NewlyDiagnosedAtrial Fibrill

ation after Acute Ischemic Strokeand Transient Ischemic Attack: Importance of Immediate and

Prolonged Continuous Cardiac Monitoring

Luciano A. Sposato, MD, MBA,*† Francisco R. Klein, MD,* Agustın Jauregui, MD,*

Marisol Ferrua, MD,* Pablo Klin, MD,‡ Rafael Zamora, MD,x Patricia M. Riccio, MD,*

and Alejandro Rabinstein, MD{

From the *Stroke

†Neurovascular Center a

nos Aires, Argentina, ‡

ment, University Hosp

Argentina, xDepartment

“Jose de San Martın,” B

of Neurology, Mayo Clin

Received May 3, 2010;

Address corresponden

Center at the Institute ofN

Foundation, Solıs 443, P

E-mail: lucianosposato@g

1052-3057/$ - see front

� 2012 by National Str

doi:10.1016/j.jstrokece

210

Atrial fibrillation (AF) is the major cause of cardioembolic stroke. It often remains

occult when asymptomatic and paroxysmal. We hypothesized that the detection

of AF after acute ischemic stroke (AIS) or transient ischemic attack (TIA) could be

improved by using continuous cardiac monitoring (CCM) immediately after admis-

sion.We sought to determine the detection rate of AF by immediate in-hospital CCM

after cryptogenic and noncryptogenic AIS or TIA in patients without a previous

diagnosis of AF.We retrospectively studied a cohort of 155 patients with cryptogenic

and noncryptogenic AIS or TIAwithout knownAF.We compared the detection rates

of newly diagnosed AF (NDAF) in patients admitted to areas with CCM and those

never admitted to these areas.We developed amultiple logistic regressionmodel for

identifying predictors of NDAF. We characterized NDAF episodes and analyzed

how the availability of CCM data changed secondary prevention strategies. We

detected NDAF in 21 patients (13.5%). Diagnostic rates of NDAF in patients who

underwent CCMand thosewho did not undergo CCMwere 18.2% and 2.2%, respec-

tively (P 5 .005). The median time from admission to recognition of NDAF was 2.0

days. Most NDAFs were paroxysmal (95.2%) and lasted less than 1 hour (85.7%).

Diabetes mellitus and infarct size were predictors of NDAF. Detection of NDAF

prompted the initiation of anticoagulation therapy in 8.2% of the patients

admitted to areas with CCM availability. Our findings suggest that immediate

and prolongedCCM significantly improves the detection of NDAF after cryptogenic

and noncryptogenic AIS or TIA, and that diabetes mellitus and infarct size are

significantly associated with NDAF. Key Words: Strokedtransient ischemic

attackdtelemetrydarrhthmyadelectrocardiographyddiagnosis.

� 2012 by National Stroke Association

Center, Institute of Neurosciences,

nd Vascular Research Unit at INECO. Bue-

Telemetry Section, Cardiovascular Depart-

ital, Favaloro Foundation, Buenos Aires,

of Internal Medicine, Hospital de Clınicas

uenos Aires, Argentina; and {Department

ic, Rochester, Minnesota.

accepted June 27, 2010.

ce to Luciano A. Sposato, MD, MBA, Stroke

eurosciences,UniversityHospital, Favaloro

iso 3, Ciudad de Buenos Aires, Argentina.

mail.com.

matter

oke Association

rebrovasdis.2010.06.010

Journal of Stroke and Cer

Atrial fibrillation (AF) constitutes a major risk factor for

ischemic stroke, and stroke patients with AF have a poorer

prognosis than those without sinus rhythm [1]. The identi-

fication of newly diagnosed AF (NDAF) in patients with

acute ischemic stroke (AIS) or transient ischemic attack

(TIA) is crucial, because oral anticoagulation significantly

decreases the risk of stroke recurrence [2]. NDAF recogni-

tion remains elusive, because 50% or more of cases may

present with asymptomatic and intermittent patterns [3-5].

Standard electrocardiography (ECG) may be insuffi-

cient for detecting NDAF. Even 24-hour ECG recording

(Holter monitoring) has shown relatively poor sensitivity

for the detection of NDAF, with a negative predictive

ebrovascular Diseases, Vol. 21, No. 3 (April), 2012: pp 210-216

NEWLY DIAGNOSED ATRIAL FIBRILLATION AFTER ISCHEMIC STROKE 211

value for freedom from AF of 25%-40% in paroxysmal AF

[5,6]. This means that that during a 1-year follow-upmon-

itoring period without detection of AF, only 1 in 3 patients

actually will be free of AF [7]. Furthermore, NDAF was

detected in 5.7% of patients with normal ECG and normal

Holter monitoring using a 7-day event loop recording in

the ambulatory setting [4].

To date, no clinical trial has assessed the usefulness of

immediate and prolonged in-hospital continuous cardiac

monitoring (CCM) in patients with cryptogenic and non-

cryptogenic AIS or TIA. The objective of the present study

was to determine the rate of detection of NDAF by imme-

diate and prolonged in-hospital CCM after cryptogenic

and noncryptogenic AIS or TIA.

Methods

This retrospective case series included all consecutive

patients aged.18 years admitted to our hospital with a di-

agnosis ofAIS or TIA between January 1, 2007, andDecem-

ber 31, 2008. Our study cohort comprised both cryptogenic

and noncryptogenic events. We chose to study noncrypto-

genic AIS and TIAs because more than one pathophysio-

logic mechanism can coexist in the same patient (eg,

extracranial carotiddisease andAF) [8]. Thus, evenpatients

with a defined noncardioembolic stroke could benefit from

antiarrhythmic treatment and anticoagulation if an NDAF

is identified. We defined a cryptogenic event as any AIS

or TIAwith no identifiable etiology on complete diagnostic

workup [9].

This study was conducted before publication of the

recent American Heart AssociationeAmerican Stroke

Association’s TIA Definition and Evaluation Statement

[10]. Consequently, we defined TIA as the sudden onset

of focal neurologic signs or symptoms lasting less than

24 hours with no associated ischemic lesions on neuroi-

maging studies. We defined ischemic stroke as the sud-

den onset of focal neurologic signs or symptoms lasting

longer than 24 hours.

PatientswithAIS andTIAwere admitted to the telemetry

ward unless there was no bed available, in which case

patientswithTIAandmildstrokewereadmitted to thegen-

eralward, and thosewithmore severe strokewereadmitted

to the intensive care unit (ICU) or the coronary care unit

(CCU). Most postoperative strokes were diagnosed in the

postsurgical care unit. Patients admitted to the latter critical

care area (CCA)were continuouslymonitored for thedetec-

tion of electrocardiographic abnormalities immediately

after admission, whereas patients not admitted to a CCA

did not undergo CCM. The telemetry ward is equipped

with an Infinity telemetry system (DragerMedical, Telford,

PA) for the receptors and an Infinitymultiviewworkstation

as a central monitoring device linked to portable telemetric

transmitters. In addition, 4 centrally supervised bedside

monitors (Infinity Vista XL) are available for unstable

patients requiring direct observation of cardiac rhythm.

The central monitoring workstation provides 24-hour

7-lead ECG monitoring, arrhythmia detection, and mea-

surement of ST-T segment changes. Two highly trained

nurseswatch the centralmonitors 24hours aday. If a parox-

ysmal event is noted, then the corresponding strip is

printed out for further assessment by a certified cardiolo-

gist. To decrease the risk of missing short paroxysmal

events, data are recorded and reviewed by the nurses at ev-

ery 8-hour shift. The CCU/ICU and postsurgical care unit

are equipped with Marquette Dash 4000 Pro (GE Health-

care, Chalfont St. Giles, UK), and Eagle 4000 (GE Health-

care) multilead ECG patient monitors, respectively. All

monitors are integrated into a central workstation at each

area. These systemsare configuredwith simultaneousmul-

tilead arrhythmia analysis software (including AF detec-

tion) for event notification, storage, and recall. One highly

trained nurse is exclusively available for each patient in

the CCU, ICU, and postsurgical care unit. Monitor alarms

are set to sound on any change in ECG rhythm or heart

rate$120 beats perminute. In either case, the nurse records

a rhythm strip for assessment by a certified cardiologist.

During the hospital stay, each patient with stroke or

TIA was assessed daily by one or more of the stroke

team members. On admission, each patient with AIS or

TIA underwent a neurologic examination and a 12-lead

ECG evaluation. Further ECGs were ordered only when

a patient became symptomatic while hospitalized in the

general ward (eg, palpitations, chest pain). In the patients

admitted to a CCA, ECG was performed daily and when-

ever a paroxysmal event was detected by CCM.

Our usual diagnostic workup includes an initial com-

puted tomography (CT) scan of the brain, blood tests,

neurovascular imaging studies (Doppler ultrasound, CT

angiography, or magnetic resonance angiography), and

transesophageal or transthoracic Doppler echocardiogra-

phy (TEE or TTE). During the study period, we could not

ordermagnetic resonance imaging (MRI) studies on a reg-

ular basis because of limited availability. For this reason,

we ordered an MRI only for selected patients. A second

brain CT scan was done in some patients during the

hospital stay or after discharge.

The study design and the distribution of patients into

the different monitoring groups are shown in Figure 1,

a flow diagram adapted from the Consolidated Standards

of Reporting Trials (CSRT) [11]. For the present analysis,

we excluded patients with a history of chronic AF or par-

oxysmal AF, and those with AF detected on the admission

ECG. The diagnosis of chronic and paroxysmal AF was

made on the basis of patient recall or by documentation

of the arrhythmia on a previous ECG when available.

AF was defined by the absence of P waves or evidence

of atrial flutter with irregular ventricular response.

NDAF was defined as AF diagnosed after AIS or TIA in

a patient with no history of chronic or paroxysmal AF.

We compared rates of NDAF in patients admitted to

aCCAand those not admitted to aCCA (ie, those admitted

Figure 1. Description of study design adapted from the CONSORT flow

diagram [11].

L.A. SPOSATO ET AL.212

only to the general ward). Following a preliminary analy-

sis indicating that most initial NDAF events were diag-

nosed within the first 3 days of CCM, we analyzed the

number of AF events detected in the CCAs and in the

general ward during the first 3 days of admission to

evaluate the diagnostic yield of CCM during that period.

We calculated the number of patients that needed to be

followed over 3 days to identify one NDAF (NNF3days).

We assessed baseline characteristics, risk factor profiles,

infarct size, Trial of ORG 10172 in Acute Stroke Treatment

(TOAST) AIS and TIA subtypes, left ventricular ejection

fraction (LVEF), and atrial area.All CTscanswere reviewed

by the same certified neurologist (L.A.S.). Infarct size was

determined by direct observation and dichotomized into

,15 mm or $15 mm [12]. Patients with CT scans without

evidence of an acute infarctionwere included in the “lesion

size ,15 mm” group. Diabetes mellitus was diagnosed if

one or both of the following criteria were met: (1) history

of diabetes mellitus reported by the patient based on a pre-

vious diagnosis, or (2) use of an oral hypoglycemic drug or

insulin before the occurrence of stroke. LVEFwas classified

as normal if $55%, moderately impaired if ,55% and

$30%, and severely impaired if ,30% [13]. Left atrial

enlargement was defined as an atrial area .20 cm2 [13].

TOAST subtypes were determined by two investigators

(M.F. and L.A.S.) before and after being apprised of the

CCM results. In case of a disagreement, a final decision

was made by consensus between the two investigators.

We compared the foregoing variables in patients with

and without a diagnosis of NDAF and also in patients ad-

mitted for at least 1 day to a CCA (with CCM) and those

not admitted to a CCA (without CCM). We used length of

stay as a proxy of duration of ECGmonitoring in the latter

group. We documented the following characteristics of

NDAFs: time from admission to diagnosis of the arrhyth-

mia, number of episodes, mean duration (stratified into

#1hour, 1e24 hours, and .24 hours), and discharge

with or without AF.We assessed the changes in secondary

prevention strategies after NDAF detection.

This study design was approved by the Institutional

Review Board and the Ethics Committee of the Favaloro

Foundation University Hospital.

Statistical Analysis

We used the c2 test or the Student t test for comparing

categorical variables and the Mann-Whitney U test for

comparing continuous variables. We performed a univari-

ate analysis for determinants of NDAF by excluding those

patients not admitted to a CCA (n 5 45). We developed

a multiple logistic regression model to identify predictors

of NDAF. Variables with P #.20 in the univariate analysis

were included in the multivariate model along with other

factors considered biologically important. A P value ,.05

was deemed statistically significant in the multivariate

model. The association between covariates and NDAF

detection was expressed as odds ratio (OR) and 95%

confidence interval (CI). SPSS 13.0 for Windows (SPSS

Inc, Chicago, IL) was used for all statistical analyses.

Results

A total of 198 patients with AIS or TIAwere admitted to

our institution during the study period (Fig 1). The study

population comprised 155 patients (114 with AIS and 41

with TIA) after the exclusion of 26 patients with chronic

AF (13.1%), 14 with paroxysmal AF (7.1%), and 3 with AF

detected on the admission ECG (1.5%). A total of 113

patientswere admitted to at least oneCCA (91 to the telem-

etry ward, 30 to the ICU, 19 to the CCU, and 29 to the post-

surgical care unit), and 45 patients were not admitted to

a CCA. Baseline characteristics are presented in Table 1.

Each patient underwent neuroimaging studies, either

one (100%) or two (49%) CT scans or MRI (11%). Overall,

148 patients (95.5%) underwent either TEE or TTE.

Seventy-five patients (48.4%) underwent TEE, 54 (49.1%)

in the CCA group and 21 (46.7%) in the non-CCA group.

The mean duration of hospital stay was 9.9 6 12.8 days

(median, 5.0 days; interquartile range [IQR], 3.0-12.0

days).

NDAFwas identified in 21 patients (OR, 13.5%; 95% CI,

9.0%-19.8%). Of these 21 cases, 20 (95.2%) were initially

diagnosed in a CCA: 9 in the postsurgical care unit, 4 in

the ICU, 0 in the CCU, and 7 in the telemetry ward.

One case of NDAF was detected in the general ward

(0.8%). The rate of diagnosis of NDAF in patients admit-

ted to a CCAwas 18%, and that in patients not admitted to

a CCA was 2.2% (P 5 .005). The overall NNF3days value

Table 1. Characteristics of the patients admitted and those not admitted to a CCA

Overall (n 5 155)

Admitted to

a CCA (n 5 110)

Not admitted to

a CCA (n 5 45) P

Baseline characteristics

NIHSS score, mean 6 SD* 7.1 6 6.2 8.0 6 6.3 4.0 6 4.4 .001

NIHSS score .8, %* 34/114 (29.8) 3/89 (37.1) 1/25 (4.0) .001

Infarct size .15 mm* 55/114 (48.2) 45/89 (50.6) 10/25 (40.0) .24

Postprocedural stroke or TIA, n (%) 38/116 (49.1) 32/110 (29.1) 6/45 (13.3) .03

Risk factor profile

Age, years, mean 6 SD 66.9 6 13.0 68.1 6 12.1 64.0 6 14.8 .22

Male sex, n (%) 111/155 (71.6) 78/110 (70.9) 33/45 (73.3) .46

Hypertension, n (%) 131/155 (84.5) 93/110 (84.5) 38/45 (84.4) .58

Diabetes mellitus, n (%) 44/155 (28.4) 34/110 (30.9) 10/45 (22.2) .19

Smoking, n (%) 68/155 (43.9) 45/110 (40.9) 23/45 (51.1) .16

Previous stroke, n (%) 34/155 (21.9) 28/110 (25.5) 6/45 (13.3) .07

Previous myocardial infarction, n (%) 27/155 (17.4) 20/110 (18.2) 7/45 (15.6) .45

Atrial area .20 cm2, n (%)y 68/147 (46.3) 50/105 (47.6) 18/42 (42.9) .37

LVEF ,30%, n (%)y 14/147 (9.5) 8/105 (7.6) 6/42 (14.3) .17

Stroke subtype before NDAF detection

Great vessel, n (%) 20/155 (12.9) 14/110 (12.7) 8/45 (17.8) .55

Cardioembolic, n (%) 33/155 (20.3) 25/110 (22.7) 8/45 (17.8) .33

Small vessel, n (%) 16/155 (10.3) 10/110 (9.1) 6/45 (13.3) .30

Other causes, n (%) 44/155 (28.4) 35/110 (31.8) 9/45 (20.0) .10

Undetermined, n (%) 42/155 (27.1) 26/110 (23.6) 16/45 (35.6) .10

Duration of CCMzDuration of monitoring, days,

mean 6 SD

9.9 6 12.8 12.2 6 14.4 4.3 6 4.2 ,.001

Duration of monitoring,

days, median (IQR)

5.0 (3.0-12.0) 7.0 (3.0-15.75) 3.0 (2.0-5.0) ,.001

Outcome and NDAF detection

In-hospital mortality, n (%) (95% CI) 10/155 (6.5) (3.5-11.5) 9/110 (8.2) (4.4-14.8) 1/45 (2.2) (0.4-11.6) .16

NDAF detection, n (%) (95% CI) 21/155 (13.5) (9.0-19.8) 20/110 (18.2) (12.1-26.4) 1/45 (2.2) (0.4-11.6) .005

Abbreviation: NIHSS, National Institutes of Health Stroke Scale.

*The 42 patients with TIA were excluded.

yEight patients had missing echocardiographic data.

zLength of stay was used as a proxy for duration of CCM in patients never admitted to a CCA.

NEWLY DIAGNOSED ATRIAL FIBRILLATION AFTER ISCHEMIC STROKE 213

was 16.7 for the CCA group (4.6 for the postsurgical care

unit, 8.6 for the ICU, and 18.7 for the telemetry ward) and

177 for the non-CCA group.

Themedian time fromadmission to recognitionofNDAF

was 2.0days (IQR, 1.0-3.5 days). In 71%of cases,NDAFwas

detected during the first 3 days of hospitalization. The time

fromadmission to detection ofNDAF is plotted in Figure 2.

NDAFwas paroxysmal in 20 patients (95.2%). The propor-

tions of patients with #3 episodes, .3 episodes, and sus-

tained NDAF (ie, AF not responsive to cardioversion)

were 61.9%, 33.3%, and 4.8%, respectively. All episodes of

AF were asymptomatic except for the one case detected in

the general ward. The duration of the AF episode was #1

hour in 17 patients, 1-24 hours in 2 patients, and.24 hours

in 2 patients. The minimum duration of a NDAF episode

was #30 seconds in 2 patients and .30 seconds in 19 pa-

tients. Eight patients returned to sinus rhythm spontane-

ously, although 2 of these patients required further

pharmacologic cardioversion with amiodarone due to

recurrent AF. A third patient required the correction of

hypokalemia; this was the only case with an electrolytic

disturbance. The remaining 13 patients were treated with

either electrical (n5 4) or pharmacologic (n5 9) cardiover-

sion. Three patients with NDAF died during the hospital

stay, 1 patient persisted with AF, and 17 patients were dis-

charged on sinus rhythm. No cases of lone AF were seen.

As shown in Table 1, the patients admitted to a CCA

had a higher proportion of severe stroke.We found no dif-

ferences in the rate of NDAF between patients with AIS

and those with TIA (15.5% vs 14.3; P5 .95). The frequency

of NDAF detection was 27.3% (95% CI, 9.8%-56.6%) in

cryptogenic events and 14.3% (95% CI, 9.5%-20.9%) in

noncryptogenic events (P 5 .47).

We included age, sex, and variables with a significance

level of P # .20 in the univariate analysis (Table 2) in the

multiple logistic regression model. We also included left

atrial enlargement, systemic hypertension, reduced

LVEF (,30%), and periprocedural cerebrovascular events

Figure 2. Time elapsed between admission and

NDAF.

L.A. SPOSATO ET AL.214

because of their biological relevance [14]. Diabetes melli-

tus and infarct size were predictors of NDAF in the mul-

tiple logistic regression model (Table 3). We found no

significant interaction effect between stroke severity and

infarct size.

NDAF detection prompted anticoagulation therapy in

9 patients admitted to a CCA (8.2%) with the following

noncardioembolic TOAST subtypes determined before

diagnosis of NDAF: 2 patients with great vessel disease,

4 with otherwise undetermined cause (1 with incomplete

evaluation and 3 with complete workup), 1 with small

vessel disease, and 2 with other etiologies.

Discussion

We found thatNDAF is common in patientswith crypto-

genic and noncryptogenic AIS or TIA undergoing immedi-

ate and prolonged in-hospital CCM. The rate ofNDAFwas

higher in patients undergoing CCM at a CCA compared

with those assessed by ECG outside of a CCA. Most

episodes of NDAF were recognized during the first 3

days of hospitalization. Diabetes mellitus and infarct size

were independent predictors of NDAF. Identification of

the arrhythmia led to important changes in interpretation

of the pathophysiologic mechanisms of cerebrovascular

events.

The 18.2% rate of NDAF seen in the patients with AIS

and TIA undergoing CCM is higher than rates reported in

patientsassessedonlyby in-hospital serialECG,Holtermon-

itoring, and event loop recording in previous studies. All to-

gether, in 5previous studies [4,15-18],NDAFwasdetected in

4.6% (95% CI, 0-12.7%) of 588 consecutive patients with

ischemic stroke assessed by Holter monitoring and event

loop recording [19]. Two recent studies identified NDAF in

10.6%-12.5% patients with AIS evaluated by in-hospital 24-

hour Holter monitoring or the combination of ECG

and Holter monitoring [20,21]. AF also was found in 23%

of patients with ischemic stroke and TIA patients using

a 21-day mobile cardiac outpatient telemetry [22] and in

20% of these patients using a 30-day cardiac event monitor

[23], although it must be noted that these studies were

restricted to cryptogenic events.

There are several possible explanations for the high rate

of NDAF detection in the present study. Given that 71% of

the NDAFs occurred within the first 3 days of hospitaliza-

tion (Fig 2), we speculate that the immediate initiation of

CCM after admission might have contributed to the high

rate of NDAF detection. In previous studies, the time

from stroke occurrence to initiation of ECG monitoring

varied between 3 days and 15 days [17], with an average

of 7.6 6 4.5 days [4,15-18,20,21]. Another plausible

explanation for the high rate of NDAF detection in the

present study is the prolonged nature of CCM. The

mean duration of CCM in our study was 12 days,

compared with 21e72 hours in previous studies using

Holter monitoring [4,16-18,20] and 4e7 days in studies

using event loop recording [4,16]. Furthermore, the

extended nature of CCM in our population allowed us

to identify 29% of NDAFs occurring up to 24 days after

admission. The highest rates of NDAF detection

reported before the present study were achieved in

patients with cryptogenic AIS and TIA undergoing

prolonged monitoring in the outpatient setting [22,23].

As reported previously [22], diabetes mellitus was an

independent predictor of NDAF in the present study.

Diabetic dysautonomy could explain this finding, given

that parasympathetic and sympathetic phenomena are

possible mechanisms of paroxysmal AF [24]. Further-

more, diabetes mellitus might contribute to the impaired

cardiovascular autonomic function seen in stroke survi-

vors [25]. Infarct size .15 mm was also a predictor of

NDAF in our multiple logistic regression model. It can

be hypothesized that larger infarcts are associated with

more severe autonomic dysregulation as well.

Surgical procedures constitute a predisposing factor

for AF and other cardiac arrhythmias [26]. As shown in

Table 1, postprocedural events were more frequent in

the patients admitted to a CCA than in those never admit-

ted to a CCA (19.1% vs 13.3%; P5 .03). This finding is not

unexpected, given that postsurgical patients are almost

always admitted to the postsurgical care unit at our

Table 2. Univariate analysis for newly diagnosed atrial fibrillation in patients admitted to a CCA

Overall

(n 5 110)

With NDAF

(n 5 20)

Without NDAF

(n 5 90) P

Baseline characteristics

NIHSS score, mean 6 SD* 8.0 6 6.3 12.3 6 6.9 7.2 6 5.9 .33

NIHSS score .8, n (%)* 33/89 (37.1) 11/15 (73.3) 22/74 (29.7) .002

Infarct size .15 mm, n (%)* 45/89 (50.6) 12/15 (80.0) 33/74 (44.6) .01

Postprocedural stroke or TIA, n (%) 32/110 (29.1) 7/20 (35.0) 25/90 (27.8) .35

Risk factor profile

Age, years, mean 6 SD 68.1 6 12.0 64.7 6 9.5 68.8 6 12.5 .63

Male sex, n (%) 78/110 (70.9) 16/20 (80.0) 62/90 (68.9) .24

Hypertension, n (%) 93/110 (84.5) 15/20 (75.0) 78/90 (86.7) .17

Diabetes mellitus, n (%) 34/110 (30.9) 10/20 (50.0) 24/90 (26.7) .04

Smoking, n (%) 45/110 (40.9) 9/20 (45.0) 36/90 (40.0) .43

Previous stroke, n (%) 28/110 (25.5) 3/20 (15.0) 25/90 (27.8) .19

Previous myocardial infarction, n (%) 20/110 (18.2) 2/20 (10.0) 18/86 (20.0) .24

Atrial area .20 cm2, n (%)y 50/105 (47.6) 11/19 (57.9) 39/86 (45.3) .23

LVEF ,30%, n (%)y 8/105 (7.6) 2/20 (10.0) 6/85 (7.1) .47

Duration of CCMzDuration of monitoring, days,

mean 6 SD

12.2 6 14.4 23.4 6 17.0 9.7 6 12.6 .03

Duration of monitoring, days,

median (IQR)

5.0 (3.0-12.3) 18.5 (11.5-33.5) 5.0 (3.0-12.0) .001

Outcome

In-hospital mortality, n (%) (95% CI) 8.2/110 (8.0) (4.4-14.8) 3/20 (15.0) (5.2-36.0) 6/90 (6.7) (3.1-13.8) .21

*Does not include the 22 TIA patients admitted to a CCA.

ySix patients had missing echocardiographic data.

zPatients not admitted to areas with CCM availability were excluded.

NEWLY DIAGNOSED ATRIAL FIBRILLATION AFTER ISCHEMIC STROKE 215

institution. Despite this finding, the variable postproce-

dural stroke or TIA was not significant in the univariate

analysis for NDAF detection (35.0% in patients with

NDAF vs 27.9% in those with sinus rhythm; P5 .35). Con-

sidering this variable’s biological and pathophysiologic

relevance, we included it in the multivariate analysis,

but it still did not reach statistical significance (Table 3).

Identification of NDAF strongly influenced the choice

of secondary prevention strategies in our cohort. Some

patients diagnosed with atherosclerotic carotid disease

were prescribed an oral anticoagulant agent after carotid

endarterectomy, and some patients originally classified as

cryptogenic were finally considered cardioembolic after

NDAF was recognized. Most previous studies were

designed to screen for NDAF in patients cryptogenic

Table 3. Determinants of NDAF in the logistic regression

model

Variable Coefficient OR (95% CI) P

Diabetes mellitus 1.34 3.8 (1.3-11.4) .02

Infarct size .15 mm 1.37 3.9 (1.3-11.8) .01

Adjusted for age, sex, stroke severity (NIHSS score), history of

hypertension, severely reduced left ventricular ejection fraction,

left atrial enlargement, and postprocedural nature of cerebrovascular

events.

ischemic strokes. However, 80% of the cases of NDAF in

the present study were diagnosed in patients with non-

cryptogenic AIS and TIAs, suggesting that NDAF also

should be investigated in patients with AIS or TIA with

other defined etiologies.

This study has several limitations. Because of the

study’s retrospective nature, we were not able to assess

some important variables, such as the results of ECG per-

formed before admission. Not having this information

could have led to underdiagnosis of previous paroxysmal

AF in those patients who could not recall a history of

AF. Most of the neuroimaging studies performed in the

present study were CT scans; the lack of MRI scans with

diffusion-weighted images hampered the reliable assess-

ment of insular involvement and the accurate determina-

tion of the number of lesions, variables that have been

proposed as a predisposing factors for AF after AIS [27].

The distribution of patients to areas with and without

availability of CCM could have generated a selection

bias, because compared with the patients never admitted

to a CCA, the patients admitted to a CCAwere older, had

a higher proportion of ischemic strokes, and had more

severe strokes. Our analysis indicated that neither age

nor stroke severity was a determinant of NDAF, however.

Moreover, the interaction term “stroke severity * infarct

size” did not reach statistical significance in the multivar-

iate logistic regression model.

L.A. SPOSATO ET AL.216

In conclusion, we found a very high rate of NDAF in

patients with cryptogenic and noncryptogenic AIS and

TIA undergoing immediate and prolonged CCM. Diabe-

tes mellitus and infarct size were associated with

NDAF. Immediate admission to a CCA for at least 3

days significantly increased the rate of NDAF detection.

These findings might have important therapeutic implica-

tions affecting a large proportion of patients with AIS and

TIA who could benefit from oral anticoagulation and

antiarrhythmic drugs. This benefit would be plausible

not only for patients with cryptogenic events, but also

for those with other defined stroke mechanisms.

Acknowledgment: We thank Professors Branco Mautner

and Gerardo Bozovich for their critical review of the manu-

script and Drs Analıa Calle, Noelia Pontello, and Francisco

Munoz Giacomelli for data collection.

References

1. Mattle HP. Long-term outcome after stroke due to atrialfibrillation. Cerebrovasc Dis 2003;16(Suppl 1):3-8.

2. EAFTStudyGroup. Secondaryprevention innon-rheumaticatrial fibrillation after transient ischaemic attack or minorstroke. Lancet 1993;342:1255-1262.

3. Lin HJ, Wolf PA, Benjamin EJ, et al. Newly diagnosedatrial fibrillation and acute stroke: The FraminghamStudy. Stroke 1995;26:1527-1530.

4. Jabaudon D, Sztajzel J, Sievert K, et al. Usefulness of am-bulatory 7-day EKG monitoring for the detection of atrialfibrillation and flutter after acute stroke and transientischemic attack. Stroke 2004;35:1647-1651.

5. Kirchhof P, Auricchio A, Bax J, et al. Recommendationsfrom a consensus conference organized by the GermanAtrial Fibrillation Competence NETwork (AFNET) andthe European Heart Rhythm Association (EHRA). EurHeart J 2007;28:2803-2817.

6. Kessler DK, Kessler KM. Is ambulatory electrocardiogra-phy useful in the evaluation of patients with recentstroke? Chest 1995;107:916-918.

7. Piorkowski C, KottkampH, TannerH, et al. Value of differ-ent follow-up strategies to assess the efficacy of circumfer-ential pulmonaryveinablation for the curative treatmentofatrial fibrillation. J Cardiovasc Electrophysiol 2005;16:1286-1292.

8. Caplan LR. Multiple potential risks for stroke. JAMA2000;283:1479-1480.

9. Sacco RL, Ellenberg JH, Mohr JP, et al. Infarcts of undeter-mined cause: The NINDS Stroke Data Bank. Ann Neurol1989;25:382-390.

10. Easton JD, Saver JL, Albers GW, et al. Definition andevaluation of transient ischemic attack. Stroke 2009;40:2276-2293.

11. Schulz KF, Altman DG, Moher D, for the CONSORTGroup. CONSORT 2010 Statement: Updated guidelinesfor reporting parallel group randomised trials. BMJ2010;340:c332.

12. Donnan GA, Norrving B, Bamford JM, et al. Subcorticalinfarction classification and terminology. CerebrovascDis 1993;3:248-251.

13. Lang RM, Bierig M, Devereux RB, et al. Recommenda-tions for chamber quantification: A report from theAmerican Society of Echocardiography’s Guidelinesand Standards Committee and the Chamber Quantifica-tion Writing Group, developed in conjunction with theEuropean Association of Echocardiography, a branch ofthe European Society of Cardiology. J Am Soc Echocar-diogr 2005;18:1440-1463.

14. Kannel WB, Wolf PA, Benjamin EJ, et al. Prevalence, inci-dence, prognosis, and predisposing conditions for atrialfibrillation: Population-based estimates. Am J Cardiol1998;82:2N-9N.

15. Rem JA, Hachinski VC, Boughner DR, et al. Value ofcardiac monitoring and echocardiography in TIA andstroke patients. Stroke 1985;16:950-956.

16. Barthelemy JC, Feasson-Gerard S, Garnier P, et al. Auto-matic cardiac event recorders reveal paroxysmal atrialfibrillation after unexplained strokes or transient ischemicattacks. Ann Noninvasive Electrocardiol 2003;8:194-199.

17. Hornig CR, Haberbosch W, Lammers C, et al. Specificcardiological evaluation after focal cerebral ischemia.Acta Neurol Scand 1996;93:297-302.

18. Schuchert A, Behrens G, Meinertz T. Impact of long-termEKG recording on the detection of paroxysmal atrialfibrillation in patients after an acute ischemic stroke.Pacing Clin Electrophysiol 1999;22:1082-1084.

19. Liao J, Khalid Z, Scallan C, et al. Noninvasive cardiacmonitoring for detecting paroxysmal atrial fibrillationor flutter after acute ischemic stroke: A systematic review.Stroke 2007;38:2935-2940.

20. Tagawa M, Takeuchi S, Chinushi M, et al. Evaluatingpatients with acute ischemic stroke with special referenceto newly developed atrial fibrillation in cerebral embo-lism. PACE 2007;30:1121-1128.

21. Douen AG, Pageau N, Medic S. Serial electrocardio-graphic assessments significantly improve detection ofatrial fibrillation 2.6-fold in patients with acute stroke.Stroke 2008;39:480-482.

22. Tayal AH, Tian M, Kelly KM, et al. Atrial fibrillationdetected by mobile cardiac outpatient telemetry in cryp-togenic TIA or stroke. Neurology 2008;71:1696-1701.

23. Elijovich L, Josephson SA, Fung GL, et al. Intermittentatrial fibrillation may account for a large proportion ofotherwise cryptogenic stroke: A study of 30-day cardiacevent monitors. J Stroke Cerebrovasc Dis 2009;18:185-189.

24. Coumel P. Neural aspects of paroxysmal atrial fibrillation.In: Falk RH, Podrid PJ, eds. Atrial Fibrillation: Mecha-nisms and Management. New York: Raven, 1992:109-125.

25. McLaren A, Kerr S, Allan L, et al. Autonomic function isimpaired in elderly stroke survivors. Stroke 2005;36:1026-1030.

26. Mariscalco G, Engstrom KG. Atrial fibrillation after car-diac surgery: Risk factors and their temporal relationshipin prophylactic drug strategy decision. Int J Cardiol 2008;129:354-362.

27. Cheshire WP, Saper CB. The insular cortex and cardiacresponse to stroke. Neurology 2006;66:1296-1297.