Methods

Top Methods Results Discussion Author & Article Info References

Study Sample and Baseline Data

The Physicians' Health Study is a randomized, double-blind, placebo-controlled, 2 x 2 factorial trial of aspirin and ß-carotene for the primary prevention of cardiovascular disease and cancer [5, 6]. From 1981 to 1984, all 261 248 U.S. male physicians who were 40 to 84 years of age were invited to enroll. Approximately half of these physicians responded, of whom 59 285 were willing to participate. Physicians were excluded if they had a history of myocardial infarction, stroke or transient ischemic attack, or cancer (except nonmelanoma skin cancer); current renal or liver disease, peptic ulcer, or gout; or contraindications to, or current use of, aspirin or ß-carotene. The 33 223 willing and eligible physicians were enrolled in an 18-week "run-in" trial that resulted in a randomly assigned sample of 22 071 men. In our analysis, we excluded the 351 participants who had a baseline history of angina pectoris or coronary revascularization.

At baseline, the participants reported on many coronary risk factors. Participants also completed an abbreviated food frequency questionnaire that asked, "How often do you usually consume alcoholic beverages?" The seven response categories ("2+/d," "daily," "5-6/week," "2-4/week," "1/week," "1-3/month," "never/rare") were interpreted as the number of drinks consumed per unit of time. Men for whom data on alcohol consumption were missing (n = 95) were excluded; 21 530 men remained for analysis. In the initial analyses, we found almost identical relative risks when either nondrinkers (those who never or rarely drank) or occasional drinkers (those who consumed one to three drinks per month) were used as the reference group. Thus, we combined these two categories into a single reference group.

Angina Pectoris and Myocardial Infarction

New cases of angina pectoris or nonfatal myocardial infarction were reported on annual questionnaires; the small number of nonresponders were contacted by telephone. Men who reported incident angina were sent a supplementary questionnaire to obtain further details. Cases of angina were confirmed by the Endpoints Committee when the participant reported a positive result on a diagnostic test (that is, exercise tolerance testing with or without thallium, exercise radionuclide ventriculography, or angiography [ 50% coronary artery stenosis]) or if the participant later had myocardial infarction, a coronary revascularization procedure (coronary artery bypass surgery or angioplasty), or cardiac death. To further validate the angina end point, we reviewed medical records from 100 randomly selected cases and found that 88 of the 90 available records (98%) corroborated at least one positive result on diagnostic testing for angina. All cases of myocardial infarction met the World Health Organization criteria [7]. For fatal myocardial infarction, we also accepted diagnoses based on autopsies and deaths confirmed by other sources (for example, hospital records and death certificates) as attributable to coronary heart disease (International Classification of Diseases, 9th revision, codes 410 to 414). Follow-up through August 1994 was 99% for nonfatal outcomes and 100% for mortality.

Statistical Analysis

We used SAS software (SAS Institute, Cary, North Carolina) to compute means or proportions of baseline risk factors for the six levels of alcohol consumption. Proportional hazards models [8] provided relative risks for angina that were adjusted for potential confounders; a similar approach was used for myocardial infarction. We examined potential confounders, including age, smoking, exercise, diabetes, and parental history of myocardial infarction. The multivariate relative risks were not adjusted for factors possibly affected by alcohol consumption-blood pressure [9], body mass index [10], and hypercholesterolemia [3]-because these factors may not be confounders but rather effects of exposure and thus should not be controlled [11]. The presence of a linear trend in relative risk across alcohol categories was tested by using an ordinal alcohol variable, with values ranging from 1 (< one drink per week) to 6 ( two drinks per day). Interaction terms were used to formally test for an alcohol-aspirin interaction during the 5-year aspirin trial [5] and to test for an alcohol-ß-carotene interaction during the 11 years of available follow-up. All relative risks are presented with 95% CIs, and all reported P values are two sided.

Results

Top Methods Results Discussion Author & Article Info References

Table 1 shows unadjusted baseline risk factors according to level of alcohol intake. Participants drank relatively small amounts of alcohol; 97% of all men reported consuming fewer than two drinks per day. Compared with the reference group, daily drinkers ( one drink per day) tended to have a less favorable cardiovascular risk profile, including older age and increased prevalence of cigarette smoking. Men who drank the most alcohol ( two drinks per day) had the highest prevalence of hypertension.

By August 1994 (the end of the 11 years of follow-up), 1368 men had developed new-onset angina and 690 had had a first myocardial infarction (Table 2). In multivariate analyses, alcohol intake had strong, independent, inverse associations with risk for each event (P < 0.001). Compared with men who consumed less than one drink per week, men who consumed two or more drinks per day had a 56% lower risk for angina and a 47% lower risk for myocardial infarction. Even inclusion of potential intermediate variables (blood pressure, body mass index, and hypercholesterolemia) did not materially change the associations; for example, men who consumed two or more drinks per day had a 53% lower risk for angina and a 44% lower risk for myocardial infarction. Preliminary analyses for each coronary end point showed no evidence of a statistically significant interaction between alcohol intake and randomly assigned treatment with aspirin or ß-carotene (P > 0.05 for all comparisons).

Because 31% of the participants who developed incident angina also had myocardial infarction, the angina analysis was repeated after men who had both events were excluded (n = 424). Alcohol intake continued to have a strong, independent, inverse association with the risk for angina only (P for trend < 0.001), and the relative risks from this analysis were almost identical to those shown in Table 2. For example, compared with men who consumed less than one drink per week, men who consumed two or more drinks per day had a 55% lower risk for developing angina only.

Discussion

Top Methods Results Discussion Author & Article Info References

In this large, prospective cohort study of U.S. male physicians, alcohol consumption had strong, independent, inverse associations with the risk for angina pectoris and myocardial infarction. These associations were seen when either nondrinkers or occasional drinkers were used as the reference group, and they were not subject to material confounding by several important predictors of coronary heart disease. Our findings are consistent with those of earlier reports of an inverse association between moderate drinking and risk for myocardial infarction [3], and they extend the apparent cardiovascular benefits of moderate drinking to angina pectoris. Our results also highlight the complexity of counseling patients about alcohol consumption, because we found an increased risk for death from cancer among participants who consumed two or more drinks per day [12]. Nonetheless, the observed cardiovascular benefits help explain our previous finding that consumption of two to six drinks per week decreased total mortality in this cohort [12].

Although moderate alcohol consumption may have some antithrombotic properties [13], the similar benefit for angina and myocardial infarction suggests that antiatherogenic mechanisms largely account for the cardiovascular benefit of moderate drinking. Experimental studies have shown that alcohol consumption increases high-density lipoprotein (HDL) cholesterol levels [14]; this pharmacologic effect remains the presumptive major mechanism for the observed risk reductions. Statistical models suggest that HDL cholesterol accounts for approximately 50% of the reduction in the rate of myocardial infarction [15-17], but this estimate does not account for error in measuring levels of alcohol intake, HDL cholesterol levels, and other risk factors [18].

Potential limitations of our study include the method of alcohol assessment, inability to distinguish among drinking patterns, possible problems with the case definition of angina, and the nature of the cohort. As in most other alcohol studies, we relied on self-reported levels of alcohol consumption. Other epidemiologic approaches to alcohol assessment are impractical [19], however, and we have found that health professionals provide reliable reports of alcohol use [20]. Furthermore, we found a strong, positive association between alcohol intake and HDL cholesterol levels (P for trend < 0.001) in the 1535 participants for whom baseline lipoprotein values were available [21]; this finding supports the rank-order validity of the alcohol data. With regard to drinking pattern effects-that is, potentially stronger inverse associations among regular, as opposed to binge, drinkers-our inability to adjust for potential differences would tend to bias the relative risks toward the null and would not explain our findings. As for angina case definition, the use of physician participants provided a unique opportunity to study this somewhat vague diagnosis. We used two validation methods to confirm cases and found that the physicians provided accurate information. Finally, the healthy-volunteer effect [11], strict exclusion criteria, and enrollment of U.S. male physicians may affect generalizability. However, there is little biological basis for suspecting that the observed relations of moderate drinking to angina and myocardial infarction would be materially different in other groups that have a higher risk for cardiovascular disease. More important, the cohort's relative homogeneity and good baseline health offer strong evidence to refute persistent claims that the benefits of moderate drinking are found only in ill populations [4].

In summary, this large, prospective cohort study shows that apparently healthy men who consume moderate amounts of alcohol have decreased risk for angina pectoris and myocardial infarction. Although cardiovascular benefits were seen with alcohol consumption levels lower than the limits suggested by the 1995 U.S. Dietary Guidelines [22]-one or fewer drinks per day for most women and two or fewer drinks per day for most men-they must be weighed against the potential hazards of even moderate alcohol consumption [23]. Furthermore, the prospect of heavier drinking and its associated problems complicates any policy recommendation that lifelong abstainers begin drinking or that occasional drinkers increase their alcohol consumption [24]. Instead, we suggest that interested patients discuss their drinking habits with their primary care providers, who can help assess the health risks and benefits of any potential behavioral change and, in that setting, can provide an individual clinical recommendation.

Drs. Glynn, Gaziano, Manson, Buring, and Hennekens: Division of Preventive Medicine, Brigham and Women's Hospital, 900 Commonwealth Avenue East, Boston, MA 02215.