International Journal of Obesity (1999) 23, 395±402
ß 1999 Stockton Press All rights reserved 0307±0565/99 $12.00
http://www.stockton-press.co.uk/ijo
The association of body weight and
anthropometry with mortality in elderly men:
The Honolulu Heart Program
S Kalmijn1*, JD Curb2,3, BL Rodriguez2,3, K Yano2,3 and RD Abbott2,3,4
1
Department of Epidemiology and Biostatistics, Erasmus University Medical School, Rotterdam, The Netherlands; 2The Honolulu Heart
Program, Kuakini Medical Center, Honolulu, HI, USA; 3Division of Clinical Epidemiology, John A. Burns School of Medicine, University
of Hawaii at Manoa, Honolulu, HI, USA and 4Division of Biostatistics and Epidemiology, University of Virginia School of Medicine,
Charlottesville, VA, USA
OBJECTIVES: To assess the relationship of body weight and anthropometry to all-cause mortality in older men.
DESIGN: A prospective cohort study of 3741 elderly Japanese-American men, enrolled in the Honolulu Heart Program.
For this report, the follow-up began at baseline examinations (1991 ± 1993), when the men were aged 71 ± 93 y.
MEASUREMENTS: Variables of interest were body mass index (BMI), waist-to-hip ratio (WHR), and the sum of the
subscapular and triceps skinfold thickness. Possible confounders included age, education, physical activity index,
smoking, alcohol consumption, systolic and diastolic blood pressure, cholesterol, glucose and insulin concentrations.
RESULTS: After an average of 4.5 y of follow-up, 766 men (21%) had died. Higher BMI was associated with lower
adjusted mortality risks (relative risk (RR)) highest vs lowest quintile-based category ˆ 0.5, 95% con®dence interval
(CI): 0.4 ± 0.6, P-trend < 0.001). Results were independent of WHR, and did not change after excluding current and
former smokers or those who died within one year of follow-up. The relation between WHR and mortality appeared to
be U-shaped, but after adjustment for BMI, a higher WHR steadily increased the risk of dying (RR highest vs lowest
category ˆ 1.5, 95%CI: 1.1 ± 2.0, P-trend ˆ 0.004). Especially in subjects with a high BMI, there was a positive
association between WHR and mortality. The results for skinfold thickness were similar to the results for BMI, but
less strong.
CONCLUSIONS: In older men, BMI and skinfold thickness showed a consistent inverse association with mortality,
even after accounting for early mortality. The WHR, on the other hand, was positively related to mortality, especially
when BMI was high. Thus, excess abdominal fat mass (FM) warrants closer concern than being overweight, in terms
of affecting mortality in the elderly.
Keywords: body mass index; skinfold thickness; waist-to-hip circumference ratio; mortality; aging; epidemiology
Introduction
The prevalence of obesity among American people
has increased substantially during the past decade.1 A
recent meta-analysis of prospective cohort studies
among middle-aged subjects showed that the association between body mass index (BMI) and all-cause
mortality was U-shaped, and that the increased mortality among underweight subjects could only partly
be explained by smoking status and pre-existing illness.2 A number of studies examined the relationship
between BMI and mortality in the elderly. The results
were inconsistent, showing either a U-shaped
association,3 ± 6 an inverse association,7 ± 10 or no association at all.11,12
It has been suggested that a measure of body fat
distribution, such as the waist-to-hip circumference
ratio (WHR), may be a better predictor of mortality
*Correspondence: Sandra Kalmijn, Department of Epidemiology
and Biostatistics, Erasmus University Rotterdam, PO Box 1738,
3000 DR Rotterdam, The Netherlands.
Received 1 July 1998; revised 26 October 1998; accepted 19
November 1998
than the BMI. However, studies that investigated the
relationship between anthropometric measurements
and mortality in the elderly are limited. Among
middle-aged subjects it was found that, when both
BMI and WHR were taken into account, the WHR
was strongly and positively related to mortality,
whereas BMI was not or even inversely related to
mortality.13 ± 15 The exact association of body weight
and body fat distribution with mortality in the elderly
remains in doubt. The purpose of the current study
was to examine these associations in a prospective
population-based study of elderly men, enrolled in the
Honolulu Heart Program. The relation of BMI, WHR
and skinfold thickness to mortality was investigated,
while taking smoking and early mortality into
account.
Methods
The Honolulu Heart Program is a population-based
prospective study of coronary heart disease (CHD)
Body weight and mortality
S Kalmijn et al
396
and stroke among a cohort of 8006 Japanese-American men, who were born between 1900±1919 and
were living on the island of Oahu, Hawaii.16 Examinations were given at the time of study enrolment
between 1965 and 1968, and periodically during a
course of follow-up which has now entered its 33rd
year. Men living in a nursing home were included in
this study. The study was approved by an institutional
review committee and informed consent was obtained
from the study participants.
Study population
The baseline examinations for this study were given
between 1991±1993, when the men were aged 71±
93 y. At that time, 3741 (80%) of the 4676 men who
were still alive were examined. Information on BMI
was available for 3594 men, on skinfold measures for
3645 men, and on WHR for 3664 men. After the
baseline examination, up to six years of follow-up (on
average 4.5 y) were available to examine the relationship of body weight and anthropometry to the risk of
death, based on a comprehensive surveillance of death
certi®cates, hospital admissions and obituary notices.
Data collection
The 1991±1993 examinations included extensive
interviews and anthropometric, physiological and
other laboratory measurements. Methods used are
described in greater detail elsewhere.17 Weight was
measured to the nearest 0.1 of a kilogram with the
subjects wearing minimal clothing. Height was
obtained to the nearest centimeter without wearing
shoes. BMI was calculated as weight (kg) divided by
height squared (m2), and used as an index of generalized obesity. Triceps and subscapular skinfold thickness, as a measure of subcutaneous fat mass (FM),
was measured to the nearest millimeter with Lange
calipers, and the sum was used in our analyses.13,14
The WHR was considered an estimation of body fat
distribution. Waist and hip circumferences were measured to the nearest centimeter; the waist circumference at the level of the umbilicus with the subject
standing, and the hip circumference at the level of the
iliac crest.
Confounding and mediating variables that were
taken into account included age, years of education,
physical activity, alcohol consumption (ml=d), cigarette smoking (number of cigarettes=d), systolic and
diastolic blood pressure (mean of two measurements),
total cholesterol, fasting glucose and fasting insulin.
Furthermore, we examined effect modi®cation by
smoking status (never, former, current). An index of
physical activity was derived by summing the number
of hours per day spent in ®ve different activity levels
(basal, sedentary, slight, moderate and heavy) after
multiplication by a weighting factor based on the
amount of oxygen required to undertake the activity.18
After an overnight fast of 12 h, blood specimens
were obtained. Total cholesterol was measured
in plasma, using the same laboratory and standard
enzymatic methods as in the Cardiovascular Heart
Study.19,20 Glucose was measured with a Kodak Ektachem 700 analyzer with reagents (Eastman-Kodak,
Rochester, NY), and insulin was measured by a
double antibody radioimmunoassay (RIA) method21
at the University of Washington (Diabetes Endocrinology Research Center Core Radioimmunoassay
Laboratory, Seattle, WA) after storage at 770 C.
Further description of these and other characteristics
considered in this report is provided elsewhere.16,22
Statistical analysis
To allow for possible nonlinear relationships with
mortality, the indices of body weight and anthropometry were categorized into quintile-based categories.
In order to compare baseline characteristics across the
quintile-based categories of BMI, age-adjusted mean
levels of these risk factors were calculated for each
category.23 Age-adjusted absolute mortality rates
across the categories were calculated using logistic
regression models.23 To assess the adjusted relative
risk of dying according to body weight and anthropometry, multivariate Cox regression models were
used.24 Dummy variables for the quintile-based categories of BMI, skinfold thickness and WHR were
entered into the model separately and together.
Adjustments were made for all possible confounders
and for age alone. Each continuous measure of body
weight and anthropometry was also entered into the
model as both a linear and quadratic term. In addition,
mortality rates and relative risks were examined
according to baseline smoking status. Because subclinical underlying disease may confound the relationship between low body weight and mortality, we
repeated the analyses after exclusion of subjects who
died within one year of follow-up. We tested for
interaction between BMI and WHR, BMI and skinfold
thickness, and between WHR and skinfold thickness.
For graphical purposes, the combined effect of BMI,
WHR and skinfold thickness on mortality was
described across tertile combinations of the
measurements.
Results
The mean age of the participants at baseline was
77.7 y (s.d. ˆ 4.6). Table 1 gives mean levels and
percentiles of BMI, the sum of the subscapular and
triceps skinfold thickness, and the WHR according to
®ve-year age groups. The mean BMI decreased from
24.2 kg=m2 in men aged < 75 y to 21.9 kg=m2 in men
aged 85 y (P < 0.001). The other two measures
declined with increasing age as well (P < 0.001).
Table 2 describes the relationship between BMI,
WHR and the other factors considered in this report.
Systolic blood pressure, total cholesterol, fasting
Body weight and mortality
S Kalmijn et al
397
Table 1 Body mass index (BMI), skinfold thickness, and waist-to-hip circumference ratio (WHR) according to ®ve-year age groups
Percentiles
Age (y)
n
Mean (s.d.)
5th
25th
50th
75th
95th
2
BMI (kg=m )
71 ± 74
75 ± 79
80 ± 84
85 ± 93
Skinfold thickness (mm)a
71 ± 74
75 ± 79
80 ± 84
85 ± 93
WHR
71 ± 74
75 ± 79
80 ± 84
85 ± 93
1070
1476
676
372
24.2
23.6
22.8
21.9
(3.0)
(3.1)
(3.0)
(3.2)
19.5
18.6
17.9
16.6
22.2
21.5
20.7
19.8
24.3
23.5
22.7
21.8
26.1
25.5
24.9
24.0
28.9
28.8
27.7
27.5
1067
1498
682
397
27.7
26.7
25.1
22.6
(8.8)
(9.2)
(8.7)
(8.0)
15.0
13.5
13.0
10.0
21.0
21.0
19.0
16.0
27.0
26.0
24.0
22.0
33.0
32.0
30.0
27.0
42.0
42.0
40.0
37.0
1075
1503
687
399
0.95
0.95
0.94
0.93
(0.05)
(0.06)
(0.06)
(0.06)
0.86
0.85
0.84
0.83
0.92
0.91
0.90
0.90
0.95
0.95
0.94
0.94
0.98
0.98
0.97
0.97
1.04
1.04
1.03
1.03
s.d. ˆ standard deviation. aSum of the subscapular and triceps skinfold thickness.
glucose and fasting insulin were signi®cantly higher
in the upper, compared to the lower, quintile-based
category of BMI and WHR, after adjustment for
age. The number of cigarettes smoked per day
decreased with increasing BMI, but did not differ
across the WHR categories. The number of years of
education decreased with increasing BMI and WHR.
Furthermore, WHR was positively associated with
alcohol consumption and inversely with physical
activity. Similar patterns of association as described
for BMI were observed with skinfold thickness,
except for the fact that alcohol consumption decreased
with increasing skinfold thickness (P-trend < 0.001).
After an average of 4.5 y of follow-up (s.d. ˆ 1.3),
766 of the 3594 men (21.3%) had died. Overall
mortality rates increased from 10.8% in subjects
aged < 75 y to 50.5% in subjects aged 85 y
(P < 0.001). Since age-adjusted absolute mortality
rates were similar among former and current smokers,
we combined these two groups to increase statistical
power. The mortality rates decreased substantially
with increasing BMI among former or current smokers (32.6% in the lowest vs 16.9% in the highest BMI
quintile, P < 0.001) (Figure 1A). Among never smokers, the same downward trend was seen, although
less clear (21.7% vs 14.5%, respectively, P ˆ 0.002).
Similarly, age-adjusted mortality rates declined signi®cantly with increasing skinfold thickness, especially in former and current smokers (Figure 1B).
The association between WHR and absolute mortality,
however, appeared to be U-shaped among former and
current smokers (P-value for quadratic term ˆ 0.02),
with the lowest mortality rate in subjects in the third
quintile (19.2%) (Figure 1C). Among never smokers,
no clear association between WHR and absolute
mortality could be discerned.
To account for underlying diseases that could affect
both weight and mortality risk, we calculated the age-
Table 2 Age-adjusted mean baseline characteristics according to quintiles of body mass index (BMI) and waist-to-hip circumference
ratio (WHR)
Quintiles
Characteristics
2
BMI (range in kg=m )
Education (y)*
Physical activity index
Alcohol consumption (ml=d)
Number of cigarettes=d******
Systolic blood pressure (mmHg)*
Plasma total cholesterol (mg=dl)*
Fasting plasma glucose (mg=dl)******
Fasting plasma insulin (mU=ml)******
WHR (range)
Education (y)******
Physical activity index******
Alcohol consumption (ml=d)*
Number of cigarettes=d
Systolic blood pressure (mmHg)*
Plasma total cholesterol (mg=dl)******
Fasting plasma glucose (mg=dl)******
Fasting plasma insulin (mU=ml)******
1st
2nd
3rd
4th
5th
12.3 ± 20.7
10.7 (3.3)**
30.7 (4.6)
10.5 (34.7)
1.95 (6.4)
147.5 (25.6)
186.2 (33.4)
108.9 (27.1)
9.5 (6.1)
0.73 ± 0.90
10.9 (3.4)
31.3 (4.6)
5.7 (13.8)
1.06 (4.2)
147.3 (24.3)
186.5 (31.7)
108.6 (27.0)
13.4 (44.5)
20.8 ± 22.6
10.5 (3.2)
30.8 (4.3)
9.2 (23.2)
1.42 (6.2)****
148.6 (24.2)
190.4 (32.8)****
110.9 (26.4)
13.4 (18.5)***
0.90 ± 0.93
10.6 (3.2)****
31.4 (4.9)
9.0 (23.5)****
1.01 (4.9)
149.8 (23.9)****
188.3 (33.0)
112.7 (29.1)***
13.8 (11.1)
22.7 ± 24.1
10.5 (3.2)
31.0 (4.7)
7.5 (18.2)****
0.70 (3.6)*****
151.8 (22.6)*****
192.1 (31.8)*****
113.1 (30.5)***
17.1 (47.3)*****
0.93 ± 0.96
10.4 (3.2)***
30.7 (4.5)***
9.3 (27.8)***
1.09 (5.5)
149.7 (23.4)****
191.4 (33.7)***
113.6 (30.8)***
16.0 (22.7)
24.2 ± 25.9
10.5 (3.2)
31.0 (4.7)
7.8 (23.4)
0.73 (4.1)*****
151.2 (22.3)***
190.6 (32.6)****
114.7 (30.2)*****
17.7 (17.4)*****
0.96 ± 0.99
10.3 (3.1)*****
30.6 (4.8)***
9.5 (31.2)***
0.79 (3.6)
151.1 (23.6)***
191.8 (33.2)***
114.9 (29.7)*****
17.3 (13.1)***
26.0 ± 39.3
10.3 (3.1)***
30.6 (4.7)
8.1 (23.1)
0.32 (2.8)*****
150.6 (21.9)****
192.0 (33.1)***
118.0 (31.0)*****
24.1 (25.7)*****
0.99 ± 1.27
10.1 (3.0)*****
29.9 (4.1)*****
9.2 (24.7)****
1.17 (5.5)
150.0 (22.4)****
191.9 (32.9)***
115.5 (29.7)*****
21.3 (29.3)*****
*P < 0.05 by test for trend; **Standard deviation between parenthesis; ***P < 0.01 for difference with lowest tertile; ****P < 0.05 for
difference with lowest tertile; *****P < 0.001 for difference with lowest tertile; ******P < 0.001 by test for trend.
Body weight and mortality
S Kalmijn et al
398
Figure 1 Age-adjusted absolute mortality rates according to
quintiles of body mass index (BMI), skinfold thickness and
waist-to-hip ratio (WHR), in former or current and never smokers.
adjusted mortality rates after exclusion of subjects
who died within one year of follow-up. After these
exclusions were made, there were no marked changes
in the patterns shown in Figure 1, although death rates
were reduced by 9% on average.
Table 3 gives the results of the Cox regression
analyses, which showed an inverse association
between BMI and mortality. After adjustment for
age, subjects in the highest quintile of BMI had a
50% lower risk of dying than those in the lowest
quintile (95% con®dence interval (CI): 0.4±0.6) with
a signi®cant trend (P < 0.001). With increasing
skinfold thickness, the mortality risk decreased as
well, although slightly less strongly (Relative risk
(RR) for highest vs lowest quintile ˆ 0.6, 95%CI:
0.5±0.7, P-trend < 0.001). The association between
WHR and mortality appeared to be U-shaped. Subjects with a WHR between 0.93±0.96 (third quintile)
had the lowest age-adjusted mortality risk compared
to the lowest quintile (RR ˆ 0.8, 95% CI: 0.7±1.0).
Adding a quadratic term yielded a P-value of 0.009.
When we included age squared or other confounding
factors in the model, the relative risks were essentially
the same as after adjustment for age alone. When we
excluded ever smokers and those who died during the
®rst year of follow-up (n ˆ 2407), the adjusted RR
was lowest in the fourth quintile of BMI compared to
the ®rst quintile (RR ˆ 0.6, 95% CI: 0.4±1.0) and the
P-value for trend was still signi®cant (P ˆ 0.005). In
this latter analysis, a quadratic effect of the WHR on
mortality was no longer statistically signi®cant
(P ˆ 0.5).
To examine the independent effects of BMI, WHR
and skinfold thickness, all indices were entered into
one model together. The association between BMI and
mortality remained an inverse one (RR highest vs
lowest
quintile ˆ 0.4,
95%CI:
0.3±0.6,
Ptrend < 0.001). On the other hand, a positive association was observed between WHR and mortality (RR
highest vs lowest quintile ˆ 1.5, 95%CI: 1.1±2.0, Ptrend ˆ 0.004), which did not change after exclusion
of former or current smokers. Skinfold thickness was
not related to mortality after adjustment for BMI and
WHR. There was a signi®cant interaction between
WHR and BMI (P < 0.001), between BMI and skinfold thickness (P ˆ 0.003), and between WHR and
skinfold thickness (P ˆ 0.003).
Figure 2 provides a graphical measure describing
these latter ®ndings. Although BMI was inversely
associated with mortality within each tertile of the
WHR (P < 0.001 for each WHR tertile), the WHR
appeared to have its strongest association with mortality among individuals in the highest tertile of BMI
(Figure 2A). The WHR was positively related to
mortality in the medium and high BMI tertiles
(P ˆ 0.02 and P < 0.001, respectively). Among subjects with a low BMI, the mortality rates were lowest
in the middle tertile of the WHR, meaning the
association was U-shaped. The quadratic term was
however not signi®cant (P ˆ 0.83). The association
between WHR and mortality in the highest and middle
BMI tertile was signi®cantly stronger, compared with
the association in the lowest tertile (P < 0.001 and
P < 0.001, respectively).
An inverse association could be observed between
BMI and mortality in the lowest two tertiles of
skinfold thickness (P < 0.001), but not when skinfold
thickness was high (P ˆ 0.54) (Figure 2B). Skinfold
thickness was inversely related to mortality in the
lowest tertile of BMI only (P < 0.001). The WHR was
positively related to an increased risk of dying in the
highest tertile of skinfold thickness (P < 0.001)
Body weight and mortality
S Kalmijn et al
399
Table 3 Adjusted relative risks (95% con®dence intervals (CI)) of all-cause mortality according to quintiles of body mass index (BMI),
skinfold thickness and waist-to-hip circumference (WHR)
Quintiles
2nd vs1st
3rd vs1st
4th vs 1st
5th vs1st
P trend
0.6
(0.5 ± 0.7)
0.5
(0.4 ± 0.7)
0.8
(0.5 ± 1.2)
0.6
(0.5 ± 0.7)
0.6
(0.4 ± 0.7)
0.8
(0.5 ± 1.3)
0.5
(0.4 ± 0.7)
0.6
(0.5 ± 0.7)
0.6
(0.4 ± 1.0)
0.5
(0.4 ± 0.6)
0.5
(0.4 ± 0.6)
0.6
(0.4 ± 1.1)
< 0.001
0.8
(0.7 ± 1.0)
0.8
(0.7 ± 1.1)
1.2
(0.7 ± 1.8)
0.7
(0.6 ± 0.9)
0.8
(0.6 ± 1.0)
1.0
(0.6 ± 1.6)
0.6
(0.5 ± 0.7)
0.7
(0.5 ± 0.9)
1.1
(0.7 ± 1.7)
0.6
(0.5 ± 0.7)
0.6
(0.5 ± 0.8)
0.6
(0.3 ± 1.1)
< 0.001
0.9
(0.7 ± 1.1)
0.9
(0.7 ± 1.1)
1.1
(0.7 ± 1.7)
0.8
(0.7 ± 1.0)
0.8
(0.6 ± 1.0)
1.0
(0.6 ± 1.6)
0.9
(0.7 ± 1.1)
0.9
(0.7 ± 1.1)
1.1
(0.7 ± 1.7)
1.1
(0.9 ± 1.3)
1.0
(0.8 ± 1.3)
1.2
(0.7 ± 1.9)
0.55
2
BMI (kg/m )
Age-adjusted
Adjusteda
Adjusteda, excluding ever
smokers and early deathsb
Skinfold thickness (mm)c
Age-adjusted
Adjusteda
a
Adjusted , excluding ever
smokers and early deathsb
WHR
Age-adjusted
a
Adjusted
Adjusteda, excluding ever
smokers and early deathsb
< 0.001
0.005
< 0.001
0.11
0.74
0.92
a
Adjusted for age, years of standard education, physical activity index, alcohol consumption, number of cigarettes/d, systolic blood
pressure, diastolic blood pressure, serum total cholesterol level, fasting serum glucose and insulin concentrations. bSubjects who died
within one year of follow-up, and former and current smokers were excluded. Data on 1187 men for BMI, 1189 men for skinfold
thickness and 1195 men for WHR. cSum of subscapular and triceps skinfold thickness.
(Figure 2C). The association in the lowest skinfold
thickness tertile seemed to be U-shaped, but the
quadratic term was not signi®cant (P ˆ 0.84). Skinfold
thickness showed an inverse relationship with mortality in the top and bottom WHR tertiles (P < 0.001 and
P ˆ 0.03, respectively) (Figure 2C). The association in
the middle tertile was not signi®cant (P ˆ 0.16).
Discussion
In this study of older Japanese-American men, mortality risk decreased with increasing BMI, which could
not be explained by current or former smoking, or by
early mortality. Moreover, this inverse association was
independent of WHR and skinfold thickness. Skinfold
thickness showed a similar inverse relation to mortality, although less consistently. On the other hand, in
subjects with a high BMI or skinfold thickness, the
WHR was positively related to mortality risk.
One of the strong points of the present study is that
the participation rate was high and the loss to followup minimal, due to the low out-migration rate of about
one per thousand per year. We studied a representative sample of the general population, since we
examined men living at home and in the nursing
home. Furthermore, we had information on different
measures of body weight and body fat distribution,
whereas most studies among elderly only examined
the association between BMI and mortality. The
indicators of body weight and body fat distribution
were measured in a standardized way, by specially
trained nurses and technicians. Many confounders and
intermediates could be taken into account. We
adjusted for current, and also for former smoking, as
it was shown that among former smokers the association between body weight and mortality was similar to
that among current smokers. Finally, we recalculated
the analyses, after excluding the men who died within
one year of follow-up, which may reduce the effect of
sub-clinical disease on the relation between low BMI
and mortality.
Although there are many studies that examine the
relationship between BMI and mortality in middleaged subjects,2 only some investigated this in the
elderly.3 ± 12 Most of the studies among subjects aged
> 65 y, observed a U-shaped association,3 ± 6 but some
of them used self-reported weight and height measures,3,6 or did not adjust for smoking status.3 A
number of studies found an inverse association
between BNI and mortality in the elderly,7 ± 10 whereas
others found no association.11,12 A previous report
from the Honolulu Heart Program10 showed that
mortality rates were highest among men with reduced
pulmonary function and a low BMI, which to some
extent could be explained by smoking behavior. A
Norwegian study among 1.7 million people showed
that the U-shaped association between BMI and mortality in middle-aged subjects changed to an almost
horizontal association in subjects aged > 75 y.11 However, no adjustments could be made for smoking or
other confounders. A study on participants from
EPESE who were aged > 70 y, observed an inverse
Body weight and mortality
S Kalmijn et al
400
Figure 2 Age-adjusted absolute mortality rates according to
tertiles of body mass index (BMI), waist-to-hip ratio (WHR) and
skinfold thickness.
association between self-reported BMI and mortality,
which disappeared after exclusion of subjects who
died within three years, or who had physical disabilities.12 Two Finnish studies observed results comparable to ours.7,8 The highest BMI was associated with
the lowest mortality risk among non-institutionalized
subjects between the ages of 84±88 y,7 and among a
general population aged 85 y.8 Both studies did not
adjust for smoking, but the frequency of current and
former smoking was low. A recent study examining
almost 80 risk factors for mortality among subjects
aged 65 y showed an independent association between
low body weight and an increased risk of dying.9
Studies on the association between WHR and allcause mortality among middle-aged subjects are
scarce,13 ± 15 and even more so among the elderly.
Findings suggested that when both BMI and WHR
were taken into account, the WHR was a better
predictor of mortality than BMI. In accordance with
our results, a study among Iowan women aged 55±
69 y, found a positive association between WHR and
mortality, whereas the association between BMI and
mortality was inverse when these measures were put
together in a multivariate regression model.15 The
study was not population-based, and BMI and WHR
were self-reported and self-measured, though. Two
population-based studies from Sweden among middleaged men13 and women14 showed an increased mortality with higher WHR after smoking and BMI had
been taken into account, whereas the association
between BMI and mortality was absent.
We found that the risk of dying decreased steadily
with rising BMI. A possible explanation for the lack
of an increased mortality risk among people with a
high BMI may be that there were few obese participants in the Honolulu Heart Program, compared to
most populations in the USA. The average BMI in the
highest quintile was 27.9 kg=m2 and the 95th percentile was 28.9 kg=m2. Several studies suggested that the
optimal BMI among older subjects is shifted upwards,
and ranges from approximately 27±31 kg=m2.25
Another reason could be selective survival and nonresponse of heavy subjects, that is, those who were
susceptible to the adverse effects of high body weight
have already died at a younger age or are at higher
risk of dying, leaving only the less susceptible heavy
subjects. Finally, among older persons, a high BMI
seems to be less strongly related to the mediating
factors that are related to mortality, such as dyslipidaemia, hypertension and insulin resistance.26 In addition, these risk factors may have an altered relative
effect on mortality in the elderly, thereby changing the
effect of body weight on mortality.
A number of explanations can be offered for the
increased mortality risk among people with a low
BMI. Leanness may be a real risk factor, because of
nutrient de®ciency, frailness (including a higher risk
of hip fractures) and reduced functional status. A low
dietary intake has been related to impairment of the
immune system in rats.27 A low body weight may be a
risk indicator as well, re¯ecting underlying illness,
which is subsequently related to mortality. Although
the exclusion of subjects who died within one year of
follow-up did not essentially change our results, the
association may still be confounded by underlying
disease.28 Especially in the elderly, subjects with a
low BMI are a heterogeneous group. Low BMI can
result from illness or loss of muscle mass, but it also
may re¯ect low FM. Therefore, it is dif®cult to
interpret the risks associated with a low BMI.
Both WHR and skinfold thickness are less determined by fat-free mass (FFM) than BMI. The WHR is
a re¯ection of body fat distribution and was positively
related to mortality. The sum of the triceps and
subscapular skinfold thickness can be seen as a
measure of subcutaneous fat, and was inversely associated with mortality, although not as strong as BMI.
Body weight and mortality
S Kalmijn et al
This implies that abdominal or visceral FM is a
more important determinant of mortality in the elderly
than subcutaneous fat. This is in agreement with the
fact that the WHR is more strongly related to the
metabolic consequences of adiposity, such as high
cholesterol, hyperinsulinaemia, and hypertension,
than BMI.29 In addition, Japanese-Americans may
be particular susceptible to the input of WHR, as a
previous paper on Japanese-American men found that
the association between WHR and coronary heart
disease was independent of BMI, whereas the association between BMI and coronary heart disease was not
independent of abdominal adiposity.30 It has been
suggested that waist circumference alone is a better
prediction of visceral adiposity than the WHR.31 In
addition, the use of a ratio does not always guarantee
adequate control for confounding and may hamper the
interpretation of the results.32 However, in this sample
of Japanese-American men, the waist circumference
yielded essentially the same results as the WHR. For
comparability with other studies, we chose to present
the results on WHR instead of waist circumference
alone.
Conclusion
This study suggests that, in the elderly, a low BMI is a
strong predictor of increased mortality, regardless of
the WHR. The mortality risk was not increased in
subjects with a high BMI or skinfold thickness. In
contrast, a high WHR was related to an increased risk
of dying in older subjects, especially when BMI or
skinfold thickness was high. Clinicians should be
aware of the fact that the elderly with a low BMI,
may have an increased mortality risk and that older
overweight people may bene®t from reducing their
abdominal FM.
Acknowledgements
This study was supported by Contract NO1-HC-05102
from the National Heart, Lung and Blood Institute, by
a Research Centers in Minorities Institutions Award
(P20 RR 11091) from the National Institute of Health,
and by the Paci®c Health Research Institute with
funding from the George F. Straub Trust and the
Robert E. Black Fund of the Hawaii Community
Foundation.
References
1 Kuczmarski RJ, Flegal KM, Campbell SM, Johnson CL.
Increasing prevalence of overweight among US adults. The
National Health and Nutrition Examination Surveys, 1960 to
1991. JAMA 1994; 272: 205 ± 211.
2 Troiano RP, Frongillo Jr EA, Sobal J, Levitsky DA. The
relationship between body weight and mortality: a quantitative
analysis of combined information from existing studies. Int J
Obes 1996; 20: 63 ± 75.
3 Allison DB, Gallagher D, Heo M, Pi-Sunyer FX, Heyms®eld
SB. Body mass index and all-cause mortality among people
age 70 and over: the Longitudinal Study of Aging. Int J Obes
1997; 21: 424 ± 431.
4 Rissanen A, HelioÈvaara M, Knekt P, Aromaa A, Reunanen A,
Maatela J. Weight and mortality in Finnish men. J Clin
Epidemiol 1989; 42: 781 ± 789.
5 Harris T, Cook EF, Garrison R, Higgins M, Kannel W,
Goldman L. Body mass index and mortality among nonsmoking older persons. JAMA 1988; 259: 1520 ± 1524.
6 Dorn JM, Schisterman EF, Winkelstein Jr W, Maurizio T.
Body mass index and mortality in a general population sample
of men and women. The Buffalo Health Study. Am J Epidemiol 1997; 146: 919 ± 931.
7 Rajala SA, Kanto AJ, Haavisto MV, Kaarela RH, Koivunen
MJ, Heikinheimo RJ. Body weight and the three-year
prognosis in very old people. Int J Obes 1990; 14: 997 ±
1003.
8 Mattila K, Haavisto M, Rajala S. Body mass index and
mortality in the elderly. BMJ 1986; 292: 867 ± 868.
9 Fried LP, Kronmal RA, Newman AB, Bild DE, Mittelmark
MB, Polak JF, Robbins JA, Gardin JM. Risk factors for 5-year
mortality in older adults. The Cardiovascular Health Study.
JAMA 1998; 279: 585 ± 592.
10 Sharp DS, Burch®el CM, Curb JD, Rodriguez BL, Enright PL.
The synergy of low lung function and low body mass index
predicting all-cause mortality among older Japanese-American
men. J Am Geriatr Soc 1997; 45: 1464 ± 1471.
11 Waaler HTh. Height, weight and mortality. The Norwegian Experience. Acta Medica Scand Suppl 1984; 679: 1 ±
56.
12 Losonczy KG, Harris TB, Cornoni-Huntley J, Simonsick EM,
Wallace RB, Cook NR, Ostfeld AM, Blazer DG. Does weight
loss from middle age to old age explain the inverse weight
mortality relation in old age? Am J Epidemiol 1995; 141: 312 ±
321.
13 Larsson B, SvaÈrdsuud K, Welin L, Wilhelmsen L, BjoÈrntorp P,
Tibblin G. Abdominal adipose tissue distribution, obesity and
risk of cardiovascular disease and death: 13 year follow up of
participants in the study of men born in 1913. BMJ 1984; 288:
1401 ± 1405.
14 Lapidus L, Bengtsson C, Larsson B, Pennert K, Rybo E,
SjoÈstroÈm L. Distribution of adipose tissue and risk of cardiovascular disease and death: a 12 year follow up of participants
in the population study of women in Gothenburg, Sweden.
BMJ 1984; 289: 1257 ± 1261.
15 Folsom AR, Kaye SA, Sellers TA, Hong C-P, Cerhan JR,
Potter JD, Prineas RJ. Body fat distribution and 5-year risk of
death in older women. JAMA 1993; 269: 483 ± 487.
16 Yano K, Reed DM, McGee DL. Ten-year incidence of
coronary heart disease in the Honolulu Heart Program: relationship to biologic and lifestyle characteristics. Am J Epidemiol 1984; 119: 653 ± 666.
17 Kagan A, Harris BR, Winkelstein W Jr, Johnson KG, Kato H,
Syme SL, Rhoads GG, Gay ML, Nichaman MZ, Hamilton HB,
Tillotson J. Epidemiologic studies of coronary heart disease
and stroke in Japanese men living in Japan, Hawaii and
California: demographic, physical, dietary and biochemical
characteristics. J Chron Dis 1974; 27: 345 ± 364.
18 Kannel WB, Sorlie P. Some health bene®ts of physical
activity: the Framingham Study. Arch Intern Med 1979; 139:
857 ± 861.
19 Erringer WH, Wahl PW, Kuller LH, Bush TL, Tracy RP,
Manolio TA, Borhani NO, Wong ND, O'Leary DH. Lipoprotein lipids in older people: results from the Cardiovascular
Health Study. Circulation 1992; 86: 858 ± 869.
20 Fried LP, Borhani NO, Enright P, Furberg CD, Gardin JM,
Kronmal RA, Kuller LH, Manolio TA, Mittelmark MB,
Newman A, et al. The Cardiovascular Health Study; design
and rationale. Ann Epidemiol 1991; 1: 263 ± 276.
21 Morgan CR, Lazarov A. Immunoassay of insulin: two antibody systems. Diabetes 1963; 21: 115 ± 126.
401
Body weight and mortality
S Kalmijn et al
402
22 Burch®el CM, Abbott RD, Sharp DS, Curb JD, Rodriguez
BL, Yano K. Distribution and correlates of lipids and
lipoproteins in elderly Japanese-American men. The Honolulu
Heart Program. Arterioscler Thromb Biol 1996; 16:
1356 ± 1364.
23 Lane PW, Nelder JA. Analysis of covariance and standardization as instances of prediction. Biometrics 1982; 38:
613 ± 621.
24 Cox DR. Regression models in life-tables. J Roy Stat Soc [B]
1972; 34: 187 ± 202.
25 Kushner RF. Body weight and mortality. Nutr Rev 1993; 51:
127 ± 136.
26 Egan BM, Bassett DR, Block WD. Comparative effects of
overweight on cardiovascular risk in young versus older men.
Am J Cardiol 1991; 67: 248 ± 252.
27 Konno A, Utsuyama M, Kurashima C, Kasai M, Kimura S,
Hirowaka K. Effects of a protein-free diet or food restriction
on the immune system of Wistar and Buffalo rats at different
ages. Mech Ageing Dev 1993; 72: 183 ± 197.
28 Allison DB, Heo M, Flanders DW, Faith MS, Williamson DF.
Examination of ``early mortality exclusion'' as an approach to
control for confounding by occult disease in epidemiologic
studies of mortality risk factors. Am J Epidemiol 1997; 146:
672 ± 680.
29 BjoÈrntorp P. The association between obesity, adipose tissue
distribution and disease. Acta Med Scand Suppl 1988; 723:
121 ± 134.
30 Huang B, Rodreiguez BL, Burch®el CM, Chyou PH, Curb JD,
Sharp DS. Associations of adiposity with prevalent coronary
heart disease among elderly men: the Honolulu Heart Program. Int J Obes 1997; 21: 340 ± 348.
31 Lemieux S, Prud'homme D, Bouchard C, Tremblay A,
DespreÂs J-P. A single threshold value of waist girth identi®es
normal-weight and overweight subjects with excess visceral
adipose tissue. Am J Clin Nutr 1996; 64: 685 ± 693.
32 Allison DB, Paultre F, Goran MI, Poehlman ET, Heyms®eld
SB. Statistical considerations regarding the use of ratios to
adjust data. Int J Obes 1995; 19: 644 ± 652.