129

HDL, HDL2, and HDL3 Subfractions, and the

Risk of Acute Myocardial Infarction
A Prospective Population Study in Eastern Finnish Men
Jukka T. Salonen, MD, PhD, MPH; Riitta Salonen, MD; Kari Seppanen, MPh;
Rainer Rauramaa, MD, PhD; and Jaakko Tuomilehto, MD, PhD

Background. We investigated the association of cholesterol concentrations in serum high

density lipoprotein (HDL) and its subfractions HDL2 and HDL3 with the risk of acute
myocardial infarction in 1,799 randomly selected men 42, 48, 54, or 60 years old.
Methods and Results. Baseline examinations in the Kuopio Ischaemic Heart Disease Risk
Downloaded from http://circ.ahajournals.org/ by guest on July 8, 2017

Factor Study were done during 1984-1987. In Cox multivariate survival models adjusted for age
and examination year, serum HDL cholesterol of less than 1.09 mmol/l (42 mg/dl) was
associated with a 3.3-fold risk of acute myocardial infarction (95% confidence intervals [CI],
1.7-6.4), serum HDL2, cholesterol of less than 0.65 mmol/1 (25 mg/dl) was associated with a
4.0-fold risk of acute myocardial infarction (95% CI, 1.9-8.3), and serum HDL3 cholesterol of
less than 0.40 mmol/l (15 mg/dl) was associated with a 2.0-fold (95% CI, 1.1-4.0) risk of acute
myocardial infarction. Adjustments for obesity, ischemic heart disease, other cardiovascular
disease, maximal oxygen uptake, systolic blood pressure, antihypertensive medication, serum
low density lipoprotein cholesterol, and triglyceride concentrations reduced the excess risks
associated with serum HDL, IIDL2, and HDL3 cholesterol in the lowest quartiles by 52%, 48%,
and 41%, respectively. Additional adjustments for alcohol consumption, cigarettes smoked
daily, smoking years, and leisure time energy expenditure reduced these excess risks associated
with low HDL, HDL2, and HDL3 cholesterol levels by another 26%, 24% and 21%, respectively.
Conclusions. Our data confirm that both total HDL and HDL2 levels have inverse associations
with the risk of acute myocardial infarction and may thus be protective factors in ischemic
heart disease, whereas the role of HDL3 remains equivocal. (Circulation 1991;84:129-139)

S everal prospective population studies have re-
ported an inverse association between circu-
lating high density lipoprotein (HDL) choles-
terol concentration and risk of ischemic heart disease
(IHD)'-11; in two studies, the inverse relation did not
achieve statistical significance,'2'13 and in two other
studies, no association was found.14"5 Most studies
have reported decreased levels of both HDL2 and
HDL3 cholesterol in patients with acute myocardial
infarction (AMI) and in those with severe coronary
artery disease.'6 The only prospective study of IHD
from which data on HDL2 and HDL3 subfractions
From the Department of Community Health and General
Practice (J.T.S.), University of Kuopio, Kuopio, the Research
Institute of Public Health (R.S.), University of Kuopio, Kuopio,
the Kuopio Research Institute of Exercise Medicine (K.S., R.R.),
Kuopio, and the Department of Epidemiology, National Public
Health Institute of Finland (J.T.), Helsinki, Finland.
Supported by grants from the Finnish Academy and the Ministry
of Education of Finland.
Address for correspondence: Professor Jukka T. Salonen, Uni-
versity of Kuopio, P.O. Box 6, 70211 Kuopio, Finland.
Received June 27, 1990; revision accepted February 19, 1991.
have been reported' observed a stronger inverse
association for HDL2 than for HDL3 cholesterol with
the risk of IHD. In none of the previous studies was
the effect of the random variability in individual HDL
cholesterol values taken into account. Nor have
thepotential confounders of the association between
HDL cholesterol and AMI risk been comprehen-
sively controlled. Because previous findings were
controversial and inconclusive, we investigated the
associations of serum HDL, HDL2, and HDL3 cho-
lesterol levels with the risk of AMI.
Methods
Subjects
The Kuopio Ischaemic Heart Disease Risk Factor
Study (KIHD) is a population study with the purpose
of investigating previously unestablished risk factors
for AMI and carotid atherosclerosis'7 in Eastern
Finnish men, the population with the highest re-
corded incidence of IHD and mortality from IHD.18
The present analysis is based on those who partici-
pated in KIHD between March 1984 and December
 130 Circulation Vol 84, No 1 July 1991
1987. The study sample included 2,155 Eastern Finn-
ish men 42, 48, 54, or 60 years old at the time of
baseline examination. Of these, 1,821 men (84.5%)
participated. Serum cholesterol levels in the main
lipoprotein fractions were assessed for 1,799 men. No
exclusions were made for the present analysis.
Lipid Analysis
The examination protocol and measurements have
been previously described in detail. 19,20 Subjects gave
blood specimens for lipoprotein separation between
8:00 and 10:00 AM on Tuesday, Wednesday, or Thurs-
day after having abstained from ingesting alcohol for
3 days, from smoking for 12 hours, and from eating
for 12 hours. After the subject had rested in the
supine position for 30 minutes, blood was drawn with
Terumo Venoject VT-100PZ vacuum tubes (Terumo
Corp., Tokyo). No tourniquet was used.
For lipoprotein analysis, serum was separated by
centrifugation at 20°C for 10 minutes at 2,000g after
coagulation at room temperature for 1 hour. Samples
Downloaded from http://circ.ahajournals.org/ by guest on July 8, 2017
were stored at 4°C for no more than 2 days before
separation of lipoproteins. All ultracentrifugations
were done at 10°C with a Kontron TGA-65 ultracen-
trifuge (Kontron, Zurich). The main fractions (very
low density lipoprotein [VLDL], low density lipopro-
tein [LDL], and HDL) were separated as described by
Carlson.21 Serum was centrifuged for 16 hours at
115,000g. VLDL was recovered as the top fraction and
HDL was recovered as the supernatant after precipi-
tation of the bottom fraction with dextran sulfate and
magnesium chloride. The cholesterol concentration in
LDL was calculated as the difference between the
bottom and HDL fractions.
The HDL2 and HDL3 subfractions were separated
during a second ultracentrifugal spin at 108,000g for
62 hours against a density of 1.125 g/cm3 as described
by Kirstein and Carlson.22 The top and bottom
fractions were separated by the tube slicing tech-
nique. In this procedure, the HDL1 subfraction, a
negligible component of total HDL, is included in the
HDL2 subfraction. Thus, the HDL subfraction, cal-
culated as the difference between total HDL and
HDL3, consists mainly of HDL2.
The cholesterol contents of all lipoprotein frac-
tions were measured enzymatically (CHOD-PAP
method, Boehringer Mannheim, Mannheim, FRG)
on the day after the last spin. All tests were run in
duplicate. A Seronorm Lipid (Nycomed, Oslo, Nor-
way) control serum sample was included in each daily
batch of cholesterol determinations. The between-
batch coefficient of variation during 1984-1987 was
2.2% for total HDL, 5.2% for LDL, 9.2% for HDL,
10.8% for HDL2, and 14.2% for HDL3 cholesterol
(210 batches). There was no linear trend in the
quarterly control mean values over time. Tracking of
serum HDL cholesterol values over time was studied
annually by taking repeat serum HDL cholesterol
measurements using the same method in random
subsamples from the study cohort of approximately
65 men. Repeat measurements of serum HDL cho-
lesterol were taken of a total of 322 men 1-5 years
after baseline examinations.
Other Risk Factor Measurements
The current number of cigarettes, cigars, and
pipefuls of tobacco smoked daily as well as the
duration of regular smoking (years); history of myo-
cardial infarction, angina pectoris, and other IHD;
and presence of hypertension and use of antihyper-
tensive medication were recorded using a self-admin-
istered questionnaire, which was checked by an in-
terviewer. The participants were reinterviewed
regarding medical history by a physician. A subject
was considered a current smoker if he had ever
smoked on a regular basis or had smoked cigarettes,
cigars, or a pipe within the past 30 days. Subjects who
had not smoked within the past 30 days were defined
as nonsmokers and given a score of 0 for number of
cigarettes, cigars, and pipefuls of tobacco smoked
daily. Smoking years were considered the sum of
years of smoking regardless of when smoking started
and stopped and whether it had occurred continu-
ously or during several periods. Alcohol consumption
was assessed by instructed and interview-checked
4-day food recording.20 Total leisure time energy
expenditure was measured with a modified Taylor
questionnaire concerning the activities within the
previous 12 months.19
Resting blood pressure was measured between
8:00 and 10:00 AM on the first examination day by one
nurse with a random-zero mercury sphygmomanom-
eter. The measuring protocol included, after a supine
rest of 5 minutes, three measurements in the supine
position, one after 1 minute of standing, and two in
the sitting position with 5-minute intervals. The
mean of all six systolic blood pressure values was
used in the present analysis. In the present study,
hypertension was defined as supine systolic blood
pressure (mean of two last measurements) of 160
mm Hg or more, sitting systolic blood pressure (mean
of two measurements) of 165 mm Hg or more, pre-
exercise sitting systolic blood pressure of 170 mm Hg
or more, supine diastolic blood pressure (mean of
two last measurements) of 95 mm Hg or more, sitting
diastolic blood pressure (mean of two measure-
ments) of 100 mm Hg or more, preexercise sitting
diastolic blood pressure of 105 mm Hg or more, or
self-reported history of hypertension or use of cur-
rent antihypertensive medication.
The respiratory gas exchange was measured on a
breath-by-breath basis with the MGC 2001 system
(Medical Graphics Corp., Minneapolis, Minn.) dur-
ing a symptom-limited exercise test. The testing
protocol consisted of a linear increase of work load
by 20 W/min.19 Highest oxygen uptake (average of
8 seconds) during the test was defined as Vo2max.
Men were determined to have IHD if they reported a
history of myocardial infarction, angina pectoris, or
other IHD; reported use of medication for angina

pectoris; or had ischemia during the maximal exer-
cise tolerance test.19 Exercise electrocardiograms
were coded manually by one cardiologist. The criteria
for ischemia were 1) ischemic electrocardiogram,
defined as horizontal or down-sloping ST depression
of 0.5 mm or more or upsloping ST depression of 1.0
mm or more; 2) typical angina pectoris pain leading
to discontinuation of exercise; or 3) maximal heart
rate during exercise of 130 beats/min or less. Diabe-
tes was defined as previous diagnosis of diabetes or
fasting blood glucose of 8.0 mmol/l or more.
Determination of Follow-up Events
A prospective registry for AMI was established in
the province of Kuopio as a part of the World Health
Organization MONICA Project in 1982.23 This reg-
istry collects detailed diagnostic information on all
suspected fatal and nonfatal AMI events among the
population, which includes the present cohort. The
events are then classified according to explicitly de-
fined, uniform diagnostic criteria into the categories
Downloaded from http://circ.ahajournals.org/ by guest on July 8, 2017
of definite AMI, possible AMI, no AMI, and insuf-
ficient data.23 This classification was based on au-
topsy, serial electrocardiographic findings, cardiac
enzymes, symptoms, and history of IHD. About half
of the fatal cases were autopsied. Serial electrocar-
diographic changes were classified according to the
Minnesota Code into five categories. Cardiac en-
zymes were routinely determined each day since
hospitalization of the patient. Aspartate aminotrans-
ferase, lactate dehydrogenase (LD), LD2, creatinine
kinase, and creatinine kinase-MB levels were used,
and the determination of enzymes was standardized
among several hospitals participating in the AMI
Registry Study in Finland. Enzymes were coded as
definite if the highest value of any of the enzymes was
twice the upper limit of the normal range and no
other cause for the elevation was apparent. Hospi-
talized patients were interviewed shortly after hospi-
tal admission. In fatal cases, data on symptoms were
obtained from medicolegal reports and, when neces-
sary, by interviewing relatives. Symptoms were coded
as typical if the chest pain lasted for at least 20
minutes in the absence of noncardiac, nonatheroscle-
rotic causes.
In the present study, fatal and nonfatal events with
the diagnostic category of definite or possible AMI
were regarded as end points. An event was regarded
as a definite AMI if at least one of the following
conditions was met: definite electrocardiographic
changes; typical, atypical, inadequately described
symptoms combined with probable electrocardio-
gram and abnormal enzymes; typical symptoms and
abnormal enzymes; or naked-eye appearance of fresh
AMI and/or recent coronary occlusion found at
necropsy regardless of other findings. Possible non-
fatal AMI was confirmed if there were typical symp-
toms combined with probable electrocardiogram or
equivocal enzymes or atypical or inadequately de-
scribed symptoms combined with probable electro-
cardiogram and equivocal enzymes. Possible AMI
Salonen et al HDL and Eastern Finnish Men 131
among fatal cases required an exclusion of any other
good evidence for another cause of death and either
typical, atypical, or inadequately described symp-
toms; evidence of chronic coronary occlusion, steno-
sis, or old myocardial scarring at necropsy without
typical, atypical, or inadequately described symp-
toms; or a good history of chronic IHD.
The coverage of the registry is checked against the
computerized national hospital discharge and death
certificate registers. We obtained the diagnostic in-
formation and dates of all heart attacks in our study
cohort by record linkage based on the uniform Finn-
ish personal identification code (social security num-
ber). No personal identification codes were missing
in either our study cohort or the heart attack registry
data. There were no losses to follow-up.
Between March 1984 and December 1988, a "def-
inite" or "possible" fatal or nonfatal AMI was regis-
tered in 97 subjects of the present study. In the case
of multiple events, the first one for each subject was
considered to be the endpoint in our present analysis.
The follow-up periods for individual subjects varied
between 1 and 43/4 years.
Statistical Methods
Three dummy (0, 1) variables were constructed
that compared the three lowest (lst, 2nd, and 3rd)
quartiles of serum HDL, HDL2, and HDL3 choles-
terol with the highest (4th) quartile: 1st versus 4th,
2nd versus 4th, and 3rd versus 4th. These were
entered simultaneously into BMDP Cox proportional
hazards models,24,25 separate models for serum HDL,
HDL2, and HDL3 cholesterol dummies, and a joint
model for both HDL2 and HDL3 dummies. Three
different sets of covariates were entered: first, age
and examination year (as three dummy variables,
1985 versus other, 1986 versus other, and 1987 versus
other); second, age, examination year (as above),
body mass index (kg/M2), history of or prevalent IHD
(yes versus no), other cardiovascular disease (yes
versus no), maximal oxygen uptake (ml/kg/min), sys-
tolic blood pressure (mean of six measurements),
antihypertensive medication (yes versus no), serum
LDL cholesterol concentration (mmol/l), and serum
triglyceride concentration (mmol/1); and third, all
variables listed above plus alcohol consumption (g/
day), cigarette pack-years, and total leisure time
energy expenditure (kcal/12 mo).
The goodness of fit of the proportional hazards
models was examined by analyzing changes in the
proportionality of hazards with time and with risk
factor levels. The results indicated that the applica-
tion of the models was appropriate. In addition, the
simplest models (including three dummy variables
for HDL, age, and examination year) were fitted
using events in only the first 21/2 or the last 21/4 years
of follow-up time. According to the deviance test for
homogeneity of coefficients, there were no significant
differences in the coefficients for HDL dummies
between these two models.
 132 Circulation Vol 84, No 1 July 1991

TABLE 1. Mean Age-Adjusted Serum HDL, HDL2, and HDL3 Cholesterol Values in Subgroups According to Their Strongest
Binary Determinants
Serum HDL cholesterol Serum HDL2 cholesterol Serum HDL3 cholesterol
No Yes* pt No Yes p No Yes p
Antihypertensive
medication 1.33 1.16 <0.001 0.88 0.74 <0.001 0.45 0.43 <0.001
(0.0080) (0.0123) (0.0077) (0.0111) (0.0019) (0.0037)
Diabetes or fasting blood
glucose of at least 8.0
mmol/l 1.30 1.15 <0.001 0.86 0.72 <0.001 0.44 0.42 0.009
(0.0070) (0.0300) (0.0070) (0.0260) (0.0020) (0.0070)
History of or prevalent
ischemic heart disease 1.32 1.26 <0.001 0.87 0.82 <0.001 0.45 0.44 0.007
(0.0089) (0.0110) (0.0084) (0.0105) (0.0022) (0.0027)
History of other
cardiovascular disease 1.30 1.23 <0.001 0.86 0.80 0.002 0.45 0.43 0.006
(0.0077) (0.0163) (0.0073) (0.0152) (0.0019) (0.0044)
Current smokert 1.30 1.26 0.011 0.86 0.83 0.023 0.45 0.45 NS
(0.0084) (0.0123) (0.0080) (0.0118) (0.0021) (0.0030)
HDL, high density lipoprotein.
Downloaded from http://circ.ahajournals.org/ by guest on July 8, 2017

*Number of men in the "yes" categories: 395 receiving antihypertensive medication, 90 with diabetes or elevated fasting blood glucose,
777 with history of or prevalent ischemic heart disease, 312 with other cardiovascular disease, and 571 current smokers.
tStatistical significance of the differences between mean values; NS denotes p>0.050.
tEver smoked on a regular basis or smoked cigarettes, cigars, or a pipe within 30 days before the examination.
Values in parentheses indicate SEM.

Risk factor-adjusted relative hazards were esti-
mated as antilogarithms of coefficients for binary (0,
1) independent variables. Their confidence intervals
were estimated based on the assumption of the
asymptomatic normality of estimates. Risk factor-
adjusted population attributable fractions for multi-
categorical risk factors were computed according to
Walter26 using the estimates of relative hazards de-
rived from the Cox models.
The attenuation of the relative hazard estimates
due to random variability over time was corrected by
replacing the baseline quartile-specific baseline se-
rum HDL cholesterol mean values with mean values
in the same groups in repeat measurements of serum
HDL cholesterol during the follow-up period. Prod-
uct-moment correlations between baseline and repeat
values were computed. This procedure corresponds to
the nonparametric method to correct for the regres-
sion dilution bias in relative hazard estimates sug-
gested by MacMahon and coworkers.27 Second, the
"usual" serum HDL cholesterol concentrations were
approximated using the simple imputation method.
The linear regression equation, based on all 322
repeat measurements, was 0.63xbaseline HDL cho-
lesterol+0.39. All tests of significance were two sided.
Results
The mean values and ranges of serum HDL,
HDL2, and HDL3 cholesterol concentrations were
1.29, 0.85, and 0.44 mmol/l and 0.52-3.05, 0.07-2.77,
and 0.13-0.83 mmol/l, respectively. The unadjusted
correlation between serum HDL2 and HDL3 choles-
terol concentrations was 0.138. The strongest corre-
lates of serum HDL, HDL2, and HDL3 cholesterol
are presented in Tables 1 and 2. Men receiving
antihypertensive medication had 12.0% lower serum
HDL cholesterol and 15.9% lower serum HDL2
cholesterol levels than those not receiving antihyper-
tensive medication (Table 1). Fasting plasma insulin
level, antihypertensive medication, alcohol consump-
tion, maximal oxygen uptake, body mass index, and
age were the strongest determinants of serum HDL
and HDL2 cholesterol levels in multivariate regres-
sion models (Table 2). Men with diabetes, IHD, or
other cardiovascular disease also had decreased se-
rum HDL, HDL2, and HDL3 cholesterol levels (Ta-
ble 1), but these conditions had no independent
association with serum HDL or its subfractions in the
multivariate regression analysis. Alcohol consump-
tion, plasma insulin concentration, antihypertensive
medication, maximal oxygen uptake, and years of
smoking were most strongly partially associated with
serum HDL3 cholesterol (Table 2).
The strongest predictors of AMI other than HDL
cholesterol in the present study cohort were current
cigarettes smoked daily, history of IHD, maximal
oxygen uptake (inverse), age, antihypertensive med-
ication, systolic blood pressure, and, if the current
systolic blood pressure was not entered into the
model, years of hypertension. Neither serum LDL
cholesterol, serum total triglycerides, body mass in-
dex, fasting blood glucose, fasting plasma insulin
concentration (mIU/l), nor diabetes (yes versus no)
had a statistically significant association with the risk
of AMI in the present analysis.
Cox models adjusted for age and examination
years compared the three lowest quartiles of HDL
cholesterol and its subfractions (as three dummy
variables) with the highest quartiles. In these models,
HDL cholesterol of less than 1.09 mmol/l (42 mg/dl)
 Salonen et al HDL and Eastern Finnish Men 133

TABLE 2. Main Behavioral and Biological Correlates of Serum HDL, HDL2, and HDL3 Cholesterol Concentrations
Serum HDL cholesterol Serum HDL2 cholesterol Serum HDL3 cholesterol
Unadjusted r Partial P* Unadjusted r Partial 1B Unadjusted r Partial /3
Fasting plasma insulin (mIU/l) -0.22 -0.10 -0.20 -0.08 -0.12 -0.09
Antihypertensive medication
(yes versus no) -0.22 -0.14 -0.20 -0.13 -0.13 -0.08
Alcohol consumption (g/day) 0.13 0.14 0.11 0.13 0.10 0.09
Maximal oxygen uptake (mI/kg/min) 0.22 0.11 0.20 0.10 0.12 0.05
Body mass index (kg/M2) -0.22 -0.12 -0.22 -0.13 ..
Age (yr) -0.02 0.10 -0.01 0.08 ... ...

Smoking years -0.06 -0.07 -0.05 .. .t -0.05 -0.05

Cigarettes, cigars, and pipefuls
smoked daily -0.06 .. .t -0.05 -0.06 -0.03 ...

Plasma fibrinogen (g/l) -0.12 -0.04 -0.12 -0.05 -0.03

Multiple r2 ... 0.130 ... 0.113 ... 0.040
HDL, high density lipoprotein.
*Standardized partial coefficient from step-up least-squares regressions models including all independent variables shown in the order of
entry to the model for serum HDL cholesterol. Correlation and regression coefficients with absolute value of at least 0.06 depart statistically
significantly from 0.
tNot entered in the step-up procedure (,B is too low to enter).
Downloaded from http://circ.ahajournals.org/ by guest on July 8, 2017

was associated with a 3.32-fold risk of AMI, serum
HDL2 cholesterol of less than 0.65 mmol/1 (25 mg/dl)
was associated with a 4.00-fold risk of AMI, and
serum HDL3 cholesterol of less than 0.40 mmol/l (15
mg/dl) was associated with a 2.04-fold risk of AMI
(Table 3). Adjustments for body mass index, diabetes
(yes versus no), and serum triglyceride and LDL
cholesterol concentrations did not influence these
relative hazards. When additional adjustments were
made for body mass index, history of or prevalent
IHD or other cardiovascular disease, maximal oxygen
uptake, systolic blood pressure, antihypertensive

TABLE 3. Excess Risk of Acute Myocardial Infarction in Three Lowest Quartiles of Serum HDL, HDL2, and HDL3 Cholesterol
Concentrations Compared With Highest Quartile
Adjusted for potential
Adjusted for potential confounders, behavioral
Serum cholesterol Adjusted for age and confounders and behavioral determinants of HDL* and the
concentration examination year determinants of HDL* other HDL subfraction
95% 95% 95%
Relative confidence Relative confidence confidence
mmol/l mg/dl hazard intervals hazard intervals Relative hazard intervals
HDL
< 1.07 <41 3.32 1.73-6.37 1.82 0.88-3.75 NA NA
1.07-1.24 41-47 2.00 1.00-3.98 1.17 0.56-2.45 NA NA
1.25-1.47 48-57 1.46 0.71-3.00 1.11 0.53-2.31 NA NA
> 1.47 >57 1.00 ... 1.00 ... NA NA
HDL2
<0.65 <25 4.00 1.93-8.26 2.20 0.98-4.90 2.09 0.95-5.12
0.65-0.80 25-30 2.56 1.19-5.49 1.45 0.64-3.26 1.44 0.63-3.24
0.81-1.01 31-39 1.97 0.89-4.32 1.45 0.64-3.26 1.43 0.61-3.21
>1.01 >39 1.00 ... 1.00 ... 1.00 ...

HDL3
<0.40 < 15 2.04 1.05-3.98 1.48 0.73-2.98 1.44 0.70-2.94
0.40-0.44 15-16 1.61 0.82-3.19 1.47 0.72-2.98 1.32 0.64-2.71
0.45-0.50 17-19 1.44 0.71-2.94 1.35 0.66-2.76 1.31 0.64-2.70
>0.50 >19 1.00 ... 1.00 ... 1.00
HDL, high density lipoprotein; NA, not applicable.
*Age (yr) (NS), examination years (three dummy variables) (NS, NS, p<0.01), body mass index (kg/m2) (NS), history of or prevalent
ischemic heart disease (yes versus no) (p<0.01), other cardiovascular disease (yes versus no) (NS), maximal oxygen uptake (ml/kg/min)
(p<O.OS), systolic blood pressure (mean of six measurements, mm Hg) (NS), antihypertensive medication (yes versus no), serum low density
lipoprotein cholesterol (mmol/l) (NS), serum triglyceride concentration (mmol/l) (NS), alcohol consumption (g/day) (NS), cigarettes
smoked daily (p<0.001), smoking years (NS), and total leisure time energy expenditure in the previous 12 months (kcal).
 134 Circulation Vol 84, No 1 July 1991

TABLE 4. Partial Excess Risks of Acute Myocardial Infarction Associated With Blood Lipids, Smoking,
and Hypertension
Adjusted for potential confounders and
Adjusted for age and examination year* behavioral determinants of HDLt
Relative hazard 95% confidence intervals Relative hazard 95% confidence intervals
Serum HDL cholesterol
(mM and mg/dl)
<1.07 (<41) 2.80 1.44-5.45 1.83 0.91-3.69
1.07-1.24 (41-47) 1.74 0.87-3.50 1.42 0.70-2.87
1.25-1.47 (48-57) 1.39 0.67-2.87 1.08 0.52-2.25
>1.47 (>57) 1.00 ... 1.00
Serum LDL cholesterol
(mM and mg/dl)
>4.77 (>184) 1.09 0.60-1.98 1.17 0.64-2.12
4.00-4.76 (154-183) 1.16 0.63-2.11 1.21 0.67-2.20
3.38-3.99 (130-153) 0.94 0.49-1.81 1.05 0.55-1.99
<3.38 (<130) 1.00 ... 1.00
Current smoking
Yest 1.96 1.30-2.94 1.79 1.18-2.71
No 1.00 ... 1.00
Downloaded from http://circ.ahajournals.org/ by guest on July 8, 2017

...

Hypertension
Yes 1.85 1.20-2.85 1.40 0.89-2.20
No 1.00 ... 1.00 ...

HDL, high density lipoprotein; LDL, low density lipoprotein.

*From Cox models including age, examination year (three dummy variables), and eight terms for risk factors shown
in the table.
tAge, examination year (three dummy variables), body mass index (kg/m2), history of prevalent ischemic heart
disease (yes versus no), other cardiovascular disease (yes versus no), maximal oxygen uptake (ml/kg/min), serum
triglyceride concentration (mmol/l), alcohol consumption (g/day), total leisure time energy expenditure in the previous
12 months (kcal), and eight terms for risk factors shown in the table.
tEver smoked on a regular basis or smoked cigarettes, cigars, or a pipe within 30 days before baseline examination.

medication (yes versus no), and serum LDL choles-
terol or serum triglyceride concentration, the excess
risks (relative hazard minus 1) associated with serum
HDL, HDL2, and HDL3 cholesterol in the lowest
quartile decreased by 52%, 48%, and 41%, respec-
tively. The decrease in excess risk was mainly a result
of history of IHD and Vo2max. Inclusion of diabetes
(yes versus no) in the model did not change the
relative hazards. Additional adjustments for alcohol
consumption (g/day), current cigarettes, cigars, and
pipefuls smoked daily, smoking years, and leisure
time energy expenditure (kcal/yr) reduced these ex-
cess risks an additional 26%, 24%, and 21%, respec-
tively (Table 3).
In a Cox model including in addition to age and
examination years both serum HDL2 and HDL3
cholesterol dummies, the three lowest quartiles of
HDL2 cholesterol were associated with 3.74-fold
(95% CI, 1.80-7.78), 2.48-fold (95% CI, 1.16-5.33),
and 1.97-fold (95% CI, 0.89-4.33) risks of AMI.
When HDL2 was simultaneously included in the
model, the impact of HDL3 cholesterol was reduced.
The respective relative hazards for the three lowest
quartiles of HDL3 cholesterol were 1.66 (95% CI,
0.84-3.27), 1.40 (95% CI, 0.70-2.78), and 1.38 (95%
CI, 0.68-2.80). In a Cox model adjusted for potential
confounders (second model in Table 3), the relative
hazards (with 95% CIs) for the three lowest quartiles
of HDL2 and HDL3 cholesterol, when entered simul-
taneously, were 2.44 (95% CI, 1.12-5.28), 1.60 (95%
CI, 0.72-3.54), and 1.62 (95% CI, 0.73-3.59) for
HDL2 and 1.46 (95% CI, 0.72-2.95), 1.34 (95% CI,
0.66-2.71), and 1.31 (95% CI, 0.64-2.70) for HDL3,
respectively. The respective relative hazards adjusted
for all potential confounders and behavioral deter-
minants of HDL are shown in Table 3.
In a Cox model that includes age, examination
year, serum HDL and LDL cholesterol levels (three
dummy variables each), smoking (within 30 days),
and hypertension, the excess risk in men in the lowest
quartile of serum HDL cholesterol was greater (rel-
ative hazard, 2.80) than that for smoking (relative
hazard, 1.96) and hypertension (relative hazard,
1.85) (Table 4). When history of or prevalent cardio-
vascular disease and maximal oxygen uptake were
added into the model, these excess risks were appre-
ciably reduced (by 18% for smoking and 53% for
hypertension). The inclusion of terms for interactions
between HDL dummies and smoking or between
HDL dummies and hypertension did not affect the
relative hazard estimates for HDL quartiles to a
significant degree. Also, the addition of interaction
terms did not improve the fit of the model, as judged
based on the change in the deviance for the entire
model.

TABLE 5. Mean Serum HDL Cholesterol Levels at Baseline and in Surveys of Subsamples 1-5 Years After Baseline Measurement
Mean concentration in each category (number of subjects with repeat measurements)
At Difference
Baseline quartile (mmol/l) baseline between quartiles
Serum HDL cholesterol
<1.07 0.95 (421)
1.07-1.24 1.16 (466) 0.21
1.25-1.47 1.36 (473) 0.20
>1.48 1.69 (437) 0.33
Range of mean serum HDL
cholesterol 0.74 ...
HDL, high density lipoprotein.
The percentages of AMIs attributable to low HDL
cholesterol were 67% when adjusted for age and
examination years, 35 % when adjusted for the poten-
tial confounding factors listed in the first footnote of
Table 3, and 24% when also adjusted for behavioral
Downloaded from http://circ.ahajournals.org/ by guest on July 8, 2017
determinants of HDL listed in the footnote of Table
3. The- respective population attributable percent-
ages for smoking and hypertension were 23% and
30% with adjustments for age and examination year
and 19% and 17% with additional adjustments for
the potential confounders defined in Table 4.
In the multivariate Cox models adjusted for age
and examination year, both serum HDL (z= -3.61,
p<0.001) and serum HDL2 (z=-3.43,p<0.001) cho-
lesterol concentrations had significant independent
associations with the risk of AMI. The association of
serum HDL3 cholesterol with AMI did not reach
statistical significance in a two-sided test (z=-1.92,
p=0.07). An increase of 1.0 mmol/l in HDL, HDL2,
and HDL3 cholesterol reduced the risk of AMI by
79%, 79%, and 94%, respectively. When adjusted for
age, examination year, maximal oxygen uptake, prev-
alent IHD, other cardiovascular disease, body mass
index, baseline serum LDL cholesterol and triglycer-
ide concentrations, systolic blood pressure, and anti-
hypertensive medication, an increase of 1.0 mmol/l in
baseline HDL cholesterol was associated with a 61%
decline of AMI risk. The respective decline associ-
ated with an increase (mmol/l) in the usual HDL
cholesterol was 76%.
The unadjusted correlations between baseline
measurement and repeat measurements 1-3 years
after baseline measurement were 0.72 (n=130) for
serum HDL cholesterol and 0.57 (n=365) for serum
LDL cholesterol concentrations. The respective cor-
relation coefficients for baseline versus 3-5-year
postbaseline measurements were 0.67 (n=192) and
0.52 (n=481). The baseline quartile-specific mean
serum HDL cholesterol concentrations for measure-
ments at baseline and 1-3 and 3-5 years after
baseline are presented in Table 5. The differences in
serum HDL cholesterol mean values between the
lowest and the highest baseline quartiles were 0.74
mmol/l at baseline and 0.44 mmol/l (41% less) an
average of 1-5 years after baseline (during the
Salonen et al HDL and Eastern Finnish Men 135
Years after baseline
Difference
1-3 3-5 1-5 between quartiles
0.98 (42) 0.97 (50) 0.98 (92)
1.15 (36) 1.11 (56) 1.13 (92) 0.15
1.29 (32) 1.23 (49) 1.26 (81) 0.13
1.49 (20) 1.38 (37) 1.42 (57) 0.16
0.51 0.41 ...
follow-up period of the present study). The average
difference between quartile specific mean values was
0.25 mmol/l for baseline measurements and 0.15
mmol/l (41% less) for measurements 1-5 years after
baseline.
The relative hazards in the quartiles of baseline
serum HDL cholesterol concentration with adjust-
ment for the potential confounding factors listed in
the footnote of Table 3 are presented in Figure 1.
The relative hazards are plotted against quartile-
specific mean values of both baseline HDL choles-
terol and HDL cholesterol during the follow-up. The
latter curve represents the association between se-
rum HDL cholesterol and AMI risk when corrected
for the regression dilution bias.
When adjusted for age, examination year, IHD,
other cardiovascular disease, systolic blood pressure,
antihypertensive medication, and serum LDL choles-
terol, an increment of 1 mg/dl in serum HDL cho-
lesterol was associated with a 2.4% (4.1% with
correction for regression dilution bias) reduction in
AMI risk, and an increment of 10% in HDL choles-
terol was associated with a 10% (corrected, 17%)
reduction in AMI risk. An additional adjustment for
cigarette pack-years decreased the respective (uncor-
rected) decrements of AMI risk to 2.2% and 9%.
Discussion
Earlier, we reported preliminary findings concern-
ing the roles of HDL2 and HDL3 subfractions with
regard to the risk of AMI.28 The present data confirm
that HDL2 subfraction has an inverse association with
the risk of AMI, whereas the role of HDL3 remains
equivocal. The association was stronger for HDL2
than HDL3, and the association of HDL3 with AMI
lost its statistical significance when adjustment was
made for HDL2. The impact of neither total HDL nor
its subfractions was weakened by adjustments for age,
serum LDL cholesterol or triglyceride concentrations,
obesity, or diabetes. Decreased serum HDL2 choles-
terol was found in men who received antihypertensive
medication, had high body mass index or high fasting
plasma insulin, or currently smoked. Elevated serum
HDL3 cholesterol was associated with high alcohol
consumption and high maximal oxygen uptake. Serum
 136 Circulation Vol 84, No 1 July 1991
Relative Hazard of AMI
2.1 0
-*--- Uncorrected
1.9 -
1.7 H
1.5 H
1.3 h
1.1 K
0.9
Downloaded from http://circ.ahajournals.org/ by guest on July 8, 2017
0.9 1 1.1 1.2
S-HDL-Cholesterol (mmol/1)
FIGURE 1. Plot of risk factor-adjusted relative hazard of acute myocardial infarction (AMI) in baseline quartile-specific serum
(S) high density lipoprotein (HDL) cholesterol mean values at baseline (uncorrected) and 1-5 years after baseline (corrected for
regression dilution bias).
HDL3 cholesterol correlated appreciably with alcohol
consumption and weakly with plasma insulin, antihy-
pertensive medication, maximal oxygen uptake, and
duration of smoking.
In the present analysis, a very concervative ap-
proach was taken, and the major behavioral determi-
nants of HDL (smoking, alcohol consumption, and
physical activity) were controlled for. Because HDL
may be a major mediating biological mechanism in
the effect of smoking, alcohol, and exercise on IHD
risk, the analytical approach may also be considered
to represent an overadjustment.
The strong age-adjusted inverse association of
total serum HDL cholesterol with the risk of AMI is
in accord with the observations of earlier prospective
population studies. The relative hazard, an estimator
of relative risk, was 3.3 when the lowest quartile (less
than 42 mg/dl) and the highest quartile (57 mg/dl or
more) were compared. In the Framingham study,
serum HDL cholesterol in the bottom quintile (less
than 35 mg/dl) was associated with a 4.1-fold in-
creased risk factor-adjusted 12-year mortality from
IHD compared with HDL cholesterol of 55 mg/dl or
more in 1,007 men free of IHD who were 50-79 years
old.29 In the same men, both the age-adjusted and the
risk factor-adjusted relative hazards of myocardial
infarction in the lowest quartile (36 mg/dl or less)
was 1.7 (95% CI, 1.0-2.9) compared with the highest
quartile (53 mg/dl or more).30
Low serum HDL cholesterol was also significantly
associated with increased mortality from IHD in the
Livermore Study,' Honolulu Heart Study,2 Framing-
ham Study,3,29'30 Tromso Heart Study,4 Oslo Heart
* Regression-corrected
1.3 1.4 1.5 1.6 1.7
Study,5 Israeli Ischaemic Heart Disease Study,6 Mul-
tiple Risk Factor Intervention Trial,7 Coronary Pri-
mary Prevention Trial,8 Lipid Research Clinics Prev-
alence Mortality Follow-up Study,9'3' Helsinki Heart
Study,10 and British Regional Heart Study.1' Statis-
tically nonsignificant inverse trends were observed in
the Minnesota Businessmen Study12 and the Finnish
cohort of the Seven Countries Study.'3 In the first
analysis of the British Regional Heart Study, a sig-
nificant inverse association was observed between
nonfasting serum HDL cholesterol and incidence of
IHD, but this association was substantially weakened
and lost its statistical significance by adjusting for the
standard coronary risk factors, including non-HDL
cholesterol.32 In a later analysis based on a longer
follow-up, the inverse association continued to be
significant after risk factor adjustment.'1 Men in the
lowest quintile (less than 0.93 mmol/1, or 36 mg/dl)
had a 2.0-fold risk compared with those in the highest
quintile (at least 1.33 mmol/l, or 51 mg/dl). In two
studies, there was no association between HDL and
IHD risk.14"15 Of these two studies, the study from
Gothenburg was based on only 18 case-control pairs,
including several men on lipid-altering medication.
Previous studies of HDL cholesterol and IHD risk
have not taken into account the attenuation of the
relation resulting from random measurement vari-
ability and the regression of HDL cholesterol values
toward the mean over time. If no correction for this
regression dilution bias is applied, the present data
agree with the estimates from four prospective US
studies that a 1-mg/dl (0.026-mmol/1) increment in
HDL cholesterol level was associated with a 2-3%

decrement in IHD risk.9 With the correction, the
respective decrement of AMI risk in our data was
approximately 4%. According to our data, the poten-
tial impact of HDL cholesterol in the prevention of
IHD is impressive. If the association between HDL
and IHD was causal, the elevation of all HDL values
of less than 1.5 mmol/l (58 mg/dl) to above this value
would theoretically prevent between 35% and 67% of
AMIs, depending on which other factors are assumed
to change with HDL cholesterol. In the present study
cohort with a short average follow-up, a low serum
HDL cholesterol concentration appears to be a
stronger risk factor for AMI than smoking, hyperten-
sion, or serum LDL cholesterol concentration. In
addition to the shortness of follow-up, the weakness
of the association between serum LDL cholesterol
and AMI risk could have been a result of the weaker
tracking (greater random variability) of LDL than
HDL cholesterol over time.
Quantitative estimates of the association between
HDL and IHD must be interpreted cautiously be-
Downloaded from http://circ.ahajournals.org/ by guest on July 8, 2017
cause differences can be caused by a number of
factors. In prospective studies, the characteristics of
the subjects, length of follow-up, choice and deter-
mination of follow-up events, blood sampling, treat-
ment of serum or plasma samples, and method of
HDL measurement may be relevant. In the present
study, blood samples were drawn from resting sub-
jects in the supine position after a 12-hour fast and
abstinence from smoking in the morning, and lipo-
protein fractions were separated from unfrozen se-
rum samples, and analyzed within 2 days. Serum
HDL was separated with a combination of ultracen-
trifugation and precipitation, and HDL3 was sepa-
rated with a second ultracentrifugal spin at the site of
blood sampling. Despite the time-consuming analyt-
ical procedure, the measurement variabilities of
HDL, HDL2, and HDL3 cholesterol measurements
were considerably large. Because random variability
in the risk factor measurements attenuates associa-
tions with disease outcomes, the greater measure-
ment variability for HDL3 than for HDL2 cholesterol
may account for a part of or the entire difference in
the strength of their association with AMI risk.
The early Livermore Study is the only prospective
population study that has investigated the association
of HDL subfractions with the risk of IHD.1 In that
study, both HDL2 and HDL3 fractions were signifi-
cantly inversely associated with the risk of an IHD
event, even though the association was stronger for
HDL2 than for HDL3 fraction. In most case-control
studies comparing subjects with and without angio-
graphically assessed coronary artery disease, lower
levels of HDL2 and HDL3 cholesterol have been
observed in cases than in controls.'6 In most studies,
there has been a greater proportionate difference in
HDL2, although there have been some exceptions.33
In addition, in the National Heart, Lung, and Blood
Institute Type II Intervention Study,34 the extent of
coronary atherosclerosis associated with both HDL
subfractions but significantly with only HDL3. The
Salonen et al HDL and Eastern Finnish Men 137
question, however, of whether only HDL2 or HDL3
subfraction is protective of IHD may be irrelevant
because both fractions participate in the reverse
cholesterol transport.35
Theoretically, serum HDL cholesterol could be an
indicator of another factor rather than a true protec-
tive factor per se. First, it reflects the levels of cigarette
smoking, physical activity, and alcohol consumption
and is associated with obesity, diabetes, and hyperten-
sion. The inverse association of serum HDL and
HDL2 cholesterol with the risk of AMI persisted in
our data, however, even after adjustments for all of
these factors in a multivariate analysis. Also, in the
Lipid Research Clinics Follow-up Study, the associa-
tion of cigarette smoking with cardiovascular disease
mortality was independent of HDL cholesterol, and
the inverse relation between alcohol consumption and
cardiovascular disease mortality was only partially
mediated through HDL cholesterol.31
It is also possible that decreased serum HDL is an
indicator of another metabolic disorder that could be
the true etiologic factor. Reduced HDL could reflect
inefficient catabolism of triglyceride-rich lipoproteins
and be associated with chylomicron remnants. Low
HDL cholesterol is associated with decreased lipopro-
tein lipase activity and increased hepatic lipase activity.
Low serum HDL cholesterol concentration can be a
consequence of either a decreased number of HDL
particles in the circulation or a reduced size of HDL
particles resulting from cholesterol depletion. These
two conditions may not have the same etiologic rele-
vance. The absence of accelerated atherogenesis in
patients with low or no HDL35-37 speaks against a
direct causal role of HDL in atherogenesis. In addition,
coronary artery disease has been reported in persons
with familial hyper-HDL2-cholesterolemia.38
In our study cohort with a very high incidence of
AMI and a short average follow-up, serum HDL
cholesterol concentration was one of the strongest
predictors of AMI. The lack of the predictive power
of serum LDL cholesterol may be a result of the short
follow-up because LDL conceivably increases the
risk of AMI through promotion of the progression of
coronary atherosclerosis as well as a result of large
random variability over time. Serum LDL cholesterol
was the strongest determinant of the progression of
carotid atherosclerosis in a 2-year follow-up study of
a subsample of the present subjects with B-mode
ultrasonography, whereas serum HDL cholesterol
had no association with atherosclerosis progression
in this small cohort.39 An identical finding was ob-
served in a 2-year coronary atherosclerosis progres-
sion study with repeat angiograms.40 On the other
hand, in two other angiography studies,34,41 HDL
cholesterol had an inverse relation with atheroscle-
rosis progression. These observations provide some
support for the hypothesis of differential pathoge-
netic mechanisms for the effects of serum LDL and
HDL cholesterol in IHD.
Because the protective impact of serum HDL
cholesterol concentration becomes detectable in such
 138 Circulation Vol 84, No 1 July 1991
a short follow-up, the effect of HDL may be mediated
through pathways other than atherogenesis; the re-
verse cholesterol transport from arteries to the liver
may not be the only antiatherogenic mechanism of
HDL. There are other possible alternative or parallel
mechanisms for the protective effect of HDL. First,
HDL may act as an antioxidant; that may explain why
it is associated in cell culture with increased prosta-
cyclin production42 and reduced platelet aggregabil-
ity.43 As an antioxidant, HDL inhibits the oxidation
of cholesterol and the formation of lipid peroxides,
which are cytotoxic for the endothelium44 and inhibit
prostaglandin synthesis at certain levels.45 Also, apo-
lipoprotein A-I has been observed to stabilize se-
creted prostacyclin.46 Second, HDL has been found
to promote fibrinolysis.47 Third, high levels of HDL
may reduce the uptake of LDL by endothelial cells by
competing for the LDL receptor.48 Fourth, HDL may
prevent the formation of LDL aggregates, which
appear to facilitate the influx of LDL to the endo-
thelial cells.49 Finally, HDL counteracts in vitro the
Downloaded from http://circ.ahajournals.org/ by guest on July 8, 2017
activation of platelets by LDL.50 Apolipoprotein A-I
has been found to increase the solubility of choles-
terol in bile.51
The role of HDL as an antioxidant and its impact
on platelet function and thrombogenesis deserve
thorough studies. The effect of HDL and its subfrac-
tions on atherosclerosis progression must be investi-
gated in large population cohorts with repetitive
assessments of atherosclerosis. In conclusion, the
present data provide confirmation of the inverse
association between serum HDL and HDL2 choles-
terol and the risk of IHD, whereas the role of HDL3
remains only suggestive. Our findings are consistent
with the concept that high HDL blood levels or
associated metabolic factors are protective for IHD.
Acknowledgments
We are grateful to Professor Kalevi Pyorala and
Dr. Pertti Palomaki, Kuopio University Hospital, for
providing access to data of the AMI Registry in the
province of Kuopio, to Dr. Esko Taskinen and Dr.
Juha Venalainen for supervising the maximal exer-
cise tests, and to Dr. Jaakko Eranen for coding the
exercise electrocardiograms.
References
1. Gofman JW, Young W, Tandy R: Ischemic heart disease,
atherosclerosis, and longevity. Circulation 1966;34:679-697
2. Rhoads GG, Gullbrandsen CL, Kagan A: Serum lipoproteins
and coronary heart disease in a population study of Hawaiian-
Japanese men. N Engl J Med 1976;294:293-298
3. Gordon T, Castelli WP, Hjortland MC, Kannel WB, Dawber
TR: High-density lipoprotein as a protective factor against
coronary heart disease: The Framingham Heart Study. Am J
Med 1977;62:707-714
4. Miller NE, Thelle DS, Forde OH, Mjos OD: The Tromso
Heart Study: High-density lipoprotein and coronary heart
disease: A case-control study. Lancet 1977;1:965-968
5. Enger SC, Hjermann I, Foss OP, Helgeland A, Holme I, Leren
P, Norum KR: High-density lipoprotein cholesterol and myo-
cardial infarction and sudden coronary death: A prospective
case-control study in middle-aged men of the Oslo Study.
Atery 1979;5:170-181
6. Goldbourt U, Holtzman E, Neufeld HN: Total and high
density lipoprotein cholesterol in the serum and risk of
mortality: Evidence of a threshold effect. Br Med J 1985;290:
1239-1243
7. Multiple Risk Factor Intervention Trial Research Group:
Relationship between baseline risk factors and coronary heart
disease and total mortality in the Multiple Risk Factor Inter-
vention Trial. Prev Med 1986;15:254-273
8. Gordon DJ, Knoke J, Probstfield JL, Superko R, Tyroler HA,
for the Lipid Research Clinics Program: High density lipopro-
tein cholesterol and coronary heart disease in hypercholes-
terolemic men: The Lipid Research Clinics Coronary Primary
Prevention Trial. Circulation 1986;74:1217-1225
9. Gordon DJ, Probstfield JL, Carrison RJ, Neaton JD, Castelli
WP, Knoke JD, Jacobs DR Jr, Bangdiwala S, Tyroler HA:
High-density lipoprotein cholesterol and cardiovascular dis-
ease: Four prospective American studies. Circulation 1989;79:
8-15
10. Manninen V, Huttunen JK, Heinonen OP, Tenkanen L, Frick
H: Relation between baseline lipid and lipoprotein values and
the incidence of coronary heart disease in the Helsinki Heart
Study. Am J Cardiol 1989;63:H42-H47
11. Pocock SJ, Shaper AG, Phillips AN: Concentrations of high
density lipoprotein cholesterol, triglycerides, and total cho-
lesterol in ischaemic heart disease. Br Med J [Clin Res]
1989;298:998-1002
12. Keys A: Alpha lipoprotein (HDL) cholesterol in the serum
and the risk of coronary heart disease and death. Lancet
1980;2:603-606
13. Keys A, Karvonen MJ, Punsar S, Menotti A, Fidanza F, Farchi
G: HDL serum cholesterol and 24-year mortality of men in
Finland. Int J Epidemiol 1984;134:428-435
14. Wiklund 0, Wilhelmsen L, Elmfeldt D, Wedel H, Valek J,
Gustafson A: Alpha-lipoprotein cholesterol concentration in
relation to subsequent myocardial infarction in hypercholes-
terolemic men. Atherosclerosis 1980;37:47-53
15. Levy RI, Klimov AN: High density lipoprotein cholesterol and
mortality in USSR and US middle-aged men: The Collabora-
tive US/USSR Mortality Follow-up Study (abstract). Circula-
tion 1987;76(suppl IV):IV-67
16. Miller NE: Associations of high density lipoprotein subclasses
and apolipoproteins with ischemic heart disease and coronary
atherosclerosis. Am Heart J 1987;113:589- 597
17. Salonen JT: Is there a continuing need for longitudinal
epidemiologic research? The Kuopio Ischaemic Heart Disease
Risk Factor Study. Ann Clin Res 1988;20:46-50
18. Keys A: Seven countries: A Multivariate Analysis of Death and
Coronary Heart Disease. Cambridge, Mass/London, Harvard
University Press, 1980
19. Salonen JT, Lakka T: Assessment of physical activity in
population studies-Validity and consistency of the methods
in the Kuopio Ischaemic Heart Disease Risk Factor Study.
Scand J Sports Sci 1987;9:89-95
20. Ihanainen M, Salonen R, Seppanen R, Salonen JT: Nutrition
data collection in the Kuopio Ischaemic Heart Disease Risk
Factory Study: Nutrient intake of middle-aged Eastern Finn-
ish men. Nutr Res 1989;9:597-604
21. Carlson K: Lipoprotein fractionation. J Clin Pathol 1973;
26(suppl 5):32-37
22. Kirstein P, Carlson K: Determination of the cholesterol con-
tent of high density lipoprotein subfractions HDL2 and HDL3,
without contamination of Lp(a), in human plasma. Clin Chim
Acta 1981;113:123-134
23. Tuomilehto J, Kuulasmaa K: WHO Monica Project: Assessing
CHD mortality and morbidity. Int J Epidemiol 1989;18(suppl
1):S38-S45
24. Cox DR: Regression models and life-tables. J Royal Stat Soc
1972;34:187-201
25. Dixon WJ, Brown MB, Engelman L, Frane JW, Hill NA,
Jennrich RI, Toporek JD: BMDP Statistical Software (1985
printing with additions). Berkeley, Calif, University of Califor-
nia Press, 1985
26. Walter SD: Estimation and interpretation of attributable risk
in health research. Biometrics 1976;32:829-849

27. MacMahon S, Peto R, Cutler J, Collins R, Sorlie P, Neaton J,
Abbott R, Godwin J, Dyer A, Stamler J: Blood pressure,
stroke, and coronary heart disease: Part 1. Prolonged differ-
ences in blood pressure: Prospective observational studies
corrected for the regression dilution bias. Lancet 1990;1:
765-774
28. Salonen JT, Seppanen K, Rauramaa R: Serum high density
lipoprotein cholesterol subfractions and the risk of acute
myocardial infarction: A population study in Eastern Finland
(abstract). Circulation 1988;78(suppl II):Il-281
29. Wilson PWF, Abbott RD, Castelli WP: High density choles-
terol and mortality: The Framingham Heart Study. Arterioscle-
rosis 1988;8:737-741
30. Abbott RD, Wilson PWF, Kannel WB, Castelli WP: High
density lipoprotein cholesterol, total cholesterol screening,
and myocardial infarction: The Framingham Study. Arteio-
sclerosis 1988;8:207-211
31. Criqui MH, Cowan LD, Tyroler HA, Bangdiwala S, Heiss G,
Wallace RB, Cohn R: Lipoproteins as mediators for the
effects of alcohol consumption and cigarette smoking on
cardiovascular mortality: Results from the Lipid Research
Clinics Follow-up Study. Am J Epidemiol 1987;126:629-637
32. Pocock SJ, Shaper AG, Philips AN, Walker M, Whitehead TP:
High density lipoprotein is not a major risk factor for
ischaemic heart disease in British men. Br Med J 1986;292:
Downloaded from http://circ.ahajournals.org/ by guest on July 8, 2017
515-519
33. Kempen HJ, van Gent CM, Buytenhek R, Buis B: Association
of cholesterol concentrations in low-density lipoprotein, high-
density lipoprotein, and high-density lipoprotein subfractions,
and of apolipoproteins AI and All, with coronary stenosis and
left ventricular function. J Lab Clin Med 1987;109:19-26
34. Brensike JF, Levy RI, Kelsey SF, et al: Effects of therapy with
cholestyramine on progression of coronary arteriosclerosis:
Results of the NHLBI Type II Coronary Intervention Study.
JAMA4 1984;69:313-324
35. Oram JF, Albers JJ, Cheung MC, Bierman EL: The effects of
subfractions of high density lipoprotein on cholesterol efflux
from cultured fibroblasts. J Biol Chem 1981;256:8348-8356
36. Schaeffer EJ, Zech LA, Schwartz DE, Brewer HB Jr: Coro-
nary heart disease prevalence and other clinical features in
familial high-density lipoprotein deficiency (Tangier disease).
Ann Intern Med 1980;93:261-266
37. Franceschini G, Sirtori CR, Capurso A, Weisgraber KH,
Mahley RW: The AAMilano apoprotein: I. Decreased high
density lipoprotein cholesterol levels with significant lipopro-
tein modifications and without clinical atherosclerosis in an
Italian family. J Clin Invest 1980;66:892-900
38. Matsuzawa Y, Yamashita S, Funahashi T, Yamamoto A, Tarui
S: Selective reduction of cholesterol in HDL2 fraction by
probucol in familial hypercholesterolemia and hyper-HDL2-
cholesterolemia with abnormal cholesteryl ester transfer. Am J
Cardiol 1988;62:66B-72B
Salonen et al HDL and Eastern Finnish Men 139
39. Salonen R, Salonen JT: Progression of carotid atherosclerosis
and its determinants: A population-based ultrasonography
study. Atherosclerosis 1990;81:33-40
40. Blankenhorn DH, Alaupovic P, Wickham E, Chin HP, Azen
SP: Prediction of angiographic change in native human coro-
nary arteries and aortocoronary bypass grafts: Lipid and
nonlipid factors. Circulation 1990;81:470-476
41. Arntzenius AC, Kromhout D, Bart JD, et al: Diet, lipoproteins
and the progression of coronary atherosclerosis: The Leiden
Intervention Trial. N Engl J Med 1985;312:805-811
42. Spector AA, Scanu AM, Kaduce TL, Figard PH, Fless GM,
Czervionke RL: Effect of human plasma lipoproteins on
prostacyclin production by cultured endothelial cells. J Lipid
Res 1985;26:288-297
43. Desai K, Hutton RA, Bruckdorfer KR, Owen JS: Aggregation
of isolated platelets is inhibited by the plasma high-density
lipoprotein subclass, HDL1. Biochem Soc Trans 1986;14:
723-724
44. Steinberg D, Parthasarathy S, Carew TE, Khoo JC, Witztum
JL: Modifications of low-density lipoprotein that increase its
atherogenicity. N Engl J Med 1989;320:915-923
45. Salmon JA, Smith DR, Flower RJ, Moncada S, Vane JR:
Further studies on the enzymatic conversion of prostaglandin
endoperoxide into prostacyclin by porcine aorta microsomes.
Biochim Biophys Acta 1978;523:250-262
46. Yoi Y, Aoyama T, Morishita H, Takahashi M, Takatsu Y,
Kawai C: Serum prostacyclin stabilizing factor is identical to
apolipoprotein A-I (Apo A-I): A novel function of Apo A-I. J
Clin Invest 1988;82:803-807
47. Saku K, Ahmad M, Glas-Greenwalt P, Kashyap ML: Activa-
tion of fibrinolysis by apolipoproteins of high density lipopro-
teins in man. Thromb Res 1985;39:1-8
48. Miller NE, Weinstein DB, Carew TE, Koschinsky T, Steinberg
D: Interaction between high density and low density lipopro-
teins during uptake and degradation by cultured human
fibroblasts. J Clin Invest 1977;60:78-89
49. Tertov VV, Sobenin IA, Gabbasov ZA, Popov EG, Orekhov
AN: Lipoprotein aggregation as an essential condition of
intracellular lipid accumulation caused by modified low den-
sity lipoproteins. Biochem Biophys Res Comm 1989;163:
489-494
50. Knorr M, Locher R, Vogl E, Vetter N, Block LH, Ferracin F,
Lefkovits H, Pletscher A: Rapid activation of human platelets
by low concentrations of low-density lipoprotein via phosphati-
dylinositol cycle. Eur J Biochem 1988;172:753-759
51. Kibe A, Holzbach RT, LaRusso NF, Mao SJT: Inhibition of
cholesterol crystal formation by apolipoproteins in supersatu-
rated model bile. Science 1984;255:514-516
KEY WORDS * clinical trials * high density lipoprotein e
acute myocardial infarction
 HDL, HDL2, and HDL3 subfractions, and the risk of acute myocardial infarction. A
prospective population study in eastern Finnish men.
J T Salonen, R Salonen, K Seppänen, R Rauramaa and J Tuomilehto

Circulation. 1991;84:129-139
Downloaded from http://circ.ahajournals.org/ by guest on July 8, 2017

doi: 10.1161/01.CIR.84.1.129
Circulation is published by the American Heart Association, 7272 Greenville Avenue, Dallas, TX 75231
Copyright © 1991 American Heart Association, Inc. All rights reserved.
Print ISSN: 0009-7322. Online ISSN: 1524-4539

The online version of this article, along with updated information and services, is located on
the World Wide Web at:
http://circ.ahajournals.org/content/84/1/129

Permissions: Requests for permissions to reproduce figures, tables, or portions of articles originally
published in Circulation can be obtained via RightsLink, a service of the Copyright Clearance Center, not the
Editorial Office. Once the online version of the published article for which permission is being requested is
located, click Request Permissions in the middle column of the Web page under Services. Further
information about this process is available in the Permissions and Rights Question and Answer document.

Reprints: Information about reprints can be found online at:

http://www.lww.com/reprints

Subscriptions: Information about subscribing to Circulation is online at:

http://circ.ahajournals.org//subscriptions/