| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Original Research Communication |
1 From the Obesity Research Center, St Luke’s–Roosevelt Hospital and the Institute of Human Nutrition, Columbia University College of Physicians and Surgeons, New York (JRF, MH, SBH, RNP, FXP-S, ZMW, JW, MH, and DG), and the Departments of Nutrition Sciences (JRF) and of Biostatistics and Clinical Nutrition Research Center (JRF and DBA), University of Alabama at Birmingham.
2 Supported in part by NIH grants R01DK51716, P30DK26687, P30DK056336, R01ES09912, and P01DK42618.
3 Reprints not available. Address correspondence to JR Fernández, Division of Physiology and Metabolism, Department of Nutrition Sciences, University of Alabama at Birmingham, Webb 435, 1530 3rd Avenue S, Birmingham, AL 35294-3360. Email: fernandj{at}shrp.uab.edu.
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONSUBJECTS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Objective: We investigated whether the relation between PBF and BMI in adult HAs differed from that of African Americans (AAs) and European Americans (EAs).
Design: We used a multiple regression model in which PBF measured with dual energy X-ray absorptiometry was predicted by the reciprocal of BMI (1/BMI; in m2/kg) in a sample of 487 men (nEA = 192, nAA = 148, and nHA = 147) and 933 women (nEA = 448, nAA = 304, and nHA = 181).
Results: For men, our results showed no significant differences between HAs and EAs, AAs and EAs, or HAs and AAs in the slope of the line relating 1/BMI to PBF. In women, there were significant differences in PBF as predicted by BMI between HAs and EAs (P < 0.002) and AAs and HAs (P = 0.020), but not between AAs and EAs. When PBF was estimated on the basis of predicting equations, the trend of the predicted PBF value in women differed according to ethnic group and BMI category. At a BMI < 30, HAs tended to have more body fat than did EAs and AAs, and at a BMI > 35, EAs tended to have more body fat than did the other groups.
Conclusions: Our results show that the relation between PBF and BMI in HA women differs from that of EA and AA women.
Key Words: Body composition • ethnicity • percentage body fat • body mass index • Hispanic Americans • Hispanics • dual-energy X-ray absorptiometry • DXA
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONSUBJECTS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Higher prevalences of overweight and obesity, as measured in terms of body mass index (BMI; in kg/m2), have been documented among HAs compared with other ethnic groups (1, 2). In a recent study, Casas et al (3) reported that healthy Hispanic women have higher adiposity and lower amounts of fat-free mass than do white women. This higher adiposity has been associated with multiple metabolic syndromes in Hispanic men and with higher fasting insulin concentrations in Hispanic men and women (4). Higher amounts of body fat in HA adults may constitute a public health concern for this population, because evidence shows that HAs are at increased risk for obesity-related conditions such as cardiovascular disease (5) and type 2 diabetes (6).
Body weight adjusted for stature is often used as an alternative to the estimation of adipose tissue mass when evaluating individuals or populations for obesity (7). Many studies have shown that BMI is a reasonable index of adiposity (8–11), given that body weight and stature are simple, inexpensive, safe, and practical measurements to acquire, particularly for groups with limited access to health care. Ethnic differences in the relation between BMI and percentage body fat (PBF) have been found. More specifically, at the same BMI, PBF differed between Asian, African American (AA), and white persons after sex, age, height, and weight were controlled for (12, 13). This relation has not been investigated in HAs. The purpose of this study was to assess the relation between BMI and PBF in HA adults, and to determine whether this relation differs from that in AA and European American (EA) adults.
|
SUBJECTS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONSUBJECTS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Descriptive characteristics, including the age, sex, and ethnicity distributions of the subjects, are shown in Table 1
. In all 11 studies, participants were recruited by asking for healthy persons; thus, all subjects defined themselves as healthy. Moreover, subjects who self-reported undernutrition, HIV-AIDS, alcoholism, or diabetes in the questionnaire were excluded from the study. The percentage of subjects in the sample who reported that they smoked was 18.9%, and 40.8% of subjects reported that they exercised for 
4 h/wk, an arbitrary threshold selected under the assumption that exercise levels exceeding this threshold may affect body composition. Because the objective of the present study was to explore the relation between PBF and BMI in the general public, we included smoking status and exercise level as covariates in the statistical analysis and did not exclude persons who engaged in these behaviors.
|
We compared the body weight value measured on the scale with a DXA-derived body weight value. The latter was calculated by adding the values for soft tissue and bone mineral content obtained with DXA. To reduce possible measuring bias, particularly at BMI values > 35, subjects who had a difference of > 2.0 kg between the DXA-measured weight and the scale-measured weight were excluded from the analyses.
Statistical analyses
Single regression models in which PBF was regressed on BMI were used to identify outliers. Subjects with PBF > 2.5 SEE (the estimated SD of the error term in the regression model) were also excluded from subsequent analyses. The effect of such exclusion was evaluated in a sensitivity analysis (see Discussion). After exclusion, a total of 487 men (nEA = 192, nAA = 148, and nHA = 147) and 933 women (nEA = 448, nAA = 304, and nHA = 181) were included in the analyses. The average age of subjects in the sample was 45.9 y (range: 18–110 y).
Multiple linear regression models were used to identify significant differences in PBF predicted from BMI for men and women of the 3 ethnic groups. The scatter plot of PBF versus BMI, showing a nonlinear trend in the distribution, is shown in Figure 1. To improve the linearity of the relation between the dependent variable and the independent variable, the reciprocal of BMI (ie, 1/BMI; in m2/kg), defined as INVBMI, was used as a predictor of PBF measured with DXA (Figure 1). A dummy coding system was used to identify ethnicity. The predictor model for each sex defined PBF as a function of ethnicity, smoking status, exercise level, age, INVBMI, and the interaction between INVBMI and ethnicity. This model is defined by the following equation:
|
 | (1) |
 |
4 h/wk exercising and 0 for other subjects. An interaction term defining age x ethnicity was added to the model to test for differences in the effect of age on the prediction equation across the 3 ethnic groups. Because we found no significant differences in the effect of age across ethnic groups in the prediction of PBF by INVBMI, this component was removed from the model for the final analysis. All analyses were performed by using the linear regression option of SPSS, version 10 (SPSS Inc, Chicago).
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONSUBJECTS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
. In this model, the estimates for bo represent EAs, and the model compares EAs with both AAs and HAs. To test whether differences existed between AAs and HAs, a second model was developed in which bo represented HAs (results not shown).
|
The results of the predictor model were used to estimate the trends of predicted PBF in a BMI range of 15–50 for women and 18–50 for men, within each ethnic group. This estimation was performed at consistent values for exercise level, smoking status, and age according to sex, ie, we used the averages for exercise (0.55 ± 0.50 for men and 0.42 ± 0.49 for women), smoking (0.27 ± 0.45 for men and 0.19 ± 0.39 for women), and age (44.7 ± 18.9 y for men and 46.5 ± 18.2 y for women).
The predicted PBF values according to BMI values between 15 and 50 for women and 18 and 50 for men are shown in Figure 2. Interestingly, in women, the trend of predicted PBF values among the ethnic groups differs according to BMI: at BMI < 30, HA tended to have higher PBF than did EA and AA, and at BMI > 35, EA tended to have higher PBF than did the other groups.
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONSUBJECTS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
The estimation of PBF from BMI by sex in each ethnic group (Figure 2
) shows almost identical PBF values in AA and HA men at BMIs of 18 to 24, in HA and EA men at BMIs of 24 to 31, and in AA and EA men at BMIs of 39 to 50. In women, we found almost identical PBF values at BMIs of 15 to 25 for AAs and EAs, 29 to 35 for EAs and HAs, and 39 to 46 for AAs and HAs. It is important to emphasize that the PBF values obtained from BMI in this study are estimates determined on the basis of our sample and that significant differences were only found in women. To assess how sensitive the results were, we tested the multiple regression models without excluding the outliers and without excluding those persons who met the exclusion criteria described previously. The results obtained were consistent across all the conditions; there were significant differences in the slopes between EA and HA women and between AA and HA women, but not between AA and EA women or between any combination of the ethnic groups in men.
Three aspects of our model deserve further discussion. These aspects are the use of INVBMI as a predictor, the inclusion of smokers in the sample, and the inclusion of exercisers in the sample. The prediction model used in this study was developed on the basis of linear multiple regression, and thus it appeared advisable to maximize the linearity of the predictor variable before modeling. The use of the inverse of BMI in the model was supported by the observation that it increased the linearity of the predictor (17).
This study investigated the prediction of PBF from BMI in the general public, and it is known that a substantial portion of the US population uses tobacco. This fact supported the inclusion of smokers in our sample, with a statistical adjustment for the effect of smoking as a covariate. According to the Centers for Disease Control and Prevention, the prevalence of cigarette smoking by adults in the United States ranges from 13.9% to 31.5%, depending on the state of residence. In our sample, 18.9% of subjects smoked, which is somewhat lower than the smoking rate of 21.9% in the state of New York (Internet: http://www2.cdc.gov/nccdphp/osh/state/rpt_epi_display.asp?rpt_id=E1). Similarly, exercise level was used as a covariate in the prediction equation, because 22.2% of the subjects in the sample reported exercising for > 4 h/wk.
We measured PBF with the DXA method, which is a well-validated tool for the measurement of body fat. The use of DXA to quantify the outcome variable reduced biases that might have resulted from measurement errors. However, the statistical analyses were limited by the lack of inclusion of other possible confounders that might have influenced the relation between BMI and PBF, such as menopausal status, hormone use, or dietary habits, and by the arbitrary selection of 4 h/wk as a cutoff for exercise level.
An important issue to consider when interpreting the results of this study is the use of the term Hispanic. Members of this ethnic group carry the cultural and genetic background of admixed populations with different combinations of Amerindian, European, and African ancestry. These populations are highly admixed, not only in terms of genetic background, but also in terms of cultural and dietary factors. Therefore, dividing the category Hispanic into subgroups with genetic, cultural, and dietary similarities would provide a more accurate research tool and deserves further exploration. Nonetheless, this investigation supports the importance of using prediction equations for PBF as a tool for studying obesity in epidemiologic and public health applications. BMI can be used by the general public to assess a person’s risk for chronic disease and illnesses, because BMI is easily calculated and is therefore accessible to the layperson. Reducing the risk of obesity-related comorbidities by empowering members of at-risk minority groups to self-assess their risk should be a priority of public health and educational leaders.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONSUBJECTS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  D. A Corley, A. Kubo, and W. Zhao Abdominal obesity, ethnicity and gastro-oesophageal reflux symptoms Gut, June 1, 2007; 56(6): 756 - 762. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 I. S Rogers, A. R Ness, C. D Steer, J. C. Wells, P. M Emmett, J. R Reilly, J. Tobias, and G. D. Smith Associations of size at birth and dual-energy X-ray absorptiometry measures of lean and fat mass at 9 to 10 y of age. Am. J. Clinical Nutrition, October 1, 2006; 84(4): 739 - 747. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
M. Ramirez-Zea, B. Torun, R. Martorell, and A. D Stein Anthropometric predictors of body fat as measured by hydrostatic weighing in Guatemalan adults. Am. J. Clinical Nutrition, April 1, 2006; 83(4): 795 - 802. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 R. C. Browning, E. A. Baker, J. A. Herron, and R. Kram Effects of obesity and sex on the energetic cost and preferred speed of walking J Appl Physiol, February 1, 2006; 100(2): 390 - 398. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
S. Chung, M.-Y. Song, H.-D. Shin, D.-Y. Kim, Q. He, S. Heshka, J. Wang, J. Thornton, B. Laferrere, F. X. Pi-Sunyer, et al. Korean and Caucasian overweight premenopausal women have different relationship of body mass index to percent body fat with age J Appl Physiol, July 1, 2005; 99(1): 103 - 107. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 Y.-J. Liu, F.-H. Xu, H. Shen, Y.-Z. Liu, H.-Y. Deng, L.-J. Zhao, Q.-Y. Huang, V. Dvornyk, T. Conway, K. M. Davies, et al. A Follow-Up Linkage Study for Quantitative Trait Loci Contributing to Obesity-Related Phenotypes J. Clin. Endocrinol. Metab., February 1, 2004; 89(2): 875 - 882. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 J-R Long, P-Y Liu, Y-J Liu, Y Lu, D-H Xiong, L Elze, R R Recker, and H-W Deng APOE and TGF-{beta}1 genes are associated with obesity phenotypes J. Med. Genet., December 1, 2003; 40(12): 918 - 924. [Full Text]
|
|
|
|
|
|
G. L Blackburn National Health and Nutrition Examination Survey: where nutrition meets medicine for the benefit of health Am. J. Clinical Nutrition, August 1, 2003; 78(2): 197 - 198. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
) and women (
).
