Panel Discussion I: Does Alcohol Consumption Prevent Cardiovascular Disease?

“Errors” in Observational Studies and the Association Between Alcohol and Coronary Heart Disease 

Wannamethee began the discussion by stating that from the evidence currently available, there may well be protection against coronary heart disease (CHD) from moderate drinking, and many mechanisms for such an action have been shown. However, she believed that the magnitude of the protection remains a question, and there is still a problem with epidemiological studies that use “nondrinkers” as controls. Even if the problem with “sick quitters” being included in this group is being addressed in newer studies, other factors, many of which we may not yet understand, may represent confounding when it comes to nondrinkers.

Mittleman stated that epidemiologists address known biases, but cannot completely eliminate all sources of bias, especially from lifestyle factors. However, the central question one should consider is: To what extent does bias affect the results of our studies? He asserted that there is good evidence that potential biases and residual confounding probably do not overcome the protective effects observed in most epidemiological studies. He further commented that future studies should also evaluate the relation between alcohol drinking pattern and the risk of CHD. While carrying out randomized trials for cardiovascular end points is daunting, he suggested that intermediate-term trials may be feasible and results from such trials would help us understand the biological mechanisms.

To respond to Fillmore's criticisms of methods in observational studies, and “errors” in the assessment of the health effects of light-to-moderate alcohol consumption, Mittleman said that most of the harmful effects from moderate drinking that she demonstrated were related to cancer and that, in fact, prospective studies also show such adverse effects. Fillmore insisted that better measures of the duration of drinking and exposure over longer periods of observation are still needed to help resolve the issue. Ellison countered by pointing out that in analyses among the generally light drinkers in the Framingham Heart Study (where there are repeated assessments of alcohol intake), similar results are usually found regardless of whether one uses alcohol data assessed at baseline, average data over time, or the most recent updated data prior to the occurrence of CHD as the exposure variable. Wannamethee stated that heavier drinkers are more likely than others to stop drinking completely at some point in their drinking career; as people age, many of them become light drinkers. She added that taking change into account may have little effect for light drinkers, but does among heavy drinkers.

Arthur Klatsky commented: “We all agree that ‘sick quitters’ are inappropriate as controls; but I am not concerned about including occasional drinkers in the referent group. Our own studies find that about 20% of the drinkers consume alcohol infrequently, perhaps once a month, and we have consistently found that these infrequent drinkers have essentially the same risk as lifetime abstainers—not at higher or lower risk.” He then asked Fillmore if she could cite any study that finds infrequent drinkers to be at higher risk, or if there was any scientific evidence suggesting that including occasional drinkers in the reference group with abstainers is really a methodologic “error,” as she had suggested.

Fillmore replied that “occasional drinkers” may include mainly individuals who have always been occasional drinkers, but also some who have decreased their consumption because of illness or infirmity. In her view, there is sufficient evidence from numerous longitudinal studies showing that the decline in consumption with increasing age and illness is a process. The meta-analytic efforts conducted by her group were a crude effort to take this process into account. Rimm said that his own studies had found that using light drinkers (rather than abstainers) as the referent group still yielded a dose-response curve of increasing protection against CHD with increasing alcohol consumption. To him, this suggested that there is a biological basis for the protective effects. Wannamethee added that, similar to the results of Klatsky, the occasional drinkers included in her own studies were not sick people. However, she stated that they differ from “abstainers” who, in some studies, may include former heavy drinkers.

Judging the Evidence that Alcohol Consumption Decreases the Risk of CHD 

Klim McPherson stated that he had come to the conference with the assumption that moderate alcohol consumption does decrease the risk of CHD, by about 10% to 15%, and what he had heard confirmed this belief. But he pointed to lessons learned from the presumed protective effects against CHD of hormone replacement therapy (HRT), which had not been supported by a large clinical trial, the Women's Health Initiative (WHI). He said that we should bear these and similar results in mind as we interpret findings on alcohol and CHD. Rimm responded that the WHI trial should remind us that a single clinical trial on alcohol and CHD will not answer all the questions. He added that the WHI was done among women at an average age of 65 years, and in reality very few women start taking HRT 15 years after menopause. When women in WHI were stratified by age, there was actually less harm and even some benefit among the youngest women. “This points out the danger of having only one clinical trial of alcohol and CHD and using its results as the ‘final answer.’ Any results from such a trial may well apply only to a subset of the population—those with a certain lifestyle, certain diet, certain genetics, certain age group, etc.—so we should not expect that a single trial will be applicable to everyone.”

Mukamal added that for HRT, or vitamin E, or almost any other lifestyle factor, overall health may be more favorable among the exposed group, making comparisons much harder. However, this is not the case for alcohol consumption. Smoking, for example, is much more common among drinkers than abstainers, yet the beneficial effects of alcohol on CHD still remain even though we realize that confounding from smoking may be incompletely controlled for in many studies.

Thomas Stuttaford added that proof that alcohol reduces the risk of CHD depends on what “standard of evidence” one uses. There are two different standards of acceptance: if it is a criminal case, the standard is “beyond reasonable doubt”; in a commercial case, it is “the balance of probability.” What matters to clinicians is very much the balance of probability. On the other hand, epidemiologists are like the criminal lawyers and want the evidence to be beyond reasonable doubt—often an impossibility. Lionel Tiger pointed out that there is almost a “disconnect” between research evidence and advice given to patients, as Stuttaford had suggested. Richard Smallwood added that those setting government policy are motivated not by the balance of probabilities or beyond reasonable doubt, but by “beyond reasonable criticism.”

Smallwood repeated the key question: “For alcohol and heart disease, are we now ‘beyond reasonable doubt’ that there is a protective effect?” Rimm responded that the answer to that question depends on whom one asks. Among cardiovascular epidemiologists, most believe it is indeed beyond a reasonable doubt based on the evidence from both epidemiological and basic science findings. He continued: “We call ourselves experts and at some point we have to conclude that there is enough evidence to say that alcohol reduces the risk of CHD. Even with 15 clinical trials over 15 or more years, we will never have enough evidence to convince some people. Besides, we all come with our own biases: people working in addiction clinics will have different biases from those working with heart disease prevention. But recall,” he added, “that stating that the science is strong enough to show such protection does not necessarily relate directly to policy decisions, as we will hear later.”

Age to Begin Drinking 

Ellison asked if there was any evidence which suggested advantages from starting alcohol intake before one reached middle age. Leighton responded that the short-term effects are only one part of the story, and there may be long-term effects of alcohol that we do not yet understand. He pointed out that atherosclerosis begins at younger ages and there is a need to control for that as early as possible. Measures such as endothelial function, although cumbersome to undertake, would greatly enhance our understanding of the effects of moderate drinking and the age at which they are seen.

Differences Among Populations 

Rehm brought up three points to be considered. First, the majority of people in the world do not drink alcohol. In “drinking countries,” there may not be enough differences between people for good comparisons. Hence more studies should look at the developing world where drinking is not the norm. Second, the decline in drinking with age is a very tricky feature. In North America, the heaviest drinking is among the young; in most European countries, however, heavy drinking continues even into the 50s and 60s, and there may be relatively less heavy drinking in early adulthood. Thus it is important to study specific cohorts in different countries. Third, the relative risk for CHD from alcohol intake goes down as people age. Therefore a 90-year-old person would not have the same relative protection against death as a younger person. Mittleman disagreed with the last statement, stating that as the risk for mortality approaches 100%, the estimated effects on the relative risk of any single factor become smaller. However, alcohol protects against CHD at all ages and, in fact, Rehm's own studies show that the beneficial effect of alcohol on the risk of CHD is not reduced among older people.

Wannamethee noted that most of the studies showing a protective effect of alcohol against CHD come from individuals who are of Caucasian origin, and such a protective effect may not apply to other groups, such as African Americans. Others stated that there are limited data available on minorities, and much more research is needed. Dwight Heath pointed out a similar problem in many parts of the developing world, in that very sophisticated and rigorous studies exist only from a few countries. Alcohol data from much of the rest of the world are often based on social surveys or hospital records, making it difficult to evaluate the relation of alcohol to disease among such populations.

Ricardo Brown said that this discussion reminded him of diabetes research. “In diabetes, we have come up with hemoglobin A1C as a biomarker of severity. In alcohol research, there is great variability in what is defined as ‘moderate drinking.’ There is also variability in what a standard drink is. None of us seems to know of a stable biomarker that would be more reliable as an indicator of moderate drinking than data from questionnaires.”

Genetic Effects 

Henk Hendriks mentioned that genetic studies have shown that polymorphisms of the alcohol dehydrogenase gene, which regulates alcohol metabolism, modify the association between alcohol and CHD, which adds to the evidence that alcohol reduces the risk of CHD. Rimm stated that although genetic studies were very important, it would be simplistic to say that one gene determines the effect of alcohol, as the body tends to compensate, and many other genetic factors will undoubtedly be found. Booyse cautioned that it is important to keep in mind that thousands of genes are activated by alcohol or polyphenols, working together to produce a final effect. The significance of individual functional changes still needs to be understood. At present, it would be an error to pick and choose among genes—it is a matter of accumulating data.

Translating Research on Alcohol and Health 

Nady el-Guebaly offered that there is a considerable difference between advice given by a doctor to a patient and that given to a population. Any discussion of moderate drinking, he observed, may be more productive at the individual level. Ellison asked if, as a basis for giving advice on drinking, there was any need for further observational epidemiological studies on alcohol and CHD. Richard Harding responded that one of the purposes of epidemiological studies is to inform mechanistic studies, such as those described by Booyse and Leighton. These findings, in turn, should be used to inform the design of future epidemiological studies, so both types of study are needed. Advice to the public should be based on the totality of evidence.

For determining what advice to give regarding drinking, Rimm pointed out difficulties in observational epidemiological studies. “Among the ‘healthiest people’ in our cohorts (who are lean, do not smoke, who exercise, and who eat a healthy diet.), the risk of CHD is very low to begin with. Among such individuals, it might be difficult to show an effect of moderate drinking. Drinking should not be viewed as a substitute for addressing risk factors such as obesity or lack of exercise.” He concluded that if a doctor does not mention to a patient to stop smoking or to lose weight but gives only one recommendation, to drink moderately, the physician would be making a mistake. Ellison challenged his conclusion: “If a person cannot stop smoking or cannot lose weight, he or she might well benefit from having a drink a day.” Rimm replied that that was true, but that other practices, such as not smoking and not being obese, impact many health conditions. “For example, smoking has dramatically strong effects on mortality, much greater than drinking.” He added that there may be the potential problem of the patient focusing only on the drinking advice and not worrying about other lifestyle factors. But he concluded that his own data have shown that even individuals who meet all other criteria for a “healthy lifestyle” have better health outcomes if they drink moderately.

Mittleman added that even though all of the genetic factors involved in the protective effect of alcohol on CHD may not be fully understood, current evidence supports an association, and it cannot all be explained by bias. “This relationship underscores the importance of reconciling basic and epidemiological research and the need for translational research that helps inform the clinician who treats patients as to the proper advice regarding alcohol consumption.”