidentally.
Eleven parameters were entered into the multiple logistic regression model: age, sex, history of hypertension (HTN), diabetes mellitus (DM), coronary artery disease (CAD), chronic heart failure (CHF), prior stroke or TIA, duration of cardiac monitoring, and echocardiogram findings of left atrial enlargement (scored according to left atrial volume index [LAVI] or left atrial diameter). Not unexpectedly, factors such as left atrial enlargement were seen more frequently in C-IS/TIA, but factors favoring non-embolic etiologies were less consistent. For example, DM was less common in C-IS/TIA as compared to stroke-only subjects (omitting the non-stroke subjects), but there were no differences in HTN between the two groups.
The time to Afib detection was significantly shorter in the C-IS/TIA group (65 days compared to 169 in the non-C-IS/TIA group) and the duration of the longest run of Afib was 120 minutes in C-IS/TIA compared to 90 minutes in non-C-IS/TIA subjects (although this did not reach statistical significance). The authors suggest that these findings indicate that when Afib occurs later and more briefly, it would have less of a causal relationship to the index stroke.
Commentary
The value of this study hinges on the accuracy of stroke subtyping, which is notoriously variable. Classification of strokes according to the commonly accepted Trial of Org 10172 in Acute Stroke Treatment (TOAST) criteria shows only about 60% concordance among experts. This uncertainty is driven by the relatively unrevealing results of many stroke workups. Despite extensive testing, about 30% of all strokes can be considered cryptogenic, a number that rises to 50% among patients younger than 60 years of age. Of note, although stroke incidence has decreased in older adults, stroke incidence in the young has grown. While this relates in part to improved detection of alternative stroke etiologies, such as cervical arterial dissection and the identification of very small infarcts on MRI, the growth of cryptogenic stroke among young patients is poorly understood.
Identifying strokes as embolic, rather than lacunar or caused by intracranial atherosclerotic disease, is based on clinical and radiographic findings not represented in a study such as ANTARCTICA. Strokes are thought to be embolic when they have certain features, such as being bilateral or simultaneously in the anterior and posterior circulation. A cortical rather than subcortical location is suggestive of embolism, but only a small portion of subcortical strokes fit the size, location, and phenomenology associated with classic lacunar infarcts. With optimal treatment of HTN, DM, and hypercholesterolemia, the presence of risk factors also are inadequate to differentiate between atherosclerotic and other etiologies.
These questions, therefore, weaken the C-IS/TIA vs. non-C-IS/TIA analysis used in ANTCARTICA. Afib detected in a presumed C-IS/TIA patient may be less relevant if the true etiology is atherosclerosis. Equally, Afib may be directly responsible for infarcts presumed to be the result of a small or large vessel atherosclerotic mechanism that actually is cardioembolic. Therapeutically, such questions are moot. Any ischemic stroke patient who has Afib detected, no matter what the presumed subtype, will be anticoagulated.
Alan Z. Segal, MD, is Associate Professor of Neurology, Weill Cornell Medicine.
References
1. Sanna T, Diener HC, Passman RS, et al. Cryptogenic stroke and underlying atrial fibrillation. N Engl J Med. 2014;370(26):2478-2486.
2. Svendsen JH, Diederichsen SZ, Højberg S, et al. Implantable loop recorder detection of atrial fibrillation to prevent stroke (the LOOP study): A randomised controlled trial. Lancet. 2021; Oct 23;398(10310):1507-1516.