Journal of Exercise Physiologyonline (JEPonline)

Fat Estimation Of Black Females 20 Journal of Exercise Physiologyonline (JEPonline) Volume 8 Number 4 August 2005 Managing Editor Tommy Boone, Ph.D. Editor-in-Chief Robert Robergs, Ph.D. Review Board Todd Astorino, Ph.D. Julien Baker, Ph.D. Tommy Boone, Ph.D. Lance Dalleck, Ph.D. Dan Drury, DPE. Hermann Engals, Ph.D. Eric Goulet, M.Sc. Robert Gotshall, Ph.D. Len Kravitz, Ph.D. James Laskin, Ph.D. Jon Linderman, Ph.D. Derek Marks, Ph.D. Cristine Mermier, Ph.D. Daryl Parker, Ph.D. Robert Robergs, Ph.D. Brent Ruby, Ph.D. Jason Siegler, Ph.D. Greg Tardie, Ph.D. Ben Zhou, Ph.D. Official Research Journal of The American Society of Exercise Physiologists (ASEP) ISSN 1097-9751 Body Composition DEVELOPMENT AND VALIDATION OF AN ANTHROPOMETRICALLY BASED PREDICTION EQUATION FOR ESTIMATING THE PERCENT BODY FAT OF POST- MENOPAUSAL BLACK FEMALES LEANNE PETRY 1, LLOYD L. LAUBACH 2, PETER W. HOVEY 3, NIKKI L. ROGERS 4, BRADFORD TOWNE 4, AND W. CAMERON CHUMLEA 4 1 University of Dayton Research Institute/Materials Engineering Division, Dayton, OH, U.S.A. 2 University of Dayton/Department of Health and Sport Science, Dayton, OH, U.S.A. 3 University of Dayton /Department of Mathematics, Dayton, OH, U.S.A. 4 Wright State University School of Medicine/Lifespan Health Research Center, Dayton, OH, U.S.A. ABSTRACT Petry L, Laubach LL, Hovey PW, Rogers NL, Towne B, Chumlea CW. Development And Validation Of An Anthropometrically Based Prediction Equation For Estimating The Percent Body Fat Of Post- Menopausal Black Females. JEPonline 2005;8(4):20-28. Anthropometric equations developed specifically for the estimation of body composition parameters in Black females are limited. Data from the Lifespan Health Research Center were used to develop a new, easy to use equation to estimate the percent body fat of post-menopausal Black females using simple and easy to collect anthropometrics. The body composition of 72 post-menopausal Black females was measured by Dual Energy X-ray Absorptiometry (DXA). Validation (N=55) and cross-validation groups (N=17) were randomly assigned. Prediction models were developed using stepwise multiple regression analyses with percent body fat as the dependent variable and various anthropometrics as the independent variables. The chosen prediction equation uses hip circumference, wrist breadth, bicep skinfold, and weight as predictors of percent body fat. The suggested prediction equation is: %BF = -214.28 + 58.58*ln (hip Circ) - 23.47*ln (wrist BB) + 7.24*ln (bicep SKF) 0.00108*weight 2 and has a model R 2 =0.82 and a SEE=1.0%. Key Words: African American, Body Composition, DXA, Ethnicity, Women

Fat Estimation Of Black Females 21 INTRODUCTION Prediction equations for estimating the body composition of White females from anthropometrics are available in the literature (1-6). These generalized equations, however, are not age-matched to postmenopausal Black females (5) and are recognized to perform less satisfactorily when applied to Black females (6). Additionally, bone breadths have been reported for White females but not for Black females (7). Compared to White females, Black females are known to have a higher bone mineral density (8), greater amount of body fat (3,9-11), and differing extremity lengths (9). In addition, the prevalence of obesity in Black females is twice that of White females (12-14). While the associations between body composition and cardiovascular disease (CVD) risk in Black and White females are still under study, Black females are at relatively greater CVD risk at all ages (14-16). Increased bone mineral density, on the other hand, is advantageous to Blacks as it inhibits the onset of osteoporosis (17,18). Collectively, these observations indicate that prediction equations specific to demographic groups most at risk of disease are needed to estimate body composition, especially in Black females. To the best of our knowledge, no anthropometric equations for the estimation of body composition parameters in post-menopausal Black females have been developed using Dual Energy X-ray Absorptiometry (DXA). Five studies using DXA methods measured the body composition of Black females (9,10,19-21), but none correlated DXA results with anthropometric data for predictive purposes. One study looked specifically at compartmentalization (19), two studies looked at the distribution of fat by comparing waist-to-hip ratios (9,10), and several other studies compared body density to bioelectrical impedance (5,6,20,21). The overall goal of this study was to address voids in the research with regard to predicting body composition measurements in different ethnic groups. The specific goals of the study were to: (1) assess the body composition of post-menopausal Black females using DXA, (2) develop a prediction equation for percent body fat from the assessment methods, and (3) provide regression equations for the prediction of body composition parameters from simple anthropometrics. METHODS Subjects Seventy-two adult females from the greater metropolitan Dayton, Ohio, U.S.A. were recruited by telephone and word-of-mouth based upon menopausal status and self-reported Black ethnicity. All subjects provided informed consent and completed menstrual history, medical history, and physical activity questionnaires. Menopause was defined as the absence of a menstrual cycle (clinical or surgical) for 12 continuous months. All subjects were post-menopausal. Subjects physical activity levels varied from sedentary to active. The majority of subjects were hospital employees, members of the academic community, or members of local churches. Procedures Subjects underwent body composition examinations that included DXA and standard anthropometric assessment (22,23). The data obtained from these examinations were then used to develop an equation to predict percent body fat from anthropometrics. Data were collected during a 4 hour visit to the Lifespan Health Research Center (LHRC) at Wright State University (WSU) School of Medicine, Dayton, Ohio, U.S.A.. The Institutional Review Boards for the University of Dayton (UD), Dayton, Ohio, U.S.A. and WSU approved the use of human subjects. Subjects received financial remuneration for their participation.

Fat Estimation Of Black Females 22 Table 1. Descriptive characteristics of validation group (N = 55). Variable Mean±SD Range Age (yrs.) 60.9±9.6 41.6 79.9 Height (cm) 162.7±5.8 150.9 177.8 Weight (kg) 82.5±15.2 50.6 117.9 DXA (% Fat) 36.9±6.9 15.6 47.6 DXA Fat Mass (kg) 31.3±10.3 7.9 57.4 DXA Fat Free Mass (kg) 51.4±7.3 36.8 72.5 BMI (kg/m 2 ) 31.2±5.7 16.8 43.3 Waist to Hip Ratio 0.89±0.06 0.76 1.06 SKF: Bicep (mm) 18.7±7.3 4.6 33.9 SKF: Tricep (mm) 26.8±6.6 10.4 38.7 SKF: Midaxillary (mm) 22.4±6.1 7.2 35.2 SKF: Suprailiac (mm) 24.7±7.2 5.1 36.0 SKF: Sum4 (mm) 92.7±22.9 35.5 131.1 Upper Arm Circ (cm) 35.2±4.9 23.2 48.5 Abdominal Circ (cm) 101.8±12.2 71.7 125.2 Hip Circ (cm) 114.0±11.4 88.5 144.0 Thigh Circ (cm) 56.6±7.1 38.2 72.0 Calf Circ (cm) 38.4±4.6 29.0 51.3 Truck Depth (cm) 29.1±4.4 18.1 38.6 BB: Bicristal (cm) 31.4±2.5 25.7 38.6 BB: Biacromial (cm) 38.4±1.7 34.9 42.6 BB: Wrist (cm) 5.3±0.3 4.6 5.8 BB: Elbow (cm) 6.8±0.5 6.1 8.1 BB: Knee (cm) 10.2±1.0 8.6 12.6 Fat mass, fat free mass, and percent fat were measured by DXA (Hologic QDR4500 Elite, Waltham, MA, Software Version 2.1). Anthropometric assessment of body composition consisted of height and weight, in addition to numerous bone breadths, circumferences, and skinfold measurements using protocols similar to those in the Anthropometric Standardization Reference Manual (22). All variables were measured to the nearest mm. Standing height was measured with a wall-mounted digital stadiometer, and body weight was measured to the nearest 0.1 kg on a digital scale. A Lange caliper was used to measure skinfold thickness (SKF) at the following four sites: bicep, tricep, suprailiac, and midaxillary. Upper arm circumference was measured with a flexible steel tape. Circumferences of the abdomen, hips, thighs, and calves were measured with a soft flexible tape. Trunk depth and five skeletal frame size measures collected as bone breadths of the wrist, elbow, knee, shoulders (biacromial), and hips (bicristal) were measured using a sliding caliper. Two measurements were taken at each anthropometric site by two experienced laboratory technicians, and the average of all four readings was used in the data analysis. Standard protocol for the LHRC is to take all measurements on the left side of the body. Statistical Analyses The total sample was divided into a validation group (N=55) and a cross-validation group (N=17). Statistical analyses were conducted using the SAS (Statistical Analysis System, Version 8) software package (24). Descriptive statistics were calculated separately for the validation group and the cross-validation group. The criterion measure for percent body fat was derived from DXA. Data analysis consisted of stepwise multiple regression, with chronological age plus all anthropometrics submitted for possible entry into the equations. The validation group was used to derive the population-specific prediction equations that were cross-validated by the second group. The best fitting polynomial equation was chosen as the representative population specific equation.

Fat Estimation Of Black Females 23 Stepwise regression was used to build several candidate models to predict percent body fat from the various anthropometrics. The DXA determined fat mass, fat free mass, and percent body fat were used as the dependent variables, and the stepwise procedures selected from the remaining independent variables to build the best linear model for predicting the dependent variables. Percent body fat is a constrained variable because it cannot be less than 0 or greater than 100. Constrained variables can cause difficulties with fitting linear models; however, there are several approaches for minimizing the effect of the constraint. The standard approaches include transforming percent fat to the natural logarithm of percent fat divided by percent lean or the arcsine of percent body fat divided by 100. An additional approach used in this study was to build separate models for fat mass and fat free mass and then predict percent body fat from the predicted fat mass and fat free mass. Six models were selected as candidates for predicting percent body fat. The composite model (Model 1) predicts percent body fat from separate predictions of fat mass and fat free mass. The direct models (Models 2 through 4) used the DXA determined percent body fat as the independent variable. One direct model (Model 2) only included the actual values of the anthropometrics in the model building process, while the other direct models (Models 3 and 4) also included the logarithm and the square of each of the anthropometrics. Elbow breadth was forced in Model 4 instead of wrist breadth because elbow breadth was an important variable in another study of body composition prediction (7). The log-odds model (Model 5) used the logarithm of percent body fat divided by one minus percent body fat as the independent variable, while the arcsine model (Model 6) used the arcsine of percent body fat divided by 100. The composite, log-odds, and arcsine models all include the straight and transformed anthropometrics. Table 2. Descriptive characteristics of cross-validation group (N=17). Variable Mean±SD Range Age (yrs.) 56.5±8.4 44.5 76.3 Height (cm) 162.5±6.2 150.7 178.1 Weight (kg) 74.7±13.7 55.4 112.6 DXA (% Fat) 33.7±6.8 15.3 41.2 DXA Fat Mass (kg) 25.7±9.0 9.3 46.5 DXA Fat Free Mass (kg) 48.7±6.1 38.2 66.4 BMI (kg/m 2 ) 28.4±5.7 19.5 42.3 Waist to Hip Ratio 0.87±0.06 0.75 0.97 SKF: Bicep (mm) 14.6±7.7 3.7 28.3 SKF: Tricep (mm) 27.1±7.7 13.0 36.9 SKF: Midaxillary (mm) 19.6±7.3 7.5 31.8 SKF: Suprailiac (mm) 23.5±6.7 10.0 37.2 SKF: Sum4 (mm) 84.6±26.6 34.2 123.6 Upper Arm Circ (cm) 33.2±4.6 25.3 40.6 Abdominal Circ (cm) 95.0±11.6 74.5 114.0 Hip Circ (cm) 108.9±10.6 94.6 134.7 Thigh Circ (cm) 55.2±5.9 48.1 71.3 Calf Circ (cm) 37.6±4.1 32.3 47.9 Truck Depth (cm) 27.2±4.4 20.6 35.9 BB: Bicristal (cm) 29.3±2.2 26.1 34.6 BB: Biacromial (cm) 37.8±1.9 35.0 41.5 BB: Wrist (cm) 5.2±0.3 4.7 5.8 BB: Elbow (cm) 6.7±0.4 6.1 7.2 BB: Knee (cm) 9.7±0.9 8.5 12.2

Fat Estimation Of Black Females 24 Table 3. Pearson s Product Moment Correlation Coefficient Matrix (r), where for *df = 53, a value of 0.270 is needed to reach p = 0.05 and for **df = 15, a value of 0.482 is needed to reach p = 0.05. Validation Group (N=55)* DXA (% Fat) Cross-Validation Group (N=17)** DXA (% Fat) DXA (% FAT) 1.00 DXA (% FAT) 1.00 DXA Fat Mass 0.89 DXA Fat Mass 0.90 DXA Fat Free Mass 0.27 DXA Fat Free Mass 0.30 Age 0.07 Age 0.16 BMI 0.76 BMI 0.80 Height -0.20 Height -0.49 Weight 0.68 Weight 0.71 Hip Circ 0.78 Hip Circ 0.69 Thigh Circ 0.72 Thigh Circ 0.63 Abdominal Circ 0.76 Abdominal Circ 0.85 Upper Arm Circ 0.73 Upper Arm Circ 0.76 Calf Circ 0.52 Calf Circ 0.51 Waist to Hip Ratio 0.23 Waist to Hip Ratio 0.56 SKF: Bicep 0.75 SKF: Bicep 0.68 SKF: Midaxillary 0.52 SKF: Midaxillary 0.67 SKF: Suprailiac 0.61 SKF: Suprailiac 0.66 SKF: Tricep 0.68 SKF: Tricep 0.72 SKF: Sum4 0.76 SKF: Sum4 0.75 BB: Biacromial 0.17 BB: Biacromial 0.17 BB: Bicristal 0.59 BB: Bicristal 0.72 BB: Elbow 0.38 BB: Elbow 0.18 BB: Knee 0.53 BB: Knee 0.47 BB: Wrist -0.10 BB: Wrist -0.29 Trunk Depth 0.77 Trunk Depth 0.81

Fat Estimation Of Black Females 25 Table 4. Mean prediction errors (ME), standard deviation of the prediction errors (SD), and standard error of the mean prediction errors (SEE) in cross-validation group (N=17) for the predictive models. Dependent Variable ME SD SEE Model 1: %BF (Composite) -0.5809 5.0200 1.2175 Model 2: %BF (Direct w/actual Independents) -0.0417 4.8151 1.1678 Model 3: %BF (Direct w/transformed Independents) -0.0140 4.2823 1.0386 Model 4: %BF (Direct w/ Elbow BB) 0.4703 4.8406 1.1740 Model 5: %BF (Log-Odds) 0.4200 4.6103 1.1182 Model 6: %BF (Arcsine) -0.0532 4.2564 1.0323 RESULTS Descriptive data for the experimental and cross-validation groups are presented in Tables 1 and 2, respectively. The final prediction model selection was based on the performance of each in predicting percent body fat for the crossvalidation data set. The correlation matrix and summary statistics for the predictions of the cross-validation group are presented in Table 3 and Table 4, respectively. The standard deviation (SD) of the prediction error is an estimate of the typical error that will be seen in field use of the model. Roughly 68% of the time the model is applied, the predicted percent body fat will fall within one standard deviation of the DXA determined percent body fat. The best performing models on the crossvalidation group were the arcsine model (Model 6) and the direct model using transformed predictors (Model 3). The standard deviations were very similar with the arcsine model (Model 6) having a slight Figure 1. Plot of residual versus predicted values for validation (N=55) and crossvalidation (N=17) groups using Model 3, the direct model with transformed independent variables. Table 5. Regression coefficients, partial explanation of variance (Partial R 2 ), and significance of independent variables for validation group (N=55) in Model 3: %BF (Direct w/transformed Independents) equation. Variable Coefficient Partial R 2 Significance Intercept -214.2770 <0.0001 ln (Hip Circ) 58.5848 0.6369 <0.0001 ln (Bicep SKF) -23.4670 0.1031 0.0097 Weight 2 7.2368 0.0516 <0.0001 ln (Wrist BB) -0.0011 0.0264 0.0078 Model R 2 0.8179 <0.0001 R e s i d u a l 12 10 8 6 4 2 0-2 -4-6 -8 15 25 35 45 55 Predicted Cross-Validation Validation

Fat Estimation Of Black Females 26 advantage. The direct model using transformed predictors (Model 3) was selected as the suggested population-specific prediction equation since it would be easier to implement in the field. When applied to a group of post-menopausal Black females, the selected model had a SEE of 1.04% and an individual SD of 4.28% (Table 4). Figure 1 depicts the adequacy of the selected model. It shows the plot of residual versus predicted values for the validation and cross-validation groups using the direct model with transformed independent variables. The relationship between the size of the residuals and the predicted values shows the residuals are unrelated to the response. These randomly scattered data indicate a good fit and therefore a properly chosen model. Table 5 summarizes the regression results for the direct model using transformed predictors containing the estimated coefficients and the p-values for each term included in the model. The suggested prediction model is: %BF = -214.28 + 58.58*ln (hip Circ) - 23.47*ln (wrist BB) + 7.24*ln (bicep SKF) 0.00108*weight 2. For example, if a 60-year-old Black female had her hip Circ, wrist BB, bicep SKF, and weight measured to be 115.0 cm, 5.2 cm, 17.9 mm, and 83.5 kg respectively, then her percent body fat calculated using the population-specific prediction equation above would be: %BF = -214.28 + 58.58*ln (115.0) - 23.47*ln (5.2) + 7.24*ln (17.9) 0.00108*83.5 2 = -214.28 + 58.58*4.74-23.47*1.65 + 7.24*2.88 0.00108*6972.25 = -214.28 + 277.67 38.73 + 20.85 7.53 = 37.98 % or approximately 38 %. DISCUSSION The authors are unaware of anthropometric equations developed specifically for the estimation of body composition parameters in post-menopausal Black females. Evaluation of prediction models developed in this study and compared with percent body fat as measured by DXA provides a new and easy to use anthropometric equation for the prediction of percent body fat of post-menopausal Black females. To the best of the authors knowledge, no other population-specific prediction equations for post-menopausal Black females have been presented or cross-validated in the literature. The equation developed in this study uses hip circumference, wrist breath, bicep skinfold, and weight as predictors of percent body fat, and has a model R 2 of 0.82 and a SEE of 1.04%. Percent body fat measured by DXA was significantly correlated with most anthropometric variables (Table 3). Significant mean differences for the cross-validation group were noted for DXA fat mass and bicristal BB variables at p=0.05. These results are not surprising given bicristal BB was somewhat more highly correlated with DXA fat mass (r=0.76) and DXA fat free mass (r=0.64) than DXA percent body fat (r=0.59). Also, studies have established the incidence of higher bone mineral densities (8) and longer extremity lengths (9) in Black females. Chumlea et al. (7) have noted that bicristal and biacromial breadths are most associated with measures of fat and lean tissues, respectively. They report a poor association of wrist breadth with body fatness, and the significance of wrist breadth in multiple regression analyses. The population-specific regression equation developed in this study also uses wrist breadth as a measure of frame size, and found it to be negatively associated with percent body fat. Elbow breadth was not significantly correlated with body composition for this population of females, supporting the findings of Chumlea et al. in a White sample (7).

Fat Estimation Of Black Females 27 CONCLUSIONS The association of body composition with the risk of diabetes and cardiovascular disease coupled with the inherently greater risk of these diseases in Black females underscores the need for accurate methods of prediction of body fat in these individuals. The equation presented here addresses a void in body composition assessment research. It is more population specific than the widely used generalized equations (1-4), and therefore, provides a valid prediction of percent body fat for postmenopausal Black women. It is applicable to post-menopausal Black females ranging in age from 41-80 engaged in a variety of different activity levels. This equation is also practical and easy to use. ACKNOWLEDGEMENTS The authors would like to thank the subjects for participation in the Miami Valley Family Aging Study NIH R01AG 18719 and the data collection staff of the LHRC at WSU. The authors would also like to thank Dr. Roger M. Siervogel of LHRC-WSU for financial assistance under grant number NIH/NICHD R01HD12252. The support of Dr. Katy E. Marre of UD is also gratefully appreciated. Address for Correspondence: Petry L, University of Dayton Research Institute, 300 College Park, Dayton Ohio USA 45469-0137; Phone: (937) 255-3505; Fax: (937) 656-7129; Email: leanne.petry@wpafb.af.mil. REFERENCES 1. Jackson AS, Pollock ML, Ward A. Generalized equations for predicting body density of women. Med Sci Sports Exerc 1980;12:175-182. 2. Durnin JVGA, Wormersley J. Body fat assessed from total body density and its estimation from skinfold thickness: measurements on 481 men and women aged from 16 to 72 years. Br J Nutr 1974;32:77-97. 3. Zillikens MC, Conway JM. Anthropometry in blacks: applicability of generalized skinfold equations and differences in fat patterning between blacks and whites. Am J Clin Nutr 1990;52:45-51. 4. Teran JC, Sparks KE, Quinn LM, Fernandez BS, Krey SH, Steffee WP. Percent body fat in obese white females predicted by anthropometric measurements. Am J Clin Nutr 1991;53:7-13. 5. Chumlea WC, Guo SS, Kuczmarski RJ, Flegal KM, Johnson CL, Heymsfield SB, Lukaski HC, Friedl K, Hubbard VS. Body composition estimates from NHANES III bioelectrical impedance data. Int J Obes 2002;26:1596-1611. 6. Sun SS, Chumlea WC, Heymsfield SB, Lukaski HC, Schoeller D, Friedl K, Kuczmarski RJ, Flegal KM, Johnson CL, Hubbard VS. Development of bioelectrical impedance analysis prediction equations for body composition with the use of a multicomponent model for use in epidemiologic surveys. Am J Clin Nutr 2003;77:331-340. 7. Chumlea WC, Wisemandle W, Guo SS, Siervogel RM. Relations between frame size and body composition and bone mineral status. Am J Clin Nutr 2002;75:1012-1016. 8. Ortiz O, Russell M, Daley TL, Baumgartner RN, Waki M, Lichtman S, Wang J, Pierson Jr RN, Heymsfield SB. Differences in skeletal muscle and bone mineral mass between black and white females and their relevance to estimates of body composition. Am J Clin Nutr 1992;55:8-13. 9. Aloia JF, Vaswani A, Mikhail M, Flaster ER. Body composition by dual energy x-ray absorptiometry in black compared to white women. Osteoporos Int 1999;10:114-119. 10. Conway JM, Chanetsa FF, Wang P. Intraabdominal adipose tissue and anthropometric surrogates in african american women with upper- and lower-body obesity. Am J Clin Nutr

Fat Estimation Of Black Females 28 1997;66:1345-1351. 11. Lovejoy JC, Smith SR, Rood JC. Comparison of regional fat distribution and health risk factors in middle-aged white and african american women: the healthy transitions study. Obes Res 2001;9:10-16. 12. Irwin ML, Ainsworth BE, Stolarczyk LM, Heyward VH. Predictive accuracy of skinfold equations for estimating body density of african american women. Med Sci Sports Exerc 1998;30:1654-1658. 13. Cote KD, Adams WC. Effect of bone density on body composition estimates in young adult black and white women. Med Sci Sports Exerc 1993;25:290-296. 14. Folsom AR, Burke GL, Byers CL, Hutchinson RG, Heiss G, Flack JM, Jacobs Jr DR, Caan B. Implications of obesity for cardiovascular disease in blacks: the CARDIA and ARIC studies. Am J Clin Nutr 1991;53:1604S-1611S. 15. Gillum RF. Overweight and obesity in black women: a review of published data from the National Center for Health Statistics. J Nat Med Assoc 1987;79:865-871. 16. Kumanyika S. Obesity in black women. Epidemiol Rev 1987;9:31-50. 17. Cohn SH, Abesamis C, Zanzi I, Aloia JF, Yasumura S, Ellis KJ. Body elemental composition: comparison between black and white adults. Am J Physiol 1977;232:E419-E422. 18. Finkelstein JS, Cleary RL, Butler JP, Antonelli R, Mitlak BH, Deraska DJ, Zamora-Quezada JC, Neer RM. A comparison of lateral versus anterior-posterior spine dual energy x-ray absorptiometry for the diagnosis of osteopenia. J Clin Endocrinol Metab 1994;78:724-730. 19. Aloia JF, Vaswani A, Ma R, Flaster E. Body composition in normal black women: the fourcompartment model. J Clin Endocrinol Metab 1996;81:2363-2369. 20. Sparling PB, Millard-Stafford M, Rosskopf LB, Dicarlo LJ, Hinson BT. Body composition by bioelectric impedance and densitometry in black women. Am J Hum Biol 1993;5:111-117. 21. Jakicic JM, Wing RR, Lang W. Bioelectrical impedance analysis to assess body composition in obese adult women: the effect of ethnicity. Int J Obes 1998;22:243-249. 22. Lohman TG, Roche AF, Martorell R. Anthropometric Standardization Reference Manual. Champaign, IL: Human Kinetics Books, 1988. 23. National Center for Health Statistics. NHANES III reference manuals and reports (CD-ROM). Hyattville, MD: National Center for Health Statistics, 1996. 24. SAS Institute Inc. SAS/STAT user's guide. Version 8. Cary, NC:SAS Institute Inc., 1999.