| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
Special Features |
National Institute of Diabetes and Digestive and Kidney Diseases, National Institutes of Health, Phoenix, Arizona 85014
Address all correspondence and requests for reprints to: Dr. Robert Lindsay, 1550 East Indian School Road, Phoenix, Arizona 85014. E-mail: rlindsay{at}mail.nih.gov or rlindsay@mail.
Abstract
Body mass index is widely used as a measure of adiposity in adults, but its use in children and adolescents is controversial. We assessed body mass index as a measure of adiposity in children and adolescents between the ages of 5 and 20 yr examined as part of the NIH survey of health in the Pima Indian population. Body mass index (measured in 985 subjects and analyzed in 3 age groups: 5–9, 10–14, and 15–19 yr, in both sexes) was compared cross-sectionally to percent fat and fat mass derived from dual energy x-ray absorptiometry and to fasting and 2-h plasma glucose, systolic and diastolic blood pressures, cholesterol, high density lipoprotein cholesterol, fasting insulin, and triglycerides.
Body mass index was strongly correlated in all age groups to both percent fat (r = 0.83–0.94; for each group, P < 0.0001) and fat mass (r = 0.96–0.98; P < 0.0001). The relationship of body mass index to percent fat was different in males and females; differences were more marked in older age groups. Body mass index, percent fat, and fat mass showed similar degrees of correlation to metabolic measures in childhood.
Body mass index is strongly associated with measures of adiposity derived from dual energy x-ray absorptiometry. Both measures show similar associations with cardiovascular risk factors in Pima Indian children.
THERE IS INCREASING clinical concern about excess adiposity in childhood. The prevalence of obesity is increasing in children, with potentially important long-term effects on metabolic and vascular diseases (1). A variety of measures are available to clinicians and researchers for assessment of adiposity in children. Body mass index (BMI; kilograms per m2) is commonly used in clinical practice because it is straightforward and relatively cheap to obtain, but its clinical interpretation remains controversial (2). Other techniques, generally confined in use to research studies [such as bioimpedance, underwater weighing, and dual energy x-ray absorptiometry (DEXA)] offer accurate measurement of adiposity, but are deemed unsuitable for routine clinical use because they are relatively expensive, rely on more complex technologies, and are technically more demanding (3). A number of recent studies have explored the relationship of BMI to adiposity by DEXA and found a high degree of correlation between these measures in Caucasian children (3, 4, 5). Others have detailed a positive relationship between BMI and individual cardiovascular risk factors in children (6, 7, 8).
The Pima Indians of Arizona suffer from unprecedentedly high levels of obesity in childhood and later life (9). The aims of this study were to examine whether BMI is correlated with more accurate measures of adiposity (derived from DEXA) in a non-Caucasian population in childhood and, further, to compare BMI and DEXA measures in their association with a variety of metabolic variables, assessing whether DEXA measures were superior in predicting potential adverse consequences of adiposity.
Subjects and Methods
The subjects were participants in the NIH population study of diabetes in the Gila River Indian Community in Arizona. All members of the community 5 yr or older were invited to a biennial examination, including a 75-g oral glucose tolerance test. Diabetes was diagnosed by WHO 1985 criteria (10), either using the results of the oral glucose tolerance test or by independent diagnosis in a clinical setting. Height and weight were measured, with the subject wearing light clothing and no shoes, for calculation of BMI (kilograms per m2). Blood pressure was measured supine with a standard sphygmomanometer using appropriately sized pediatric or adult cuffs.
Glucose was measured by the glucose oxidase method with a glucose analyzer (Beckman Coulter, Inc., Fullerton, CA) Plasma insulin concentrations were determined by a single, commercially available, RIA analyzer (Concept4, ICN Biomedicals, Inc., Horsham, PA), and fasting triglyceride and cholesterol in plasma were quantified on an autoanalyzer II (Technicon Instruments, Tarrytown, NY) from fasting samples as previously described (11).
From 1993–1999, a total of 985 subjects had undergone DEXA examination between the ages of 5 and 20. DEXA scans were performed using a whole body scanner (DPX-L, Lunar Corp., Madison, WI), calibrated daily against phantoms according to the standard Lunar Corp. procedure and analyzed using Lunar Corp. software (version 1.3z). Where body size exceeded the scanning area of the machine, body composition was estimated on one side of the body as described previously (12). DEXA has previously been found to be highly accurate in predicting body composition compared with underwater weighing in adults in this population (12).
Subjects were included if they were over 5 yr and less than 20 yr of age at the time of examination. To allow assessment at different stages of childhood, data were analyzed in three 5-yr age groups (5–9, 10–14, and 15–19 yr) in both sexes. Where subjects had been examined more than once in an age group, only the first examination was included. Differences between males and females were modeled in a general linear model with adjustment for age at examination. The relationship of percent fat and fat mass to BMI in different sex and age groups was modeled by regression against either linear or linear and quadratic, terms for percent fat and fat mass, depending on significance of these terms. Correlations between different measures of adiposity (BMI, percent fat, and fat mass) and between these adiposity measures and metabolic variables were assessed by Pearson partial correlation adjusted for age as a covariate using standard software (SAS Institute, Inc., Cary, NC), and values of Pearson partial correlation coefficients (r) were reported. Metabolic variables were measured in the majority of those with measurement of adiposity: systolic and diastolic blood pressures in 979 individuals (99%), fasting triglycerides and high density lipoprotein (HDL) cholesterol in 883 individuals (90%), and total cholesterol in 881 individuals (89%). A total of 18 subjects (1.8%) were diagnosed with diabetes either before or at the time of examination and were excluded from correlation analysis of fasting glucose, 2-h glucose, and fasting insulin. Two-hour glucose is not included where children refused the glucose load. Fasting glucose measurements were thus available in 887 individuals (90%), 2-h glucose in 749 (76%), and fasting insulin in 605 (61%). To approximate normal distributions for correlation analysis, log-transformed values of triglycerides and fasting insulin were used.
Results
Relationship of BMI to body composition by DEXA
Among female subjects, values of BMI, percent body fat, fat mass,
and fat-free mass increased in successively older age groups
(P < 0.05; Table 1
). By
contrast, although BMI, fat mass, and fat-free mass increased in
successively older age groups in males (P < 0.05),
percent fat was slightly lower in those aged 15–19 yr compared with
10- to 14-yr-old males. This suggested that the relationship of BMI to
percent fat was likely to vary with age in males.
|
). These differences became greater with
age.
Percent fat was highly correlated to BMI in both sexes and in each age
group (Fig. 1), with linear correlation
coefficients above 0.8 (males: age group 5–9, r = 0.94; 10–14,
r = 0.84; 15–19, r = 0.89; females: age group 5–9, r =
0.92; 10–14, r = 0.86; 15–19, r = 0.83: all
P < 0.0001). Fat mass was even more closely correlated
to BMI (Fig. 2), with linear correlation
coefficients all above 0.95 (males: age group 5–9, r = 0.96;
10–14, r = 0.97; 15–19, r = 0.96; females: age group 5–9,
r = 0.96; 10–14, r = 0.98; 15 to 19, r = 0.96: all
P < 0.0001).
|
|
), and percent fat increased
asymptotically to maximum levels of around 50%, achieved with BMIs of
approximately 35- 40 kg/m2. In keeping with this,
a quadratic term for BMI was highly significant (P <
0.0001) in each age and sex group. By contrast, the relationship of fat
mass and BMI appeared linear in all age and sex groups (Fig. 2),
addition of quadratic terms for fat mass were not statistically
significant in any group (P > 0.05). As well as the absolute differences between the sexes described above, the relationship of BMI to both percent fat and fat mass were different in males and females. Thus, when the relationship of fat mass to BMI was modeled within age groups, terms not only for gender but also the interaction of gender and BMI in the two older age groups were statistically significant (interaction term gender x BMI, age 5–9, P = 0.1; age 10–14, P < 0.02; age 15–19, P < 0.001). In a similar fashion, when the relationship of percent fat to BMI was modeled within age groups (including in this case terms for gender and both linear and quadratic for BMI), the interaction of gender and BMI was significant in all age groups (interaction term gender x BMI, P < 0.005).
To examine the extent to which the relationship of BMI with percent fat
and fat mass varies with age in each sex, regression lines for each age
group are presented for males and females in Fig. 3. In females the relationship of BMI to
both percent fat and fat mass did not change markedly in older age
groups, following similar curvilinear and linear relationships,
respectively (Fig. 3). Regression lines from data for females in age
groups 10–14 and 15–19 yr were almost superimposable for both percent
fat and fat mass. In the youngest age group the levels of BMI found in
the upper parts of the range of this group (25–30
kg/m2) were associated with lower fat mass, and
higher percent fat than might be predicted in the older age groups.
|
). Equivalent levels of BMI predicted a
10–15% lower percent fat in older (age 15–19 yr) vs.
younger (age 5–9 yr) age groups. For fat mass the general pattern was
very similar to that for females: the relationships of BMI to fat mass
in age groups 10–14 and 15–19 yr were very similar, whereas in the
youngest age group, higher levels of BMI found in the upper parts of
the range of this group (25–30 kg/m2) were
associated with lower fat mass than might be predicted in the older age
groups. In support of this general interpretation that the relation of
BMI to both percent fat and fat mass was different at different ages,
when the relationship of percent fat was modeled against BMI (linear
and quadratic) in all ages combined (age 5–19 yr), age was highly
significantly and negatively associated with percent fat in males (ß
coefficient, -1.2%/yr; P < 0.0001). In females, age
was also negatively associated but with a smaller ß coefficient and
of only marginal significance (ß coefficient, -0.1%/yr;
P = 0.02). In both cases, interaction terms of age and
BMI (age x BMI) were also significant (P < 0.01)
when added to the model. For fat mass, age was positively associated
with fat mass over and above effects of BMI; the relationship was
highly significant in females (P < 0.0001), but not in
males (P = 0.1). Again, interaction terms of age
x BMI proved significant (P < 0.01) when added to the
above models. Relationship of BMI and percent body fat to cardiovascular risk factors
Adiposity per se (as represented by all three measures:
BMI, percent fat, and fat mass) displayed varying strengths of
relationship with the metabolic risk factors. Correlation to the
various adiposity measures was strongest and most consistent across age
and sex groups with HDL cholesterol (range of r = -0.29 to
-0.53; all P < 0.0005) and fasting insulin (range of
age-adjusted Pearson partial correlation coefficient, r =
0.31–0.67; all P < 0.05) and was most weakly and
least consistently with total cholesterol (range of r =
-0.02–0.3; P = 0.8–0.004). Notwithstanding these
differences, the different measures of adiposity appeared to have very
similar relationships with the metabolic risk factors. In all but one
measure in one group (percent fat vs. fasting glucose in
females 10–14 yr: r = 0.16; P = 0.06) there was a
significant correlation to both fasting and 2-h glucose as well as
fasting insulin (Fig. 4). For
triglycerides and HDL cholesterol (Fig. 5) again the correlation of measures of
adiposity was significant apart from one measure in one group (percent
fat vs. triglycerides in females aged 15–19 yr,
P = 0.09). No single measure of adiposity (BMI, percent
fat, or fat mass) appeared to have consistently higher correlation
coefficients (Figs. 4 and 5). Other metabolic variables not presented
in Figs. 4 and 5 had lower (and not always significant) correlation
coefficients, but again these were similar between the various measures
of adiposity: diastolic blood pressure (BMI, r = 0.04–0.29;
percent fat, r = 0.03–0.35; fat mass, r = 0.04–0.35),
systolic blood pressure (BMI, r = 0.14–0.43; percent fat, r
= 0.03–0.44; fat mass, r = 0.13–0.45), cholesterol (BMI, r
= -0.01 to 0.31; percent fat, r = -0.03 to 0.30; fat mass,
r = -0.02 to 0.31). No single measure of adiposity (BMI, percent
fat, or fat mass) had consistently higher correlation coefficients
(data not shown).
|
|
BMI is closely associated with measures of adiposity derived from DEXA and bears a similar relationship to measures of cardiovascular and metabolic risk in Pima Indian children.
Our results are in agreement with previous studies in Caucasian children, suggesting that BMI is highly related to adiposity (3, 4, 5, 13), that BMI has a linear relationship with total fat mass (3, 4, 5), and that simple correlations of BMI are greater with fat mass than with percent body fat (3, 5). We have now confirmed this in a larger sample in a different ethnic group.
In addition, and importantly, we confirm that BMI and more direct measures of adiposity derived from DEXA are not always equivalent, with important effects of gender and age influencing this relationship. Body compositions of males and females of similar BMI are different even in prepubertal children (14). This gender difference is also apparent in our study; even at the youngest ages, males and females of equivalent BMI have significantly different body composition, with males having greater lean body mass and smaller fat mass, on the average. Similarly, the changing relationship of BMI and adiposity with age has been noted in a number of previous studies (3, 13). Pietrobelli et al. (3) found that older children tend to have greater fat mass and lower percent fat for equivalent BMI, a pattern that is also present in the Pima population in this study. The relationship of BMI and adiposity may also depend on two factors not addressed in our study. The relationship of BMI and direct measures of adiposity changes with sexual maturity in children (15) and is potentially influenced by ethnicity, as is the case in adults (16). As data on stage of puberty have not been collected as part of the longitudinal study, we were unable to examine influences of sexual maturity on the relationship of BMI and adiposity.
BMI does have certain advantages over assessment of adiposity using DEXA. Measurement of BMI is cheaper, technically far easier and, given that variability on repeated measurements of height and weight should be low, likely to be more precise than either percent fat or fat mass. Nevertheless, BMI remains an indirect measure of adiposity and for that reason more direct measures, such as those offered by DEXA, will be preferred in many studies examining adiposity in children. Where BMI is used it will be important to assess the accuracy of the estimate of adiposity supplied by BMI against such direct measures.
Our results differ in some respects from earlier studies, probably relating to the higher numbers of overweight children we include. Pietrobelli et al. (3) suspected, but failed to demonstrate significantly, the presence of a curvilinear relationship of BMI and percent body fat as detailed in our findings. In our report mean BMI is 5–7 kg/m2 higher than in their study (mean BMI of postpubertal boys and girls in their study, 23.9± 5.4 and 26.5± 7.7 kg/m2, respectively); the inclusion of children with very high BMIs makes clear the curvilinear relationship of BMI and percent fat. This raises a broader issue. The greater adiposity of Pima children may render our results less applicable to leaner populations; compared with Center for Disease Control standards for the United States population (17), on the average, BMI in the children in this report lies above the 80th percentile of for ages 5–9 yr and above the 90th percentile in older ages. However, as detailed above, most of our findings are in agreement with earlier studies.
We have also explored an alternative measure of how BMI performs as a correlate of metabolic disease, examining both BMI and DEXA measures of adiposity in relation to metabolic and cardiovascular risk factors in this population. Several studies have examined relationships of BMI to metabolic risk factors in children and detailed significant relationships with blood pressure (6, 7, 8), lipids (7, 8), and serum insulin (18), with such relationships being apparent in both black and white populations (7, 8). To our knowledge no study has examined relationships of BMI and DEXA measures of adiposity in such a large sample. Daniels et al. (15) detailed relationships of BMI and percent fat by DEXA to cardiovascular risk factors in a sample of 127 black and white children. Interestingly, BMI appears to have been more strongly associated with some risk factors, notably blood pressure, than percent fat, but the sample size was relatively small and ethnically heterogeneous. The particular question of the relative strength of relationships of BMI and percent fat to cardiovascular risk factors was not analyzed in detail. Our most important observation in this regard is that although the relationship of adiposity to the individual risk factors varies greatly, the relationships of BMI, percent fat, and fat mass to cardiovascular risk factors are broadly similar.
Our data suggest that BMI is a good measure of adiposity in Pima children, as previously reported in other populations (3, 4, 5). BMI is also closely associated with the more direct measures of adiposity provided by DEXA and has as close an association to cardiovascular risk factors. As such BMI would appear to be a reasonable surrogate for percent fat by DEXA in cross-sectional analysis, particularly in epidemiological studies. It should also be noted, however, that the measures are not always equivalent, and that the relationship of BMI to adiposity is dependent on age and gender. There is, therefore, an important role for direct measures of adiposity where there is the possibility that the relationship of BMI to adiposity may not be homogenous in the population studied, perhaps due to the effects of disease, ethnicity, or timing of puberty.
Acknowledgments
We thank the members of the Gila River Indian Community for their continued support and participation in this study, and the staff of the Diabetes and Arthritis Epidemiology Section for their help in conducting this study.
Footnotes
Abbreviations: BMI, Body mass index; DEXA, dual energy x-ray absorptiometry; HDL, high density lipoprotein.
Received November 16, 2000.
Accepted April 20, 2001.
References
This article has been cited by other articles:
 |
|
 |
|
  D. S Freedman, P. T Katzmarzyk, W. H Dietz, S. R Srinivasan, and G. S Berenson Relation of body mass index and skinfold thicknesses to cardiovascular disease risk factors in children: the Bogalusa Heart Study Am. J. Clinical Nutrition, July 1, 2009; 90(1): 210 - 216. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 M. S. Thearle, J. C. Bunt, W. C. Knowler, and J. Krakoff Childhood Predictors of Adult Acute Insulin Response and Insulin Action Diabetes Care, May 1, 2009; 32(5): 938 - 943. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 H Wang, J Necheles, M Carnethon, B Wang, Z Li, L Wang, X Liu, J Yang, G Tang, H Xing, et al. Adiposity measures and blood pressure in Chinese children and adolescents Arch. Dis. Child., September 1, 2008; 93(9): 738 - 744. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 L. E.S. Mayer, C. A. Roberto, D. R. Glasofer, S. F. Etu, D. Gallagher, J. Wang, S. B. Heymsfield, R. N. Pierson Jr., E. Attia, M. J. Devlin, et al. Does Percent Body Fat Predict Outcome in Anorexia Nervosa? Am J Psychiatry, June 1, 2007; 164(6): 970 - 972. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 R. A. Niinimaki, A. H. Harila-Saari, A. E. Jartti, R. M. Seuri, P. V. Riikonen, E. L. Paakko, M. I. Mottonen, and M. Lanning High Body Mass Index Increases the Risk for Osteonecrosis in Children With Acute Lymphoblastic Leukemia J. Clin. Oncol., April 20, 2007; 25(12): 1498 - 1504. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 J. Nething, K. Ringwald-Smith, R. Williams, M. L. Hancock, and G. A. Hale Establishing the Use of Body Mass Index as an Indicator of Nutrition Risk in Children With Cancer JPEN J Parenter Enteral Nutr, January 1, 2007; 31(1): 53 - 57. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 L. A. Moreno, M. G. Blay, G. Rodriguez, V. A. Blay, M. I. Mesana, J. L. Olivares, J. Fleta, A. Sarria, M. Bueno, and AVENA-Zaragoza Study Group Screening performances of the international obesity task force body mass index cut-off values in adolescents. J. Am. Coll. Nutr., October 1, 2006; 25(5): 403 - 408. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 E. W. Demerath, C. M. Schubert, L. M. Maynard, S. S. Sun, W. C. Chumlea, A. Pickoff, S. A. Czerwinski, B. Towne, and R. M. Siervogel Do Changes in Body Mass Index Percentile Reflect Changes in Body Composition in Children? Data From the Fels Longitudinal Study Pediatrics, March 1, 2006; 117(3): e487 - e495. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 B. Gance-Cleveland and M. Bushmiaer Arkansas School Nurses' Role in Statewide Assessment of Body Mass Index to Screen for Overweight Children and Adolescents The Journal of School Nursing, April 1, 2005; 21(2): 64 - 69. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 J. D. Veldhuis, J. N. Roemmich, E. J. Richmond, A. D. Rogol, J. C. Lovejoy, M. Sheffield-Moore, N. Mauras, and C. Y. Bowers Endocrine Control of Body Composition in Infancy, Childhood, and Puberty Endocr. Rev., February 1, 2005; 26(1): 114 - 146. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
D. S. Freedman, L. K. Khan, M. K. Serdula, W. H. Dietz, S. R. Srinivasan, and G. S. Berenson The Relation of Childhood BMI to Adult Adiposity: The Bogalusa Heart Study Pediatrics, January 1, 2005; 115(1): 22 - 27. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
M. G Neovius, Y. M Linne, B. S Barkeling, and S. O Rossner Sensitivity and specificity of classification systems for fatness in adolescents Am. J. Clinical Nutrition, September 1, 2004; 80(3): 597 - 603. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
R. W Taylor, I. E Jones, S. M Williams, and A. Goulding Body fat percentages measured by dual-energy X-ray absorptiometry corresponding to recently recommended body mass index cutoffs for overweight and obesity in children and adolescents aged 3-18 y Am. J. Clinical Nutrition, December 1, 2002; 76(6): 1416 - 1421. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
A. D. Salbe, C. Weyer, I. Harper, R. S. Lindsay, E. Ravussin, and P. A. Tataranni Assessing Risk Factors for Obesity Between Childhood and Adolescence: II. Energy Metabolism and Physical Activity Pediatrics, August 1, 2002; 110(2): 307 - 314. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 M. Horlick Body Mass Index in Childhood-- Measuring a Moving Target J. Clin. Endocrinol. Metab., September 1, 2001; 86(9): 4059 - 4060. [Full Text] [PDF]
|
|
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| Endocrinology | Endocrine Reviews | J. Clin. End. & Metab. |
| Molecular Endocrinology | Recent Prog. Horm. Res. | All Endocrine Journals |
