Atrial fibrillation in the Malmo ¨diet and cancer study: a study of occurrence, risk factors and diagnostic validity J. Gustav Smith •Pyotr G…. [630825]

CARDIOVASCULAR DISEASE
Atrial fibrillation in the Malmo ¨diet and cancer study: a study
of occurrence, risk factors and diagnostic validity
J. Gustav Smith •Pyotr G. Platonov •
Bo Hedblad •Gunnar Engstro ¨m •Olle Melander
Received: 9 August 2009 / Accepted: 5 November 2009 / Published online: 21 November 2009
/C211Springer Science+Business Media B.V. 2009
Abstract The validity of atrial fibrillation (AF) diagnoses
in national registers for use as endpoints in prospective
studies has not been evaluated. We studied the validity ofAF diagnoses in Swedish national hospital discharge and
cause of death registers and the occurrence of and risk
factors for AF in a middle-aged Swedish population usingthese registers. Our study included the 30,447 individuals
(age 44–73) who attended baseline visits in 1991–1996 of
the Malmo ¨Diet and Cancer study. Individuals with a first
AF diagnosis were identified by record linkage with
national registers. A subset of cases was randomly selected
for validation by examination of electrocardiograms andpatient records. Electrocardiograms were available in 98%
of the validation sample (95% definitive AF, 3% no AF).
The 2% with ECGs unavailable had probable AF. BaselineAF prevalence was 1.3%, higher in men and increased with
age. During 11.2 years of follow-up 1430 first AF diag-
noses occurred. Risk factors were age, hypertension, BMI,diabetes, history of heart failure, history of myocardial
infarction and, in men but not women, current smoking.
The strongest risk factors were history of heart failure
(hazard ratio men 4.5, women 8.7) and myocardial infarction(hazard ratio men 2.0, women 1.8). The largest population
attributable risks were observed for hypertension (men
38%, women 34%) and obesity (men 11%, women 10%).In conclusion, case misclassification of AF in national
registers is small, indicating feasibility of use in prospec-
tive studies. Hypertension and obesity account for largeportions of population risk in middle-aged individuals with
low prevalence of manifest cardiac disease.
Keywords Atrial fibrillation /C1Epidemiology /C1
Cardiovascular risk factors /C1Cohort study
Abbreviations
AF Atrial fibrillationECG Electrocardiogram
PAR Population attributable risk
MDCS The Malmo ¨Diet and Cancer Study
BMI Body mass index
HDR The Swedish Hospital Discharge Register
CDR The Swedish Cause of Death RegisterICD The International Classification of Diseases
HR Hazard ratio
95% CI 95% confidence interval
Background
Atrial fibrillation is the most common cardiac arrhythmia
with a lifetime risk at 40 years as high as one in four in the
late twentieth century [ 1] and a major cause of morbidity
and mortality through increased risk of heart failure andischemic stroke [ 2–5]. Recent years have seen a shift in the
spectrum of cardiac disease with decreasing incidence andJ. G. Smith ( &)/C1B. Hedblad /C1G. Engstro ¨m/C1O. Melander
Department of Clinical Sciences, Lund University,Clinical Research Center (CRC), Malmo ¨University Hospital,
Entrance 72, Bldg 91, Floor 12, 205 02 Malmo ¨, Sweden
e-mail: [anonimizat]
J. G. Smith /C1P. G. Platonov
Department of Clinical Sciences, Lund University, Lund,Sweden
J. G. Smith
Program in Medical and Population Genetics, Broad Instituteof Harvard and Massachusetts Institute of Technology,Cambridge, MA, USA
123Eur J Epidemiol (2010) 25:95–102
DOI 10.1007/s10654-009-9404-1

mortality of acute myocardial infarction and increasing rates
of atrial fibrillation and heart failure [ 6,7]. The causes of this
trend are incompletely known, motivating the use of pro-
spective, population-based studies of these diseases toidentify risk factors, monitor incidence trends and study their
relation to risk factors. To identify clinical disease end-
points in such studies, national or hospital discharge registersare commonly used with the advantages of low cost, sim-
plicity and resistance to survival bias. However, such study
designs may be susceptible to misclassification bias. Previ-
ous studies have reported high validity of diagnoses of
myocardial infarction and heart failure in national registers[8–11] but to our knowledge, the validity of diagnoses of
atrial fibrillation has not been reported. We postulated a
higher validity of atrial fibrillation diagnoses than for myo-cardial infarction and heart failure as the diagnosis rests
entirely on one easily obtained and interpreted investigation,
the electrocardiogram (ECG), as it has done since early in thetwentieth century when ECGs of patients with atrial fibril-
lation were first reported shortly after the initial development
of the ECG by Willem Einthoven [ 12–14]. Consequently,
atrial fibrillation diagnoses are also readily validated retro-
spectively if ECGs are archived.
We describe the ascertainment of individuals with atrial
fibrillation in a large, community-based cohort study from
Malmo ¨, Sweden, by record linkage to national registers
using Swedish personal identification numbers. We vali-date the method by review of ECGs and patient records for
identified cases and compare the results in terms of prev-
alence and risk factor associations with previous reportsfrom cohorts using different ascertainment schemes. Finally,
we provide estimates of the population attributable risks
conferred by identified risk factors.
Methods
Study cohort
The Malmo ¨Diet and Cancer study (MDCS) is a prospec-
tive cohort study in which all men born between 1923 and
1945 and women born between 1923 and 1950 from thecity of Malmo ¨in southern Sweden (total population of
about 230 000 individuals) were invited to participate. The
participation rate was approximately 40%. A total of30,447 men and women underwent a baseline examination
between 1991 and 1996 [ 15]. Individuals underwent mea-
surement of anthropometric dimensions (BMI, waist andhip circumference) and blood pressure and filled out a
questionnaire on health, lifestyle, socioeconomic factors
and medications [ 16]. Blood pressure was measured using
a mercury-column sphygmomanometer after 10 minutes of
rest in the supine position. Baseline hypertension wasdefined as blood pressure C140/90 mm Hg or use of
antihypertensive medications. Baseline diabetes mellitus
was defined as self-reported history of a physician’s diag-
nosis of diabetes or use of antidiabetic medications.Obesity was defined as BMI [30. Current smoking was
ascertained from the baseline questionnaire. All individuals
provided written informed consent and the study wasapproved by the regional ethics committee.
National registersThe Swedish Hospital Discharge Register (HDR) and the
Swedish Cause of Death Register (CDR) are administered by
the Swedish National Board of Health and Welfare. Data
collection in the HDR was started in the 1960s and includesdates of admission and discharge as well as primary and
contributory diagnoses from all public hospitals in Sweden.
Reporting to the HDR has been compulsory since 1987 but theonly hospital in Malmo ¨(Malmo ¨University Hospital) has
reported since 1969. The CDR includes diagnoses from death
certificates since 1952, regardless if death occurred outside ofSweden. Diagnoses in the HDR are coded as primary or
contributory and in the CDR as underlying or contributory
cause of death, both using the International Classification ofDisease (ICD). The 8th edition (ICD-8) was used until the end
of 1986, the 9th edition (ICD-9) between 1987 and 1996 and
the 10th edition (ICD-10) from 1997 until present.
Ascertainment of atrial fibrillation and other
cardiac diseases
A first diagnosis of atrial fibrillation or atrial flutter were
ascertained until 31 December 2005 by linkage of Swedishpersonal identification numbers to HDR and CDR using
diagnosis codes 427.92 for ICD-8, 427D for ICD-9 and I48
for ICD-10. As in previous studies [ 17,18], the outcome of
atrial fibrillation was defined as either a diagnosis of atrial
fibrillation or atrial flutter, given the close interrelationship
of these diseases [ 19
]. Myocardial infarction was defined as
code 410 for ICD8 and ICD-9 and I21 for ICD-10. Heart
failure was defined as codes 427.00, 427.10, 428.99 for
ICD-8, 428 for ICD-9 and I50 and I11.0 for ICD-10.
Case validation
We randomly selected a hundred individuals with atrial
fibrillation or atrial flutter for validation. ECGs were retrieved
from the ECG database (GE MUSE, GE Healthcare) of theScania region in southern Sweden and were all reviewed by a
trained arrhythmologist (PP) and a physician in training at the
cardiology department (JGS). Atrial fibrillation was defined aslack of consistent P-waves preceding the QRS complex and
irregular R-R intervals and atrial flutter was defined as96 J. G. Smith et al.
123

presence of flutter waves [ 14]. Patient records and any ECGs
from the local ECG database of the hospital in the neighboring
city, Lund University Hospital (Siemens Megacare), were
reviewed for individuals where ECGs were unavailable orwith inconclusive ECG findings.
Statistical analysisWe examined prevalence, defined as a diagnosis of atrial
fibrillation prior to the baseline visit, per gender and 5-year
age group at the MDCS baseline visit. We excluded pre-
valent cases and studied the association of incident atrialfibrillation to previously reported risk factors [ 20]a t
baseline including age, sex, myocardial infarction, heart
failure, hypertension, diabetes , smoking and obesity. Kaplan–
Meier plots were constructed to visualize cumulative
incidence during follow-up per sex and 5-year age category
at baseline. Sex-specific Cox proportional hazards modelswere fitted with time to first diagnosis of atrial fibrillation
as dependent variable and each risk factor adjusted for age
as independent variables. Follow-up extended until 31December 2005 and was censored at death or emigration.
P-values \0.05 were considered significant. Log-cumu-
lative hazard plots were examined for the validity of pro-portional hazards assumptions. To examine the impact of
uncertainty in diagnosis dates relative to disease onset we
also fitted unconditional logistic regression models forcomparison of results. Population attributable risk (PAR)
estimates [ 21] were calculated using the standard formula
where P
eis the population risk factor prevalence and RR is
the relative risk, which is equivalent to the hazard ratio
here:
PAR ¼PeðRR/C01Ț
1țPeðRR/C01Ț
Adjustment for age was performed using a weighted sum
approach [ 22] in which PARs were calculated per 5-yearage group and summed weighted for the proportion of
cases in each 5-year age group ( wi):
PAR ¼XI
i¼1wiPAR i
To identify independent risk factors we fitted sex-specific
multivariable-adjusted Cox proportional hazards models
incorporating all risk factors with backwards elimination.
All analyses were performed in SPSS version 16 (SPSSInc, Chicago, IL).
Results
Individuals with missing date of baseline visit were
excluded ( n=6), leaving 18,323 women and 12,118 men
aged 44–73 years who were examined at the baseline visit.
Baseline characteristics are shown in Table 1. 312 indi-
viduals developed atrial fibrillation prior to the baseline
visit. During follow-up (332,240 person-years, mean fol-
low-up time 11.2 years) we identified 807 men and 623women with a first diagnosis in the HDR ( n=1423) or the
CDR ( n=7), corresponding to a cumulative incidence of
4.7%. Atrial fibrillation was coded as the underlying causeof death in two individuals and as contributory cause of
death in five individuals of whom one individual died from
a stroke, one died of myocardial infarction, one died of‘‘injuries or poison’’, one died of cancer and one died of
neurological disease.
Validation of atrial fibrillation diagnoses in MDCS
The hundred cases that were randomly selected for valida-
tion had a similar baseline age- and sex-distribution as the
full sample of atrial fibrillation cases. Using the Scania ECGdatabase we were able to confirm ECGs of 94 individuals as
Table 1 Baseline characteristics of MDCS
Men Women Total
Age (years) 59.1 (7.0) 57.3 (7.9) 58.0 (7.6)
Body mass index 26.3 (3.6) 25.5 (4.3) 25.8 (4.0)
Systolic blood pressure (mm Hg) 144.0 (19.5) 139.2 (20.2) 141.1 (20.1)Diastolic blood pressure (mm Hg) 88.0 (9.9) 84.0 (9.8) 85.6 (10.0)Antihypertensive treatment (%) 19.5 15.9 17.3Current smoking (%) 28.7 28.1 28.3History of diabetes (%) 4.2 2.5 3.1History of heart failure (%) 0.5 0.2 0.3History of myocardial infarction (%) 4 0.7 2
Continuous variables are presented as mean (SD) and categorical variables are presented as number (percentage). Information on BMI, blood
pressure and antihypertensive treatment at the baseline visit was available in most individuals ( n=30,352) but information from the ques-
tionnaire was missing for a substantial proportion on current smoking ( n=28,559) and diabetes ( n=28,500)Atrial fibrillation in the Malmo ¨diet and cancer study 97
123

definitive atrial fibrillation or flutter. For the remaining 6
individuals paper ECGs were retrieved from patient records
which made validation of 1 individual as definitive atrial
fibrillation possible. However, 3 individuals did not haveatrial fibrillation or flutter; 1 individual instead had sinus
tachycardia and pulmonary embolism and 2 individuals had
different supraventricular tachycardias of whom one sub-sequently underwent a successful catheter ablation for
concealed Wolff–Parkinson–White syndrome and the other
currently is waiting for ablation. We were unable to retrieve
ECGs for two individuals but according to records both had
been seen by several physicians including consultant phy-sicians who clearly stated diagnoses of atrial fibrillation.
Both were admitted primarily for other causes, specifically
pneumonia and heart failure, and atrial fibrillation wascoded as contributory rather than primary.
Occurrence of atrial fibrillation in MDCSThe prevalence of atrial fibrillation at the baseline visit
between 1991 and1996 was 10.25 per 1,000 individualsand higher in men (16 per 1,000) than in women (6 per
1,000). Prevalence increased markedly with age from 2 per
1,000 in ages 45–49 to 29 per 1,000 in ages 70–74 asshown in Fig. 1. For estimates from higher age groups we
also studied the prevalence up to the end of follow-up in2005. 4,953 individuals had then reached the age category
75–79 in which the prevalence was 92 per 1,000 and
higher among men (118 per 1,000) than women (75 per
1,000). 1486 individuals had reached the age category80–82 years in which the prevalence was 124 per 1,000
individuals, 171 per 1,000 among men and 98 per 1,000
among women.
The incidence rate was 4.3 incident cases per 1,000
person-years of follow-up, was higher in men (6.3 per
1,000) than in women (3.1 per 1,000) and showed a marked
increase with age from 0.8 per 1,000 in ages 45–49 to 12.4
per 1,000 in ages 70–74. The cumulative incidence of afirst episode of atrial fibrillation during follow-up is shown
per age category in Fig. 2. 71.9% of cases were over
60 years of age at diagnosis.
Risk factors for atrial fibrillation in MDCS
Information on BMI, blood pressure and antihypertensive
treatment at the baseline visit was available in most indi-
viduals ( n=30,352) but information from the question-
naire was missing for a higher proportion on current
smoking and diabetes for which information was available
in 28,559 and 28,500 individuals, respectively. Male sexwas associated with an increased risk of atrial fibrillation in
Cox proportional hazards models (HR 2.0, 95% CI =1.8–2.2).
All risk factors were statistically significant predictors inboth men and women except current smoking which was
not a significant predictor in women (Table 2). Results
from logistic regression were similar to results from Coxproportional hazards models (data not shown). The highest
hazard ratios were observed for heart failure and myocar-
dial infarction but the highest population attributable riskswere observed for hypertension and obesity, due to high
prevalence of these risk factors. Risk estimates for hyper-
tension changed only marginally when using the definitionblood pressure C160/95 mm Hg or use of antihyperten-
sive medications but PAR estimates decreased resulting
from a substantial decrease in prevalence as shown inTable 2. BMI was associated with atrial fibrillation in men
(HR per unit 1.08, 95% CI =1.06–1.10) and women (HR
per unit 1.06, 95% CI =1.04–1.08).
Independent risk factors for atrial fibrillation in MDCS
In stepwise multiple regression models for men, effect
estimates for age (HR 1.1, 95% CI =1.09–1.11), hyper-
tension (HR 1.6, 95% CI =1.33–1.97), obesity (HR 1.8,
Fig. 1 Prevalent atrial fibrillation by gender and age group at
baseline. Shown are numbers of prevalent cases of atrial fibrillationper age group and gender. Numbers of individuals in each group isshown in boxes . Numbers of individuals were higher in younger age
groups; 5,875 (45–49), 6,360 (50–54), 5,593 (55–59), 6,152 (60–64),
3,649 (65–69) and 2,812 (70–74), respectivelyFig. 2 Cumulative incidence of first atrial fibrillation during a mean
follow-up of 11.2 years per age category at baseline in men ( a) and
women ( b). Age categories from bottom to top in both panels are 45–49,
50–54, 55–59, 60–64. 65–69 and 70–74 yearsc98 J. G. Smith et al.
123

Atrial fibrillation in the Malmo ¨diet and cancer study 99
123

95% CI =1.47–2.10), history of heart failure (HR 2.8,
95% CI =1.43–5.54) and history of myocardial infarction
(HR 2.0, 95% CI =1.56–2.59) were all slightly lower
whereas current smoking (HR 1.3, 95% CI =1.10–1.53)
was slightly higher. In women, effect estimates were
slightly lower for age (HR 1.1, 95% CI =1.11–1.14),
hypertension (HR 1.6, 95% CI =1.31–1.99), obesity (HR
1.5, 95% CI =1.25–1.85) and heart failure (HR 7.4, 95%
CI=3.05–17.94) in multivariable-adjusted models. Dia-
betes was not significantly associated with atrial fibrillation
in men (HR 1.1, 95% CI =0.84–1.59) and neither smok-
ing (1.2, 95% CI =0.98–1.44), diabetes (1.4, 95%
CI=0.95–2.05) nor myocardial infarction (HR 1.01, 95%
CI=0.47–2.20) was significantly associated with atrial
fibrillation in women.
Discussion
We have described the ascertainment of individuals with a
first diagnosis of atrial fibrillation from a prospective
cohort study using Swedish national registers and the
validity of atrial fibrillation diagnoses in these registers.Using these data we determined the prevalence and risk
factors for atrial fibrillation.
As expected, the validity of a diagnosis of atrial fibril-
lation in registers is high, indicating only a small impact of
case misclassification bias. Validity was higher than for
heart failure and myocardial infarction which have beenreported previously as 80–82 and 86–97%, respectively
[8–11]. One previous study examined the validity of hos-
pital discharge diagnoses of atrial fibrillation and observedslightly higher predictive values, with a confirmed diag-
nosis in 209 of 212 subjects [ 18]. However, that study only
included elderly individuals, who are likely to have a muchlower incidence of other supraventricular tachyarrhythmiasrelative to atrial fibrillation. Other supraventricular
arrhythmias were observed in two of three individuals
incorrectly diagnosed as atrial fibrillation in our study. Ourdata includes information from the CDR on the few indi-
viduals who died during the first hospitalization for atrial
fibrillation but does not provide any information on thenumber of asymptomatic and undiagnosed cases, poten-
tially leading to control misclassification bias, conferring
falsely low prevalence estimates and decreased statisticalpower in studies of risk factors. The only way to detect
such cases would be through long-term ECG recording
which is not feasible in epidemiologic studies of this scale.
The total prevalence observed in our study, about 1%, is
in agreement with estimates from previous studies in thegeneral American population [ 23,24]. The age category-
specific prevalence estimates were similar to those observed
in a screening with 24-h ambulatory ECG registrations inmen from Malmo ¨[25]. Age category-specific estimates
were also similar to those observed in the general Ameri-
can population but with lower estimates in the 70–74 yearscategory at baseline [ 23], which could reflect the recruit-
ment bias towards healthy individuals in MDCS, and
higher estimates in categories 75–79 and [80 after follow-
up [26] which could reflect lack of participation bias with
the use of national register data rather than follow-up visits
for endpoint ascertainment. Furthermore, previous studiesused different sets of exclusions, such as atrial fibrillation
after cardiac surgery or association with hyperthyroidism,
which further complicates comparisons. As in previousstudies, we found atrial fibrillation to be a disease mainly
affecting older adults with about 70% of cases [60 years at
diagnosis [ 24]. The prevalence of atrial fibrillation below
50 years of age was low, \0.5%, and increased gradually
across middle age to over 10% at 80 years similar to pre-
vious studies [ 23,24]. Like in previous studies the preva-
lence was higher in men across all age categories.Table 2 Risk factors for onset of first episode of atrial fibrillation
Variable Men Women
Prevalence (%) HR PAR (%) Prevalence (%) HR PAR (%)
Age – 1.109 (1.10–1.12) – – 1.14 (1.13–1.15) –
Body mass index [30 13.4 1.88 (1.59–2.23) 11 14.2 1.66 (1.38–2.00) 10
Hypertension, C140/90 mm Hg 68.1 1.78 (1.48–2.14) 38 56.3 1.74 (1.42–2.13) 34
Hypertension, C160/95 mm Hg 42.8 1.85 (1.60–2.13) 29 33.0 1.69 (1.43–1.99) 25
Current smoking 28.8 1.2 (1.02–1.42) 5 26.6 1.12 (0.92–1.35) –History of diabetes 4.1 1.39 (1.02–1.90) 1.9 2.5 1.67 (1.15–2.43) 2.5History of heart failure 0.3 4.53 (2.34–8.75) 2 0.1 8.70 (3.60–20.94) 1.4
History of myocardial infarction 3.8 2.03 (1.65–2.49) 5 0.6 1.84 (1.26–2.69) 0.6
Shown are hazard ratios (HR) with 95% confidence intervals from Cox proportional hazards models and corresponding population attributable
risks (PAR). All models were adjusted for age. PARs were age-adjusted using a weighted sum approach. Prevalent cases of atrial fibrillation wereexcluded ( n=312). Two definitions of hypertension were compared, defined as use of antihypertensive medications and either blood pres-
sure C140/90 or C160/95 mm Hg100 J. G. Smith et al.
123

Risk factors for atrial fibrillation have been described in
previous studies and besides age and sex include manifest
cardiac disease (heart failure, valvular disease, myocardial
infarction) as well as cardiometabolic risk factors hyper-tension, diabetes, smoking and obesity [ 17,18,27–29]. Our
large sample allowed identification of large numbers of
atrial fibrillation providing our study with high statisticalpower and precision in risk estimates. We found manifest
cardiac disease to confer the largest relative risks, followed
by hypertension, obesity and diabetes. Smoking was only
significant in men, likely due to lack of power in sex-
specific analyses. Diabetes was a substantially weaker riskfactor in men, as previously reported for heart failure [ 30].
Effect estimates for hypertension and BMI were high and
remarkably similar to previous studies with age-adjustedhazard ratios of 1.7–1.8 for hypertension [ 17] and 1.06–
1.08 per unit of BMI [ 27–29]. To examine the population
burden of each risk factor we also estimated populationattributable risks for dichotomous risk factors, an estimate
of the proportion of atrial fibrillation that would not have
occurred if the population had been in the low risk categoryunder the assumption of causality. Although stronger risk
factors, manifest cardiac diseases were rare in our sample
and so contributed little to population risk. Large portionsof population risk were explained by BMI and hypertension
as both showed high risk estimates and prevalence. The
impact of hypertension decreased somewhat when definedas blood pressure C160/95 mm Hg and use of antihyper-
tensive medications. However, the true impact of blood
pressure on atrial fibrillation may be even higher as a largestudy recently found increased risk of atrial fibrillation also
at prehypertensive levels [ 31]. One previous study, from
the Framingham Heart Study investigators, has reportedPAR estimates for atrial fibrillation. In that study the PARs
for cardiometabolic risk factors was similar to our study
but the risk explained by manifest cardiac diseases washigher, likely resulting from the higher prevalence of car-
diac diseases in the older sample. The effect estimates
decreased substantially with adjustment for cardiometa-bolic risk factors. Thus, our findings in middle-aged indi-
viduals with low prevalence of clinical cardiac disease
together with previous findings from the FraminghamHeart Study indicate benefits from targeting cardiometa-
bolic risk factors in middle-aged individuals without car-
diac disease for prevention of atrial fibrillation. Preventionof atrial fibrillation is gaining increasing interest, as high-
lighted by a recent National Heart, Lung, and Blood
Institute workshop [ 20].
Much of population risk of atrial fibrillation remains
unexplained. For comparison, the Interheart study sug-
gested that known and modifiable risk factors explain up to90–94% of risk of myocardial infarction [ 32]. It is possible
that undiagnosed or subclinical cardiac disease such asasymptomatic ventricular dysfunction or silent myocardial
infarction may contribute substantially to population risk of
atrial fibrillation. Also, risk factors such as valvular disease
or thyroid disease that were not available in the presentstudy, or newer risk factors such as heritability, obstructive
sleep apnea or inflammation [ 33] might explain substantial
proportions of PAR.
Our study has several limitations why the generaliz-
ability of occurrence estimates and risk factor burden must
be considered with some caution. First, the present study
recruited forty percent of the eligible population, likely
biased towards healthier individuals, and was restricted tomiddle-aged individuals. Second, as discussed above our
study did not include information on asymptomatic or
undiagnosed cases. Finally, it does not need to be reiteratedthat the date of onset may be uncertain, potentially leading
to false occurrence estimates.
In conclusion, our results show that atrial fibrillation
diagnoses can be ascertained with high validity from
national hospital discharge registers and suggests that large
portions of the population-attributable risk can be explainedby hypertension and obesity in middle-aged individuals
with low prevalence of manifest cardiac disease. These
findings motivate future studies of incidence trends andnovel risk factors for atrial fibrillation and suggest benefits
for reduction of occurrence of atrial fibrillation by cardio-
metabolic risk factor management.
Acknowledgments MDCS was made possible by grants from the
Malmo ¨city council. J. Gustav Smith gratefully acknowledges finan-
cial support from the Swedish Heart Lung Foundation. Pyotr Platonov
was supported by governmental funding for clinical research within
the Swedish NHS. Olle Melander was supported by grants from theSwedish Medical Research Council, the Swedish Heart and LungFoundation, the Medical Faculty of Lund University, Malmo ¨Uni-
versity Hospital, the Albert Pa ˚hlsson Research Foundation, the
Crafoord foundation, the Ernhold Lundstro ¨ms Research Foundation,
the Region Skane, the Hulda and Conrad Mossfelt Foundation, theKing Gustaf V and Queen Victoria Foundation, the Lennart Hanssons
Memorial Fund, the Wallenberg Foundation.
Disclosures Gunnar Engstro ¨m is employed as senior epidemiologist
by AstraZeneca R&D.
References
1. Lloyd-Jones DM, Wang TJ, Leip EP, et al. Lifetime risk for
development of atrial fibrillation: the Framingham heart study.
Circulation. 2004;110:1042–6.
2. Stewart S, Hart CL, Hole DJ, McMurray JJ. A population-based
study of the long-term risks associated with atrial fibrillation:20-year follow-up of the Renfrew/Paisley study. Am J Med.
2002;113:359–64.
3. Wang TJ, Larson MG, Levy D, et al. Temporal relations of atrial
fibrillation and congestive heart failure and their joint influence
on mortality: the Framingham heart study. Circulation. 2003;107:
2920–5.Atrial fibrillation in the Malmo ¨diet and cancer study 101
123

4. Shinbane JS, Wood MA, Jensen DN, Ellenbogen KA, Fitzpatrick
AP, Scheinman MM. Tachycardia-induced cardiomyopathy: a
review of animal models and clinical studies. J Am Coll Cardiol.1997;29:709–15.
5. Wolf PA, Abbott RD, Kannel WB. Atrial fibrillation as an
independent risk factor for stroke: the Framingham study. Stroke.1991;22:983–8.
6. Braunwald E. Shattuck lecture–cardiovascular medicine at the
turn of the millennium: triumphs, concerns, and opportunities. N
Engl J Med. 1997;337:1360–9.
7. Miyasaka Y, Barnes ME, Gersh BJ, et al. Secular trends in
incidence of atrial fibrillation in Olmsted County, Minnesota,
1980 to 2000, and implications on the projections for future
prevalence. Circulation. 2006;114:119–25.
8. Hammar N, Alfredsson L, Rose ´n M, Spetz CL, Kahan T, Ysberg
AS. A national record linkage to study acute myocardial infarc-
tion incidence and case fatality in Sweden. Int J Epidemiol.
2001;30(Suppl 1):S30–4.
9. Lindblad U, Ra ˚stam L, Ranstam J, Peterson M. Validity of reg-
ister data on acute myocardial infarction and acute stroke: the
Skaraborg hypertension project. Scand J Soc Med. 1993;21:3–9.
10. Ingelsson E, Arnlo ¨v J, Sundstro ¨m J, Lind L. The validity of a
diagnosis of heart failure in a hospital discharge register. Eur J
Heart Fail. 2005;7:787–91.
11. Merry AHH, Boer JMA, Schouten LJ, Feskens EJ, Verschuren WM,
Gorgels AP, et al. Validity of coronary heart diseases and heartfailure based on hospital discharge and mortality data in the
Netherlands using the cardiovascular registry Maastricht cohort
study. Eur J Epidemiol. 2009;24:237–47.
12. Einthoven W. Le telecardiogramme. Arch Int de Physiol. 1906;4:
132–64.
13. Hering HE. Das Elektrocardiogramm des Pulsus Irregularis per-
petuus. Deutsches Archiv fu ¨r Physiologie. 1908;94:205–8.
14. Fuster V, Ryde ´n LE, Cannom DS, et al. ACC/AHA/ESC 2006
guidelines for the management of patients with atrial fibrillation:
a report of the American College of Cardiology/American HeartAssociation task force on practice guidelines and the EuropeanSociety of Cardiology committee for practice guidelines (writing
committee to revise the 2001 guidelines for the management of
patients with atrial fibrillation): developed in collaboration withthe European heart rhythm association and the Heart rhythmsociety. Circulation. 2006;114:e257–354.
15. Berglund G, Elmsta ˚hl S, Janzon L, Larsson SA. The Malmo ¨diet and
cancer study. Design and feasibility. J Intern Med. 1993;233:45–51.
16. Persson M, Hedblad B, Nelson JJ, Berglund G. Elevated Lp-PLA2
levels add prognostic information to the metabolic syndrome on
incidence of cardiovascular events among middle-aged nondiabeticsubjects. Arterioscler Thromb Vasc Biol. 2007;27:1411–6.
17. Benjamin EJ, Levy D, Vaziri SM, D’Agostino RB, Belanger AJ,
Wolf PA. Independent risk factors for atrial fibrillation in a
population-based cohort. The Framingham Heart Study. JAMA.1994;271:840–4.18. Psaty BM, Manolio TA, Kuller LH, et al. Incidence of and risk
factors for atrial fibrillation in older adults. Circulation.
1997;96:2455–61.
19. Waldo AL, Feld GK. Inter-relationships of atrial fibrillation and
atrial flutter. J Am Coll Cardiol. 2008;51:779–86.
20. Benjamin EJ, Chen P, Bild DE, et al. Prevention of atrial fibril-
lation: report from a National Heart, Lung, and Blood Instituteworkshop. Circulation. 2009;119:606–18.
21. Levin ML. The occurrence of lung cancer in man. Acta Unio Int
Contra Cancrum. 1953;9:531–41.
22. Walter SD. The estimation and interpretation of attributable
fraction in health research. Biometrics. 1976;32:829–49.
23. Go AS, Hylek EM, Phillips KA, Chang Y, Henault LE, Selby JV,
et al. Prevalence of diagnosed atrial fibrillation in adults: nationalimplications for rhythm management and stroke prevention: theanticoagulation and risk factors in atrial fibrillation (ATRIA)
study. JAMA. 2001;285:2370–5.
24. Feinberg WM, Blackshear JL, Laupacis A, Kronmal R, Hart RG.
Prevalence, age distribution, and gender of patients with atrial
fibrillation. Analysis and implications. Arch Intern Med. 1995;155:
469–73.
25. Engstro ¨m G, Hedblad B, Juul-Mo ¨ller S, Tyde ´n P, Janzon L.
Cardiac arrhythmias and stroke. Increased risk in men with high
frequency of atrial ectopic beats. Stroke. 2000;31:2925–9.
26. Furberg CD, Psaty BM, Manolio TA, Gardin JM, Smith VE,
Rautaharju PM. Prevalence of atrial fibrillation in elderly subjects(the Cardiovascular Health Study). Am J Cardiol. 1994;74:
236–41.
2 7 . W a n gT J ,P a r i s eH ,L e v yD ,D ’ A g o s t i n oR BS r ,W o l fP A ,V a s a nR S ,
et al. Obesity and the risk of new-onset atrial fibrillation. JAMA.
2004;292:2471–7.
28. Frost L, Hune LJ, Vestergaard P. Overweight and obesity as risk
factors for atrial fibrillation or flutter: the Danish diet, cancer, andhealth study. Am J Med. 2005;118:489–95.
29. Rosengren A, Hauptman PJ, Lappas G, Olsson L, Wilhelmsen L,
Swedberg K. Big men and atrial fibrillation: effects of body sizeand weight gain on risk of atrial fibrillation in men. Eur Heart J.2009;30:1113–20.
30. Levy D, Larson MG, Vasan RS, Kannel WB, Ho KKL. The
progression from hypertension to congestive heart failure. JAMA.1996;275:1557–62.
31. Conen D, Tedrow UB, Koplan BA, Glynn RJ, Buring JE, Albert CM.
Influence of systolic and diastolic blood pressure on the risk of
incident atrial fibrillation in women. Circulation. 2009;119:
2146–52.
32. Yusuf S, Hawken S, Ounpuu S, et al. Effect of potentially
modifiable risk factors associated with myocardial infarction in52 countries (the INTERHEART study): case-control study.Lancet. 2004;364:937–52.
33. Schoonderwoerd BA, Smit MD, Pen L, van Gelder IC. New risk
factors for atrial fibrillation: causes of ‘not so lone atrial fibril-lation’. Europace. 2008;10:668–73.102 J. G. Smith et al.
123