37
Joint Effects of Serum Triglyceride and LDL
Cholesterol and HDL Cholesterol
Concentrations on Coronary Heart Disease
Risk in the Helsinki Heart Study
Implications for Treatment
Vesa Manninen, MD; Leena Tenkanen, MSci; Pekka Koskinen, MD; Jussi K. Huttunen, MD;
Matti Manttari, MD; Olli P. Heinonen, MD, DSci; and M. Heikki Frick, MD
Downloaded from http://circ.ahajournals.org/ by guest on June 14, 2017
Background. We studied the joint effect of baseline triglyceride and lipoprotein cholesterol
levels on the incidence of cardiac end points in the trial group (n=4,081) of the Helsinki Heart
Study, a 5-year randomized coronary primary prevention trial among dyslipidemic middleaged men. The relative risks (RR) were calculated using Cox proportional hazards models with
a dummy variable technique that allows simultaneous study of subgroup combinations from the
placebo and treatment groups.
Medwxs and Results. In the placebo group (n=2,045), the low density lipoprotein cholesterol
(LDL-C)/high density lipoprotein cholesterol (HDL-C) ratio was the best single predictor ofcardiac
events. This ratio in combination with the serum triglyceride level revealed a high-risk subgroup:
subjects with LDL-C/HDL-C ratio >5 and triglycerides >2.3 mmol/1 had a RR of 3.8 (95% CI,
2.2-6.6) compared with those with LDL-C/HI)L-C ratio <5 and triglyceride concentration c2.3
mmol/1. In subjects with triglyceride concentration >2.3 mmoUll and LDL-C/H1)L-C ratio <5, RR
was close to unity (1.1), whereas in those with triglyceride level s2.3 mmoVIl and LDL-C/HDL-C
ratio >5, RR was 1.2. The high-risk group with LDL-C/HDL-C ratio >5 and triglyceride level >2.3
mmol/1 profited most from treatment with gemfibrozil, with a 71% lower incidence of coronary heart
disease events than the corresponding placebo subgroup. In all other subgroups, the reduction in
CHD incidence was substantially smaller.
Conclusions. Serum triglyceride concentration has prognostic value, both for assessing
coronary heart disease risk and in predicting the effect of gemfibrozil treatment, especially
when used in combination with HDL-C and LDL-C. (Circulation 1992;85:37-45)
A
positive association between serum triglyceride concentration and risk of coronary heart
disease (CHD) has been observed in most
case-control studies and in a number of prospective
studies.1-5 However, this association has often disappeared when adjustment has been made for other
risk factors, particularly the level of high density
lipoprotein cholesterol (HDL-C),5-8 and it has thus
been suggested that serum triglycerides do not have a
causal role in atherosclerosis.9 This conclusion has
From the First Department of Medicine, Helsinki University
Central Hospital, and Department of Public Health, University of
Helsinki, and the National Public Health Institute, Helsinki,
Finland.
Address for correspondence: Vesa Manninen, MD, First Department of Medicine, Helsinki University Central Hospital,
Haartmaninkatu 4, SF-00290 Helsinki, Finland.
Received January 9, 1991; revision accepted September 3, 1991.
recently been called into question for several reasons.
First, due to multicollinearity, it is not appropriate to
adjust for HDL-C when studying the association
between serum triglycerides and CHD.10 Second, an
independent association between triglycerides and
CHD may have no biological meaning, as the metabolisms of triglyceride-rich very low density lipoprotein (VLDL), and high density lipoprotein (HDL)
See p 365
and low density lipoprotein (LDL) are closely interrelated.6 Some of the analytical difficulties can be
surmounted by studying the joint effects of triglycerides with HDL-C and low density lipoprotein cholesterol (LDL-C) on the risk of CHD1' and by introducing their interaction into the statistical modeling.12
The present contribution is based on further analyses of the Helsinki Heart Study data.13-15 We have
38
Circulation Vol 85, No 1 January 1992
Incidence of cardiac events
(per 1000 person-years)
Incidence of cardiac events
(per 1000 person-years)
Incdence of cardiac events
(per 1000 person-years)
16
16
16
Total Cholester
LDL CholestercD]
14
14
14
12
12
12
10 I
10
10
8
8
8
6
6
6
4
4
4
2
2
2
42
0
1
TOT. CHOL . 7.8
Incidence of cardiac events
(per 1000 person-years)
Downloaded from http://circ.ahajournals.org/ by guest on June 14, 2017
16
-
HDL
Cholestern
14
12
0
TOT. CHO
-
24
32
WL.S5
LD:> 5
mnolfl
Triglycerdes
I|
0
TG. 3
Incidence of cardiac events
(per 1000 person-years)
Inddence of cardiac events
16
16
-
Total-C/HDL-C.f
14
12
12
-
10
-
-
10
8
8
8
6
6
6
4
4
4-
2
2
2-
0 1
342
HDL .1.06
HDL < 1.08
24
TOTIHDL < 6
132
TOT/DL > 6
0
LDL-C/HDL-C
-
14
I
mmn~
(per 1000 person-years)
10 |
0
TG>2.3
3
138
LDUHDL < 5
18
LDUHDL > 5
Placebo
Gemfibrozil
1.
Bar graphs show univariate associations of baseline lipoprotein cholesterol levels and triglycerides with coronary heart
FIGURE
disease incidence in the gemfibrozil and placebo groups of the Helsinki Heart Study. Numbers of events in each subgroup are shown
at the base of each bar. LDL, low density lipoprotein; HDL, high density lipoprotein; C, cholesteroL
*
studied the joint effect of baseline triglycerides and
lipoprotein cholesterol concentrations on the risk of
CHD and on the risk reduction associated with
gemfibrozil treatment.
Methods
The Helsinki Heart Study was a 5-year, placebocontrolled, double-blind clinical trial designed to test
the hypothesis that lowering serum LDL-C and triglyceride levels and elevating serum HDL-C level with
gemfibrozil, a lipid-modulating drug belonging to the
group of fibric acid derivatives, reduces the incidence of
CHD in middle-aged dyslipidemic men. The design and
execution of the trial have been described in detail
elsewhere.'3-15 Briefly, the participants for the study
were selected from 23,531 men aged 40-55 years who
were employed by two state agencies and five industrial
companies in Finland. Volunteers were eligible for the
trial if their serum non-HDL cholesterol (i.e., LDL-C
plus VLDL cholesterol) was .5.2 mmol/l at two successive measurements and if they had no evidence of
CHD or other major disease. The mean baseline cholesterol, LDL-C, HDL-C, and triglyceride levels of the
trial population were 6.98, 4.88, 1.23, and 2.02 mmol/l,
respectively. The trial participants were randomly allocated either to gemfibrozil (n =2,046) or placebo
(n=2,035).
The follow-up examinations performed at 3-month
intervals included laboratory measurements and an
interview on possible adverse effects, hospitalizations, main illnesses, and any symptoms suggesting
myocardial infarction. Routine electrocardiograms
were taken annually and whenever the participants
reported symptoms suggesting myocardial infarction.
Definite fatal and nonfatal myocardial infarctions
and cardiac death were the end points. The endpoint assessments were made without knowledge of
the treatment group.
Manninen et al Joint Effect of Baseline Triglyceride and Lipoprotein Levels on CHD
Downloaded from http://circ.ahajournals.org/ by guest on June 14, 2017
Fasting serum samples were required only at the
second screening and semiannually during the follow-up for serum triglyceride measurement. Total
cholesterol concentration was determined from serum, and HDL-C was measured after precipitation of
VLDL and LDL with dextran sulphate magnesium
chloride by an enzymatic method (Boehringer Mannheim, kit No. 236691). The serum concentration of
triglycerides was measured as glycerol after enzymatic hydrolysis with lipase/esterase (Boehringer
Mannheim, kit No. 124966). The LDL-C concentration was calculated using the formula LDL-C=total
cholesterol minus HDL-C minus triglycerides divided
by 2.2.16 Triglyceride values of 8.1 mmol/1 or more
were excluded from LDL calculations. Computed
LDL-C values below 2.6 mmol/1 were also excluded.
Triglyceride values from the second screening visit
were taken to represent the baseline level of serum
triglycerides and were also used in calculating baseline LDL-C values. Cholesterol and triglyceride concentrations are given in millimoles per liter. To
convert values to milligram per deciliter, they should
be divided by 0.02586 and 0.01129, respectively.
Baseline lipid values were dichotomized for the
present analyses. The cut-off points were total cholesterol, 7.8 mmol/l; LDL-C, 5.0 mmol/1; triglycerides, 2.3 mmol/l; and LDL-C/HDL-C ratio, 5. These
lipid levels follow the Nordic recommendations for
drug treatment of hypercholesterolemia.17 Total cholesterol, LDL-C and triglyceride levels, and LDL-C/
HDL-C ratio were above these cut-off points in 16,
42, 26, and 21% of the trial population, respectively.
The cut-off point for HDL-C was 1.08 mmol/l (the
upper limit of the lowest tertile of the baseline
HDL-C distribution).
Smoking habits were recorded by an interview at
the first screening visit, and a dichotomous variable
(current smokers, nonsmokers, and past smokers)
was used in the analyses.
Statistical Methods
The analyses were done on an "intention to treat"
principle. Crude incidence rates are reported when
39
describing the joint effects of lipids and lipoproteins
the incidence of CHD events. To control for
confounding variables, the risk patterns were also
studied using Cox proportional hazards models18"19
with age, smoking, and systolic blood pressure as
covariates. In these analyses, we used categorized
variables and a dummy variable technique that allows
simultaneous study of several subgroups from the
placebo and the gemfibrozil groups.20 For example,
when studying the joint effect of treatment and high
and low levels of LDL-C and triglycerides (2x2x2
groups), the placebo group with low LDL-C and low
triglycerides was chosen as the reference and a
system of dummy variables was generated for the
remaining seven combinations. These dummy variables were equal to 1 if the combination was present
and 0 if the combination was absent.
To study the short-term and long-term predictive
power of baseline lipoprotein levels, follow-up time
was divided into two periods (the first 2 years and the
last 3 years) and the risks were assessed separately for
both periods. All analyses reported here are based on
baseline lipid and lipoprotein concentrations.
on
Results
Univariate Association Between Serum Lipid Levels
and CHD Incidence
Univariate associations between CHD incidence
and the dichotomized lipid and lipoprotein variables
are shown in Figure 1. In this selected hyperlipidemic
population, both total cholesterol and LDL-C were
poor indicators of CHD risk. On the other hand, low
HDL-C concentration (<1.08 mmol/l) was associated with high risk (RR, 1.73 compared with subjects
with HDL-C >1.08 mmol/l) (Figure 1, Table 1).
Similarly, subjects with serum triglycerides >2.3
mmol/l had a significantly increased risk of CHD
(RR, 1.81) (Figure 1, Table 2). Excess CHD risk
associated with elevated triglyceride and low HDL-C
levels was largely eliminated by gemfibrozil treatment
(Table 1).
TABLE 1. Relative Risk of Cardiac Events in Relation to Baseline Serum High Density Lipoprotein Cholesterol
Concentration in the Helsinki Heart Study
HDL-C concentration (mmol/1)
. 1.08
< 1.08
n=651 (placebo), 694 (gemfibrozil)
n=1,384 (placebo), 1,352 (gemfibrozil)
Total 5-year follow-up
1.73 (1.12-2.66)
1
Placebo
0.92 (0.55-1.54)
0.79 (0.50-1.23)
Gemfibrozil
First 2 years of follow-up
1.44 (0.68-3.05)
1
Placebo
1.49 (0.72-3.11)
0.91 (0.44-1.85)
Gemfibrozil
Last 3 years of follow-up
1.90 (1.12-3.22)
1
Placebo
0.60 (0.28-1.26)
0.72 (0.41-1.28)
Gemfibrozil
Risks were estimated using Cox regression models with age, smoking status, and systolic blood pressure as covariates.
Risk in subjects in the placebo group with high density lipoprotein cholesterol (HDL-C) concentration .1.08 mmol/l
was set at 1. In parentheses, 95% confidence intervals are given.
40
Circulation Vol 85, No 1 January 1992
TABLE 2. Relative Risk of Cardiac Events in Relation to Baseline
Helsinki Heart Study
Serum Triglycerides Concentration in the
Triglyceride concentration (mmol/l)
c2.3
>2.3
n= 1,529 (placebo), 1,506 (gemfibrozil)
n=506 (placebo), 540 (gemfibrozil)
Total 5-year follow-up
Placebo
1
1.81 (1.16-2.81)
Gemfibrozil
0.83 (0.55-1.25)
0.76 (0.43-1.36)
First 2 years of follow-up
Placebo
1
3.97 (1.85-8.33)
Gemfibrozil
1.84 (0.91-3.75)
1.10 (0.49-3.51)
Last 3 years of follow-up
Placebo
1
1.17 (0.65-2.10)
Gemfibrozil
0.52 (0.31-0.90)
0.60 (0.29-1.24)
Risks were estimated using Cox regression models with age, smoking status, and systolic blood pressure as covariates.
Risk in subjects in the placebo group with high density lipoprotein cholesterol concentration .1.08 mmol/1 was set at
1. In parentheses, 95% confidence intervals are given.
Downloaded from http://circ.ahajournals.org/ by guest on June 14, 2017
The risk associated with low HDL-C level in the
placebo group was similar during the first 2 years and
the last 3 years of follow-up (Table 1). The pattern
seen in the gemfibrozil group was quite different.
Here, subjects with low baseline HDL level had an
increased risk only during the first 2 years but not
during the last 3 years, suggesting that the gemfibrozil-induced elevation of HDL-C level had eliminated the extra risk after 2 years of treatment. High
baseline triglyceride concentration (>2.3 mmol/l)
predicted the risk of CHD only during the first 2
years in the placebo group (Table 2). This extra risk
was eliminated by gemfibrozil treatment.
The association between baseline triglyceride concentration and CHD incidence in the placebo group
was J-shaped, with the lowest incidence in subjects
with triglyceride levels between 1.2 and 1.5 mmol/l
(Figure 2). The incidence of cardiac end points in the
two highest quintiles was approximately 2.5 times
greater than that in the second quintile but only
approximately 1.5 times higher than the incidence in
the lowest quintile. Despite the lack of monotony, the
relative risk of CHD was significantly higher in the
two highest quintiles of triglycerides (>1.9 mmol/l)
than in the three lowest quintiles (Table 3). When
adjusted for age, smoking, hypertension, and either
total cholesterol or LDL-C, using Cox models, the
subjects in the two highest quintiles had approximately 75% higher CHD risk than subjects in the
three lowest quintiles. When HDL-C was added to
the models, relative risk decreased and was no more
statistically significant (NS). Essentially similar results were obtained when 2.3 mmol/l was used as the
cut point for hypertriglyceridemia.
The pattern was quite similar when serum triglyceride level was studied as a continuous variable.
When controlled for age, smoking, and hypertension,
triglycerides were significantly (p<0.05) associated
with CHD risk in the placebo group, but when
HDL-C was included in the models, this association
was no longer significant. Essentially similar results
were observed both when serum triglyceride concentration was used as such and after a logarithmic
transformation. At first sight, these results would
imply that baseline triglyceride concentration was not
an independent risk factor for CHD. However, because the correlation between baseline values of
HDL-C and triglycerides (after a logarithmic transformation) was not negligible (r= -0.44), further
analyses were carried out in an attempt to eliminate
the effects of multicollinearity.
Joint Effects of Serum Triglycerides and Lipoprotein
Cholesterol on CHD Incidence
Analysis of the joint effect of triglycerides and total
cholesterol demonstrated that a high level of total
Incidence of cardiac events
(per 1000 person-years)
10
5
0
Q1
02
Q3
Q4
Q5
of
FIGURE 2. Bar graph shows incidence of coronary heart
disease by quintiles of baseline triglyceride values in the
placebo group. Cut-offpoints of quintiles were 1.20, 1.51, 1.90,
and 2.50 mmolll, respectively. Number of events in each
subgroup are shown by numbers at the base of each bar.
Manninen et al Joint Effect of Baseline Triglyceride and Lipoprotein Levels on CHD
A Incidence of cardiac events
(per 1000 person-years)
35
and Total
of carfia ever
BIlncideno
(per 1000 pers-years)
3 -Triglycerides and ILDL Cholesterol
30
30
25
25
20
20
15
15
10
10
5
5
0
0
TG
.2.3
TG >23
<2.3
>2.3
TOTAL CHOLESTEROL>73
TG
TOTAL CHOLESTEROLS73
Downloaded from http://circ.ahajournals.org/ by guest on June 14, 2017
C
Cholesterol
35
6
LDL> 5D
ent
carda
D (pr
rdn1000 ofpsrron-years)
and HDL Cholesterol
35
30
25
25
20
20
15
15
10
10
5
5
0
*
18
LfLSSO
30
c
-
TG <2.3 TG >23
TG
of cardiac events
Incidence
(per 1000 person-years)
41
TO 52.3 TO >2.3
TO <2.3 TO >2.3
HDL21.
HDL< 1.08
Triglcerides and LDL-C/HDL-C
9
02
TO 52.3 TB >2.3
LDHDLig5.0
TO3
23 TO >23
LDUHL>5Z
Placebo
Gemfibrozil
FIGURE 3. Bargraphs show joint effect of baseline triglycerides and serum lipoprotein cholesterol levels on coronary heart disease
incidence in the gemfibrozil and placebo groups of the Helsinki Heart Study. Number of events in each subgroup is shown by
numbers at the base of each bar. LDL, low density lipoprotein; HDL, high density lipoprotein; C, cholesterol.
cholesterol was associated with elevated CHD risk
only when the triglyceride level was also high (>2.3
mmol/l) (Figure 3). Similar results were seen when
the risks were calculated using LDL-C instead of
total cholesterol (Figure 3). Data obtained after
adjustment for age, smoking, and hypertension using
Cox models confirmed the visual impression from the
crude incidence rates. Relative risk was highest for
subjects with both LDL-C and triglycerides above the
cut-off points (RR, 2.37) and was not elevated in
subjects with high LDL-C and low triglyceride level
(RR, 1.06) (Table 4). The joint effects of HDL-C and
triglycerides on CHD incidence are shown in Figure
3 and Table 5. The risk was particularly high in
subjects with high triglyceride and low HDL-C levels.
As shown in Table 6 and Figure 3, the LDL-C/
HDL-C ratio was a strong predictor of CHD risk,
especially in subjects with high triglyceride levels;
thus, the highest risk in any of the subgroups examined in this report was observed in subjects with
LDL-C/HDL-C ratio >5 and triglyceride concentration >2.3 mmol/l. This group also benefitted most
from gemfibrozil treatment; the incidence was reduced by more than 70% compared with the placebo
group (p<0.005). No difference was observed in
all-cause mortality between the gemfibrozil- and placebo-treated subjects in this subgroup. The all-cause
mortality in the placebo group was 5.82 per 1,000
person-years, with four deaths (all cardiovascular)
and 3.88 per 1,000 person-years in the gemfibrozil
group, with three deaths (two cardiovascular deaths
and one accident).
Finally, we analyzed the LDL-C/HDL-C ratio in
subjects with this ratio >5 and serum triglycerides
42
Circulation Vol 85, No 1 January 1992
TABLE 3. Relative Risk of Cardiac Events by Baseline Levels of Serum Triglcerides in Placebo Group of the
Helsinki Heart Study
Serum triglyceride concentration (mmol/l)2*
. 1.9
> 1.9
<2.3
>2.3
Covariates
(n = 1,217)
(n = 1,529)
(n=818)
(n =506)
*Model A: triglycerides,
1
1.75
1
1.86
age, smoking, systolic
(1.13-2.71)
(1.19-2.91)
blood pressure
Model A plus
1
1.72
1
1.80
total cholesterol
(1.11-2.66)
(1.15-2.83)
Model A plus
1
1.85
1
2.01
LDL cholesterol
(1.19-2.88)
(1.27-3.18)
Model A plus
1
1.29
1
1.40
HDL cholesterol
(0.80-2.07)
(0.87-2.25)
Risks were estimated using Cox regression model with age, smoking status, and systolic blood pressure as covariates.
In each model, the lower category was used as the reference group with unity risk. In parentheses, 95% confidence
intervals are given. LDL, low density lipoprotein; HDL, high density lipoprotein.
*The two cut points used in categorization (1.9 and 2.3 mmol/1) were chosen based on quintile analysis of the present
cohort (cut point between the third and fourth quintile, see also Figure 2) and the Nordic recommendations for drug
treatment of hypertriglyceridemia (see Reference 17).
Downloaded from http://circ.ahajournals.org/ by guest on June 14, 2017
below or above 2.3 mmol/l. Although the ratio was
slightly higher in the hypertriglyceridemic subgroup
(6.2 versus 5.9), the difference was not large enough
to explain the observed difference in the CHD risk.
As an LDL-C/HDL-C ratio >5 may result either
from a high LDL-C level or from a low HDL-C level
(or both), the distributions of baseline LDL-C,
HDL-C, and triglycerides were examined in the
groups formed on the basis of the LDL-C/HDL-C
ratio and triglyceride concentration (Table 7). Ninety
percent of the subjects in the group with LDL-C/
HDL-C ratio >5 and triglycerides >2.3 mmol/l had
low levels of HDL-C, whereas 49% in that group had
high LDL levels. These data suggest that both the
elevated risk of CHD and the treatment effect depend more on the level of HDL-C than on the level
of LDL-C in this hyperlipidemic population.
Discussion
The main finding in the present report was the
strong interdependence of LDL-C, HDL-C, and triglyceride concentrations as predictors of CHD risk
and treatment benefit due to gemfibrozil. The risk
associated with the levels of HDL-C and LDL-C was
dependent on the level of triglycerides, and vice
versa. The LDL-C/HDL-C ratio had more prognostic
value than LDL-C and HDL-C alone, and hypertriglyceridemia was a strong indicator of the short-term
CHD risk, especially when the LDL-C/HD-C ratio
was also high.
Our results are compatible with several earlier
studies. The CHD risk associated with serum triglyceride concentration disappeared in most studies
when adjustment was made for other CHD risk
factors, especially HDL-C.6-8 On the other hand, an
independent effect of triglycerides on CHD risk was
observed both in the Framingham7 and Milwaukee8
Cardiovascular Data Register cohorts when triglyceride level was considered jointly with HDL-C or
with the total cholesterol HDL-C ratio. In the Paris
prospective study,12 interaction of the effects of total
cholesterol and triglycerides on CHD risk was significant, suggesting that the effects of triglyceride and
total cholesterol levels are interdependent. All these
observations suggest that much information is lost if
the independent association between serum triglycerides or lipoprotein cholesterol levels and the risk of
CHD is evaluated without taking into account their
joint effects.
Extensive evidence from animal studies, clinical
trials, and observational epidemiological studies has
confirmed the causal role of LDL-C in atherosclero-
TABLE 4. Relative Risk of Cardiac Events in Combined Categories of Baseline
Cholesterol and Triglycerides
Total 5-year follow-up
Placebo
Serum
LDL-C 55.0 (mmol/1)
TG .2.3
TG >2.3
TG
(mmolll)
(mmoIl/)
(mmol/l)
Low
Density
Lipoprotein
LDL-C >5.0 (mmol/l)
.2.3
TG >2.3
(mmol/l)
RR
1
1.61 (0.89-2.89)
1.06 (0.62-1.80)
2.37 (1.21-4.63)
n
841
339
687
140
Gemfibrozil
RR
0.67 (0.37-1.20)
0.49 (0.20-1.17)
1.07 (0.62-1.84)
1.50 (0.71-3.16)
n
819
346
686
164
Risks were estimated using Cox regression models with age, smoking status, and systolic blood pressure as covariates.
Those in the placebo group who had low density lipoprotein cholesterol (LDL-C) <5.0 mmol/l and triglycerides (TG)
<2.3 mmol/l were the reference group with unity risk. In parentheses, 95% confidence intervals are given.
Manninen et al Joint Effect of Baseline Triglyceride and Lipoprotein Levels on CHD
TABLE 5. Relative Risk of Cardiac Events in Combined Categories of Baseline Serum High
Cholesterol and Triglycerides
Total 5-year follow-up
Placebo
Density Lipoprotein
HDL-C .1.08 (mmolll)
TG .2.3
TG >2.3
HDL-C <1.08 (mmol/l)
TG .2.3
TG >2.3
(mmoI/l)
(mmol/l)
(mmol/l)
43
(mmol/l)
RR
n
RR
n
1
1.31 (0.63-2.72)
1.33 (0.74-2.40)
2.43 (1.43-4.12)
1,166
218
363
288
Gemfibrozil
0.82 (0.50-1.35)
0.86 (0.38-1.94)
1.10 (0.59-2.04)
0.80 (0.37-1.73)
1,106
246
400
294
Risks were estimated using Cox regression models with age, smoking status, and systolic blood pressure as covariates.
Those in the placebo group with triglycerides (TG) <2.3 mmolI and high density lipoprotein cholesterol (HDL-C) .1.08
mmol/1 were the reference group with unity risk. In parentheses, 95% confidence intervals are given. RR, relative risk.
Downloaded from http://circ.ahajournals.org/ by guest on June 14, 2017
SiS.21,22 LDL-C also emerged as a risk factor for CHD
in the Helsinki Heart Study, although the association
found between baseline levels of LDL-C and risk of
CHD was relatively weak (Figure 1). In contrast,
there was a strong negative association between
HDL-C and CHD risk. This association persisted
after multivariate analyses with adjustment for total
cholesterol, LDL-C, and triglycerides. The relatively
weak association between LDL-C and CHD risk may
be due to the study inclusion criteria. Exclusion of
men with non-HDL cholesterol <5.2 mmol/l eliminated most subjects with low LDL-C but did not
influence the distribution of HDL-C.23 In fact, when
our results are compared with the data from
Framingham subjects with LDL-C >4.0 mmol/l, risk
patterns are fairly similar.24
In several studies, the LDL-C/HDL-C ratio has
emerged as the best single lipid predictor of CHD
risk.25 The present results suggest that the association between LDL-C/HDL-C ratio and CHD is influenced by the serum triglyceride concentration. Moreover, using the LDL-C/HDL-C ratio >5 and serum
triglycerides >2.3 mmol/1 as the cut-off points, it was
possible to define a subgroup with a particularly high
risk and with an approximately 75% reduction in
CHD incidence during gemfibrozil treatment. Even
more important, most of the treatment benefit observed during the trial was confined to this group,
which comprised about 10% of the initial study
cohort. It should be noted, however, that the cut-off
points chosen are to some extent arbitrary: lower
values of triglycerides and LDL-C/HDL-C ratio
would also have defined a high-risk group, although
not as marked as the present one. For example,
subjects with triglycerides >1.9 mmol/l and LDL-C/
HDL-C ratio >4.5 comprise a 17% subgroup of the
trial population that has a 53% reduction in CHD
incidence as a result of treatment with gemfibrozil.
Caution is also necessary in the interpretation of
these findings, as they are based on a post hoc
analysis of subgroups not defined in the original study
plan. It is important to note that our conclusions are
based on a cohort of initially healthy, hyperlipidemic
middle-aged men and are not necessarily generalizable to other populations.
Several mechanisms could explain the strong joint
effect of LDL-CIHDL-C ratio and hypertriglyceridemia on the rate of CHD. Based on what is known
about the physiological functions of lipoproteins, the
high LDL-C/HDL-C ratio is likely to affect the
progression of atherosclerosis. On the other hand,
there is increasing evidence that hypertriglyceridemia influences various hemostatic functions and
may therefore be involved in events leading to thrombosis formation.26 The remarkably strong treatment
effect in the group defined by high LDL-C/HDL-C
ratio and hypertriglyceridemia may be due to the
effects of gemfibrozil on both processes.27
Further support for the hypothesis that serum
triglycerides and HDL cholesterol do not influence
TABLE 6. Relative Risk of Cardiac Events in Combined Categories of Baseline Serum Low Density Lipoprotein
Cholesterol to High Density Lipoprotein Cholesterol Ratio and Triglycerides
LDL-C/HDL-C <5.0
TG .2.3
TG >2.3
(
Total 5-year follow-up
Placebo
mmol/l) (mmol/l)
1.05 (0.56-2.00)
341
0.72 (0.35-1.47)
356
LDL-C/HDL-C >5.0
TG c2.3
TG >2.3
(mmol/l)
(mmol/l)
1.19 (0.61-2.32)
3.82 (2.20-6.63)
266
138
0.76 (0.47-1.22)
1.24 (0.65-2.35)
1.08 (0.46-2.55)
Gemfibrozil
304
154
1,201
Risks were estimated using Cox regression models with age, smoking status, and systolic blood pressure as covariates.
Those in the placebo group with serum low density lipoprotein cholesterol (LDL-C) to high density lipoprotein
cholesterol (HDL-C) ratio .5.0 mmol/1 and triglycerides (TG) .2.3 mmol/l were the reference group with unity risk.
In parentheses, 95% confidence intervals are given. RR, relative risk.
RR
1
n
RR
n
1,262
44
Circulation Vol 85, No 1 January 1992
TABLE 7. Percentage Distribution of Tertiles of Baseline Low Density Lipoprotein Cholesterol and High Density
Lipoprotein Cholesterol Levels in the Placebo Group
LDL-CIHDL-C >5
LDL-C/HDL-C .5
TG .2.3 (mmol/l)
TG >2.3 (mmol/l)
TG .2.3 (mmol/l)
TG >2.3 (mmol/l)
n=1,262
n=341
n=266
n=138
LDL-C
34
65
3
Ti
18
38
T2
23
22
33
28
12
75
T3
49
100
100
100
Total
100
HDL-C
41
73
Ti
13
90
37
38
T2
24
9
50
21
3
T3
1
100
100
100
Total
100
LDL, low density lipoprotein; C, cholesterol; HDL, high density lipoprotein; TG, triglycerides.
Tertile limits were LDL-C Ti <4.49, 4.49 <T2 .5.20, T3 >5.20; HDL-C Ti <1.08, 1.08 <T2 <1.32, T3 .1.32.
Downloaded from http://circ.ahajournals.org/ by guest on June 14, 2017
the same mechanisms in the development of myocardial infarction was obtained from differences in risk
patterns during the first 2 and the last 3 years of the
present trial. Low HDL-C level predicted the risk in
the placebo group, especially during the last 3 years.
Moreover, gemfibrozil-induced reduction in risk in
subjects with low HDL cholesterol was seen only
during that last part of the intervention. In contrast,
high serum triglyceride level predicted the risk in the
placebo group only during the first 2 years of the
study. This extra risk was totally eliminated by gemfibrozil treatment. Although based on relatively small
numbers, these observations are compatible with the
hypothesis that serum triglycerides are involved in
thrombus formation in already-compromised coronary arteries.26
It is interesting to note that the lipid criteria used
to identify the subgroup with high risk are closely
similar to the lipoprotein pattern described for the
syndrome with small dense LDL.28 These criteria
include moderately elevated LDL-C and VLDL levels and low HDL-C levels. This pattern is common
among subjects with myocardial infarction and appears to be influenced by a common allele at a single
gene locus.29 Whether gemfibrozil treatment specifically reduces the risk associated with this syndrome
remains to be studied.
Summary
The present data suggest that elevated serum
triglyceride concentration is a marker of elevated
CHD risk, especially in subjects with a high LDL-C/
HDL-C ratio. Moreover, our findings suggest that by
using LDL-C/HDL-C ratio and fasting triglyceride
concentration as criteria, it is possible to define a
subgroup with an over-70% reduction in CHD risk
during gemfibrozil therapy. In fact, most of the effect
associated with gemfibrozil treatment was confined
to this subgroup comprising about 10% of the trial
population. These observations have important clinical implications, as they suggest that relatively sim-
ple laboratory measurements can be used to identify
a small patient group that is likely to benefit from
long-term drug intervention. Such tailoring of drug
therapy is particularly important in the light of
potential adverse effects associated with life-long
treatment with lipid-lowering drugs.
Acknowledgments
We thank Dr. Bjarne Hansen for inspiring discussions and Laura Fellman and Marja Nordberg
for skillful technical assistance in preparation of
the manuscript.
References
1. Castelli WP: The triglyceride issue: A view from Framingham.
Am Heart J 1986;112:432-437
2. Kaukola S, Manninen V, Halonen PI: Serum lipids with
special reference to HDL cholesterol and triglycerides in
young male survivors of acute myocardial infarction. Acta Med
Scand 1980;208:41-43
3. Gotto AM, Gorry GA, Thompson JR, Cole JS, Trost R,
Yeshurun D, DeBakey ME: Relationship between plasma
lipid concentrations and coronary artery disease in 496
patients. Circulation 1977;56:875-883
4. Barbir M, Wile D, Trayner I, Aber VR, Thompson GR: High
prevalence of hypertriglyceridaemia and apolipoprotein
abnormalities in coronary artery disease. Br Heart J 1988;60:
397-403
5. Austin MA: Epidemiologic associations between hypertriglyceridemia and coronary heart disease, in: Seminars in Thrombosis and Hemostasis. New York, Thieme Medical Publishers,
Inc, 1988, pp 137-142
6. Austin MA: Plasma triglyceride as a risk factor for coronary
heart disease: The epidemiologic evidence and beyond. Am J
Epidemiol 1989;129:249-259
7. Castelli WP, Doyle TG, Hames CG, Castelli WP, Dogle JT,
Gordon T, Hames CG, Hjortland MC, Hulley SB, Kagan A,
Zukel WJ: HDL cholesterol and other lipids in coronary heart
disease: The cooperative lipoprotein phenotyping study. Cir-
culation 1977;55:767-772
8. Freedman DS, Gruchow HW, Anderson AJ, Rimm AA,
Barboriak JJ: Relation of triglyceride levels to coronary artery
disease: The Milwaukee Cardiovascular Data Registry. Am J
Epidemiol 1988;127:1118-1130
9. Hulley SB, Rosenman RH, Bawol RD, Brand RJ: Epidemiology as a guide to clinical decisions: The association between
Manninen et al Joint Effect of Baseline Triglyceride and Lipoprotein Levels
10.
11.
12.
13.
14.
Downloaded from http://circ.ahajournals.org/ by guest on June 14, 2017
15.
triglyceride and coronary heart disease. N Engl J Med 1980;
302:1383-1389
Lippel K, Tyroler H, Eder H, Gotto A, Vahoung G: Relationship of hypertriglyceridemia to atherosclerosis. Arteriosclerosis
1981;1:406-417
Abbot RD, Carrol R: Interpreting multiple logistic regression
coefficients in prospective observational studies. Am J Epidemiol 1984;119:830-836
Cambien F, Jacqueson A, Richard JL, Warnet JM, Ducimetiere P, Claude JR: Is the level of serum triglyceride a
significant predictor of coronary death in "normocholesterolemic" subjects? Am J Epidemiol 1986;124:624-632
Minttari M, Elo 0, Frick MH, Haapa K, Heinonen OP,
Heinsalmi P, Helo P, Huttunen JK, Kaitaniemi P, Koskinen P,
Maenpaa H, Malkdnen M, Norola S, Pasternack A, Pikkarainen J, Romo M, Sjoblom T, Nikkila EA: The Helsinki
Heart Study: Basic design and randomization procedure. Eur
Heart J 1987;8(suppl 1):1-29
Frick MH, Elo 0, Haapa K, Heinonen OP, Heinsalmi P, Helo
P, Huttunen JK, Kaitaniemi P, Koskinen P, Maenpiiaa H,
Malkdnen M, Minttari M, Norola S, Pasternack A, Pikkarainen J, Romo M, Sjdblom T, Nikkila EA: The Helsinki
Heart Study: Primary prevention trial with gemfibrozil middleaged men with dyslipidemia: Safety of treatment, changes in
risk factors, and incidence of coronary heart disease. N Engl J
Med 1987;317:1237-1245
Manninen V, Elo 0, Frick MH, Haapa K, Heinonen OP,
Heinsalmi P, Helo P, Huttunen JK, Kaitaniemi P, Koskinen P,
Maenpaa H, Malkonen M, Manttari M, Norola S, Pasternack
A, Pikkarainen J, Romo M, Sjdblom T, Nikkila EA: Lipid
alterations and decline in the incidence of coronary heart
disease in the Helsinki Heart Study. JAMA 1988;260:641-651
16. Friedewald WT, Levy RI, Fredrickson DS: Estimation of the
concentration of low-density lipoprotein cholesterol in plasma,
without use of the preparative ultracentrifugation. Clin Chem
1972;18:499-502
17. National Board of Health and Welfare Drug Information
Committee, Sweden: Workshop on Treatment of Hyperlipidemia. Uppsala, Almqvist & Wiksell, 1989
18. Cox DR, Oakes D: Analysis of survival data, in: Monographs
on Statistics and Applied Probability. London, Chapman & Hall,
Ltd, 1984
on
CHD
45
19. Dixon WJ: BMDP Statistical Software. Berkeley, Calif, University
of California Press, 1985
20. Kahn HA, Sempos M: Statistical methods in epidemiology, in:
Monographs in Epidemiology and Biostatistics, vol 14. New
York, Oxford University Press, 1989
21. Goldstein JL, Kita T, Brown MS: Defective lipoprotein receptors and atherosclerosis: Lessons from an animal counterpart
of familial hypercholesterolemia. N Engl J Med 1983;309:
288-296
22. Tyroler HA: Review of lipid-lowering clinical trial in relation
to observational epidemiologic studies. Circulation 1987;76:
515-522
23. Manninen V, Huttunen JK, Tenkanen L, Heinonen OP,
Minttairi M, Frick MH: High-density lipoprotein cholesterol
as a risk factor for coronary heart disease in the Helsinki Heart
Study, in Miller NE (ed): High Density Lipoproteins and
Atherosclerosis II. Amsterdam, Excerpta Medica, ICS 826,
1989, pp 35-42
24. Gordon T, Castelli WP, Hjortland MC, Kannel WB, Dawber
TR: High density lipoprotein as a protective factor against
coronary heart disease: The Framingham Study. Am J Med
1977;62:707-714
25. Assmann G, Schulte H, Funke H, von Eckardstein A, Seedorf
U: The prospective cardiovascular Munster (PROCAM)
study: Identification of high-risk individuals for myocardial
infarction and the role of high-density lipoprotein, in Miller
NE (ed): High Density Lipoproteins and Atherosclerosis II.
Amsterdam, Excerpta Medica, ICS 826, 1989, pp 51-65
26. Crutchley DJ, McPhee GV, Terris MF, Canossa-Terris MA:
Levels of three hemostatic factors in relation to serum lipids.
Arteriosclerosis 1989;9:934- 939
27. Andersen P, Smith P, Seljeflot I, Brataker S, Arnesen H:
Effects of gemfibrozil on lipids and haemostatis after myocardial infarction. Thromb Haemost 1990;63:174-177
28. Austin MA, Breslow JL, Hennehens CH, Buring JE, Willet
WC, Krauss RM: Low-density lipoprotein subclass patterns
and risk of myocardial infarction. JAMA 1988;260:1917-1921
29. Austin MA, King M-C, Vranizan KM, Krauss RM: Atherogenic lipoprotein phenotype: A proposed genetic marker for
coronary heart disease risk. Circulation 1990;82:495-506
KEY WoRDS
o
triglycerides * low density lipoprotein
high density lipoprotein * risk factors * clinical trials
-
Joint effects of serum triglyceride and LDL cholesterol and HDL cholesterol
concentrations on coronary heart disease risk in the Helsinki Heart Study. Implications
for treatment.
V Manninen, L Tenkanen, P Koskinen, J K Huttunen, M Mänttäri, O P Heinonen and M H
Frick
Downloaded from http://circ.ahajournals.org/ by guest on June 14, 2017
Circulation. 1992;85:37-45
doi: 10.1161/01.CIR.85.1.37
Circulation is published by the American Heart Association, 7272 Greenville Avenue, Dallas, TX 75231
Copyright © 1992 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/85/1/37
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/