Original Paper Eur Neurol 2014;71:313–318 DOI: 10.1159/000357561

Received: June 21, 2013 Accepted: November 24, 2013 Published online: March 26, 2014

Silent Atrial Fibrillation after Ischemic Stroke or Transient Ischemic Attack: Interest of Continuous ECG Monitoring Vanessa Fernandez a Yannick Béjot b Marianne Zeller c Joëlle Hamblin a Benoit Daubail b Agnes Jacquin b Maud Maza a Claude Touzery a Yves Cottin a Maurice Giroud b a Cardiology Department, CHU Dijon, b Neurology Department and Dijon Stroke Registry, EA4184, CHU Dijon, and c Laboratory of Cardiometabolic Pathophysiology and Pharmacology, INSERM U866, University of Burgundy, Dijon, France

Key Words Atrial fibrillation · Stroke · Transient ischemic attack · Continuous ECG monitoring

Abstract Background: Since atrial fibrillation (AF) may be undiagnosed when asymptomatic and paroxysmal, we aimed to investigate the incidence and determinants of silent AF in patients with acute ischemic stroke or transient ischemic attack (TIA). Methods: Consecutive patients admitted to the Stroke Unit of the University Hospital of Dijon, France, for acute ischemic stroke or TIA were prospectively enrolled from March to December 2012. Silent AF was assessed by continuous electrocardiography (ECG) monitoring for 24 h after admission. An echocardiography was performed at day 2 ± 1 to measure left ventricular ejection fraction (LVEF) and left auricular (LA) dimensions. Results: Among the 187 patients included, 19 (10%) developed silent AF. Patients with silent AF were markedly older (76 vs. 66 years, p < 0.002), with lower creatinine levels (90 vs. 80 μmol/l, p = 0.030) and were less often smokers (5 vs. 24%, p = 0.058) than patients without silent AF. They also showed a trend towards more frequent

© 2014 S. Karger AG, Basel 0014–3022/14/0716–0313$39.50/0 E-Mail [email protected] www.karger.com/ene

hypertension and a recent history of infection. Patients with silent AF had a larger indexed LA volume (37.4 vs. 30.8 ml/m3, p = 0.057) and LA diameter (23.2 vs. 20.8 mm/m2, p = 0.059). LVEF in the two groups was similar. In multivariate analysis, only age remained an independent estimate of silent AF. Conclusion: Silent AF detected by continuous ECG monitoring is common and closely associated with older age. Further studies are needed to investigate the interest of systematically screening for silent AF for secondary prevention after ischemic stroke/TIA. © 2014 S. Karger AG, Basel

Introduction

Atrial fibrillation (AF) is frequent in patients with ischemic stroke. AF confers a fivefold risk of stroke, and one in five strokes is attributed to AF [1, 2]. However, because AF is often asymptomatic and paroxysmal, it is frequently undetected in clinical routine [2–5]. Moreover, paroxysmal AF carries the same risk of cardioembolism as persistent AF [6] and leads to ischemic stroke in approximately 25–30% of cases. Some patients with ‘crypMarianne Zeller, PhD Laboratory of Cardiometabolic Physiopathology and Pharmacology INSERM U866, Faculty of Medicine, University of Burgundy 7 Bd Jeanne d’Arc, FR–21000 Dijon (France) E-Mail marianne.zeller @ u_bourgogne.fr

togenic’ stroke or transient ischemic attack (TIA) may in fact have clinically undetected, asymptomatic AF [7]. Indeed, several studies suggest that paroxysmal AF is not always detected by standard 12-channel electrocardiography (ECG) and/or 24-hour Holter ECG recording [8, 9]. In contrast, a recent study demonstrated that continuous ECG monitoring (CEM) in stroke units improves the detection of paroxysmal AF in patients with stroke [10]. Current guidelines recommend that patients with stroke should be treated in stroke units, and must undergo a 12-channel ECG on admission, with CEM for at least the first 24 h to detect rhythm disorders [11–14]. The aim of this study was to investigate in patients with acute stroke and TIA (1) the rate of silent AF and (2) the determinants of silent AF including left atrium (LA) parameters.

Patients with history of atrial fibrillation n = 35 Patients with symptomatic fibrillation n=5 Continuous ECG monitoring n = 187

Patients without atrial fibrillation n = 168

Patients with silent atrial fibrillation n = 19

Fig. 1. Flowchart of the study. n refers to number of events.

Methods This prospective study included all consecutive patients hospitalized in the Stroke Unit of the University Hospital of Dijon, France, from March 2012 to December 2012 for acute ischemic stroke or TIA. To be included in the final analysis, patients had to meet the following prespecified criteria: (1) age ≥18 years; (2) stroke/TIA diagnosis by a neurologist specialized in stroke; (3) no previously known AF in the medical history; (4) no AF on the admission ECG; (5) availability of ECG data for CEM analysis ≥24 h, and (6) patients with shock or requiring brain surgery were transferred to a specialized intensive care unit and were no longer followed once they had been discharged from the Stroke Unit. Stroke was defined by a sudden neurological disorder due to brain lesion from a vascular origin. TIA was defined by a short episode of brain dysfunction with clinical symptoms lasting 200 ms [13]. CEM was started immediately after admission using the CSM Philips IntelliVue MP50 system. This monitoring system includes the following rhythm alarms: (1) flat ECG; (2) ventricular fibrillation; (3) couplets or pulsus bigeminus, and (4) preset upper and lower heart rate thresholds (120 and 40/min). Silent AF was defined as asymptomatic episodes of AF lasting at least 30 s. Symptomatic AF was defined as rapid AF resulting in clinical symptoms of heart failure, palpitations or the need for urgent cardioversion. Blood samples were collected on admission for biological parameters. An echocardiographic evaluation was performed to assess left ventricular ejection fraction (LVEF) and LA size. LA diameter, surface area, and volume were indexed to the body area [14]. Recent infection (ULN Creatinine, μmol/l HDL-C, mmol/l LDL-C, mmol/l Total cholesterol, mmol/l Glucose, mmol/l TSH, mIU/l

No AF (n = 168)

Silent AF (n = 19)

p

42 (26) 47 (28) 79.9 ± 28.8 1.3 ± 0.6 2.8 ± 1.2 5.0 ± 1.3 6.8 ± 2.3 2.4 ± 6.7

5 (26) 4 (21) 90.0 ± 26.7 1.3 ± 0.4 2.7 ± 0.7 4.5 ± 0.8 6.4 ± 1.6 1.8 ± 1.7

0.959 0.511 0.030 0.247 0.990 0.114 0.913 0.959

Values are n (%) or mean ± SD. CRP = C-reactive protein; HDL-C = high-density lipoprotein cholesterol; LDL-C = low-density lipoprotein cholesterol; TSH = thyroid-stimulating hormone; ULN = upper limit of normal.

No AF (n = 168)

Silent AF (n = 19)

p

110 ± 43 157 ± 23 121 ± 15 88 ± 34 90 ± 24 75 ± 19 0 18 ± 24 14 (74) 18 (6 – 90) 76 ± 19 105 ± 26 63 ± 16

0.143 0.886 0.227 0.048 0.052 0.978 NA NA 0.116 0.013 0.231 0.075 0.672

Values are n (%), mean ± SD or n (range). DBP = Diastolic blood pressure; HR = heart rate; NA = non-applicable; SBP = systolic blood pressure; SVES = supraventricular extrasystoles; VF = ventricular fibrillation; VT = ventricular tachycardia.

Table 4. Echocardiography parameters (mean ± SD)

LVEF, % Indexed LA diameter, mm/m2 Indexed LA surface, cm2/m2 Indexed LA volume, ml/m3

Eur Neurol 2014;71:313–318 DOI: 10.1159/000357561

No AF (n = 168)

Silent AF (n = 19)

p

59 ± 10 20.8 ± 3.9 9.7 ± 2.6 30.8 ± 9.1

59 ± 7.7 23.2 ± 4.4 11 ± 2.9 37.4 ± 7.9

0.582 0.059 0.350 0.057

315

Color version available online

1.0

Sensitivity

0.8

0.6

0.4

0.2

0

0

0.2

0.4

0.6

0.8

1.0

1 – specificity

Fig. 2. ROC curve for age as a predictor of silent AF (AUC 0.72,

95% CI 0.61–0.82, p = 0.002).

towards larger LA in the silent-AF group, no significant difference was observed for indexed LA volume (37.4 vs. 30.8 ml/m3, p = 0.057) or LA diameter (23.2 vs. 20.8 mm/m2, p = 0.059) for the two groups. In order to identify factors associated with the occurrence of silent AF, we performed a backward logistic regression analysis including all of the variables that were associated with silent AF in univariate analysis (i.e. age, creatinine, hypertension, ACE inhibitor, smoking, infection). Only age remained an independent predictor of silent AF (OR 1.05, 95% CI 1.01–1.09). The threshold age that maximized combined specificity and sensitivity for predicting silent AF on the ROC curves was 78 years (AUC 0.72, 95% CI 0.61–0.82, p = 0.002), with sensitivity at 63% and specificity at 74% (fig. 2).

Discussion

CEM, a standard procedure for routine clinical practice in stroke units, provided the major findings of our preliminary study: (1) silent AF was common at the acute phase of ischemic stroke or TIA; (2) no significant relationship was observed between silent AF and echocardiography parameters, and (3) only older age was an independent correlate for developing silent AF. 316

Eur Neurol 2014;71:313–318 DOI: 10.1159/000357561

Incidence of AF and Detection Strategies after Stroke or TIA The incidence of new AF varies widely depending on cardiac monitoring devices, recruitment criteria, stroke characteristics, duration of cardiac monitoring and the time between stroke onset and the start of ECG monitoring [16]. In a recent study, Seet et al. [16] performed a systematic review of studies that assessed the incidence of newly diagnosed AF after stroke. The average incidence in an unselected population was 5.3% (range 1.0–13.9%) with two detection methods (i.e. ambulatory ECG (24 h) or inpatient CEM), the duration of screening ranged from 24 h to 7 days and populations ranged from 96 to 465 patients. Our findings, with an incidence of 10.2%, are in agreement with these results. Our work, based on CEM monitoring, is in agreement with a recent study which showed that CEM was more efficient than 24-hour Holter ECG for the detection of AF [10]. The above study, based on a population of 469 unselected patients, found an incidence of 8.7%. Although the duration of screening was limited to 24 h after the acute stage in most studies, the rare investigations that carried out ECG recording for a longer period identified a higher incidence of AF [17, 18]. Overall, in these studies the duration of monitoring ranged from 22 h to 14 months. In a recent study using 30-day monitoring of cardiac events, 218 patients with a diagnosis of ischemic stroke or TIA were analyzed. Among these patients, 36 (16.5%) were classified as suffering from cryptogenic stroke and 20% were diagnosed with AF [18]. In our study, ECG monitoring duration was limited to 24 h as the minimal length of stay in the Stroke Unit. Overall, even with short period of recording, we found that silent AF was not rare in patients with acute stroke. When patients with previous stroke are monitored, it is pertinent to wonder if the observed arrhythmia could be a consequence of ischemic lesions in the brain, rather than a cause. ECG abnormalities, including atrial arrhythmia, have been reported in patients with no underlying cardiac disease [19, 20]. Moreover, some studies have reported a significantly higher incidence of atrial tachyarrhythmia, particularly in patients with anterior circulation infarction [21]. Significance of Paroxysmal AF Detected after Stroke New technology now allows the identification of very brief episodes of paroxysmal AF, the significance of which is uncertain. It is not known, for example, whether patients with paroxysmal AF included in previous trials could represent a selected population with high AF burFernandez et al.

den (sufficient to be detected by routine ECG) resulting in a stroke risk comparable to that in patients with chronic AF [22, 23]. Nevertheless, in the recent ASSERT study, subclinical atrial tachyarrhythmia, without overt clinical AF, occurred frequently in patients implanted with a pacemaker and was associated with a significantly increased risk of ischemic stroke or systemic embolism [24– 26]. However, only 12% of patients had a prior stroke or TIA. To the best of our knowledge, no data are currently available on silent AF in patients with acute stroke. Echocardiography Parameters and Paroxysmal AF Detected after Stroke Echocardiography is an important examination in the evaluation of stroke. AF is a progressive condition that begins with hemodynamic and/or structural changes in the LA and evolves throughout the paroxysmal and persistent stage [25]. LA enlargement has been widely related to AF, both in chronic AF and in paroxysmal AF [27, 28]. In the present study, patients with silent AF showed only a trend towards LA enlargement (indexed LA volume 37.4 vs. 30.8 ml/m3, p = 0.057). In contrast, a previous study demonstrated a significantly greater indexed LA volume in patients with paroxysmal AF than in control patients (33.2 vs. 26.7 ml/m3, p = 0.004) [29]. These apparent discrepancies with our results can be explained by (a) patients were compared with controls; (b) our population was much older; (c) the relatively small sample, and (d) subnormal values reflecting an early stage. Previous studies have failed to demonstrate an association between both baseline and follow-up LA size and the onset of arrhythmia or episodes of paroxysmal AF. This supports the hypothesis that LA enlargement occurs only if AF becomes chronic [30, 31]. Moreover, Suarez et al. [32] reported that paroxysmal AF may induce slowly progressive LA enlargement. In addition, Kallmünzer et al. [33] recently demonstrated that the risk of significant cardiac arrhythmia after an acute cerebrovascular event is highest during the first 24 h of care and then progressively declines over the first 3 days. In our study, NIHSS was not associated with silent AF. However, the study of Kallmünzer et al. [33] considered all acute cerebrovascular events and all clinically significant arrhythmias such as ventricular arrhythmia, sinus arrest, second- or third-degree atrioventricular block and AF. Study Limitations Our study has several limitations. First of all, it was a single-center study and ECG recording was limited to the first 24 h. Then, the echocardiography analysis was conducted without left atrial deformation analysis by Silent Atrial Fibrillation after Ischemic Stroke

speckle tracking and/or left atrial function. Moreover, we only compared silent AF patients with non-silent AF patients, in the absence of a control group. In the present study, we included all type of infarctions including lacular, atherothrombotic and cryptogenic strokes. This design was chosen because a significant rate of stroke initially diagnosed as lacunar or atherothrombotic during their stay in the Stroke Unit may be finally attributable to cardioembolic stroke at the end of their hospital stay. As suggested in recent studies, it has been demonstrated that patients with atherothrombosis often suffer from AF and that the population is associated with more frequent cardiovascular outcomes including strokes [34]. Moreover, Wessels et al. [35] showed that approximately one third of strokes clinically defined as lacunar were finally identified by MRI as cardioembolic strokes. Most patients (n = 114) were initially diagnosed as cryptogenic strokes, based on the TOAST classification, of whom 14 developed silent AF. In this study, patients were included early on admission in the Stroke Unit (