Author + information
- Received January 14, 2014
- Revision received March 8, 2014
- Accepted March 11, 2014
- Published online July 22, 2014.
- ∗Unit of Nutritional Epidemiology, Institute of Environmental Medicine, Karolinska Institutet, Stockholm, Sweden
- †Department of Cardiology, Karolinska University Hospital, Stockholm, Sweden
- ↵∗Reprint requests and correspondence:
Dr. Susanna C. Larsson, Unit of Nutritional Epidemiology, Institute of Environmental Medicine, Karolinska Institutet, Box 210, SE-17177 Stockholm, Sweden.
Background Although high alcohol consumption has been associated with increased risk of atrial fibrillation (AF), the role of light to moderate drinking remains unclear.
Objectives The study sought to investigate the association between alcohol consumption and AF risk in a prospective study of Swedish men and women and to conduct a meta-analysis of prospective studies to summarize available evidence.
Methods We followed 79,019 men and women who, at baseline, were free from AF and had completed a questionnaire about alcohol consumption and other risk factors for chronic diseases. Incident AF cases were ascertained by linkage to the Swedish Inpatient Register. For the meta-analysis, studies were identified by searching PubMed through January 10, 2014, and by reviewing references of pertinent publications. Study-specific relative risks (RRs) were combined using a random effects model.
Results Over 859,420 person-years of follow-up (1998 to 2009), 7,245 incident AF cases were identified in our own cohort study. The association between alcohol consumption and AF did not differ by sex (p for interaction = 0.74). Compared with current drinkers of <1 drink/week (12 g alcohol/drink), the multivariable RRs of AF were 1.01 (95% confidence interval [CI]: 0.94 to 1.09) for 1 to 6 drinks/week, 1.07 (95% CI: 0.98 to 1.17) for 7 to 14 drinks/week, 1.14 (95% CI: 1.01 to 1.28) for 15 to 21 drinks/week, and 1.39 (95% CI: 1.22 to 1.58) for >21 drinks/week. Results were similar after excluding binge drinkers. In a meta-analysis of 7 prospective studies, including 12,554 AF cases, the RRs were 1.08 (95% CI: 1.06 to 1.10) for 1 drink/day, 1.17 (95% CI: 1.13 to 1.21) for 2 drinks/day, 1.26 (95% CI: 1.19 to 1.33) for 3 drinks/day, 1.36 (95% CI: 1.27 to 1.46) for 4 drinks/day, and 1.47 (95% CI: 1.34 to 1.61) for 5 drinks/day, compared with nondrinkers.
Conclusions These findings indicate that alcohol consumption, even at moderate intakes, is a risk factor for atrial fibrillation.
Atrial fibrillation (AF)/atrial flutter (AFL), the most common cardiac arrhythmia, is accompanied with a 4- to 5-fold increased risk for stroke, tripling of the risk for heart failure, doubling of the risk for dementia, and 40% to 90% increase in the risk for all-cause mortality (1). Recently, much attention has been focused on new treatments for AF, whereas disease prevention has received little attention (1). It is important to identify modifiable risk factors for AF to decrease the disease burden. There are only a few established risk factors for AF, including increasing age, sex (men have higher risk), cardiac disease, elevated blood pressure, type 2 diabetes, obesity, and cigarette smoking (1).
High alcohol consumption also has been associated with an increased AF risk (2–7). However, few studies have detected an association between light to moderate alcohol drinking and AF (3,6,7). Previous studies had limited statistical power to detect weak associations with light to moderate alcohol drinking. Two meta-analyses of case-control and prospective studies observed a dose-response relation between alcohol consumption and AF risk, and the association was similar for men and women (8,9). Data on specific alcoholic beverages in relation to AF risk are sparse (4,5).
We investigated the relation between consumption of total alcohol and specific alcoholic beverages and the incidence of AF in a large, prospective study of Swedish men and women. We also examined the association between binge drinking and AF risk. Moreover, we performed an updated meta-analysis of prospective studies, which are less prone to bias compared with case-control studies, to examine the dose-response association of alcohol consumption with risk of AF and to assess whether light to moderate alcohol consumption increases the risk.
Our study population included participants from the COSM (Cohort of Swedish Men) and the SMC (Swedish Mammography Cohort) cohorts. In late 1997, 48,850 men (COSM), born between 1918 and 1952 and living in Västmanland and Örebro Counties, and 39,227 women (SMC trial), born between 1914 and 1948 and living in Västmanland and Uppsala counties, completed a 350-item questionnaire that elicited information on alcohol consumption and other risk factors for chronic diseases. We excluded those with an erroneous or a missing National Registration Number (n = 540), those with a prior diagnosis of AF (recorded in the Swedish Inpatient Register; n = 2,130) or cancer (recorded in the Swedish Cancer Registry; n = 4,403), those who died before start of follow-up (n = 81), and those with missing data on alcohol consumption (n = 1,904). This left a total of 79,019 participants, including 43,841 men and 35,178 women, 45 to 83 years of age, for the current analysis. This study was approved by the Regional Ethical Review Board at Karolinska Institutet in Stockholm, Sweden.
Case ascertainment and follow-up
AF cases were ascertained by record linkage to the Swedish Inpatient Register. Diagnoses in the Inpatient Register are coded according to the Swedish International Classification of Diseases (ICD) system (modified from the World Health Organization ICD classification system). Cases of AF and AFL were identified using the code I48 from ICD-10. AF was defined as either a diagnosis of AF or AFL, due to their close interrelationship and the difficulties of making a distinction between them. The validity of AF diagnoses in the Swedish registers has been evaluated in the Malmö Diet and Cancer study (10). The AF diagnoses were validated by examination of electrocardiograms (available in 98% of the sample) and patient records in a random sample of 100 AF cases identified by record linkage with national registers. It was found that 95% were definitive AF and 3% of the individuals had no AF. Two percent with unavailable electrocardiograms had probable AF. Information on dates of deaths for deceased participants was obtained from the Swedish Death Registry. Participants were followed up from January 1, 1998, to the date of diagnosis of AF, death, or the end of follow-up (December 31, 2009), whichever came first.
Assessment of alcohol consumption
Alcohol consumption was assessed in 1997 with a food frequency questionnaire on which participants reported their average consumption of 96 foods and beverages, including 6 alcoholic beverages, during the past year. Participants were asked to indicate how often they consumed class I beer (alcohol by volume, <2.25%), class II beer (2.8% to 3.5%), class III beer (>3.5%), wine, strong wine, and liquor, as well as the amount of beer, wine, and liquor consumed on a single occasion. The amount consumed was reported in an open-ended question. For frequency of consumption of alcoholic beverages, there were 9 pre-defined categories, ranging from never to ≥3 times per day. We calculated weekly alcohol consumption for each participant by multiplying the frequency of consumption of each alcoholic beverage by the amount consumed. We assumed that 1 standard drink contains 12 g alcohol. This amount corresponds to approximately 660 ml class I beer, 500 ml class II beer, 330 ml class III beer, 150 ml wine, 80 ml strong wine, or 40 ml liquor. Beer was calculated by summing up drinks of class I, II, and III beer. Total wine included wine and strong wine. The food frequency questionnaire used in the current study has been validated, and the Spearman correlation coefficients between estimates from the food frequency questionnaire, and the mean of 14 24-h recall interviews was 0.81 for alcohol (ethanol) (11).
Assessment of other risk factors
Information on education, body weight, height, physical activity, smoking, family history of myocardial infarction (MI) before 60 years of age, aspirin use, and history of hypertension was obtained through a self-administered questionnaire. Information on diabetes, coronary heart disease (CHD), and heart failure was obtained through linkage of the study population to the Swedish National Diabetes Register (for diabetes) and the Inpatient Register, which was complemented with self-reported data about diabetes on the questionnaire. Body mass index (BMI) was calculated as the body weight in kilograms divided by the square of height in meters. The validity of self-reported BMI compared with clinical measures is high (r = 0.9) in the Swedish population (12). We calculated pack-years of smoking history by multiplying the number of packs of cigarettes smoked per day at different ages (in 5 age-spans and during the last year), by the number of years of smoking.
We categorized participants into 7 categories of total alcohol consumption and into 4 categories of liquor, wine, and beer consumption (Table 1). We also grouped participants according to binge drinking, defined as having 5 or more drinks of liquor, wine, or beer on a single occasion or a calculated average consumption of 5 or more drinks per day.
Cox proportional hazards regression models were used to estimate relative risks (RRs) with 95% confidence intervals (CIs) of AF. All analyses were adjusted for age (in months) and sex as stratum variables. Multivariable models were further adjusted for education and for known risk factors for AF, including smoking, BMI, family history of MI before 60 years of age, a history of CHD or heart failure (ascertained through the Swedish Inpatient Register), a history of diabetes, and a history of hypertension (more details in Table 1). We also conducted analyses excluding those with a history of CHD or heart failure at baseline. In analyses of different alcoholic beverages, liquor, wine, and beer were mutually adjusted by inclusion in the same multivariable model.
Test for trend across categories of alcohol consumption in current drinkers was performed by assigning the median value for each category of alcohol consumption and modeling this variable as a continuous variable. We tested for differences in associations for beverage types using a chi-square test. All statistical analyses were conducted using SAS (version 9.2, SAS Institute, Cary, North Carolina). A 2-tailed p < 0.05 was considered statistically significant.
We performed a meta-analysis that included findings from our own prospective study (COSM and SMC trials) and results from previously published, prospective studies of alcohol consumption in relation to AF risk. We followed the MOOSE guidelines for conducting and reporting of meta-analyses of observational studies (13). Studies were identified by a computerized search of the PubMed database through January 10, 2014. We used the search terms alcohol consumption, alcohol drinking, or alcohol intake combined with atrial fibrillation or flutter. The references of relevant papers also were reviewed for further studies. No language restrictions were imposed. Eligibility criteria for inclusion in the meta-analysis were: 1) prospective design; 2) the exposure was alcohol consumption; 3) the outcome was incidence of AF or AF and AFL combined; and 4) RRs with 95% CIs were reported for at least 3 categories of alcohol consumption to be able to estimate a dose-response trend. We did not include studies of AF recurrence. We conducted a dose-response meta-analysis by using the same methods as described in previous meta-analyses (14,15). To standardize alcohol consumption, we used a common scale (i.e., alcoholic drinks/day). In 2 studies that reported alcohol consumption in grams/day (2,3), we converted alcohol consumption into drinks/day assuming that 1 drink contains 12 g of alcohol. The RR estimates from individual studies were combined using a random effects model (16). When results were presented separately for men and women (2 studies) (3,4), we combined the RR estimates, using a random effects model and included the pooled estimate in the meta-analysis. In a sensitivity analysis, we combined the RR estimates by using a fixed effects model. To evaluate a potential nonlinear association of alcohol consumption with AF risk, we used a restricted cubic spline model with 3 knots at percentiles 25%, 50%, and 75% of the distribution. A p value for nonlinearity was calculated by testing the null hypothesis that the coefficient of the second spline is equal to 0.
We performed subgroup analyses by study location (United States or Europe) and sex. Statistical heterogeneity among studies was evaluated by using the p and I2 statistics (17). Publication bias was examined with Egger’s test (18). We used Stata (version 12.0, StataCorp, College Station, Texas) for the statistical analyses.
Over 12 years of follow-up (859,420 person-years), we ascertained 7,245 incident AF cases (4,488 in men and 2,757 in women). Compared with men and women with light alcohol consumption, those with high consumption (>21 drinks/week) were younger and more likely to have a postsecondary education, to be current smokers, and to have a history of diabetes, hypertension, CHD, or heart failure (Online Table 1).
The associations of alcohol drinking status and consumption of total alcohol and different alcoholic beverages with risk of AF are presented in Table 1. Because the association between alcohol consumption and AF did not differ by sex (p for interaction = 0.74), all analyses were conducted for men and women combined. Compared with current alcohol drinkers of less than 1 drink/week, consumption of 15 to 21 drinks/week and >21 drinks/week was associated with a statistically significant 14% and 39%, respectively, increased risk of AF. The association between alcohol consumption and AF was somewhat stronger after excluding those with a diagnosis of CHD or heart failure at baseline (Table 1). For specific alcoholic beverages, consumption of more than 14 drinks/week of liquor or wine was associated with increased risk of AF (Table 1). There was no association with beer. When comparing the RRs in the highest categories, the RR for liquor differed statistically significantly from that of beer (p = 0.03), whereas no statistically significant difference was observed between liquor and wine (p = 0.54) or wine and beer (p = 0.10).
Binge drinking (consumption of ≥5 drinks on a single occasion) was identified in 18% of current drinkers and was associated with an increased AF risk after adjustment for age, sex, education, smoking, BMI, family history of MI before 60 years of age, and histories of CHD, heart failure, diabetes, and hypertension (multivariable RR: 1.13; 95% CI: 1.05 to 1.22). The association between binge drinking and AF was similar after further adjustment for frequency of consumption of each alcoholic beverage (multivariable RR: 1.12; 95% CI: 1.04 to 1.21) or for total alcohol consumption (drinks/week) (multivariable RR: 1.11; 95% CI: 1.03 to 1.20). Binge drinking of liquor (multivariable RR: 1.10; 95% CI: 1.02 to 1.19) and wine (multivariable RR: 1.24; 95% CI: 1.08 to 1.43), but not beer (multivariable RR: 1.02; 95% CI: 0.79 to 1.30), was associated with increased AF risk in a multivariable model adjusted for the other AF risk factors, mutually for liquor, wine, and beer, and for frequency of consumption of each alcoholic beverage.
When examining the association between total alcohol consumption and AF risk after excluding binge drinkers, the association was slightly attenuated but remained; multivariable RRs (95% CI) were 1.11 (95% CI: 0.98 to 1.26) for 15 to 21 drinks/week and 1.34 (95% CI: 1.17 to 1.53) for >21 drinks/week compared with current drinkers of <1 drink/week.
Seven prospective studies (2–7), including our own study, with a total of 12,554 cases of AF were eligible for inclusion in the meta-analysis (Online Fig. 1). Characteristics of the included studies are presented in Table 2. Alcohol consumption was assessed through a self-administered questionnaire in 6 studies (2,3,5–7) and by interviews in 1 study (4).
In a meta-analysis of all studies, we observed a linear dose-response relationship between alcohol consumption and AF risk, with no evidence of departure from linearity (p = 0.85). All studies reported a positive association, with an overall 8% (6% to 10%) increase in AF risk per 1 drink/day increment in alcohol consumption (Central Illustration). There was no heterogeneity among studies (p = 0.93) and no evidence of publication bias (p = 0.70). Results were identical when we combined the results by using a fixed effects model. Excluding our own study, which contributed 40% of the statistical weight, the overall results did not change (RR: 1.08; 95% CI: 1.05 to 1.10). Results were similar for studies conducted in the United States (RR: 1.09; 95% CI: 1.04 to 1.14) and Europe (RR: 1.08; 95% CI: 1.06 to 1.10). When stratified by sex, the RRs of AF were 1.08 (95% CI: 1.06 to 1.11) for men (5 studies) and 1.08 (95% CI: 1.03 to 1.13) for women (6 studies), without heterogeneity among studies (men: p = 0.89; women: p = 0.71).
In this large prospective study, both moderate (1 to 3 drinks/day) and high (>3 drinks/day) alcohol consumption was associated with increased AF risk, and the associations persisted after excluding binge drinkers. With regard to specific alcoholic beverages, consumption of liquor and wine but not beer was significantly positively associated with AF risk. Our finding for total alcohol consumption was confirmed in a dose-response meta-analysis, including our own study and 6 other prospective studies, showing an overall 8% increased AF risk for each 1 drink/day increase of alcohol consumption. Results from individual studies were consistent, thus providing strong support for a causal association between alcohol consumption and risk of AF.
In our study cohort, we observed an association between binge drinking and increased AF risk. Although the link between binge drinking, as may occur during weekends or holidays, and cardiac arrhythmias has been known for a long time (19), few large prospective studies have examined the association between binge drinking and AF risk. Results from a study of 30,433 adults (≥55 years of age) with a history of cardiovascular disease or diabetes showed that among moderate drinkers, binge drinkers (>5 drinks/day at 1 time) had a statistically significant increased risk (29%) of AF compared with non-binge drinkers (7). As in our study, the positive association of moderate and high alcohol consumption with AF risk remained after excluding binge drinkers. These findings suggest that not only binge drinkers but also regular drinkers of moderate amounts of alcohol have an increased risk of developing AF.
Result from our meta-analysis is consistent with findings from 2 previous meta-analyses showing an 8% increase in AF risk for each 10 g/day increase of alcohol consumption (9) or per drink (equivalent of 12 g pure alcohol) and day (8). Those 2 meta-analyses included both prospective and case-control studies. We only included prospective studies, which are less susceptible to systematic bias compared with case-control studies. The current meta-analysis also included twice as many AF cases as the previous dose-response meta-analyses.
There are several speculations by which mechanisms alcohol consumption may increase AF risk. Studies in animals and humans have shown that acute and chronic alcohol ingestion leads to depression of heart function and also may result in cardiac conduction abnormalities and morphologic changes (19). A study in healthy men showed that acute moderate alcohol consumption was associated with an increased interatrial electromechanical delay (20). Long-term heavy alcohol consumption could lead to dilated cardiomyopathy with both supraventricular arrhythmias such as AF but also ventricular arrhythmias with an increased risk of sudden cardiac death (21,22). The association between alcohol consumption and AF also may be related to a shortening of the right atrial effective refractory periods (23). Other potential mechanisms include alterations in oxidative stress (24) and vagal activity (25), electrolyte imbalances (26), and hypertension (27).
Major strengths of our study include its prospective design, large sample size, large number of AF cases, and the nearly complete follow-up of participants through computerized linkage to population-based Swedish registries. Limitations include the observational design and that information on some of covariates, including smoking, BMI, and history of hypertension, was obtained through self-report only. Thus, we cannot rule out the possibility of residual confounding. Another limitation is that alcohol consumption was self-reported and measured only once. This will inevitably lead to some measurement error in the assessment of alcohol consumption. Because this study was prospective, any measurement error in alcohol consumption would be nondifferential and most likely lead to an attenuation of the true association between alcohol consumption and AF risk. Finally, the incidence AF in our study may be underestimated, because we only identified hospitalized and symptomatic cases.
One strength of our meta-analysis is the large sample size, thus providing high statistical power to detect even weak associations of light to moderate alcohol consumption with AF risk. Moreover, the prospective design of the included studies avoided the potential influence of recall and selection bias, which could be a problem in retrospective studies. Another limitation is that we cannot rule out the possibility that publication bias may have affected our results. Although we found no evidence of publication bias, tests for publication bias have low statistical power, especially when the number of studies is small.
Alcohol consumption is positively associated with risk of AF. Even moderate consumption of alcohol, which lowers the risk of other cardiovascular diseases (28), seems to slightly increase the risk of AF.
COMPETENCY IN MEDICAL KNOWLEDGE: Consumption of large amounts of alcohol, binge drinking, and even moderate alcohol intake are associated with an increased risk of developing AF.
COMPETENCY IN PATIENT CARE: The protective effect of consuming small to moderate amounts of alcohol to reduce the risk of ischemic heart disease should be balanced against the increased risk of developing AF.
TRANSLATIONAL OUTLOOK: The mechanisms by which consistent alcohol intake leads to the development of AF, and specifically whether this involves proarrhythmic toxicity, atrial myocytes degeneration, or associated factors or comorbidities in susceptible individuals requires additional studies.
For a supplemental table and figure, please see the online version of this article.
This study was supported by a research grant from the Swedish Research Council. All authors have reported that they have no relationships relevant to the contents of this paper to disclose.
- Abbreviations and Acronyms
- atrial fibrillation
- atrial flutter
- body mass index
- coronary heart disease
- confidence interval
- International Classification of Diseases
- myocardial infarction
- relative risk
- Received January 14, 2014.
- Revision received March 8, 2014.
- Accepted March 11, 2014.
- American College of Cardiology Foundation
- Benjamin E.J.,
- Chen P.S.,
- Bild D.E.,
- et al.
- Mukamal K.J.,
- Tolstrup J.S.,
- Friberg J.,
- et al.
- Liang Y.,
- Mente A.,
- Yusuf S.,
- et al.
- Kodama S.,
- Saito K.,
- Tanaka S.,
- et al.
- Messerer M.,
- Johansson S.E.,
- Wolk A.
- Larsson S.C.,
- Orsini N.
- Larsson S.C.,
- Orsini N.
- Egger M.,
- Davey Smith G.,
- Schneider M.,
- et al.
- Denison H.,
- Jern S.,
- Jagenburg R.,
- et al.
- Pearson T.A.