The impact of moderate use.

Both public policy and individual decisions concerning alcohol use are made, in part, on the basis of its health effects. While there is little disagreement about the adverse consequences of heavy alcohol use, the impact of light to moderate alcohol use is a matter of great controversy.

This paper will discuss three aspects of the controversy surrounding moderate alcohol use: First, we will review the major health-related findings associated with moderate alcohol use; Next, we will discuss potential reasons for the variability in studies that deal with alcohol and health; Third, we will propose a framework for resolving the controversy.

The bulk of the studies concerning moderate alcohol consumption and health outcomes revolve around cardiovascular mortality and morbidity, breast cancer, and the effects of maternal alcohol use on fetal health. There are a smaller number of studies relevant to the relationship between alcohol use and both cognitive function and gall bladder disease.

In none of these areas is there unanimity of findings. The majority of studies support a protective effect of moderate alcohol use on cardiovascular morbidity and mortality. There is also evidence to support a dose-response relationship between alcohol consumption and the risk of breast cancer. The majority of studies that examine the relationship between moderate maternal alcohol use and fetal outcome suggest either no significant impact or a slight decrease in birthweight. In other areas, there are little data to suggest either a significant positive or negative health impact of moderate alcohol use.

The variability in the above study findings may be accounted for by a number of factors: ascertaining actual alcohol exposure is difficult and there is no standard definition of "light" and "moderate." Next, both retrospective and case-control studies are prone to a number of design biases. Moreover, alcohol use is associated with a host of other behaviors that may directly affect health.

Since decisions concerning alcohol use will be made long before the "definitive" study is done, how does one resolve the controversy and minimize uncertainty? Meta-analysis - a data analytic tool that pools the results of multiple studies - allows a means of exploring the robustness of a finding. This will be demonstrated using the relationship between alcohol and breast cancer. Decision analysis -a quantitative tool for comparing multiple decisions under circumstances of uncertain outcome - allows one to weight the consequences of making choices about alcohol consumption. This paper will illustrate the principals of decision analysis by evaluating an individual's decisions surrounding use versus non-use of alcohol when a variety of health effects are simultaneously considered. The two tool may offer insight into the issue of conflicting opinions.

ALCOHOL LONGEVITY AND

LIFESTYLES: DATA, DISAGREEMENTS,

AND DECISIONS

The search for the secret of longevity

- the Holy Grail of the physician - is a never-ending exercise in the unravelling of complex relationships. The wise researcher recognizes that the human organism is not simply a collection of biochemical processes. Given that recognition, our understanding of those individual factors that contribute to, or detract from, a prolongation of graceful aging has nonetheless grown considerably in recent years. In part, that knowledge has come from observing the effect of social change, rather than medical advances, on human longevity. In part, that knowledge has come from carefully performed epidemiologic studies that have then led to a laboratory experiments that further illuminate those edpidemiologic investigations. While we can still only speculate on what the potential human life span might actually be, we have already demonstrated the beneficial effects of reducing the prevalence of common infectious diseases and have a clearer understanding of the need to reduce cardiovascular disease and cancer. We have only just begun to understand the role of psychological factors - such as stress and social relationships - in mediating disease and modifying life expectancy.

This presentation will focus particularly on the specific relationship between alcohol and longevity. For the most part, this means evaluating epidemiologic studies. It will discuss the controversies that have arisen over such studies and propose several means for resolving those controversies.

The discussion will center on the relationship between moderate alcohol consumption and longevity - the area that has attracted the most controversy. Even the definition of moderate alcohol consumption is controversial - neither the precise amount nor the pattern of consumption (regular vs intermittent) had been standardized in studies relating alcohol to health. What is most clear is that moderate drinking is somewhere between no drinking and heavy drinking. Whether there should be a fourth category of consumption - "light" - is also problematic. Given that the bulk of studies focus their attention on average consumption per unit of time (e.g. daily, weekly, or monthly) our definition of moderate will include average daily consumption of up to 25 gm of alcohol (or the equivalent of two drinks of wine, beer or spirits). Average consumption calculations must be relied upon, since there has been little research relating different patterns of drinking to health outcomes.

We used a computerized literature search as the starting point for this review - focusing on English language studies that evaluated the relationship between alcohol consumption and health outcome. Articles identified via the initial search were used to identify other relevant articles. Articles were included for review if they met the following criteria: (1) adequate description of study population (in terms of methods of selection, sample size, age and sex); (2) adequate description of means of ascertaining alcohol consumption; (3) use of independently-determined, valid measures (e. g. death, hospitalization or surgery) of outcome.

Papers were identified in three areas: alcohol and mortality (particularly cardiovascular mortality), alcohol and cancer (primarily breast cancer), and alcohol and other kinds of morbidity hospitalization, gall bladder disease and cognitive function).

I. Alcohol and Mortality

A total of 22 papers evaluating the relationship between alcohol and mortality were identified. Eleven of these were prospective studies, eight were case-control studies and three were cross-sectional studies. Nearly all demonstrated increased mortality with heavy alcohol use; sixteen demonstrated the lowest mortality levels in moderate drinkers, five showed no effect of moderate consumption on mortality (although one of these showed diminished cardiovascular mortality but unaffected total mortality in moderate drinkers) and one focused its attention primarily on heavy drinkers. In general the majority of studies showed that moderate drinkers have a relative risk of cardiovascular mortality - compared to non-drinkers - of 0.5 to 0.7 (a 30% to 50% reduction in mortality).

The major criticisms of the studies that suggest that alcohol has a protective effect on mortality include the following: (1) studies that show worse outcomes in abstainers may do so because they have inadequately controlled for important differences between abstainers and drinkers, such as prior ill health, income, personality type or other factors that might actually produce health differences between the two groups. (2) a tendency to under-report alcohol consumption may blur the distinctions between non-drinkers, moderate drinkers and heavy drinkers.

Arguments in support of the protective effect of moderate alcohol consumption include: (1) the results are robust - they have been replicated in a wide variety of studies involving many different populations and a number of different analytic techniques; (2) there is a plausible mechanism for this relationship - in that alcohol consumption elevates high-density lipoprotein levels; (3) at least some studies have controls for prior ill health and other cardiac risk factors and have still demonstrated a protective effect of moderate alcohol consumption.

Much of the controversy in this area concerns the actions that might be taken as a result of those studies. Should physicians recommend moderate alcohol for cardiovascular protection in the same way that they commonly prescribe exercise, low-fat and high-fibre diets and abstention from smoking? Many authors have voiced the concern that such a recommendation would not have the desired effect because patients might substitute alcohol for other more healthful activities and that some might eventually consume large, more dangerous amounts of alcohol. Neither of these hypotheses have been tested, however.

II. Alcohol and Cancer

For the most part, most studies that have suggested an increased risk of certain cancers (liver, oropharyngeal, breast, esophageal) have done so in heavy drinkers. Almost none have shown that moderate drinkers have an elevated cancer risk. The one exception may be women who consume moderate amounts of alcohol; some studies suggest an increased risk of breast cancer.

To address the issue, 23 studies were identified; there were seven prospective studies and 16 case control studies. Twelve studies indicated a positive relationship between alcohol consumption in the risk of breast cancer, nine indicated no increased risk and two provided uninterpretable results. Those studies that showed a positive relationship between alcohol consumption and breast cancer indicated a relative risk of approximately 1. 5 (50 % increase in risk) for moderate drinkers compared to non-drinkers. In general, prospective studies were more likely to show an increased risk of cancer than were case-control studies, and wine conferred the lowest risk of breast cancer.

The major criticisms of the studies that suggest a positive relationship between alcohol and breast cancer is that the demonstrated relationship is inconsistent and relatively weak. Moreover, under-reporting of alcohol consumption would bias the findings toward a worse outcome in moderate drinkers. In addition, we know far less about factors that contribute to breast cancer than we do about those that contribute to heart disease. As such, many studies of breast cancer may be less than optimally controlled for variations in diet, coffee consumption or smoking. Finally, there has been no identifiable mechanism by which alcohol might cause breast cancer.

Arguments in favor of the breast cancer hypothesis include the observation that the most carefully designed studies - the longitudinal, prospective ones - showed the strongest, most consistent relationship between alcohol and breast cancer. These studies also have attempted to control for other known risk factors for breast cancer.

As with cardiovascular disease, significant controversy surrounds suggestions for actions. Should women abstain from alcohol as a means of protecting themselves from breast cancer? As we will discuss later, such a decision almost certainly does not rest solely on the consideration of a single disease.

III. Alcohol and Morbidity

Althought the major thrust of this review is on alcohol and longevity, a brief discussion of other health outcomes is included because they may play a role in decisions about alcohol consumption.

A review of the health-related literature shows that at least three areas have been directly studied with respect to moderate alcohol consumption: hospitalization rates, gall bladder disease and cognitive function. While any discussion of alcohol also raises issues of accidents and suicide, these are most relevant to heavy alcohol use. Moreover, those studies that have evaluated moderate alcohol consumption and mortality do not suggest that moderate drinkers are at increased risk of accidental or self-induced death. Moreover, at least one large study has indicated that moderate drinkers have lower hospitalization rates than nondrinkers - even when the data are adjusted for age, race, income and smoking.

Similarly, several prospective studies have demonstrated a negative relationship between alcohol consumption and the risk of gall bladder disease.

The most problematic area of study is the relationship between cognitive function and moderate alcohol use. A very small number of investigators have suggested that social drinkers have defects in cognitive function at times when they are not drinking. However, the methods of testing have not been well validated, study results have not been easily replicable and the magnitude of the negative effect of alcohol - even in the most negative studies - is small. Moreover, most epidemiologic studies indicate that moderate drinkers are, on the average, of higher intelligence and socioeconomic status than either non-drinkers or heavy drinkers. Finally, even studies that have evaluated the relationship between acute alcohol use and brain function provide conflicting results, depending on which kind of performance is tested. Some show diminished function while others show enhanced function. Overall, studies of cognition, memory and moderate alcohol use do not suggest a definite relationship - either positive or negative.

IV. Toward Diminution of

the Controversy

At every step in the analysis of alcohol's relationship to the length and quality of life, there is controversy. Some of this stems from the nature of the studies and the data they have produced, and some stems from the implications of those studies. An ideal solution - one that will never occur - would be a randomized, controlled, long-term study of the effects of moderate alcohol consumption on a variety of measurable outcomes. Short of that, there are two potential, complementary solutions: better study design and use of newer data analytic techniques. The latter approach has two components. The first involves the analysis of data from combinations of existing studies called meta-analysis. The second component (called decision analysis) addresses the problem of making decisions (in this case, the decisions concern choices surrounding alcohol use) when there are multiple, uncertain outcomes.

V. Improved Study Design

The most useful epidemiologic studies are prospective and longitudinal. Case-control studies are problematic in that the selection of control patients - since it occurs after the fact - often introduces biases that may be difficult to identify or quantify. Criteria for patient selection must be clearly understood; the most generalizable studies are those that have the fewest restrictions for entry and are community based. A careful assessment of actual alcohol exposure is important, as are definable outcome measures, adequate sample size and adequate study duration. The best studies are designed to test specific hypotheses, rather than generate them; studies that engage in "data-dredging" are likely to commit type I errors (finding differences when they don't actually exist), while studies that excessively stratify their subjects are likely to commit type Il errors (missing differences when they actually do exist). Hypotheses that are based on plausible biologic mechanisms and that might be further tested in the laboratory are more likely to be validated.

If we apply all of these criteria to existing studies, virtually none can be classified as perfect. Does that mean they are all useless? Not at all. While none provides definitive answers about cause and effect, they show trends - especially when viewed in the aggregate.

Meta-analysis

When a single study is inadequate or when multiple studies provide conflicting or puzzling results, a clearer picture may emerge when the data from several studies are pooled and viewed as a whole. This is the essence of the emerging field of meta-analysis. This approach does not simply involve the lumping of data, but requires a careful understanding of the methodology used for performing each individual study. Generally, studies selected for a meta-analysis must first pass some basic selection criteria - usually criteria indicative of a well-conceived, valid and generalizable study. Different studies may receive different weights according to study quality" and sample size. Data may be subdivided into smaller categories and then pooled - which can add statistical power to the overall anaylsys. The advantages of meta-anaylsys are that it may identify reasons for disagreements among studies and it can more clearly resolve differences between them. It does not guarantee a definitive or valid answer, however.

When applied to the questions of breast cancer and cardiovascular mortality in connection with moderate alcohol use, meta-analysis confirms the trends suggested by the majority of published studies. There appears to be a statistically significant relationship between alcohol consumption and the risk of breast cancer. It becomes most apparent at or above a consumption level of 24 gm/day and is not significant below 12 gm/day.

On the other hand, meta-analysis supports the U or J-shaped curve relationship between alcohol consumption and carciiovascular disease. People who consume between 10 and 30 gm of alcohol per day have the lowest risk - this risk reduction is less for women and most apparent for male smokers, but the relationship holds true in all cases. Overall, such an analysis affirms the overall improved mortality outlook for moderate drinkers -irrespective of sex or age.

Decision Analysis

How did the above data translate into action? How does a woman decide about alcohol consumption when faced with the competing outcomes of diminished cardiovascular risk and increased breast cancer risk? What if there is a history of alcohol abuse in the family? How does one factor in issues of enjoyment or pleasure? Decision analysis may play a role. Although not a new technique, decision analysis has entered the health field only within the last three decades and has only become relatively prominent within the last 10 years. Simply put, decision analysis is a data analytic tool that allows one to more clearly evaluate complex decision issues. It is quantitative, explicit, and requires input about the probabilities of uncertain events and the values of the outcomes that may follow. It does not necessarily guarantee the right decision, but a carefully-done decision analysis will often perform as well as a skilled decision maker. Most importantly, it helps to clarify which components of a decision are the most important. Implicit in the last statement is the assumption that even the most complex decision problems are often driven by a smaller subset of issues. Furthermore, decision analysis often allows one to highlight the importance of psychological or perceptual factors which may play a role in a decision. For instance, some individuals may make decisions designed to minimize risk or avoid regret, should the decision not work out. Others may choose the option that will maximize certain benefits - even at a somewhat greater risk. A recognition of different decision-making styles or perceptual biases may illuminate reasons behind controversy. A key issue in using decision analysis concerns identities of the decision maker. An individual decision concerning choosing to drink or not to drink will be very different from a policy decision that deals with decisions relevant to advertising or availability of alcohol. While time does not allow a complete and formal decision analysis of any of the multiple decisions involved in the alcohol controversy, some observations concerning individual decisions about alcohol are in order. First, it should be clear that there is no single correct decision about alcohol consumption. Individual factors such as sex, age, heredity and other cardiovascular risk factors will all have a modulating effect on the risks and benefits of alcohol use. Second, it appears highly unlikely that individuals make decisions about alcohol consumption purely on a health basis - to do so would imply an excessive reliance on uncertain data. In all probability, the most important factor driving the decision to drink is likely to be personal preference and taste. Health issues become secondary. Hence, a physician would be loathe to recommend a daily glass of wine to someone who doesn't like the taste. Conversely, an individual who enjoys wine might take heart that not only is such a pleasure not dangerous, but it may actually be beneficial.

Conclusion

It is unlikely that the perfect" study on alcohol and health will ever come along, especially since it would still be open to varying interpretations, depending on the biases of the interpreters. Even so, we still have a great deal to learn about the contribution of risk factors such as diet, heredity and personality to cancer. Moreover, research on beverage-specific differences in cancer and heart disease risks are still in their infancy. In the meantime, there may be ways via a clearer understanding of the psychology of decision making and the techniques of decision analysis and meta-analysis to resolve some of the apparent conflicts over the risks and benefits of moderate alcohol use.