By Michael H. Crawford, MD, Editor

SYNOPSIS: An analysis of the UK Biobank database revealed low levels of alcohol consumption, especially with wine and spirits, is associated with the lowest incidence of atrial fibrillation.

SOURCE: Tu SJ, Gallagher C, Elliott AD, et al. Risk thresholds for total and beverage-specific alcohol consumption and incident atrial fibrillation. JACC Clin Electrophysiol 2021 Jul 19;S2405-500X(21)00524-7. doi: 10.1016/j.jacep.2021.05.013. [Online ahead of print].

Binge drinking alcohol can precipitate atrial fibrillation (AF), but little is known about low to moderate consumption, the importance of the beverage type, or sex of the drinker. Tu et al analyzed the UK Biobank database to provide information on these uncertainties.

The UK Biobank contains information about more than 500,000 subjects age 40-69 years who were enrolled between 2006 and 2010. Alcohol consumption data were obtained from questionnaires and corrected for variable consumption over time in the 12% of the population who provided repeat questionnaires during the study period. Investigators excluded subjects with a history of AF and those who had quit drinking alcohol to reduce the effect of reverse causality. Alcohol consumption was measured as standard U.K. drinks/week (standard U.K. drink is 8 g of alcohol; the standard U.S. drink is 14 g of alcohol). Binge drinking was defined as ≥ 6 drinks per day for women and ≥ 8 drinks/day for men. Frequent drinking was defined as ≥ 3 days/week. The primary outcome was incident AF (atrial flutter was included). Various covariates or comorbidities detected on the initial questionnaire or hospital records were used to adjust the data. Sensitivity analyses were conducted for those who developed AF in the first two years by excluding those with known heart disease, including ex-drinkers, and the consumption of other beverages. Exploratory analyses were conducted on the effect of sex, binge drinking, and frequent drinking.

The final population included 403,281 subjects (mean age = 58 years), of whom 52% were women and 94% were white. Median alcohol consumption was eight standard U.K. drinks/week, and 5.5% did not drink alcoholic beverages. After dividing alcohol consumption into quartiles, the higher the number of drinks, the lower the percentage of women and the higher the percentage of smokers. The highest quartile (> 28 drinks/week) included more men and more comorbidities.

Over the 11-year follow-up, 21,312 people developed AF over 4.5 million person-years. The primary outcome of incident AF exhibited a J-shaped curve, with a nadir at five drinks/week. The lowest quartile (one to seven drinks/week) exhibited a lower risk of AF than that observed in non-drinkers (HR, 0.91; 95% CI, 0.86-0.96).

Higher consumption was associated with increasing risk of AF, but it varied with the type of alcoholic beverage. The lowest risk in beer drinkers was at zero drinks/week, whereas in red wine, white wine, and spirits drinkers, the lowest risk was at less than 10 drinks/week, 8 drinks/week, and four drinks/week, respectively, with a nadir at five drinks/week, four drinks/week, and one drink/week, respectively. In the low consumption quartile (one to seven drinks/week), of 70,683 subjects, 18,550 drank beer, 28,220 drank red wine, and 16,934 drank white wine. Compared to beer drinkers, wine drinkers were at lower risk for AF (HR, 0.83; 95% CI, 0.75-0.92 for red wine and HR, 0.89; 95% CI, 0.79-1.0 for white wine).

Sensitivity analyses did not alter the results. Higher risk was observed in binge drinkers and frequent drinkers. Analysis by sex only showed an increased risk with women who drank spirits. The authors concluded low levels of alcohol consumption were associated with the lowest AF risk, and low levels of wine and spirits may not be associated with increased AF risk. Any consumption of beer may increase risk.


When considering heart disease in general, a J-shaped curve relationship between alcohol consumption and risk has been demonstrated and an upper limit of alcohol consumption to realize this heart disease risk reduction has been proposed (14 drinks/week). Regarding AF specifically, there has been controversy with some studies showing a linear relationship between alcohol consumption and the risk of incident AF.1 This UK Biobank study was designed to address this controversy and examine the importance of the type of beverage and sex on the risk of AF. The authors confirmed previous data showing binge drinking is associated with AF. However, at the low to moderate drinking level, there is a J-shaped curve, with low levels exhibiting lower risk, but at an alcohol consumption level lower than for heart disease protection generally (< 7 drinks/week vs. < 14 drinks/week). In addition, in their exploratory analyses, Tu et al demonstrated that at low levels of wine or spirit consumption, there was almost no increased risk of AF, but any beer consumption raised the risk. Since the beverage type analyses were not adjusted for multiple testing, these results should be considered exploratory. Finally, these results were not related to sex.

The strengths of this study include the large database with beverage-specific information. Also, the authors excluded ex-drinkers and those with a history of AF to mitigate any reverse causality. In addition, Tu et al corrected for regression dilution bias by adjusting for measurement error and long-term variability in the data.

There also were some weaknesses. Although the data were collected prospectively, it was analyzed retrospectively and there could have been residual confounding and reverse causality despite the measures taken to reduce these biases. For example, beer and spirit drinkers were more morbid than wine drinkers. Self-reporting is likely to underestimate alcohol consumption, although adjustment based on the subgroup that was questioned a second time should have helped abrogate this concern.

There were no data on the type of AF (i.e., paroxysmal, persistent, permanent) and no outpatient data were captured. All the disease-specific data were based on hospital ICD codes. The population studied was almost all white, and the UK Biobank may suffer from the healthy volunteer bias. Finally, there were no data on prevalent AF since patients with known AF were excluded.

Considering the size and strengths of this study, it would seem that advising against all alcohol consumption to prevent AF probably is not warranted. Given the other potential heart disease preventive effects of alcohol, low to moderate consumption could be safe and might be beneficial. However, the data are not strong enough to recommend non-drinkers start drinking.


  1. Csengeri D, Sprünker NA, Di Castelnuovo A, et al. Alcohol consumption, cardiac biomarkers, and risk of atrial fibrillation and adverse outcomes. Eur Heart J 2021;42:1170-1177.