How much alcohol and how often? Population based case-control study of alcohol consumption and risk of a major coronary eventBMJ 1997; 314 doi: https://doi.org/10.1136/bmj.314.7088.1159 (Published 19 April 1997) Cite this as: BMJ 1997;314:1159
- a Department of Statistics and Centre for Clinical Epidemiology and Biostatistics, University of Newcastle, New South Wales 2308, Australia
- Correspondence to: Mr McElduff
- Accepted 24 January 1997
Objective:To quantify the effects of quantity and frequency of alcohol consumption on risk of acute myocardial infarction and coronary death.
Setting:Lower Hunter region of New South Wales, Australia, 1983-94.
Subjects:Men and women aged 35-69 years.
Main outcome measure:Acute myocardial infarction or coronary death.
Results:Alcohol consumption patterns were compared between 11 511 cases of acute myocardial infarction or coronary death and 6077 controls randomly selected from the same study population. After adjusting for the effects of age, smoking, and medical history, men and women who consumed one or two drinks of alcohol on five or six days a week had a reduction in risk of a major coronary event compared with men and women who were non-drinkers (odds ratios: men 0.31 (95% confidence interval 0.22 to 0.45); women 0.33 (0.18 to 0.59)). A similar reduction in risk was found after excluding non-drinkers who were formerly moderate to heavy drinkers. An acute protective effect of alcohol consumption was also found for regular drinkers who consumed one or two drinks in the 24 hours preceding the onset of symptoms (odds ratios: men 0.74 (0.51 to 1.09); women 0.43 (0.20 to 0.95)).
Conclusions:Frequency and quantity of alcohol consumption are important in assessing the risk of a major coronary event. Risk is lowest among men who report one to four drinks daily on five or six days a week and among women who report one or two drinks daily on five or six days a week.
Alcohol consumption has been associated with a reduced risk of coronary heart disease
Broad categories of average weekly consumption of alcohol do not take into account the importance of frequency of consumption
A new study shows that men and women who consume one or two alcoholic drinks a day on five or six days a week have a substantially reduced risk of coronary heart disease
Alcohol consumption is associated with an acute protective effect for 24 hours
Adverse physical and social effects of alcohol consumption should prevent consumption of alcohol being recommended as a health measure
Several studies have shown that moderate consumption of alcohol is associated with a reduced risk of coronary heart disease.1 2 3 4 5 Other studies have shown little or no association.6 7 In these studies subjects were categorised either by the average number of alcoholic drinks consumed per week or in broad groups of light, moderate, or heavy consumption. These methods tend to group together people who have completely different drinking habits–for example, those who have two drinks a day six days a week and those who have a dozen drinks on one day of the week.
We conducted a case-control study to quantify the joint effects of frequency and quantity of alcohol consumed on the risk of a major coronary event. We also investigated the suggestion by Jackson et al that moderate consumption of alcohol has an acute protective effect.8
Subjects and methods
This study was a product of the World Health Organisation's MONICA project, which monitored trends and determinants in cardiovascular disease in well defined populations in more than 20 countries over 10 years. One such population was subjects aged 35-69 years in Newcastle, Australia.
Cases were defined as patients who had coronary events which satisfied the criteria for non-fatal definite myocardial infarction, non-fatal possible myocardial infarction, fatal definite myocardial infarction, fatal possible myocardial infarction, or coronary death with insufficient information for further classification.9 Information on cases was obtained by following up all suspected coronary events occurring in the study population. This entailed interviewing patients while still in hospital to obtain information on symptoms, medical history, and other variables. Cardiac enzyme activities were extracted from hospital notes and electrocardiograms copied and coded according to the Minnesota code. Details of fatal cases were obtained from death certificates and postmortem records and from doctors, relatives, or other informants. For this study cases were included for the whole period 1984-94.
Controls were participants in risk factor prevalence studies conducted as part of the MONICA project. Those studies were conducted in June to December 1983, June to December 1988 and June to November 1989, and June to December 1994.
For each risk factor study stratified random samples of the study population were selected from the electoral rolls. (In Australia registration on the electoral roll is compulsory for most people.) In 1983 the risk factor survey did not include people aged 65-69 and in 1988-9 and 1994 the sampling fraction was greater for the older age groups. People chosen for the sample were invited to attend study centres to complete a self administered questionnaire and have physical measurements and a blood sample taken. Extensive systems of reminders and follow up were used to encourage participation. The response rate in 1983 was 68%; in 1988-9, 63%; and in 1994, 64%. People who could not attend the centres were asked to complete a brief questionnaire. In 1994 the brief questionnaire included the same questions on frequency and quantity of alcohol consumption as the main questionnaire. This resulted in an increase in the response rate for these items in 1994 from 64% to 75%.
Smoking, age, hypertension, hypercholesterolaemia, diabetes, and a history of angina, acute myocardial infarction, and stroke are associated with an increased risk of acute myocardial infarction. If they are also associated with alcohol consumption, then they are potential confounders to the relation between alcohol consumption and acute myocardial infarction. We therefore adjusted for each of these factors in the analyses.
Subjects were stratified in five year age groups from 35-39 to 65-69. Subjects were deemed to have a history of heart disease if they answered “yes” to either, “Have you ever been told you have had a heart attack/myocardial infarction?” or “Have you ever been told you have angina?” Subjects were deemed to have high blood pressure if they answered “yes” to, “Have you ever been told by a doctor or other medical person that you have high blood pressure?” A similar question was used to ascertain hypercholesterolaemia.
Information about alcohol consumption was obtained by two questions. To determine the frequency of alcohol consumption subjects were asked, “How often do you usually drink alcohol?” Response categories were every day, five or six days a week, three or four days a week, one or two days a week, less than once a week, rarely, and never. The question on quantity was, “On a day when you do drink alcohol, how many drinks do you usually have?” Subjects were asked to respond in terms of standard drinks (10 g alcohol), using the categories more than 20, 13-20, 9-12, 5-8, 3 or 4, 1 or 2, and “I don't drink.” Very few men took more than 20 drinks a day, and they were therefore grouped with those who took 13-20 drinks daily. Similarly for women the categories of more than 20, 13-20, 9-12, and 5-8 were combined to form one category of more than five drinks a day. It was more difficult to obtain information for those who died than for survivors. Data on alcohol consumption were not available for 524 (6.2%) of the 8482 cases who survived and 1422 (46.9%) of the 3029 cases who died.
In the 1994 risk factor prevalence survey an additional category was added to the possible responses to the question on frequency of alcohol consumption. This was, “Used to be a moderate to heavy drinker.” The same response category was also included from 1986 for the cases. Using this information we conducted a subgroup analysis with cases from 1991 to 1994 and with controls from the 1994 risk factor survey. In this analysis we separated people who used to be moderate to heavy drinkers from the group of non-drinkers.
To investigate the claim of an acute protective effect of alcohol consumption, we calculated crude and adjusted odds ratios for regular drinkers who consumed 1 or 2, 3 or 4, 5-8, and 9 or more drinks in the 24 hours before the onset of symptoms compared with regular drinkers who did not consume any alcohol in the period. As before, the categories of 5-8 and 9 or more drinks were combined for women. Controls were regular drinkers who participated in the risk factor prevalence surveys, and the exposure variable was their alcohol consumption in the 24 hours before the interview. Regular drinkers were those who reported drinking at least once a week.
Initial exploratory analysis entailed comparing cases and controls for factors known to be associated with the risk of a major coronary event. Χ2 Tests were applied to differences in the proportions of cases and controls who had a previous myocardial infarction, angina, stroke, high blood pressure, high cholesterol concentration, or diabetes and to test for an association between case-control status and age group or cigarette smoking.
To measure any difference in risk of a major coronary event associated with alcohol consumption after adjusting for the effects of age, smoking, previous myocardial infarction, angina, stroke, history of high blood pressure, cholesterol concentration, and diabetes odds ratios and 95% confidence intervals were calculated by logistic regression. The base category for alcohol consumption in the analysis was, “I don't drink.” Other categories were defined by cross tabulation of the frequency and quantity categories. Logistic regression was performed with the genmod procedure in sas.10
A total of 11 511 cases were registered by the Newcastle MONICA project during 1984-94 and 6077 controls participated in the risk factor prevalence studies. Of these subjects, 1946 (16.9%) cases and 12 (0.2%) controls were excluded from analysis because of insufficient information on the quantity or frequency of alcohol consumed. Cases excluded from analysis were more likely to be 65-69 years old (34.5% v 29.3%; P<0.001), less likely to have survived the event (26.9% v 83.2%; P<0.001), and as likely to be male (70.2% v 69.9%; P=0.79) as cases who were not excluded.
Table 1 shows that for both men and women cases were significantly more likely than controls to be older, current smokers, and have a history of diabetes, high blood pressure, myocardial infarction, angina, and stroke. The age and sex distribution of controls was determined by the design of the study, so that adjustment for age and sex differences was necessary for all analyses.
Compared with subjects who did not drink alcohol there was a significant reduction in risk of a major coronary event for men who took one to four drinks daily and women who took one or two drinks daily less than once a week up to five or six days a week (table 2). There was an increased risk of a major coronary event for men and women who took one or two drinks a day rarely and for men who took more than 13 drinks a day on one or two days a week or every day.
After removing former moderate to heavy drinkers from the non-drinking group there remained a reduction in risk of a major coronary event for men who took one to four drinks a day on five or six days a week and for women who took one or two drinks a day on three or four days a week (table 3). Men who took nine or more drinks a day on one or two days a week or every day had an increased risk of a major coronary event compared with men who did not drink and were never moderate to heavy drinkers.
Women had a reduced risk of a major coronary event in the 24 hours after consuming one or two alcoholic drinks compared with regular drinkers who consumed no alcohol in the period (table 4). There was a possible reduced risk for men but it was not significant.
This study shows that moderate regular consumption of alcohol over five or six days a week is associated with a reduction in risk of a major coronary event. The increased risk for “binge” drinkers compared with non-drinkers is in contrast with the reduction in risk for those drinking a similar amount a week spread over more days. For example, men who took nine or more drinks a day on one or two days a week consumed similar amounts to those who took three or four drinks a day on five or six days a week but the odds ratios for the two groups were substantially different (2.62 (95% confidence interval 1.12 to 6.17) v 0.46 (0.27 to 0.80)).
The biological effects of alcohol depend on how much and how often alcohol is consumed.11 12 An increase in blood pressure in drinkers is influenced more by the frequency of consumption than by the quantity consumed.11 Moderate consumption of alcohol causes temporary changes in the fibrinolytic system, which returns to normal within 24 hours.12 This explains why people who consumed alcohol on five or six days a week had a lower risk of a major coronary event than those who consumed alcohol once a week. It also helps to explain why those who consumed large amounts on one or two days a week did not gain the same benefit as those who consumed similar amounts over five or six days. However, the observation that those who consumed alcohol every day did not seem to have the same beneficial effect suggests that the biological mechanisms of alcohol consumption are more complex than fibrinolytic changes alone. The authors also claimed that the pattern of increase in circulating tissue type plasminogen activator activity, particularly in the morning (13 hours after consumption), may have a protective effect at a time when a large proportion of heart attacks occur.12 This could explain the protective effect of alcohol consumption in the 24 hours before onset.
Suh et al found a positive association between high density lipoprotein cholesterol concentration and consumption of alcohol.13 They, however, concluded that the effect of alcohol consumption on high density lipoprotein cholesterol only partly explained the reduction in coronary deaths.
Some critics of epidemiological studies that have shown a reduction in the risk of coronary heart disease with moderate alcohol consumption claim that the effect is due to former heavy drinkers or people who are otherwise ill becoming non-drinkers.6 In all our analyses we controlled for the effects of history of high blood pressure, angina, stroke, previous myocardial infarction, high cholesterol concentration, and diabetes. This adjustment substantially improved the fit of the model but had little effect on the point estimate for each category of alcohol consumption. Similar results were observed after excluding all cases and controls with a history of acute myocardial infarction, angina, or stroke. Even when former moderate to heavy drinkers were excluded from the analysis there seemed to be a reduction in the risk of a major coronary event for those who consumed a moderate amount of alcohol on three to six days a week. The analysis which excluded previously moderate to heavy drinkers showed no increase in risk among occasional drinkers, and among women who consumed one or two drinks only rarely was there a significant reduction in risk. Thus many previously moderate to heavy drinkers may have reported being occasional drinkers.
Possible confounding factors
A weakness of this study was the large number of cases from whom we did not obtain information on the pattern of alcohol consumption. Of the 1946 cases excluded, 73.1% had died within 28 days after the onset of symptoms. This group may bias the results either in favour of or against a protective effect of alcohol consumption, depending on whether they were more or less likely to be regular drinkers than those included in the analysis.
The validity of self reported alcohol consumption is a possible source of concern in this paper. Romelsjo et al showed that the quantity-frequency approach, as used in this paper, resulted in underreporting of alcohol consumption by all sections of the community, women underreporting more than men.14 General underreporting or overreporting of the quantity of alcohol consumed does not affect the ordinal validity of this study, though it could bias the estimate of threshold levels for “safe” drinking.15 If women underreported their alcohol consumption more than men this could explain why the reduction in risk for women found in this study was less than the reduction in risk for men.
There was a significant reduction in risk of a major coronary event for women who consumed one or two drinks in the 24 hours before the onset of symptoms and a non-significant reduction in risk for men. This is consistent with the claim by Jackson et al of an acute protective effect of moderate alcohol consumption.8 This acute protective effect could be a result of changes in fibrinolytic factors which occur within two hours of alcohol consumption12 and are known to reduce blood clots rather than some cases not drinking in the 24 hour period due to non-specific prodromal symptoms as suggested by Jackson et al.
To compare our data with results from other studies we multiplied the average value of each frequency category by the average value of each quantity category to obtain a crude measure of the average number of alcoholic drinks consumed a week. Dividing this result by 7, we categorised subjects as consuming none, less than 1, 1 or 2, 2-4, 4-7, or more than 7 drinks a day. Comparing the risk of acute myocardial infarction for each of these categories with the risk for those who were non-drinkers, we found a similar U shaped curve as reported elsewhere, with the lowest risk for men who consumed two to four alcoholic drinks a day (odds ratio 0.63; 95% confidence interval 0.53 to 0.76) and for women who consumed two to four alcoholic drinks a day (odds ratio 0.54; 0.36 to 0.82). Though these results are consistent with those of other investigators,1 3 4 they obscure the different effects of frequency and quantity of alcohol shown in tables 2 and 3.
Despite the results of this and other studies caution is needed in promoting alcohol consumption because the adverse effects of abuse may well outweigh any potentially beneficial effect in reducing heart disease. This paper is intended to clarify understanding of the biological effect alcohol consumption has on coronary heart disease and provide a better understanding of the aetiology of the disease.
Funding: Commonwealth Department of Health and Family Services, National Heart Foundation of Australia, and National Health and Medical Research Council of Australia.
Conflict of interest: None.