Estimates of body composition with dual-energy X-ray absorptiometry in adults 1 3

Transcription

1 Original Research Communications Estimates of body composition with dual-energy X-ray absorptiometry in adults 1 3 Chaoyang Li, Earl S Ford, Guixiang Zhao, Lina S Balluz, and Wayne H Giles ABSTRACT Background: Little is known about the distributions of percentage body fat (PBF), total body fat (TBF), and fat-free mass (FFM) in the adult population in the United States. Objectives: We sought to estimate the means and percentile cutoffs of PBF, TBF, and FFM and to assess the differences by sex, age, race-ethnicity, and body mass index in US adults. Design: Data from the National Health and Nutrition Examination Survey (NHANES), which were collected during the 6-y period from 1999 to 2004 and comprise a large nationally representative sample of the US population, were analyzed (n = 6559 men and 6507 nonpregnant women). TBF and FFM were measured by using dual-energy X-ray absorptiometry. PBF was calculated as TBF divided by total mass multiplied by 100. Results: There were large differences between men and women in unadjusted mean PBF (28.1% compared with 40.0%, P, 0.001), TBF (25.4 compared with 30.8 kg, P, 0.001), and FFM (62.3 compared with 44.0 kg, P, 0.001); the sex differences persisted across all body mass index categories after adjustment for age and race-ethnicity (all P, 0.001). The common percentile cutoffs of PBF, TBF, and FFM were estimated by sex, race-ethnicity, and age groups. Equations for the estimation of PBF (R 2 = 0.85), TBF (R 2 = 0.94), and FFM (R 2 = 0.94) according to demographic characteristics and simple anthropometric measures were generated. Conclusion: The estimates of means and percentile cutoffs for PBF, TBF, and FFM, on the basis of NHANES dual-energy X-ray absorptiometry data, provide a reference in the US adult population. Am J Clin Nutr 2009;90: INTRODUCTION The prevalence of overall obesity as measured by body mass index (BMI) has increased in the United States (1). Excess body fat is associated with increased cardiometabolic risks in the general population and in people with type 2 diabetes (2 4). BMI, calculated as weight (kg) divided by height (m) squared, has been used commonly to assess overall obesity in clinical settings and public health surveys. However, because BMI does not measure body fat directly and poorly distinguishes between fat mass and lean or bone mass, the use of BMI as an index of body fat for a person may be inaccurate. Dual-energy X-ray absorptiometry (DXA) is one of the most widely accepted methods to directly assess total and regional body fat and fat-free mass (FFM), which includes lean soft issues and bone mineral (5). This technique is noninvasive and easily applicable to persons with and without diseases, and it yields accurate and reliable results (6). Percentage body fat (PBF), which is calculated as total body fat (TBF) divided by total mass multiplied by 100, is a direct measure of a person s relative body fat. There exist significant differences in body fat across sex, age, and race-ethnicity (7). Previous National Health and Nutrition Examination Survey (NHANES) studies have mainly assessed the distributions of anthropometric data (ie, weight, height, waist circumference, and skinfold thickness) and used BMI as an approximate index for relative body fat. Although bioelectrical impedance analysis (BIA) was used to estimate body composition in the United States in the third NHANES ( ) (8), estimates based on BIA may be inaccurate because it is not a direct method for the assessment of body composition (9, 10). Moreover, estimates of the percentile cutoffs of body composition may provide a reference for the US population. To the best of our knowledge, there are no population-based quantitative estimates of means and percentile cutoffs of directly measured PBF, TBF, and FFM with DXA among adults in the United States thus far. The objectives of this study were 1) to analyze the DXA data from the NHANES and assess differences in the means of PBF, TBF, and FFM by sex, age, race-ethnicity, and BMI and 2) to estimate sex-, age-, and race-ethnicity specific percentile cutoffs of PBF, TBF, and FFM. METHODS Study design and population In NHANES , a cross-sectional sample was recruited through the use of a complex, stratified, multistage probability design to represent the noninstitutionalized civilian US population. People aged 60 y, non-hispanic blacks, and Mexican Americans were oversampled to ensure accurate esti- 1 From the Division of Adult and Community Health, National Center for Chronic Disease Prevention and Health Promotion, Centers for Disease Control and Prevention, Atlanta, GA. 2 The findings and conclusions in this report are those of the authors and do not necessarily represent the official position of the Centers for Disease Control and Prevention. 3 Address correspondence to C Li, Centers for Disease Control and Prevention, 4770 Buford Highway, MS K66, Atlanta, GA cli@ cdc.gov. Received May 28, Accepted for publication September 12, First published online October 7, 2009; doi: /ajcn Am J Clin Nutr 2009;90: Printed in USA. Ó 2009 American Society for Nutrition 1457

2 1458 LI ET AL mates in these groups. NHANES underwent ethical approval by the National Center for Health Statistics Institutional Review Board and Research Ethics Review Board and included written informed consent from each participant. After they were interviewed at home, participants were invited to attend the mobile examination center, where they were examined. Details about the plan and operation of the survey have been published (11). In the present study, we limited the analyses to men and nonpregnant women aged 20 y who had complete or imputed data on PBF, TBF, and FFM (n = 13,066). The final sample represents adults aged 20 y in the United States. DXA Detailed descriptions about the DXA data have been published (12, 13). In brief, whole-body DXA scans were administered in the NHANES mobile examination center to eligible participants during the 6-y period from 1999 to The participants with the following conditions were excluded from the DXA examination: 1) pregnancy assured by positive urine pregnancy test and/or by self-report at the time of the DXA examination, 2) self-reported history of radiographic contrast material (barium) use in past 7 d, 3) self-reported nuclear medicine studies in the past 3 d, and 4) self-reported weight (.300 lb or 136 kg) or height (.6 ft 5 in or 196 cm) in excess of the DXA table limit. The DXA scans provided bone and soft tissue measurements for the total body and regions (both arms and both legs, the trunk, and the head), including total mass (g), bone mineral content (g), bone area (cm 2 ), bone mineral density (g/cm 2 ), fat mass (g), and lean mass (g). Total PBF was calculated as TBF mass divided by total mass 100. Whole-body DXA scans were acquired with the use of a Hologic QDR-4500A fan-beam densitometer (Hologic Inc, Bedford, MA) with the Hologic DOC software (version 8.26:a3). The participants wore paper gowns and were in a supine position on the tabletop with their feet in a neutral position and hands flat by their sides. The DXA examinations were administered by certified radiology technologists. The participants were scanned with an X-ray source that used fan-beam scan geometry. The radiation exposure from DXA is extremely low (,10 lsv). Hologic Discovery software (version 12.1) was used to analyze the scans. Quality assurance and quality control of DXA scans A high-level quality control was maintained during the entire process of the DXA data collection and scan analysis. Staff from the National Center for Health Statistics and the NHANES data collection contractors monitored the acquisition performance of radiology technologists by in-person observations in the field. The technologists were trained annually to reinforce correct techniques and appropriate protocol. In addition, technologist performance codes were recorded and tracked by the NHANES quality control center at the Department of Radiology of the University of California, San Francisco (UCSF). Each participant and phantom scan was reviewed and analyzed by the UCSF with the use of standard radiologic techniques and study-specific protocols developed for NHANES. Expert review was conducted by the UCSF on all analyzed participant scans to verify the accuracy and consistency of the results. Multiple imputations Because the validity of the DXA data were influenced by age, weight, and height, invalid and missing data could not be considered a random subset of the original sample. To avoid bias due to nonrandom missing data, multiple imputations of the DXA data were performed. With the exception of pregnant women and those with amputations other than fingers or toes, all participants with invalid or missing data were included in the multiple imputation process. SAS-callable imputation and variance estimation software (IVEware), developed by the Survey Methodology Program at the Institute of Survey Research, University of Michigan, was used to impute NHANES DXA data. The IVEware module IMPUTE performs multiple imputations of missing values with the use of the sequential regression imputation method. Five complete records that contained valid and/or imputed values were created for each survey participant to allow the assessment of variability due to imputation. Use of the imputed data sets provides complete DXA data for all participants and ensures TABLE 1 Descriptive characteristics of participants aged 20 y in the analytic sample: National Health and Nutrition Examination Survey (NHANES) Men Women 2 Sample size [n (weighted %)] 6559 (48.8) 6507 (51.2) Race-ethnicity [n (weighted %)] Non-Hispanic white 3320 (72.0) 3234 (71.6) 3 Non-Hispanic black 1245 (10.0) 1309 (11.5) 4 Mexican American 1488 (8.0) 1433 (6.4) 5 Other race-ethnicity (10.0) 531 (10.6) 3 Age [n (weighted %)] y 2183 (41.0) 2018 (36.2) y 2024 (38.9) 2049 (38.7) y 1851 (17.2) 1846 (20.2) 5 80 y 501 (3.0) 594 (4.9) 5 BMI [n (weighted %)] 7,25.0 kg/m (31.1) 2147 (38.3) kg/m (40.7) 1961 (28.6) kg/m (19.0) 1256 (17.6) kg/m (9.2) 1060 (15.5) 5 Percentage body fat (%) Total body fat (kg) Fat-free mass (kg) Measured weight (kg) Measured height (cm) BMI (kg/m 2 ) Triceps skinfold thickness (mm) Two-sample t tests were used to test for equality between men and women. 2 Nonpregnant women only. 3 NS. 4,5 Significantly different from men: 4 P, 0.01, 5 P, Includes multiracial, other Hispanic, and all other race-ethnicities. 7 Data on BMI were missing for 165 participants (82 men and 83 women). 8 Weighted mean 6 SE (all such values).

3 BODY-COMPOSITION ESTIMATES IN ADULTS 1459 Statistical analyses We examined the distributions of PBF, TBF, and FFM by sex, race-ethnicity, age, and BMI. A 2-sample t test was used to assess the differences in the percentages of demographic characteristics and in the means of DXA data and anthropometric measures between men and women. Linear regression analyses were performed to estimate the predicted marginals (adjusted means) of PBF, TBF, and FFM adjusted for race-ethnicity and age stratified by sex and BMI, and to generate equations for the estimation of PBF, TBF, and FFM according to demographic characteristics and simple anthropometric measures. Pairwise comparisons of the unadjusted and adjusted means of PBF, TBF, FIGURE 1. Population distributions (weighted % of US adults) of percentage body fat (A), total body fat (B), and fat-free mass (C) measured by dual-energy X-ray absorptiometry among US adults aged 20 y, by sex [National Health and Nutrition Examination Survey (NHANES) ]. a more accurate SEE. A detailed description of the imputation procedures can be found in the Technical Documentation of Multiple Imputation of NHANES DXA Data (13). Age was categorized into 4 groups: 20 39, 40 59, 60 79, and 80 y. To estimate percentile cutoffs, the last 2 age groups were combined to form 1 age group (ie, 60 y) to ensure adequate sample size. Race-ethnicity was self-reported and was classified as non-hispanic white, non-hispanic black, Mexican American, and other race-ethnicity. Weight (kg) and height (cm) were measured with the use of standardized protocol and instruments (12). BMI (in kg/m 2 ) was calculated with the use of measured weight and height, and was categorized into 4 groups (,25, , , and 35) according to World Health Organization criteria (14). Triceps skinfold thickness was measured at the midpoint on the posterior surface of the right upper arm to the nearest 0.1 mm. FIGURE 2. Weighted means (and 95% CIs) of percentage body fat (A), total body fat (B), and fat-free mass (C) measured by dual-energy X-ray absorptiometry among US adults aged 20 y, by age, in men and women [National Health and Nutrition Examination Survey (NHANES) ].

4 1460 LI ET AL TABLE 2 Mean percentage body fat according to BMI categories among US adults: National Health and Nutrition Examination Survey (NHANES) Total percentage body fat BMI (in kg/m 2 ), Mean SE Mean SE Mean SE Mean SE Mean SE Men % body fat Crude Adjusted NHW 28.3 a a a a a 0.2 NHB 25.8 b b b b a 0.4 MEX 28.9 a a a a a 0.8 Other 27.9 a a a a a y 26.1 a a a a a y 28.7 b b b a a y 30.9 c c c b b y 30.6 c d d b a,b 1.0 Women Crude Adjusted NHW 39.7 a a a a a 0.2 NHB 40.9 b,c b b b b 0.3 MEX 41.6 b c a a b 0.3 Other 39.9 a,c a,c a a,b b y 37.8 a a a a a,b y 40.6 b b b a,c a y 42.5 c c c b b y 40.6 b c c b,c a,b n = 12,901 (165 participants with missing data on BMI were excluded). NHW, non-hispanic white; NHB, non-hispanic black; MEX, Mexican American. There were significant interactions between sex and BMI (P, 0.001), between race-ethnicity and BMI in men (P = 0.002) and women (P, 0.001), and between age and BMI in men (P, 0.001) and women (P, 0.001). Values within categories of race-ethnicity or age groups in the same column with different superscript letters are significantly different, P = (equivalent to P, 0.05 after Bonferroni correction for multiple comparisons); t tests were used to test for equality. 2 Adjusted for age and race-ethnicity. 3 Adjusted for age. 4 Adjusted for race-ethnicity. 5 Significantly different from men, P, and FFM between any 2 groups within 4 racial-ethnic groups, 4 age groups, and 4 BMI levels were performed with the use of a t test with a linear contrast in linear regression models. We estimated selected percentiles of PBF, TBF, and FFM (5, 10, 15, 25, 33.3, 50, 66.7, 75, 85, 90, and 95) by sex, race-ethnicity, and age. In the subgroup analyses, we present only the subtotal, rather than age-specific estimates, for participants with other race-ethnicity (Asian/Pacific Islanders, Native Americans, multiracial persons, and other Hispanic persons) because of small sample size. All analyses were conducted 5 times, once for each imputation data set. The combined estimate of PBF, TBF, and FFM from all 5 analyses was the mean of the 5 individual estimates (15, 16). The within-imputation variance (W) was calculated as the mean of 5 individual variance estimates, and the between-imputation (B) variance was calculated as the sample variance of the 5 individual estimates. The total variance (T) combines the within- and between-imputation variances as follows: T = W + (6/5) B. Results with a P value,0.001, equivalent to P, 0.05 after Bonferroni correction for the multiple comparisons, were considered to be statistically significant in 2-tailed t tests. Data analyses were conducted with the use of SAS, version 9.1 (SAS, Cary, NC) and SUDAAN software, release 9.0 (RTI International, Research Triangle Park, NC) to account for the complex sampling design. RESULTS A total of 13,066 adults (6559 men and 6507 nonpregnant women) 20 y of age who had complete data on PBF, TBF, and FFM measures were included in the analyses. Of these, 165 had missing data on BMI; thus, the analyses that involved BMI were based on 12,901 participants. The characteristics of the men and women in the analytic sample are described in Table 1. The population distributions (weighted %) of PBF, TBF, and FFM are shown in Figure 1. There are apparent differences in the distributions of the weighted % between men and women. The means of PBF, TBF, and FFM by age groups are shown in Figure 2. It appeared that PBF and TBF increased with age before age 55 y and decreased thereafter in women (P for linear

5 BODY-COMPOSITION ESTIMATES IN ADULTS 1461 TABLE 3 Mean total body fat according to BMI categories among US adults: National Health and Nutrition Examination Survey (NHANES) BMI (in kg/m 2 ) Total body fat, Mean SE Mean SE Mean SE Mean SE Mean SE kg kg total body fat Men Crude Adjusted NHW 26.0 a a a a a 0.6 NHB 23.8 b b b b a,b 1.3 MEX 24.2 b c b b b 1.3 Other 23.2 b a,c b b a,b y 23.7 a a a a a y 26.5 b b b a b y 27.4 c c c b c y 23.5 a c b,c a,b d 1.9 Women Crude Adjusted NHW 30.5 a a,b a a a 0.6 NHB 35.4 b a b a a 0.8 MEX 31.0 a b b b b 0.7 Other 28.5 c a,b b b a,b y 28.8 a a a a,b a y 32.5 b b b a a y 32.3 c c a b b y 26.7 d a,b c c c n = 12,901 (165 participants with missing data on BMI were excluded). NHW, non-hispanic white; NHB, non-hispanic black; MEX, Mexican American. There were significant interactions between sex and BMI (P, 0.001), between race-ethnicity and BMI in men (P = 0.003) and women (P, 0.001), and between age and BMI in men (P, 0.001) and women (P, 0.001). Values within categories of race-ethnicity or age groups in the same column with different superscript letters are significantly different, P = (equivalent to P, 0.05 after Bonferroni correction for multiple comparisons); t tests were used to test for equality. 2 Adjusted for age and race-ethnicity. 3 Adjusted for age. 4 Adjusted for race-ethnicity. 5 Significantly different from men, P, trend:,0.001). In men, the turning point occurred at age 65 y (P for linear trend:,0.001). FFM tended to decrease with age in both men and women (P for linear trend:,0.001). On average, women had 12% higher PBF than that of the men (P, 0.001) (Table 2), and the differences persisted after adjustment for race-ethnicity and age and across the 4 BMI categories. In men, non-hispanic blacks had a lower PBF than non-hispanic whites, Mexican Americans, and persons with other race-ethnicity (all P, 0.001). In women, both non-hispanic blacks and Mexican Americans had a higher PBF than non- Hispanic whites and women with other race-ethnicity (P, 0.001). However, non-hispanic black women had a lower PBF than non-hispanic white women within each category of BMI, which suggests possible confounding and/or interaction between race-ethnicity and BMI on PBF. There was a significant interaction between sex and BMI (P, 0.001), between raceethnicity and BMI in men (P = 0.002) and women (P, 0.001), and between age and BMI in men (P, 0.001) and women (P, 0.001) on mean PBF. On average, women had 5 kg greater TBF than that of the men (P values, 0.001) (Table 3), and the differences persisted after adjustment for race-ethnicity and age and across the 4 BMI levels. In men, non-hispanic whites had a higher TBF than non- Hispanic blacks, Mexican Americans, and persons with other race-ethnicity (all P, 0.001). In women, non-hispanic blacks had a higher TBF than non-hispanic whites, Mexican Americans, and persons with other race-ethnicity (all P, 0.001). However, non-hispanic black women had a lower TBF than non-hispanic white women within each category of BMI, which suggests possible confounding and/or interaction between raceethnicity and BMI on TBF. There was a significant interaction between sex and BMI (P, 0.001), between race-ethnicity and BMI in men (P = 0.003) and women (P, 0.001), and between age and BMI in men (P, 0.001) and women (P, 0.001) on mean TBF. On average, men had 18 kg greater FFM than that of the women (P values, 0.001) (Table 4), and the differences persisted after adjustment for race-ethnicity and age. The differences

6 1462 LI ET AL TABLE 4 Mean fat-free mass according to BMI categories among US adults: National Health and Nutrition Examination Survey (NHANES) BMI (in kg/m 2 ) Total fat-free mass, Mean SE Mean SE Mean SE Mean SE Mean SE kg kg fat-free mass Men Crude Adjusted NHW 63.1 a a a a a 0.8 NHB 64.4 a a a b a 1.2 MEX 57.4 b b b c b 1.0 Other 57.6 b b b c a y 63.1 a a a a a y 63.5 a a a b b y 59.3 b b b c c y 51.8 c c c d d 1.9 Women Crude Adjusted NHW 43.9 a a a a a 0.4 NHB 48.5 b b b b b 0.5 MEX 41.9 c c c c c 0.4 Other 41.2 c c c c a,c y 44.5 a a a a a y 45.2 a a b a a y 42.2 b b c b b y 37.8 c c d c c n = 12,901 (165 participants with missing data on BMI were excluded). NHW, non-hispanic white; NHB, non-hispanic black; MEX, Mexican American. There were significant interactions between sex and BMI (P, 0.001), between race-ethnicity and BMI in men (P = 0.04) and women (P, 0.001), and between age and BMI in men (P, 0.001) and women (P = 0.02). Values within categories of race-ethnicity or age groups in the same column with different superscript letters are significantly different, P = (equivalent to P, 0.05 after Bonferroni correction for multiple comparisons); t tests were used to test for equality. 2 Adjusted for age and race-ethnicity. 3 Adjusted for age. 4 Adjusted for race-ethnicity. 5 Significantly different from men, P, appeared to widen as BMI increased (15.7, 18.8, 21.1, and 23.8 kg greater FFM at BMI categories of,25, 25 29, 30 34, and 35, respectively). In men, both non-hispanic whites and non-hispanic blacks had a greater FFM than Mexican Americans, and persons with other race-ethnicity (all P, 0.001). In women, non-hispanic blacks had a greater FFM than non- Hispanic whites, Mexican Americans, and persons with other race-ethnicity (all P, 0.001). There was a significant interaction between sex and BMI (P, 0.001), between race-ethnicity and BMI in men (P = 0.04), and between age and BMI in men (P, 0.001) and women (P = 0.02) on mean FFM. However, there was no significant interaction between race-ethnicity and BMI in women (P = 0.08). The estimated percentile cutoffs of PBF by race-ethnicity and age groups in men and women are presented in Table 5. The differences in PBF across race-ethnicity and age groups appeared to be larger at lower percentiles (eg, 5, 10, 15, 25, and 33.3) than at upper percentiles (eg, 50, 66.7, 75, 85, 90, and 95). At the 90th percentile, there were small differences in PBF across racial-ethnic groups in both men (35 36%) and women (47 49%). The equations used to estimate PBF, TBF, and FFM according to demographic characteristics and simple anthropometric measures are shown in Table 6. The R 2 of all 3 equations ranged from 0.85 to 0.94, which indicates that 85 94% of the variations in PBF, TBF, and FFM may be accounted for by the combination of sex, age, race-ethnicity, BMI, triceps skinfold thickness, weight, and height. The estimated percentile cutoffs of TBF by race-ethnicity and age groups in men and women can be found under "Supplemental data" in the online issue. The differences in TBF across raceethnicity and age groups appeared to be larger at lower percentiles than at upper percentiles in men, but the differences were smaller at lower percentiles than at upper percentiles in women. The estimated percentile cutoffs of FFM by race-ethnicity and age groups in men and women can also be found under "Supplemental data" in the online issue. In contrast to PBF, the

7 BODY-COMPOSITION ESTIMATES IN ADULTS 1463 TABLE 5 Percentile cutoff estimates of percentage body fat by sex, race-ethnicity, and age among US adults aged 20 y: National Health and Nutrition Examination Survey (NHANES) Percentile n % body fat Men Total y y y y Non-Hispanic white y y y Non-Hispanic black y y y Mexican American y y y Other race-ethnicity Women Total y y y y Non-Hispanic white y y y Non-Hispanic black y y y Mexican American y y y Other race-ethnicity indicates that the estimate is not provided because the sample size was,30 at given subgroups. 2 Percentiles for participants with other race-ethnicity could not be estimated by age group because of small sample size. differences in FFM across race-ethnicity and age groups appeared to be smaller at lower percentiles than at upper percentiles. DISCUSSION By using a nationally representative sample of US adults, we described the distributions of PBF, TBF, and FFM and provided their quantitative estimates according to sex, age, race-ethnicity, and BMI. Our results clearly showed that women had, on average, 12% higher PBF, 5 kg greater TBF, and 18 kg less FFM than men, and the differences persisted across 4 BMI categories. There were relatively small, but statistically significant, differences in PBF, TBF, and FFM by race-ethnicity and age group. Whole-body DXA scans were not administered in the NHANES mobile examination center until in In NHANES III ( ), BIA data provided approximate body composition measures in Americans aged y (8). It is difficult to compare the absolute numbers of DXA in NHANES with those of the BIA data in NHANES III ( ) because 2 different methods were used to measure at the 2 different time periods. However, the patterns of the differences in body composition by sex, age, and race-ethnicity in the current study are similar to those reported previously, such that women had greater TBF and PBF than men, and non-hispanic black women and Mexican American women had greater TBF and PBF than non- Hispanic white women at all ages (8). Nevertheless, our results, based on NHANES DXA data, may provide more reliable and accurate estimates of PBF, TBF, and FFM than previous reports based anthropometric measures and BIA data. As shown in validation studies, BIA may underestimate PBF

8 1464 LI ET AL TABLE 6 Equations for the estimation of percentage body fat (PBF), total body fat (TBF), and fat-free mass (FFM) according to demographic characteristics and anthropometric measures Equation R 2 PBF (%) TBF (kg) FFM (kg) PBF = 0.56 BMI (kg/m 2 ) triceps skinfold thickness (mm) age (y) (if female) 1.76 (if African American) (if Mexican American) (if other race-ethnicity) TBF = 0.60 weight (kg) 0.25 height (cm) age (y) (if female) 1.26 (if African American) (if Mexican American) FFM = 0.40 weight (kg) height (cm) 0.06 age (y) (if African American) 0.32 (if Mexican American) or FFM weight (kg) TBF systematically compared with both the underwater weighing method (the gold standard for body composition) and DXA (9, 10). Our results highlight sex differences in body fat in the general adult population. Similar to previous findings in children, young adults, and adults (8, 9, 17 20), women in this study generally had greater estimated mean PBF and TBF than men. Our estimates, with a large population-based sample and DXA, conform to previous results with smaller samples or less accurate methods and provide reliable quantitative estimates of the difference in PBF and TBF between men and women. BMI has been used widely to define overall obesity in both pediatric and adult populations because of its simplicity, clinical availability, strong association with health risks and consequences, and high correlation with PBF (14, 21, 22), although it has been considered an inaccurate measure of body fat. To date, the universal cutoffs of and 30 have been used to define overweight and obesity in both men and women (14). A recent study has shown that BMI corresponds adequately well with PBF by sex and age subgroups (22). However, in agreement with previous findings (7, 8), our results suggest that a man and a woman may have a similar BMI but his or her relative or absolute body fat may differ appreciably. Note that the differences in PBF and TBF across age and raceethnic groups were more pronounced among men and women with a BMI,25 than among those with a BMI 25. Our results are consistent with the findings of a previous study (23) in which the researchers showed that mean values of PBF in the normal and overweight BMI groups were higher in older women than in younger women. Decrease in muscle density and increase in the accumulation of body fat mass with age (24, 25) may partially explain the differences in PBF between the younger and older adults. Consistent with other reports (8, 17 19), our results that non- Hispanic blacks had smaller mean PBF than non-hispanic whites across all BMI categories in both men and women suggest that racial-ethnic variations in body composition may exist. However, in comparison with differences in PBF between men and women (12%), differences in PBF between non-hispanic whites and non- Hispanic blacks across each category of BMI ( %) were relatively small in this study. It is unclear whether such small, albeit statistically significant, racial-ethnic differences in PBF may have clinical or public health significance in the diagnosis, assessment, prediction, and prevention of obesity-related disorders. Future research is warranted to assess the sex and ethnic variations in the association between PBF and cardiometabolic risks. The estimated percentile cutoffs of PBF, TBF, and FFM at commonly used percentiles by race-ethnicity and age in men and women may provide a useful reference in clinical settings and public health services. The equations provided in this article could be used to estimate a person s relative and absolute body fat according to his or her sex, age, race-ethnicity, body weight, height, BMI, and triceps skinfold thickness. Our equations are simple and have a predictive validity that is similar to or better than the equations generated in previous studies (26 29). To date, there is no agreement about the optimal cutoff of PBF for overweight and obesity. Our data suggest that the cutoffs of PBF at the 50th percentile for men (28%) and women (41%) (Table 5) corresponded approximately to the mean PBF in the BMI category of (Table 2). The cutoffs of PBF at the 75th percentile for men (32%) and women (45%) corresponded approximately to the mean PBF in the BMI category of The cutoffs of PBF at the 90th percentile for men (36%) and women (48%) corresponded approximately to the mean PBF in the BMI category of 35. Future studies may be needed to determine sex-specific cutoffs of PBF in relation to cardiometabolic risk factors, type 2 diabetes, and cardiovascular disease morbidity and mortality. Interpretation of our results is subject to 2 limitations. First, although DXA has been validated by the 4-compartment model and has been considered a reliable, accurate, and convenient method for the assessment of body composition, it has not been established as a gold standard for PBF because of variations with different manufacturers and beam configurations (30). Thus, caution is warranted when our results are compared with the findings of other studies. Second, multiply imputed data for participants with missing or invalid body-composition data (weighted %: 18.2%) may not perfectly represent the actual values. However, a large number of variables, including demographic, socioeconomic, and geographic variables, body measurements, health status, dietary intake, use of medications, blood test results, and variables related to the NHANES design, were included in the imputation models (15); therefore, the effect of imputed data on the validity of the body composition estimates could be minimal.

9 BODY-COMPOSITION ESTIMATES IN ADULTS 1465 In conclusion, by using DXA data from a nationally representative sample, we described the distributions and showed the differences in body composition by sex, age, race-ethnicity, and BMI categories. Our results highlighted the large sex differences and relatively small differences in PBF, TBF, and FFM by age and race-ethnicity across different BMI categories. Because direct measurement of body composition with the use of DXA is expensive, the estimated percentiles of PBF, TBF, and FFM in conjunction with the estimation equations according to demographic characteristics and simple anthropometric measures could be helpful to clinicians and researchers when they assess the relative and absolute body fat of their patients or study participants. The authors responsibilities were as follows CL, ESF, GZ, LSB, and WHG: study concept and design; CL and ESF: acquisition of data; CL, ESF, GZ, LSB, and WHG: analysis and interpretation of data; CL: draft of manuscript; CL, ESF, GZ, LSB, and WHG: critical revision of the manuscript for important intellectual content; CL and ESF: statistical analysis; ESF and LSB: administrative, technical, or material support; and ESF: study supervision. CL had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. None of the authors had a conflict of interest. REFERENCES 1. Ogden CL, Carroll MD, Curtin LR, McDowell MA, Tabak CJ, Flegal KM. Prevalence of overweight and obesity in the United States, JAMA 2006;295: Englyst NA, Crook MA, Lumb P, et al. Percentage of body fat and plasma glucose predict plasma sialic acid concentration in type 2 diabetes mellitus. Metabolism 2006;55: Katzmarzyk PT, Gagnon J, Leon AS, et al. Fitness, fatness, and estimated coronary heart disease risk: the HERITAGE Family Study. Med Sci Sports Exerc 2001;33: Segal KR, Dunaif A, Gutin B, Albu J, Nyman A, Pi-Sunyer FX. Body composition, not body weight, is related to cardiovascular disease risk factors and sex hormone levels in men. J Clin Invest 1987;80: Mazess RB, Barden HS, Bisek JP, Hanson J. Dual-energy X-ray absorptiometry for total-body and regional bone-mineral and soft-tissue composition. Am J Clin Nutr 1990;51: Plank LD. Dual-energy X-ray absorptiometry and body composition. Curr Opin Clin Nutr Metab Care 2005;8: Gallagher D, Visser M, Sepulveda D, Pierson RN, Harris T, Heymsfield SB. How useful is body mass index for comparison of body fatness across age, sex, and ethnic groups? Am J Epidemiol 1996;143: Chumlea WC, Guo SS, Kuczmarski RJ, et al. Body composition estimates from NHANES III bioelectrical impedance data. Int J Obes Relat Metab Disord 2002;26: Fogelholm M, van Marken LW. Comparison of body composition methods: a literature analysis. Eur J Clin Nutr 1997;51: Volgyi E, Tylavsky FA, Lyytikainen A, Suominen H, Alen M, Cheng S. Assessing body composition with DXA and bioimpedance: effects of obesity, physical activity, and age. Obesity (Silver Spring) 2008;16: Centers for Disease Control and Prevention. National Health and Nutrition Examination Survey. Available from: about/major/nhanes/datalink.htm (cited 27 August 2008). 12. Centers for Disease Control and Prevention. National Health and Nutrition Examination Survey: body composition procedures manual. Available from: (cited 27 August 2008). 13. Centers for Disease Control and Prevention. The dual energy X-ray absorptiometry (DXA) multiple imputation data files and technical documentation. Available from: major/nhanes/dxx/dxa.htm (cited 27 August 2008). 14. World Health Organization. Obesity: preventing and managing the global epidemic. Report of a WHO consultation. World Health Organ Tech Rep Ser 2000;894: Centers for Disease Control and Prevention. National Health and Nutrition Examination Survey: technical documentation for the dual energy X-tay absorptiometry (DXA) multiple imputation data files. Available from: pdf (cited 27 August 2008). 16. Rubin DB. Multiple imputation for nonresponse in surveys. New York, NY: John Wiley, Ellis KJ. Body composition of a young, multiethnic, male population. Am J Clin Nutr 1997;66: Ellis KJ, Abrams SA, Wong WW. Body composition of a young, multiethnic female population. Am J Clin Nutr 1997;65: Shaw NJ, Crabtree NJ, Kibirige MS, Fordham JN. Ethnic and gender differences in body fat in British schoolchildren as measured by DXA. Arch Dis Child 2007;92: Schutz Y, Kyle UU, Pichard C. Fat-free mass index and fat mass index percentiles in Caucasians aged y. Int J Obes Relat Metab Disord 2002;26: Cole TJ, Bellizzi MC, Flegal KM, Dietz WH. Establishing a standard definition for child overweight and obesity worldwide: international survey. BMJ 2000;320: Flegal KM, Shepherd JA, Looker AC, et al. Comparisons of percentage body fat, body mass index, waist circumference, and waist-stature ratio in adults. Am J Clin Nutr 2009;89: Movsesyan L, Tanko LB, Larsen PJ, Christiansen C, Svendsen OL. Variations in percentage of body fat within different BMI groups in young, middle-aged and old women. Clin Physiol Funct Imaging 2003; 23: Ryan AS, Nicklas BJ. Age-related changes in fat deposition in mid-thigh muscle in women: relationships with metabolic cardiovascular disease risk factors. Int J Obes Relat Metab Disord 1999;23: Goodpaster BH, Carlson CL, Visser M, et al. Attenuation of skeletal muscle and strength in the elderly: The Health ABC Study. J Appl Physiol 2001;90: Larsson I, Henning B, Lindroos AK, Naslund I, Sjostrom CD, Sjostrom L. Optimized predictions of absolute and relative amounts of body fat from weight, height, other anthropometric predictors, and age 1. Am J Clin Nutr 2006;83: Lean ME, Han TS, Deurenberg P. Predicting body composition by densitometry from simple anthropometric measurements. Am J Clin Nutr 1996;63: Peterson MJ, Czerwinski SA, Siervogel RM. Development and validation of skinfold-thickness prediction equations with a 4-compartment model. Am J Clin Nutr 2003;77: van der Ploeg GE, Gunn SM, Withers RT, Modra AC. Use of anthropometric variables to predict relative body fat determined by a four-compartment body composition model. Eur J Clin Nutr 2003;57: Williams JE, Wells JC, Wilson CM, Haroun D, Lucas A, Fewtrell MS. Evaluation of Lunar Prodigy dual-energy X-ray absorptiometry for assessing body composition in healthy persons and patients by comparison with the criterion 4-component model. Am J Clin Nutr 2006;83: