HDL Cholesterol and Other Lipids
in Coronary Heart Disease
The Cooperative Lipoprotein Phenotyping Study
WILLIAM P. CASTELLI, M.D., JOSEPH T. DOYLE, M.D.,
TAVIA GORDON, CURTIS G. HAMES, M.D., MARTHANA C. HJORTLAND, PH.D.,
STEPHEN B. HULLEY, M.D., ABRAHAM KAGAN, M.D., AND WILLIAM J. ZUKEL, M.D.
Downloaded from http://circ.ahajournals.org/ by guest on June 17, 2017
SUMMARY The relation between coronary heart disease (CHD)
prevalence and fasting lipid levels was assessed by a case-control
study in five populations with a total of 6859 men and women of
black, Japanese and white ancestry drawn from subjects aged 40
years and older from populations in Albany, Framingham, Evans
County, Honolulu and San Francisco.
In each major study group mean levels of high density lipoprotein
(HDL) cholesterol were lower in persons with CHD than in those
without the disease. The average difference was small - typically
3-4 mg/dl - but statistically significant. It was found in most agerace-sex specific groups. The inverse HDL cholesterol-CHD association was not appreciably diminished when adjusted for levels of low
density lipoprotein (LDL) cholesterol and triglyceride. LDL, total cholesterol and triglycerides were directly related to CHD prevalence;
surprisingly, these findings were less uniformly present in the various
study groups than the inverse HDL cholesterol-CHD association.
DURING THE PAST TWO DECADES considerable
progress has been made delineating the role of the plasma
lipoproteins in the development of coronary heart disease
(CHD). Interest has focused chiefly on the very low density
and low density lipoproteins (VLDL and LDL); there has
been relatively little interest in the role of the high density
lipoproteins (HDL), which ordinarily carry about 20% of the
total plasma cholesterol. (In electrophoretic terms, HDL
and LDL correspond to alpha and beta, while VLDL corresponds to prebeta.) The neglect of HDL cholesterol is
curious since as early as 1951 Barr et al. reported that
healthy men had higher levels of alpha (or high density)
lipoprotein than did men with CHD.1 This early observation
was confirmed in subsequent cross-sectional studies;2-7
moreover, women, who have less CHD than men, were
noted to have higher levels of this lipoprotein.5
The Cooperative Lipoprotein Phenotyping Study of subjects drawn from epidemiologic studies of five diverse populations provides an excellent data base for examining the
role of the various lipid fractions in coronary heart disease.
In this report fasting levels of HDL, LDL and total
cholesterol, and triglyceride are related to CHD prevalence.
Service employees in Albany, New York; a general population of black and white men and women in Evans County,
Georgia; a general population of men and women in
Framingham, Massachusetts; and general populations of
men of Japanese ancestry living in Honolulu and San Francisco. In Albany and Framingham entire cohorts were invited to participate in the Cooperative Lipoprotein Phenotyping Study. The other three studies invited only random
samples of their total study population but supplemented
this by calling in all study subjects who were known from
prior examinations to have CHD. Details of the design are
given elsewhere.1" The CHD cases from each study were
contrasted with non-cases (controls) from the same study
population.
Plasma was obtained after an overnight fast of at least 12
hours. The fast was confirmed by a query at tne time the participant appeared. An additional check on fasting was made
by refrigerating an aliquot of plasma overnight and examining it the following day. If a surface wisp of chylomicrons
appeared, it was presumed that the participant had not
fasted. Only fasting specimens were used.
The procedures for preparing specimens were specified by
the Laboratory of Molecular Disease of the National Heart
and Lung Institute, which also provided training in the
procedures. The participating laboratories used common
primary standards as well as specimens from common pools
supplied from the Lipid Standardization Laboratory of the
Center for Disease Control for standardizing cholesterol and
triglyceride measurements. These specimens were then used
for determining levels of total, HDL, LDL and VLDL
cholesterol and triglycerides. Lipid determination for the
San Francisco and Honolulu studies were made at a common laboratory (in San Francisco); otherwise each study
had its own laboratory. Details of collection, preparation,
and lipid determination methods together with particulars
respecting quality control have been published previously.18
CHD cases were defined in each center according to the
established study criteria. 12 Sixty-five per cent of the cases
had definite myocardial infarction (MI) diagnosed by ECG;
48% had angina pectoris (AP) based on the Rose questionnaire or clinical assessment; and 7% had coronary in-
Methods
Data from five study populations participating in the
Cooperative Lipoprotein Phenotyping Study served as the
basis for this report. The overall design and methods of these
studies, all of which were derived from ongoing prospective
studies of cardiovascular disease, have been described
previously.""' Briefly they were: a population of male Civil
From the National Institutes of Health, National Heart, Lung, and Blood
Institute, Heart Disease Epidemiology Study, Framingham, Massachusetts;
the Cardiovascular Health Center of the New York State Department of
Health, Division of Cardiology, Department of Medicine, Albany Medical
College, Albany, New York; The National Institutes of Health, Bethesda,
Maryland; Claxton, Georgia; Institutes of Medical Sciences, San Francisco,
California; and Honolulu Heart Study, Honolulu, Hawaii.
Address for reprints: Tavia Gordon, National Heart, Lung, and Blood
Institute, Landow Bldg., Room C841, 7910 Woodmont Avenue, Bethesda,
Maryland 20014.
Received September 10, 1976; revision accepted January 4, 1977.
767
VOL 55, No
CIRCULATION
768
TABLE 1. Mean HDL Choleterol Levels and Number of
Persons in Study Groups by Age, Race and Sex
Study group
40-49
Mean HDL cholesterol (mg/dl)
Men
48.3
Albany
Framingham
Honolulu
46.3
San Francisco
52.3
Evans County-white
61.8
Evans County-black
Women
Framingham
61.3
Evans County-white
66.4
Evans County-black
Downloaded from http://circ.ahajournals.org/ by guest on June 17, 2017
Number of subjects
Men
Albany
Framingham
Honolulu
San Francisco
Evans County-white
Evans County-black
Women
Framingham
Evans County-white
Evans County-black
Age groups
50-59
60-69
5,
MAY 1977
examined, the method of Mantel'5 was employed to test the
association between one lipid and CHD controlling for the
other lipid (figs. 1-3).
70+
48.7
44.8
44.4
45.9
48.3
59.0
49.6
45.5
46.0
48.8
48.1
56.3
51.4
45.5
46.6
49.7
50.9
51.0
58.6
64.3
53.3
56.8
57.7
66.4
55.3
53.9
67.5
Results
Table 1 gives the mean levels of HDL cholesterol by age,
race and sex in the five study populations. Levels for women
are about 10 mg/dl higher than for men and mean levels for
blacks about 10 mg/dl higher than for whites. Although the
black populations are very small, sex trends are consistent,
i.e., black women have mean levels approximately 10 mg/dl
higher than black men.
HDL Cholesterol Level in CHD Cases
differences in HDL cholesterol levels between
CHD and those without are given in table 2.
For all studies combined the differences are significant at a
5% level for the age group 40-49 and at a 1% level for the
three remaining age groups. Mean HDL cholesterol levels
for persons with CHD are consistently lower than for persons without CHD. This is true not only across the studies,
but (with one minor exception) within each population when
taken over all age groups. In fact, except in instances where
the numbers in certain age groups were too small for construction of reliable statistics, this differential was also noted
for individual age and sex groups.
A lower HDL cholesterol level was evident for both
angina pectoris and myocardial infarction. When the data
for all random samples of men 50-69 were pooled, persons
without CHD had higher levels than those with either angina
pectoris (AP) or myocardial infarction (MI) and both contrasts were statistically significant. The contrast was greater
for AP than MI, AP cases having lower HDL levels than MI
cases. The difference was statistically significant at a 5%
level.
The mean
persons with
94
0
0
319
22
9
361
506
1182
221
25
19
728
423
621
84
14
10
122
284
54
20
8
9
0
16
16
614
27
23
575
407
14
10
14
8
sufficiency by clinical criteria. All available CHD cases were
used in case-control comparisons for individual studies.
Where studies were pooled, only cases within the probability
samples were included.
Differences between the mean values for all CHD
prevalence cases and the probability samples of non-cases
for the individual age, sex, race categories (tables 2-3) were
tested for significance using the conventional t-test. Regression coefficients were computed for the bivariate and multivariate discriminant function by the method of R.A.
Fisher.'4 For purposes of assessing the relative strength of
association between specified lipids and CHD, unit-free
coefficients (standardized coefficients) were obtained by
multiplying each coefficient by the standard deviation of the
lipid (table 5). When cross-classification of two lipids was
Levels of Total and LDL Cholesterol, and Triglyceride, in CHD Cases
The relations of other lipids to CHD prevalence are shown
in table 3. For most study groups, the mean level in subjects
with CHD is higher than the mean level in those without disease. The magnitude of the average differences for total and
TABLE 2. The Relaion of HDL Cholesterol to CHD Prevalence
Men
Evans County
Age
Totalt
Albany
Framingham
Honolulu San Francisco
White
Mean difference in HDL cholesterol level (mg/dl); CHD cases noncasest
-3.46
- 17.19**
40-49 -6.05* -13.08
-2.32
-3.20
-2.61
-3.01**
-10.07*
50-59 - 3.17**
3.88
-6.23** -4.22**
-5.86
60-69 -4.03** -2.14
-4.11**
-6.42
-5.16
-4.93*
-18.89*
-7.59*
70+
All ages -3.81** -2.93** -4.85** -3.78**
-3.34
-7.96*
Number of CHD cases
2
14
2
0
40-49
18§
48
142
14
19
285
35
50-59
383
74
4
14
116
109
60-69
203
12
28
57
16
3
70+
*P <0.05
**P <0.01
Black
Framingham
-10.95*
50.00
-1.59
Women
Evans County
Black
White
-
16.32
-5.46*
-2.25
-3.34*
0
3
4
3
16
51
73
-1.33
12.24
-10.51
-16.43** -28.71**
10.04
-22.17*
-3.84
- 11.03*
1
4
7
8
tMean level of all CHD cases minus mean level in the probability sample of noncases for the specified age-race-sex group of the specified study.
, where ni = number of cases, di = mean difference, and n = Mnm.
t2 nidi
n
Tested for significance by means of a summary t-statistic computed on mean differences similarly weighted.
§Omits women for Evans County.
1
4
4
3
HDL CHOLESTEROL/Castelli
et
769
al.
TABLE 3. The Relation of Specified Lipids to CHD Prevalence
Men
Age
Total$
Albany
Framingham
Honolulu San Francisco
Evans County
White
Black
Mean difference in total cholesterol level (mg/dl); CHD cases-noncasest
40-49
4.03
11.11
36.64
- 78.10**
50-59 10.84**
9.77
7.74
8.55*
24.80*
26.05*
60-69
7.88**
5.74
-1.36
11.79**
7.04
30.00*
70+
-.58
-4.54
2.01
-1.67
-13.33
-39.67*
All agest 6.82**
4.00
2.15
9.26**
13.99*
16.20*
Mean difference in LDL cholesterol level (mg/dl); CHD cases-noncasest
40-49
2.29
7.24
28.88*
-58.91**
50-59
9.57
4.50
10.09**
13.11
14.43
9.74**
60-69
7.06**
5.18
-0.30
12.34**
0.55
31.21**
70+
.02
-2.13
5.88
0.17
-10.25
- 74.72**
All ages4 6.22**
4.19
2.96
10.41**
3.21
13.66
Mean difference in triglyceride level (mg/dl); CHD cases-noncasest
40-49 47.8
-32.7
46.9
135.4**
50-59 18.8
8.6
29.5
7.7
72.9
94.7
60-69 22.1**
17.1*
31.6**
10.3
34.5
20.1
70+
19.8**
3.2
11.8
28.0
346.9
10.3
All agest 21.0**
12.8
24.7**
10.0
80.5**
47.7*
- 5.77*
15.83*
-34.13
-5.64
8.51
-25.31**
-32.50
-17.32
-11.5
-49.6
22.7
-16.5
Framingham
6.82
17.46**
0.55
7.43
Women
Evans County
White
Black
40.36
-23.00
46.08
18.45
7.75
-18.64
-35.63
-12.54
9.09
13.54*
-0.57
5.67
42.78
-35.29*
33.54
9.62
-0.16
-0.29*
-18.93
-4.88
4.0
32.7**
19.5*
22.5**
1.8
145.7
19.5
59.2
-52.2
42.8
-4.4
-4.2
Downloaded from http://circ.ahajournals.org/ by guest on June 17, 2017
*P <0.05.
<0.01.
seP
tMean level of all CHD cases minus mean level in the probability sample of noncases for the specified age-race-sex group of the specified study.
where ni = number of cases, di = mean difference, n = 2ni. Tested for significance by means of a summary t-statistic computed on mean
tz nin
difference similarly weighted.
LDL cholesterol, over all ages and across the five studies, is
about 6 mg/dl; for triglyceride it is about 21 mg/dl. The
associations shown in table 3, however, are less uniform than
those shown in table 2 for HDL cholesterol and CHD.
Prevalence Rates by Lipid Level
The prevalence of CHD by level of HDL cholesterol is
given in table 4 for the pooled studies. Prevalence rates for
subjects with very low levels are about twice those for subjects with intermediate levels. There is no discernible
gradient when subjects with intermediate levels are compared with those having high levels.
Correlations Among Lipids
To assess the association of HDL cholesterol to other
lipids known to be related to CHD, correlation coefficients
were computed. The correlations between HDL cholesterol
and other cholesterol fractions are of a low order of magnitude, but their direction is consistent across populations; the
associations of HDL cholesterol with total cholesterol are
all positive, and those with LDL cholesterol are all negative.
There is a moderately strong negative correlation between
HDL cholesterol and triglyceride, which is also consistent
across populations. Triglyceride has a small negative correlation with LDL cholesterol (except for women), and a
moderate positive correlation with total cholesterol. By contrast, there is a very strong positive correlation between
LDL and total cholesterol.
These intercorrelations are critical in assessing the casecontrol data. The multivariate linear discriminant function
provides a statistical method useful for disentangling the
role of the three lipids relative to CHD prevalence, taking
these intercorrelations into account.
Results of discriminant analysis of CHD cases versus noncases using HDL and LDL cholesterol and triglycerides are
given in table 5. Bivariate analysis, in which the coefficient
for each lipid is adjusted for age, may be contrasted with
multivariate analysis, in which the coefficient for each lipid
is adjusted for age and the other two lipids. Standardized
bivariate coefficients for HDL cholesterol are negative and
significantly different from zero at a 5% level (or less) in five
of the six major study groups. The coefficients are slightly
diminished in absolute magnitude in the multivariate
analysis, but the inverse CHD-HDL cholesterol association
persists in all but one of the studies. LDL and triglyceride
have positive coefficients in both bivariate and multivariate
analysis, although in the bivariate case they are not as consistently significantly different from zero as was HDL
cholesterol.
When the data for men 50-69 were pooled, standardized
discriminant coefficients for HDL and LDL cholesterol were
almost the same in absolute magnitude, indicating similarly
strong associations with CHD prevalence; whereas no
association between triglyceride and CHD was found (table
5). Bivariate and multivariate results for each lipid from the
pooled sample were almost identical.
To explore this further, two-way tables were prepared for
each pair of lipids (HDL cholesterol, LDL cholesterol and
triglycerides) (figs. 1-3). For each lipid, levels were chosen
so as to divide the populations into three groups of roughly
TABLE 4. Prevalence of CHD by HDL Cholesterol Level,
Men aged 60-69
HDL
Cholesterol level
All levels
Less 25
25-34
35-44
45-54
55-64
65-74
75 plus
CHD
Number
Total population
Rate/1000
383
9
78
133
91
45
17
10
4165
50
631
1406
1168
578
215
117
180.0
123.6
94.6
77.9
77.9
79.1
85.5
92.0
VOL 55, No 5, MAY 1977
CIRCULATION
770
TABLE 5. Discriminant Analysis of CHD Cases versus Noncases among Persons 50-69, Using Specified Lipids
Bivariate standardized coefficients
Cholesterol
HDL
LDL
Triglyceride
Study group
Men
Albany
Framingham
Honolulu
San Francisco
Evans County, white
Pooled random sample
Women
Framingham
Multivariate standardized coefficients
Cholesterol
HDL
LDL
Triglyceride
Numbr
of cases
-0.19*
-0.37**
-0.30**
-0.09
-0.68**
- 0.24**
0.20*
0.06
0.95**
0.29
0.59*
0.20**
0.16*
0.29**
0.06
0.45*
0.56*
0.11*
-0.14
-0.30**
-0.28**
0.17
-0.39
-0.21**
0.21**
0.09
0.32**
0.43*
0.74**
0.22**
0.13
0.20*
0.06
0.61**
0.74**
0.09
151
122
251
23
28
383
-0.26*
0.29*
0.39**
-0.09
0.22*
0.32*
67
*P <0.05.
**P <0.01.
Note: Bivariate includes age and specified lipids.
Multivariate includes age and the three lipids.
Downloaded from http://circ.ahajournals.org/ by guest on June 17, 2017
equal size. While this analytical method is not as powerful as
multivariate discriminant analysis it is sensitive to special interactions that are not allowed for in discriminant analysis.
Data for men aged 50-69 were pooled from all five studies.
Data for blacks and for women were omitted, since they
were substantially different (table 1). Because Albany,
Framingham and Honolulu had most of the CHD cases this
part of the analysis is dominated by their results.
There is a very regular increase of CHD prevalence rates
with increasing LDL cholesterol level at each level of HDL
cholesterol (fig. 1). Similarly, for each level of LDL
cholesterol, subjects with low levels of HDL cholesterol had
a greater CHD prevalence rate than subjects with moderate
or high levels of HDL cholesterol. The inverse relationship
between HDL cholesterol and CHD, when taken over the
three levels of LDL cholesterol, is significant (P < 0.001) by
a method of Mantel,I5 as are the positive trends of CHD
prevalence on LDL cholesterol level. While the HDL trends
are not completely regular, the irregularities can be attributed to sampling variability and are not evidence of significant interaction between HDL and LDL.
Figure 2 presents a similar analysis of CHD prevalence in
men by levels of triglyceride and HDL cholesterol for the
pooled data. The prevalence of CHD among subjects with
low levels of triglyceride are similar to those of subjects with
intermediate or high levels; i.e., there is no discernible trend
of CHD prevalence by triglyceride level when HDL level is
taken into account. However, an inverse association between
CHD prevalence and HDL cholesterol within each of the
three triglyceride levels is apparent and when taken over the
three levels is significant at a level of less than 1%.
Cross-classification of triglyceride with LDL cholesterol
level (fig. 3) leads to the conclusion that either lipid has a
statistically significant association with CHD prevalence
when the other is held constant. In particular, there is a
statistically significant trend of CHD prevalence rates by triglyceride level when LDL cholesterol level is taken into account whereas figure 2 shows that this trend is not seen when
HDL level is taken into account.
In general, then, when contingency tables are constructed
for the three lipids considered two at a time, HDL and LDL
cholesterol emerge as consistently significant factors in
CHD prevalence whereas triglyceride is less consistently a
significant factor. Multivariate discriminant analysis for the
pooled sample yields similar logistic regression coefficients
for HDL and LDL cholesterol but for triglyceride no
statistically significant relation to CHD was demonstrable in
multivariate analysis.
250
HDL
225
200
Cholesterol, mgidl =
Numbers On
Bars
40
m
40-53
e
>53
Indicate Population At Risk
w
Ul
z
z
-j
a-0
ia
a.
in
0
I
V.
<
Trend
Ot
CHD On Lipid:
LDL CHOLESTEROL, mg/di
HDL p-0.001; LDL p'0,001
FIGURE 1. Prevalence of CHD by levels of LDL and HDL
cholesterol in men aged 50-69. Pooled data from Cooperative
Lipoprotein Phenotyping Studies.
89
89-206
> 207
TRIGLYCERIDE, mg/dl
Trenid Of CHD Oil Lipid: HDL p1z0.01, Triglyceride NS
FIGURE 2. Prevalence of CHD by levels of triglyceride and HDL
cholesterol in men aged 50-69. Pooled data from Cooperative
Lipoprotein Phenotyping Studies.
HDL CHOLESTEROL/Castelli et al.
250
225
LLDL Cholesterol, mg/dl
LI < 140
m
140-179
o 180
,
Numbers On Bars Indicate Population At Risk
200
175
z
0
150
125
100
I
1.)
75
50
25
0
< 89
89-206
> 207
TRIGLYCERIDE, mg/dl
Tri(i
Of CHD Oii Lip d
LDL ;) 0.001
Trir(lycerldie i) 0.025
FIGURE 3. Prevalence of CHD by levels of triglyceride and LDL
cholesterol in men aged 50-69. Pooled data from Cooperative
Lipoprotein Phenotyping Studies.
Downloaded from http://circ.ahajournals.org/ by guest on June 17, 2017
Discussion
Two forms of analysis were used. In tables 2, 3 and 5
analyses are made for each specific study. This has the advantage of assuring that population and laboratory
differences are fully allowed for. Since specimens from cases
and noncases were intermingled in the local laboratories, the
comparisons of their means were unbiased. In table 4 and
figures 1-3, the data for men in the age range 50-69 from the
main study populations are pooled. Such pooling could, conceivably, introduce artifactual features but does have the advantages accruing from a larger data base and simplifying
the presentation. Moreover, the results are consistent with
the study-specific analyses.
The analyses using cross-classification are supplemented
by the analytically more powerful discriminant analysis.
Not only are the results of the two approaches consistent,
but the results of analysis by cross-classification tend to confirm the linear assumptions of the discriminant model.
The most important finding that emerges from these data
concerns the inverse relationship between HDL cholesterol
and CHD prevalence. The basic observation was made in
several early studies" but this report is the first describing a
wide diversity of populations and using modern quantitative
techniques for measuring lipoproteins. The results show that
the inverse HDL-CHD association is characterized by a
high degree of generality and strength.
The generality of the observation is striking in view of the
small magnitude of the difference in mean level of HDL
cholesterol between subjects with CHD and those without
the disease. The difference is typically 3-4 mg/dl, about 7%
of value, and less than the usual technical error of the
measurement, but it is found in all age groups studied, in
practically all populations studied, for both categories of
CHD studied, and within and between sexes. In fact, the uniformity of the finding even exceeds that for the direct relationship observed among the study groups between total
cholesterol concentration and CHD prevalence.
The independence of the association between HDL
cholesterol and CHD has been a point of concern because
there is a moderate inverse association between HDL
cholesterol and triglyceride.2' 891', "If In the studies reported
771
here, co-variance among the various lipid factors was controlled both by analysis of cross-tabulations and by discriminant analyses. Both analytic approaches leave no doubt that
the inverse association between HDL cholesterol and CHD
largely persists even when other lipid factors are considered;
that is, knowledge of HDL cholesterol appears to provide
risk information beyond that available from the usual lipid
risk factors.
The strength of the inverse association between HDL
cholesterol and CHD is such that there is a twofold gradient
of CHD prevalence between subjects at the higher and lower
ends of the distribution of HDL cholesterol concentration.
Examination of pooled data for men aged 50-69 shows that
this gradient of risk occurs in the lower half of the HDL
cholesterol concentration distribution. Thus, excess CHD is
associated with a low level of HDL cholesterol, but there
appears to be no advantage to having a higher than average
level. However, prospective data from Framingham (unpublished) show a decreasing CHD incidence even at aboveaverage levels.
HDL cholesterol levels may provide partial explanation of
hitherto unexplained population differences in CHD prevalence and incidence rates. Rates are lower in Evans County
blacks than whites, unexplained by the major risk factors of
blood pressure and serum cholesterol." The distinctly higher
HDL cholesterol levels in blacks than whites may in part account for this. On the other hand, Honolulu Japanese men
have only half the CHD incidence of men in Framingham"'
but the HDL cholesterol levels of these two populations are
practically the same.
The epidemiologic data presented in this report include
only information on prevalence of prior episodes of CHD.
The question of a precursive association can only be
answered by prospective studies of the relationship of HDL
cholesterol to CHD incidence. Such a study was reported in
1966 by Gofman who found that lower HDL levels were
followed by greater CHD incidence among young men.19 A
similar prospective finding among middle-aged Israeli men
has also been reported.20 Unpublished data from Framingham also show a lower HDL cholesterol level preceding the
appearance of CHD.
The finding that HDL cholesterol concentration is inversely related to subsequent development of CHD supports but does not prove the possibility that HDL elevation
may prevent the development of CHD. If this were the case,
there would be potential for identifying factors that could
favorably influence health by elevating HDL levels. The
fragmentary information on what maneuvers will lead to an
increase in HDL cholesterol levels suggest that physical activity,21' 22 weight loss2z and a low carbohydrate intake28' 24
may be beneficial.
If high HDL levels are shown to be protective, the
mechanism may be the one proposed by Miller and Miller."
They suggested that plasma HDL is a transport mechanism
for carrying cholesterol from the peripheral tissues to the
liver where it is catabolized and excreted. Hence, higher
levels of HDL cholesterol would be associated with less
atherosclerosis. Some evidence supporting this hypothesis
was presented, but as yet, there has not been sufficient test of
its validity.
The data in this report show a direct relationship between
fasting plasma triglyceride concentration and prevalence of
CIRCULATION
772
Downloaded from http://circ.ahajournals.org/ by guest on June 17, 2017
CHD. This relationship is found in most of the study groups
but when other lipids are considered is only equivocally
significant. However, even this limited finding is important
since controversy still exists concerning whether triglyceride
concentration is a risk factor for CHD, separate from its
association with cholesterol. However, it should be noted
that we have not adjusted for co-variance between triglyceride concentration and other relevant variables such as body
weight and the presence of diabetes.
While the full implications of these findings remain for
future work to elucidate, the virtue of partitioning total
cholesterol in assessing CHD risk is unequivocally demonstrated. Clearly if one fraction (HDL cholesterol) has a
negative association with the risk of CHD while the other
two (LDL and VLDL cholesterol) have positive associations with CHD risk, then the arithmetic sum, i.e., total
cholesterol, must be a less sensitive indicator of risk than an
appropriately weighted algebraic sum. (From a practical
point of view fasting triglyceride is the appropriate method
of measuring VLDL cholesterol since practically all the fasting triglyceride is carried in the VLDL portion and the two
measures are strongly correlated.) However, the laboratory
making the measurements must have good precision; otherwise the rate of misclassification may be intolerably high.
Still there is now no question from a scientific point of view
that partitioning total cholesterol is preferable to a single
measure of total cholesterol in assessing CHD risk; and a
lipid profile based on HDL and LDL cholesterol and triglyceride is a logically preferable method of measuring the
CHD risk associated with lipid characteristics. The appropriate statistical weighting, however, is yet to be determined
and should come from prospective data rather than casecontrol studies.
References
1. Barr DP, Russ EM, Eder HA: Protein-lipid relationship in human
plasma. II. In atherosclerosis and related conditions. Am J Med 11: 480,
1951
2. Nikkila E: Studies on the lipid-protein relationships in normal and
pathological sera and the effect of heparin on serum lipoproteins. Scan J
Clin Lab Invest 5 (suppl 8): 1, 1953
3. Jencks WP, Hyatt MR, Jetton MR, Mattingly TW, Durrum EL: A study
VOL 55, No 5, MAY 1977
of serum lipoproteins in normal and atherosclerotic patients by paper
electrophoresis techniques. J Clin Invest 9: 980, 1956
4. Brunner D, Lobl K: Serum cholesterol, electrophoretic lipid pattern, diet
and coronary artery disease: A study in coronary patients and in healthy
men of different origin and occupations in Israel. Ann Intern Med 49:
732, 1958
5. Keys A, Fidanza F: Serum cholesterol and relative body weight of coronary patients in different populations. Circulation 22: 1091, 1960
6. Carlson LA: Serum lipids in men with myocardial infarction. Acta Med
Scand 167: 399, 1960
7. Rhoads GG, Gulbrandsen CL, Kagan A: Serum lipoproteins and coronary heart disease in a population study of Hawaii Japanese men. N
Engl J Med 294: 293, 1976
8. Keys A: Blood lipids in man - a brief review. J Am Diet Assoc 51: 508,
1968
9. Hilleboe HE, James G, Doyle JT: Cardiovascular health center. I. Project design for public health research. Am 3 Public Health 44: 851, 1954
10. Cassel JC (ed): Evans County cardiovascular and cerebrovascular
epidemiologic study. Arch Intern Med 128: 883, 1971
11. Dawber TR, Meadors GF, Moore FE: Epidemiological approaches to
heart disease: The Framingham Study. Am J Public Health 41: 279, 1951
12. Kagan A, Harris BR, Winkelstein W Jr: Epidemiologic studies of coronary heart disease and stroke in Japanese men living in Japan, Hawaii,
and California: Demographic, physical, dietary and biochemical
characteristics. J Chronic Dis 27: 345, 1974
13. Castelli WP, Cooper GR, Doyle JT, Garcia-Palmieri M, Gordon T,
Hames C, Hulley SB, Kagan A, McGee D, Vicic W, Zukel WJ: Distribution of triglyceride and total LDL and HDL cholesterol in several populations: The cooperative lipoprotein phenotyping study. J Chronic Dis,
in press
14. Fisher RA: The use of multiple measurement in taxonomic problems.
Ann Eugen 7: 179, 1936
15. Mantel N: Chi-square tests with one degree of freedom; extensions of the
Mantel-Haenszel procedures. J Am Statist Assoc 58: 690, 1963
16. Miller GJ, Miller NE: Plasma-high-density-lipoprotein concentration
and development of ischemic heart disease. Lancet 1: 16, 1975
17. Tyroler HA, Hames CG, Krishan I, Heyden S, Cooper G, Cassel J:
Black-white differences in serum lipids and lipoproteins in Evans County.
Preventive Med 4: 541, 1975
18. Gordon T, Garcia-Palmieri MR, Kagan A, Kannel WB, Schiffman J:
Differences in coronary heart disease in Framingham, Honolulu and
Puerto Rico. J Chron Dis 27: 329, 1974
19. Gofman JW, Young W, Tandy R: Ischemic heart disease, atherosclerosis and longevity. Circulation 34: 679, 1966
20. Medalie JH, Kahn HA, Neufeld HN, Riss E, Goldbourt U: Five-year
myocardial infarction incidence. II. Association of single variables by age
and birthplace. J Chron Dis 26: 329, 1973
21. Lopez SA, Vial R, Balart L, Arroyave G: Effect of exercise and physical
fitness on serum lipids and lipoproteins. Atherosclerosis 20: 1, 1974
22. Wood PD, Klein H, Lewis S, Haskell WL: Plasma lipoprotein concentrations in middle-aged runners. Circulation 50 (suppl I1I):III-1 15, 1974
23. Wilson DE, Lees RJ: Metabolic relationships among the plasma
lipoproteins. Reciprocal changes in the concentrations of very low and
low density lipoproteins in men. J Clin Invest 51: 1051, 1972
24. Hulley SB, Wilson WS, Burrows MI, Nichaman MZ: Lipid and lipoprotein responses of hypertriglyceridemic outpatients to a low-carbohydrate
modification of the A.H.A. fat-controlled diet. Lancet 2: 551, 1972
HDL cholesterol and other lipids in coronary heart disease. The cooperative lipoprotein
phenotyping study.
W P Castelli, J T Doyle, T Gordon, C G Hames, M C Hjortland, S B Hulley, A Kagan and W J
Zukel
Downloaded from http://circ.ahajournals.org/ by guest on June 17, 2017
Circulation. 1977;55:767-772
doi: 10.1161/01.CIR.55.5.767
Circulation is published by the American Heart Association, 7272 Greenville Avenue, Dallas, TX 75231
Copyright © 1977 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/55/5/767
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/