American Journal of Epidemiology
Copyright O 1997 by The Johns Hopkins University School of Hygiene and Public Health
All rights reserved
Vol. 145, No 7
Printed In USA
Relation between Body Mass Index and Body Fat in Black Population
Samples from Nigeria, Jamaica, and the United States
Amy Luke,1 Ram6n Durazo-Arvizu,1 Charles Rotimi,1 T. Elaine Prewitt,1 Terrence Forrester,2 Rainford Wilks,2
Olufemi J. Ogunbiyi,3 Dale A. Schoeller,4 Daniel McGee,1 and Richard S. Cooper1
blacks; body composition; body mass index; obesity
The prevalence of obesity is increasing in both industrialized nations and those undergoing alterations
in diet and activity patterns as a result of increasing
westernization (1). It has been shown that increases in
the prevalence of obesity within a population often
precede a rise in the occurrence of chronic diseases,
such as diabetes mellitus and hypertension (2). Also of
concern to public health researchers in developing
countries is the decrease in body weight and body fat
associated with inadequate food and energy supplies
on a national level (3). To reliably measure and monitor levels of obesity and/or undernutrition within and
between populations, easy-to-use methods of determining body composition are needed.
To date, the most commonly used method of determining relative fatness has been the body mass index
(BMI). Although an often-cited definition of obesity is
the 95th percentile of BMI from the First National
Health and Nutrition Examination Survey (NHANES
I) (4), the use of BMI in this context has been controversial, because the correlation between BMI and
body fat typically is less than 0.60 (5-9). BMI, expressed as weight/height2, was not originally intended
as a measure of adiposity but rather as a means of
comparing body weights independently of height (10).
Adiposity, however, irrespective of weight, is believed
to be a primary risk factor for diabetes and cardiovascular disease (11), providing a rationale for the use of
methods which measure body composition directly.
Debate over the value of BMI for the estimation of
body fat has recently led some investigators to recommend the use of new technologies for the direct measurement of body fat levels in epidemiologic research
(8). For epidemiologic studies with large samples,
bioelectrical impedance analysis (BIA) appears to be
an ideal tool. Based on the principles governing the
electrical impedance of body tissues, BIA provides a
rapid, noninvasive, and relatively accurate and precise
estimation of total body water, from which body composition is derived (12).
The purpose of this study was to compare the relations between BMI and percentage of body fat, as
measured by BIA, in populations with similar genetic
backgrounds but different environments. BIA was validated and subsequently used to estimate body com-
Received for publication February 22, 1996, and in final form
October 29, 1996.
Abbreviations. BIA, bioelectrical impedance analysis; BMI, body
mass index; Cl, confidence interval; NHANES, National Health and
Nutrition Examination Survey.
1
Department of Preventive Medicine and Epidemiology, Stntch
School of Medicine, Loyola University, Maywood, IL
2
Tropical Metabolism Research Unit, University of the West Indies, Mona, Kingston, Jamaica
3
University College Hospital, University of Ibadan, Ibadan, Nigeria.
4
Department of Medicine, Biological Sciences Division, University of Chicago, Chicago, IL
620
Downloaded from http://aje.oxfordjournals.org/ at Pennsylvania State University on February 26, 2014
Body mass index (BMI) is the most commonly used measure of obesity. Recently, some investigators have
advocated direct measurement of adiposity rather than use of the BMI. This study was undertaken to
determine the ability of BMI to predict body fat levels in three populations of West African heritage Irving in
different environments. A total of 1,054 black men and women were examined in Nigeria, Jamaica, and the
United States during 1994 and 1995. A standardized protocol was'used to measure height, weight, waist and
hip circumferences, and blood pressure at all sites; percentage of body fat was estimated using bioelectrical
impedance analysis. Percentage of body fat and BMI were highly correlated within site- and sex-specific
groups, and the resulting r2 ranged from 0.61 to 0.85. The relation was quadratic in all groups except Nigerian
men, in whom it was linear. The regression coefficients were similar across sites, yet the mean body fat levels
differed significantly (p < 0.001) as estimated by the intercept, making intersite comparison difficult. Compared with BMI, percentage of body fat was not a better predictor of blood pressure or waist or hip
circumference. Am J Epidemiol 1997; 145:620-8.
Body Mass Index and Body Fat in Black Populations
position among three distinct populations of West
African descent, and the relation between BMI and
body fat content was determined for each site- and
sex-specific group.
Well-described relations exist between BMI and
blood pressure (13, 14) and between BMI and measures of regional fat distribution (15, 16). It is not
obvious, however, whether it is weight or adiposity
which influences these relations. For this reason, we
also assessed the degree to which BMI and percentage
of body fat explained the between-person variance
observed in blood pressure and anthropometric measurements in these populations.
Sampling procedure
The study was conducted at three field sites of the
International Collaborative Study on Hypertension in
Blacks: Idikan, Ibadan, Nigeria (n = 314); Spanish
Town, Jamaica (n = 242); and Maywood, Illinois,
United States (n = 335). An additional 181 individuals (65 men and 116 women) from the US site were
participants in a clinical trial of the reduction of dietary fat, the Fat Reduction Intervention Trial in African Americans. Data presented here were obtained at
baseline, prior to the intervention. Participant recruitment for the intervention trial was conducted using the
same sampling strategy as that used in the hypertension survey.
The cluster sampling technique—probability proportional to size—was used at all sites to recruit equal
proportions of adult men and women aged 25 years or
over. Clusters (e.g., city blocks in the United States,
enumeration districts in Jamaica, and villages in rural
Nigeria) were identified, and the approximate sizes of
the populations were ascertained. After random selection of appropriate clusters, all black adults residing
within the clusters were asked to participate in the
survey. The final number of clusters selected was
determined by the sizes of the clusters and the participation rate at each site. Details on the sampling strategy are given elsewhere (17).
Body measurements
BIA, anthropometric data, and blood pressure were
measured using standardized procedures by field staff
local to each site who had been trained by centrally
certified supervisors (17). Body composition was determined by BIA. BIA measures the impedance of
body tissues to the flow of an applied mild alternating
current. Impedance is a function of two components,
the resistance of the tissues and the additional opposition, called reactance, due to the capacitant effect of
Am J Epidemiol
Vol. 145, No. 7, 1997
cell membranes, tissue interfaces, and nonionic tissues
(12). The measured impedance of body tissues provides an estimate of total body water from which
fat-free mass and fat mass can be calculated.
All BIA measurements were performed using a
single-frequency (50-kHz) battery-operated bioimpedance analyzer (model BIA 101Q; RJL Systems, Clinton Township, Michigan). A tetrapolar placement of
electrodes was used, with current electrodes placed on
the dorsal surfaces of the right hand and foot at the
distal metacarpals and metatarpals, respectively, and
the detector electrodes placed at the pisiform prominence of the right wrist and between the medial and
lateral malleoli at the right ankle (18).
The deuterium dilution method (19) was used for
validation of BIA estimation of body fat levels. In a
separate sample of 15 adults in Nigeria, 21 in Jamaica,
and 8 in the United States, representing a BMI range
of 16 to 39, total body water was determined using
BIA and previously published predictive equations
(20) and was compared with total body water measured using deuterium dilution. With the deuterium
dilution method, an oral dose of deuterium oxide (0.08
g per kg of body weight, 99.9 percent deuterium;
Cambridge Isotopes, Worcester, Massachusetts) was
administered and was allowed to equilibrate with body
water. Urine was sampled before deuterium oxide
administration and following equilibration, 4 - 6 hours
after dosing, and was analyzed for deuterium content.
Using simple dilution principles, total body water was
calculated from the increase in the deuterium content
of the urine from baseline to equilibration. Stable
isotopic analyses were performed by the Clinical Nutrition Research Unit of the University of Chicago. For
the BIA measurements, total body water was calculated from resistance, height, weight, and sex using
published equations. Although several of the most
commonly cited equations provided good estimates of
total body water, the best fit was obtained using the
equations of Kushner and Schoeller (20). The mean
deviation of the BIA-derived total body water from the
isotopically determined body water was —0.2 kg (95
percent confidence interval (CI) —0.7 to 0.3), corresponding to +0.2 percent (95 percent CI - 1 . 2 to 0.8)
body fat, for all sites combined. (Data for the individual sites were: +0.4 percent (95 percent CI —1.2 to
2.0) fat in Nigeria, 0 percent (95 percent CI -1.5 to
1.5) in Jamaica, and +0.3 percent (95 percent CI —1.3
to 1.9) in the United States.) No BIA-derived total
body water X site interaction was observed (p >
0.50). Thus, the same equations were used for all data
collected at the three field sites, with subsequent extrapolation to fat-free mass, fat mass, and percentage
of fat mass.
Downloaded from http://aje.oxfordjournals.org/ at Pennsylvania State University on February 26, 2014
MATERIALS AND METHODS
621
622
Luke et al.
Standing height was measured without shoes to the
nearest 0.1 cm at all sites. In the United States, weight
was measured to the nearest 0.25 pound (0.1 kg) using
a balance scale (Health-o-Meter, Inc., Bridgeview,
Illinois). In both Jamaica and Nigeria, weights were
measured to the nearest 0.1 kg using a calibrated
electronic scale with digital readout (Seca Corporation, Columbia, Maryland). BMI was calculated as
weight (kg)/height (m)2. Waist circumference was
measured to the nearest 0.1 cm at the narrowest part of
the torso as seen from the anterior aspect. In obese
subjects for whom it was difficult to identify a narrowing of the torso, the smallest horizontal circumference in the area between the lowest rib and the iliac
crest was utilized. Hip circumference was measured to
the nearest 0.1 cm at the point of maximum extension
of the buttocks. Repeat measurements were obtained
for waist and hip circumferences, and a third measurement was made if the difference between the first two
readings was more than 0.5 cm.
Blood pressure was measured three times using a
standard mercury manometer with the participant
seated (17). A 30-second radial pulse was taken and,
with the antecubital fossa at heart level, an initial
systolic pulse disappearance level was measured.
Three subsequent auscultatory readings were taken,
with additional pulse measurements made between
readings.
Statistical methods
Basic descriptive statistics, including mean values,
proportions, and Pearson's correlation coefficients,
were calculated. Multivariate linear regression was
performed taking percentage of body fat as the response variable and BMI as the explanatory variable.
The model was also fitted with additional covariates to
test for interactions. First, a dummy variable for sex
was considered for each site to study the interaction
between sex and BMI. Second, two dummy variables
for site were defined, which in turn were used to test
the interaction between site and BMI (23). All statistical analyses were carried out using Stata for Windows (24).
The study protocol was approved by the Institutional Review Board of Loyola University's Stritch
School of Medicine; by the University of the West
Indies, Kingston, Jamaica; and by University College
Hospital, Ibadan, Nigeria.
RESULTS
An obvious trend was observed for all anthropometric variables across sites (table 1). For both sexes,
height, weight, waist and hip circumference, BMI, and
percent body fat were lowest in Nigeria and greatest in
the United States, with intermediate values in Jamaica.
The relation between BMI and percent body fat is
presented for men and women in figure 1. Among
women, BMI and BMI2 were both determined to be
significant, irrespective of site (table 2). Among the
men, however, the relation varied by site. In the
United States and Jamaica, the relation between BMI
and percent body fat was also quadratic. In Nigeria,
where there was only one male with a BMI greater
than 35, the best-fitting mathematical relation was
linear. In addition, because of questions concerning
the validity of BIA at the extremes of BMI (25), we
also truncated the data from all sites at BMI = 35 and
reanalyzed the regression models. The quadratic relation was maintained among women in all populations
and among the Jamaican and US men; the levels of
correlation were maintained to within 5 percent for all
site- and sex-specific groups. Utilizing the quadratic
model for the full range of data explains 75 percent, 82
Am J Epidemiol
Vol. 145, No. 7, 1997
Downloaded from http://aje.oxfordjournals.org/ at Pennsylvania State University on February 26, 2014
Of the original 314 participants in Nigeria, 16 (one
woman and 15 men) were omitted from data analysis
because the BIA resistance values obtained resulted in
negative values for percentage of body fat. The mean
BMI of the 16 omitted participants was 3.9 units lower
than the mean for that population, and several had total
cholesterol levels less than 2.8 mmol/liter; thus, these
individuals were extremely lean and were probably
chronically undernourished (21). Data on two male
participants from Jamaica were omitted because of
negative BIA values, and no data were omitted from
the United States (n — 516). The equations that best
fitted our validation population and were subsequently
used for all three sites overestimated total body water
somewhat in extremely lean persons, resulting in implausibly low levels of body fat. In our validation
group, six men had extremely low levels of body fat as
measured using deuterium dilution (less than 10 percent, with a minimum of 4 percent). In these men, the
applied BIA equations estimated total body water with
the same accuracy as in the validation group as a
whole, and mean deviation from the isotopically determined total body water was only —0.3 kg. Thus, it
was not possible to determine the exact level of body
fat or the corresponding BMI at which overestimation
of total body water occurred among the Nigerian and
Jamaican participants. Because of the lack of individuals with extremely low levels of body fat at the US
site, further testing of the prediction equation could
not be carried out. Equations developed for a clinically
malnourished population, i.e., patients with Crohn's
disease (22), could not be validated in our population,
and thus were not used for the extremely lean individuals.
Body Mass Index and Body Fat in Black Populations
623
NIGERIA
i n <q oq
i--: r--: r»:
Tf V CO
o oo t^
T^ a t^
T~ T - CM
oq •* q
a> r-' T^
•tf -3- oq
d ui
T-"
£33
10
20
30
40
50
60
50
60
50
60
Body Mass Indsx
T-_ l>» 0 5
o i CM »SOU)
O ^) Is*;
co Tf in
O p O)
in cb r~.'
(s. CO O
co p a>
JAMAICA
u.
T-'
n
C\J
CM
s
CM
co in r-pi r^ co
o
2
m
c
s
CM CM CO
I
co in in
co in W
i-
i - CM
S
o co
r-» oo
Q.
0
(OSr
y) S
"5
S
rf
S
20
30
40
Body Mass Index
i - CM CD
(-. 00 10
il
10
CO 0 0
UNITED STATES
tn co co
GO 0 0 CO
1 ^ CO CO
I
I
353
e
CD
a
if
CO
0.
r- 2 a>
If1
CO Cfe CD
•o
a
Ill
Ul
CD
CD
3
$
C/3
SCO
I sill
o
Am J Epidemiol'
Vol. 145, No. 7, 1997
CD 6
-5T3
III
20
30
40
Body (Mass Index
FIGURE 1. Relations between body mass index and percentage
of body fat among men (+) and women (O) in Nigena, Jamaica, and
the United States, 1994-1995. TTie relations differ significantly between men and women at all sites (p < 0.001).
Downloaded from http://aje.oxfordjournals.org/ at Pennsylvania State University on February 26, 2014
0
O CO
624
Luke et al.
TABLE 2. Sex- and site-specific regression coefficients for the relation between body mass index*
(BMI) and percentage of body fat, 1994-1995t
Sex and
sits
Men
Nigeria
Jamaica
United States
Women
Nigeria
United States
P.
(BMP)
SEE*
(%tat)
-21.39
-31.96
1.51
2.39
-0.018§
4.7
4.7
0.61
0.62
-19.74
-36.61
1.66
3.05
-0.028
3.7
3.6
0.73
0.78
-2.84
-25.81
1.11
2.69
-0.025
4.6
4.3
0.66
0.70
-17.00
-44.24
1.78
4.01
-0.043
5.4
5.2
0.74
0.75
2.70
-30.81
1.21
3.65
-0.042
4.2
3.6
0.75
0.82
14.83
-11.18
0.88
2.53
- 0 025
3.7
3.0
0.78
0.85
* Weight (kgyhelght (m)*
t All p values were less than 0.001.
$ SEE, standard error of the estimate
§ BMI 2 was not a significant covariata in the model.
percent, and 85 percent of the variance in percent body
fat in Nigerian, Jamaican, and US women, respectively. Among Jamaican and US men, 78 percent and
70 percent of the variance was explained by BMI and
BMI2, respectively. Sixty-one percent of the variance
in percent body fat among Nigerian men was described by BMI. Standard errors for the prediction of
percent body fat ranged from 3.6 percent to 4.7 percent
among the men and from 3.0 percent and 5.4 percent
among the women (table 2).
Age was not consistently found to be a significant
variable in the regression of percent body fat on BMI.
This observation did not change with the truncated
data (maximum BMI = 35). Age was found to be a
significant predictor among Nigerian men {p < 0.01),
where the relation was positive, and among Jamaican
women {p < 0.05), where the relation was negative.
In neither case, however, was r 2 changed by the addition of age. Therefore, age was not included in any
of the results presented here.
Sex was a significant predictor variable at each site
(p < 0.001). Regression coefficients for the combined
data from men and women are presented in table 3.
The mean levels of percent body fat varied across BMI
values uniformly for each sex at a given site (figure 1);
i.e., within each site, the difference in percent body fat
for men and women was approximately constant
across the range of BMI values. The shape of the
relation between percent body fat and BMI was not
modified by sex, which implies that there was no
sex X BMI interaction within each site. Women, on
average, had 11 percent higher percent body fat values
than men at any given level of BMI (p < 0.001).
There was a clear trend across levels of BMI. US
men and women had the greatest mean percentage of
body fat, followed by Jamaicans; Nigerians had the
lowest mean (p < 0.001) (figure 2). It is apparent
from figure 2 that the sex-specific functional relation
between percent body fat and BMI was modified by
site. That is, there was a BMI X site interaction across
the sexes. However, this finding was shown to be
statistically significant for women but not for men—
i.e., we had enough power to detect an interaction
between site and BMI for women only. The model
containing BMI, BMI2, sex, and site explained over 90
percent of the variance at all sites combined (table 3).
Percent body fat did not account for more of the
variance in waist circumference, hip circumference,
waist:hip ratio, or blood pressure than did BMI
(table 4).
DISCUSSION
Some investigators have recommended the abandonment of BMI as a measure of obesity in favor of
more direct measures of body fat (6, 8), and several
published reports have provided evidence to support
this contention (5-9, 26-31). We found, however, that
within populations of American, Jamaican, and Nigerian blacks, BMI was a relatively good predictor of
level of body fat. At the same time, mean levels of
body fat varied substantially at similar levels of BMI
Am J Epidemiol
Vol. 145, No. 7, 1997
Downloaded from http://aje.oxfordjournals.org/ at Pennsylvania State University on February 26, 2014
Jamaica
P,
(BMI)
Intercept
Body Mass Index and Body Fat in Black Populations
625
TABLE 3. Regression coefficients for the relation between body mass index* (BU) and percentage of body fat, adjusted for sex
and site, 1994-19951
No. of
part&pants
Site
Nigeria
298
Jamaica
240
United States
516
All sites combined
1,054
Intercept
SEE*
(%fat)
(BUI)
(BMP)
(S«t)
-45.17
^t3.86
3.59
3.23
-0.033
-0.030
10.34
7.1
50
0.62
0.81
^45.70
^0.18
4.11
3.44
-0.044
-0.038
11.48
6.5
3.6
0.67
0.90
-22.58
-24 04
£81
2.61
-0.026
-0.025
11.47
6.4
3.5
0.64
0.89
-47.98
^7.38
-40.75
4.11
3.78
3.15
-0.042
-0.040
-0.033
10.96
11.20
7.3
5.1
4.1
0.70
0.85
0.91
(Site§)
4.02
between populations. The differences observed between populations were directly analogous to the difference observed between men and women at each
site. At any level of BMI, women had greater mean
levels of body fat than did men (p < 0.001). At any
BMI, the mean level of body fat increased with increasing industrialization (p < 0.001). Although age
was not a significant predictor of percent body fat after
data were adjusted for BMI, this finding has been
observed previously (9, 28, 31-33).
The cross-group differences observed in the present
study are clearly the result of population- and sexspecific differences in the ratio of fat-free mass to fat
mass at any given weight. Men have a greater proportion of fat-free mass than women, Nigerians have a
greater proportion than Jamaicans, and Jamaicans
have a greater proportion than African Americans.
Body composition is strongly influenced by both genetic and environmental factors (11). While the difference in body composition between men and women is
due in large part to genetic and hormonal differences
between the sexes, the differences observed between
sites among these genetically similar populations must
be due primarily to environment. The environmental
factors most responsible for the determination of body
composition are energy intake and energy expenditure
in the form of physical activity level (34).
Between 61 percent and 85 percent of the variance
in percent body fat could be explained by BMI for
each of the six site- and sex-specific groups in the
present study. The finding of a quadratic relation between BMI and percent body fat has been described
WOMEN
MEN
60
50
>.
•o
o
m
cs
LL.
>>
•o
40
m
30
//
o
cen1
c
/ '
V
o
u
a.
o
//
/
20
a.
10
0
()
Body Mass Index
10
20
30
40
50
6
Body Mass Index
FIGURE 2. Regression lines for the relations between body mass Index and percentage of body fat in Nigeria (. .), Jamaica ( — ) , and the
United States (-4, by sex, 1994-1995. The relations drffer significantly within each sex across the sites (p < 0.001).
Am J Epidemiol
Vol. 145, No. 7, 1997
Downloaded from http://aje.oxfordjournals.org/ at Pennsylvania State University on February 26, 2014
• Weight (kgVhelght (m)s.
t All p values were less than 0.001.
i Men = 0, women = 1.
§ Nigeria = 0, Jamaica = 1, United States = 2.
11 SEE, standard error of the estimate
626
Luke et a).
TABLE 4. Estimated correlations of body mass index* (BMI) and percentage of body fat with blood pressure and waist and hip
circumference, by sex and she, 1994-1995$,§
Women
Variable
Nigeria
BMI
Waist circumference
Hip circumference
Systolic blood pressure
Dtastofic Wood pressure
0 90"
0 93"
0.24
016
Men
Jamaica
*
0.77"
0.81 • •
0 24
0.14
BMI
United States
*
0 87"
0.77"
0.91" 0.82"
0.16
0 15
0.20
0.16
Nigeria
BMI
%
fat
BMI
0 91"
0.93"
0.21
0.07
0.85"
0 87"
0.21
0.10
0.89"
0.89"
0 09
0.31
Jamaica
United States
*
BMI
%
fat
BMI
%
fat
0.79"
0.76"
0.09
0 24
0.69"
0 64"
0 24
016
0.76"
0 72"
0 24
0.11
0 93"
0 93"
0 24*
022
0 83"
0.82"
0.23*
0.20
• p < 0.05; *• p< 0 01.
t Weight (kgVheight (m)«.
t Data were adjusted tor age
§ No significant difference was found between correlation coefficients for body mass Index and percentage of body fat
the US and Jamaican sites, a decrease in food supply
had occurred over the previous 3 years in Nigeria (3,
37). Not only did food resources differ between the
US, Jamaican, and Nigerian populations but physical
activity patterns varied enormously. Energy intake and
physical activity are both crucial determinants of body
composition (34) and most likely explain the shift in
the relation between BMI and body fat.
The question of the value of BMI in predicting
adiposity is further confused by the fact that percent
body fat is not necessarily a better predictor of blood
pressure or anthropometric variables associated with
chronic illness. Our results support previous findings
that BMI is similarly or better correlated with blood
pressure and plasma glucose levels than is body fat
level (38).
One of the advantages of using BMI to define obesity was the availability of national data sets which
could be used to establish standards, such as the 85th
percentile of BMI in NHANES I designating overweight and the 95th percentile designating obesity (4).
Currently, there are no large national data sets with
data on body composition and no consensus on acceptable levels of body fat; suggested upper limits
range from 25 percent to 33 percent for women and
from 20 percent to 25 percent for men (39-42). BIA
was used to estimate body composition in NHANES
in (1988-1991); however, as of this writing those data
have not yet been published.
In addition to the question of whether or not to
directly measure adiposity in epidemiologic research,
there is the issue of which method to use for its
estimation. BIA appears to be a good method for
estimating body fat at the population level. While it is
a relatively accurate and inexpensive method to employ, there are questions about the reliability of BIAderived estimates of body composition at the extremes
of body fat distribution. Clearly, in the present study
we were confronted with erroneous estimates of body
fat among extremely lean individuals in the Nigerian
and Jamaican samples. While we cannot be totally
Am J Epidemiol
Vol. 145, No. 7, 1997
Downloaded from http://aje.oxfordjournals.org/ at Pennsylvania State University on February 26, 2014
before (7, 8); however, the strength of the relations in
the present study for the most part exceeds that of
those reported from these studies. We recognize, however, that the correlations were somewhat inflated by
the inclusion of height and weight in the calculation of
percent body fat. In studies of primarily nonblack
populations, r 2 has ranged from 0.35 to 0.71 (5-8, 27,
31). Gallagher et al. (9) recently reported coefficients
of determination of 0.44 and 0.58 for black men and
women, respectively. In their linear model, both BMI
and age were significant coefficients. Most studies to
date have utilized methods of determining body composition that are considerably more accurate than BIA
but are unavailable to most institutions, are quite
costly, and are virtually impossible to use at remote
field sites—i.e., hydrodensitometry, isotope dilution,
or dual photon absorptiometry. While the measurement of skinfold thickness for the estimation of body
fat is a viable alternative to BIA, it is subject to
significant inter- and intraobserver variability (32, 35).
Using isotope dilution, we were able to validate the
less well established BIA method in the field settings
of the three international sites. It seems reasonable to
conclude, therefore, that BMI is a valid proxy for
measures of adiposity within these black populations.
While the relation between BMI and percent body
fat is strong within populations, average body composition differed substantially across sites, making intersite comparisons of BMI difficult to interpret. For
example, a BMI of 25 represented mean body fat
levels of 25.8 percent, 22.2 percent, and 16.4 percent
among US, Jamaican, and Nigerian men, respectively
(p < 0.001). Although comprehensive studies using
informative genetic markers which are currently available have yet to be carried out, the historical record
indicates that the populations we studied share a common West African ancestry (36). In contrast, the current environments of the three populations are very
different in terms of the level of industrialization and
its consequences for individual lifestyles. While food
availability was not an issue for most participants from
Body Mass Index and Body Fat in Black Populations
ACKNOWLEDGMENTS
This research was partially supported by grants
HL45508, HL51508, and DK 26678 from the National
Institutes of Health.
REFERENCES
1. Popkin BM. The nutrition transition in low-income countries'
an emerging crisis. Nutr Rev 1994,52:285-98.
2 Poulter NR, Khaw K, Hopwood BE, et al. Determinants of
blood pressure changes due to urbanization: a longitudinal
study. J Hypertens Suppl 1985;3(suppl 3)5375-7.
3 Cooper R, Rotimi C, Kaufman J, et al. Nutritional status in
rural Nigerians. (Letter). Lancet 1996;347331-2.
4. National Institutes of Health Consensus Development Panel
on the Health Implications of Obesity. Health implications of
obesity. Ann Intern Med 1985,103:1073-7.
5. Gam SM, Leonard WR, Hawthorne VM. Three limitations of
the body mass index. (Editorial) Am J Clin Nutr 1986;44:
996-7.
6 Bouchard C. Genetics of human obesities: introductory notes.
In. Bouchard C, ed. The genetics of obesity. Boca Raton, IT-:
CRC Press, 1994.1-15.
7 Smalley KJ, Knerr AN, Kendrick ZV, et al. Reassessment of
body mass indices. Am J Clin Nutr 1990;52:4O5-8.
8. RoubenoffR, DallalGE, Wilson PW Predicting body fatness.
the body mass index vs estimation by bioelectncal impedance.
Am J Public Health 1995;85:726-8.
9 Gallagher D, Visser M, Sepiilveda D, et al. How useful is body
mass index for comparison of body fat across age, sex, and
ethnic groups? Am J Epidemiol 1996;143.228-39.
10. Garrow JS, Webster J. Quetelet's index (W/H2) as a measure
Am J Epidemiol
Vol. 145, No. 7, 1997
of fatness. In J Obes 1985;9.147-53.
11. Bouchard C, ed The genetics of obesity Boca Raton, FLCRC Press, 1994
12. Foster KR, Lukaski HC. Whole-body impedance—What does
it measure? Am J Clin Nutr 1996;64{suppl):388S-96S.
13. Trye P, Jackson R, Stewart A, et al. Trends and determinants
of blood pressure in Auckland, New Zealand 1982-94. N Z
Med J 1996;109:179-81.
14. Kaufman JS, Durazo-Arvizu R, Rotimi CN, et al. Obesity and
hypertension prevalence in populations of African origin. Epidemiology 1996;7:398-405
15. Hodge AM, Dowse GK, Gareeboo H, et al. Incidence, increasing prevalence, and predictors of change in obesity and fat
distribution over 5 years in the rapidly developing population
of Mauritius. Int J Obes Relat Metab Disord 1996;20-137-46.
16. Rotimi CN, Cooper RS, Ataman SL, et al Distribution of
anthropometnc variables and the prevalence of obesity in
populations of West African origin: the International Collaborative Study on Hypertension m Blacks (ICSHIB). Obes Res
1995,3(suppl 2).95S-105S.
17. Ataman SL, Cooper R, Rotimi C, et al. Standardization of
blood pressure measurement in an international comparative
study. J Chn Epidemiol 1996,49:869-77.
18. Lukaski HC, Bolonchuk WW, Hall CB, et al Validation of
tetrapolar bioelectrical impedance method to assess human
body composition. J Appl Physiol 1986;60:1327-32.
19. Schoeller DA, van Santen E, Peterson DW, et al. Total body
water measurement in humans with 18O and 2H labeled water
Am J Clin Nutr 1980;33:2686-93.
20 Kushner RF, Schoeller DA. Estimation of total body water by
bioelectncal impedance analysis. Am J Chn Nutr 1986;44.
417-24.
21. Gonzilez J, Periago JL, Gil A, et al. Malnutrition-related
polyunsaturated fatty acid changes in plasma lipid fractions of
cirrhotic patients. Metabolism 1992,41:954-60.
22. Royall D, Greenberg GR, Allard JP, et al. Critical assessment
of body-composition measurements in malnourished subjects
with Crohn's disease- the role of bioelectric impedance analysis. Am J Clin Nutr 1994;59.325-30
23. Draper NR, Smith H. Applied regression analysis. 2nd ed.
New York, NY- John Wiley and Sons, Inc, 1981.
24. Stata Corporation. Stata reference manual, release 4: statistics,
graphics, data management. College Station, TX: Stata Corporation, 1995.
25 Kaminsky LA, Whaley MH. Differences in estimates of percent body fat using bioelectncal impedance. J Sports Med
Phys Fitness 1993;33.172-7.
26. Wang J, Thornton JC, Russell M, et al. Asians have lower
body mass index (BMI) but higher percent body fat than do
whites: comparisons of anthropometnc measurements. Am J
Chn Nutr 1994;60.23-8
27. Revicki DA, Israel RG. Relationship between body mass
indices and measures of body adiposity. Am J Public Health
1986;76:992-4.
28. Deurenberg P, Weststrate JA, Seidell JC. Body mass index as
a measure of body fatness- age- and sex-specific prediction
formulas. Br J Nutr 1991 ;65-105-14.
29. Wellens Rl, Roche AF, Khamis HJ, et al Relationships between the body mass index and body composition. Obes Res
1996;4-35-44.
30. Heitmann BL. Evaluation of body fat estimated from body
mass index, skinfolds and impedance- a comparative study.
Eur J Clin Nutr 1990;44:831-7.
31. Baumgartner RN, Heymsfield SB, Roche AF. Human body
composition and the epidemiology of chronic disease. Obes
Res 1995;3:73-95.
32. Lohman TG. Advances in body composition assessment.
Champaign, IL. Human Kinetics, 1992.
33. Mazanegos M, Wang ZM, Gallagher D, et al. Differences
between young and old females in the five levels of body
composition and their relevance to the two-compartment
chemical model. J Gerontol 1994;49:M201-8.
Downloaded from http://aje.oxfordjournals.org/ at Pennsylvania State University on February 26, 2014
confident of the validity of the technique and the
predictive model used in the very lean or the very
obese, we believe the method to be reliable across the
normal range of BMI values found in most healthy
populations. Thus, while some investigators have
raised concerns about the use of BIA in epidemiologic
research (43-45)—i.e., requirements for the standardization of protocols (46), the profusion of predictive
equations in the literature (43), and inconsistencies
between analyzers from different manufacturers (47,
48)—we believe it is currently the best option for
measuring body composition in the field.
In summary, BMI is the most commonly used measure of obesity in epidemiologic research. In three
different populations of West African heritage from
the United States, Jamaica, and Nigeria, BMI was
shown to provide good within-population estimates of
body fat levels. However, the relations between BMI
and body fat levels differed significantly across the
three populations, making it impossible to compare the
prevalence of obesity based on BMI between populations. In addition to sex, current environment, probably including dietary patterns and level of physical
activity, can strongly influence the relation between
BMI and body fat at the population level.
627
628
Luke et al.
caloric requirements of men. Am J Clin Nutr 1987,46:875-85.
43. Hammond J, Rona RJ, Chinn S. Estimation in community
surveys of total body fat of children using bioelectrical impedance or skinfold thickness measurements. Eur J Clin Nutr
1994;48:164-71.
44. Young RE, Sinha DP. Bioelectrical-impedance analysis as a
measure of body composition in a West Indian population
Am J Clin Nutr 1992,55:1045-50
45. Deurenberg P, Snut HE, Kusters CS. Is the bioelectncal
impedance method suitable for epidemiological field studies?
Eur J Chn Nutr 1989,43:647-54.
46. Gudivaka R, Schoeller D, Ho T, et al. Effect of body position,
electrode placement and time on prediction of total body water
by multifrequency bioelectrical impedance analysis Age Nutr
1994,5:111-17.
47. Smye SW, Sutchffe J, Pitt E. A comparison of four commercial systems used to measure whole-body electncal impedance. Physiol Meas 1993;14473-8.
48. Graves JE, Pollock ML, Colvin AB, et al. Comparison of
different bioelectncal impedance analyzers in the prediction of
body composition. Am J Hum Biol 1989;l:603-ll.
Am J Epidemiol
Vol. 145, No. 7, 1997
Downloaded from http://aje.oxfordjournals.org/ at Pennsylvania State University on February 26, 2014
34. Ravussin E, Fontveille AM, Swinburn BA, et al. Risk factors
for the development of obesity. Ann N Y Acad Sci 1993;683:
141-50.
35. Lukaski HC. Methods for the assessment of human body
composition: traditional and new. Am J Clin Nutr 1987,46:
537-56.
36. Harris JE Global dimensions of the African diaspora 2nd ed.
Washington, DC: Howard University Press, 1993.
37. Obasohan AO Body weight and heart failure in hypertensives: a Nigerian experience. Trop Cardiol 1995;21:47-50.
38. Spiegelman D, Israel RG, Bouchard C, et al. Absolute fat
mass, percent body fat, and body-fat distribution: Which is the
real detenninant of blood pressure and serum glucose9 Am J
Clin Nutr 1992,55:1033-44.
39. Owen OE, Kavle E, Owen RS, et al. A reappraisal of caloric
requirements in healthy women. Am J Clin Nutr 1986;44:
1-19.
40. Whitney EN, Hamilton EM, Rolfes SR. Understanding nutrition. New York, NY: West Publishing, 1990:363.
41. Bray GA. Fat distribution and body weight. Obes Res 1993;
1:203-6.
42. Owen OE, Holup JL, D'Alessio DA, et al A reappraisal of the