Despite being appropriately anticoagulated, a considerable proportion of ischemic stroke patients with atrial fibrillation (AF) experience recurrence (1). Factors [p]predictive of recurrence in patients with cardioembolic stroke (CE) include older age, presence of diabetes, large lesion, and left atrial enlargement (2). Another factor that may be associated with recurrence is the mechanism of stroke; more specifically, AF may be an innocent bystander, with alternative mechanisms precipitating ischemic stroke (2). A quarter of CE patients have concomitant high-grade carotid stenosis (3), whereas 10–15% of ischemic stroke patients with AF present with lacunar infarction. Approximately one-third of patients with AF seem to have cerebral infarction not associated with the cardioembolic source (AF-unrelated ischemic stroke) (4).

Predicting the risk of AF-related and AF-unrelated stroke may be important to determine the secondary prevention strategy. The CHA2DS2-VASC score is widely used to determine the risk of CE (5). However, the components of the CHA2DS2-VASC score are also risk factors for atherosclerosis and small vessel disease, both of which may increase the risk of AF-unrelated stroke recurrence (6). Therefore, there have been efforts to focus on the cardiac source itself to predict true AF-related stroke recurrence; previously, findings of echocardiography were related with the type of stroke recurrence (7). The characteristics of AF itself are also associated with the risk of stroke recurrence. The type of AF (persistent vs. paroxysmal AF) and its time of diagnosis (before vs. after stroke), which are associated with stroke recurrence, may also influence the type of stroke recurrence among ischemic stroke patients with AF (8). As ischemic stroke lesion patterns represent the stroke mechanism, it may also influence the type of stroke recurrence (9). The present study comprehensively investigated and compared the clinical and radiological factors associated with AF-related and AF-unrelated recurrences among ischemic stroke patients with AF.

Of the 3,213 patients enrolled in the K-ATTENTION study, 974 were excluded, including 101 patients without follow-up data, 870 patients without neuroimaging data, and three patients with incomplete workup. Thus, 2,239 patients were enrolled in this analysis (Figure 1). The mean follow-up duration was 1.46 years.

Of the 2,339 enrolled patients, 115 (5.1%) experienced recurrence of ischemic stroke. Risk factors did not differ significantly between patients with and without ischemic stroke recurrence, except that the prevalence of dyslipidemia was significantly higher in patients with stroke recurrence. The percentage diagnosed with AF before the index stroke was higher in patients with ischemic stroke recurrence (Table 1). Multivariable analysis showed that recurrence of any stroke was independently associated with AF diagnosed before the index stroke [hazard ratio (HR) = 1.82, 95% confidence interval (CI) 1.23–2.7; p = 0.002] and lesion patterns of subcortical infarction (HR = 2.38, 95% CI 1.24–4.59; p = 0.009) and small scattered lesions in a single vascular territory (HR = 2.00, 95% CI 1.10–3.64; p = 0.024).

Of the 115 patients with ischemic stroke recurrence, 75 (65.2%) had AF-related and 40 (34.8%) had AF-unrelated ischemic stroke recurrence. Although risk factors did not differ significantly between these two groups, the type of AF differed: the prevalence of persistent AF being higher in patients with AF-related stroke and the prevalence of paroxysmal AF being higher in patients with AF-unrelated stroke (Table 1). Regarding the initial lesion pattern, corticosubcortical infarction pattern was the most common in patients without stroke recurrence. In comparison, the pattern of confluent lesions with additional small lesions in a single vascular territory was the most common in patients with AF-related stroke recurrence, and the pattern of small scattered lesions in a single vascular territory was the most common in patients with AF-unrelated ischemic stroke recurrence. The recurrences of AF-related and AF-unrelated stroke after the index stroke according to the type of AF are shown at Figure 2.

This study found that AF diagnosed before the index stroke and the lesion patterns were independently associated with recurrence of ischemic stroke. Two-thirds of ischemic stroke recurrences were AF-related, whereas the remaining one-third was regarded as AF-unrelated. Regarding the type of stroke recurrence, persistent AF was associated with AF-related stroke recurrence, whereas paroxysmal AF was associated with AF-unrelated stroke recurrence.

AF is frequently detected by cardiac monitoring during patient admission during the acute stage (11). It is unclear whether AF observed in this stage is the cause of stroke or is triggered by the stroke (neurogenic AF) (12). A previous study found that the incidence of stroke recurrence was low in patients with AF detected after stroke, suggesting that their pathophysiology may differ (13). Similarly, the present study showed that AF diagnosed before stroke was more frequent in patients with ischemic stroke recurrence than those without. The influence of AF detected after stroke due to an autonomic or inflammatory response may differ from that of longstanding AF diagnosed before stroke (14). However, it did not differ according to the type of stroke recurrence.

The present study found that the type of AF was associated with the type of stroke recurrence. Although the rate of early ischemic recurrence was found to be higher in patients with persistent than with paroxysmal AF, this difference was no longer observed after adjustment for relevant risk factors (15). However, post-hoc analysis of the major non-vitamin K-dependent oral anticoagulant trials showed that the rates of thromboembolic events and deaths were lower in patients with paroxysmal than with persistent AF (16). In the present study, which assessed ischemic stroke recurrence according to stroke mechanism, persistent AF was associated with AF-related ischemic stroke recurrence. On the other hand, paroxysmal AF was relatively more associated with AF-unrelated ischemic stroke recurrence, which may be explained by the chance that paroxysmal AF may not be directly be the cause of AF-unrelated ischemic stroke recurrence.

Single small subcortical lesions are common in small vessel disease, and small scattered lesions in a single vascular territory are common in ischemic stroke caused by a corresponding artery stenosis (9). Although the type of index stroke and the presence of concomitant carotid stenosis were not associated with AF-unrelated stroke recurrence, lesion patterns representing embolism from a concomitant vascular stenosis were associated with AF-unrelated stroke recurrence. Non-stenotic atherosclerotic plaques may have contributed to ischemic lesions in a single vascular territory (17). Interestingly, the conventional CHA2DS2-VASC score was not significantly associated with AF-unrelated or AF-related stroke. Because factors associated with AF itself and imaging characteristics showed greater associations with AF-related or AF-unrelated stroke, a prediction model including these factors may increase its ability to predict each type of stroke recurrence.

The present study had several limitations stemming from its prospectively registered observational study design. First, our data are prone to a possible residual bias with methodological shortcomings. Several factors important for stroke recur, such as the presence of peripheral artery disease or the reason for not receiving anti-thrombotics, were not obtained. Second, though we have divided recurrent stroke to AF-related and unrelated strokes, there may be some controversies. However, DWI lesion pattern well-reflects the stroke mechanism, and the stroke mechanism was determined by local investigators and was once more confirmed centrally. Third, because this study was a non-randomized comparison of an observational registry, the relationship between treatment and outcome may have been influenced by measured or unmeasured confounding factors. For example, although treatment with antiplatelet agents at discharge was associated with AF-unrelated stroke recurrence, patients who were at risk of a non-CE recurrence may have received antiplatelet agents.

Despite these limitations, our study showed that the time of AF diagnosis was associated with stroke recurrence and that the type of AF affected the type of stroke recurrence among ischemic stroke patients with AF. AF diagnosed before the index stroke increases the risk of stroke recurrence. Persistent AF was more associated with AF-related stroke recurrence, whereas paroxysmal AF was more associated with AF-unrelated stroke. Ischemic lesions in a single vascular territory predicted the recurrence of AF-unrelated stroke.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

The studies involving human participants were reviewed and approved by Kyung Hee University Hospital. Written informed consent for participation was not required for this study in accordance with the national legislation and the institutional requirements.

BK contributed to the conception and design of the work, the acquisition, analysis and interpretation of data, and drafting of the content. Y-HH, M-SP, J-TK, K-HC, J-MJ, SY, CK, KO, T-JS, Y-JK, K-YP, J-MK, J-HP, JC, J-WC, OB, and G-MK contributed to the analysis of the work and critically revised the content. SH contributed to the acquisition and analysis of the data and critically revised the content. W-KS contributed to the conception and design of the work, the acquisition, analysis and interpretation of data, and critically revision of the content. All authors contributed to the article and approved the submitted version.

This study was supported by a grant from the Korean Neurological Association (KNA-17-MI-10: W-KS).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.