Although moderate alcohol drinkers have lower rates of incident coronary artery disease than abstainers, much less is known about the health effects of different patterns of alcohol use in women with established coronary artery disease. In the Determinants of Myocardial Infarction Onset Study, 1,253 women hospitalized for acute myocardial infarction (MI) at 64 centers nationwide from 1989 to 1996 were followed for mortality through December 31, 2007. Of the women, 761 (61%) reported abstention in the year before their MIs, 280 (22%) reported consumption of <1 serving/week, 75 (6%) reported consumption of 1 to 3 servings/week, and 137 (11%) reported consumption of ≥3 servings/week. Using Cox proportional-hazards models, the associations between total weekly volume of consumption, drinking days per week, drinks per drinking day, and beverage type with 10-year mortality were investigated, adjusting for clinical and socioeconomic potential confounders. Compared with abstention, adjusted hazard ratios were 0.66 (95% confidence interval 0.50 to 0.86) for <1 serving/week, 0.65 (95% confidence interval 0.38 to 1.11) for 1 to 3 servings/week, and 0.65 (95% confidence interval 0.38 to 1.11) for ≥3 servings/week (p for trend = 0.008). No differences were found by beverage type, and generally inverse associations of drinking frequency and quantity with mortality were found. In conclusion, in women who survive MI, moderate drinking is associated with a decreased risk for mortality, with no clear differences on the basis of pattern or beverage type. These results suggest that women who survive MI need not abstain from alcohol, but any derived benefit would appear to occur well below currently recommended limits in alcohol consumption.
We studied mortality after acute myocardial infarction (AMI) as a function of weekly total alcohol consumption, consumption pattern, and beverage type, in the year before AMI in women enrolled in the Determinants of Myocardial Infarction Onset Study (the Onset study). This multicenter, prospective cohort study included chart reviews and face-to-face interviews with patients who were hospitalized with confirmed AMI.
Methods
The first phase of the Onset study was conducted at 45 community hospitals and tertiary care medical centers in the United States from August 1989 to September 1994 and was then expanded to 64 medical centers through September 1996. Altogether, 1,259 women enrolled in the study. We excluded patients with missing information on usual alcohol consumption (n = 1) and those with histories of alcoholism who reported current abstention (n = 5), leaving 1,253 patients for analysis. The institutional review board of each center approved this protocol, and all participants provided informed consent. For the analyses in the present study, approval was obtained from the Beth Israel Deaconess Medical Center Committee on Clinical Investigations.
Trained research interviewers identified all eligible patients by reviewing coronary care unit admission logs and patient charts. For inclusion, patients were required to have creatine kinase levels higher than the upper limit of normal for the clinical laboratory at each center, positive MB isoenzymes, an identifiable onset of pain or other symptoms typical of AMI, and the ability to complete a structured interview. Interviewers used a structured data abstraction and questionnaire form. Patients were interviewed during their initial inpatient admission for AMI after they had been medically stabilized.
Participants reported average frequency during the past year of consumption (and corresponding numbers of drinks) of wine, beer, and liquor individually. We determined each patient’s average weekly ethanol consumption from wine, beer, and liquor on the basis of the average ethanol content for a serving of each beverage type reported in the early 1990s (i.e., 13.2 g for beer, 10.8 g for wine, and 15.1 g for liquor). We defined a standard serving of alcohol as 13.7 g of ethanol and categorized average alcohol consumption as none or <1, 1 to <3, or ≥3 servings/week. We also subdivided the <1 serving/week category into <1 serving/month and from 1 to <1 serving/week to assess for any dose-response pattern among the lightest drinkers. Patients who reported drinking a certain beverage type but did not say how much they drank on each occasion were assigned the median value of 1 drink/drinking occasion for each beverage type (n = 70). Results were similar when these patients were excluded from analyses (results not shown).
To assess drinking patterns, we analyzed drinking frequency in drinking days per week and drinking quantity in drinks per drinking day. Patients were asked how often they drank each of the 3 beverage types individually, so for the drinking days per week analyses, we summed up the days per week that patients reported drinking beer, wine, or liquor, assuming these were drunk on separate days. As a sensitivity analysis, we made the opposite assumption: that all beverages were consumed on the same day(s). Drinking days per week were categorized into abstainers (0 days/week) or <1, 1 to <3, or 3 to 7 days/week. For drinks per drinking day analyses, we again assumed that all drinks were consumed on separate days and made the opposite assumption as a sensitivity analysis. These were categorized into ≤1 or >1 drink/drinking day. For the frequency and volume per drinking occasion analyses, we performed a sensitivity analysis by adding total alcohol intake as a covariate. For beverage-specific analyses, weekly consumption of each beverage type was categorized into 0, <1, or ≥1 drinks/week. As a measure of heavy episodic drinking during the preceding year, patients reported the usual frequency with which they consumed ≥3 drinks in 1 to 2 hours for each of the beverage types.
Other information collected included age, gender, medical history from chart review, smoking history, and prescription and nonprescription medication use. During the chart review, interviewers recorded complications of congestive heart failure or ventricular arrhythmias on the basis of clinical diagnoses documented in medical records and all creatine kinase values available at the time of chart review. We used 1990 United States census data to derive median household income from census block groups. We defined noncardiac co-morbidity as any diagnosis of cancer, respiratory disease, renal failure, or stroke recorded in medical records. We derived body mass index on the basis of self-reported height and weight. Patients were asked their usual frequency of heavy physical activity using a validated instrument. As with previous Onset study analyses, we categorized the usual frequency of physical activity as activity ≥6 METs in the following frequencies: <1, 1 to 4, or >4 times/week.
We searched the National Death Index for deaths of Onset study participants through December 31, 2007, and requested death certificates from state offices of vital statistics records for all probable matches using a previously validated algorithm. Three physicians blinded to exposure data independently verified the determination of each death. Disagreements among raters were resolved by discussion. The outcome measure in all analyses was all-cause mortality after 10 years of follow-up.
We performed univariate comparisons of continuous and binary variables using analysis of variance and chi-square or Fisher’s exact tests, respectively. We used Cox proportional-hazards models to examine the independent effect of alcohol use on mortality. We performed separate analyses for weekly total alcohol consumption, drinking days per week, drinking sessions per day, beverage-specific analyses, and heavy episodic drinking. In all models, we adjusted for age, body mass index (as linear and quadratic terms), previous AMI (yes, no, or uncertain), previous congestive heart failure, previous angina, diabetes mellitus, hypertension, noncardiac co-morbidity, previous medication use (aspirin, β blockers, calcium channel blockers, digoxin, and angiotensin-converting enzyme inhibitors individually), current or previous smoking, frequency of physical activity (in 3 categories), household income (in quartiles), education (in 3 categories), marital status (married or single), race, and measures of index AMI treatment and severity (peak creatine kinase level, receipt of thrombolytic therapy, and congestive heart failure and ventricular tachycardia during hospitalization). We used indicator variables for missing education (n = 26), marital status (n = 15), and income (n = 23). For patients missing body mass index (n = 16), we assigned the mean value.
We tested hazard ratios (HRs) for linear trend across alcohol consumption categories. We tested the proportionality of hazards using time-varying covariates and Schoenfeld residuals and found no significant violations. We present HRs from Cox models with 95% confidence intervals. All probability values are 2 sided.
Results
Characteristics of the women according to alcohol consumption are listed in Table 1 . The median consumption in the heavier consumption group was 7.6 servings/week, and only 31 women in the study drank >14 servings/week. Higher alcohol consumption was associated with younger age, current or former smoking, higher household income, and higher educational attainment. It was inversely associated with cardiac morbidity and the use of cardiac medications.
| Average Alcohol Consumption (Servings/Week) | |||||
|---|---|---|---|---|---|
| Characteristic | None | <1 | ≥1 to <3 | ≥3 | p Value |
| (n = 761) | (n = 280) | (n = 75) | (n = 137) | ||
| Age (years) | 68.1 ± 11.8 | 64.5 ± 12.3 | 60.0 ± 14.1 | 61.9 ± 12.6 | <0.0001 |
| White race | 655 (86%) | 251 (90%) | 66 (88%) | 123 (90%) | 0.36 |
| Married | 325 (43%) | 133 (48%) | 38 (51%) | 77 (56%) | 0.019 |
| Income ($) | 35,938 ± 17,010 | 38,000 ± 16,560 | 39,612 ± 15,741 | 42,724 ± 19,015 | 0.0002 |
| Education | <0.001 | ||||
| Less than high school | 231 (30%) | 58 (21%) | 14 (19%) | 16 (12%) | |
| Completed high school | 365 (48%) | 138 (49%) | 33 (44%) | 63 (46%) | |
| Some college | 148 (19%) | 78 (28%) | 25 (33%) | 58 (42%) | |
| Body mass index (kg/m 2 ) | 27.9 ± 6.2 | 27.9 ± 5.7 | 26.3 ± 4.8 | 25.3 ± 4.5 | <0.0001 |
| Smoking status | <0.001 | ||||
| Current | 174 (23%) | 95 (34%) | 29 (39%) | 67 (49%) | |
| Former | 215 (28%) | 104 (37%) | 24 (32%) | 46 (34%) | |
| Physical activity (times/week) | 0.014 | ||||
| <1 | 720 (94%) | 251 (90%) | 69 (91%) | 124 (91%) | |
| 1–4 | 29 (4%) | 25 (9%) | 7 (9%) | 9 (7%) | |
| ≥4 | 12 (2%) | 4 (1%) | 0 (0%) | 4 (3%) | |
| Morbidity | |||||
| Hypertension | 419 (55%) | 128 (46%) | 33 (44%) | 60 (44%) | 0.0060 |
| Diabetes mellitus | 283 (37%) | 56 (20%) | 9 (12%) | 13 (9%) | <0.0001 |
| Previous myocardial infarction | 226 (30%) | 54 (19%) | 15 (20%) | 18 (13%) | <0.0001 |
| Angina | 235 (31%) | 68 (24%) | 12 (16%) | 24 (18%) | 0.0005 |
| Congestive heart failure | 60 (8%) | 5 (2%) | 1 (1%) | 1 (1%) | <0.0001 |
| Regular use of | |||||
| Angiotensin-converting enzyme inhibitors | 153 (20%) | 29 (10%) | 7 (9%) | 14 (10%) | <0.0001 |
| Aspirin | 272 (36%) | 106 (38%) | 28 (37%) | 44 (32%) | 0.71 |
| β blockers | 192 (25%) | 63 (23%) | 15 (20%) | 32 (24%) | 0.65 |
| Calcium channel blockers | 246 (32%) | 71 (25%) | 15 (20%) | 24 (18%) | 0.0005 |
| Digoxin | 70 (9%) | 19 (7%) | 5 (7%) | 4 (3%) | 0.068 |
| Index hospitalization | |||||
| Thrombolytic use | 229 (30%) | 109 (39%) | 25 (33%) | 69 (50%) | <0.0001 |
| Congestive heart failure | 145 (19%) | 44 (16%) | 13 (17%) | 24 (18%) | 0.66 |
| Ventricular tachycardia | 45 (6%) | 21 (8%) | 7 (9%) | 15 (11%) | 0.15 |
| Peak creatine kinase level (U/L) | 1,243 ± 1,633 | 1,401 ± 1,565 | 1,242 ± 1,380 | 1,589 ± 2,317 | 0.12 |
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
