Independent Risk Factors for Atrial

Fibrillation in a Population-Based Cohort

The

Framingham

Emelia J.

Benjamin, MD, ScM; Daniel Levy, MD; Sonya M. Vaziri, MD, MPH; Ralph B. D'Agostino, PhD;
Belanger, MA; Philip A. Wolf, MD

Albert J.

Heart

Study

Objective.\p=m-\Todetermine the independent risk factors for atrial fibrillation.

Design.\p=m-\Cohortstudy.
Setting.\p=m-\TheFramingham Heart Study.
Subjects.\p=m-\Atotal of 2090 men and 2641 women members of the original cohort, free of a history of atrial fibrillation, between the ages of 55 and 94 years.
Main Outcome Measures.\p=m-\Sex-specificmultiple logistic regression models to
identify independent risk factors for atrial fibrillation, including age, smoking, diabetes, electrocardiographic left ventricular hypertrophy, hypertension, myocardial infarction, congestive heart failure, and valve disease.
Results.\p=m-\Duringup to 38 years of follow-up, 264 men and 298 women developed atrial fibrillation. After adjusting for age and other risk factors for atrial fibrillation, men had a 1.5 times greater risk of developing atrial fibrillation than women.
In the full multivariable model, the odds ratio (OR) of atrial fibrillation for each decade of advancing age was 2.1 for men and 2.2 for women (P<.0001). In addition,
after multivariable adjustment, diabetes (OR, 1.4 for men and 1.6 for women), hypertension (OR, 1.5 for men and 1.4 for women), congestive heart failure (OR, 4.5
for men and 5.9 for women), and valve disease (OR, 1.8 for men and 3.4 for women)
were significantly associated with risk for atrial fibrillation in both sexes. Myocardial
infarction (OR, 1.4) was significantly associated with the development of atrial
fibrillation in men. Women were significantly more likely than men to have valvular
heart disease

as a

risk factor for atrial fibrillation. The multivariable models

were

largely unchanged after eliminating subjects with valvular heart disease.

Conclusion.\p=m-\Inaddition to intrinsic cardiac causes such as valve disease and

congestive heart failure, risk factors for cardiovascular disease also predispose to
atrial fibrillation. Modification of risk factors for cardiovascular disease may have the
added benefit of diminishing the incidence of atrial fibrillation.
(JAMA. 1994;271:840-844)

From The Framingham (Mass) Heart Study (Drs

Benjamin, Levy, Vaziri, D'Agostino, and Wolf and Mr
Belanger); the Cardiology Section, Boston (Mass) City
Hospital, Boston University School of Medicine (Dr
Benjamin); National Heart, Lung, and Blood Institute,
Bethesda, Md (Dr Levy); Department of Mathematics,
Boston University (Dr D'Agostino and Mr Belanger);
Departments of Preventive Medicine (Drs Benjamin,
Levy, and Wolf) and Neurology (Dr Wolf), Boston University School of Medicine; and Divisions of Cardiology
and Clinical Epidemiology, Beth Israel Hospital, Boston
(Dr Levy).
Presented in part at the 33rd Annual Conference on
Cardiovascular Disease Epidemiology, American
Heart Association, Santa Fe, NM, March 19, 1993.
Reprint requests to Framingham Heart Study, 5
Thurber St, Framingham, MA 01701 (Dr Benjamin).

ATRI AL fibrillation occurs with increas

ing frequency as people age and is re

sponsible for substantial morbidity and
mortality.1 Present in only 0.5% of 50- to
59-year-old subjects, the lifetime preva
lence of atrial fibrillation is nearly 9%
among 80- to 89-year-old subjects.2 Non-

valvular atrial fibrillation is associated

with approximately a threefold to five
fold increased risk for stroke after ad
justing for other stroke risk factors; by
the ninth decade of life, atrial fibrilla
tion is the most important risk factor for

Downloaded From: http://jama.jamanetwork.com/ by a Rijksuniversiteit Groningen User on 09/11/2014

stroke.2 Furthermore, atrial fibrillation

has a deleterious impact on longevity,3

with an approximate doubling of all-

mortality.4
successfully prevent atrial fibril
lation, physicians must understand the
factors that predispose to its develop
ment. Earlier reports have emphasized
cause

To

valvular heart

disease, coronary heart

disease, congestive heart failure, and hy
pertensive heart disease as risk factors
for atrial fibrillation.1'412 However, these

studies lacked multivariable analyses

and thus were limited in their ability to
identify independent contributors to the
development of atrial fibrillation. The
purpose of this study was to examine
the risk factors for atrial fibrillation in
subjects of the Framingham Heart
Study, for whom nearly 40 years of fol
low-up are available.
METHODS

Study Population
The study is based on the original
cohort of the Framingham Heart Study,

which

was

initiated in 1948 with sub

jects aged 28 to 62 years to prospectively observe the risk factors for car
diovascular disease. Details of the study
design have been published previously.13

Routine biennial examinations were per

formed on the original sample of 5209
men and women. All subjects gave in
formed consent. The current investiga
tion was based on 38 years of follow-up.
Because atrial fibrillation was rare in
younger subjects and there were few
subjects over 94 years old, analysis was
restricted to subjects aged 55 to 94 years
at baseline. In addition, the 18 subjects
with atrial fibrillation at examination 1
were excluded.

Clinical Variables

At each Framingham Heart Study

clinic examination, subjects' medical his
tories, physical examinations, and elec
trocardiograms were obtained to ascer
tain symptoms and findings suggestive
of cardiovascular disease. The records
of all interim hospitalizations for car
diovascular disease were sought and sub
sequently reviewed. Subjects were con
sidered to have had atrial fibrillation if
atrial fibrillation or atrial flutter was
verified on review of the electrocardio
gram by one of two Framingham Heart

Study cardiologists. Electrocardiograms

from the following sources were evalu
ated: routine Framingham Heart Study
clinic examinations, hospital records, and
outside physician records. A separate
analysis of subjects with atrial fibrilla
tion present on routine Framingham
Heart Study clinic electrocardiograms
was undertaken to examine the risk fac
tors for nonacute atrial fibrillation, and
to determine if the inclusion of electro

cardiograms performed during hospital-

ization resulted in ascertainment bias.

Potential risk factors were assessed
at clinic examinations. Myocardial in
farction and congestive heart failure
were determined according to previously
published criteria14 by a panel of three
physicians after review of Framingham
Heart Study and outside hospital and
physician records. Subjects were clas
sified as having myocardial infarction or
congestive heart failure if the condition
was diagnosed at or prior to the day
atrial fibrillation was documented. Valve
disease was defined as any diastolic mur
mur or at least a grade 3/6 systolic mur
mur on Framingham Heart Study physi
cal examination. Because echocardiography was unavailable for the first 30
years of the study, it was not utilized to
establish the diagnosis of valve disease.
Hypertension was defined for each ex
amination as a systolic blood pressure of
at least 160 mm Hg or a diastolic blood
pressure of at least 95 mm Hg on each
of two readings by Framingham Heart
Study physicians, or the use of antihypertensive medication. Diabetes mellitus was diagnosed if the subject had a
fasting blood glucose level of 7.77 mmol/L
or higher (>140 mg/dL); a random, nonfasting blood glucose level of 11.11
mmol/L or higher (>200 mg/dL); or was
taking insulin or an oral hypoglycmie

agent at the current examination or at

any prior Framingham Heart Study ex

amination. Forced expiratory volume in

1 second (FEVj) was used as an indica
tor of lung function. Alcohol use was
ascertained by self-report and catego
rized as ounces of ethanol consumed per
week. Body mass index was calculated

as the weight (in kilograms) divided by

the height (in meters), squared. Electrocardiographic left ventricular hyper
trophy was considered present if the
subject fulfilled voltage criteria for hy
pertrophy and manifested lateral repolarization changes.15 Because electrocardiographic left ventricular hypertrophy
cannot be interpreted in the setting of
left bundle-branch block, subjects were
excluded from the multivariable analy
ses if left bundle-branch block was

present.
Statistical

Analyses
Logistic regression based on pooled
biennial person-examination data was

used to evaluate the association of po

tential risk factors with the development
of atrial fibrillation.16,17 (Each examina
tion with its 2-year follow-up period was
considered a separate person-examina
tion.) At each biennial examination, all
subjects aged 55 to 94 years were evalu
ated for a history of atrial fibrillation.
Incident atrial fibrillation was defined as
the occurrence of atrial fibrillation docu
mented by electrocardiogram up to the
time of the next examination. Putative
risk factors were measured at each bi
ennial examination. Missing data were
imputed by substituting the most recent
values as long as they were obtained
within the two preceding examinations;
this occurred for 10% of hypertension,
10% of left ventricular hypertrophy, and
25% of smoking variables. Potential clini
cal risk factors for atrial fibrillation were
examined first by sex-specific bivariate
models adjusting for age. Variables that
were significant in the bivariate models
were entered into sex-specific multivari
able models. Results of the regressions
were expressed as odds ratios (with 95%
confidence intervals) and population-at
tributable risks: [prevalence(odds ratio
l)]/[prevalence(odds ratio 1) + l].18
To evaluate the interaction of sex with
the risk factors, the data for men and
women were combined and sex-inter
action terms were examined for all po
tential risk factors. Backwards elimina
tion was employed to remove all non
significant interactions (P>.05 was used
for elimination).
-

RESULTS
Characterization of Subjects
During up to 38 years of follow-up of
2090 men (who contributed 18 245 personexaminations of follow-up) and 2641
women (who contributed 26 277 personexaminations of follow-up) with no his
tory of atrial fibrillation, there were 562
incident cases of atrial fibrillation. Among
subjects with atrial fibrillation, 298 (53%)
were women and 264 (47%) were men.

Downloaded From: http://jama.jamanetwork.com/ by a Rijksuniversiteit Groningen User on 09/11/2014

Incidence of atrial fibrillation, Framingham Heart

Study 38-year follow-up. Darker bars indicate men;
and lighter bars, women. Vertical axis indicates bi
ennial rate per 1000 person-examinations (approxi
mately 2000 person-years).
The biennial incidence of atrial fibrilla
tion increased with age (Figure), rang
ing from 6.2 and 3.8 cases per 1000 personexaminations in men and women, respec
tively, aged 55 to 64 years, to 75.9 and
62.8 cases per 1000 person-examinations
in men and women aged 85 to 94 years.
Table 1 shows the sex-specific, age-ad
justed characteristics of subjects accord
ing to incident atrial fibrillation status
and the age-adjusted odds ratios of de
veloping atrial fibrillation for each vari
able. Subjects who developed atrial fi
brillation were older and more likely to
have diabetes mellitus, left ventricular

hypertrophy, hypertension, myocardial

infarction, congestive heart failure, and
valvular heart disease than subjects with
out atrial fibrillation. Body mass index (a
measure of obesity) and alcohol use were
not significantly related to the develop
ment of atrial fibrillation in the age-ad
justed models; hence, they were not in
cluded in the multivariable models.

Multivariable Analyses
In both sexes, age, diabetes, hyper
tension, congestive heart failure, and val
vular heart disease were significant pre
dictors of atrial fibrillation (Table 2).
For each advancing decade of age the
odds ratio of developing atrial fibrilla
tion was 2.1 in men and 2.2 in women.
Congestive heart failure was associated
with an increased risk of developing
atrial fibrillation with an odds ratio of
4.5 in men and 5.9 in women. The pres
ence of valvular heart disease was as
sociated with an increased risk of de
veloping atrial fibrillation (odds ratio of
1.8 in men and 3.4 in women). Myocar
dial infarction was significantly associ
ated with the development of atrial fi
brillation in men (odds ratio, 1.4).
The population-attributable risk analy
ses are presented in Table 3. Populationattributable risk incorporates both the

Table 1.Risk Factors for

Developing Incident Atrial Fibrillation (AF)*

Ment

Variable

Age (mean), y
Cigarette smoking,

AF
72

Woment
Odds Ratio

(95% CI)

65

33.7

Diabetes, %
ECG LV hypertrophy, %
Hypertension. %
Myocardial infarction, %
Congestive heart failure, %
Valve disease, %
Body mass index (mean)
Ethanol, oz/wk

No AF

16.3

10.2

10.7

4.4

44.1

30.9

25.5

13.0

3.2
16.7

6.7

26.2

26.0

5.4

5.1

Table 2.Multivariable Risk of Developing Incident

Atrial Fibrillation (2-Year Pooled Logistic Regression)

1.0(0.8-1.4)
1.7||(1.2-2.3)
3.01) (1.9-4.8)
1.811(1.4-2.3)
2.2I1 (1.6-2.8)
6.111(4.5-8.4)
2.211(1.6-3.1)
1.03(0.99-1.06)
1.01 (0.99-1.03)

AF

No AF

75

66

28.5

23.4
7.5
3.8
40.7
4.6

Variable

29.5

8.7

26.0

25.7

1.5

1.8

0.95(0.89-1.02)

13.6
51.7
13.0

2.9

expressed as age-adjusted % (categorical data) and means (continuous data); odds ratios are
age-adjusted from 2-year pooled logistic regression, subjects with compared with those without AF. CI indicates
confidence interval; ECG, electrocardiographs; and LV, left ventricular.
tThere were 264 Incident cases of AF among men In 18 245 follow-up person-examinations.
iThere were 298 Incident cases of AF among women in 26 277 follow-up person-examinations.
P=s.05.
IIPs.01.
IJPs.0001.
"Variables

are

odds ratio of the risk factor and the risk

factor's prevalence in the study popula
tion. It reflects the percentage of atrial
fibrillation cases that potentially would
be reduced if a causal risk factor could be
eliminated from the population. The at
tributable risk of congestive heart fail
ure for atrial fibrillation was 10% in men
and 12% in women. The attributable risk
for atrial fibrillation resulting from myo
cardial infarction was greater in men (5%)
than in women (1%), while the attribut
able risk from valve disease was approxi
mately threefold higher in women (18%)
than in men (5%). In part because of the
high prevalence of hypertension in the
elderly (31% of men, 40% of women), its
population-attributable risk for atrial fi
brillation was 14% in both men and
women. Electrocardiographic left ven
tricular hypertrophy and diabetes each
accounted for less than 5% of the attrib
utable risk for atrial fibrillation.
Impact of Sex.After adjusting for
age, smoking, diabetes, left ventricular

hypertrophy, hypertension, myocardial

infarction, congestive heart failure, and

valvular heart disease, men had a 1.5

times greater likelihood of developing
atrial fibrillation than women (95% con
fidence interval, 1.3 to 1.8; P<.0001). To
explore potential sex differences in the
risk factors for atrial fibrillation, the
models were reanalyzed with sex-inter
action terms. While the models for men
and women were similar, valvular heart
disease was a significantly (P<.01) more
potent risk factor for the development
of atrial fibrillation in women than in
men.

Nonacute Atrial Fibrillation.To as

risk factors for nonacute atrial fi
brillation, a separate multivariable
analysis identified the risk factors for
atrial fibrillation diagnosed on routine
Framingham Heart Study clinic eleesess

trocardiograms (Table 4).

While the
number of atrial fibrillation cases was
much smaller, and hence some of the
risk factors no longer achieved statis
tical significance, the odds ratios of the
clinical risk factors for atrial fibrillation
were similar to those found in the entire
sample, except for myocardial infarc
tion, which had an odds ratio of less than
one.

Additional

(95% Confidence Interval)

(95% CI)

1.4 (1.0-2.0)
2.1H(1.5-2.8)
3.811 (2.6-5.6)
1.7H(1.3-2.2)
2.411(1.7-3.4)
8.111(6.1-10.7)
3.611 (2.8-4.6)
1.02(1.00-1.05)

15.5

Odds Ratio

Odds Ratio

Analyses

Nonvalvular Atrial Fibrillation.

Since there is little doubt that valvular
heart disease predisposes to atrial fi
brillation, Table 5 presents multivari
able models restricting the analysis to
subjects without clinically apparent
valve disease (thereby eliminating 136
cases of atrial fibrillation). In men and
women without valvular heart disease,
diabetes was no longer significantly as
sociated with the development of atrial
fibrillation, while age, hypertension, and
congestive heart failure remained sig
nificant predictors. In women without
valvular heart disease, myocardial in
farction also was significantly associated
with the development of atrial fibrilla
tion.
Lung Disease.Because of the sug
gestion of a relation between lung dis
ease and atrial fibrillation,56 the asso
ciation of FEVi and incident atrial fi
brillation was examined. When FEVj
was added to the full multivariable mod
els, it was not significantly associated
with atrial fibrillation in either sex (odds
ratio, 1.00/0.10 L in both sexes).
COMMENT
The incidence of atrial fibrillation ap
proximately doubled for every 10-year
increment in age in the Framingham
Heart Study cohort. In both sexes, car
diovascular disease risk factors (ie, dia-

Downloaded From: http://jama.jamanetwork.com/ by a Rijksuniversiteit Groningen User on 09/11/2014

I-1
Men*
Womenf

(1.8-2.5)
1.1(0.8-1.5)
1.4 (1.0-2.0)

2.2 (1.9-2.6)

1.4(0.9-2.4)
1.5|| (1.2-2.0)
1.4 (1.0-2.0)

1.3(0.9-2.1)
1.4 (1.1-1.8)
1.2(0.8-1.8)

4.5$ (3.1-6.6)
1.811(1.2-2.5)

5.9$ (4.2-8.4)
3.4$ (2.5-4.5)

2.1

Age (/10 y)
Cigarette smoking

*Atrial fibrillation was diagnosed in 226 men in 16 529

follow-up person-examinations.
tAtrial fibrillation was diagnosed in 244 women in 23 763
follow-up person-examinations.
$P==.0001.
Ps.05.

Diabetes

Elect rocardiographic
left ventricular

hypertrophy
Hypertension
Myocardial infarction
Congestive heart
failure
Valve disease

1.4(1.0-2.0)
1.6|| (1.1-2.2)

IIPS.01.

betes and hypertension) and structural

cardiac disease (ie, congestive heart fail
ure and valvular heart disease) also were
risk factors for atrial fibrillation.
In the present study, men were 1.5
times more likely to develop atrial fi
brillation than women, even after ad
justing for other risk factors for atrial
fibrillation. The reason for the greater
susceptibility of men to atrial fibrilla
tion remains unclear. With rare excep
tion,6 most prior publications also have
noted that men are at greater risk for
atrial fibrillation than women.1'6,9111920
In a retrospective hospital-based analy
sis,5 men were 1.2 times more likely to
have atrial fibrillation than women. Men
were noted to have a 5.4 times greater
risk of having atrial fibrillation than
women in the Coronary Artery Surgery
Study registry of patients catheterized
for suspected coronary artery disease.20
In a population-based study, men were
2.7 times more likely to have atrial fi
brillation than women; however, the au
thors did not examine whether the
higher prevalence of atrial fibrillation in
men merely reflected the greater bur
den of atrial fibrillation risk factors in
men.11
In the current study, women were
significantly more likely than men to
have valvular heart disease as a risk
factor for atrial fibrillation, despite the
similar prevalence of valvular heart dis
ease in men and women (7% and 9%,
respectively, in the total sample). With
the exception of that by Stroud et al,9
other studies have found a similar fe
male predominance in the cases of atrial
fibrillation associated with valvular heart
disease, although these studies did not
use multivariable analyses.5611 The bio
logic mechanism for the greater suscep
tibility of women with valvular heart
disease to the development of atrial fi-

brillation may be secondary to the fe

male predisposition to rheumatic mitral
stenosis21 and the propensity of mitral
stenosis to cause atrial fibrillation more
frequently than other valve lesions.
Previous studies have identified car
diac and noncardiac prcipitants of atrial
fibrillation. Potential cardiac etiologies
of atrial fibrillation have been noted to
include valvular heart disease (rheu
matic, myxomatous, senile calcific, and
congenital valve abnormalities)511; myo
cardial disease (acute myocardial infarc
tion and ischemie heart disease,10,12 con
gestive heart failure,11 hypertensive
heart disease,10 myocarditis,5"7 and hy
pertrophie cardiomyopathy22); conduc
tion system disease (Wolff-ParkinsonWhite syndrome)6; congenital heart dis
ease (atrial septal defect, patent ductus
arteriosus, and so forth)6; and pericardial disease6 (including pericarditis and
constriction). The noncardiac etiologies
of atrial fibrillation have been reported
to include thyrotoxicosis,510 alcohol
use,512 severe infections,5 and pulmonary
pathology (lung cancer, pulmonary in
fections, pulmonary embolism, and
chronic lung disease).58 While clinical
series provide insights into the poten
tial risk factors for atrial fibrillation in
the individual patient, they are limited
in their generalizability to the broader
population for several reasons. Most se
ries are retrospective and hospital-based
and may select for sicker patients with
atrial fibrillation.510 Furthermore, be
cause these studies are generally de
scriptive and lacking in antecedent clini
cal risk factor information, they are
unable to control for confounding risk
factors.510 Two population-based stud
ies have examined atrial fibrillation; how
ever, their analyses were limited by
small case numbers and a failure to con
trol for confounding variables.11,12 The
current study population included a large
number of cases of atrial fibrillation docu
mented both in the hospital and in an

ambulatory setting during nearly 40

years of follow-up. In addition, potential
risk factors were assessed prospectively,
thereby reducing ascertainment bias.
Prior Framingham Heart Study re
ports have explored risk factors for atrial
fibrillation, noting that valvular heart
disease, cardiac failure, hypertensive
heart disease, and coronary heart dis
ease were predictors of atrial fibrilla
tion.1,4 However, these investigations

based on far fewer cases and

younger subjects (with a mean age 10
years younger),1,4 used case-control
methodology (matching on age and sex),4
and did not utilize multivariable analy
were

ses.1,4 The main discrepancy between the

current and

prior Framingham Heart

Study analyses was in the estimated im-

Table 3.Population-Attributable Risk (Prevalence

of Variable In %) of the Risk Factors Associated
With Atrial Fibrillation*
Variable

Men
2

Cigarette smoking
Diabetes

Electrocardiographic left
ventricular hypertrophy
Hypertension
Myocardial infarction
Congestive heart failure
Valve disease

(34)
4 (10)
2

(4)
(31)
5(13)
10(3)
5(7)

14

person-examination
the

same as

Odds Ratio*

Women
8
4

(95% Confidence Interval)

(24)
(8)

1 (4)
14(40)
1(4)
12(3)
18(9)

*Two-year pooled logistic regression. Follow-up

are

Table 4.Multivariable Risk of Nonacute Atrial

Fibrillation Detected on Routine Framingham Heart
Study Clinic Electrocardiogram

and number with atrial fibrillation

in Table 2.

pact of congestive heart failure on atrial

fibrillation risk in men. In a prior analy

sis, cardiac failure was associated with
an age-adjusted risk ratio of 17.5 in men
and 5.7 in women.4 In the current analy
sis, after full multivariable adjustment,
the odds ratio of developing atrial fi
brillation with congestive heart failure
was 4.5 in men and 5.9 in women. Con
gestive heart failure frequently occurs
in the setting of valvular heart disease,
myocardial infarction, hypertension, and
left ventricular hypertrophy, so the im
pact of heart failure in the current mul
tivariable analyses may have been di
minished by its correlation with other
variables in the models. The attenuated
risk of congestive heart failure in the
current study, as opposed to the earlier
studies,14 also may reflect the aging of
the cohort. Because the current analy
ses used full multivariable models with
3931 to 4644 additional cases of atrial
fibrillation, the current analyses pro
vided more stable and reliable estimates
of the impact of a variety of risk factors
on the incidence of atrial fibrillation.
Limitations of the Study
The ascertainment of atrial fibrilla
tion is problematic, since some cases of
atrial fibrillation may remain undetec
ted if they were brief or not severe. For
the principal analyses, subjects with
atrial fibrillation on either Framingham
Heart Study or outside electrocardio
grams were included. Using hospitalization electrocardiograms for the diag
nosis of atrial fibrillation selected for
sicker subjects with more symptomatic
atrial fibrillation. Conversely, analyses
restricted to subjects with atrial fibril
lation on routine clinic examination (nonacute atrial fibrillation) selected for
chronic and more benign atrial fibrilla
tion cases (subjects with fatal or com
plicated atrial fibrillation would be un
available for follow-up). Nevertheless,
the results from the multivariable mod
els of all atrial fibrillation cases and atrial
fibrillation cases documented on Fram
ingham Heart Study clinic electrocar
diograms were comparable, except for
myocardial infarction.

Downloaded From: http://jama.jamanetwork.com/ by a Rijksuniversiteit Groningen User on 09/11/2014

Variable

Age(/10y)
Cigarette smoking
Diabetes

Electrocardiographic

I-1
Ment
Women$

1.9 (1.5-2.5)
1.3(0.8-2.0)
1.7|| (1.1-2.8)

1.7 (1.4-2.1)
1.1(0.7-1.8)
1.4(0.9-2.4)

1.9(1.0-3.8)
1.2 (0.8-1.8)
0.6(0.3-1.1)

1.2(0.6-2.3)
1.3 (0.9-1.9)
0.7(0.3-1.3)

4.6 (2.6-8.2)
1.7|| (1.0-2.8)

5.4 (3.3-8.9)
3.9 (2.6-5.9)

left ventricular

hypertrophy
Hypertension
Myocardial infarction
Congestive heart
failure
Valve disease

*Odds ratios are based on multiple logistic regression

analyses

of

2-year pooled

observations.

tNonacute atrial fibrillation

was diagnosed in 109

follow-up person-examinations.
tNonacute atrial fibrillation was diagnosed in 119
women in 23929 person-examinations.
men

In 16 694

P==.0001.
IIPs.05.

Table 5.Multivariable Risk of Atrial Fibrillation in

Subjects Without Clinical Valve Disease*
Odds Ratio

(95% Confidence Interval)

Variable

Age (/10 y)
Cigarette smoking
Diabetes

Electrocardiographic

Woment

Menf
2.2 (1.8-2.7)
1.0 (0.7-1.4)
1.1(0.8-1.7)

2.6 (2.1-3.2)
1.5 (1.0-2.2)
1.5(1.0-2.3)

1.6(0.9-3.0)
1.6|| (1.2-2.2)
1.4(1.0-2.1)

1.7|| (1.2-2.4)
2.11| (1.3-3.4)

5.0 (3.2-7.8)

4.3 (2.7-7.1)

left ventricular

hypertrophy
Hypertension
Myocardial infarction
Congestive heart
failure

1.2(0.6-2.4)

*Two-year pooled logistic regression.

t Atrial fibrillation was diagnosed in 180 men in 15 421

follow-up person-examinations.
tAtrial fibrillation was diagnosed in 154 women in
21 627 person-examinations.
P==.0001.

IIP.01.

Because of the potential for misclassification, chronic and paroxysmal atrial

fibrillation were combined as end points,
despite plausible differences in etiology.
Another potential area of misclassifica-

tion was in the clinical ascertainment of

valvular heart disease. This study be
gan over 40 years ago, prior to the ad
vent of echocardiography. Echocardiographic predictors of atrial fibrillation
have been reported in another Framing
ham Heart Study investigation.23 A fur
ther potential limitation of the study is
its generalizability. The study sample
was largely white, and the age was re
stricted to subjects 55 to 94 years of age.
The findings may not be generalizable
to other racial groups or younger or older

subjects.
Finally, the lack of statistical signifi
cance of several suspected risk factors
for atrial fibrillation should not be con
strued to mean that the factors are not
possible risk factors in the individual
patient. For example, myocardial in-

farction was not shown to be a signifi

cant risk factor for atrial fibrillation in

Clinical

in the whole sample (Table 2).

Subjects with myocardial infarction may
develop congestive heart failure that in
turn precipitates atrial fibrillation; be
cause congestive heart failure'and myo

The incidence of atrial fibrillation in

creases dramatically with age and is a
significant source of disability and death
in the elderly. The treatment of atrial
fibrillation with anticoagulant or antiarrhythmic medications may diminish
the symptoms and complications asso
ciated with it.2528 However, anticoagu
lant and antiarrhythmic agents are fre

women

cardial infarction are colinear, adjusting

for congestive heart failure may have
attenuated the known association be
tween myocardial infarction and atrial
fibrillation. In another instance, binge
alcohol consumption may precipitate
atrial fibrillation in susceptible individu
als.24 The lack of an association between
alcohol consumption and atrial fibrilla
tion may reflect the relatively few in
dividuals affected by "holiday heart," or
the extreme alcohol use required to pre
cipitate atrial fibrillation.

Implications

quently contraindicated,2728 particularly

in the elderly, and their use is associ
ated with significant morbidity.29 31 The
safest and most effective means of pre
fi
brillation is to prevent its occurrence.
Multivariable analyses demonstrate that

venting the complications of atrial

age,

diabetes, hypertension, congestive

failure, and valve disease are in-

heart

dependent risk factors for atrial fibril

lation in both sexes. If these precursors

of atrial fibrillation contribute to the in
cidence of atrial fibrillation in a causal
manner, the analyses of population-at
tributable risk suggest that modifica
tion of cardiovascular disease risk fac
tors and the prevention of structural
cardiac disease may have the added ben
efit of diminishing the incidence of atrial
fibrillation.
This study was supported in part through Na
tional Institutes of Health/National Heart, Lung,
and Blood Institute contract NO1-HC-38038 and by
National Institute of Neurological Disorders and
Stroke grant 2-R01-NS-17950-11.
The authors greatly appreciate the contribu
tions of Rodney M. Falk, MD, and David M. Pollak
for reviewing the manuscript; William Kannel, MD,
for insight into the analyses; and Mary Healy, Lois
Abel, Karen Mutalik, and Deborah Dumphy for as
sistance in the collection of atrial fibrillation cases.

References
1. Kannel WB, Abbott RD, Savage DD, McNamara
PM. Coronary heart disease and atrial fibrillation:
The Framingham Study. Am Heart J. 1983;106:
389-396.
2. Wolf PA, Abbott RD, Kannel WB. Atrial fibrillation as an independent risk factor for stroke: The
Framingham Study. Stroke. 1991;22:983-988.
3. Gajewski J, Singer RB. Mortality in an insured
population with atrial fibrillation. JAMA. 1981;245:
1540-1544.
4. Kannel WB, Abbott RD, Savage DD, McNamara
PM. Epidemiologic features of chronic atrial fibrillation: the Framingham Study. N Engl J Med. 1982;
306:1018-1022.
5. Aberg H. Atrial fibrillation: a review of 463 cases
from Philadelphia General Hospital from 1955 to
1965. Acta Med Scand. 1968;184:425-431.
6. Davidson E, Weinberger I, Rotenberg Z, Fuchs
J, Agmon J. Atrial fibrillation: cause and time of
onset. Arch Intern Med. 1989;149:457-459.
7. GodtfredsenJ. Atrial fibrillation: course and prognosis\p=m-\afollow-up study of 1212 cases. In: Kulbertus HE, Olsson SB, Schlepper M, eds. Atrial Fibrillation. M\l=o"\lndal, Sweden: AB Hassle; 1982:134\x=req-\
145.
8. McEachern D, Baker BM. Auricular fibrillation:
its etiology, age incidence and production by digitalis therapy. Am J Med Sci. 1932;183:35-48.
9. Stroud WD, Laplace LB, Reisinger JA. The etiology, prognosis and treatment of auricular fibrillation. Am J Med Sci. 1932;183:48-60.
10. Sawyer CG, Bolin LB, Stevens EL, Daniel LB,
O'Neil NC, Hayes DM. Atrial fibrillation: its etiology, treatment and association with embolization.
South Med J. 1958;51:84-93.
11. \l=O"\nundarson PT, Thorgeirsson G, Jonmundsson
E, Sigfusson N, Hardarson T. Chronic atrial fibrillation\p=m-\epidemiologicfeatures and 14-year follow\x=req-\
up: a case control study. Eur Heart J. 1987;8:521-527.
12. Lake FR, Cullen KJ, de Klerk NH, McCall MG,
Rosman DL. Atrial fibrillation and mortality in an
elderly population. Aust N Z J Med. 1989;19:321-326.

13. Dawber TR, Meadors GF, Moore FE. Epidemiological approaches to heart disease: the Framingham Study. Am J Public Health. 1951;41:279-286.

14. Shurtleff D. Some characteristics related to the

incidence of cardiovascular disease and death:
Framingham Study, 18-year follow-up. In: Kannel
WB, Gordon T, eds. The Framingham Study: An

Epidemiological Investigation of Cardiovascular

Disease. Washington, DC: Dept of Health, Education, and Welfare; 1974: section 30. DHEW publication NIH 74-599.
15. Kannel WB, Gordon T, Offutt D. Left ventricular hypertrophy by electrocardiogram: prevalence,
incidence, and mortality in the Framingham Study.
Ann Intern Med. 1969;71:89-105.
16. Cupples LA, D'Agostino RB, Anderson K, Kannel WB. Comparison of baseline and repeated measure covariate techniques in the Framingham Heart
Study. Stat Med. 1988;7:205-218.
17. D'Agostino RB, Lee ML, Belanger AJ, Cupples
LA, Anderson K, Kannel WB. Relation of pooled
logistic regression to time dependent Cox regression analysis: the Framingham Study. Stat Med.

1990;9:1501-1515.

18. Walker SD. The estimation and interpretation

of attributable risk in health research. Biometrics.

1976;32:829-849.

19. Sherman DG, Goldman L, Whiting RB, Jurgensen K, Kaste M, Easton D. Thromboembolism
in patients with atrial fibrillation. Arch Neurol.

1984;41:708-710.

20. Cameron A, Schwartz MJ, Kronmal RA, Kosinski AS. Prevalence and significance of atrial fibrillation in coronary artery disease (CASS Registry). Am J Cardiol. 1988;61:714-717.
21. Wood P. An appreciation of mitral stenosis,
part I: clinical features. BMJ. 1954;1:1051-1063.
22. Robinson K, Frenneaux MP, Stockins B,
Karatasakis G, Poloniecki JD, McKenna WJ. Atrial
fibrillation in hypertrophic cardiomyopathy: a longitudinal study. J Am Coll Cardiol. 1990;15:1279\x=req-\
1285.

Downloaded From: http://jama.jamanetwork.com/ by a Rijksuniversiteit Groningen User on 09/11/2014

SM, Larson MG, Benjamin EJ, Levy D.

Echocardiographic predictors ofnonrheumatic atrial
fibrillation: Framingham Heart Study. Circulation.
23. Vaziri

1994;89:724-730.

24. Ettinger PO, Wu CF, De la Cruz C, Weisse AB,

Ahmed SS, Rega TJ. Arrhythmias and the 'holiday
heart': alcohol associated cardiac rhythm disorders.
Am Heart J. 1978;95:555-562.
25. Petersen P, Godtfredsen J, Boysen G, Andersen ED, Andersen B. Placebo-controlled, randomized trial of warfarin and aspirin for prevention of
thromboembolic complications in chronic atrial fibrillation: the Copenhagen AFASAK Study. Lancet. 1989;1:175-179.
26. The Boston Area Anticoagulation Trial for Atrial
Fibrillation Investigators. The effect of low-dose
warfarin on the risk of stroke in patients with nonrheumatic atrial fibrillation. N Engl J Med. 1990;
323:1505-1511.
27. Stroke Prevention in Atrial Fibrillation Investigators. Stroke Prevention in Atrial Fibrillation
Study: final results. Circulation. 1991;84:527-539.
28. Ezekowitz MD, Bridgers SL, James KE, et al,
for the Veterans Affairs Stroke Prevention in Nonrheumatic Atrial Fibrillation Investigators. Warfarin in the prevention of stroke associated with
nonrheumatic atrial fibrillation. N Engl J Med. 1992;
327:1406-1412.
29. Coplen SE, Antman EM, Berlin JA, Hewitt P,
Chalmers TC. Efficacy and safety of quinidine
therapy for maintenance of sinus rhythm after cardioversion: a meta-analysis of randomized control
trials. Circulation. 1990;82:1106-1116.
30. Albers GW, Sherman DG, Gress DR, Paulseth
JE, Petersen P. Stroke prevention in nonvalvular
atrial fibrillation: a review of prospective randomized trials. Ann Neurol. 1991;30:511-518.
31. Fihn SD, McDonell M, Martin D, et al, for the

Warfarin Optimized Outpatient Follow-up Study

Group. Risk factors for complications of chronic
aticoagulation: a multicenter study. Ann Intern Ned.

1993;118:511-520.