Critique 093: Is moderate alcohol consumption associated with an increased risk of atrial fibrillation among patients with cardiovascular disease? ————– 8 October 2012

Liang Y, Mente A, Yusuf S, Gao P, Sleight P, Zhu J, Fagard R, Lonn E, Teo KK; for the ONTARGET and TRANSCEND Investigators.  Alcohol consumption and the risk of incident atrial fibrillation among people with cardiovascular disease.  CMAJ 2012. DOI:10.1503/cmaj.120412.

Authors’ Abstract

Background: Moderate alcohol consumption may reduce cardiovascular events, but little is known about its effect on atrial fibrillation in people at high risk of such events.  We examined the association between moderate alcohol consumption and the risk of incident atrial fibrillation among older adults with existing cardiovascular disease or diabetes.

Methods: We analyzed data for 30 433 adults who participated in 2 large antihypertensive drug treatment trials and who had no atrial fibrillation at baseline.  The patients were 55 years or older and had a history of cardiovascular disease or diabetes with end-organ damage.  We classified levels of alcohol consumption according to median cut-off values for low, moderate and high intake based on guidelines used in various countries, and we defined binge drinking as more than 5 drinks a day.  The primary outcome measure was incident atrial fibrillation.

Results: A total of 2093 patients had incident atrial fibrillation.  The age- and sex- standardized incidence rate per 1000 person-years was 14.5 among those with a low level of alcohol consumption, 17.3 among those with a moderate level and 20.8 among those with a high level.  Compared with participants who had a low level of consumption, those with higher levels had an increased risk of incident atrial fibrillation (adjusted hazard ratio [HR] 1.14, 95% confidence interval [CI] 1.04–1.26, for moderate consumption; 1.32, 95% CI 0.97–1.80, for high consumption).  Results were similar after we excluded binge drinkers.  Among those with moderate alcohol consumption, binge drinkers had an increased risk of atrial fibrillation compared with non–binge drinkers (adjusted HR 1.29, 95% CI 1.02–1.62).

Interpretation: Moderate to high alcohol intake was associated with an increased incidence of atrial fibrillation among people aged 55 or older with cardiovascular disease or diabetes.  Among moderate drinkers, the effect of binge drinking on the risk of atrial fibrillation was similar to that of habitual heavy drinking.

Forum Comments

Background:  Results from previous studies are mixed on the effects of light to moderate drinking on the risk of atrial fibrillation (AF), a common condition that increases the risk of stroke.  The present study evaluates the development of AF among subjects who have been diagnosed with CVD or diabetes according to their alcohol consumption following diagnosis of their underlying condition.

Comments on the present study:  This study was done only among subjects who already have cardiovascular disease (CVD) or diabetes, so the results may not be applicable to the general population.  Further, using such subjects raises questions about the characteristics of the subjects – those who developed CVD despite being drinkers may be different from those that were never drinkers who developed CVD.  (Discussed below under “Index event bias.”)

The authors apparently used 12-15 grams of alcohol as a “drink.”  The range of “low” alcohol was less than 1 drink per week, so this group included non-drinkers and occasional drinkers.  The large majority of subjects were in the low group (n=18,775) or moderate groups (n=11,139).  There were only 519 subjects in the high group.  Drinkers in the high group were predominantly male and smokers, were much more likely to have peripheral vascular disease, less likely to have hypertension or diabetes, and much more likely to be in studies from Europe or Australia/Asia than from the Americas.

The authors report that the risk of death among their subjects during follow up was 12.5% of those in the low alcohol group, 9.9% of those considered to be moderate drinkers, and 10.6% of those in the high alcohol group.  In comparison with subjects in the low alcohol group, the fully adjusted hazard ratios for death in the moderate group was 0.79 (CI=0.73, 0.85) and for those in the high-consumption group it was 0.84 (CI=0.64, 1.10).  When adjusting for such competing risks in a regression model, the hazard ratio for AF was 1.16 (CI=1.05, 1.27) for the moderate group and 1.33 (CI=0.98-1.92) for the high-alcohol group.  Thus, these results suggest that even moderate drinkers may have a slightly increased risk of AF.

Binge drinking was associated with higher risk of AF.  Most subjects who could be classified as binge drinkers (n=1,204) were included in the moderate or high groups (which totaled 10,454), but the authors mainly present their binge/no binge comparisons for the moderate and high groups combined.  The authors state that the estimated hazard ratios (HR) for AF when binge drinkers were excluded were very different for moderate drinkers (HR=1.13) than for heavy drinkers (1.54), so combining them into a single comparison group seems unusual.

While this appears to be a well-done study by respected scientists, Forum reviewers had two major concerns about the paper: the wide range of alcohol consumption referred to as “moderate,” and potential bias in the results by a phenomenon known as index event bias.

Range of alcohol consumption considered to be “moderate” drinking:  Of concern to reviewers is the wide range chosen by the authors for the category of “moderate” drinking, which included subjects reporting from 1 drink/week up through those reporting 21 drinks/week for men and 14 drinks/week for women.  It would have been preferable to have results given for subjects reporting no more than 14 drinks/week for men and 7 drinks/week for women, as these are the guidelines for “moderate” or sensible drinking in the United States and many other countries.  As pointed out by Forum reviewer Stockley, the upper level of alcohol consumption considered to be “moderate” in this paper also exceeds such limits in Australia.  There is a conspicuous absence of data in this paper on the effects on the risk of AF of subjects consuming “moderately” by these more common drinking guidelines.

Reviewer Svilaas commented: “The results (and conclusions) are weakened by the wide ‘moderate’ definition of alcohol consumption.  Those patients with higher consumption also were more frequent current/former smokers.  This might also affect the outcome, despite the statistical adjustment for smoking.”  He adds: “A strength of the study is the high number of study participants.  However, as the participants already have cardiovascular disease and/or diabetes, the results cannot be applied to a healthy population of the same age.”

Forum reviewer Djoussé comments: “Given the likelihood of under-reporting of alcohol use, the ‘moderate’ group might include a substantial number of heavy drinkers.  Despite sound statistical methods utilized by the authors, this generous classification of moderate drinking dilutes the clinical and public health utility of the results, as most investigators agree that heavy drinking has more harmful than beneficial effects, whereas light-to-moderate drinking is generally found to confer primarily health benefits.”

Index event bias:  There was also concern by Forum reviewers that the results of these analyses may suffer from what is known in epidemiology as index event bias or collider bias.  Since the exposure being evaluated (alcohol) affects the development of CHD and diabetes, it is likely that the subjects in this study (of whom 77% had coronary artery disease and 37% had diabetes at baseline) differed according to their previous alcohol intake (being either a drinker or a non-drinker prior to developing CVD or diabetes).  This is an epidemiologic problem that has increasingly become recognized and discussed.  This topic was well described recently by Dahabreh and Kent,1 who point out problems when evaluating a risk factor (e.g., alcohol) among subjects selected for follow up of a particular disease in which not only the occurrence of the disease is affected by the risk factor but the course of the disease (e.g., including the development of AF) may also be affected by the same risk factor.

This is analogous to findings that have shown that while obesity is an important risk factor for the development of CVD, it may have the opposite effect on the subsequent course among subjects who have already developed CVD (the “obesity paradox”).2 Similarly, among other diseases showing such an effect, an “aspirin paradox” has been described in that aspirin tends to decrease the risk of developing coronary disease initially but not for cardiovascular events subsequent to the development of coronary disease.3

Forum reviewer Zhang comments on this phenomenon as follow:  “If moderate alcohol consumption reduces the risk of CVD and diabetes (as has been shown in almost all studies), subjects in this analysis who drank alcohol moderately and still developed CVD or diabetes presumably had other genetic or environmental risk factors that contributed to their disease.  If these other factors also cause AF and were not appropriately adjusted for, then index event bias could occur.  In general, this type of bias would attenuate any protective effect of moderate alcohol towards the null, but magnify the potential harmful effect away from the null; even effects in the opposite direction could occur.”

The interpretation of the results of this study depends on our clear understanding of the causal pathway of the effects of moderate alcohol consumption on CVD and diabetes, as well as on AF.  The implications of the present study will be better defined after further research on alcohol and AF among subjects who have already developed CVD and/or diabetes.

There is very strong and consistent scientific data on the effects of heavy bouts of alcohol consumption resulting in AF as part of the “holiday heart syndrome.”4 While the arrhythmias in this syndrome can sometimes be serious, they are usually transient, and clear after a short period of abstinence.  In terms of the effects on the risk of AF of moderate drinking (especially when “moderate’ refers to most common drinking guidelines), Forum reviewers believe that problems in the present study weaken the conclusions of the authors.  The association of truly light-to-moderate drinking with the risk of AF remains unclear.

References from Forum critique

1.  Dahabreh IJ, Kent DM.  Commentary: Index event bias as an explanation for the paradoxes of recurrence risk research.  JAMA 2011;305, reprint No. 8.

2.  Gruberg L, Weissman NJ, Waksman R, et al.  The impact of obesity on the short-term and long-term outcomes after percutaneous coronary intervention: the obesity paradox?  J Am Coll Cardiol 2002;39:578-584.

3.  Rich JD, Cannon CP, Murphy SA, Qin J, Giugliano RP, Braunwald E.  Prior aspirin use and outcomes in acute coronary syndromes.  J Am Coll Cardiol. 2010;56:1376-1385.

4.  Ettinger PO, Wu CF, De La Cruz C, Weisse AB, Ahmed SS, Regan TJ.  Arrhythmias and the “Holiday Heart”: alcohol-associated cardiac rhythm disorders. Am Heart J 1978; 95:555–562.

Forum Summary

An analysis of the association of alcohol consumption with the development of atrial fibrillation (AF) among subjects with coronary heart disease, stroke, diabetes, or other manifestations of cardiovascular disease (CVD) was based on subjects in two large antihypertensive drug treatment trials.  Previous research in the general population has suggested an increase in the risk of the development of AF for heavy drinkers, and the present study shows such an association among subjects who already have CVD.

Among subjects in this study (all of whom had previous CVD), the authors also report that even “moderate” drinkers had a higher risk of AF than low-alcohol consumers, although the risk of death during follow up of moderate drinkers (9.9%) was lower than that of subjects reporting low-levels of drinking(12.5%).  Excluding binge drinkers, the estimated risk of AF was about 13% higher in moderate drinkers than among subjects classified as low-alcohol consumers.

While the multiple analyses in this paper were done appropriately, Forum reviewers were concerned about two aspects of this study.  First, there was concern about the wide range the authors chose for the category of “moderate” drinkers, which included subjects reporting from 1 drink per week up through those reporting 21 drinks per week for men and 14 drinks per week for women.  It would have been useful to present results also for subjects who met more common definitions for moderate drinking: no more than 14 drinks/week for men and 7 drinks/week for women, the values used to define moderate drinking in the United States, Australia, and many other countries.  Unfortunately, there is a conspicuous absence of data in this paper of the effects on the risk of AF of subjects consuming alcohol “moderately” by these standards.

Another major concern about these analyses relates to potential bias in the estimates from what is known as index event bias or collider bias.  Given that alcohol intake prior to enrollment in this study may well have related to subjects’ development of cardiovascular disease, it is problematic to judge the effects of alcohol after a cardiovascular event on the subsequent risk of AF, a condition frequently associated with CVD.

All subjects in these analyses had already developed CVD (the prerequisite for being in the present study).  Thus, the study included those who consumed alcohol prior to the diagnosis of CVD or diabetes and those who did not drink prior to these diagnoses.  Given that moderate alcohol consumption has been shown to reduce substantially the risk of both CVD and diabetes, it can be assumed that subjects in this study who developed CVD despite being drinkers had other risk factors contributing to the disease that overcame any “protection” afforded by alcohol consumption.  Unless adjusted for, these other risk factors could well affect the subsequent course of subjects following the onset of CVD, including the development of AF.  A similar phenomenon has been seen for obesity (the “obesity paradox”), aspirin use (the “aspirin paradox”), and other exposures, where the associations with CVD seen prior to the initial development of the disease differ from those seen after the development of CVD.

It is clear from many previous studies in the general population that heavy alcohol intake and binge drinking increase the risk of developing atrial fibrillation.  This cardiac arrhythmia is a common component of the “holiday heart” syndrome that may occur after very heavy bouts of drinking.  As for the effects on the risk of atrial fibrillation from moderate drinking, as studied in this paper, there were a number of concerns from Forum reviewers about the analyses and results.  They raise questions about the conclusions of the authors that even “moderate” drinking results in an increased risk of atrial fibrillation after the development of cardiovascular disease.  Especially when defined as no more than 14 drinks per week for men or 7 drinks per week for women, the association between “moderate” alcohol consumption and atrial fibrillation remains unclear.

*                      *                      *

Contributions to this critique by the International Scientific Forum on Alcohol Research were made by the following members:

Harvey Finkel, MD, Hematology/Oncology, Boston University Medical Center, Boston, MA, USA

Gordon Troup, MSc, DSc, School of Physics, Monash University, Victoria, Australia

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

Yuqing Zhang, MD, DSc, Epidemiology, Boston University School of Medicine, Boston, MA, USA

Creina Stockley, PhD, MBA, Clinical Pharmacology, Health and Regulatory Information Manager, Australian Wine Research Institute, Glen Osmond, South Australia, Australia

Luc Djoussé, MD, DSc, Dept. of Medicine, Division of Aging, Brigham & Women’s Hospital and Harvard Medical School, Boston, MA, USA

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