Risk Factors for Thromboembolism in
Patients With Paroxysmal
Atrial Fibrillation
Hiroshi Inoue, MD, and Hirotsugu Atarashi, MD,
for the Research Group for Antiarrhythmic Drug Therapy*
There is some controversy concerning which clinical
characteristics predict thromboembolism and whether
treatment with class I antiarrhythmic drugs reduces
thromboembolim in patients with paroxysmal atrial fi-
brillation (AF). This retrospective, multicenter study was
undertaken to determine risk factor or factors for throm-
boembolism in patients with paroxysmal AF. Seven hun-
dred forty patients with paroxysmal AF (mean age 56
years) without prior thromboembolic events were fol-
lowed retrospectively. Cerebral thromboembolism, in-
cluding transient ischemic attack and embolism of pe-
ripheral arteries, were selected as primary end points.
Independent risk factors were determined with multivar-
iate analysis, and event-free survival curves were esti-
mated. During 3.4-year follow-up period, primary end
points occurred in 55 patients (2.2% per year). Patients
Ifortparoxysmal
remains still controversial whether patients with
atrial fibrillation (AF) have lower risk
thromboembolism compared with those with
chronic AF.1– 8 Large, prospective studies7,8 have
shown that risk for thromboembolism does not differ
between patients with paroxysmal AF and those with
chronic AF. Some studies have tried to identify inde-
pendent risk factor or factors for thromboembolism in
patients with paroxysmal AF.3,9 Corbalàn et al9 iden-
tified history of hypertension and left atrial enlarge-
ment as independent risk factors for systemic embo-
lism. Petersen and Godtfredsen,3 however, showed
that transition from paroxysmal AF to chronic AF was
the only independent risk factor for thromboembo-
lism. If this is so, treatment with antiarrhythmic drugs
to suppress AF or prevent transition from paroxysmal
to chronic AF would reduce the risk for systemic
embolism in patients with paroxysmal AF. Currently,
a prospective study is under way to address this is-
sue.10 To determine risk factors for thromboembolism
in patients with paroxysmal AF and whether class I
antiarrhythmic drugs could reduce the thromboem-
From the Second Department of Internal Medicine, Toyama Medical
and Pharmaceutical University, Toyama; and the First Department of
Internal Medicine, Nippon Medical School, Tokyo, Japan. This study
was supported by a research grant from Eisai Co. Ltd, Tokyo, Japan.
Manuscript received January 20, 2000; revised manuscript received
and accepted April 26, 2000.
Address for reprints: Hiroshi Inoue, MD, Second Department of
Internal Medicine, Toyama Medical and Pharmaceutical University,
2630 Sugitani, Toyama 930 – 0194, Japan.
*See Appendix for a list of investigators.
852 ©2000 by Excerpta Medica, Inc. All rights reserved.
The American Journal of Cardiology Vol. 86 October 15, 2000
with thromboembolism had a higher prevalence of un-
derlying heart disease (p <0.01), less frequent treat-
ment with antiarrhythmic drugs (p <0.01), and received
diuretics more often (p <0.01) compared with patients
without thromboembolism. Age (>65 years, RR 3.33,
p ⴝ 0.0001) and gender (male, RR ⴝ 2, p ⴝ 0.0291)
emerged as predictors of thromboembolism by multivar-
iate analysis with Cox’s proportional hazard model.
Treatment with antiarrhythmic drugs (RR ⴝ 0.57, p ⴝ
0.0578) and aspirin (RR ⴝ 0.52, p ⴝ 0.1094) showed
trends toward reducing thromboembolic risks. It is sug-
gested that elderly men (>65 years) with paroxysmal
AF are at risk for thromboembolism, but the risk tended
to be reduced by treatment with antiarrhythmic drugs
and aspirin. 䊚2000 by Excerpta Medica, Inc.
(Am J Cardiol 2000;86:852– 855)
bolic events, we retrospectively followed ⬎700 pa-
tients with paroxysmal AF.
METHODS
Patient population: The study group consisted of
740 patients with paroxysmal AF (506 men, 234
women; mean age 56 years). The study patients were
recruited from 1,970 patients, including inpatients and
outpatients who had been treated for AF at the ar-
rhythmic clinic of the participating institutions (see
Appendix) from January 1991 to December 1993.11
AF was documented by electrocardiogram in each
patient; it was also noted if each patient spontaneously
converted to sinus rhythm or converted with adequate
treatment. Patients who had prior symptomatic stroke
or systemic embolism at the initial examination were
excluded from the present study. From medical
records, clinical characteristics including underlying
heart disease, New York Heart Association functional
class, diabetes mellitus, cardiothoracic ratio on chest
x-ray, transthoracic echocardiographic findings, and
medication were determined. Among echocardio-
graphic variables, left atrial dimension, left ventricular
end-diastolic dimension, and fractional shortening of
the left ventricle were determined. Because of the
retrospective nature of the study, treatment of patients
was selected ad libitum by physician’s decision. The
last drugs given when events occurred (see the fol-
lowing), or at the end of follow-up period if any event
did not occur, were selected for analyses in the present
study because of simplicity for statistical analyses.
Medication analyzed included warfarin, antiplatelet
0002-9149/00/$–see front matter
PII S0002-9149(00)01105-X
 TABLE 1 Underlying Diseases and Thromboembolism TABLE 2 Antithrombotic Drugs and Thromboembolism
No. of Patients With No. of Patients With
Thromboembolism No. of Patients Thromboembolism (%/year)
No. of Patients (%/year)
None 421 35 (2.4)
No heart disease 253 7 (0.8)† Aspirin* 175 7 (1.2)
Coronary artery disease 92 11 (3.5) Warfarin† 73 5 (2.0)
Systemic hypertension 211 22 (3.0) Ticlopidine 42 2 (1.4)
Mitral valve disease 63 8 (3.7) Others 29 6 (6.1)
Sick sinus syndrome 78 10 (3.7)
Other heart diseases* 57 5 (2.6) *Used alone and in combination with other antithrombotic drugs, except for
Diabetes mellitus 58 8 (4.9) warfarin.
†
Used alone and in combination with other antithrombotic drugs.
*Including cardiomyopathies and congenital heart disease.
†
p ⬍0.01 versus those with underlying heart disease.

TABLE 3 Underlying Heart Diseases and Thromboembolic

Risk in Patients Not Treated With Antithrombotic Drugs
drugs (aspirin, ticlopidine, and others), digitalis, di- No. of Patients With
uretics, and class I antiarrhythmic drugs. Thromboembolism
No. of Patients (%/year)
End points: The primary end points were symptom-
atic cerebral thromboembolism and systemic embo- No heart disease 160 4 (0.7)*
lism. Diagnostic criteria for embolism were clinical Coronary artery disease 36 8 (6.5)
Systemic hypertension 120 16 (3.9)
signs of sudden-onset ischemia in the cerebral arteries Mitral valve disease 29 4 (4.1)
or in the peripheral arteries. Computed tomography Sick sinus syndrome 30 4 (3.9)
was performed in 23% of the patients, and therefore
*p ⬍0.001 versus patients without underlying heart disease.
was not employed as a diagnostic criterion of cere-
brovascular events. It was difficult to distinguish
thrombotic from embolic stroke retrospectively; all
strokes, including transient ischemic attack and cere- TABLE 4 Clinical Characteristics of Patients With and
Without Thromboembolism
bral infarction, were analyzed as primary end points
(i.e., cerebral thromboembolism). Thromboembolism
Statistical analysis: The data are presented as Variable Absent (n ⫽ 685) Present (n ⫽ 55)
mean ⫾ SD. Mean values between 2 groups were
compared with t test, and nonparametric variables Men (%) 68 75
Age (yrs) 56 ⫾ 13 61 ⫾ 12
were compared with chi-square test. The independent Lone AF (%) 36 13‡
risk factors for thromboembolism were determined Diabetes mellitus (%) 7 15
with Cox’s proportional hazard model, and regression Antithrombotic drugs (%) 44 36
variables were estimated with standard maximum Class I antiarrhythmics (%) 51 31‡
Digitalis (%) 35 36
likelihood methods. Explanatory variables included Diuretics (%) 13 26‡
age, gender, coronary artery disease, mitral valve dis- New York Heart Association 82 71
ease, hypertension, sick sinus syndrome, diabetes mel- functional class I (%)
litus, New York Heart Association functional class, Cardiothoracic ratio (%)* 50 ⫾ 7 53 ⫾ 6
Left atrial dimension (mm)† 37 ⫾ 7 37 ⫾ 9
aspirin, warfarin, digitalis, diuretics, and class I anti- Left ventricular end-diastolic 48 ⫾ 7 48 ⫾ 8
arrhythmic drugs. The threshold for stepwise inclu- dimension (mm)†
sion of a specific variable into the model was arbi- Left ventricular fractional 36 ⫾ 9 35 ⫾ 10
trarily set at p ⬍0.20. A p value of ⬍0.05 was re- shortening (%)†
garded as statistically significant. *Data from 664 patients.
†
Data from 508 patients.
‡
p ⬍0.01 versus patients without thromboembolism.
RESULTS
During the mean follow-up period of 3.4 years,
thromboembolism occurred in 55 patients (2.2% per
year). Among these patients, cerebral infarction oc- mg in about 2/3 of patients. Mean daily dosage of
curred in 37 patients, transient ischemic attack in 15, warfarin was 2.5 mg. The thromboembolic risk of
and embolism in the peripheral arteries in 3. The risk patients who did not receive antithrombotic treatment
of thromboembolism varied depending on the under- was related to their underlying heart disease (Table 3).
lying heart disease (Table 1). Patients without any Clinical characteristics of patients with thrombo-
apparent heart diseases (i.e., lone AF) had lower risk embolism and those without thromboembolism are
of thromboembolism compared with those with un- listed in Table 4. Among antiarrhythmic drugs given,
derlying heart disease. Of 740 patients, 421 patients disopyramide was used most frequently (19.2% of 740
did not receive antiplatelet drugs or warfarin, and the patients). The other drugs used were cibenzoline (7%),
remaining 319 received different types of antithrom- pilsicainide (7%), aprindine (6.6%), flecainide (6.2%),
botic drugs (Table 2). Daily dosage of aspirin was 81 propafenone (2.6%), quinidine (2%), and procain-

ARRHYTHMIAS AND CONDUCTION DISTURBANCES/PAROXYSMAL AF AND THROMBOEMBOLISM 853

 TABLE 5 Risk Factors for Thromboembolism Analyzed With
Cox’s Proportional Hazard Model
RR (95% CI) p Value
Age ⱖ65 yrs 3.33 (1.92–5.81) 0.0001
Men 2.00 (1.07–3.72) 0.0291
Class I antiarrhythmics 0.57 (0.32–1.02) 0.0578
Diuretics 1.75 (0.94–3.28) 0.0774
Aspirin 0.52 (0.23–1.16) 0.1094
CI ⫽ confidence intervals.
FIGURE 1. Event-free survival curves predicted with PHREG pro-
cedure. Four groups (A to D) of male patients were used for this
prediction. Group A showed better outcome, whereas group D
showed the worst outcome. AS ⴝ aspirin; AAD ⴝ class I antiar-
rhythmic drugs; DU ⴝ diuretics.
amide (0.8%). Thromboembolism occurred in 12.3%
of 228 patients with paroxysmal AF who had not
received antiarrhythmic drugs, antiplatelet drugs, and
warfarin. This risk was quite similar to that of 380
patients with chronic AF (12.4%) who had not re-
ceived warfarin and antiplatelet drugs in our previous
study,11 from which the patient population of the
present study was selected.
Multivariate regression analysis revealed age and
gender as independent risk factors for thromboembo-
lism in patients with paroxysmal AF (Table 5). Al-
though the p value was ⬎0.05, class I antiarrhythmic
drugs, diuretics, and aspirin showed trends in modifi-
cation in thromboembolic risk of patients with parox-
ysmal AF. Event-free survival curves for thromboem-
bolism in men with paroxysmal AF were estimated
using the proportional-hazards regression procedure
of the SAS package (SAS Institute, Cary, North Caro-
lina) (Figure 1). When treated with class I antiarrhyth-
mic drug and aspirin, younger patients had a better
outcome in terms of stroke and peripheral artery em-
bolism.
Gastrointestinal bleeding occurred in 4 patients:
2 patients not given antithrombotic drugs and 2
treated with aspirin. Cerebral bleeding occurred in a
patient on warfarin. Mortality was quite low; cardiac
death occurred in 2 patients and noncardiac death in
another 2.
854 THE AMERICAN JOURNAL OF CARDIOLOGY姞 VOL. 86
DISCUSSION
The major findings of the present, retrospective
study on a relatively large number of patients with
paroxysmal AF are as follows. First, more than half of
the patients who did not have prior symptomatic
thromboembolism did not receive antithrombotic
treatment. Second, patients without underlying heart
diseases had lower risk for thromboembolism even
when not given antithrombotic treatment. Third, pa-
tients with paroxysmal AF had increased risk for
thromboembolism when they were male, older (ⱖ65
years), and treated with diuretics. In contrast, the
thromboembolic risk tended to be reduced when class
I antiarrhythmic drug or aspirin was given.
It is still controversial whether paroxysmal AF has
a lower incidence of thromboembolism than chronic
AF.1– 8 In the Framingham study,1 the annual rate of
stroke was higher for subjects with chronic AF (5.4%)
compared with those with paroxysmal AF (1.3%).
Petersen and Godtfredsen3 showed similar results. In
contrast, analysis of the pooled data of 5 large pro-
spective studies showed similar risk of stroke between
paroxysmal and chronic AF.12 The reason for different
results might relate to differences in the study design
and clinical characteristics of the study population,
including age, underlying heart disease, and so forth.
A meta-analysis of large trials on AF determined
independent risk factors for thromboembolism in pa-
tients with nonrheumatic AF.12 There were, however,
only limited studies on the risk of thromboembolism
in patients with paroxysmal AF. Petersen amd Godt-
fredsen3 reviewed medical charts of 426 consecutive
patients (median age 66 years) with paroxysmal AF
who had been admitted to their institute from 1940 to
1967. They found that transition from paroxysmal to
chronic AF emerged as the only risk factor for throm-
boembolism, suggesting that patients with paroxysmal
AF might benefit from antiarrhythmic treatment in
terms of prophylaxis of thromboembolism.12
Corbalán et al9 determined risk factors for systemic
embolism in 63 consecutive patients with symptom-
atic, nonvalvular paroxysmal AF. Patients with history
of congestive heart failure, coronary artery disease,
and sick sinus syndrome were excluded from their
analyses. With multiple stepwise logistic regression
analysis, history of hypertension and left atrial en-
largement were identified as independent risk factors
for systemic embolism. The present retrospective
study did not exclude such patients as Corbalán did,
and identified different risk factors from those of the
previous studies.3,9 Recently, Stroke Prevention in
Atrial Fibrillation investigators determined indepen-
dent predictors of ischemic stroke in patients with
intermittent AF who received aspirin.13 Advancing
age, hypertension, and prior stroke emerged as inde-
pendent predictors.
The present, retrospective data indicates antithrom-
botic treatment should be given to older, male patients
with paroxysmal AF, and class I antiarrhythmic drugs
would be effective in reducing thromboembolic risk.
Treatment with diuretics would increase risk for
stroke (RR 1.75, p ⫽ 0.0774), possibly due to de-
OCTOBER 15, 2000
pressed left ventricular function and elevated hemat-
ocrit levels.14,15
APPENDIX
The following investigators participated in the present study. Hokkaido
University: Masayuki Sakurai, MD; Yamagata University: Kozue Ikeda, MD;
University of Tsukuba: Keisuke Kuga, MD; Nihon University: Ichiro Watanabe,
MD, Hiroshi Yagi, MD; Juntendo University: Yuji Nakazata, MD; University of
Tokyo: Yuji Murakawa, MD; Tokyo Medical and Dental University: Fumio
Suzuki, MD; Keio University: Hideo Mitamura, MD; Kanto Central Hospital:
Akira Nozaki, MD; Toho University: Kaoru Sugi, MD; Tokyo Metropolitan
Hiroo Hospital: Harumizu Sakurada, MD; Showa University: Yoichi Kobayashi,
MD; National Nagoya Hospital: Rinya Kato, MD; Toyama Medical and Phar-
maceutical University: Akira Fujiki, MD; Kyoto University: Minoru Horie, MD;
Kagawa Prefectural Shiratori Hospital: Shigenobu Bando, MD; Kyushu Univer-
sity: Akiko Chishaki, MD; Oita Medical University: Tetsunori Saikawa, MD;
Kumamoto University: Ken Okumura, MD.
1. Kannel WB, Abbot RD, Savage DD, McNamara PM. Coronary heart disease
and atrial fibrillation: the Framingham Study. Am Heart J 1983;106:389 –396.
2. Roy D, Marchand E, Gagné P, Chabot M, Cartier R. Usefulness of anticoag-
ulant therapy in the prevention of embolic complications of atrial fibrillation. Am
Heart J 1986;112:1039 –1043.
3. Petersen P, Godtfredsen J. Embolic complications in paroxysmal atrial fibril-
lation. Stroke 1986;17:622– 626.
4. Wiener I. Clinical and echocardiographic correlates of systemic embolization
in nonrheumatic atrial fibrillation. Am J Cardiol 1987;59:177.
ARRHYTHMIAS AND CONDUCTION DISTURBANCES/PAROXYSMAL AF AND THROMBOEMBOLISM
5. Takahashi N, Seki A, Imataka K, Fujii J. Clinical features of paroxysmal atrial
fibrillation. An observation of 94 patients. Jpn Heart J 1981;22:143–149.
6. Shimomura K, Ohe T, Uehara S, Matsuhisa M, Kamakura S, Sato I. Signifi-
cance of atrial fibrillation as a precursor of embolism. Am J Cardiol 1989;63:
1405–1407.
7. The Stroke Prevention in Atrial Fibrillation Investigators. Predictors of throm-
boembolism in atrial fibrillation: I. Clinical features of patients at risk. Ann Intern
Med 1992;116:1–5.
8. The Boston Area Anticoagulation Trial for Atrial Fibrillation Investigators.
The effect of low-dose warfarin on the risk of stroke in patients with nonrheu-
matic atrial fibrillation. N Engl J Med 1990;323:1505–1511.
9. Corbalán R, Arriagada D, Braun S, Tapia J, Huete I, Kramer A, Chávey A.
Risk factors for systemic embolism in patients with paroxysmal atrial fibrillation.
Am Heart J 1992;124:149 –153.
10. The Planning, and Steering Committees of the AFFIRM Study for the NHLBI
AFFIRM Investigators. Atrial fibrillation follow-up investigation of rhythm man-
agement: the AFFIRM Study design. Am J Cardiol 1997;79:1198 –1202.
11. Research Group for Antiarrhythmic Drug Therapy. Atrial fibrillation and
thromboembolism: a multicenter cooperative study (abstr in English). J Cardiol
1998;31:227–238 .
12. Atrial Fibrillation Investigators. Risk factors for stroke and efficacy of
antithrombotic therapy in atrial fibrillation. Analysis of pooled data from five
randomized controlled trials. Arch Intern Med 1994;154:1449 –1457.
13. Hart RG, Pearce LA, Rothbart RM, McAnulty JH, Asinger RW, Halperin JL,
for the Stroke Prevention in Atrial Fibrillation Investigators. Stroke with inter-
mittent atrial fibrillation: incidence and predictors during aspirin therapy. J Am
Coll Cardiol 2000;35:183–187.
14. Tohgi H, Yamanouchi H, Murakami M, Kameyama M. Importance of the
hematocrit as a risk factor in cerebral infarction. Stroke 1978;9:369 –374.
15. Carter C, McGee D, Reed D, Yano K, Stemermann G. Hematocrit and the
risk of coronary heart disease. Am Heart J 1983;105:674 – 679.
855