Substantial epidemiological evidence links moderate alcohol

Download (0)

Full text


Drinking Frequency, Mediating Biomarkers, and Risk of

Myocardial Infarction in Women and Men

Kenneth J. Mukamal, MD, MPH, MA; Majken K. Jensen, MSc; Morten Grønbæk, MD, PhD, DMSc;

Meir J. Stampfer, MD, DrPH; JoAnn E. Manson, MD, DrPH; Tobias Pischon, MD, MPH; Eric B. Rimm, ScD

Background—The associations of drinking frequency and quantity with risk of myocardial infarction have not been studied

among women, and the degree to which specific risk factors mediate the inverse association of drinking frequency with risk of myocardial infarction is uncertain.

Methods and Results—We conducted nested case-control studies of 32 826 women enrolled in the Nurses Health Study followed up from 1990 to 1998 and 18 225 men enrolled in the Health Professionals Follow-Up Study followed up from 1994 to 2000. A total of 249 women and 266 men with incident myocardial infarction were matched on age, smoking, and date of entry to 498 female and 532 male control participants. We determined the risk of myocardial infarction related to frequency and quantity of alcohol intake and the change in risk before and after adjustment for putative cardiovascular risk factors. Among both women and men, drinking frequency tended to be associated with lower risk of myocardial infarction, with the lowest risks among those who drank 3 to 7 days per week. Further adjustment for levels of high-density lipoprotein cholesterol, hemoglobin A1c, and fibrinogen attenuated 75% of the association of

frequent drinking with risk among women and fully attenuated the association among men.

Conclusions—Alcohol intake at least 3 to 4 days per week is associated with a lower risk of myocardial infarction among women and men, an association apparently attributable to the relationship of alcohol with HDL cholesterol, fibrinogen, and hemoglobin A1c. Because the effects of alcohol on HDL cholesterol, fibrinogen, and insulin sensitivity have been

confirmed in randomized trials, our findings support the hypothesis that the inverse relation of alcohol use and myocardial infarction is causal.(Circulation. 2005;112:1406-1413.)

Key Words:alcohol women men epidemiology myocardial infarction


ubstantial epidemiological evidence links moderate

alco-hol consumption with a lower risk of myocardial infarc-tion (MI). This evidence includes results from geographic

comparisons,1large cohort studies2–5and several

meta-anal-yses.6 –9However, the role of pattern of alcohol consumption

has not been well studied. We recently found that frequency of alcohol intake was the primary determinant of lower risk of

MI among men,10but we know of no comparable studies in

women, who differ from men both in their risk of MI and

their metabolism of ethanol.11

See p 1379

The lower risk of MI among moderate drinkers has been attributed in great part to their higher levels of HDL

choles-terol (HDL-C).12HDL-C is strongly associated with lower

risk of MI in epidemiological studies13and appears to have

beneficial vascular effects when augmented experimental-ly.14,15In both the Honolulu Heart Program and Multiple Risk Factors Intervention Trial, adjustment for HDL-C levels attenuated the estimated effect of alcohol by approximately 50%, which suggests that approximately half of the benefit attributable to moderate drinking was mediated by HDL-C.16,17We know of no study that has formally evaluated the role of other factors that could mediate the lower risk of MI among moderate drinkers.

The role of mediating factors in the inverse association of moderate drinking and risk of MI is particularly important because no long-term randomized trial has been performed to

Received January 21, 2005; revision received May 12, 2005; accepted May 23, 2005.

From the Division of General Medicine and Primary Care (K.J.M.), Beth Israel Deaconess Medical Center, Boston, Mass; Departments of Nutrition (M.K.J., M.J.S., T.P., E.B.R.) and Epidemiology (M.J.S., J.E.M., T.P., E.B.R.), Harvard School of Public Health, Boston, Mass; Centre for Alcohol Research (M.K.J., M.G.), National Institute of Public Health, Copenhagen, Denmark; Channing Laboratory (M.J.S., J.E.M., E.B.R.) and Division of Preventive Medicine (J.E.M.), Department of Medicine, Brigham and Women’s Hospital, Boston, Mass; and Department of Epidemiology (T.P.), German Institute of Human Nutrition, Potsdam-Rehbruecke, Germany.

The online-only Data Supplement, which contains an additional table, can be found at CIRCULATIONAHA.105.537704/DC1.

A short abstract of this work was presented at the 44th Annual American Heart Association Conference on Cardiovascular Disease Epidemiology and Prevention in San Francisco, Calif, on March 5, 2004, and published in abstract form (Circulation2004;109:e71– e144).

Reprint requests to Kenneth J. Mukamal, MD, MPH, MA, Division of General Medicine and Primary Care, Beth Israel Deaconess Medical Center, 330 Brookline Ave, RO-114, Boston, MA 02215. E-mail

© 2005 American Heart Association, Inc.

Circulationis available at DOI: 10.1161/CIRCULATIONAHA.105.537704



confirm this association, nor is one likely in the near future. However, multiple short-term randomized trials of alcohol administration in humans demonstrate effects of moderate alcohol consumption on several potential mediators. For example, a meta-analysis of more than 40 such trials con-firmed that consumption of 30 g of alcohol daily (about 2 standard drinks) raises levels of HDL-C by 0.10 mmol/L and

lowers levels of fibrinogen by 0.22 ␮mol/L.18 Moderate

alcohol use also improves insulin sensitivity19,20and possibly

markers of inflammation21 in experimental trials. Evidence

that biomarkers directly affected by alcohol mediate the relation of alcohol and risk of MI would support the hypoth-esis that this relation is causal.

To address these issues, we studied the relation of alcohol consumption to risk of MI among participants in 2 large, well-characterized cohort studies, women enrolled in the Nurses Health Study (NHS) and men enrolled in the Health

Professionals Follow-Up Study (HPFS).22These studies

in-clude prospectively collected information on drinking pat-terns, dietary and lifestyle factors, and serum levels of biomarkers. In these analyses, we assessed the associations of quantity and frequency of alcohol consumption with risk of MI among both women and men and the role of a wide array of potentially mediating biomarkers in these associations.


Study Population

The NHS cohort was established in 1976. The study population consists of 121 700 married female registered nurses aged 30 to 55 years residing in 1 of 11 larger US states. Women have received follow-up questionnaires biennially to update information on expo-sures and newly diagnosed illnesses. Since 1980, participants have updated information on diet, alcohol, and vitamin supplements through a food frequency questionnaire (FFQ) every 4 years.

The HPFS began in 1986, when 51 529 male health professionals 40 to 75 years of age completed the initial 6-page HPFS question-naire. The population includes 29 683 dentists, 3745 optometrists, 2218 osteopathic physicians, 4185 pharmacists, 1600 podiatrists, and 10 098 veterinarians. Biennial follow-up has mirrored the NHS.

Assessment of Biomarkers

Blood samples were requested from all active participants and collected from 32 826 NHS members in 1989 to 1990 and 18 225 HPFS members in 1993 to 1994. With the exception of a modestly lower prevalence of smoking, those who returned blood samples did not differ substantially from those who did not in both cohorts, including average alcohol intake of 6.5 versus 6.3 g/d among women and 12.8 versus 12.2 g/d among men. Participants underwent local phlebotomy and returned samples to our laboratory via overnight courier. On arrival, whole blood samples were centrifuged and stored in cryotubes as plasma, buffy coat, and red blood cells in the vapor phase of liquid nitrogen freezers.

The measurement of biomarkers in this sample has been described previously. The markers generally showed excellent stability and reproducibility during simulated transport and storage.23

Homocys-teine levels were only assayed among women, and adiponectin (a plasma protein associated with insulin sensitivity24,25) levels were

only assayed among men. Candidate biomarkers assessed in both women and men include LDL cholesterol, triglycerides, total cho-lesterol, apolipoprotein B, soluble tumor necrosis factor receptors 1 and 2 (sTNFR1 and sTNFR2), interleukin-6, hemoglobin A1c (HbA1c), soluble intercellular adhesion molecule-1, soluble vascular

cell adhesion molecule-1, high-sensitivity C-reactive protein,26

li-poprotein(a), and fibrinogen.

Assessment of Alcohol Consumption

We assessed diet with a 131-item semiquantitative FFQ administered every 4 years that included separate items for beer, white wine, red wine, and liquor, as described previously.10We previously validated

estimated alcohol consumption against two 1-week dietary records collected approximately 6 months apart from 173 NHS participants and 136 HPFS participants residing in eastern Massachusetts,27with

Spearman correlation coefficients between these 2 measures of 0.90 in women and 0.86 in men. Estimated mean alcohol intake among men was 12.8 g/d with dietary records and 12.5 g/d with the FFQ; estimated mean intake among women was 9.0 g/d with both methods. In 1986 and 1988, NHS and HPFS participants reported the number of days per week that they typically drank any form of alcohol. The correlation coefficient between this measure and diet records among men was 0.79.28

In these analyses, we used average alcohol intake from the 1990 FFQ among women and the 1994 FFQ among men (ie, at the time of the blood draw) and frequency of alcohol intake assessed in 1988, using previous assessments when missing data occurred.

Assessment of MI During Follow-Up

We used World Health Organization criteria for MI, which require typical symptoms and either diagnostic ECG changes or elevated cardiac enzymes (including cardiac-specific troponin).29 Fatal MI

included fatal cases defined by World Health Organization criteria with hospital records, or if coronary heart disease was listed as the cause of death on the death certificate, was the most plausible cause, and evidence of previous coronary heart disease was available. We included sudden cardiac deaths, defined as death within 1 hour of symptom onset with no previous serious illness and no other plausible cause. Physicians reviewing medical records were unaware of participants’ reported alcohol intake.

Case-Control Sampling

We performed parallel nested case-control studies within the samples of women and men who provided blood samples. In the NHS, we identified 249 women free of cardiovascular disease or cancer in 1990 who sustained an incident MI before June 30, 1998. In the HPFS, we identified 266 men free of cardiovascular disease in 1994 who sustained an incident MI before January 31, 2000. Of the 266 cases in HPFS, 196 had nonfatal MI and 70 had fatal CHD; the corresponding numbers in the NHS were 212 nonfatal cases and 37 fatal events.

For each incident case of MI, we randomly selected 2 women or men free of cardiovascular disease matched on age (in 5-year increments), smoking (in 5 categories), and month of blood return using risk-set sampling.30 In the NHS, matching criteria also

in-cluded fasting status and reported problems with blood drawing. In analyses of drinking frequency, we deleted 4 women (1 case and 3 controls) but no men who were missing this information.

Statistical Analysis

We calculated ORs from conditional logistic regression models as measures of relative risk, stratifying on the matching variables. In multivariable analyses, we additionally adjusted for body mass index (in quintiles), parental history of MI at or before age 60 years, diabetes, hypertension, hypercholesterolemia, physical activity (in 5 categories), aspirin use, supplemental use of vitamin E, energy intake (in quintiles), and energy-adjusted intake of saturated fat, trans fatty acids, omega-3 fatty acids, folate, and dietary fiber (each in quin-tiles). Nutritional variables, anthropomorphic measures, aspirin use, and self-reported cardiovascular risk factors have been validated.31

We also adjusted for current hormone replacement therapy use among women. We used covariate information from the time of the blood draw, defined as 1994 in the HPFS and 1990 in the NHS. We used information from previous questionnaires when covariate data from the time of blood draw were missing. We conducted tests of linear trend across increasing categories of alcohol consumption by treating the midpoints of consumption in categories as a continuous


variable; we squared this variable after centering to test quadratic trend.

We tested potential mediators of the association of alcohol use and risk of MI by comparing the regression coefficients associated with drinking frequency from multivariable models with coefficients from models additionally adjusted for each mediator in quintiles, consis-tent with previous work.16The regression coefficients are the natural

log of the ORs for each level of intake and are scaled linearly. Because these analyses require a single coefficient to describe the primary exposure, we incorporated drinking frequency as a contin-uous variable to reflect our findings in men and to include the full range of data; sensitivity analyses that incorporated drinking fre-quency in categories (that better specify the model form for women) are shown in the online-only Data Supplement Table. We also assessed the effects of adjustment for groups of mediators, including lipids [HDL-C, LDL cholesterol, total cholesterol, triglycerides, lipoprotein(a), and apolipoprotein B], inflammatory markers (interleukin-6, soluble intercellular adhesion molecule-1, soluble vascular cell adhesion molecule-1, sTNFR1, sTNFR2, and C-reactive protein), and markers implicated in randomized trials (HDL-C, fibrinogen, and HbA1cor adiponectin).

To maintain consistent sample sizes in the NHS, we assigned indicator variables when levels of a given mediator were missing (n⫽7 to 49). Analyses that instead deleted women with missing mediators or that assigned such women median population values yielded substantially similar results that are not shown here. We also assigned indicator variables for 90 women with missing triglyceride levels; analyses that used the population median gave similar results, and both approaches yielded results consistent with our previous findings on triglycerides and risk of MI.32Only 4 men in the HPFS

were missing values for any mediator, and hence these men were deleted only from analyses that included the missing mediator.

To determine whether the magnitude with which biomarkers appeared to mediate the relationship of alcohol intake with risk of MI was consistent with known effects of alcohol from randomized trials,18we determined the relation of alcohol intake with candidate

biomarkers across the range of drinking frequency in linear sion models, adjusting for the same covariates as in logistic regres-sion models. We used robust variance methods to accommodate deviations from normality of the conditional outcome variable.33


Characteristics of Cases and Controls

Table 1 shows the characteristics of case and control partic-ipants in the HPFS and NHS; the characteristics of these participants according to alcohol use have been reported

previously.5,10 As expected, traditional cardiac risk factors

were more common, HDL-C levels were lower, and LDL cholesterol levels were higher among cases than among controls. The age-adjusted correlations between drinking frequency and average quantity of use were 0.75 among women and 0.64 among men, with little difference between cases and controls. There was little variability in the amount of alcohol consumed per drinking day across the range of

frequency of use, with median intakes of 18.2 to 20.1 g (⬇1.5

drinks) per drinking day among control women and 22.4 to

26.2 g (⬇2 drinks) per drinking day among control men.

Drinking Pattern and Risk of MI

Figure 1 shows the relative associations of average alcohol consumption and drinking frequency with risk of incident MI in the NHS. Figure 2 shows similar results in the HPFS. For both men and women, drinking frequency was associated with risk of MI.

Among men, there was a significant graded inverse relation

of MI risk with drinking frequency (Ptrend 0.04), with the

lowest risk among men who drank 5 to 7 days per week. In

previous analyses of the entire HPFS cohort,10the risks were

similar among men who drank 3 to 4 and 5 to 7 days per week.

Among women, the risk was lowest among those who drank 3 to 4 days per week, with a slightly higher risk among those who drank more frequently. We did not find significant results in a test for quadratic trend in drinking frequency

(P⫽0.13) or in a comparison of intake 3 to 4 versus 5 to 7

days per week (P⫽0.38) among women; the corresponding

probability value from a test for linear trend was 0.02. The relationship between drinking frequency and MI risk was generally similar among participants who did and did not take aspirin (data not shown).

Mediators of the Relation of Drinking Frequency and Risk of MI

We added potentially mediating biomarkers one at a time (and then in groups) into multivariable regression models to assess the magnitude with which each biomarker or group of biomarkers changed the association of frequent drinking with risk of MI (Table 2 and Data Supplement Table).

Among women, multivariable analysis suggested an OR of 0.76 (95% CI 0.59 to 0.97), which corresponds to a regression

coefficient of ⫺0.28 per drinking day. HDL-C alone

ac-counted for 36% of this association, and HbA1c alone

ac-counted for a similar proportion. Other biomarkers only modestly attenuated the regression coefficient. When

HDL-C, HbA1c, and fibrinogen were included together, these

3 biomarkers explained 75% of the inverse association of frequent alcohol intake and risk of MI.

Among men, the OR associated with frequent intake of alcohol (in drinking days per week) was 0.86 (95% CI 0.74 to 1.00), which corresponds to a regression coefficient of

⫺0.15. Further adjustment for HDL-C alone dropped the

regression coefficient to⫺0.07, which suggests that

approx-imately half of the inverse association was attributable to HDL-C among men. Individual adjustment for adiponectin,

fibrinogen, triglycerides, or HbA1cin the multivariable

mod-els attenuated the association by⬇30%. Adjustment for the

combination of HDL-C, fibrinogen, and HbA1cfully

attenu-ated the regression coefficient in men.

Because the relationship of drinking frequency and risk was not linear among women, we performed sensitivity analyses that assessed mediators in a comparison of

con-sumption 3 or more days per week to concon-sumption⬍1 day

per week, reflecting the plateau in risk at 3 to 4 days per week (Data Supplement Table). In these analyses, adjustment for

the combination of HDL-C, fibrinogen, and HbA1c reduced

the association of frequent drinking with risk by 61% among women and eliminated it among men (adjusted coefficient 0.01).

Because the association of frequent drinking and risk of MI differed little between models that included only matching factors and larger multivariable models (Table 2 and Data Supplement Table), we also examined potentially mediating biomarkers in models with only matching factors. These models with fewer covariates again yielded very similar results. Adjustment for HDL-C, fibrinogen, and adiponectin


(as an alternate measure of insulin sensitivity also directly

affected by alcohol19,34) yielded results similar to HDL-C,

fibrinogen, and HbA1c, as did the replacement of fibrinogen

with C-reactive protein (which is also lowered by alcohol


Because frequency and overall quantity of drinking were closely correlated among women, we examined the relation-ship of quantity of intake with risk of MI. Compared with abstention, intake of at least 5.0 g/d was associated with a relative risk of 0.56 (95% CI 0.34 to 0.92). Adjustment for HDL-C alone reduced the association of this quantity of drinking with risk by 24% compared with 55% for the

combination of HDL-C, fibrinogen, and HbA1c.

Relation of Alcohol Use to Biomarkers Implicated in Randomized Trials

We next assessed the relation of drinking frequency with levels of the 3 biomarkers for which randomized trials provided the strongest evidence for an effect of alcohol:

HDL-C, fibrinogen, and HbA1c(Table 3). Across the

distri-bution of alcohol use, we found a range in HDL-C levels of roughly 0.2 to 0.3 mmol/L, consistent with the range expected

from clinical trials.18 The differences in HbA

1c levels were

also within a credible range, with levels among controls 0.1 to 0.2 percentage points lower among frequent drinkers than among abstainers. The distribution in fibrinogen levels (which appeared to mediate less of the association of alcohol TABLE 1. Baseline Characteristics of Cases of MI and Controls in the NHS and HPFS

Women Men Variable Cases (n⫽249) Controls (n⫽495) Cases (n⫽266) Controls (n⫽532) Alcohol, g/d 0.9 (3.5) 1.8 (8.6) 5.4 (14.4) 6.9 (17.3) Drinking frequency, d/wk 1.0 (1.0) 1.0 (3.0) 1.5 (4.0) 2.0 (5.0) Age, y 62 (9.3) 62 (9.9) 66 (12.6) 66 (12.4) Physical activity, METS/wk 11.0 (19.2) 11.5 (18.9) 22.8 (37.8) 26.6 (37.2) BMI 25.0–29.9 kg/m2, % 29.0 31.1 51.5 43.6 BMIⱖ30.0 kg/m2, % 26.2 14.1 11.7 11.1 Vitamin E supplement use, % 10.9 14.8 33.5 40.0 Aspirin use, % 46.0 50.1 39.1 35.0 Diabetes, % 19.8 6.7 9.4 4.5 Hypertension, % 57.7 29.1 42.1 30.8 Hypercholesterolemia, % 53.6 39.6 49.3 40.6 Postmenopausal hormone use, % 30.1 36.2 N/A N/A Cholesterol-lowering medication use, % 4.0 2.8 8.7 6.8 Biomarkers Cholesterol, mmol/L 6.07 (1.24) 5.78 (1.42) 5.57 (1.37) 5.26 (1.22) Triglycerides, mmol/L 1.45 (1.21) 1.25 (0.88) 1.72 (1.37) 1.34 (1.06) HDL-C, mmol/L 1.28 (0.50) 1.51 (0.56) 1.06 (0.36) 1.14 (0.39) LDL-C, mmol/L 3.74 (1.15) 3.42 (1.28) 3.53 (1.05) 3.22 (1.03) CRP, mg/L 3.1 (6.2) 2.2 (4.2) 1.7 (2.4) 1.1 (1.9) Fibrinogen,␮mol/L 10.52 (3.56) 9.97 (3.62) 11.94 (3.01) 11.37 (2.76) Creatinine,␮mol/L 55.4 (14.7) 55.4 (13.7) 71.1 (16.2) 70.9 (14.2) Lp(a),␮mol/L 0.461 (1.887) 0.398 (0.841) 0.681 (1.526) 0.551 (1.402) ApoB, g/L 1.05 (0.40) 0.99 (0.35) 1.03 (0.31) 0.93 (0.29) sTNFR1, pg/mL 1328 (491) 1222 (422) 1390 (546) 1390 (543) sTNFR2, pg/mL 2546 (963) 2365 (898) 2878 (1063) 2814 (1028) sICAM, ng/mL 283 (86) 261 (81) 342 (116) 323 (88) sVCAM, ng/mL 678 (220) 657 (192) 1329 (448) 1283 (380) IL-6, pg/mL 2.0 (1.7) 1.7 (1.5) 1.9 (2.0) 1.5 (1.9) Adiponectin, mg/L N/A N/A 14.3 (10.7) 16.5 (11.2) HbA1c, % 5.8 (0.8) 5.6 (0.5) 5.7 (0.1) 5.6 (0.4) Folate, ng/g Hb 1150 (475) 1178 (487) 1192 (346) 1182 (352) Homocysteine,␮mol/L 10.7 (4.1) 10.0 (4.1) N/A N/A METS indicates metabolic equivalents; BMI, body mass index; LDL-C, LDL cholesterol; CRP, C-reactive protein; Lp(a), lipoprotein(a); ApoB, apolipoprotein B; sICAM, soluble intercellular adhesion molecule-1; sVCAM, soluble vascular cell adhesion molecule-1; and IL-6, interleukin-6.

Median values with interquartile distance are shown for continuous covariates.


with risk) was somewhat broader, with differences across

drinking categories of 0.2 to 1.0␮mol/L.


In this nested case-control study, frequency of alcohol intake was associated with lower risk of MI among both women and

men. Among women, HDL-C, fibrinogen, and HbA1c (a

marker of glucose tolerance) accounted for 75% of the association. Among men, this association appeared to be virtually entirely attributable to these factors.

The results of the present study provide support for a causal relation between alcohol use and risk of MI, a relation that would be difficult to test in a long-term randomized trial. We found associations between alcohol use and candidate inter-mediates that were of credible magnitude and that appeared to account for most or all of the observed alcohol-MI relation in statistical models.

In both this and previous analyses, we found drinking frequency to be inversely associated with risk of MI, but not necessarily in a linear fashion. Among women in the present

study and men in a larger previous study,10there appeared to

be a clear threshold in risk at 3 to 4 drinking days per week, a level of consumption that is also consistent with a previous

meta-analysis of alcohol use and nonfatal MI.7Interestingly,

Dorn and colleagues recently found that greater drinking frequency had a strong inverse cross-sectional association with central adiposity among both women and men, whereas a higher quantity of intake per drinking day had a positive

relation.35Because central adiposity is linked to low levels of

HDL-C, high levels of inflammation-sensitive proteins

(in-cluding fibrinogen36 –38), and glucose intolerance (because all

are part of the metabolic syndrome), it is intriguing to hypothesize that alcohol use could act early in the pathway

that leads to these linked abnormalities.39 Indeed, average

alcohol intake has been inversely associated with prevalence

of metabolic syndrome in some studies.40,41However, other

direct mechanisms may contribute to the effect of alcohol on

HDL-C levels.42,43

We found inverse relations of alcohol use and risk of MI among both men and women, although the association was generally stronger among women. At the same time, mea-sured biomarkers accounted for a larger proportion of the association among men. These observations could be due to chance or to a greater degree of residual confounding among women. However, they may also suggest that alcohol has particular effects on cardiovascular risk markers that are stronger among women than men and for which we did not have data. For example, small, dense LDL particles may be particularly linked to risk of coronary heart disease among

women,44,45 and limited experimental and epidemiological

evidence suggests that alcohol use could increase levels of

larger LDL particles via effects on lipoprotein lipase.46,47

The present results provide indirect support for the hypoth-esis that ethanol itself, rather than nonalcoholic components of some alcoholic beverages, is the factor responsible for the relation of moderate drinking and risk of MI. Because this

Figure 1.Adjusted ORs for incident MI among women according to average daily alcohol intake (A) or frequency of alcohol intake (B). Probability values shown are from tests of linear trend. The ORs are conditioned on matching factors of age, smoking, and date of blood sam-ple return and adjusted for body mass index, parental history of MI, diabetes, hypertension, hypercholesterolemia, physical exertion, aspirin use, supple-mental use of vitamin E, energy intake, and energy-adjusted intake of saturated fat, trans fatty acids, omega-3 fatty acids, folate, and dietary fiber. I-bars indicate 95% CIs.

Figure 2.Adjusted ORs for incident MI among men according to average daily alcohol intake (A) or frequency of alcohol intake (B). Legend as in Figure 1.


relation appears to be mediated predominantly by biomarkers linked to ethanol intake itself, it is unlikely that other components influence the relation to an important degree. However, because wine is the most frequently consumed beverage among women but not men, it is possible that nonalcoholic constituents of wine may also contribute to the greater risk reduction found in women.

Specific limitations of the present study warrant discus-sion. As in any observational study, our results could be influenced by differences between participants in factors other than alcohol consumption for which we did not adjust. In this study, we adjusted for age, smoking, diet, exercise, body mass index, aspirin use, and cardiovascular risk factors, and our populations were homogeneous with respect to occupational class and sex. To have influenced our results on mediators, any remaining confounder would need to be associated with both alcohol consumption and risk of MI and generally be unrelated to the other covariates in our

multivari-able models, yet still associated with a measured biomarker. In such a scenario, inclusion of that measured biomarker in our models would effectively partially adjust for the uncon-trolled confounder and not solely test its mediating potential. The fact that the magnitude of the relations between the 3 key biomarkers and drinking frequency was comparable to the magnitude seen in randomized trials argues against (but cannot disprove) this possibility.

We studied 2 groups of predominantly white health pro-fessionals, and hence, we cannot necessarily generalize our results to other populations with different educational levels, incomes, or distributions of race and ethnicity. However, the relationship of alcohol intake with risk of MI reported here is

quite similar to that found in a wide variety of populations.7

Although we documented ⬎500 incident cases of MI

among⬎50 000 individuals in the present study, our power

to detect differences, particularly for individual categories of consumption, was limited in some cases. Likewise, drinking TABLE 2. ORs, 95% CIs, and Regression Coefficients for Risk of MI Associated With Drinking

Frequency (in Days per Week) Among Women and Men

Women Men

Model OR (95% CI) ␤(%)6 OR (95% CI) (%)¶ Basic* 0.70 (0.57–0.85) ⫺0.36 0.86 (0.76–0.97) ⫺0.15 Multivariable† 0.76 (0.59–0.97) 0.28 0.86 (0.74–1.00) 0.15 ⫹LDL-C 0.74 (0.57–0.95) ⫺0.30 (⫹7) 0.86 (0.74–1.00) ⫺0.15 (0) ⫹Total cholesterol 0.73 (0.57–0.94) 0.31 (11) 0.85 (0.73–0.99) 0.16 (7) ⫹HDL-C 0.83 (0.65–1.08) ⫺0.18 (⫺36) 0.94 (0.80–1.00) ⫺0.07 (⫺53) ⫹Triglycerides 0.78 (0.61–1.00) 0.25 (11) 0.90 (0.77–1.05) 0.11 (27) ⫹sTNFR1 0.77 (0.60–0.99) ⫺0.26 (⫺7) 0.86 (0.74–1.00) ⫺0.15 (0) ⫹sTNFR2 0.77 (0.60–0.99) 0.27 (4) 0.86 (0.74–1.01) 0.15 (0) ⫹IL-6 0.77 (0.60–0.99) ⫺0.26 (⫺7) 0.87 (0.75–1.01) ⫺0.14 (⫺7) ⫹HbA1c 0.84 (0.65–1.09) ⫺0.17 (⫺39) 0.90 (0.77–1.04) ⫺0.11 (⫺27) ⫹sICAM 0.76 (0.59–0.98) ⫺0.27 (⫺4) 0.88 (0.75–1.02) ⫺0.13 (⫺13) ⫹sVCAM 0.79 (0.61–1.01) 0.24 (14) 0.89 (0.76–1.03) 0.12 (20) ⫹CRP 0.76 (0.59–0.98) ⫺0.27 (⫺4) 0.89 (0.76–1.03) ⫺0.12 (⫺20) ⫹Lp(a) 0.77 (0.60–0.99) 0.26 (7) 0.86 (0.74–1.00) 0.15 (0) ⫹ApoB 0.73 (0.56–0.95) ⫺0.31 (⫹11) 0.88 (0.76–1.03) ⫺0.12 (⫺20) ⫹Fibrinogen 0.77 (0.60–0.98) 0.27 (4) 0.90 (0.77–1.05) 0.11 (27) ⫹Homocysteine 0.76 (0.59–0.97) ⫺0.28 (0) 䡠 䡠 䡠 䡠 䡠 䡠 ⫹Adiponectin 䡠 䡠 䡠 䡠 䡠 䡠 0.90 (0.77–1.05) 0.10 (33) ⫹Lipids‡ 0.76 (0.57–1.02) ⫺0.27 (⫺4) 0.96 (0.81–1.13) ⫺0.05 (⫺67) ⫹Inflammatory§ 0.82 (0.63–1.07) 0.20 (29) 0.90 (0.76–1.06) 0.11 (27) ⫹HDL-C, HbA1c, fibrinogen 0.94 (0.71–1.22) ⫺0.07 (⫺75) 1.01 (0.85–1.19) 0.01 (⫺107) All biomarkers 0.87 (0.65–1.16) 0.14 (50) 0.98 (0.82–1.16) 0.02 (87) Abbreviations as in Table 1.

*The basic model includes the matching factors of age, smoking, and date of blood sample return.

†The multivariable model further adjusts for body mass index, parental history of MI at or before age 60 years, diabetes, hypertension, hypercholesterolemia, physical activity, aspirin use, supplemental use of vitamin E, energy intake, and energy-adjusted intake of saturated fat, trans fatty acids, omega-3 fatty acids, folate, and dietary fiber.

‡Lipids include HDL-C, LDL cholesterol, cholesterol, triglycerides, lipoprotein(a), and apolipoprotein B. §Inflammatory markers include interleukin-6, sICAM, sVCAM, sTNFR1, sTNFR2, and CRP.

储When included simultaneously, all biomarkers were modeled as continuous variables.

¶The␤-coefficient (also called the regression coefficient) is the natural log of the OR. The percent change in the regression coefficient with adjustment for each additional factor or factors (compared with the multivariable model) is shown in parentheses.


frequency and average quantity of use were particularly strongly correlated among women, which reflects their lim-ited variability in intake per drinking day. Drinking frequency appeared to be more strongly associated with risk than was average quantity of intake among men, a finding more clearly

seen in the full HPFS cohort.10

We measured biomarkers at a single point in time, and we were not able to incorporate correction for measurement error

in the 2 cohorts.48This may have led us to underestimate the

true association of these markers with risk (and hence the degree to which they mediate effects of alcohol). We also relied on single determinations of alcohol consumption and could not update this variable during follow-up, and in the case of men, our assessment of drinking frequency occurred a few years before the onset of follow-up. Both of these facts may have led us to underestimate the association of frequent alcohol use with risk of MI based on a previous comparison

of baseline and updated measures of alcohol use.10

Although we relied on self-reported alcohol consumption in the present study, we have previously validated the measures of quantity and frequency of alcohol consumption in these

popu-lations.27,28As in most prospective cohort studies, the NHS and

HPFS studies contain few drinkers who consume alcohol heavily on a regular or even episodic basis. We could not, therefore, separately assess binge drinking, although other

stud-ies suggest it may be associated with an increased risk of MI,49

and our results can only be generalized to the frequent intake of moderate amounts of alcohol.

Several of the mediators that we studied correlate and interact with each other. For example, although fibrinogen is part of the thrombotic pathway, it also clusters with other

inflammation-sensitive proteins.36 Thus, adjustment for fibrinogen may, in

essence, partially adjust for more traditional inflammatory mark-ers, and adjustment for C-reactive protein instead of fibrinogen led to similar results. This does not alter the fact that moderate drinking influences several biomarkers directly, but it suggests that the causal pathway between alcohol use and risk of MI may be more complex than it appears at first glance. We also made the conservative decision to adjust for prevalent diabetes, but this may have led us to underestimate both the association of drinking frequency with risk of MI and the mediating effect of insulin sensitivity, because moderate drinking is associated with

a lower risk of incident diabetes in these cohorts.50,51

In summary, we found inverse relations of frequency of alcohol intake with risk of MI that were at least as strong as the relation of average alcohol use with MI risk. Levels of

HDL-C, HbA1c, and fibrinogen, which have been causally

linked to frequent alcohol use in randomized trials, appeared to mediate the bulk of the alcohol-MI relation. The present results provide support for the hypothesis that the observed relationship of alcohol intake and risk of MI is causal, although definitive confirmation of this will require long-term, randomized outcome trials. In the interim, clinical decisions about appropriate alcohol use should ultimately reflect individual susceptibility to the entire balance of risks and benefits attributable to alcohol use.


This study was supported by grants AA11181, HL35464, and CA55075 from the National Institutes of Health, Bethesda, Md.


Eric Rimm has received honoraria for occasional speaking engage-ments at research conferences from industry-related organizations (the Distilled Spirits Council and the National Beer Wholesalers Association).


1. LaPorte RE, Cresanta JL, Kuller LH. The relationship of alcohol con-sumption to atherosclerotic heart disease.Prev Med. 1980;9:22– 40. 2. Camargo CA Jr, Stampfer MJ, Glynn RJ, Grodstein F, Gaziano JM,

Manson JE, Buring JE, Hennekens CH. Moderate alcohol consumption and risk for angina pectoris or myocardial infarction in U.S. male phy-sicians.Ann Intern Med. 1997;126:372–375.

3. Gaziano JM, Gaziano TA, Glynn RJ, Sesso HD, Ajani UA, Stampfer MJ, Manson JE, Hennekens CH, Buring JE. Light-to-moderate alcohol con-sumption and mortality in the Physicians’ Health Study enrollment cohort.J Am Coll Cardiol. 2000;35:96 –105.

4. Jackson R, Scragg R, Beaglehole R. Alcohol consumption and risk of coronary heart disease.BMJ. 1991;303:211–216.

5. Stampfer MJ, Colditz GA, Willett WC, Speizer FE, Hennekens CH. A prospective study of moderate alcohol consumption and the risk of coronary disease and stroke in women.N Engl J Med. 1988;319:267–273. 6. Cleophas TJ. Wine, beer and spirits and the risk of myocardial infarction:

a systematic review.Biomed Pharmacother. 1999;53:417– 423. 7. Maclure M. Demonstration of deductive meta-analysis: ethanol intake

and risk of myocardial infarction.Epidemiol Rev. 1993;15:328 –351. 8. Corrao G, Rubbiati L, Bagnardi V, Zambon A, Poikolainen K. Alcohol

and coronary heart disease: a meta-analysis. Addiction. 2000;95: 1505–1523.

9. Di Castelnuovo A, Rotondo S, Iacoviello L, Donati MB, De Gaetano G. Meta-analysis of wine and beer consumption in relation to vascular risk.

Circulation. 2002;105:2836 –2844. TABLE 3. Adjusted Levels of Potential Mediators Among Men

and Women According to Drinking Frequency

Drinking Days per Week

⬍1 1–2 3–7 Women Cases HDL-C, mmol/L 1.30 1.37 1.42 Fibrinogen,mol/L 10.61 10.91 10.41 HbA1c, % 6.2 6.4 6.3 Controls HDL-C, mmol/L 1.45 1.55 1.74 Fibrinogen,mol/L 10.64 10.32 9.61 HbA1c, % 5.8 5.7 5.6 Men Cases HDL-C, mmol/L 1.01 1.04 1.22 Fibrinogen,␮mol/L 12.52 12.32 11.61 HbA1c, % 6.2 6.0 5.8 Controls HDL-C, mmol/L 1.06 1.11 1.30 Fibrinogen,␮mol/L 12.11 11.97 11.23 HbA1c, % 5.8 5.7 5.7 Results adjusted for age, smoking, body mass index, parental history of MI at or before age 60 years, diabetes, hypertension, hypercholesterolemia, average daily exertion, aspirin use, supplemental use of vitamin E, energy intake, and energy-adjusted intake of saturated fat, trans fatty acids, omega-3 fatty acids, folate, and dietary fiber.


10. Mukamal KJ, Conigrave KM, Mittleman MA, Camargo CA Jr, Stampfer MJ, Willett WC, Rimm EB. Roles of drinking pattern and type of alcohol consumed in coronary heart disease in men.N Engl J Med. 2003;348: 109 –118.

11. Baraona E, Abittan CS, Dohmen K, Moretti M, Pozzato G, Chayes ZW, Schaefer C, Lieber CS. Gender differences in pharmacokinetics of alcohol.Alcohol Clin Exp Res. 2001;25:502–507.

12. Linn S, Carroll M, Johnson C, Fulwood R, Kalsbeek W, Briefel R. High-density lipoprotein cholesterol and alcohol consumption in US white and black adults: data from NHANES II. Am J Public Health. 1993;83:811– 816.

13. Shah PK, Kaul S, Nilsson J, Cercek B. Exploiting the vascular protective effects of high-density lipoprotein and its apolipoproteins: an idea whose time for testing is coming, part I.Circulation. 2001;104:2376 –2383. 14. Spieker LE, Sudano I, Hurlimann D, Lerch PG, Lang MG, Binggeli C,

Corti R, Ruschitzka F, Luscher TF, Noll G. High-density lipoprotein restores endothelial function in hypercholesterolemic men.Circulation. 2002;105:1399 –1402.

15. Bisoendial RJ, Hovingh GK, Levels JH, Lerch PG, Andresen I, Hayden MR, Kastelein JJ, Stroes ES. Restoration of endothelial function by increasing density lipoprotein in subjects with isolated low high-density lipoprotein.Circulation. 2003;107:2944 –2948.

16. Langer RD, Criqui MH, Reed DM. Lipoproteins and blood pressure as biological pathways for effect of moderate alcohol consumption on coronary heart disease.Circulation. 1992;85:910 –915.

17. Suh I, Shaten BJ, Cutler JA, Kuller LH; the Multiple Risk Factor Inter-vention Trial Research Group. Alcohol use and mortality from coronary heart disease: the role of high-density lipoprotein cholesterol.Ann Intern Med. 1992;116:881– 887.

18. Rimm EB, Williams P, Fosher K, Criqui M, Stampfer MJ. Moderate alcohol intake and lower risk of coronary heart disease: meta-analysis of effects on lipids and haemostatic factors.BMJ. 1999;319:1523–1528. 19. Sierksma A, Patel H, Ouchi N, Kihara S, Funahashi T, Heine RJ, Grobbee

DE, Kluft C, Hendriks HF. Effect of moderate alcohol consumption on adiponectin, tumor necrosis factor-alpha, and insulin sensitivity.Diabetes Care. 2004;27:184 –189.

20. Davies MJ, Baer DJ, Judd JT, Brown ED, Campbell WS, Taylor PR. Effects of moderate alcohol intake on fasting insulin and glucose con-centrations and insulin sensitivity in postmenopausal women: a ran-domized controlled trial.JAMA. 2002;287:2559 –2562.

21. Sierksma A, van der Gaag MS, Kluft C, Hendriks HF. Moderate alcohol consumption reduces plasma C-reactive protein and fibrinogen levels: a randomized, diet-controlled intervention study.Eur J Clin Nutr. 2002;56: 1130 –1136.

22. Rimm EB, Giovannucci EL, Willett WC, Colditz GA, Ascherio A, Rosner B, Stampfer MJ. Prospective study of alcohol consumption and risk of coronary disease in men.Lancet. 1991;338:464 – 468.

23. Pai JK, Curhan GC, Cannuscio CC, Rifai N, Ridker PM, Rimm EB. Stability of novel plasma markers associated with cardiovascular disease: processing within 36 hours of specimen collection.Clin Chem. 2002;48: 1781–1784.

24. Pischon T, Hotamisligil GS, Rimm EB. Adiponectin: stability in plasma over 36 hours and within-person variation over 1 year. Clin Chem. 2003;49:650 – 652.

25. Chandran M, Phillips SA, Ciaraldi T, Henry RR. Adiponectin: more than just another fat cell hormone?Diabetes Care. 2003;26:2442–2450. 26. Pai JK, Pischon T, Ma J, Manson JE, Hankinson SE, Joshipura K, Curhan

GC, Rifai N, Cannuscio CC, Stampfer MJ, Rimm EB. Inflammatory markers and the risk of coronary heart disease in men and women.N Engl J Med. 2004;351:2599 –2610.

27. Giovannucci E, Colditz G, Stampfer MJ, Rimm EB, Litin L, Sampson L, Willett WC. The assessment of alcohol consumption by a simple self-administered questionnaire.Am J Epidemiol. 1991;133:810 – 817. 28. Feskanich D, Rimm EB, Giovannucci EL, Colditz GA, Stampfer MJ,

Litin LB, Willett WC. Reproducibility and validity of food intake mea-surements from a semiquantitative food frequency questionnaire.J Am Diet Assoc. 1993;93:790 –796.

29. Alpert JS, Thygesen K, Antman E, Bassand JP. Myocardial infarction redefined: a consensus document of The Joint European Society of Cardiology/American College of Cardiology Committee for the redefi-nition of myocardial infarction.J Am Coll Cardiol. 2000;36:959 –969. 30. Prentice RL, Breslow NE. Retrospective studies and failure time models.

Biometrika. 1978;65:153–158.

31. Mukamal KJ, Ascherio A, Mittleman MA, Conigrave KM, Camargo CA Jr, Kawachi I, Stampfer MJ, Willett WC, Rimm EB. Alcohol and risk for ischemic stroke in men: the role of drinking patterns and usual beverage.

Ann Intern Med. 2005;142:11–19.

32. Shai I, Rimm EB, Hankinson S, Curhan G, Manson JE, Rifai N, Stampfer MJ, Ma J. A multivariate assessment of lipid parameters as predictors of CHD among women: potential implications for clinical guidelines. Cir-culation. 2004;109:e97– e98.

33. White HA. A heteroskedasticity-consistent covariance matrix estimator and a direct test for heteroskedasticity.Econometrica. 1980;48:817– 838. 34. Pischon T, Girman CJ, Rifai N, Hotamisligil GS, Rimm EB. Association between dietary factors and plasma adiponectin concentrations in men.

Am J Clin Nutr. 2005;81:780 –786.

35. Dorn JM, Hovey K, Muti P, Freudenheim JL, Russell M, Nochajski TH, Trevisan M. Alcohol drinking patterns differentially affect central adi-posity as measured by abdominal height in women and men.J Nutr. 2003;133:2655–2662.

36. Sakkinen PA, Wahl P, Cushman M, Lewis MR, Tracy RP. Clustering of procoagulation, inflammation, and fibrinolysis variables with metabolic factors in insulin resistance syndrome. Am J Epidemiol. 2000;152: 897–907.

37. Bonora E, Kiechl S, Willeit J, Oberhollenzer F, Egger G, Bonadonna RC, Muggeo M. Metabolic syndrome: epidemiology and more extensive phe-notypic description: cross-sectional data from the Bruneck Study.Int J Obes Relat Metab Disord. 2003;27:1283–1289.

38. Ford ES. The metabolic syndrome and C-reactive protein, fibrinogen, and leukocyte count: findings from the Third National Health and Nutrition Examination Survey.Atherosclerosis. 2003;168:351–358.

39. Goude D, Fagerberg B, Hulthe J. Alcohol consumption, the metabolic syndrome and insulin resistance in 58-year-old clinically healthy men (AIR study).Clin Sci (Lond). 2002;102:345–352.

40. Park YW, Zhu S, Palaniappan L, Heshka S, Carnethon MR, Heymsfield SB. The metabolic syndrome: prevalence and associated risk factor findings in the US population from the Third National Health and Nutrition Examination Survey, 1988 –1994.Arch Intern Med. 2003;163: 427– 436.

41. Solymoss BC, Bourassa MG, Lesperance J, Levesque S, Marcil M, Varga S, Campeau L. Incidence and clinical characteristics of the metabolic syndrome in patients with coronary artery disease. Coron Artery Dis. 2003;14:207–212.

42. Lin RC, Miller BA, Mei MH, Ochs S, Lumeng L. Effects of alcohol feeding on synthesis and secretion of apolipoproteins by regenerating rat sciatic nerve.Alcohol Clin Exp Res. 1992;16:899 –903.

43. Lakshman MR, Chirtel SJ, Chambers LC, Campbell BS. Hepatic syn-thesis of apoproteins of very low density and high density lipoproteins in perfused rat liver: influence of chronic heavy and moderate doses of ethanol.Alcohol Clin Exp Res. 1989;13:554 –559.

44. Kuller L, Arnold A, Tracy R, Otvos J, Burke G, Psaty B, Siscovick D, Freedman DS, Kronmal R. Nuclear magnetic resonance spectroscopy of lipoproteins and risk of coronary heart disease in the cardiovascular health study.Arterioscler Thromb Vasc Biol. 2002;22:1175–1180. 45. Stampfer MJ, Krauss RM, Ma J, Blanche PJ, Holl LG, Sacks FM,

Hennekens CH. A prospective study of triglyceride level, low-density lipoprotein particle diameter, and risk of myocardial infarction.JAMA. 1996;276:882– 888.

46. Taskinen MR, Nikkila EA, Valimaki M, Sane T, Kuusi T, Kesaniemi A, Ylikahri R. Alcohol-induced changes in serum lipoproteins and in their metabolism.Am Heart J. 1987;113:458 – 464.

47. Williams PT, Krauss RM. Associations of age, adiposity, menopause, and alcohol intake with low-density lipoprotein subclasses. Arterioscler Thromb Vasc Biol. 1997;17:1082–1090.

48. Spiegelman D, McDermott A, Rosner B. Regression calibration method for correcting measurement-error bias in nutritional epidemiology.

Am J Clin Nutr. 1997;65:1179S–1186S.

49. Trevisan M, Dorn J, Falkner K, Russell M, Ram M, Muti P, Freudenheim JL, Nochajaski T, Hovey K. Drinking pattern and risk of non-fatal myocardial infarction: a population-based case-control study.Addiction. 2004;99:313–322.

50. Conigrave KM, Hu BF, Camargo CA Jr, Stampfer MJ, Willett WC, Rimm EB. A prospective study of drinking patterns in relation to risk of type 2 diabetes among men.Diabetes. 2001;50:2390 –2395.

51. Stampfer MJ, Colditz GA, Willett WC, Manson JE, Arky RA, Hennekens CH, Speizer FE. A prospective study of moderate alcohol drinking and risk of diabetes in women.Am J Epidemiol. 1988;128:549 –558.




Related subjects :