Association Between Consumption of Beer, Wine, and Liquor and Plasma Concentration of High-Sensitivity C-Reactive Protein in Women Aged 39 to 89 Years

Association Between Consumption of Beer, Wine, and Liquor and Plasma Concentration of High-Sensitivity C-Reactive Protein in Women Aged 39 to 89 Years Emily B. Levitan, MS a,e, Paul M. Ridker, MD, MPH a,c, JoAnn E. Manson, MD, DrPH a,b,e, Meir J. Stampfer, MD, DrPH a,b,e,f, Julie E. Buring, ScD a,d,e, Nancy R. Cook, ScD a, and Simin Liu, MD, ScD a,e, * Although cross-sectional studies have shown an inverse or U-shaped relation between alcohol consumption and plasma concentration of high-sensitivity C-reactive protein (hs-crp), the associations between specific types of alcoholic beverages beer, wine, and liquor and hs-crp concentrations are less clear. Plasma concentrations of hs-crp were measured in 11,815 participants in the Women s Health Study who had never used postmenopausal hormones. Alcohol intake was measured using a semiquantitative food-frequency questionnaire. Alcohol consumption had an inverse association with geometric mean hs-crp concentrations (nondrinkers 1.43 mg/l, 0.1 to 6 g alcohol/day 1.37 mg/l, 6.1 to 12 g alcohol/day 1.29 mg/l, >12 g alcohol/day 1.28 mg/l, p for trend 0.003). In age-adjusted analyses, beverage preference was a significant predictor of geometric mean hs-crp concentration. However, after adjustment for body mass index (BMI), beer drinkers who consumed 6.1 to 12 g alcohol/day had a geometric mean hs-crp concentration of 1.03 mg/l, wine drinkers 1.09 mg/l, liquor drinkers 1.28 mg/l, and combination drinkers 1.09 mg/l (p 0.43). The association between alcohol and hs-crp concentration appears to be mediated primarily by ethanol and was independent of the type of alcoholic beverage consumed once BMI was taken into account. 2005 Elsevier Inc. All rights reserved. (Am J Cardiol 2005;96:83 88) Most studies examining the relation between alcohol and high-sensitivity C-reactive protein (hs-crp) have focused on a single type of alcoholic beverage in dietary trials or the total intake of alcohol in cross-sectional studies. Recently, several studies have reported associations with CRP that appear to be similar for beer, wine, and liquor. 1,2 It remains uncertain whether the relation between alcohol and hs-crp, if any, may differ by type of alcoholic beverage and pattern of consumption, particularly for the lower alcohol consumption typical of moderate drinkers in the United States. To further understand these issues, we examined the crosssectional associations of the 3 main types of alcoholic beverages beer, wine, and liquor with plasma concentrations of hs-crp in a large population of middle-aged and a Division of Preventive Medicine, b Channing Laboratory, and c Division of Cardiology, Brigham and Women s Hospital and Harvard Medical School; d Department of Ambulatory Care and Prevention, Harvard Medical School; and Departments of e Epidemiology and f Nutrition, Harvard School of Public Health, Boston, Massachusetts. Manuscript received December 28, 2004; revised manuscript received and accepted March 3, 2005. The Women s Health Study was supported by research Grants HL43851 and CA47988 from the National Institutes of Health, Bethesda, Maryland. * Corresponding author: Tel: 617-732-8108; fax 617-731-3843. E-mail address: sliu@rics.bwh.harvard.edu (S. Liu). older women who had never used postmenopausal hormones. Methods The Women s Health Study is a randomized, double-blind, placebo-controlled trial of small-dose aspirin and Vitamin E for the prevention of cardiovascular disease and cancer. The trial participants are 39,876 female health professionals aged 39 to 89 years who have no history of heart disease, cancer (other than nonmelanoma skin cancer), or stroke at the beginning of follow-up. The institutional review board of Brigham and Women s Hospital approved the Women s Health Study, and all participants provided written informed consent. Information on usual diet, including alcohol intake, was provided at baseline by 39,310 (99%) of the randomized participants, who completed a 131-item, validated, semiquantitative food-frequency questionnaire at baseline. 3,4 For each food, a commonly used unit or portion size was specified on the questionnaire, and the participants were asked how often on average during the previous year they had consumed that amount. The portion sizes for beverages containing alcohol were 1 glass, bottle, can for beer and light beer, 4 oz. glass for red wine and white wine, and 1 drink or shot for liquor. Nine responses were possible, ranging from never 0002-9149/05/$ see front matter 2005 Elsevier Inc. All rights reserved. www.ajconline.org doi:10.1016/j.amjcard.2005.03.031

84 The American Journal of Cardiology (www.ajconline.org) or less than once per month to 6 or more times per day. Women who did not respond to any of the alcohol questions were excluded. Of women who answered 1 of the alcohol questions, nonresponse to an alcohol question was recoded as never or less than once per month in accordance with previous validation studies. 5 Nutrient scores for total energy, total fat, saturated fat, carbohydrate, protein, cholesterol, folate, dietary fiber, and dietary glycemic load (dietary glycemic index multiplied by grams of carbohydrate per serving) were computed by multiplying the frequency of consumption of each unit of food from the semiquantitative food-frequency questionnaire by the nutrient content of the specified portion size according to food-composition tables from the US Department of Agriculture 6 and other sources. Nutritional variables were adjusted for total energy using the residual method. 5 A detailed description of the semiquantitative food-frequency questionnaire and procedures used to calculate nutrient intake as well as data on reproducibility and validity of the semiquantitative food-frequency questionnaire in a similar cohort were previously reported. 5 In a similar cohort of women, Spearman s correlation coefficient between total alcohol consumption as measured by four 1-week diet records and a semiquantitative food-frequency questionnaire was 0.90. 7 The correlation coefficient for beer was 0.81, for wine 0.83, and for liquor 0.80. 8 Blood samples from 28,345 participants (71% of the population) were collected at baseline into tubes containing ethylenediaminetetraacetic acid and stored in liquid nitrogen until analysis. hs-crp was assayed with a validated, high-sensitivity assay (Denka Seiken Co., Ltd., Tokyo, Japan) with a coefficient of variation of 7.8%. 9 For this analysis, we included participants with available information on diet, hs-crp, age, body mass index (BMI), parental history of myocardial infarction at 60 years, history of diabetes mellitus, history of hypertension, history of high cholesterol, smoking status, and physical activity levels. Women who used postmenopausal hormones were excluded from the analysis because of the effects of postmenopausal hormones on hs-crp. 10 Thirty percent of the total population (11,815 women) met all inclusion criteria. The women included in this analysis were younger and had a lower prevalence of hypertension and high cholesterol than those who were excluded. We first calculated means and SDs or proportions of covariates for this population of women by level of alcohol consumption (nondrinkers, 0.1 to 6, 6.1 to 12, and 12 g alcohol/day). These alcohol cutoffs approximately correspond to up to 0.5, 0.5 to 1, and 1 drinks/day. Because hs-crp was not normally distributed, medians and interquartile ranges were calculated. Geometric means were computed by linearly regressing the natural logarithm of hs-crp concentrations on total alcohol intake and then taking the antilog of the resulting mean logarithmic hs-crp concentration. We used 3 models to examine the relation. The first adjusted for age (continuous), the second adjusted for age and BMI (continuous), and the third adjusted for age, BMI, parental history of myocardial infarction at 60 years (yes or no), history of diabetes mellitus, history of hypertension ( 140/90 mm Hg), history of high cholesterol ( 240 mg/dl), smoking status (current, past, or never), physical activity levels (rarely or never, 1,1to3,or 4 times/week), and dietary variables (continuous), including glycemic load and intakes of total fat, saturated fat, carbohydrate, dietary fiber, folate, cholesterol, and total energy. We performed linear and quadratic trend tests by calculating the median alcohol intake for each level of alcohol consumption and including the median and median squared as continuous variables in the models. Subgroup analyses were performed by cardiovascular risk predictors: cigarette smoking (ever or never); hypertension; diabetes mellitus; high cholesterol; and BMI 25 or 25 kg/m 2, further adjusted for BMI as a continuous variable. We assessed interaction between these characteristics and category of total alcohol consumption by creating cross-product terms and including them in the models. We used 4 degree-of-freedom F tests to determine whether interactions were statistically significant. Among women who consumed alcohol, we calculated means or proportion of covariates by beverage preference (beer, wine, or liquor). The association between alcoholic beverage preference and hs-crp was examined by limiting the analysis to women who reported some alcohol consumption and cross-classifying beverage choice (beer, wine, liquor, or a combination) and the amount of alcohol consumed. Three models similar to those described for the total alcohol analysis were used to examine the relations. Analysis of variance was used to test for the significance of the type of alcohol consumed as a predictor of the natural logarithm of hs-crp concentration within level of alcohol use. A p value of 0.05 was considered significant. We performed a sensitivity analysis excluding women who consumed 30 g alcohol/day because earlier studies have suggested that high alcohol consumption is associated with increased hs-crp concentrations. Statistical analyses were conducted with SAS software version 8.2 (SAS Institute Inc., Cary, North Carolina). Results The characteristics of the population are described in Table 1. The overall pattern for the relation between alcohol intake and hs-crp concentrations was inverse (Table 2). The inverse association was present only in women with BMIs 25; in leaner women, hs-crp concentrations were substantially less and did not vary materially by alcohol intake. Also, the inverse relation was stronger in current and past smokers and in women without a history of hypertension or diabetes mellitus. Although the quadratic trends were statistically significant in the age-adjusted models, in multivariate-adjusted models, only the current and past smokers showed a U-shaped relation between smoking and alcohol consump-

Preventive Cardiology/C-Reactive Protein, Beer, Wine, and Liquor 85 Table 1 Baseline characteristics of 11,815 participants in the Women s Health Study Variable* Alcohol Consumed (g/d) Nondrinkers 0.1 6 6.1 12 12 (n 4,661) (n 4,429) (n 1,427) (n 1,298) Age (yrs) 52.8 7.4 51.8 6.8 52.0 6.9 52.8 7.3 BMI (kg/m 2 ) 27.2 5.9 26.0 4.9 24.8 4.3 24.6 4.1 History of hypertension 26.0% 20.4% 19.8% 20.9% History of high cholesterol 27.1% 22.9% 20.9% 23.7% History of diabetes mellitus 4.4% 1.7% 1.1% 0.9% Parental history of MI at 60 yrs 14.4% 15.0% 15.1% 14.9% Smoker Current 11.5% 10.5% 10.7% 16.6% Past 24.8% 38.4% 46.0% 47.8% Never 63.7% 51.1% 43.2% 35.7% Exercise Rarely or never 43.1% 34.5% 33.1% 31.7% 1 times/wk 20.5% 21.0% 19.1% 20.3% 1 3 times/wk 27.4% 34.0% 35.0% 33.0% 4 times/wk 9.0% 10.5% 12.9% 15.1% Nutrients Total energy intake (kcal/d) 1,725 554 1,730 521 1,780 522 1,879 519 Total fat intake (g/d) 58.8 12.4 58.1 11.3 57.1 11.2 55.1 11.2 Saturated fat intake (g/d) 20.2 5.1 20.0 4.7 19.5 4.7 18.8 4.7 Carbohydrate intake (g/d) 227 35 223 31 214 31 198 34 Protein intake (g/d) 81.1 15.0 82.1 13.4 80.8 12.7 77.1 13.0 Dietary cholesterol intake (mg/d) 226 76 226 66 224 65 220 67 Dietary folate intake ( g/d) 408 221 418 214 422 209 413 207 Dietary glycemic load 122 21 118 19 112 18 103 20 Dietary fiber intake (g/d) 18.9 6.2 18.7 5.3 18.5 5.3 17.0 5.1 * Mean SD or percentage. All nutrients energy adjusted except total energy intake. Dietary glycemic index multiplied by grams of carbohydrate per serving. MI myocardial infarction. tion (p 0.01). In multivariate-adjusted formal tests of significance, all of the interactions with alcohol intake modeled as a categorical variable were significant at the 0.05 level except the interaction between smoking and alcohol. Wine drinkers had a smaller mean BMI than beer or liquor drinkers (25.5 kg/m 2 for wine drinkers compared with 26.1 kg/m 2 for beer drinkers and 27.1 kg/m 2 for liquor drinkers) and were less likely to be sedentary (36% of wine drinkers rarely or never exercised compared with 41% of beer drinkers and 45% of liquor drinkers) or to be current smokers (9% of wine drinkers compared with 18% of beer drinkers and 27% of liquor drinkers). Wine drinkers also had greater mean intakes of folate and fiber and smaller mean intakes of total fat, saturated fat, and cholesterol than beer and liquor drinkers. The age-adjusted geometric mean hs-crp estimates were significantly different between drinkers of beer, wine, liquor, and multiple types of alcohol at levels of 0.1 to 6 and 6.1 to 12 g ethanol/day (Table 3). However, after adjustment for BMI, the alcohol preference groups no longer had significant differences in geometric mean hs-crp concentrations within strata of alcohol intake. Within each category of alcoholic beverage preference, alcohol intake and hs-crp also appeared to be inverse, although in most cases, the relation did not reach statistical significance. The results did not change markedly when women who consumed 30 g alcohol/day were excluded. Discussion In this large population of 11,815 women, we found an inverse relation between alcohol consumption and plasma concentrations of hs-crp, but this inverse relation was limited to overweight women. We found no difference in plasma concentrations of hs-crp across types of alcoholic beverage consumed after taking into account BMI and other lifestyle factors. Although alcohol use did not explain as much variation in hs-crp as BMI or physical activity, these results suggest that dietary intake may influence systemic inflammation. In a smaller cross-sectional study previously conducted in the United States, participants who did not drink or drank 1 time per month had a median hs-crp concentration of 2.60 mg/l, whereas participants who consumed 2 drinks/ day had a median hs-crp concentration of 1.80 mg/l (p 0.001). 11 An inverse relation was also demonstrated in a nationally representative sample of the United States, which evaluated the odds of elevated CRP ( 75th percen-

86 The American Journal of Cardiology (www.ajconline.org) Table 2 Plasma concentration of high-sensitivity C-reactive protein (hs-crp) (mg/l) according to cardiovascular risk predictors Variable Alcohol Consumed (g/d) p trend p interaction None 0.1 6 6.1 12 12 Age-adjusted geometric mean (95% CI) Full cohort 1.63 (1.57 1.68) 1.33 (1.28 1.37) 1.09 (1.03 1.16) 1.09 (1.02 1.17) 0.0001 Normal weight 0.79 (0.76 0.83) 0.79 (0.75 0.83) 0.73 (0.68 0.79) 0.77 (0.72 0.83) 0.28 0.0001 Overweight 2.74 (2.64 2.85) 2.27 (2.17 2.37) 2.09 (1.92 2.29) 2.05 (1.86 2.25) 0.0001 Never smoker 1.50 (1.44 1.57) 1.23 (1.19 1.31) 1.05 (0.95 1.15) 1.02 (0.92 1.14) 0.0001 0.0001 Ever smoker 1.86 (1.76 1.97) 1.41 (1.34 1.48) 1.13 (1.05 1.23) 1.13 (1.05 1.23) 0.0001 Normotensive 1.35 (1.30 1.40) 1.15 (1.10 1.19) 0.95 (0.89 1.02) 0.95 (0.89 1.02) 0.0001 0.0001 Hypertensive 2.80 (2.63 2.99) 2.28 (2.12 2.46) 1.89 (1.65 2.16) 1.86 (1.62 2.12) 0.0001 Normoglycemic 1.56 (1.50 1.61) 1.30 (1.25 1.34) 1.09 (1.02 1.16) 1.08 (1.02 1.16) 0.0001 0.0001 Diabetic 4.15 (3.56 4.83) 4.97 (3.86 6.41) 1.51 (0.87 2.61) 2.61 (1.34 5.05) 0.007 Normal cholesterol 1.52 (1.46 1.58) 1.22 (1.17 1.27) 1.02 (0.95 1.09) 0.97 (0.90 1.05) 0.0001 0.0001 High cholesterol 1.97 (1.85 2.10) 1.72 (1.60 1.85) 1.41 (1.24 1.61) 1.60 (1.41 1.82) 0.0004 Age- and BMI-adjusted geometric mean (95% CI) Full cohort 1.43 (1.39 1.47) 1.36 (1.32 1.40) 1.30 (1.23 0.37) 1.34 (1.27 1.41) 0.02 Normal weight* 0.80 (0.76 0.83) 0.78 (0.74 0.81) 0.74 (0.69 0.79) 0.79 (0.73 0.84) 0.58 0.0001 Overweight 2.56 (2.47 2.65) 2.35 (2.26 2.45) 2.33 (2.15 2.52) 2.29 (2.10 2.50) 0.02 Never smoker 1.34 (1.29 1.39) 1.32 (1.26 1.37) 1.27 (1.17 1.38) 1.27 (1.15 1.39) 0.16 0.0001 Ever smoker 1.59 (1.51 1.67) 1.40 (1.34 1.46) 1.32 (1.23 1.42) 1.38 (1.29 1.48) 0.005 Normotensive 1.22 (1.18 1.26) 1.15 (1.11 1.19) 1.10 (1.04 1.17) 1.13 (1.06 1.20) 0.02 0.0001 Hypertensive 2.45 (2.32 2.59) 2.39 (2.24 2.55) 2.29 (2.04 2.56) 2.38 (2.12 2.68) 0.54 Normoglycemic 1.38 (1.34 1.42) 1.32 (1.28 1.36) 1.28 (1.21 1.35) 1.31 (1.24 1.38) 0.07 0.0001 Diabetic 3.89 (3.39 4.46) 5.33 (4.25 6.69) 1.99 (1.21 3.26) 3.63 (2.00 6.58) 0.20 Normal cholesterol 1.32 (1.28 1.37) 1.25 (1.21 1.29) 1.21 (1.14 1.28) 1.20 (1.12 1.28) 0.007 0.0001 High cholesterol 1.80 (1.70 1.90) 1.76 (1.65 1.87) 1.65 (1.47 1.84) 1.89 (1.69 2.12) 0.68 Multivariate-adjusted geometric mean (95% CI) Full cohort 1.43 (1.39 1.48) 1.37 (1.33 1.41) 1.29 (1.22 1.36) 1.28 (1.20 1.37) 0.003 Normal weight* 0.80 (0.76 0.84) 0.79 (0.75 0.82) 0.74 (0.69 0.80) 0.76 (0.70 0.83) 0.25 0.0001 Overweight 2.55 (2.46 2.65) 2.38 (2.28 2.47) 2.30 (2.12 2.50) 2.23 (2.01 2.47) 0.02 Never smoker 1.34 (1.29 1.39) 1.33 (1.27 1.38) 1.27 (1.17 1.37) 1.23 (1.10 1.37) 0.10 0.07 Ever smoker 1.58 (1.50 1.66) 1.42 (1.36 1.49) 1.33 (1.24 1.43) 1.33 (1.22 1.45) 0.005 Normotensive 1.23 (1.18 1.27) 1.16 (1.12 1.20) 1.10 (1.03 1.16) 1.09 (1.01 1.17) 0.007 0.03 Hypertensive 2.45 (2.31 2.59) 2.42 (2.27 2.58) 2.30 (2.05 2.58) 2.28 (1.98 2.63) 0.32 Normoglycemic 1.40 (1.35 1.44) 1.33 (1.29 1.37) 1.26 (1.20 1.33) 1.24 (1.16 1.33) 0.004 0.004 Diabetic 3.80 (3.30 4.38) 5.44 (4.31 6.89) 2.29 (1.36 3.88) 3.95 (2.07 7.55) 0.50 Normal cholesterol 1.32 (1.28 1.37) 1.26 (1.21 1.30) 1.20 (1.13 1.27) 1.17 (1.09 1.27) 0.01 0.05 High cholesterol 1.83 (1.72 1.94) 1.79 (1.68 1.90) 1.62 (1.45 1.82) 1.70 (1.49 1.95) 0.22 * Stratified by BMI 25 or 25 kg/m 2 and additionally adjusted for BMI as a continuous variable. Adjusted for age (continuous); BMI (continuous); parental history of myocardial infarction at 60 years; history of diabetes mellitus; history of hypertension ( 140/90 mm Hg); history of high cholesterol ( 240 mg/dl); smoking status (current, past, never); physical activity levels (rarely or never, 1,1to3,and 4 times/week); and dietary factors (continuous) including glycemic load, total fat, saturated fat, carbohydrate, dietary fiber, folate, cholesterol, and total energy intake, except for the variable used as a stratification factor. CI confidence interval. tile) by frequency of alcohol use (odds ratio 0.67 for subjects who consumed 61 drinks/month, approximately 2 drinks/day, compared with nondrinkers, 95% confidence interval 0.48 to 0.93, p 0.01). 12 In contrast, 2 crosssectional studies in Germany found a U-shaped relation between alcohol consumption and hs-crp. 13,14 In these studies, the amount of alcohol consumed in the top category was much greater ( 40 and 80 g/day) than in this population, which may explain the differences in the observed association. As in the present study, other cross-sectional studies that have examined the associations of different types of alcoholic beverages with hs-crp have found that the associations do not appear to differ greatly. 1,2 Overall, our results confirm the inverse relation between alcohol and hs-crp concentrations in the range of alcohol commonly consumed by residents of the United States and that hs-crp concentrations are similar between beverage preference groups after adjusting for potential lifestyle confounders. Additionally, our results suggest that the inverse association may be limited to overweight patients. Alcohol use and beverage preference in our population of predominately white, middle-aged women were similar to concentrations reported in a nationally representative sample of the United States. 12 Validation studies in a similar

Preventive Cardiology/C-Reactive Protein, Beer, Wine, and Liquor 87 Table 3 Plasma concentration of high-sensitivity C-reactive protein (hs-crp) concentration (mg/l) by beverage preference among drinkers Variable Alcohol Consumed (g/day) p trend 0.1 6 6.1 12 12 Median (interquartile range) Beer only 1.37 (0.60 3.07) 1.17 (0.43 2.82) 0.92 (0.53 3.62) n 420 50 35 Wine only 1.29 (0.54 3.10) 0.98 (0.41 2.46) 1.21 (0.44 2.62) n 1,973 282 108 Liquor only 2.01 (0.77 4.15) 1.88 (0.65 3.80) 1.51 (0.75 3.41) n 286 103 67 Multiple types 1.32 (0.55 3.01) 1.02 (0.45 2.48) 1.11 (0.45 2.39) n 1,750 992 1,088 Age-adjusted geometric mean (95% CI) Beer only 1.38 (1.23 1.54) 1.10 (0.78 1.53) 1.18 (0.82 1.70) 0.18 Wine only 1.27 (1.20 1.33) 0.96 (0.83 1.10) 1.15 (0.93 1.41) 0.005 Liquor only 1.65 (1.44 1.89) 1.53 (1.21 1.93) 1.40 (1.07 1.82) 0.47 Multiple types 1.30 (1.23 1.37) 1.09 (1.01 1.17) 1.08 (1.01 1.16) 0.0001 p value for heterogeneity 0.004 0.01 0.31 Age- and BMI-adjusted geometric mean (95% CI) Beer only 1.33 (1.21 1.46) 1.03 (0.77 1.39) 1.14 (0.82 1.57) 0.70 Wine only 1.31 (1.25 1.36) 1.09 (0.96 1.23) 1.16 (0.97 1.40) 0.79 Liquor only 1.32 (1.17 1.48) 1.28 (1.04 1.57) 1.24 (0.98 1.57) 0.23 Multiple types 1.31 (1.25 1.37) 1.07 (1.00 1.14) 1.09 (1.03 1.16) 0.55 p value for heterogeneity 0.99 0.43 0.69 Multivariate-adjusted geometric mean* (95% CI) Beer only 1.29 (1.18 1.42) 0.97 (0.72 1.30) 1.02 (0.74 1.41) 0.35 Wine only 1.31 (1.26 1.37) 1.10 (0.97 1.24) 1.14 (0.95 1.37) 0.33 Liquor only 1.28 (1.14 1.44) 1.17 (0.95 1.44) 1.09 (0.86 1.38) 0.59 Multiple types 1.32 (1.26 1.38) 1.08 (1.01 1.15) 1.11 (1.04 1.17) 0.05 p value for heterogeneity 0.97 0.76 0.95 * Adjusted for age (continuous); BMI (continuous); parental history of myocardial infarction at 60 years; history of diabetes mellitus; history of hypertension ( 140/90 mm Hg); history of high cholesterol ( 240 mg/dl); smoking status (current, past, never); physical activity levels; and dietary factors (continuous) including glycemic load, total fat, saturated fat, carbohydrate, dietary fiber, folate, cholesterol, and total energy intake. CI confidence interval. population have demonstrated that alcohol is well measured by the semiquantitative food-frequency questionnaire compared with diet records, although the 2 methods depend on self-report. 7,8 The under-reporting of alcohol intake, especially among heavier consumers, may overestimate the association with hs-crp in the small intake range. Our questionnaire assesses usual intake, which may not reflect the alcohol intake in the days or weeks before the blood draw, leading to the potential misclassification of alcohol use in the relevant time period and the possible underestimation of the true relation. Additionally, we did not have information regarding weight loss in the period before the blood draw, which may bias results if it is associated with short-term changes in alcohol intake. The hs-crp measurements were performed using a high-sensitivity assay allowing the examination of hs-crp as a continuous variable. Although the large sample size affords a reliable measure of hs-crp concentrations, the cross-sectional nature of our analysis cannot prove a causal relation. Although the relation between alcohol and hs-crp remained significant after extensive control for nutritional and lifestyle confounding, we cannot rule out some unmeasured or residual confounding. However, a crossover feeding study in middle-aged subjects found a statistically significant 35% reduction in hs-crp concentrations when beer (40 g ethanol/day for men and 30 g ethanol/day for women) was added to a controlled diet, 15 although this was not seen in an earlier feeding study (23 g ethanol/day from wine) in young men. 16 1. Wannamethee SG, Lowe GD, Shaper G, Whincup PH, Rumley A, Walker M, Lennon L. The effects of different alcoholic drinks on lipids, insulin and haemostatic and inflammatory markers in older men. Thromb Haemost 2003;90:1080 1087. 2. Imhof A, Woodward M, Doering A, Helbecque N, Loewel H, Amouyel P, Lowe GD, Koenig W. Overall alcohol intake, beer, wine, and systemic markers of inflammation in Western Europe: results from three MONICA samples (Augsburg, Glasgow, Lille). Eur Heart J 2004;25:2092 2100. 3. Liu S, Manson JE, Lee IM, Cole SR, Hennekens CH, Willett WC, Buring JE. Fruit and vegetable intake and risk of cardiovascular disease: the Women s Health Study. Am J Clin Nutr 2000;72:922 928. 4. Liu S, Buring JE, Sesso HD, Rimm EB, Willett WC, Manson JE. A prospective study of dietary fiber intake and risk of cardiovascular disease among women. J Am Coll Cardiol 2002;39:49 56. 5. Willett WC. Nutritional Epidemiology. 2nd Ed. New York: Oxford University Press, 1998.

88 The American Journal of Cardiology (www.ajconline.org) 6. US Department of Agriculture. Composition of Foods Raw, Processed, and Prepared, 1963-1988. Agricultural Handbook No. 8. Washington, District of Columbia: US Government Printing Office, 1989. 7. Willett WC, Stampfer MJ, Colditz GA, Rosner BA, Hennekens CH, Speizer FE. Moderate alcohol consumption and the risk of breast cancer. N Engl J Med 1987;316:1174 1180. 8. Giovannucci E, Colditz G, Stampfer MJ, Rimm EB, Litin L, Laura S, Willett WC. The assessment of alcohol consumption by a simple self-administered questionnaire. Am J Epidemiol 1991;133:810 817. 9. Roberts WL, Moulton L, Law TC, Farrow G, Cooper-Anderson M, Savory J, Rifai N. Evaluation of nine automated high-sensitivity C- reactive protein methods: implications for clinical and epidemiological applications. Part 2. Clin Chem 2001;47:418 425. 10. Lowe G, Upton M, Rumley A, McConnachie A, O Reilly DS, Watt GC. Different effects of oral and transdermal hormone replacement therapies on factor IX, APC resistance, t-pa, PAI and C-reactive protein: a cross-sectional survey. Thromb Haemost 2001;86:550 556. 11. Albert MA, Glynn RJ, Ridker PM. Alcohol consumption and plasma concentration of C-reactive protein. Circulation 2003;107:443 447. 12. Stewart SH, Mainous AG III, Gilbert G. Relation between alcohol consumption and C-reactive protein levels in the adult US population. J Am Board Fam Pract 2002;15:437 442. 13. Imhof A, Froehlich M, Brenner H, Boeing H, Pepys MB, Koenig W. Effect of alcohol consumption on systemic markers of inflammation. Lancet 2001;357:763 767. 14. Koenig W, Sund M, Frohlich M, Fischer HG, Lowel H, Doring A, Hutchinson WL, Pepys MB. C-reactive protein, a sensitive marker of inflammation, predicts future risk of coronary heart disease in initially healthy middle-aged men: results from the MONICA (Monitoring Trends and Determinants in Cardiovascular Disease) Augsburg Cohort Study, 1984 to 1992. Circulation 1999;99:237 242. 15. 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. 16. Mezzano D, Leighton F, Martinez C, Marshall G, Cuevas A, Castillo O, Panes O, Munoz B, Perez DD, Mizon C, et al. Complementary effects of Mediterranean diet and moderate red wine intake on haemostatic cardiovascular risk factors. Eur J Clin Nutr 2001;55:444 451.