Critique 247: Effects of alcohol consumption on risk of stroke and systemic embolism among subjects with atrial fibrillation – 15 March 2021

Reddiess P,  Aeschbacher S, Meyre P, Coslovsky M, Kühne M, Rodondi N, et al, for the BEAT-AF and Swiss-AF investigators.  Alcohol consumption and risk of cardiovascular outcomes and bleeding in patients with established atrial fibrillation.  CMAJ 2020;193:E117-E123. doi: 10.1503/cmaj.200778

Authors’ Abstract

BACKGROUND: Little is known about the association between alcohol consumption and risk of cardiovascular events in patients with established atrial fibrillation (AF). The main aim of the current study was to investigate the associations of regular alcohol intake with incident stroke or systemic embolism in patients with established AF.

METHODS: To assess the association between alcohol consumption and cardiovascular events in patients with established AF, we combined data from 2 comparable prospective cohort studies that followed 3852 patients with AF for a median of 3.0 years. Patients were grouped into 4 categories of daily alcohol intake (none, > 0 to < 1, 1 to < 2 and ≥ 2 drinks/d). The primary outcome was a composite of stroke and systemic embolism. Secondary outcomes were all-cause mortality, myocardial infarction, hospital admission for acute heart failure, and a composite of major and clinically relevant nonmajor bleeding.  Associations were assessed using time-updated, multivariable-adjusted Cox proportional hazards models.

RESULTS: Mean age (± standard deviation) was 71 ± 10 years (28% were women and 84% were on oral anticoagulants).  We observed 136 confirmed strokes or systemic emboli. Compared with nondrinkers, adjusted hazard ratios for the primary outcome event were 0.87, 95% confidence interval (CI) 0.55–1.37 for > 0 to < 1 drinks/d; 0.70, 95% CI 0.39–1.25 for 1 to < 2 drinks/d; and 0.96, 95% CI 0.56–1.67 for ≥ 2 drinks/d (p for linear [quadratic] trend 0.71 [0.22]).  There was no significant association between alcohol consumption and bleeding, but there was a nonlinear association with heart failure (p for quadratic trend 0.01) and myocardial infarction (p for quadratic trend 0.007).

INTERPRETATION: In patients with AF, we did not find a significant association between low to moderate alcohol intake and risk of stroke or other cardiovascular events.  Our findings do not support special recommendations for patients with established AF with regard to alcohol consumption.

Forum Comments

Forum members realized that there are no data in this paper on the effect of alcohol on the initial development of atrial fibrillation (AF), as all subjects had some level of AF at the beginning of the study.  They noted that levels of alcohol intake for each subject were taken at baseline and then at each annual examination to update alcohol exposure; this was done to take into account changes in alcohol intake over time (even though the authors state that “there was very little change in the distribution of alcohol intake over time”).  Further, the authors included in their analyses an appropriate, long list of potential confounders or modifiers of effect.  Thus, it is considered that, overall, this was a well-done study.  However, Forum members thought it unfortunate that the effects of alcohol consumption on the presence or absence of AF during follow up were not reported.

Forum member Finkel stated:  “I think that recurrence/persistence of AF is a big issue.  It is believed by many to be precipitated by alcohol, even in small quantities, either as a paroxysmal or stable arrhythmia (Voskoboinik, et al, 2016; Voskoboinik, et al, 2020).  While this study seems well done, I note that it included subjects who at baseline were diagnosed with ‘paroxysmal, persistent, or permanent atrial fibrillation,’ and I find no information on the relationship of quantity or pattern of drinking with subsequent control of AF.  This is a big issue, for AF is common and may have major sequelae.  Most physicians probably agree that excessive drinking may predispose toward AF, and that moderation or abstinence may help in its alleviation.  Acknowledging the other substantial cardiovascular benefits of moderate drinking, thoughtful physicians believe that the risk of adverse events associated with light to moderate alcohol consumption in gradual aliquots is often exaggerated.  Some modify their views.”

Reviewer Ellison agreed with Finkel that it is problematic that there are no results in this paper on the effects of alcohol intake on AF itself.  “While subjects who ‘had only short, potentially reversible AF episodes’ were not included in the two parent studies, the current paper reports that almost one-half of all subjects in their analyses (56.5% in one study and 44.8% in the other) had a diagnosis of ‘paroxysmal AF’ at baseline.  It would have been helpful to know specifically if alcohol consumption was related to subsequent persistence or disappearance of AF, based on the results of the yearly follow-up examinations.

“The primary outcomes of this study were the risk of two important adverse entities associated with AF, major stroke or systemic embolism and major bleeding episodes; the authors also reported effects of alcohol intake on hospitalization for heart failure, all-cause mortality, and myocardial infarction.  It was noted that while the authors do not emphasize them, the most striking results of this study are the marked decrease in risk of all-cause mortality (fully adjusted HRs were 0.61 for light and 0.49 for moderate drinkers) and myocardial infarction (HRs 0.41, 0.39, respectively) associated with light and moderate drinking, in comparison with abstinence.  For moderate drinkers, there was also a significant decrease in acute heart failure.”

Reviewer Skovenborg wrote:  “I agree that the results of the prospective cohort studies of patients with AF are plausible and encouraging. The group of nondrinkers had a low level of exercise, education and health perception and a high level of diabetes, renal failure and current falls; however, the relevant adjustments seems to be adequate and thorough.

“The lack of increased risk of bleeding from alcohol intake in the patients taking anticoagulants for AF is reassuring but not surprising.  The same results were found in a study of 1,244 men enrolled in the Post-Coronary Artery Bypass Graft (CABG) Trial who had undergone previous coronary bypass surgery (Mukamal, et al).  Moderate drinking did not adversely influence the safety of low-dose warfarin among men in this randomized trial, as measured by INR levels.  There is even experimental evidence that (1) 10 and even 20 oz of daily mealtime wine has no effect on therapeutic hypoprothrombinemia (O’Reilly, 1979) and (2) intake of 296 mL/day of fortified wine during fasting has no effect on therapeutic levels of hypoprothrombinemia of normal humans (O’Reilly, 1981).  The code word regarding use of alcoholic beverages for patients on oral anticoagulants is regular intake.  Binge drinking is never sensible and even more dangerous combined with anticoagulant medication.  I did note, however, that the AF patients in the two cohort studies described here did not include a large number of heavy or binge drinkers, so no data are available for such patients.”

Forum member de Gaetano noted: “This study clearly shows that, independently of other clinical outcomes, which can be influenced by different variables and factors, in patients with an established diagnosis of AF all-cause mortality is significantly reduced by moderate alcohol consumption, following a J shaped curve.  This is indeed the great message of this paper.  In all our studies we have constantly considered total mortality as a key outcome to be considered in alcohol consumption research.”

Statistical problems in separating continuing alcohol consumption prior to AF with the initiation of alcohol intake following the diagnosis of AF:  While the results of this paper are very interesting, reviewer Zhang pointed out problems in judging effects of alcohol among subjects who may, or may not, have been regular drinkers prior to the baseline observations.  He wrote: “One should always consider potential collider bias, a kind of selection bias (Dahabreh & Kent).  This may happen when studying a risk factor (here, alcohol) in relation to an outcome (here, bleeding or stroke), conditioning on a potential intermediate variable (here, atrial fibrillation).  If alcohol consumption is associated with the risk of AF, and AF may affect the risk of bleeding or stroke, then the obtained effect estimate from such kinds of analysis may represent an ‘indirect effect’ (i.e., the effect of alcohol on the risk of bleeding/stroke but not necessarily through AF), and this effect estimate does not represent the total effect of alcohol on the risk of bleeding/stroke among AF patients. Unless one assesses the incident alcohol exposure, i.e., people who started to drink alcohol after the diagnosis of AF (which is unlikely) or changed their alcohol consumption after developing AF, studies using alcohol information measured before AF diagnosis and assessing its relation to the risk of bleeding or stoke generates an indirect effect estimate; this estimate tends to bias the results towards the null or sometimes in an opposite direction.”

Added reviewer Ellison: “I agree with this problem.  In other words, the presence of AF at baseline may relate to the subject’s previous alcohol exposure, which could have led perhaps to an increase in the risk of AF, or perhaps a decrease in the risk of any underlying heart disease.  Thus, the absence or presence of previous alcohol use among subjects with AF at baseline may have resulted to two different groups: those whose AF was somehow related to previous drinking and those whose AF was not related to previous drinking (as they were abstainers).  This makes it difficult to judge alcohol during follow up as a risk factor for subsequent AF and may even lead to paradoxical effects.”  As stated by Dahabreh & Kent, “Because risk factors often have congruent effects on the index and recurrent events, this negative association will tend to bias any estimation of the effect [of the exposure] on recurrence risk toward the null, unless there is a thorough accounting for all shared risk factors.”

Having to rely on self-reported alcohol consumption.  The present analyses, as are essentially all epidemiologic studies of the health effects of alcohol consumption, must rely on self-reported drinking.  However, given that our estimation of the exposure to alcohol is known to be imprecise, it is remarkable that in almost all epidemiologic studies over many decades, self-reported alcohol levels have consistently been shown to relate to certain outcomes (cardiovascular disease, all-cause mortality), almost uniformly showing a J-shaped relationship (a decrease in risk with moderate drinking, usually an increase with heavier drinking).

Still, there have been attempts to obtain more precise estimates of exposure to alcohol.  Klatsky and his colleagues, using data from the Kaiser-Permanente cohort, have shown that investigators can often use other information collected in their databases that relates to excessive alcohol intake (such as diagnoses of alcoholic use disorders, alcoholic cirrhosis, etc.) to estimate whether or not individuals are under-reporting their alcohol intake or giving realistic amounts (Klatsky, et al, 2006; Klatsky, et al, 2014).  These investigators have concluded that many of the adverse effects attributed to light or moderate drinking (including most cancers and hypertension) occur only among subjects estimated to be under-reporters of their intake.

In reviewing the present paper, Forum member Mattivi writes:  “I suggest that the time has come to include in observational cohort studies additional lab tests to complement the self-reported alcohol intake, in search of the possible presence of heavy or binge drinkers who are under-reporting their intake. Several biomarkers for intake of ethanol have been suggested, and possibly the detection of ethyl glucuronide in hair (alone or in combination with ethyl palmitate) is the most promising, since it is reflecting the consumption over a long period of time. As a biomarker of long-term alcohol use, in addition to validating the self-reported measures of alcohol use it gives the additional possibility of detecting the presence of harmful drinking patterns (Iglesias, et al) and it has been shown capable of supporting the identification of severe and high intake (Bastiani, et al).”


Bastiani MF, Lizot LLF, Da Silva ACC, Hahn RZ, Dries SS, Perassolo MS, Antunes MV, Linden R.  Improved measurement of ethyl glucuronide concentrations in hair using UPLCMS/MS for the evaluation of chronic ethanol consumption. Forensic Science International 2020;306:110071.

Dahabreh IJ, Kent DM.  Commentary: Index Event Bias as an Explanation for the Paradoxes of Recurrence Risk Research.  JAMA 2011;305:822-823.

Iglesias K, Lannoy S, Sporkert F, Daeppen J-B, Gmel G, Baggio S.  Performance of self reported measures of alcohol use and of harmful drinking patterns against ethyl glucuronide hair testing among young Swiss men. PLoS ONE 2020;15:e0244336. pone.0244336.

Klatsky AL, Gunderson EP, Kipp H, Udaltsova N, Friedman GD (2006) Higher prevalence of systemic HTN among moderate alcohol drinkers: exploring the role of under-reporting. J Stud Alcohol 2006;67:421–428.

Klatsky AL, Udaltsova N, Li Y, Baer D, Tran HN, Friedman GD.  Moderate alcohol intake and cancer: the role of underreporting.  Cancer Causes Control 2014;25:693–699.  DOI 10.1007/s10552-014-0372-8

Mukamal KJ, Smith CC, Karlamangla AS, Moore AA.  Moderate alcohol consumption and safety of lovastatin and warfarin among men: the post-coronary artery bypass graft trial.  Am J Med 2006;119:434-440.)

O’Reilly RA.  Lack of effect of mealtime wine on the hypoprothrombinemia of oral anticoagulants.  Am J Med Sci 1979;277:189-194.

O’Reilly RA.  Lack of effect of fortified wine ingested during fasting and anticoagulant therapy. Arch Intern Med 1981;141:458-459.

Voskoboinik A, Prabbu S, Ling L-B, et al. Alcohol and atrial fibrillation: A sobering review. JACC 2016;68:2567-2576.

Voskoboinik A, Kalman JM, Da Silva A, et al. Alcohol abstinence in drinkers with atrial fibrillation. N Engl J Med 2020;382:20-28.

Forum Summary

The main aim of the current study was to investigate the associations of regular alcohol intake with incident stroke or systemic embolism in patients with established atrial fibrillation (AF), most of whom (84%) were on anti-coagulant therapy.  The authors used combined data from two prospective studies of subjects with AF, followed for an average of 3 years.  While they excluded subjects with “only short, potentially reversible AF episodes (e.g., as after cardiac surgery or sepsis)” about one half of subjects in the study were diagnosed as having paroxysmal AF.  Reported alcohol at baseline and at subsequent examinations was classified as none, > 0 to < 1 drink/day, 1 to < 2 drinks/day, and ≥ 2 drinks/day.

The primary outcome was a composite of stroke and systemic embolism.  Secondary outcomes were all-cause mortality, myocardial infarction, hospital admission for acute heart failure, and a composite of major and clinically relevant nonmajor bleeding.  In their analyses, they adjusted for age, sex, education, hypertension, history of heart failure, history of diabetes, BMI, smoking status, physical activity, history of stroke, history of coronary heart disease, oral anticoagulation, history of renal failure, AF type, and health perception.  With annual assessments, the investigators updated all covariates over time, if appropriate.

The main results indicated no significant effect of alcohol consumption on the risk of the major outcomes (stroke or systemic embolism).  However, there were significant decreases in risk for several secondary outcomes.  For hospital admissions for heart failure, in comparison with non-drinkers the HR for moderate drinkers was 0.60 (CI 0.41-0.87); further, among moderate drinkers the HR for myocardial infarction was 0.39 (CI 0.20 – 0.78), and for all-cause mortality the HR was 0.49 (CI 0.35-0.69).

Forum members had two major concerns with the paper.  First, the authors failed to report specifically how alcohol consumption was associated with increased or decreased presence of AF during follow up, even though approximately one-half of subjects had only paroxysmal AF at baseline. Determining if alcohol consumption affected subsequent AF would be information of importance to practitioners who are advising their patients with AF regarding alcohol consumption.

Secondly, it was pointed out by Forum members that the results may have been affected by what is known as collider bias: as moderate drinking shows an inverse association with coronary heart disease (which is strongly related to the presence of AF), subjects who consumed alcohol prior to the baseline diagnosis of AF were probably different from those who had no prior alcohol exposure, and the results of combining these two groups (as was done in these analyses) may have been biased.  Such bias usually tends to result in estimates of effect going toward the null.

While the reported results of no real effect of alcohol consumption on the risk of stroke or systemic embolism among patients with AF is encouraging, to decrease the risk of bias in future studies it will be important that such research attempts to separate subjects according to their prior use of alcohol before baseline measurements are assessed.  The marked reduction in the risk of other major cardiovascular outcomes and total mortality, found in this study to be associated with light to moderate drinking, matches the results seen in most previous studies.

*                      *                      *                      *                      *                      *

Comments from the following members were provided for this critique by the International Scientific Forum on Alcohol Research:

Giovanni de Gaetano, MD, PhD, Department of Epidemiology and Prevention, IRCCS Istituto Neurologico Mediterraneo NEUROMED, Pozzilli, Italy

R. Curtis Ellison, MD, Professor of Medicine, Section of Preventive Medicine & Epidemiology, Boston University School of Medicine, Boston, MA, USA

Ramon Estruch, MD, PhD, Hospital Clinic, IDIBAPS, Associate Professor of Medicine, University of Barcelona, Spain

Harvey Finkel, MD, Hematology/Oncology, Retired (Formerly, Clinical Professor of Medicine, Boston University Medical Center, Boston, MA, USA)

Fulvio Mattivi, MSc, CAFE – Center Agriculture Food Environment, University of Trento, via E. Mach 1,  San Michele all’Adige, Italy

Erik Skovenborg, MD, specialized in family medicine, member of the Scandinavian Medical Alcohol Board, Aarhus, Denmark

Creina Stockley, PhD, MSc Clinical Pharmacology, MBA; Principal, Stockley Health and Regulatory Solutions; Adjunct Senior Lecturer, The University of Adelaide, Adelaide, Australia

Arne Svilaas, MD, PhD, general practice and lipidology, Oslo University Hospital, Oslo, Norway

Pierre-Louis Teissedre, PhD, Faculty of Oenology–ISVV, University Victor Segalen Bordeaux 2, Bordeaux, France

David Van Velden, MD, Dept. of Pathology, Stellenbosch University, Stellenbosch, South Africa

Yuqing Zhang, MD, DSc, Clinical Epidemiology, Massachusetts General Hospital, Harvard Medical School, Boston, MA, USA.