- Eric B Rimm, assistant professor of epidemiology and nutritiona,
- Arthur Klatsky, senior consultant in cardiologyb,
- Diederick Grobbee, professor of epidemiologyc,
- Meir J Stampfer, professor of epidemiology and nutritiona
- a Department of Nutrition, Harvard School of Public Health, Boston, MA 02115, USA
- b Kaiser Permanente Medical Center and the Division of Research Group, Oakland CA, USA
- c Department of Epidemiology and Biostatistics, Erasmus University Medical School, Rotterdam, Netherlands
- Correspondence to: Dr Rimm.
- Accepted 18 December 1995
Objectives: To review the effect of specific types of alcoholic drink on coronary risk.
Design: Systematic review of ecological, case-control, and cohort studies in which specific associations were available for consumption of beer, wine, and spirits and risk of coronary heart disease.
Subjects: 12 ecological, three case-control, and 10 separate prospective cohort studies.
Main outcome measures: Alcohol consumption and relative risk of morbidity and mortality from coronary heart disease.
Results: Most ecological studies suggested that wine was more effective in reducing risk of mortality from heart disease than beer or spirits. Taken together, the three case-control studies did not suggest that one type of drink was more cardioprotective than the others. Of the 10 prospective cohort studies, four found a significant inverse association between risk of heart disease and moderate wine drinking, four found such an association for beer, and four for spirits.
Conclusions: Results from observational studies, where alcohol consumption can be linked directly to an individual's risk of coronary heart disease, provide strong evidence that all alcoholic drinks are linked with lower risk. Thus, a substantial portion of the benefit is from alcohol rather than other components of each type of drink.
We examined the relation between specific alcoholic drinks and reduction of risk of coronary heart disease by summarising published reports from ecological, case-control, and cohort studies
Most ecological studies suggested that wine was more effective in reducing risk of mortality than beer or spirits, whereas the three case-control studies together did not suggest that one type of drink was more cardioprotective than others
Of the 10 prospective cohort studies, four found a significant inverse association between risk of heart disease and moderate wine drinking, four found the association for beer, and four found it for spirits.
The evidence suggests that all alcoholic drinks are linked with lower risk, so that much of the benefit is from alcohol rather than other components of each type of drink
The inverse association between moderate alcohol consumption and coronary heart disease is well established. Evidence for a causal interpretation comes from over 60 ecological, case-control, and cohort studies. Previous reviews have concluded that men and women who drink one to two drinks a day have the lowest risk of coronary heart disease.1 2 3 4 In a recent meta-analysis of cohort studies Maclure found a summary relative risk of coronary heart disease of 0.83 (95% confidence interval 0.77 to 0.89) for moderate drinkers (2-3 drinks a day) compared with teetotallers.2
Several reports give individual relative risks attributable to consumption of beer, wine, and spirits. Although the possible additional benefits of wine—especially red wine—have received considerable attention in the media (Sixty Minutes, CBS Television, 17 November 1991, 5 November 1995), whether any specific type of alcoholic drink has particular benefit has not been systematically addressed. In this review we examine the relation between specific alcoholic drinks and reduction of risk of coronary heart disease by summarising published reports from ecological, case-control, and cohort studies.
We searched the Medline database for epidemiological investigations of alcohol and coronary heart disease published after 1965. We supplemented our search through citations in review articles,1 2 3 4 proceedings of meetings and symposia, and journals dedicated to tracking alcohol related research. We selected studies that provided specific information on consumption of beer, wine, or spirits in relation to risk of coronary heart disease. We also included results from the few investigations in which the alcohol consumption in the study population was generally limited to only one or two of the three main types of drink.
Several observational studies gave relative risks of coronary heart disease for specific types of alcoholic drink. However, differences in categorisation of the amounts consumed and the extent of control for confounding factors precluded a useful meta-analysis or summary statistic. We discuss the strengths and weaknesses of each study design and provide an overall qualitative conclusion from the evidence.
Since 1979, 12 studies have examined the association between per capita consumption (or, more accurately, disappearance data) of specific alcoholic drinks and mortality from heart disease across countries or over time (table 1). Only mortality is used in the analyses summarised since there are no standardised rates for morbidity from coronary heart disease. St Leger et al published the first ecological analysis showing a strong inverse association between average per capita consumption of wine and mortality from ischaemic heart disease.5 The inverse association was pronounced for wine in both men and women, less strong for spirits, and non-existent for beer. The other studies generally reported similar results (table 1).
case-CONTROL AND COHORT STUDIES
Three case-control studies provided estimates of the relative risk of coronary heart disease in relation to intake of specific types of alcoholic drink (table 2). Hennekens et al16 and Rosenberg et al17 reported that each type of drink was associated with a reduction in risk of coronary heart disease. Among men spirits were most strongly associated with a reduced risk of death from coronary heart disease,16 while among women only wine was significantly associated with a reduction in risk of non-fatal myocardial infarction.17 Kaufman et al did not find an inverse association between total alcohol consumption and coronary heart disease18: each drink type was marginally positively associated with risk of myocardial infarction, with wine showing the strongest association. Taken together, these case-control studies do not suggest that one specific type of drink may be more cardioprotective.
Ten separate prospective cohort studies have provided results on risk of coronary heart disease in association with specific types of alcoholic drink (table 3). Only two of the studies did not simultaneously account for the effects of each alcoholic drink.19 21 Four of the studies reported a significant inverse association between wine consumption and coronary heart disease,24 25 29 31 four reported such an association for beer,19 23 29 31 and four reported an association for spirits.22 23 26 29 To determine whether one type of drink was most advantageous, we identified studies in which only one of the three drink types was strongly associated with reduced coronary heart disease. Two studies found wine to have the strongest inverse association with coronary heart disease,25 31 none found this for beer, and two found it for spirits.22 26 In many of these studies the other types of drink also showed trends towards protection; the significance for any single type of drink was partly a function of the distribution of intake of different drinks in the population studied.
Ecological studies of alcohol consumption and coronary heart disease are based on existing data typically collected by government agencies or international surveillance programmes. In addition to easy accessibility of data, ecological analyses provide a unique opportunity to study associations across a wide variation of alcohol intake. These studies generally showed a strong inverse association between consumption of wine and mortality from heart disease, while the association was weaker or non-existent for spirits and beer.
The thoughtful analysis by Renaud et al illustrates the strengths and weaknesses of drawing definitive conclusions from ecological data.13 Whereas controlling for dairy fat strengthened the association between wine and coronary heart disease, other characteristics not accounted for might have been related to wine consumption and could have explained the reported correlations.32 33
Ecological studies (cross cultural and time trend) have other important limitations. In some countries a small proportion of the population may consume a large proportion of a specific type of alcoholic drink so that average per capita consumption may be an inaccurate representation. In other countries, like France, average per capita consumption of wine may be more representative as this is drunk by much of the population. Furthermore, what may seem to be a moderate or low level of consumption of spirits on the basis of per capita per day may mask excessive consumption by a small proportion of the population. In countries like the United States28 wine drinkers tend to be of higher socioeconomic status, have a healthier lifestyle, and have better access to health care; thus, the stronger inverse association between wine consumption and mortality from coronary heart disease could be explained in part by a lower case fatality rate.
St Leger et al reported a strong correlation between per capita wine consumption and mortality from coronary heart disease (-0.70 for men) but weaker correlations for cigarette smoking and mortality from coronary heart disease (0.28) and cerebrovascular mortality (0.08).5 However, later studies have shown strong associations between smoking and death from heart and cerebrovascular disease.34 35 36 This illustrates the importance of confirming results from ecological data with results from studies with more rigorous methods.
CASE-CONTROL AND COHORT STUDIES
The best evidence to judge whether an association exists between a type of alcoholic drink and coronary heart disease comes from observational data, in which individual intake can be linked directly to coronary heart disease. In addition, depending on the study design, both morbidity and mortality can be examined.
Although case-control studies of diet and chronic disease may produce biased results due to incomparability of controls37 or differential recall of past intake,38 recall of past alcohol consumption is not as problematic as is recall of many other exposures.39 Furthermore, most case-control studies of total alcohol consumption and coronary disease have found inverse associations similar to those from prospective studies.1
Almost all cohort studies find a strong inverse association between total alcohol intake and coronary heart disease,2 but no consistent pattern has emerged for specific types of drink. Even in the three largest studies, where the absolute consumption of all drink types would be greatest and the data would provide the most stable relative risks, all three types of drink conferred a reduction in coronary heart disease.25 26 28 From all the cohort studies—with data collected from more than 305000 men and women followed up for over 1.8 million person years—we conclude that if any type of drink does provide extra cardiovascular benefit apart from its alcohol content, the benefit is likely to be modest at best or possibly restricted to certain subpopulations.
REASONS FOR DIFFERENCES BETWEEN STUDIES
If no single type of drink provides all or most of the cardiovascular benefit, then why do results from several individual cohort studies suggest a stronger association for one particular type of drink? The differences between studies may be due to different drinking patterns or aspects of lifestyle correlated with choice of drink in particular populations. In the studies where only one type of drink was significantly associated with reduced risk of coronary heart disease, that drink was usually consumed by much of the population, typically at levels of one or two glasses a day. This pattern of widespread “healthy” drinking is more likely to take place with meals than is heavy or episodic drinking by a small percentage of the population.
In the health professionals follow up study, in which spirits were the most commonly consumed type of drink and were the most strongly protective, total alcohol consumption was strongly correlated with total number of days alcohol was consumed (r=0.89).26 This suggests that spirits were consumed most days of the week and were not restricted to heavy weekend consumption. Conversely, in the Copenhagen city heart study, in which consumption of spirits did not reduce coronary heart disease, only 8.5% of the men and 4% of the women reported drinking spirits on average once a day or more.31 Because this small sample of consumers of spirits may have had different drinking patterns (and other lifestyle characteristics) from the rest of the population in Copenhagen, it may explain the absence of a cardioprotective effect for spirits in this population.
Klatsky et al studied drinking patterns and characteristics correlated with choice of drink among 85000 men and women who specified a preference for one of the three main drink types.40 Compared with people who preferred spirits (after adjustment for age), wine consumers were less likely (relative risk=0.7) and beer drinkers were more likely (relative risk=1.2) to develop coronary heart disease. These differences were almost completely eliminated after further adjustment for sex, race, daily intake of alcohol, cigarette smoking, coffee drinking, and education. This is strong evidence that characteristics correlated with choice of drink may explain the different relative risks associated with types of drink in different populations. The behavioural characteristics that are correlated with choice of drink will vary widely among different populations and socioeconomic groups and will change over time.
An ideal study of the effects of types of alcoholic drink would have maximum variation in alcohol consumption from specific types of drink, with additional information on drinking patterns associated with specific drink types, but minimal variation in other risk factors for coronary heart disease such as socioeconomic group, smoking, diet, and obesity. A complete analysis could then include the simultaneous assessment of both drinking patterns and total alcohol intake. The Copenhagen city heart study had adequate variation in alcohol intake and drink type,31 but the participants were drawn from all inhabitants of a defined area of Copenhagen and their large variation in other factors like diet, occupation, and drinking patterns might explain why wine seemed to be most strongly associated with reduced risk of coronary heart disease.
Observational studies of alcohol consumption in populations where alcoholic drinks are limited to one or two types provide a unique opportunity to test the possibility that only one type of drink is largely responsible for the reduction in coronary heart disease. For example, in an angiographic study linking alcohol intake to reduced risk of coronary stenosis, Baboriak et al reported that few patients drank wine, suggesting that the association was due mainly to beer and spirits.41 Ducimetiere et al reported a similar association for alcohol in an angiographic study of a population where most of the alcohol consumed was from wine.42 In the Italian rural cohorts of the seven countries study the lowest mortality from cardiovascular disease occurred in the group of men who consumed an average 77.8 g of alcohol a day, almost exclusively wine.27 Conversely, the Honolulu heart study found a significant inverse association between coronary heart disease and beer consumption but no such association for wine.19 However, in this study only 15% of the population drank wine, with a median intake of half a glass a month. These studies suggest that alcohol itself rather than a particular type of drink is responsible for the reduction in coronary heart disease.
MECHANISM OF ALCOHOL'S PROTECTIVE EFFECT
Several short term experimental studies have shown that alcohol (not specific to drink type) increases the serum concentration of high density lipoprotein cholesterol.43 44 This mechanism provides a biological basis for a causal relation between alcohol consumption and lower rates of coronary heart disease.45 In epidemiological studies of total alcohol consumption and coronary heart disease in which measurements of high density lipoprotein cholesterol concentration are also available46 47 48 49 it was estimated that about half of the beneficial effect of alcohol was due to its effect on high density lipoprotein cholesterol. However, this calculation does not take into account laboratory and biological variability in high density lipoprotein concentrations or potential confounding by other lifestyle factors (such as diet, obesity, or physical activity). Thus, the true proportion of the effect attributable to high density lipoprotein is likely to be greater. Other potential mechanisms include an effect of alcohol on platelet function50 and on tissue plasminogen activator and other components of clotting and fibrinolysis.51 52 These mechanisms might partly explain an apparent acute protective effect of recent alcohol consumption (for example, previous evening).53
Each type of alcoholic drink has many non-alcohol components. Wine is the best studied54 and has been found to contain antioxidants,55 56 vasorelaxants,57 and stimulants to antiaggregatory mechanisms.52 For example, Maxwell et al found that 10 subjects who drank wine with meals had higher serum antioxidant activity than others who abstained from wine.56 However, little is known about any association between circulating serum antioxidant activity and risk of coronary heart disease, and comparisons were not made with other alcohol drinks.
Although most ecological studies support the hypothesis that wine consumption is most beneficial, the methodological problems of these studies limit their usefulness in drawing conclusions. Most of the differences in findings regarding specific drink types are probably due to differences in patterns of drinking specific types of alcoholic drink and to differing associations with other risk factors. Results from observational studies, where individual consumption can be assessed in detail and linked directly to coronary heart disease, provide strong evidence that a substantial proportion of the benefits of wine, beer, or spirits are attributable primarily to the alcohol content rather than to other components of each drink.
Funding This report was funded by the International Life Sciences Institute (ILSI Europe Alcohol Task Force).
Conflict of interest None.