Fax +41 61 306 12 34
E-Mail karger@karger.ch
www.karger.com
Research Report
Eur Addict Res 2009;15:143–149
DOI: 10.1159/000213641
Patterns of Alcohol Consumption
and Acute Myocardial Infarction:
A Case-Crossover Analysis
M.G. Gerlich
a, b
A. Krämer
b
G. Gmel
c
M. Maggiorini
d
T.F. Lüscher
d
H. Rickli
e
G.R. Kleger
e
J. Rehm
a, f–h
a
Research Institute for Public Health and Addiction, Zurich , Switzerland;
b
Department of Health Medicine,
University of Bielefeld, Germany;
c
Swiss Institute for the Prevention of Alcohol and Drug Problems, Lausanne,
d
Department of Internal Medicine, University Hospital Zurich, Zurich,
e
Department of Internal Medicine,
Cantonal Hospital St. Gallen, St. Gallen, Switzerland;
f
Department of Public Health Sciences, University of Toronto,
g
Centre for Addiction and Mental Health, Toronto, Ont., Canada;
h
Epidemiological Research Unit,
Technische Universität Dresden, Clinical Psychology and Psychotherapy, Dresden, Germany
Introduction
Regular moderate alcohol consumption without
heavy drinking has been shown to be a protective factor
for cardiovascular disease including acute myocardial
infarction (AMI) in particular
[1, 2] . Other protective
factors include regular physical activity and daily con-
sumption of fruits and vegetables, whereas hyperlipid-
emia, smoking, hypertension, diabetes mellitus and obe-
sity have been identified as major modifiable risk factors
for AMI
[3, 4] . Furthermore, psychosocial stress has been
linked to an increased risk for AMI
[5, 6] , and mental ill-
ness, such as depression, has been shown to be an inde-
pendent predictor of coronary heart disease (CHD)
events
[7, 8] .
Studies have also explored the influence of certain ex-
posures shortly before the event, and factors such as heavy
physical exertion
[9–11] , anger [12, 13] and consumption
of cocaine
[14] or marijuana [15] have been observed to
be potential triggers of AMI onset. While extensive re-
search has examined the influence of usual or regular
alcohol use on CHD outcomes, the influence of alcohol
consumption prior to the event has received little atten-
tion. Some case reports of men under the age of 40 have
Key Words
Acute myocardial infarction ⴢ Coronary heart disease ⴢ
Alcohol ⴢ Drinking patterns ⴢ Binge drinking ⴢ Case crossover
Abstract
Background: Alcohol consumption has been causally relat-
ed to the incidence of coronary heart disease, but the role of
alcohol before the event has not been explored in depth.
This study tested the hypothesis that heavy drinking (binge
drinking) increases the risk of subsequent acute myocardial
infarctions (AMI), whereas light to moderate drinking occa-
sions decrease the risk. Methods: Case-crossover design of
250 incident AMI cases in Switzerland, with main hypotheses
tested by conditional logistic regression. Results: Alcohol
consumption 12 h before the event significantly increased
the risk of AMI (OR 3.1; 95% CI 1.4–6.9). Separately, the effects
of moderate and binge drinking before the event on AMI
were of similar size but did not reach significance. In addi-
tion, AMI patients showed more binge drinking than compa-
rable control subjects from the Swiss general population.
Conclusions: We found no evidence that alcohol consump-
tion before the event had protective effects on AMI. Instead,
alcohol consumption increased the risk.
Copyright © 2009 S. Karger AG, Basel
Published online: April 22, 2009
E
u
r
o
p
e
a
n
A
d
d
i
c
t
i
o
n
c
R
e
e
s
a
r
h
Miriam G. Gerlich
Research Institute for Public Health and Addiction
Konradstrasse 32, CH–8031 Zurich (Switzerland)
Tel. +41 44 448 11 60, Fax +41 44 448 11 70
E-Mail miriam.gerlich@isgf.uzh.ch
© 2009 S. Karger AG, Basel
1022–6877/09/0153–0143$26.00/0
Accessible online at:
www.karger.com/ear
Gerlich/Krämer/Gmel/Maggiorini/
Lüscher/Rickli/Kleger/Rehm
Eur Addict Res 2009;15:143–149
144
discussed heavy acute alcohol intake as a potential trigger
for AMI
[16 –19] , but only a few studies in the literature
have systematically investigated the effect of alcohol ex-
posure shortly before CHD events using a case-crossover
or case-control design [20–22] . Moreover, the results of
these controlled studies have been mixed.
The influence of binge drinking is of particular inter-
est in this analysis. In a recent study of alcohol drinking
patterns and myocardial infarction (MI) in women, those
who drank at least once a month in a way that ‘their
speech was slurred or they became unsteady on their feet’,
a behavior classified as ‘intoxication’, had nearly a three-
fold risk of MI when compared to lifetime abstainers.
Among current drinkers, women who drank to intoxica-
tion once or more per month had a sixfold increased like-
lihood of MI
[23] .
As reviewed by McKee and Britton
[24] and further
sources
[1, 2] , a number of possible mechanisms have
been discussed in the literature explaining the acute ef-
fects of binge drinking on CHD events. These include
increased blood clotting and reduced threshold for ven-
tricular fibrillation: (i) Binge drinking has been shown to
increase low-density lipoproteins, which in turn are
linked to cardiovascular morbidity and mortality. Con-
trary to low or moderate steady drinking, heavy irregular
drinking has not been associated with increased levels of
high-density lipoproteins, which are linked to favorable
cardiovascular outcomes. (ii) Binge drinking has been as-
sociated with an increased risk of thrombosis, occurring
after cessation of a heavy drinking episode. (iii) Binge
drinking appears to predispose drinkers to histological
changes in the myocardium and conducting system, as
well as to a reduction in the threshold for ventricular fi-
brillation. Kauhanen et al.
[25] suggested that ‘heavy
acute intake of beer may involve acute triggers of severe
pathophysiological events in the myocardium or the cor-
onary arteries, or both. These events might include ar-
rhythmia, ischaemia, and possibly thrombotic processes’.
Binge drinking has been shown to increase the risk of
calcification of blood vessels
[26] as well as that of major
coronary events, independent of previous history of CHD
[25, 27–29] .
A cardioprotective effect of light to moderate drinking
has been shown when compared with lifetime abstainers
as a control group in regular alcohol use
[23, 28, 30–32] ,
as well as shortly before CHD events
[20, 21] . Based on
these epidemiological findings, this study tested the hy-
pothesis that heavy drinking occasions (binge drinking)
increase the risk of subsequent AMI, whereas light to
moderate drinking occasions decrease the risk.
M e t h o d s
Study Population and Data Collection
Patients were recruited in three major hospitals in the Ger-
man-speaking part of Switzerland – a university hospital, a coun-
ty hospital and a city hospital. Interviews based on a standardized
questionnaire were conducted in the hospitals by trained staff of
the Research Institute or by trained hospital staff. Interview train-
ing included test interviews with staff of the research institute as
well as supervised interviews in the hospitals. Through collabora-
tion with the staff it was also possible to track all 673 patients who
were hospitalized for first time AMI (ICD I21; for detailed defini-
tion see below) in the study hospitals between July 26, 2005 and
November 15, 2006. We used the following definition of AMI
which is compatible with the national register for Acute Myocar-
dial Infarction in Switzerland (AMIS-plus database): symptoms
and/or ECG changes compatible with acute coronary syndrome
and with the enzymes creatine kinase or creatine kinase-muscle
brain at least twice the upper limits of normal.
256 interviews were conducted, 6 of which could not be in-
cluded in the analysis as essential data for testing the main hy-
potheses (date and time of AMI) were missing. As a result, the
study group consisted of 250 patients. The major reason why not
all 673 patients were eligible for interviews was the early reloca-
tion of 239 patients (35.5%) within 24 h to another hospital, main-
ly because of regional responsibility and bed availability. These
patients were, by definition, not eligible for our study. We ar-
ranged with the hospitals that patients were not to be interviewed
on their first day in hospital, in order to not disturb medical ex-
aminations, and in consideration of the patients with recent ma-
jor complications. In some cases the health status of the patient
allowed an earlier timing of the interview; the situation was dis-
cussed with the medical staff, referring to each case individually.
The other exclusion criteria were: no domicile in Switzerland
(tourists) (5 persons, 0.7%); insufficient skills in German lan-
guage (29, 4.3%); mental or physical inability to participate in an
interview (e.g. cerebral damages, dementia, aphasia, hypoxia, pa-
tient too confused) (55 persons, 8.2%) or death after admission
(12, 1.8%), involvement in another interview study (17 persons,
2.5%) or other (e.g. the patient left the hospital without official
discharge) (4 persons, 0.6%). The eligible sample consisted there-
fore of 312 patients. Of those, 56 patients (18%) refused to par-
ticipate in the interview study. As such, the overall response rate
among eligible patients was 82%.
Since we collected the time of the event as well as the time of
the interview, the corresponding time span between AMI and in-
terview could be gathered for each interview. The interviews were
conducted in half of the study population within 38 h after the
AMI. 55% of the patients were referred directly to the study hos-
pital and 45% were transferred from another hospital to the study
hospital. The patients who refused to participate in the interview
did not differ in gender distribution with the patients who par-
ticipated in the interviews, but they were on average 8 years
older.
S t u d y D e s i g n
The case-crossover methodology used in the present study
was introduced by Maclure
[33] in the early 1990s to study tran-
sient effects on the risk of rare acute events. The case-crossover
design basically tries to answer the question: Was this event trig-
Patterns of Alcohol Consumption and
Acute Myocardial Infarction
Eur Addict Res 2009;15:143–149
145
gered by something unusual that happened just before the inci-
dent
[34]?
We used this design to assess the change in risk of AMI during
a ‘hazard period’ after exposure to alcohol and other potential
triggers of AMI onset. An important feature of the case-crossover
design is that the control information for each patient is based on
his or her own past exposure experience. Thus, there is no exter-
nal control group as in the case-control design, but each case
serves as its own control. It is possible to define one or more con-
trol points or time periods, respectively. This design was invented
to circumvent potential control bias by using controls from the
general population or by using other patients as controls
[34] .
The 12 h immediately preceding the AMI were considered a
hazard period. The patient was first asked to report time of infarc-
tion. Alcohol consumption in this hazard period was compared
with the corresponding time period 1 week prior. Patients were
then asked to report what they were doing during the control pe-
riod, if there were unusual events and if they were drinking alco-
hol, what type, the amount and the time of consumption, as exact
as possible.
In addition, we compared the drinking patterns of our patients
with the drinking patterns of the Swiss general population of sim-
ilar sex and age in order to explore whether our sample differed
with regard to alcohol consumption.
Alcohol Consumption
Participants were asked in detail about their drinking behavior:
frequency, including binge drinking, type (e.g. wine, beer, spirits),
and amount on a usual drinking occasion in deciliters (dl).
Drinking Status. Patients who drank at least once a month dur-
ing the 3 months prior to the interview were defined as current
drinkers. Patients who did not drink at least once a month during
the 3 months prior, but drank once a month during some period
of their life, were defined as former drinkers. Patients who drank
alcohol during some period of their life, but always less than once
a month, were classified as occasional drinkers. Patients who nev-
er drank alcohol in their life (apart from sips) were defined as
lifetime abstainers. These categories were used in a landmark
study by Jackson et al.
[21, 35] , who first examined the short-term
influence of alcohol intake on the risk of CHD events via the case-
crossover design.
Volume of Ethanol. Using the amount of alcoholic beverages
in deciliters as well as the type of beverage, standard glasses with
12 g ethanol were calculated corresponding to 1.2 dl wine, 3 dl
beer, 0.38 dl spirits or 0.8 dl fortified wine (e.g. port), respec-
tively.
Heavy Drinking Occasions (Binge Drinking). Definitions of the
term binge drinking vary. The common international definition
of Wechsler and colleagues
[36] characterizes binge drinking as
an episodic, heavy drinking occasion, with 4 or more drinks of
alcohol for women (minimum of 48 g ethanol) and 5 or more
drinks for men (minimum of 60 g ethanol)
[37] . Since the largest
scale survey in Switzerland, The Swiss Health Survey
[38] , defined
heavy drinking occasions as 6 or more drinks for women and 8 or
more drinks for men, we additionally asked the patients about this
standard in order to compare the heavy drinking patterns of our
sample with those of the general population of Switzerland. The
categories regarding the frequency of binge drinking in the prior
year were as follows: never, less than once a month, every month,
every week, (almost) every day
[38] .
Basic Medical Data
Basic medical data were obtained from the medical records of
the patients with informed consent. Information was obtained on
the presence of arterial hypertension (documented pressure
1 140/90), hyperlipidemia, diabetes mellitus and prior angina pec-
toris as well as regular medication before the event. Further data
such as smoking history, weight, height, family history of MI, etc.
was asked about in our questionnaire. Our study was approved by
the responsible cantonal research ethics boards of the participat-
ing hospitals.
Statistical Analysis
Conditional logistic regression was used to estimate the odds
ratios for alcohol consumption in general and binge drinking,
within the time before the AMI compared to the same time span
1 week prior
[39] . Conditional regression tries to answer the ques-
tion of a causal impact of drinking before the event. The sample
size was determined a priori, in order to detect a difference in
proportions of 10% when the proportion of discordant pairs was
expected to be 30% and the method analysis was a McNemar’s test
of equality of paired proportions with a 0.05 two-sided signifi-
cance level. Our sample size of 250 patients corresponded to a
power of 83%.
Logistic regression was carried out to additionally compare
the impact of different social and other characteristics on drink-
ing before the AMI.
Furthermore, we compared the usual drinking patterns of our
sample with the Swiss Health Survey conducted in the year 2002
[38] . This sample comprised of 19,706 individuals older than 15
years and living in Switzerland (Swiss or foreign nationality), who
were interviewed by phone and 16,141 of whom also completed a
written questionnaire. Comparisons between the patterns of al-
cohol consumption between our sample and the general popula-
tion were carried out by first computing comparable rates for dif-
ferent drinking categories, which were directly standardized to
the sex and age distribution of the AMI patients
[40] . The follow-
ing age categories were used: 15–39, 40–49, 50–59, 60–69, 70+.
These rates were then compared using table analysis, resulting in
a likelihood ratio
2
to test for overall differences between sam-
ples and standardized residuals to test for deviation of a cell from
expected value. For the latter, an absolute value of 1 2, correspond-
ing to a 5% significance level, was fixed as criterion for signifi-
cance
[41] .
R e s u l t s
Description of the Sample
In table 1 , sociodemographic characteristics of the
study population are shown. The study group consisted
mainly of men (79.2%) with an average age of 59.7 years
(range 29–88).
In table 2 , the risk profile at admission is presented.
Most AMI occurred in the morning hours, 33.2% (83)
were between 06:
00 and 11: 59 h. The beginning of the
week was overrepresented as time of occurrence of the
event. None of the patients had a cardiac rupture, 52% of
Gerlich/Krämer/Gmel/Maggiorini/
Lüscher/Rickli/Kleger/Rehm
Eur Addict Res 2009;15:143–149
146
the patients were hypertensive and 35.6% had the diag-
nosis of hyperlipidemia.
Of the study population, 74.8% (187) were current
drinkers. 16.4% of the sample reported daily drinking,
with 3 patients (1.2% of the study group) drinking more
than once a day. Among the weekly non-daily drinkers,
most people drank 1–4 times a week. Only 5 patients (2%
of the whole study group) drank 5–6 times a week. Ta-
ble 3 also provides information about the beverage choice
in men and women, mainly wine and beer for men, and
wine for women. 15 out of the 159 current male drinkers
(9.4%) drank 5 or more glasses on a usual drinking occa-
sion. In current female drinkers, 2 out of 28 (7.2%) drank
4 or more glasses normally. About three quarters of the
current drinkers did not change the quantity of drinking
in the last 12 months.
Drinking Patterns in the Hazard and Control Period
Drinking any alcohol in the 12-hour hazard period
increased the risk for the event threefold (OR 3.1; 95% CI
1.4–6.9). 25 individuals were exposed in the hazard pe-
riod and 8 individuals in the control period.
Moderate drinking was defined as drinking up to
24 g of pure alcohol for women and up to 36 g for men.
Drinking moderately in the 12-hour hazard period in-
creased the risk of AMI marginally significant compared
with no alcohol consumption (OR 2.3; 95% CI 0.97–5.2).
18 persons were exposed in terms of moderate drinking
in the hazard period and 8 persons were exposed in the
control period.
Table 1. Characteristics of the study group (n = 250)
Characteristics n
1
(valid %)
Sex
Male 198 (79.2)
Female 52 (20.8)
Age, years
59.7812.4
<50 55 (22.0)
50–59 64 (25.6)
60–69 72 (28.8)
70+ 59 (23.6)
Nationality
Swiss 202 (80.8)
Other 48 (19.2)
Marital status
Never married 23 (9.2)
Divorced 46 (18.4)
Married 163 (65.2)
Widowed 18 (7.2)
Education (highest certificate)
No school leaving certificate 3 (1.2)
Obligatory school (9 years) 32 (13.0)
Vocational education 127 (51.4)
Higher school education (12 years) 7 (2.8)
Higher education
2
78 (31.6)
Employment status
Full time (100%) 105 (42.2)
Part time 20 (8.0)
Pension 105 (42.2)
Other 19 (7.6)
1
Differences to 250 are due to missing values.
2
Upper vocational school, school of applied sciences, uni-
versity.
Table 2. Risk profile at admission (n = 250)
Health anamnesis n
1
(valid %)
Time of AMI
00:00–05:59 h 53 (21.2)
06:00–11:59 h 83 (33.2)
12:00–17:59 h 67 (26.8)
18:00–23:59 h 47 (18.8)
Day of AMI
Monday 40 (16.9)
Tuesday 53 (21.2)
Wednesday 24 (9.6)
Thursday 21 (8.4)
Friday 26 (10.4)
Saturday 31 (12.4)
Sunday 55 (22.0)
Smoking
Never 74 (29.6)
Current smoker 110 (44.0)
Former smoker 66 (26.4)
Medical history
Angina pectoris 12 (4.9)
Hypertension 130 (52.0)
Hyperlipidemia 89 (35.6)
Diabetes mellitus 23 (9.2)
Obesity
2
39 (15.6)
BMI (average)
26.784.1
Family history of MI
3
Yes 29 (11.7)
Medication use before MI (regular use)
Aspirin
4
39 (17.7)
Calcium channel blocker 14 (6.6)
-Blockers
40 (18.6)
ACE inhibitors 26 (12.1)
1
Differences to 250 are due to missing values.
2
Obesity defined as BMI ≥30 kg/m
2
.
3
MI of parents under the age of 60.
4
Aspirin or other inhibitors of platelet aggregation.
Patterns of Alcohol Consumption and
Acute Myocardial Infarction
Eur Addict Res 2009;15:143–149
147
We did not find any significant differences regarding
binge drinking before AMI, using the international defi-
nition of 4 or more for women and 5 or more glasses of
alcohol for men (see Method section). In the 12-hour pe-
riod before the AMI, 3 individuals were exposed in terms
of binge drinking and 1 person was exposed in the con-
trol period (OR 3.0; 95% CI 0.3–28.8).
In a logistic regression analysis, our results were not
influenced significantly by the following parameters:
age, gender, smoking status, family history of AMI, hy-
pertension, hyperlipidemia, diabetes mellitus, prior un-
stable angina pectoris, physical exertion shortly before
the event, e.g. jogging or carrying heavy things, heavy
psychological exposure in the 12 h before AMI, e.g. death
of a family member or a conflict assessed as heavy emo-
tional strain by the patient, or cocaine use in the 12 h
prior to the AMI.
Comparison of Drinking Patterns with the General
Population
Patterns of alcohol consumption varied significantly
between AMI patients and the general population (likeli-
hood ratio
2
: 40.0; d.f. = 3; p ! 0.001). Compared to the
general population, AMI patients had less lifetime ab-
stention (standardized rates: 5.2 vs. 11.8%), were more of-
ten former drinkers (7.6 vs. 4.8%), more often irregular,
i.e. less than weekly, drinkers (29.6 vs. 16.2%) and less of-
ten weekly drinkers (57.6 vs. 67.0%). The standardized
residuals were significant for the categories of abstention
and irregular drinking. In addition, binge drinking oc-
casions were significantly more frequent among AMI pa-
tients: both ‘less than monthly’ (20.7 vs. 10.9%) and
‘monthly or more frequent’ binge drinking occasions
were about twice as prevalent in our study population (6.8
vs. 3.4%). Both rates had significant standardized residu-
als of 4.4 and 2.7, respectively.
Table 3. Alcohol consumption
Drinking characteristics Men (n = 198)
n (valid % of men)
Women (n = 52)
n (valid % of women)
Drinking status (last 3 months)
a. Lifetime abstainers 6 (3.0) 7 (13.5)
b. Former drinkers
1
14 (7.1) 5 (9.6)
c. Less than once a month (occasional drinkers) 19 (9.6) 12 (23.1)
d. Less than weekly, at least once a month 35 (17.7) 8 (15.4)
e. Weekly non-daily drinkers 88 (44.4) 15 (28.8)
f. Daily drinkers 36 (18.2) 5 (9.6)
Kind of beverage usually consumed in the last 3 months
2
Beer 82 (51.6) 8 (28.6)
Wine 136 (85.5) 25 (89.3)
Fortified wine
3
3 (1.9) 1 (3.6)
Spirits 34 (21.4) 7 (25.0)
Amount of glasses at a usual drinking occasion last 3 months
4
Less than 2 glasses 63 (39.6) 17 (60.7)
2–2.9 glasses 36 (22.6) 6 (21.4)
3–3.9 glasses 27 (17.0) 3 (10.7)
4–4.9 glasses 18 (11.3) 1 (3.6)
5 or more glasses 15 (9.4) 1 (3.6)
Compared to the last 12 months, how much did you drink in the last few weeks?
4
More 12 (7.5) 3 (10.7)
Less 27 (17.0) 7 (25.0)
About the same 120 (75.5) 18 (64.3)
1
Drank in the past at least once a month, but not in the last 3 months.
2
More than one answer is possible, so the total exceeds 100%.
3
Drinks with about 15–20% alcohol, e.g. port wine, sherry.
4
% related to categories d, e, f of drinking status (current drinkers).