The Relation of Alcohol Intake to

Coronary

Heart Disease and All^Cause in a

Mortality

Beer ^Drinking Population

Ulrich Keil,1'2 Lloyd E. Chambless,1'3 Angela Doring,2 Birgit Filipiak2 andjutta Stieber2

Epidemiologic studies indicate that light to moderate alcohol
consumption from beer, wine, or spirits is associated with a
reduction in all-cause mortality, owing primarily to a reduced
risk of coronary heart disease (CHD). To find out whether this
protective effect of small to moderate amounts of alcohol could
be confirmed in Germany, where much of the alcohol con-
sumed is taken in the form of beer, we studied the relation
between alcohol and CHD and total mortality in a population
of southern Germany. We conducted a prospective cohort
study firom 1984 to 1992 among 1,071 men and 1,013 women,
age 45-64 years at baseline, from the Augsburg region. Eighty-
seven per cent of men and 56% of women reported drinking
alcohol at baseline. Among drinkers, men had an average
alcohol intake of 42 gm per day, of which 33 gm per day came
from beer. Women who drank had an average alcohol intake of
16 gm per day and derived about half of it from beer and the
other half from wine. During the 8 years of follow-up, 96
deaths (all causes) and 62 incident CHD events (nonfatal and
fatal) occuned in men, and 45 deaths (all causes) occurred in
women. Adjusting for a number of potential confounders, in
men the adjusted hazard rate ratio (HRR) of CHD events for
drinkers as compared with nondrinkers was 0.51 [95% confi?
dence interval (CI) = 0.27-0.95]; this protective effect starts
with the 0.1-19.9 gm per day alcohol category and does not
change much with higher intake. In men, the adjusted total
mortality HRR for drinkers as compared with nondrinkers was
0.59 (95% CI = 0.36-0.97). The total mortality HRRs for the
different alcohol groups compared with nondrinkers show a
U-shaped curve, with the lowest HRR of 0.46 (95% CI =
0.20-0.80) for the 20-39.9 gm per day alcohol group and an
HRR of 1.04 (95% CI = 0.54-2.00) for the >80 gm per day
alcohol group. In women, the total mortality HRR for those
drinking up to 19.9 gm per day as compared with nondrinkers
was 0.46 (95% CI = 0.22-0.96). (Epidemiology 1997;8:150-
156)

Keywords: alcohol, beer-drinking population, L-shaped coronary heart disease curve, U-shaped total mortality curve, cohort
study.

Light to moderate alcohol intake from beer, wine, or of alcohol intake and increased risk for higher levels of
spirits1 is associated with a reduction in all-cause mor? alcohol use.
tality,2"4 owing primarily to a reduced risk of coronary Although the relations between alcohol and CHD
heart disease (CHD).5"12 The relation between alcohol and total mortality have been found in many popula?
intake and incident CHD has been described as L tions and are generally accepted, there is still some
shaped,6 which means that small amounts of alcohol debate on the mechanisms of the protective effect of
have a protective effect that is not diminished when the alcohol on CHD. Alcohol is known to have an effect on
intake is increased. The relation between alcohol intake several important CHD risk factors, such as blood pres?
and overall mortality has been described as U shaped,2,3 sure,913-15 high-density-lipoprotein (HDL)-cholester-
indicating a protective effect of light to moderate levels Qi 8,16-19 low-density-lipoprotein (LDL)-cholesterol,19,20
and fibrinogen21 and clotting factors.18,22,23
Another point of debate is the accurate assessment of
Fromthe 'Instituteof Epidemiology
andSocialMedicine,Universityof Munster, alcohol intake in epidemiologic studies. Accurate mea?
Munster,Germany; 2GSF-Institute
of Epidemiology,
Neuherberg,Germany; and surements are a prerequisite for making reasonable rec?
3Departmentof Biostatistics,
Universityof NorthCarolina,ChapelHill,NC. ommendations concerning levels of alcohol intake that
Addresscorrespondence to: UlrichKeil,Instituteof Epidemiology
and Social offer more protection than harm.9,24 There is also the
Medicine,Universityof Munster,Domagkstrasse3, D-48129Munster,Germany. problem of distinguishing between the effects of alcohol
and those of various accompanying life-style factors such
The studywasfinancedby the GSF-Forschungszentrum
furUmweltund Ge-
sundheitGmbH,Munich,andsupported bygrantsfromthe FederalMinistryof as smoking or physical activity.8
Science,ResearchandTechnology,Bonn.
Education, The alcohol-CHD and alcohol-total mortality associ?
ations have not previously been investigated in Ger?
SubmittedMay7, 1996;finalversionacceptedNovember14, 1996.
many. We studied these relations in a primarily beer-
? 1997by Epidemiology
Resources
Inc. drinking population of Bavaria, using the data of the

150

Lippincott Williams & Wilkins

is collaborating with JSTOR to digitize, preserve, and extend access to
Epidemiology ®
www.jstor.org
Epidemiology March 1997, Volume 8 Number 2 ALCOHOL, CORONARY DISEASE, AND MORTALITY 151

MONICA (Monitoring Trends and Determinants of Determination of Putative Confounding and Effect-
Cardiovascular Diseases) Augsburg cohort study from Modifying Variables
1984 to 1992. This study provides validated alcohol Body mass index (BMI) was calculated as weight (kg)
intake data,25 incident CHD events with high diagnostic per height squared (m2). For determination of body
accuracy,26-29 and total mortality data. weight and height, participants were asked to remove
shoes and heavy clothing. A smoker was defined as
Methods someone who reported smoking one cigarette or more
Design of the First MONICA Augsburg Survey, per day. Blood pressure was measured with the Hawksley
random zero sphygmomanometer35 three times under
1984-1985
standardized conditions.26,36 All results concerning blood
The designs of the multinational World Health Organi?
zation MONICA project26'30-31and the MONICA Augs? pressure values are based on the first and fifth phase of
the Korotkoff sounds and on the calculation of the mean
burg project32,33 have been described in detail elsewhere.
of the second and third blood pressure measurements. If
The main objective of the first MONICA Augsburg
on which the MONICA Augsburg
a participant was on antihypertensive medication, or if
survey, 1984-1985,
he or she had blood pressure values above 160 or 95
cohort study is based, was the investigation of the prev?
alence and distribution of cardiovascular risk factors in mmHg, then the participant was considered hyperten-
sive.
the study area, which comprises the city of Augsburg and
two adjacent counties, covering a population of 532,987 Nonfasting blood samples were drawn under standard?
ized conditions.34 The serum total cholesterol and HDL-
inhabitants, in 1984-34 A two-stage cluster sample of
cholesterol analyses were carried out with an autoana-
5,312 persons of German nationality was drawn from the
lyzer by a clinical laboratory (Zentralklinikum,
population of 282,279 inhabitants age 25-64 years.33,34
The data were gathered through interview and physical Augsburg, Germany). An enzymatic method (CHOD-
examination. A response of 79.3% was achieved; that is, PAP; Boehringer Mannheim, Mannheim, Germany)
was applied; HDL-cholesterol was precipitated with
4,022 of the 5,069 eligible people (5,312 minus those
phosphotungstate/Mg2+. Internal and external quality
who had died after sampling, minus errors in the popu?
control was performed according to the WHO
lation register, etc) participated in the survey.34 The
MONICA Manual.26
cohort study analyses were restricted to those in the
Educational attainment was estimated by recording
45-64 year age range at baseline (1,071 men and 1,013
years of schooling completed; the variable was then
women).
dichotomized into ^10 years and >10 years.15 Sport
activity was estimated by the questions: How often do
Estimation of Alcohol Consumption
you participate in sports in the winter? ... in the sum?
Each subject participating in the 1984-1985 survey was mer? Answers could be (1) regularly, more than 2 hours
asked how much beer, wine, and spirits he or she had
per week; (2)^egularly, 1-2 hours per week; (3) less than
drunk on the previous workday and over the previous 1 hour per week; or (4) no sport activity.15
weekend. The questions on alcohol use were as follows:
1. How much beer, wine, and spirits did you drink the
previous weekend (Saturday and Sunday)? Morbidity and Mortality Follow-up of the Survey
2. How much beer, wine, and spirits did you drink Participants
during the previous workday (or on the previous Thurs- A follow-up of the 1,074 men and 1,013 women, age
day, if Friday was the previous workday)? 45-64 years, who had participated in the 1984-1985
Total alcohol intake was calculated by multiplying
survey was conducted, assessing fatal and nonfatal CHD
weekday consumption by 5 and adding this figure to and total mortality (vital status) over a median period of
weekend consumption. The following conversions were 7.9 years and a maximal period of 8.2 years. We were
then made: able to use the MONICA Augsburg coronary event
1 liter beer = 40 gm alcohol registry to assess fatal and nonfatal CHD events because,
as a population-based registry, it covers the population
1 liter wine = 100 gm alcohol
from which the sample of the 1984-1985 survey was
1 shot spirits (0.02 liter) = 6.2 gm alcohol
drawn. The registry monitors fatal and nonfatal CHD
Finally, an average number of grams of alcohol intake events outside and inside hospitals of the study area.
per day was derived. Detailed descriptions of the procedures and the data
This 7-day recall method was validated against a quality of the MONICA Augsburg coronary event reg?
7-day diet record method in 899 male participants, age istry have been given elsewhere.27,28,32
45-64 years, of the 1984-1985 survey, who had also The WHO MONICA diagnostic categories (derived
completed an additional 7-day diet record and who are from electrocardiographic tracings and enzyme, symp?
part of the cohort study. The recall method produced a tom, and necropsy findings) inciuded as CHD events in
mean value of 35.5 gm per day and the 7-day record this cohort study are (1) definite and possible nonfatal
method a mean value of 34.7 gm per day.25 The Pearson acute myocardial infarction and (2) fatal CHD (com-
correlation coefficient between the two methods was bining definite and possible fatal coronary events and
0.75. fatal cases with insufficient data). Detailed descriptions
152 KEIL ET AL Epidemiology March 1997, Volume 8 Number 2

of the definitions and applications of these diagnostic

categories have been published.26,29'30 We considered a
CHD event as incident if it was the first event during
follow-up in a person reporting no history of acute myo?
cardial infarction in the 1984-1985 survey. We assessed
total mortality by regularly checking the vital status of
all cohort members with the population registries inside
and outside the study area; this procedure guaranteed
that the vital status of cohort members who had moved
out of the study area could also be assessed.
We excluded 3 of the 1,074 men because of missing
alcohol intake data. The remaining 1,071 men provided
7,857 person-years, with a median follow-up time of 7.9 Noalcohol 1-19 20-39 40-59 60-79 80-99 100-119 >= 120
years; 96 deaths accumulated over this time period. Of intake
the 1,071 men, we excluded 43 who had preexisting Alcohol Intake in Grams/Day
CHD at baseline from the incident CHD follow-up,
FIGURE 1. Alcohol intake at baseline by categories.
leaving 1,028 men with 7,594 person-years and a median
follow-up time of 7.9 years; 62 incident CHD events
developed over this time period.
women, we defined only three categories, nondrinkers
There were 1,013 women, none of whom were missing
and those who drank 0.1-19.9 gm per day and >20 gm
alcohol intake data. They provided 7,746 person-years
per day (Table 2).
over a median follow-up time of 7.9 years; 45 deaths
Table 2 shows baseline characteristics of the cohort
occurred over this follow-up period.
for men and women by level of alcohol intake. Mean
There were no missing data for smoking, hyperten?
HDL-cholesterol and total cholesterol increase with
sion, or age; 16 men and 13 women had missing data in
higher levels of alcohol intake in men and women. The
one of the following variables: BMI, serum total choles?
same is true for systolic blood pressure. There is hardly
terol, HDL-cholesterol, education, or sporting activity.
any relation between level of alcohol intake and BMI in
men, whereas female nondrinkers have a larger BMI.
Statistical Methods The relations between alcohol intake and smoking are J
We computed descriptive statistics (means, percentiles, shaped for men and U shaped for women. The percent?
and proportions) for baseline characteristics, by sex and age of men with 10 years of education or less generally is
follow-up event status. We adjusted event rates for po? slightly higher in drinkers than in nondrinkers. The
tential confounders by Poisson regression,37 We used percentage of women in this category is slightly lower in
Cox proportional hazard models38 to give hazard rate drinkers than in nondrinkers.
ratios (HRRs) comparing different levels of alcohol in? Table 3 shows the number of incident CHD events,
take to nondrinkers. We examined assumption of pro- total follow-up years, and the unadjusted CHD inci?
portionality of hazards by fitting the model stratified by dence rates by alcohol intake at baseline for men. The
two alcohol strata, drinkers vs nondrinkers, then plot- incidence rate is more than twice as high for nondrink?
ting the log[? log(survival)] curves to check parallel- ers than for drinkers. The relation can be described as L
ism.39 The curves were close to parallel after 3 years. To shaped, with about the same CHD risk reduction for all
decrease the likelihood that persons with impending alcohol intake groups.
symptomatic CHD had altered their alcohol intake as a Figure 2 depicts the age- and smoking-adjusted CHD
consequence of poor health, we also fit models after incidence rates along with the total mortality rates for
eliminating the first 3 years of follow-up. men by alcohol intake. The CHD incidence rates de?
crease from nondrinkers to those who drink up to 19.9

Results gm per day and remain at about the same low level in the
higher alcohol intake groups (L-shaped relation). The
Figure 1 shows the distribution of alcohol intake for men alcohol-total mortality relation shows a U-shaped pat-
and women. Eighty-seven per cent of men (927/1,071)
and 56% of women (564/1,013) reported drinking alco?
hol at baseline (Figure 1; Table 1). Among drinkers, TABLE 1. Means and Medians of Beer, Wine, Spirits, and
men had an average alcohol intake of 41.8 gm per day; Total Alcohol Intake (gm per Day) for Drinkers Age 45-64
one-half of the men who drank derived at least 94% of Years
their alcohol intake from beer. Women who drank had
an average alcohol intake of 16.4 gm per day and derived
7.5 and 7.9 gm per day of this intake from beer and wine,
respectively (Table 1). On the basis of this distribution,
we defined five alcohol categories for men: nondrinkers
and those who drank 0.1-19.9 gm per day, 20-39.9 gm
per day, 40-79.9 gm per day, and >80 gm per day. For
Epidemiology March 1997, Volume 8 Number 2 ALCOHOL, CORONARY DISEASE, AND MORTALITY 153

TABLE 2. Characteristics of the Cohort at Baseline by Level of Alcohol Intake

tern, with those drinking 20-39.9 gm per day having the similar HRRs for the different alcohol intake groups.
lowest risk and those drinking ^80 gm per day having Fitting separate Cox models adjusting for age; age and
about the same all-cause mortality risk as nondrinkers. smoking; and age, smoking, hypertension, BMI, and
Table 4 provides unadjusted and adjusted CHD event total cholesterol does not materially change the HRRs
hazard rate ratios (HRRs) with 95% confidence inter? for all drinkers vs nondrinkers or for all of the different
vals (CI) for drinkers compared with nondrinkers and alcohol intake groups vs nondrinkers. When HDL-cho-
for the different alcohol intake categories relative to lesterol is added to the model, the HRRs are slightly
nondrinkers. The results come from Cox proportional increased. We also added the variables education and
hazard models, adjusting for various sets of potential sport activity15 to this model, but doing so did not
confounders. The unadjusted CHD HRR for drinkers vs change the respective HRR estimates.
nondrinkers is 0.48 (95% CI = 0.26-0.86), with very A simpler model, adjusting only for age and smoking,
and comparing drinkers with nondrinkers, has an HRR
of 0.46 (95% CI = 0.26-0.84). We used this simpler
TABLE 3. Number of Coronary Heart Disease (CHD)
model to check the proportional hazards assumption,
Incident Events in Men, Total Follow-up Years, and Unad?
justed CHD Incidence Rates by Alcohol Intake plotting log[? log(survival)] curves for each of the two
alcohol intake strata. These curves are approximately
Incident parallel and distinct after 3 years, consistent with the
Fatal and Unadjusted
Total Nonfatal CHD proportional hazards assumption for that period, but are
Alcohol Intake Follow-up CHD Events Incidence closer together for the first 3 years. Since there are only
(gm/day) Years (N = 62) Rates* two events among nondrinkers in the first 3 years, there
are insufficient data to compare the HRRs between
these periods. If the first 3 years of follow-up are ignored,
the HRR comparing drinkers with nondrinkers is 0.37
(95% CI = 0.19-0.74), deviating even further from
* Eventsper 1,000person-years. unity (not shown in Table 4).
Table 5 gives the number of deaths, total follow-up
years, and unadjusted all-cause mortality rates for men
and women by alcohol intake. The relation is U shaped
20 A I? AllCauseMortality
Rates|
?-- CHDIncidence
Rates for both men and women, a finding illustrated for ad?
justed total mortality rates for men in Figure 2 and for
16 total mortality HRRs in Table 6. These HRRs for alco?
hol intake categories relative to nondrinkers and adjust?
ing for various potential confounders show a protective
12
effect of alcohol intake in the three lowest categories of
drinkers, with the risk returning approximately to the
nondrinker level for the group consuming ^80 gm per
day. When the variables total cholesterol, education,
and sport activity were added to this model, the respec?
tive HRR estimates did not change much. Furthermore,
No alcohol <20 20-39 40-79 >80
Intake when we removed 52 men with self-reported diabetes
Alcohol Intake in grams/day mellitus from the cohort, we found that the respective
FIGURE 2* Age- and smoking-adjusted coronary heart HRRs were nearly identical. Table 7 gives the unad?
disease (CHD) incidence rates and all-cause mortality rates justed and adjusted HRRs for women, with the ratio
in men (per 1,000 person-years) by alcohol intake categories. being 0.5 (95% CI = 0.22-0.96) for the 0.1-19.9 gm per
154 KEIL ET AL Epidemiology March 1997, Volume 8 Number 2

TABLE 4. Coronary Event Hazard Rate Ratios (HRRs) in Men (and 95% Confidence Intervals) for Alcohol Intake
Categories Relative to Nondrinkers, from Cox Proportional Hazards Models, Adjusting for Various Sets of Potential
Confounders (Covariates)

* Referentcategory.

day alcohol intake category vs nondrinkers. When the
variables total cholesterol, education, and sport activity
were added to this model, the respective HRR estimates
did not change much. When we removed 40 women
with self-reported diabetes mellitus from the cohort, the
respective HRRs changed from 0.5 to 0.6.
Discussion
There is considerable evidence that light to moderate
drinkers have lower rates of CHD and total mortality
than nondrinkers.2"8 There is also evidence that heavy
drinkers have greater rates of total mortality than ab?
stainers or light to moderate drinkers.2,3 Some uncer-
tainty still remains whether CHD rates increase with
higher levels of alcohol intake.8,40 Some studies also
report that wine22 has a particularly important protective
effect, although the collective epidemiologic evidence
suggests that all alcoholic beverages are similarly protec?
tive.1,8,41
The MONICA Augsburg cohort study represents the
first report on the relation between alcohol and CHD
and total mortality from Germany, the country with the
largest beer consumption in Europe. First, our study
confirms that the protective effect of light to moderate
drinking on CHD in men is seen in a mainly beer-
drinking population. The risk reduction for CFID was

TABLE 5. Number of Deaths, Total Follow-up Years, and Unadjusted Death Rates by Alcohol Intake

Men Women

* Eventsper 1,000person-years.
t Forwomen,rateis 5.1 foralcoholintake^20 gm perday.

TABLE 6. AU-Cause Mortality Hazard Rate Ratios (HRRs) in Men (and 95% Confidence Intervals) for Alcohol Intake
Categories Relative to Nondrinkers, from Cox Proportional Hazards Models, Adjusting for Various Sets of Potential
Confounders (Covariates)
Epidemiology March 1997, Volume 8 Number 2
TABLE 7. All-Cause Mortality Hazard Rate Ratios (HRRs) in Women (and 95% Confidence Intervals) for Alcohol Intake
Categories Relative to Nondrinkers, from Cox Proportional Hazards Models, Adjusting for Various Sets of Potential
Confounders (Covariates)
near 50%, close to what has been reported in recent
reviews.10,11 Furthermore, higher amounts of alcohol did
not increase the risk for CHD. Our study also confirms
an L-shaped pattern.6 Participants with a daily alcohol
intake of ^80 gm per day are rare in our study, so we
cannot draw conclusions on the CHD risk of this high-
intake group. In the model that also adjusts for HDL-
cholesterol (Table 4), we see a decrease in the protective
effect of alcohol in all intake categories; this is to be
expected, because HDL-cholesterol is thought to be a
mediator of the protective effect of alcohol on CHD risk.
Further protective mechanisms such as lowering of fi-
brinogen, platelet aggregation, and other clotting factors
could not be studied, because these variables were not
determined in the 1984-1985
survey.
Intake of small to moderate amounts of alcohol is also
associated with a lower total mortality risk in men and
women.
A strength of this study was our ability to validate the
7-day recall data in 899 men of the cohort with prospec?
tive 7-day diet record data on alcohol intake. The high
correlation between the two methods puts confidence in
our "exposure" data.25 The 7-day diet record data also
show that food habits in terms of fat intake, polyunsat-
urated/saturated fatty acid ratio, and carbohydrate and
protein intake of drinkers and nondrinkers are not ma-
terially different in this male population.
Obviously, small amounts of alcohol have a protective
effect on CHD and total mortality. Nevertheless, a shift
of the alcohol distribution curve to higher values might
invariably increase the percentage of problem drink?
ers41-43 and the prevalence of hypertension,13-15 hemor-
rhagic stroke,44 and noncardiovascular diseases45 in the
population.
Acknowledgments
We thankthe MONICAAugsburgcoronaryevent registerteam underthe
directionof HanneloreLowelfor the identification
of fatalandnonfatalcoro?
naryeventsin the cohort.We alsothankMarianna Chambless forpreparing
tablesandfiguresandCarmenEweforpreparing the manuscript.
References
1. RimmEB,KlatskyA, GrobbeeD, Stampfer MJ.Reviewof moderate
consumption heartdisease:is the effectdueto
andreducedriskof coronary
beer,wine,or spirits?BMJ1996;312:731-736.
ALCOHOL, CORONARY DISEASE, AND MORTALITY 155
2. Poikolainen K.Alcoholandmortality: a review.J Clin Epidemiol1995;48:
455-465.
3. DuffyJ. Alcohol consumptionand all-causemortality.Int J Epidemiol
1995;24:100-105.
4. Doll R, Peto R, Hall E, WheatleyK, GrayR. Mortalityin relationto
consumption of alcohol:13 yearsobservations on maleBritishdoctors.BMJ
1994;309:911-918.
5. JacksonR, ScraggR, BeagleholeR. Alcohol consumptionand risk of
coronaryheartdisease.BMJ1991;303:211-216.
6. MaclureM. Demonstration of deductivemeta-analysis: ethanolintakeand
riskof myocardial infarction.Epidemiol Rev 1993;15:328-351.
7. KlatskyA. Epidemiology of coronaryheartdisease:influenceof alcohol.
AlcoholClin ExpRes 1994;18:88-96.
8. RenaudS, CriquiMH,FarchiG, Veenstra J.Alcoholdrinkingandcoronary
heartdisease.In: VerschurenPM, ed. HealthIssuesRelatedto Alcohol
Consumption. International LifeSciencesInstitute(ILSI).Washington DC
andBrussels: ILSIPress,1993;81-123.
9. KeilU. Alcoholconsumption andits relationto hypertensionandcoronary
heartdisease.Atherosclerosis Rev 1990;21:43-52.
10. MansonJE,TostesonH, RidkerPM,Satterfield S, HebertP,O'ConnorGT,
BuringJE,HennekensCH. The primarypreventionof myocardial infarc?
tion.N EnglJ Med1992;326:1406-1416.
11. Rich-Edwards JW, MansonJE, HennekensCH, BuringJE.The primary
preventionof coronaryheartdiseasein women.N EnglJ Med 1995;332:
1758-1766.
12. RimmEB,GiovannucciEL,WillettWC,ColditzGA, AscherioA, Rosner
B, StampferMJ. Prospectivestudyof alcoholconsumptionand risk of
coronaryheartdiseasein men.Lancet1991;338:464-468.
13. KeilU, SwalesJ, GrobbeeD. Alcoholintakeandits relationto hyperten?
sion.Cardiovasc RiskFactors1993;3:189-200.
14- KeilU, SwalesJD,GrobbeeDE.Alcoholintakeand its relationto hyper?
tension.In: Verschuren PM, ed. HealthIssuesRelatedto AlcoholCon?
LifeSciencesInstitute(ILSI).Washington DC and
sumption.International
Brussels:ILSIPress,1993;17-42.
15. KeilU, Chambless L, FilipiakB, HartelU. Alcoholandbloodpressure and
its interactionwith smokingandotherbehavioural variables:
resultsfrom
the MONICAAugsburg survey1984-1985.J Hypertens1991;9:491-498.
16. GazianoJM,BuringJE,BreslowJL,Goldhaber SZ,RosnerB, Vandenburgh
M, WillettWC,HennekensCH. Moderate alcoholintake,increased levels
of high-density lipoprotein andits subfractions, anddecreased riskof myo?
cardialinfarction. N EnglJ Med1993;329:1829-1834.
17. Suh I, ShatenBJ,CutlerJA, KullerLH.Alcoholuse and mortalityfrom
coronary heartdisease:the roleof highdensitylipoprotein cholesterol.Ann
InternMed1992;116:881-887.
18. VeenstraJ. Effectsof alcoholon blood lipidsand plateletfunction.In:
VeenstraJ, van der Heij D, eds. Alcohol and Cardiovascular Disease.
Wageningen, The Netherlands: PudocPress,1992;65-69.
19. CriquiMH,CowanLD,TyrolerHA, Bangdiwala S, HeissG, WallaceRB,
Cohn R. Lipoproteins as mediatorsfor the effectsof alcoholconsumption
and cigarettesmokingon cardiovascular mortality:resultsfromthe lipid
researchciinicsfollow-upstudy.Am J Epidemiol1987;126:629-637.
20. Hein HO, SuadicaniP, GyntelbergF. Alcohol consumption, serumlow
the densitylipoproteincholesterolconcentration, and riskof ischaemicheart
disease:six yearsfollow-upin the Copenhagen malestudy.BMJ1996;312:
736-741.
21. KrobotK,HenseHW,CremerP, EberleE, KeilU. Determinants of plasma
fibrinogen:relationto bodyweight,waist-to-hip ratio,smoking,alcohol,age,
and sex: resultsfromthe secondMONICAAugsburgSurvey1989-90.
Arterioscler Thromb1992;12:780-788.
alcohol 22. RenaudS, de Lorgeril M. Wine,alcohol,platelets,andthe Frenchparadox
forcoronaryheartdisease.Lancet1992;339:1523-1526.
23. RidkerPH, VaughanDE,StampferMJ,GlynnRJ,HennekensCH. Asso-
156 KEIL ET AL Epidemiology March 1997, Volume 8 Number 2

ciation of moderatealcohol consumptionand plasmaconcentrationof Project,RegionAugsburg, Manualof Operations,Survey.Munich:GSF-

endogenous tissue-typeplasminogenactivator.JAMA1994;272:929-933.
24. AlankoT. An overviewof techniquesandprobiemsin the measurement of
alcoholconsumption. In:SmartR, CappellH, GlaserF,SchmidtW, Sellers
E,eds.Research Advancesin AlcoholandDrugProbiems. vol. 8. NewYork:
Plenum,1984;209-226.
25. SchaefflerV, DoringA, WinklerG, Keil U. Erhebungen der Alkohol-
aufnahme:Vergleichverschiedener Methoden.ErnahrUmsch 1991;38:
490-494.
26. WHOMONICAProject.MONICAManual.Geneva:WorldHealthOr?
ganizationCardiovascular DiseaseUnit, 1990.
27. LowelH, LewisM, HormannA, KeilU. Casefinding,dataqualityaspects
and comparability of myocardialinfarctionregisters:resultsof a South
Germanregisterstudy.J Clin Epidemiol1991;44:249-260.
28. LowelH, HermanB,LewisMA,HoltzH, QuietzschD, Hormann
A, Greiser
E, VoigtG, KeilU. Registration methodsandestimatesof morbidityand
mortalityof acutemyocardial infarction:resultsfromEastandWestGer?
many.Ann Epidemiol1993;3:S69-S78.
29. WHO MONICAProject.Myocardial Infarctionand CoronaryDeathsin
the WorldHealthOrganization MONICAProject:
Procedures,
Registration
EventRates,andCaseFatalityRatesin 38 Populations from21 Countriesin
FourContinents.Circulation1994;90:583-612.
30. WHOMONICAProjectPrincipalInvestigators: The WorldHealthOrga?
nizationMONICAProject(Monitoring TrendsandDeterminants in Car?
diovascularDisease):A MajorInternationalCollaboration.JClinEpidemiol
1988;41:105-114.
31. WHOMONICAProject.A Worldwide Monitoring SystemforCardiovas?
cularDiseases.WorldHealthStat Ann 1989:27-249.
32. Keil U, KoenigW, LowelH, CairnsV, JorcikJ, HormannA, HeissG,
SchomannC. MONICAProject,RegionAugsburg, Manualof Operations,
Myocardial InfarctionRegister.Munich:GSF-Bericht 21, 1985.
33. KeilU, CairnsV, DoringA, HartelU, JorcikJ, PerzS, StieberJ. MONICA
Bericht20, 1985.
34. KeilU, StieberJ, DoringA, Chambless L,HartelU, FilipiakB, HenseHW,
TietzeM, Gostomzyk J. The cardiovascularriskfactorprofilein the study
areaAugsburg: resultsof the firstMONICAsurvey1984/85.Acta Med
ScandSuppl1988;728:119-128.
35. WrightBM,DoreCF.The randomzero-sphygmomanometer. Lancet1970;
1:337-338.
36. KirkendallWM,FeinleibM, FreisED.Recommendations forhumanblood
pressuredeterminationby sphygmomanometers. Hypertension1981,3:
510A-519A.
37. WilcoskyTC, ChamblessLE.A comparison of directadjustment and re?
gressionadjustment of epidemiologicmeasures.J ChronDis 1985;38:849-
856.
38. Kleinbaum DG, KupperLL,MullerKE.AppliedRegression Analysisand
OtherMultivariable Methods.2nded. Boston:PWS-Kent,1988.
39. Kalbfleisch
JD,PrenticeRL.The StatisticalAnalysisof FailureTimeData.
New York:JohnWileyandSons, 1980.
40. MarmotM, Brunner E.Alcoholandcardiovascular disease:the statusofthe
U-shapedcurve.BMJ1991;303:565-568.
41. Stampfer MJ,RimmEB,WalshDC. Commentary: alcohol,the heart,and
publicpolicy.Am J PublicHealth1993;83:801-804.
42. ShaperAG. Editorial: alcohol,the heart,andhealth.Am J PublicHealth
1993;83:799-801.
43. Rose G, Day S. The populationmean predictsthe numberof deviant
individuals.
BMJ1990;301:1031-1034.
44. StampferMJ,ColditzGA, Willett WC, SpeizerFE, HennekensCH. A
prospectivestudyof moderate alcoholconsumption andthe riskof coronary
diseaseandstrokein women.N EnglJ Med1988;319:267-273.
45. FuchsCS, Stampfer MJ,ColditzGA, Giovannucci EL,MansonJE,Kawachi
I, HunterDJ,Hankinson SE,HennekensCH,RosnerB, SpeizerFE,Willett
WC. Alcoholconsumption and mortalityamongwomen.N EnglJ Med
1995;332:1245-1250.