0021-9681/87$3.00+ 0.00
Pergamon Journals Ltd
J Cbron Dts Vol. 40, No. 5, pp. 421428, 1987
Printed in Great Britain
THE ASSOCIATION OF BODY FAT DISTRIBUTION
WITH
HYPERTENSION,
HYPERTENSIVE
HEART DISEASE,
CORONARY HEART DISEASE, DIABETES AND
CARDIOVASCULAR
RISK FACTORS IN MEN AND
WOMEN AGED 18-79 YEARS
RICHARD F. GILLUM
Office of Analysis and Epidemiology Program, National Center for Health Statistics, Hyattsville,
MD 20782, U.S.A.
(Received
form 1 I August 1986)
in revised
Abstract-To
confirm the reported association of body fat distribution
with cardiovascular
disease,
diabetes, blood pressure and serum cholesterol, data from the 1960-62 Health Examination
Survey
were analyzed. In this sample drawn from the noninstitutionalized
population
of the United States
aged 18-79, mean values of two indices of upper versus lower body fat distribution
increased
steadily with age. Men had higher values than women, and black women had higher values than
white women. Higher values of the indices were significantly associated with higher blood pressure,
post-load serum glucose and greater prevalence of definite hypertension
and definite hypertensive
heart disease independent
of multiple confounders.
Associations
with higher serum cholesterol and
definite coronary heart disease prevalence were independent
of overall ponderosity
but not of age
and multiple other confounders.
Greater abdominal
relative to lower body fat deposits were
independently
associated with increased cardiovascular
risk in men and women, blacks and whites.
Obesity
Cardiovascular
Epidemiologic
methods
diseases
Hyperlipidemia
INTRODUCTION
SEVERAL studies
have reported
an increased
ratio of waist to hip girth to be associated with
increased occurrence
of cardiovascular
disease
and diabetes independent
of overall ponderosity
[l-6]. This study attempted
to replicate these
findings in a large sample of persons drawn
from
the
United
States
population
in
1960-1962.
This sample permitted
the examination of the association
by race and sex in a
representative
population
sample.
METHODS
The first cycle of the Health Examination
Survey was conducted
on a nationwide
multistage probability
sample of 7710 adults. The
sample was drawn from the noninstitutionalized
population
aged 18-79 yr of the United States
421
Hyperglycemia
Blood pressure
excluding Alaska and Hawaii [7]. Of the 7710
selected for the sample, 6672 persons
were
examined during the period from October 1959
to December
1962. Excluded
from present
analyses were 140 women who reported pregnancy and persons whose race was other than
white or black. Included in this report were 2669
white men, 358 black men, 2931 white women
and 448 black women. Details of the plan,
sampling,
response, and operation
were published previously,
as were procedures
used to
obtain
informed
consent
and to maintain
confidentiality
of obtained information
[7,8].
A Census Bureau interviewer
obtained personal and health information
at a household
interview before the examination.
The examination
in a mobile center included
a selfTechnicians
administered
medical
history.
measured height to the nearest mm; weight to
the nearest half pound; and standing waist girth,
422
RICHARD F. GILLUM
sitting
seat breadth,
sitting
thigh clearance
height, triceps and subscapular
skinfold thickness to the nearest mm as described elsewhere
[7,9, lo]. Seat breadth was the distance across
the greatest lateral protrusion
on each side of
the buttocks,
using light but sure contact of
the anthropometer
to compress
the clothing
but not the body. Thigh clearance height was
the distance from the top of the sitting surface
to the junction
of the abdomen
and thigh on
the right side with the anthropometer
crossbar in firm contact to compress the clothing.
Ponderal
index (PI) was computed
as heightinches/(weight-pounds)‘:3.
A 14 x 17” PA film of
the chest was taken at a 6 ft distance.
The
physician
made three blood pressure
determinations in the left arm over about 30 min with
the person sitting, recording diastolic pressure
as the cessation of sounds [1 11. The average of
the three blood pressure readings was used in
this analysis. The physician also performed
a
detailed cardiovascular
examination.
A technician obtained
a 12-lead electrocardiogram.
The purpose of these analyses was to replicate
findings reported for the ratio of waist girth to
hip girth. Since hip girth was not measured,
it was first estimated
by assuming
it to
approximate
an ellipse with major axis equal
to seat breadth
and minor
axis equal to
thigh clearance height using the formula:
hip
girth = 4.443 x [(thigh clearance
height/2)2 +
(seat breadth/2)2]“2.
To avoid confusion
of
methods with those of other reports in which
hip girth was measured, the termfat distribution
index was adopted for the following ratios.
One fat distribution
index (FDIl) was computed as waist circumference
divided by the
above indicator
of hip size. Validation
of this
approach was not possible in adults aged 18-79.
However, the FDIl was compared to measured
waist-to-hip
ratio (WHR) in a national sample
of 953 black and white 17- and 18-yr-olds
examined
in the HES Cycle III (19661970),
which used the same anthropometric
techniques. In this sample FDIl was well correlated
with WHR in males (r = 0.89, p = 0.001) and
less well correlated
with WHR
in females
(r = 0.74, p = 0.0001). FDIl averaged 0.09 f
SE 0.001 greater than WHR in males and
0.07 f SE 0.002 greater in females. A second
fat distribution
index (FD12) was derived as
follows. In the 17-18-yr-olds,
hip girth was
estimated from linear regression in which hip
girth was the dependent
variable
and thigh
clearance
and seat breadth,
with or without
their interaction,
were independent
variables.
For
males,
FD12 = waist
girth/(4.99 + 1.58
thigh
clearance + 1.81 x seat breadth).
For
females FD12 = waist girth/( 17.34 + 2.75 thigh
clearance + 0.95 x seat breadth). FD12 was also
well correlated
with WHR
in 17-18-yr-old
males
(r = 0.90, p = 0.0001)
and
females
(r = 0.82, p = 0.0001). Further FD12 averaged
only 0.0012 f SE 0.0009 greater than WHR in
males and 0.0006 f SE 0.0014 smaller than
WHR in females. As for FDIl,
validation
in
18-79-yr-olds
was not possible.
Blood
glucose
was
measured
by
the
Somogyi-Nelson
method on 3 ml of blood collected in vacutainers
containing
30 mg of sodium fluoride 1 hr after a 50 g oral glucose load
[ 121. Subjects with a clear history of diabetes
were excluded from the glucose tolerance test.
The persons diagnosed as having definite known
diabetes either reported they were on medication for diabetes or had elevated blood glucose
levels on the GTT [12]. Serum cholesterol was
measured
by a ferric chloride method at the
central Public Health Service laboratory
[7].
Details of the cardiovascular
examination,
disease classification and criteria were published
[13]. Briefly the diagnostic
criteria
were as
follows.
Hypertension
Dejinite hypertension.
160 mmHg or over sysor over diastolic.
Borderline
hypertension.
Below 160 mmHg
systolic and below 95 mmHg diastolic, but not
simultaneously
below both 140 and 90 mmHg.
Normotension.
Below both 140 mmHg systolic and 90 mmHg diastolic.
When aortic insufficiency was present or the
heart rate was under 60, hypertension
or borderline hypertension
was defined by the diastolic
pressure.
tolic or 95 mmHg
Hypertensive
Definite.
heart disease
One of the following:
(1) Hypertension
plus left bundle
branch
block or left ventricular
hypertrophy
(LVH) by ECG. (By voltage criteria when
35 yr of age or over. If under 35yr left
ventricular
or subendocardial
ischemia
must have been present in addition
to
LVH by voltage criteria. No person under
35 had hypertension
or borderline hypertension with this combination
of ECG
findings.)
Body Fat Distribution and Cardiovascular Disease
(2) Hypertension
plus LVH or general cardiac enlargement (GCE) by X-ray.
(3) A history of hypertension currently on
medication for hypertension, and LVH or
GCE by X-ray and/or LVH by ECG.
Suspect. One of the following:
(1) Borderline hypertension plus LVH by
ECG and/or LVH or GCE by X-ray.
(2) Borderline hypertension plus LVH or
GCE by X-ray.
Coronary heart disease
Definite. One of the following:
(1) Myocardial infarction on ECG and/or
definite angina (judgment of examining
physician). Angina was not ascribed to
coronary heart disease if aortic stenosis or
syphilitic heart disease was present.
(2) History of myocardial infarction in judgment of examining physician and either
LV ischemia or myocardial infarction on
ECG outside criteria.
Suspect. One of the following:
(1) History of myocardial infarction in judgment of examining physician with no
evidence of myocardial infarction or left
ventricular ischemia on the ECG.
(2) Suspect angina (judgment of examining
physician).
Sample weights were not used in this analysis.
Therefore, the results are not presented as
population estimates for the United States.
Descriptive statistics were computed by
standard methods. Pearson product moment
correlations were computed for body fat distribution indices and continuous variables.
The effect of potential confounders on these
associations was controlled by multiple linear
regression analysis [ 141.The association of body
fat distribution indices with dichotomous prevalence variables was evaluated by multiple logistic regression [15]. Age-adjustment was done by
analyses of covariance [16].
RESULTS
Means and medians of both indices increased
steadily with age in all sex, race groups, except
for no change or a decrease after age 55-64 in
black men and black women. In whites and
blacks, both indices were greater in men than in
women at each age indicating greater abdominal
423
relative to lower body obesity in males. For each
sex, the distributions were skewed toward
higher values and peaked. Mean and median fat
distribution indices by race were similar in black
men (BM) and white men (WM) but higher in
black women (BW) than white women (WW)
both overall and within IO-yr age groups up to
65-74 yr. After age 65, means and medians were
similar in black and white women and lower in
black than white men. Age-adjusted means of
FDIl and FD12 were as follows: WM 1.04,
0.89, BM 1.04, 0.88; WW 0.87, 0.65, BW 0.93,
0.68. In women the racial difference was highly
significant (p = 0.0001) for both indices. The
sex ratio (men/women) was greater in whites
than blacks.
The association of fat distribution with
several other demographic variables was examined within age, sex, race groups. Among men,
urban dwellers had similar fat distribution to
rural dwellers. Among women, rural dwellers
generally had higher indices: farm dwellers had
the highest values. Region was not consistently
related to fat distribution in either sex. Among
white and black women in each age group,
family income and educational attainment were
inversely related to FDIl and FD12 (JJ < 0.01).
Among men, education but not income was
inversely related to the indices in whites
(p < 0.01) but not blacks. In the 35544 and
45-54yr age groups, indices were only minimally higher in menopausal compared with
premenopausal women.
Univariate associations with risk factors
(SBP)
was
Systolic
blood
pressure
significantly correlated with fat distribution indices within sex-race groups and within most
age-sex groups. Correlation coefficients were as
follows: FDIl, WM 0.34, BM 0.35, WW 0.47,
BW 0.44; FD12, WM 0.35, BM 0.36, WW 0.52,
BW 0.48. Mean SBPs in first vs fourth quartile
of FDIl were 124 vs 142 in men and 118 vs 149
in women. Diastolic blood pressure (DBP) was
also sigificantly correlated with fat distribution
indices within sex-race groups and within most
age-sex groups. Correlation coefficients were as
follows: FDIl, WM 0.32, BM 0.31, WW 0.38,
BW 0.36; FD12, WM 0.34, BM 0.33, WW 0.43,
BW 0.42. Mean DBPs in first vs fourth quartiles
of FDIl were 74 vs 85 in men and 73 vs 87 in
women. Ponderal index and relative body
weight were significantly but not highly correlated with the indices in all sex, race and
age-sex groups: e.g. for FDIl with ponderal
RICHARD F. GILLUM
424
index, WM -0.58,
BM -0.61,
WW -0.55,
BW -0.39; relative body weight, WM 0.57, BM
0.61, WW 0.51, BW 0.40. The indices were
significantly
correlated
with serum cholesterol
in all sex-race
and in 18-34 and 35-54 yr
age-sex groups with coefficients for all ages as
follows: FDIl WM 0.24, BM 0.24, WW 0.31,
BW 0.30; FD12, WM 0.25, BM 0.25, WW 0.35,
BW 0.33. Correlations
were generally
lower
within age-sex
groups.
Mean cholesterol
in
quartiles 1 vs 4 of FDIl was 200 vs 230 in men
and 203 vs 243 in women. The indices were
significantly
correlated
with post-load
serum
glucose in all sex-race and many age-sex groups
as follows: FDIl WM 0.25, BM 0.27, WW 0.29,
BW 0.27, FDI2 WM 0.25, BM 0.26, WW 0.32,
BW 0.28. Correlations
were generally
lower
within age-sex groups. Mean post-load serum
glucose in quartile 1 vs quartile 4 of FDIl was
106 vs 134 in men and 114 vs 148 in women.
Subjects were also cross-classified
by quartile
of waist girth and quartile
of lower body
measurements
(the denominators
of FDIl and
FD12). For both men and women, the highest
means were generally found in subjects in the
highest quartiles of waist girth and the lowest
quartiles of lower body measurements
for systolic and diastolic blood pressure, serum cholesterol and serum glucose. The lowest levels were
Table
found in subjects in the lowest quartile of waist
girth with little variation
among quartiles of
lower body measurements.
Within the upper
three quartiles of waist girth, mean levels of
systolic blood pressure and serum glucose generally decreased
with increasing
lower body
measurements,
while diastolic blood pressure
and serum cholesterol
varied little. Within all
quartiles
of lower body measurements,
mean
levels of blood pressure,
serum glucose and
cholesterol increased with increasing waist girth.
In contrast,
relative body weight, triceps and
subscapular
skinfold thickness and arm girth
were directly related to lower body measurements independent
of waist girth and vice versa.
Association with risk factors after controlling for
confounding by multiple regression analysis
FDI 1 remained significantly
associated with
SBP and DBP after controlling
for age, sex,
race, ponderal
index, diabetes diagnosis
and
education:
SBP b = 27.96,
t = 19.14, p =
0.0001; DBP b = 16.83, t = 8.58, p =O.OOOl.
The indices were no longer significantly
associated with serum cholesterol
after controlling
for age, sex, race, ponderal
index, diabetes
diagnosis
and education:
b = 7.87, t = 1.04,
p = 0.2985. FDIl remained significantly associated with post-load serum glucose after control-
1. Age-adjusted
meant fat distribution
index by sex-race
cardiovascular
disease status (number of cases)
White
men
Hypertension:
Definite
1.0771
Borderline
Normotensive
Hypertensive
Definite
heart
(52)
1.027
(210)
1.074*
(172)
1.034
(1757)
1.049**
(71)
1.038
(156)
1.048”
(96)
1.065
1.043
(2512)
1.048’
(10)
1.040
(336)
1.083’
(37)
1.043
(2627)
I.5591
(3)
1.040
(355)
(100)
Diabetes mellitusf
Definite
None
1.058*
(346)
1.062
(458)
1.034
(1865)
disease3
None
White
women
0.921*
(433)
0.895
(357)
0.856
(2140)
and
Black
women
0.960*
(129)
0.911
(49)
0.908
(270)
disease3
None
Coronary
heart
Definite
Black
men
group
it
0.925*
0.959’
(264)
0.858
(2108)
(105)
0.902
(220)
0.890**
(55)
0.868
(2813)
0.943**
0.936*
0.926**
(56)
0.869
(2870)
(8)
0.922
(430)
(14)
0.923
(433)
*Difference among disease groups statistically
significant, p < 0.01.
**Difference among disease groups NS, p > 0.05.
tAge adjustment
by analysis of covariance.
$Cases with suspect disease, race other than white or black, or missing values
were excluded, hence Ns and totals vary.
Body Fat Distribution and Cardiovascular Disease
ling for age, sex, race and ponderal index;
b = 56.65, t = 8.02, p = 0.0001. The results were
similar for FD12.
Univariate associations with disease endpoints
Hypertension. Mean fat distribution index
adjusted for age was higher in persons with
definite hypertension than in normotensives
with an intermediate value for persons with
boderline hypertension in both sexes (Table 1).
The estimates for WM and BW in Table 1 must
be viewed with caution because of significant
interactions of age with hypertension class. All
data in Table 1 are for FDIl. Results of
significance tests were essentially the same for
FD12. Prevalence of definite hypertension by
quartile of FDIl was consistent with this association. The unadjusted relative risk of quartile 1
vs 4 was 8.8 in men and 24.5 in women.
Hypertensive heart disease. Mean FDIl adjusted for age was higher for cases of hypertensive heart disease (HTHD) than for persons
with no heart disease or no hypertensive heart
disease (Table 1). The estimates in Table 1 must
be viewed with caution because of a significant
interaction of age with HTHD. Analyses with
FD12 yielded essentially the same significance
testing results. Prevalence of HTHD by quartile
of FDIl was consistent with this association.
The unadjusted relative risk of quartile 1 vs 4
was 9.7 in men and 43.2 in women.
Coronary heart disease. The mean FDIl tended to be higher for cases of CHD than for
persons with no CHD in both sexes (Table 1).
FD12 was significantly higher among cases than
non cases in white women (0.673 vs 0.648,
p = 0.006). Other trends did not attain statistical significance. Prevalence of definite CHD by
quartile of FDIl also showed this trend. The
unadjusted relative risk of quartile 1 vs 4
(definite CHD vs no CHD) was 3.9 in men and
6.3 in women.
Diabetes. Mean FDIl adjusted for age was
higher for cases of diabetes than for noncases in
both sexes (Table 1). Similar results were obtained with FD12. Prevalence of diabetes by
quartile of FDIl confirmed this association.
Unadjusted relative risk of quartile 1 vs 4 was
5.9 in men and 6.3 in women.
Association of fat distribution with disease after
controlling multiple confounders by logistic multiple regression
FDIl
definite
remained significantly associated with
hypertension vs normotension
after
425
controlling age and ponderal index among most
sex-race groups: WM unadjusted FDIl beta
9.53, SE 0.71, p = 0.0000, adjusted FDIl beta
3.31, SE 0.95, p = 0.0005; BM unadjusted beta
8.20, SE 1.60, p = 0.0000, adjusted beta 1.88, SE
2.12, p = 0.377; WW unadjusted beta 12.34, SE
0.66, p = 0.0000, adjusted beta 4.08, SE 0.80,
p = 0.0000; BW unadjusted beta 10.37, SE 1.34,
p = 0.0000,
adjusted beta 4.76, SE 1.56,
p = 0.0023. Results were similar using FD12.
Ponderal index was associated with hypertension independent of FDI but was not an
important confounder.
FDI 1 remained significantly associated with
definite hypertensive heart disease vs no heart
disease after controlling age and ponderal index
among most sex-race groups: WM unadjusted
FDIl beta 10.39, SE 0.95, p = 0.0000, adjusted
FDIl beta 2.73, SE 1.31, p =0.0369; BM unadjusted beta 6.06, SE 1.74, p = 0.0005, adjusted beta -0.37, SE 2.61, p = 0.888; WW
unadjusted beta 13.07, SE 0.77, p = 0.0000,
adjusted beta 4.40, SE 0.96, p = 0.0000; BW
unadjusted beta 12.23, SE 1.56, p = 0.0000,
adjusted beta 4.74, SE 1.91, p = 0.013. Results
were similar using FD12.
FDI 1 remained generally not significantly
associated with definite CHD vs no CHD after
controlling age, race, SBP, serum cholesterol
and diabetes: men, unadjusted FDIl beta 5.90,
SE 1.OO,p = 0.0000; adjusted beta 2.44, SE 1.49,
p = 0.103; women, unadjusted FDIl beta 6.34,
SE I .04, p = 0.0000, adjusted FDIl beta - 0.25,
SE 1.46, p = 0.866. Results were similar using
FD12 except for a significant beta after adjustment (p = 0.028) in men. For definite and suspect CHD vs no CHD, the beta remained
significant for both indices after adjustment
only in men, e.g. unadjusted FDII beta 5.73,
SE 0.80, p = 0.0000, adjusted FDIl beta 2.63,
SE 1.19, p = 0.028. For definite myocardial
infarction only, the adjusted betas were all
nonsignificant in both sexes.
FDIl remained significantly associated with
definite diabetes vs no diabetes after controlling
age, race and ponderal index: men unadjusted
FDIl beta 8.47, SE 1.54, p = 0.0000, adjusted
FDIl beta 6.10, SE 2.23, p = 0.006; women
unadjusted beta 7.45, SE 0.98, p = 0.0000, adjusted beta 5.35, SE 1.32, p = 0.0001. Consistent
results were obtained with FD12.
DISCUSSION
In a large sample of the United States population, greater waist girth relative to hip and
426
RICHARD F. GILLUM
thigh measurements
was independently
associated with increased
prevalence
of definite
hypertension,
definite hypertensive heart disease
and diabetes mellitus. Greater abdominal
fat
distribution
was also independently
associated
with higher blood pressure and post-load serum
glucose concentration.
Fat distribution
was
not consistently
associated with coronary heart
disease prevalence
or serum cholesterol
concentration
independent
of age and
other
confounders.
These findings confirm those of other crosssectional studies of fat distribution,
risk factors
and cardiovascular
disease, extending
the previous observations
to larger numbers of men, to
nonobese persons from the general population,
and to blacks [l-6, 17-211. Of particular
interest, however, are comparisons
with prospective
studies of disease incidence.
A study of the
incidence of coronary heart disease and stroke
in Gothenburg,
Sweden, found significant associations of the ratio of waist to hip girth with
coronary
heart disease and stroke in men and
with myocardial infarction (but not angina pectoris) and stroke in women independent
of age,
body mass index and smoking [5,6]. The associations were no longer significant after controlling blood pressure and serum cholesterol
in
men and, for stroke, after controlling
serum
triglycerides
in women.
Differences
in study
design, fat distribution
measurement,
diagnostic
criteria, population,
and potential confounders
controlled,
may account for the failure of the
present analysis to replicate these findings. The
present findings are quite consistent
with the
Swedish incidence data for diabetes [19].
The mechanisms
by which increased abdominal fat deposits may affect blood pressure remain unknown.
Serum insulin may play a role
[22,23]. The abdominal
and upper body segment obesity pattern is associated with hypertrophy and greater sensitivity to lipolytic stimuli
of adipocytes in the abdominal
region, increased
plasma free fatty acid levels, with decreasing
insulin effect and hyperinsulinemia
and hypertriglyceridemia
[ 1-6, 17-l 91. The direct access of
abdominal
adipose tissue to the portal circulation may be important.
The role of sex hormones requires
further
study [ 171. The relationship
is unclear of abdominal
obesity to
central obesity of the upper trunk, which has
also been linked to elevated blood pressure and
metabolic
abnormalities
[18]. In the present
analysis, the ratio of waist to lower body measurements was significantly
associated with sys-
tolic and diastolic blood pressure independent
of upper body skinfolds (sum of subscapular
and triceps) in each sex-race group. Perhaps
abdominal
adipose tissue evolved early among
vertebrates
to be metabolically
active whereas
subcutaneous
adipose tissue evolved primarily
to serve homoiothermy
as an energy store and
insulator
[24]. In any case, a comprehensive
theory encompassing
the several body typing
systems that have been related to cardiovascular
risk would be a major contribution
[17-261.
Age, sex, race, income, education
and ponderal index were associated with fat distribution, consistent
with both environmental
and
genetic influences [17]. The greater relative abdominal obesity combined with greater overall
ponderosity
in black compared to white women
[9, IO] may contribute
to higher mortality patterns from obesity-related
diseases (hypertensive
disease, diabetes,
coronary
heart disease) in
black women than in white women [27].
The chief limitation
of this and other crosssectional studies is that of possible bias in the
ascertainment
of body fat distribution
prior to
disease occurrence which arises from the necessary assumption
that fat distribution
at the time
of examination
accurately reflects prior levels.
Another possible source of bias is the unknown
relationship
of chosen indices of fat distribution
to the relative mass of intradominal
and lower
body fat. Since WHR had been shown to be
related to cardiovascular
risk, indices were computed that were likely highly correlated
with
WHR. Unfortunately,
despite the high correlations
in 17-yr-olds,
lesser correlations
in
adults might be associated with biased results in
this study. Changes in posture and abdominal
wall tone with aging and obesity might be
important
in this regard. The consistency
of
findings with the diabetes incidence study and
other studies of hypertension
is reassuring
in
this regard. However, the lack of consistency of
findings with the CHD incidence study could be
due to bias. The diagnostic methods and criteria
used in 1960 might contribute
to such bias.
Different results might be obtained
if current
diagnostic
techniques
such as exercise stress
testing for CHD or echocardiography
for hypertensive
heart disease were used. Data on
reliability of observers has been published [13].
Most important
potential
confounders
were
controlled
except cigarette
smoking,
family
history,
and diet, which were unavailable.
However, controlling
smoking history did not
eliminate or increase the significant association
Body Fat Distribution
and Cardiovascular
of WHR with coronary heart disease and stroke
in the Swedish incidence study [5,6]. The large
sample size provided good statistical power for
most analyses. However power was limited for
coronary heart disease and diabetes analyses in
blacks and in women, so negative results may
have been due to chance. Positive results were
not likely due to chance since most significant p
values were well below 0.01. While causal inferences are premature,
it seems likely that relatively greater abdominal
obesity increases risk
of cardiovascular
disease in part by elevating
blood pressure, serum glucose and serum lipids
[ 171. Any effect on cardiovascular
disease occurrence independent
of these variables could be a
direct effect of a common cause (perhaps male
sex hormones) or an indirect effect mediated by
mrshanisms
as yet unknown.
Future research should include: (1) Further
evaluation of profiles of sex and other hormones
affecting fat metabolism
and blood pressure in
persons with abdominal
as compared to lower
body obesity. (2) Further evaluation
of in vitro
metabolic properties of fat cells from abdominal
vs lower body sites. (3) Further evaluation
of
the effects of weight reduction on fat distribution and metabolic and risk factor patterns in
subjects with abdominal
vs lower body obesity.
These should include insulin, renin, mineralocorticoid
hormones,
sympathetic
nervous activity, and blood volume. (4) Evaluation
of the
association of fat distribution
with HDL cholesterol, other lipid fractions and apolipoproteins.
(5) Evaluation
of the effects of puberty
and
menopause (natural or surgical) on fat distributions in longitudinal
studies. (6) Further evaluation of familial aggregation
of fat distribution
patterns and their association
with family history of hypertension,
cardiovascular
disease and
diabetes. (7) Further evaluation
of the association of fat distribution
with risk factors in
prepurbertal
children
and
adolescents.
(8)
Evaluation
of patterns of obesity in populations
with high prevalence of diabetes such as Pima
Indians.
(9) Measurement
of waist and hip
girth, and multiple skinfold sites in addition to
the conventional
measurements
of overall overweight and obesity in future national
health
examination
surveys and in prospective
epidemiologic studies of cardiovascular
disease and
diabetes to confirm the independent
association
of fat distribution
with disease incidence. (10)
Methodologic
studies to determine
the feasibility of self-measurement
and reporting
of
waist and hip girth in mail and telephone
Disease
427
health surveys. (11) Further
investigation
of
socio-economic,
cultural
and
nutritional
determination
of far distribution
patterns.
REFERENCES
1.
2.
3.
4.
5.
6.
I.
8.
9.
10.
11.
12.
13.
14.
15.
16.
Kissebah
AH, Vydelingum
N, Murray
R et al.:
Relation of body fat distribution
to metabolic complications of obesitv. J Chin Endocr Metab 54: 25&260,
1982
Hartz AJ, Rupley DC, Kalkhoff
RD, Rimrn AA:
Relationshin
of obesitv to diabetes: Influence of obesity level and body fat distribution.
Prev Med 12:
351-357, 1983
Hartz AJ, Rupley DC, Rimm AA: The association
of
girth measurements
with disease in 32,856 women. Am
J Epidemiol 119: 71-80, 1984
Krotkiewski
M, Bjorntorp
P, Sjostrom L, Smith U:
Impact of obesity on metabolism
in men and women.
Importance
of regional adipose tissue distribution.
J CIin Invest 72: 115&1162,
1983
Larsson B, Svardsudd
K, Welin L et al.: Abdominal
adipose tissue distribution,
obesity, and risk of cardiovascular disease and death: 13 year followup of participants in the study of men born in 1913. Br Med J
288: 1401-1404,
1984
Lapidus L, Bengtsson D, Larsson B et al.: Distribution
of adipose tissue and risk of cardiovascular
disease
and death: a 12 year followup of participants
in the
population
study of women in Gothenburg,
Sweden.
Br Med J 289: 1257-1261,
1984
U.S. National Health Survey: Plan and Initial Program
of the Health Examination -Survey. Washington,
DC:
Public Health Service. Health Statistics Series A4, 1962
(PHS publication
no. 584-A4)
National
Center for Health Statistics: Cycle I of the
Health Examination Survey: Sample and Response.
United States 1960-1962. Washington,
DC: Public
Health Service, Vital and Health Statistics, 1964, Series
11, No. 1 (PHS publication
no. 1000)
National Center for Health Statistics: Weight, Height
and Selected Body Dimensions of Adults. United States
1960-1962. Washington,
DC: Public Health Service,
Vital and Health Statistics, 1965, Series 11, No. 8 (PHS
publication
no. 1000)
National Center for Health Statistics: Skinfolds, Body
Girths, Biacromial Diameter and Selected Anthropometric Indices of Adults. United States, 1960-1962.
Washington,
DC: Public Health Service, Vital and
Health Statistics, 1970, Series 11, No. 35 (PHS publication no. 1000)
National
Center for Health Statistics: Hypertension
and Hypertensive Heart Disease in Adults, United
States 1960-1%2.
Washington,
DC: Public Health
Service, Vital and Health Statistics,
1966, Series 11,
No. 13 (PHS publication
no. 1000)
National Center for Health Statistics: Glucose Tolerance of Adults, United States l%O-1%2. Washington,
DC: Public Health Service, Vital and Health Statistics,
1964, Series 11, No. 2 (PHS publication
no. 1000)
National Center for Health Statistics: Heart Disease in
Adults. United States l-1%2.
Washington,
DC:
Public Health Service, Vital and Health Statistics,
1964, Series 11, No. 6 (PHS publication
no. 1000)
An&age
P: Statistical Methods in Medical Research.
New York: Wiley, 1971, p. 302
Harrell F: The LOGIST Procedure.
In SAS Supplemental Library User’s Guide, 1980 edn. Cary NC: SAS
Institute,
1980, p. 83
Freund RJ, Littel RC: SAS for Linear Models. Cary,
NC: SAS Institute, 1980, p. 191
428
17.
18.
19.
20.
21.
RICHARD F. GILLUM
Stern MP, Haffner SM: Body fat distribution
and
hyperinsulinemia
as risk factors for diabetes and cardi&ascular
disease. Arteriosclerosis 6: 123-I 30, 1986
Blair D. Habicht
J-P. Sims EAH.
Svlvester
D.
Abraham
S: Evidence fir an increased’ ri& for hyper:
tension with centrally located body fat and the effect
of race and sex on this risk. Am J Epidemiol 119:
526-540, 1984
Ohlson LO, Larsson
B, Svardsudd
K et a/.: The
influence of body fat distribution
on the incidence of
diabetes mellitus. 13.5 year followup of the participants in the study of men born in 1913. Diabetes 34:
1055-1058,
1985
Seidell JC, Bakx JC, DeBoer
E, Deurenberg
P.
Hautrast JGAJ: Fat distribution
of overweight persons
in relation to morbidity
and subjective health. Int J
Obesity 9: 363-374, 1985
Weinsier RL, Norris DJ, Birch R et al.: The relative
contribution
of body fat and fat pattern
to blood
pressure level. Hypertension 7: 578-585, 1985
22.
23.
24.
25.
26.
27.
Weinsier RL, Norris DJ, Birch R et al.: Serum insulin
and blood pressure in an obese population.
Int J
Obesity 10: 11-17, 1986
Fournier AM, Gadia MT, Kubrusly
DB, Skyler JS,
Sosenko JM: Blood pressure, insulin, and glycemia in
nondiabetic
subjects. Am J Med 80: 861-864, 1986
Vague J, Fenasse R: Comparative
anatomy of adipose
tissue. In Handbook of Physiology.
Reinhold
AE,
Cahill GF Jr (Eds). Washington,
DC: American
Physiological
Society, 1965, p. 25-36
Vague J: The degree of masculine differentiation
of
obesities: a factor determining
predisposition
to diabetes,
atherosclerosis,
gout,
and
uric calculous
diseases. Am J Clin Nutr 4: 2(t34, 1956
Damon A, Damon ST, Harpending
HC, Kannel WB:
Predicting coronary heart disease from body measurements. J Chron Dis 21: 781-802, 1969
National Center for Health Statistics: Health United
States, 1984. Washington,
DC: Government
Printing
Office, 1984 (DHHS publication
no. (PHS) 85-1232)