Percentage of body fat cutoffs by sex, age, and race-ethnicity in the US adult population from NHANES

Transcription

1 Percentage of body fat cutoffs by sex, age, and race-ethnicity in the US adult population from NHANES Moonseong Heo, Myles S Faith, Angelo Pietrobelli, and Steven B Heymsfield ABSTRACT Background: To date, there is no consensus regarding adult cutoffs of percentage of body fat or estimated cutoffs on the basis of nationally representative samples with rigorous body-composition measurements. Objective: We developed cutoffs of percentage of body fat on the basis of the relation between dual-energy x-ray absorptiometry measured fat mass and BMI (in kg/m 2 ) stratified by sex, age, and race-ethnicity by using NHANES data. Design: A simple regression (percentage of body fat = b 0 + b BMI) was fit for each combination of sex (men and women), 3 age groups (18 29, 30 49, and y of age), and 3 race-ethnicity groups (non-hispanic whites, non-hispanic blacks, and Mexican Americans). Model fitting included a consideration of complex survey design and multiple imputations. Cutoffs of percentage of body fat were computed that corresponded to BMI cutoffs of 18.5, 25, 30, 35, and 40 on the basis of estimated prediction equations. Results: R 2 ranged from 0.54 to 0.72 for men (n = 6544) and 0.58 to 0.79 for women (n = 6362). In men, the percentage of body fat that corresponded to a BMI of 18.5, 25, 30, 35, and 40 across age and racial-ethnic groups ranged from 12.2% to 19.0%, 22.6% to 28.0%, 27.5% to 32.3%, 31.0% to 35.3%, and 33.6% to 37.6%, respectively; the corresponding ranges in women were from 24.6% to 32.3%, 35.0% to 40.2%, 39.9% to 44.1%, 43.4% to 47.1%, and 46.1% to 49.4%, respectively. The oldest age group had the highest cutoffs of percentage of body fat. Non-Hispanic blacks had the lowest cutoffs of percentage of body fat. Cutoffs of percentage of body fat were higher in women than in men. Conclusions: Cutoffs of percentage of body fat that correspond to the current US BMI cutoffs are a function of sex, age, and race-ethnicity. These factors should be taken into account when considering the appropriateness of levels of percentage of body fat. Am J Clin Nutr 2012;95: INTRODUCTION Excess fatness is a major global public health concern (1) that is associated with both mortality (2, 3) and medical comorbidities such as diabetes and cardiovascular disease (4, 5). Excess fatness in most epidemiologic studies has been defined on the basis of overweight or obese BMI (in kg/m 2 ) criteria developed by the National Heart, Lung, and Blood Institute (6, 7). However, it has often been debated whether BMI represents body fat adequately (8 11). For example, although the relation of mortality and comorbidities with BMI is well recognized as J-shaped (3), risk of mortality could monotonically increase with fat mass (12 14). Therefore, it remains an open question as to whether the definition of human obesity should be based on BMI or body fat. There have been considerable efforts aimed at linking BMI with the percentage of body fat on relatively small sample sizes with varied body-composition measurements (15 21). However, among these efforts, Gallagher et al (21) used the percentage of body fat measured by DXA 5, which has been accepted as a reference method of body-composition measurement for largescale studies and proposed an approach for the development of guidelines that predict the percentage of body fat on the basis of BMI from a relatively large, if convenient, sample size. For this reason, estimates of percentage of body fat of Gallagher et al (21) have served as percentage of body fat ranges for research purpose (see, eg, reference 22). Nevertheless, there exists no consensus of percentage of body fat criteria to define obesity or excess percentage of body fat (23) despite that the American Association of Clinical Endocrinology/American College of Endocrinology suggested 25% and 35% of body fat as cutoffs for obesity in men and women, respectively (24). The NHANES had begun to use DXA scanners to measure total body fat and, thus, the percentage of body fat since 1999 when it became a biannual regular survey. This introduction of a reference measurement method in the NHANES offers an important opportunity to further develop and update predictions of the percentage of body fat on the basis of BMI with in vivo body fat measured by using DXA from an unprecedented nationally representative large-size sample. In this article, as an advanced, but still initial, step toward cutoff guidelines of 1 From the Albert Einstein College of Medicine, Bronx, NY (MH); the University of North Carolina Chapel Hill, Chapel Hill, NC (MSF); the Verona University Medical School, Verona, Italy (AP); and the Pennington Biomedical Research Center, Baton Rouge, LA (AP and SBH). 2 Part of results from this study were presented at the Ninth International Symposium on In Vivo Body Composition Studies held in Hangzhou, China, on May 21-24, No funding was received for this study. 4 Address correspondence to M Heo, Department of Epidemiology and Population Health, Albert Einstein College of Medicine, 1300 Morris Park Avenue, Belfer 13th Floor, Bronx, NY moonseong.heo@ einstein.yu.edu. 5 Abbreviations used: DXA, dual-energy x-ray absorptiometry; lowess, locally weighted scatterplot smoothing; MEX, Mexican American; NHB, non-hispanic black; NHW, non-hispanic white. Received August 22, Accepted for publication December 15, First published online February 1, 2012; doi: /ajcn Am J Clin Nutr 2012;95: Printed in USA. Ó 2012 American Society for Nutrition

2 PERCENTAGE OF BODY FAT CUTOFFS 595 TABLE 1 Sample sizes stratified by sex, race-ethnicity, and age used for this study from NHANES Men Women Race-ethnicity and age n 2 % 2 % 3 n 2 % 2 % 3 NHW y y y Total MEX y y y Total NHB y y y Total Total y y y Total All weighted percentages were estimated with the NHANES sampling design effects taken into account. MEX, Mexican American; NHB, non- Hispanic black; NHW, non-hispanic white. 2 Unweighted. 3 Weighted. percentage of body fat, we present updated prediction equations for the percentage of body fat on the basis of BMI and develop cutoffs of percentage of body fat in reference to BMI cutoffs of 18.5, 25, 30, 35, and 40. These cutoffs define underweight, normal, overweight, and obesity class I, II, and III status ranges (6, 7). These developed cutoffs of percentage of body fat were stratified by sex, age, and race-ethnicity by using the NHANES data survey from 1999 to SUBJECTS AND METHODS We used adult subjects aged 18 y from the US NHANES data set, which is a combined data set of 3 biannual cross-sectional waves from , , and Subjects were measured in vivo for regional body fat distributions by using DXA scanners in mobile examination centers. A Hologic QDR-4500A fan-beam densitometer (Hologic Inc) was used throughout the NHANES with Hologic software version 8.26:a3* (Hologic Inc). Because the algorithm of the QDR-4500A densitometer (Hologic Inc) overestimated lean mass by 5 6 1% (25), the NHANES DXA lean mass was decreased by 5%, and an equivalent kilogram weight was added to the fat mass without affecting the total mass (26). However, selfreported pregnant women (even those women with negative pregnancy test results), and subjects heavier than 300 lb (136 kg) or taller than 6 ft 5 in (195 cm) were excluded from the DXA measurement on the basis of NHANES criteria (27). Anthropometric measures were also available including measured height and weight with BMI calculated as weight (in kg) divided by height (in m 2 ), which was used for the current study. With respect to missing data, ;22% of subjects in the NHANES had at least one missing regional body fat measurement that was due to invalid DXA scanning (28). Reasons for missing data included, but were not limited to, meeting aforementioned exclusion criteria, the presence of certain nonremovable objects (eg, prostheses), excess X-ray noise, and positional problems (28). For this reason, the NHANES generated 5 imputed data sets for missing DXA regional body-composition measurements; imputation for missing DXA data in later NHANES surveys are currently ongoing (28). Nevertheless, imputation was not applied to pregnant women or to subjects with amputated body parts and, thus, these subjects were excluded from the current study. Also excluded were subjects who had incomplete anthropometric BMI measurements and highly variable imputed DXA measurements. All of these exclusions resulted in a total sample of n = 12,906 subjects (6544 men and 6362 women) who had complete information, observed or imputed, on all variables used for this study. FIGURE 1. DXA-measured percentage of body fat compared with BMI in men and women on the basis of n = 1000 randomly selected subjects for each sex from the first imputed data set. The solid line represents the locally weighted scatterplot smoothing fit to all subjects for each sex. DXA, dual-energy x-ray absorptiometry.

3 596 HEO ET AL FIGURE 2. DXA-measured percentage of body fat compared with 1 4 BMI (1/BMI) in men and women on the basis of n = 1000 randomly selected subjects for each sex from the first imputed data set. The solid line represents the locally weighted scatterplot smoothing fit to all subjects for each sex. DXA, dual-energy x-ray absorptiometry. For the development of the percentage of body fat and total body fat in reference to BMI status, we stratified the sample by sex, race-ethnicity, and 3 age groups. The percentage body fat differs between men and women (21) and is also associated with ethnicity (29). In this study, we considered the following NHANES race-ethnicity classifications (with the NHANES variable name RIDRETH1) categorized by the NHANES: NHW, NHB, and MEX who are Hispanic and of Mexican origin. Other Hispanics and other racial-ethnic and multi-racial groups were excluded because of very small sample sizes. Age was categorized into the following 3 groups: 18 29, 30 49, and y. This age categorization was based on the following considerations: 1) total potassium declines after ;30 y of age in both men and women (30); 2) age of 50 y is approximately when women progress into menopause, which affects body fat mass (31); and 3) unweighted sample sizes are approximately equal across the 3 age groups for both men and women in all ethnic groups except NHW who include more older subjects than other groups. However, weighted sample sizes and percentages at the US adult population level were not TABLE 2 Subject characteristics 1 Variables Race-ethnicity NHW NHB MEX P Men n Age (y) , Weight (kg) , Height (cm) , BMI (kg/m 2 ) Fat (kg) , Percentage of body fat , Women n Age (y) , Weight (kg) , Height (cm) , BMI (kg/m 2 ) , Fat (kg) , Percentage of body fat All means 6 SEMs were estimated with NHANES sampling design effects taken into account. P values were computed on the basis of testing the significance of 2 dummy variables that represented the 3 ethnic groups in the survey linear regression models. For the imputed variables fat and percentage of body fat, means 6 SEMs and P values were based on the pooling of results from 5 sets of National Center for Health Statistics imputed data. Minimum and maximum values of fat and percentage of body fat were obtained from the entire 5 imputed data sets. 2 Mean 6 SEM (all such values). 3 Minimum maximum (all such values).

4 PERCENTAGE OF BODY FAT CUTOFFS 597 necessarily comparable across the 3 age groups; details information in regard to the sample size is presented in Table 1. Statistical analysis The NHANES implemented a multistage, complex survey design to increase the representativeness of the US adult population. Therefore, by adhering to the analytic guidelines suggested by the NHANES (32), we took into account sampling strata, the primary sampling unit, and individual sampling weights for descriptive and inferential statistical analyses by using SAS PROC SURVEYFREQ, SURVEYMEANS, and SURVEYREG with SAS v9.1.3 (SAS Institute Inc). Statistical analyses were applied to each of 5 imputed data sets, which yielded 5 sets of results from which final results were obtained on the basis of a pooling method proposed by Rubin (33) by applying SAS PROC MIANALYZE (SAS Institute Inc). For descriptive statistics, we reported the mean and SEM. For inferential statistics, we reported estimated regression coefficients and SEs. Graphical presentations and a lowess fit (34) were performed by using S-plus v8.1.1 (TIBCO Software Inc). Determinations of ranges of percentage of body fat First, we inspected a scatter plot of percentage of body fat compared with BMI for men and women and showed that the relation between them was apparently nonlinear as shown by the lowess fit (Figure 1). The nonlinear pattern of the relation was consistent across the 9 combinations of age and ethnicity strata in both men and women. To linearize the relation, we transformed BMI into the inverse BMI (ie, 1 4 BMI). The linearity of 1 4 BMI was apparent on the basis of the lowess fits (Figure 2). Of note, Figures 1 and 2 were based on n = 1000 randomly selected subjects for each sex from the first imputed data set. However, the lowess fits were based on entire subjects for each sex. Again, the linear pattern was consistent across the combinations of the strata. Therefore, we fit the simple linear model Percentage of body fat ¼ b 0 þ b BMI with SAS PROC SURVEYREG (SAS Institute Inc). Goodness of fit was quantified by R 2, which was a square of the correlation r between percentage of body fat and 1 4 BMI. We determined cutoffs of percentage of body fat by replacing BMI with corresponding BMI cutoffs in the prediction equations: Percentage of body fat ¼ b 0 þ b :5 ð2þ and Percentage of body fat ¼ b 0 þ b Percentage of body fat ¼ b 0 þ b Percentage of body fat ¼ b 0 þ b Percentage of body fat ¼ b 0 þ b Finally, on the basis of estimated cutoffs, we estimated the sex-age-race-ethnicity specific population-level prevalence of ð1þ ð3þ ð4þ ð5þ ð6þ TABLE 3 Prediction equations on the basis of estimated coefficients in the regression form of percentage of body fat = b 0 + b BMI 1 Age and race-ethnicity n b 0 6 SE b 1 6 SE R 2 Men y NHW NHB MEX y NHW NHB MEX y NHW NHB MEX Women y NHW NHB MEX y NHW NHB MEX y NHW NHB MEX All estimated b 0 and b 1 coefficients were significant with P, All results were based on the pooling of results from the 5 sets of National Center for Health Statistics imputed data and were obtained with NHANES sampling design effects taken into account. MEX, Mexican American; NHB, non-hispanic black; NHW, non-hispanic white. subjects who were conditionally BMI percentage of body fat discordant. Specifically, we estimated the prevalence of subjects whose percentage of body fat was greater than or equal to the sexage-race-ethnicity specific cutoffs that corresponding to a BMI of 25and 30 in subjects those who had a BMI 25 and 30, respectively. RESULTS Descriptive statistics Subject characteristics are described in Table 2. NHW and MEX subjects were the oldest and youngest, respectively, in both sexes. In men, the percentage of body fat was the lowest in the NHB group despite that BMI was comparable across the 3 racial-ethnic groups, and NHB subjects were older than MEX subjects. In women, however, the difference in percentage of body fat was not biologically meaningful, if significant, across the 3 groups despite that the NHB group had the greatest BMI, and MEX subjects were the youngest. Prediction equations Estimated regression coefficients of Equation 1 along with the sample size used in each stratum categorized by sex, age,

5 598 HEO ET AL and ethnicity are show in Table 3. The predicted relation between the percentage of body fat and BMI is depicted in Figure 3, which shows that the relation depended on age and raceethnicity in both men and women. Intercept estimates and slope estimates differed across the strata. However, the estimated regression coefficient b 1 was significantly,0 across all strata defined by sex, age, and race-ethnicity, which supported that the percentage of body fat is an increasing function of BMI with an upper limit. All of the estimated coefficients were significant at a 2-sided P, , and the goodness of fit R 2 ranged from 0.54 to 0.72 for men and 0.58 to 0.79 for women across sex, age, and race-ethnicity. Although R 2 was slightly higher in women, negative changes in the percentage of body fat per unit change in 1 4 BMI (and, thus, positive changes per BMI) were greater in men in MEX subjects regardless of age and in and y-old NHW subjects (Table 3). In contrast, the changes were smaller in men in NHB subjects regardless of age. Predicted percentage of body fat and its cutoffs Cutoffs of percentage of body fat predicted on the basis of the estimated equation presented in Table 3 are shown in Table 4. Across age and racial-ethnic groups, cutoffs of percentage of body fat in men that corresponded to a BMI of 18.5, 25, 30, 35, and 40 ranged from 12.2% to 19.0%, 22.6% to 28.0%, 27.5% to 32.3%, 31.0% to 35.3%, and 33.6% to 37.6%, respectively. In women, cutoffs for these same BMI values ranged from 24.6% to 32.3%, 35.0% to 40.2%, 39.9% to 44.1%, 43.4% to 47.1%, and 46.1% to 49.4%, respectively (Table 4). For example, in men, fat cutoffs of percentage of body fat that corresponded to a BMI of 30 ranged from 27.5% (18 29-y-old NHBs) to 32.3% (50 84-y-old NHWs), whereas for women, cutoffs ranged from 39.9% (18 29-y-old NHBs) to 44.1% (50 84-y-old MEXs). Additional inspection of Table 4 and Figure 3 revealed the following patterns. First, cutoffs of percentage of body fat were highest in the oldest age group regardless of sex and race-ethnicity. Cutoffs of percentage of body fat monotonically increased with age in NHWand NHB subjects, except for in NHW men with a BMI of 40. However, in MEX subjects the mid y-old group had the lowest cutoffs of percentage of body fat for BMI 30 for both men and women (Table 4). Furthermore, in MEX men, the youngest group (ie, age y) had the highest predicted percentage of body fat for BMI.45 (Figure 3). Second, with respect to racialethnic differences in percentage of body fat, NHB subjects had a lower percentage of body fat for the same BMI compared with that of NHW and MEX subjects regardless of age and sex. Third, cutoffs of percentage of body fat were higher in women than in men for any given BMI regardless of age and race-ethnicity (Figure 3). Finally, with regard to the predicted percentage of body fat BMI association, changes in the predicted percentage of body fat per BMI unit change were relatively smaller for larger BMIs. This result was due to the slopes of the tangent lines of the estimates curves that declined with increasing BMI. This relation was also reflected on the lowess curves, which showed that the percentage of body fat bent downward after a BMI of ;30 for both men and women (Figure 1). As could be inferred from Table 4 or from the prediction equations, rates of change in percentage of body FIGURE 3. Predicted relation between percentage of body fat and BMI (in kg/m 2 ) by age and race-ethnicity in men and women. From left to right, vertical lines represent BMI of 18.5, 25, and 30, respectively. MEX, Mexican American; NHB, non-hispanic black; NHW, non-hispanic white.

6 PERCENTAGE OF BODY FAT CUTOFFS 599 TABLE 4 Cutoffs of percentage of body fat in reference to BMI cutoffs (in kg/m 2 ) in men and women 1 Cutoffs of percentage of body fat Age and race-ethnicity BMI of 18.5 BMI of 25 BMI of 30 BMI of 35 BMI of 40 Men y NHW NHB MEX y NHW NHB MEX y NHW NHB MEX Women y NHW NHB MEX y NHW NHB MEX y NHW NHB MEX Cutoffs were determined on the basis of estimated prediction equations presented in Table 2. MEX, Mexican American; NHB, non-hispanic black; NHW, non-hispanic white. fat per BMI unit between a BMI of 25 and 30 were much smaller than those between a BMI of 18.5 and 25 for any combination of the 3 demographic factors. Furthermore, for every BMI interval, percentage of body fat changes per BMI unit decreased with increasing age within each race-ethnicity for both men and women. Prevalence of subjects who were conditionally BMI percentage of body fat discordant The population level weighted prevalence estimates of subjects who were conditionally BMI percentage of body fat discordant is presented in Table 5. In general, prevalence estimates were associated with the 3 demographic factors. The prevalence ranged from 10.0% to 31.7% in subjects who had BMI 25 and from 7.0% to 22.8% in subjects who had BMI 30. The prevalence was higher in older subjects regardless of sex and BMI. The prevalence was the lowest in the NHW group for women regardless of age and BMI levels. For the lower BMI of 25, the prevalence was the lowest in NHB group for men. Overall, the prevalence depended on the 3 demographic factors considered in this study. DISCUSSION The principal finding of this study was that cutoffs of percentage of body fat that corresponded to BMI cutoffs substantially varied depending on age, sex, and race-ethnicity, which are perhaps the most representative demographic factors. Specifically, the cutoff of percentage of body fat was the highest for the same BMI in the oldest age group, was the lowest in NHB subjects, and was higher in women (Tables 3 and 4). The relation between BMI and the predicted percentage of body fat was associated with different racial-ethnic backgrounds (Figure 3). These findings were consistent with those from numerous previous studies. For example, ethnic differences in DXA-measured percentage of body fat for BMI in young men and women were also noted by Jackson et al (35) in a recent study. Therefore, these factors should be taken into consideration in clinical settings when body-composition measurements are evaluated. At the same time, unlike BMI, anticipated clinical guidelines concerning body fat should be flexible on the basis of those factors. To this end, we believe that the developed prediction equations (Table 3) with high R 2 may have primary clinical usefulness because they are very simple to apply and useful to predict percentage of body fat for any given BMI. Furthermore, more careful clinical attention and/or an additional medical examination may be warranted for subjects who are conditionally BMI percentage of body fat discordant whose BMI is lower but with a higher percentage of body fat as presented in Table 5. The finding that the percentage of body fat is a nonlinear function of BMI does not necessarily imply that allocations of weight gain to body composition are not necessarily proportional as far as fat increases are concerned. Strong linear relations between BMI and both fat mass and fat-free mass were observed

7 600 HEO ET AL TABLE 5 Population-level sex-age-race-ethnicity specific prevalence estimates of subjects who were conditionally BMI percentage of body fat discordant 1 Race-ethnicity NHW NHB MEX Percentage of body fat greater than or equal to cutoffs corresponding to BMI (in kg/m 2 ) of 25 in subjects with BMI 25 Men aged y y y Women aged y y y Percentage of body fat greater than or equal to cutoffs corresponding to BMI of 30 in subjects with BMI 30 Men aged y y y Women aged y y y All values are prevalences (%) 6 SEs. All results were based on the pooling of results from 5 sets of National Center for Health Statistics imputed data and were obtained with NHANES sampling design effects taken into account. MEX, Mexican American; NHB, non-hispanic black; NHW, non-hispanic white. in the NHANES data (data not shown). This finding suggested that weight gain may proportionally be distributed to fat mass and fat-free mass although the proportionality likely depends much on age. Nevertheless, the percentage of body fat appears to change at a BMI of ;30 kg/m 2, after which the percentage of body fat tends to plateau in men and women (Figures 1 and 3). Furthermore, the percentage of body fat appears to be 50% and 60% in men and women, respectively (Figure 1). Indeed, the estimated intercepts (Table 3) serve as upper bounds, or asymptotes, for the percentage of body fat at very large BMI values. Therefore, the percentage of body fat is limited by ;55% and ;65%, which were the greatest intercept estimates, in men and women, respectively, regardless of ethnicity, age, and BMI. Of note, the addition of the 14 BMI 2 term to the simple Equation 1 models did not substantially, although significant in some cases, increase R 2 in either men or women. Increases ranged from to in men and to in women. However, the majority of previous prediction equations had been based on an assumed linear relation between the percentage of body fat and BMI (16 20), although studies tried to fit nonlinear equations. For example, Jackson et al (15) used the log BMI with percentage of body fat on the basis of underwater weighing, and Gallagher et al (21) used 1 4 BMI with the DXAmeasured percentage of body fat. The approach of Gallagher et al (21) is the same as that applied in the current study and, thus, makes comparison feasible. Compared with the cutoffs reported in Gallagher et al (21), the current cutoffs of percentage of body fat tend to be higher, especially in younger groups, regardless of age, sex, and ethnicity. For example, the percentage of body fat that corresponded to a BMI of 25 for subjects aged y was estimated at ;33% and ;35% in White and African American women, respectively, in Gallagher et al (21) [Table 3 in Gallagher et al (21)], whereas our newly developed cutoffs of percentage of body fat are ;35% and ;37% in NHWs and NHBs, respectively, for people aged y (Table 4). The underlying reasons for the overall increase in percentage of body fat for the same BMI are unknown even if the samples were different between our study and the study of Gallagher et al (21). Potential causes include variation in DXA system calibration (given that the NHANES DXA-measured fat mass was increased by 5% as previously mentioned), samplecharacteristic differences, and ecologic factors such as secular trends in diet and activity levels and their translation into bodycomposition effects. Flegal et al (36) also reported the relation between the percentage of body fat and BMI by using the NHANES for subjects aged 20 y. However, Flegal et al (36) applied an empirical approach on the basis of descriptive statistics, which was not model based and, therefore, differs from other approaches in the literature including ours used in the current study. Specifically, Flegal et al (36) matched weighted percentiles of the percentage of body fat with those of BMI in an ageand sex-specific fashion by using percentiles of the percentage of body fat as the reference (in contrast, we used BMI cutoffs as references). Although the approach of Flegal et al (36) could provide more accurate BMI values that correspond to given intervals of percentage of body fat and age, it does not provide

8 PERCENTAGE OF BODY FAT CUTOFFS 601 a prediction equation for a given BMI value. Despite these different approaches, both approaches agree on a similar conclusion that the BMI percentage of body fat relation varies by age, sex, and race-ethnicity. The DXA method is not available in most primary care clinical settings, although it is primarily used in research settings. Anthropometric indicators such as BMI and skinfold thickness measurements are frequently used to derive estimates of percentage of body fat, perhaps primarily because of the little cost and ease of measurement (37). For this reason, these methods and others, such as bioimpedance analysis, are used for screening purposes followed by more precise and sophisticated, and yet costly, methods that include DXA, computed tomography, and MRI (38). Although DXA systems often provide slightly different values for the percentage of body fat, as do all crossevaluated body-composition methods, the NHANES DXAmeasurement and -calibration approaches are extremely well documented in publications and on the Web and, thus, are comparable and reliable. Nevertheless, because the ranges of percentage of body fat in the current report were developed with a specific DXA system and related calibration (39), it is important that the estimates of percentage of body fat provided by other available clinical and research methods, including other DXA systems, are calibrated to the values acquired in NHANES. Otherwise, the ranges of NHANES percentage of body fat will not be reliable when applied in these settings. Therefore, although correlations among DXA systems for values of percentage of body fat are significant (40), an important need exists for standardization of the percentage of body fat, which we believe remains as an important issue in the field. The current study had limitations. First, the developed ranges of percentage of body fat were not developed on the basis of comorbidity and mortality outcomes. Therefore, risk ratios across cutoffs of percentage of body fat are not known, and it is also unknown if the developed cutoffs are optimal for maximizing sensitivity and specificity in diagnosing comorbidities. Once developed on the basis of comorbidity and mortality outcomes, such cutoffs of percentage of body fat will undoubtedly have greater clinical use. Nevertheless, the choice of comorbid health outcomes is both important and complicated; it will require longterm follow-up of NHANES subjects for reliable assessments of mortality and morbidity. Second, although NHANES data included pediatric sample with subjects,18 y old, we did not attempt to develop prediction equations because the development of weight status criteria on the basis of cutoffs of percentage of body fat in the pediatric population should be based on population percentiles of percentage of body fat (41) as was adopted for the development of weight-status criteria on the basis of BMI percentiles. Because the pediatric population, unfortunately, does not have well-established cutoffs of percentage of body fat that define normal weight and obesity (42), a future study should address this issue. Third, the percentage of body fat BMI relation may depend on the age within each age stratum; especially in older participants within the y-old age stratum as noted by Flegal et al (36). In conclusion, NHANES cutoffs of percentage of body fat as developed in the current report can serve as a useful research and clinical tool. Future studies are appropriate for continually updating the DXA database on nationally representative samples and to further explore topics such as the development of bodycomposition ranges in relation to morbidity and mortality in both adult and pediatric populations. We are grateful to the anonymous reviewers for their valuable suggestions that resulted in a great improvement of the manuscript. The authors responsibilities were as follows MH and SHB: conceived and designed the study; MH: acquired data, conducted statistical analyses, and had primary responsibility for the final content of the manuscript; MH, MSF, and SBH: analyzed and interpreted data; MH, AP, and SBH: drafted the manuscript; and all authors: read, provided critical revisions to, and approved the final manuscript. None of the authors had a conflict of interest. REFERENCES 1. Finucane MM, Stevens GA, Cowan MJ, Danaei G, Lin JK, Paciorek CJ, Singh GM, Gutierrez HR, Lu Y, Bahalim AN, et al; Global Burden of Metabolic Risk Factors of Chronic Diseases Collaborating Group. National, regional, and global trends in body-mass index since 1980: systematic analysis of health examination surveys and epidemiological studies with 960 country-years and 9.1 million participants. Lancet 2011;377: Flegal KM, Graubard BI, Williamson DF, Gail MH. Cause-specific excess deaths associated with underweight, overweight, and obesity. JAMA2007;298: Berrington de Gonzalez A, Hartge P, Cerhan JR, Flint AJ, Hannan L, MacInnis RJ, Moore SC, Tobias GS, Anton-Culver H, Freeman LB, et al. Body-mass index and mortality among 1.46 million white adults. N Engl J Med 2010;363: Mokdad AH, Ford ES, Bowman BA, Dietz WH, Vinicor F, Bales VS, Marks JS. Prevalence of obesity, diabetes, and obesity-related health risk factors, JAMA 2003;289: Must A, Spadano J, Coakley EH, Field AE, Colditz G, Dietz WH. The disease burden associated with overweight and obesity. JAMA 1999; 282: National Institutes of Health. National Heart, Lung, and Blood Institute Clinical Guidelines on the Identification, and Treatment of Overweight and Obesity in Adults: The Evidence Report. Bethesda, MD: National Institutes of Health, Pi-Sunyer FX. NHLBI Obesity Education Initiative Expert Panel on the identification, evaluation, and treatment of overweight and obesity in adults the evidence report. Obes Res 1998;6:51S 209S. 8. Romero-Corral A, Somers VK, Sierra-Johnson J, Thomas RJ, Collazo- Clavell ML, Korinek J, Allison TG, Batsis JA, Sert-Kuniyoshi FH, Lopez- Jimenez F. Accuracy of body mass index in diagnosing obesity in the adult general population. Int J Obes (Lond) 2008;32: Bergman RN, Stefanovski D, Buchanan TA, Sumner AE, Reynolds JC, Sebring NG, Xiang AH, Watanabe RM. A better index of body adiposity. Obesity (Silver Spring) 2011;19: Okorodudu DO, Jumean MF, Montori VM, Romero-Corral A, Somers VK, Erwin PJ, Lopez-Jimenez F. Diagnostic performance of body mass index to identify obesity as defined by body adiposity: a systematic review and meta-analysis. Int J Obes (Lond) 2010;34: Green DJ. Is body mass index really the best measure of obesity in individuals? J Am Coll Cardiol 2009;53:526; author reply Allison DB, Faith MS, Heo M, Kotler DP. Hypothesis concerning the U-shaped relation between body mass index and mortality. Am J Epidemiol 1997;146: Allison DB, Zhu SK, Plankey M, Faith MS, Heo M. Differential associations of body mass index and adiposity with all-cause mortality among men in the first and second National Health and Nutrition Examination Surveys (NHANES I and NHANES II) follow-up studies. Int J Obes Relat Metab Disord 2002;26: Zhu S, Heo M, Plankey M, Faith MS, Allison DB. Associations of body mass index and anthropometric indicators of fat mass and fat free mass with all-cause mortality among women in the first and second national health and nutrition examination surveys follow up studies. Ann Epidemiol 2003;13: Jackson AS, Stanforth PR, Gagnon J, Rankinen T, Leon AS, Rao DC, Skinner JS, Bouchard C, Wilmore JH. The effect of sex, age and race on estimating percentage body fat from body mass index: The Heritage Family Study. Int J Obes Relat Metab Disord 2002;26: Jackson AS. Research design and analysis of data procedures for predicting body density. Med Sci Sports Exerc 1984;16:

9 602 HEO ET AL 17. Jackson AS, Pollock ML, Ward A. Generalized equations for predicting body density of women. Med Sci Sports Exerc 1980;12: Deurenberg P, Yap M, van Staveren WA. Body mass index and percent body fat: a meta analysis among different ethnic groups. Int J Obes Relat Metab Disord 1998;22: Deurenberg P, Weststrate JA, Seidell JC. Body-mass index as a measure of body fatness - age-specific and sex-specific prediction formulas. Br J Nutr 1991;65: 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: Gallagher D, Heymsfield SB, Heo M, Jebb SA, Murgatroyd PR, Sakamoto Y. Healthy percentage body fat ranges: an approach for developing guidelines based on body mass index. Am J Clin Nutr 2000; 72: Oreopoulos A, Padwal R, McAlister FA, Ezekowitz J, Sharma AM, Kalantar-Zadeh K, Fonarow GC, Norris CM. Association between obesity and health-related quality of life in patients with coronary artery disease. Int J Obes (Lond) 2010;34: Ho-Pham LT, Campbell LV, Nguyen TV. More on body fat cutoff points. Mayo Clin Proc 2011;86:584; author reply AACE/ACE Obesity Task Force. AACE/ACE position statement on the prevention, diagnosis, and treatment of obesity. Endocr Pract 1998;4: Schoeller DA, Tylavsky FA, Baer DJ, Chumlea WC, Earthman CP, Fuerst T, Harris TB, Heymsfield SB, Horlick M, Lohman TG, et al. QDR 4500A dual-energy X-ray absorptiometer underestimates fat mass in comparison with criterion methods in adults. Am J Clin Nutr 2005;81: National Center for Health Statisitic. Documentation, codebook, and frequencies: dual-energy x-ray absortiometry. National Health and Nutrition Examination Survey Available from: cdc.gov/nchs/data/nhanes/dxa/dxx_c.pdf (cited 30 November 2011). 27. National Center for Health Statisitic. Technical documentation for the dual energy x-ray absorptiometry (DXA) multiple imupation data files. National Health and Nutrition Examination Survey. Available from: techdoc.pdf (cited 30 November 2011). 28. Schenker N, Borrud LG, Burt VL, Curtin LR, Flegal KM, Hughes J, Johnson CL, Looker AC, Mirel L. Multiple imputation of missing dual-energy X-ray absorptiometry data in the National Health and Nutrition Examination Survey. Stat Med 2011;30: Fernández JR, Heo MS, Heymsfield SB, Pierson RN Jr, Pi-Sunyer FX, Wang ZM, Wang J, Hayes M, Allison DB, Gallagher D. Is percentage body fat differentially related to body mass index in Hispanic Americans, African Americans, and European Americans? Am J Clin Nutr 2003;77: He Q, Heo M, Heshka S, Wang J, Pierson RN Jr, Albu J, Wang Z, Heymsfield SB, Gallagher D. Total body potassium differs by sex and race across the adult age span. Am J Clin Nutr 2003;78: Guo SS, Zeller C, Chumlea WC, Siervogel RM. Aging, body composition, and lifestyle: the Fels Longitudinal study. Am J Clin Nutr 1999;70: National Center for Health Statisitic. Analytic and reporting guidelines. The National Health and Nutrition Examination Survey. Available from: 04/nhanes_analytic_guidelines_dec_2005.pdf (cited 30 November 2011). 33. Rubin DB. Multiple imputation for nonresponse in surveys. New York, NY: Wiley & Sons, Cleveland WS. Robust locally weighted regression and smoothing scatterplots. J Am Stat Assoc 1979;74: Jackson AS, Ellis KJ, McFarlin BK, Sailors MH, Bray MS. Crossvalidation of generalised body composition equations with diverse young men and women: the Training Intervention and Genetics of Exercise Response (TIGER) Study. Br J Nutr 2009;101: Flegal KM, Shepherd JA, Looker AC, Graubard BI, Borrud LG, Ogden CL, Harris TB, Everhart JE, Schenker N. Comparisons of percentage body fat, body mass index, waist circumference, and waist-stature ratio in adults. Am J Clin Nutr 2009;89: Brodie D, Moscrip V, Hutcheon R. Body composition measurement: A review of hydrodensitometry, anthropometry, and impedance methods. Nutrition 1998;14: Heymsfield S, Lohman TG, Wang Z, Going SB, eds. Human body composition. 2nd ed. Champaign, IL: Human Kinetics, Kelly TL, Wilson KE, Heymsfield SB. Dual energy x-ray absorptiometry body composition reference values from NHANES. PLoS ONE 2009;4:e Ioannidou E, Padilla J, Wang J, Heymsfield SB, Thornton JC, Horlick M, Gallagher D, Pierson RN Jr. Pencil-beam versus fan-beam dualenergy X-ray absorptiometry comparisons across four systems: appendicular lean soft tissue. Acta Diabetol 2003;40:S Mueller WH, Harrist RB, Doyle SR, Labarthe DR. Percentiles of body composition from bioelectrical impedance and body measurements in US adolescents 8-17 years old: Project HeartBeat! Am J Hum Biol 2004;16: Pietrobelli A, Boner AL, Tato L. Adipose tissue and metabolic effects: new insight into measurements. Int J Obes (Lond) 2005;29:S