BACKGROUND AND OBJECTIVES

Although the limitations of BMI have long been recognized, there are recent concerns that it is not a good screening tool for adiposity. We therefore examined the cross-sectional relation of BMI to adiposity among 6923 8- to 19-year-olds in the National Health and Nutrition Survey from 2011 through 2018.

METHODS

Participants were scanned with dual-energy x-ray absorptiometry. Adiposity was expressed as fat mass index (FMI, fat mass kg ÷ m2) and percentage of body fat (%fat). Lean mass was expressed as lean mass index (LMI, lean mass ÷ m2). Regression models and 2 × 2 tables were used to assess the relation of BMI to FMI, %fat, and LMI.

RESULTS

Age and BMI accounted (R2) for 90% to 94% of the variability of FMI and LMI in each sex. Associations with %fat were weaker (R2s ∼0.70). We also examined the screening abilities of a BMI ≥ Centers for Disease Control and Prevention 95th percentile for high levels of adiposity and LMI. Cut points were chosen so that prevalences of high values of these variables would be similar to that for high BMI. Of participants with a high BMI, 88% had a high FMI, and 76% had a high %fat. Participants with a high BMI were 29 times more likely to have a high FMI than those with lower BMIs; comparable relative risks were 12 for high %fat and 14 for high LMI.

CONCLUSIONS

Despite its limitations, a high BMI is a very good screening tool for identifying children and adolescents with elevated adiposity.

What’s Known on This Subject:

Although BMI is widely used as a screening tool for high adiposity, it has recently been criticized for (1) not distinguishing between fat and lean mass and (2) not being a good screening tool for high adiposity.

What this study adds:

BMI is a very good screening tool for high adiposity. Participants with a high BMI were 29 times more likely to have a high fat mass index than those with lower BMIs.

BMI and other weight-for-height indices have long been used as screening tools to identify children and adults who likely have excess adiposity.1 Most children and adolescents with a high BMI for their sex and age2 are likely to have increased adiposity.3–5 Moreover, a high BMI in childhood and adolescence is associated with adverse levels of cardiovascular disease risk factors, the initial stages of atherosclerosis, and a high BMI in adulthood.6,7 

BMI is based only on weight and height, which can be a poor indicator of adiposity among those with normal or relatively low adiposity.8 The limitations of BMI, such as its inability (1) to distinguish between fat and lean mass, and (2) to characterize body fat distribution, have been appreciated by researchers.9 Further, some investigators have concluded that BMI has low sensitivity for detecting high adiposity,10 and there are differences in the relation of BMI to adiposity by racial and ethnic identity.4,5,11,12 For example, at the same BMI, Black children have less adiposity than white children. These limitations have led to attempts to find alternatives more strongly related to adiposity and the risk for various adverse health outcomes.13–15 

Previous studies of percentage of body fat (%fat), based on dual-energy x-ray absorptiometry (DXA), have shown that about three-fourths of children and adolescents with a BMI ≥ Centers for Disease Control and Prevention (CDC) 95th percentile had elevated %fat,16 and almost all had a fat mass index (FMI, fat mass ÷ height2) ≥75th percentile for their sex and age.4 Although these studies, based on data from 1995 to 2008, showed that children with high BMI are likely to have high adiposity, they did not address the agreement between BMI and FMI across the full range of adiposity, nor did they evaluate the relation of BMI to adiposity among children with very high BMIs (eg, ≥97 percentile).

Therefore, the current study of 8- to 19-year-olds uses nationally representative US data from 2011 to 2018 to more fully characterize (1) the ability of BMI to predict DXA-calculated adiposity and lean mass across the entire BMI distribution and (2) the screening characteristics of a high BMI for high adiposity. The current analyses also use the newly developed extended BMI z-scores (BMIz)17 to examine the relation of very high BMIs to adiposity and lean mass.

DXA scans were acquired for 8- to 19-year-olds examined from 2011 to 2012 through 2017 to 2018 in the National Health and Nutrition Survey (NHANES). This nationally representative, cross-sectional survey of the civilian, noninstitutionalized population contains interviews and examinations,18 and is based on a complex, multistage sampling design. NHANES underwent institutional review board approval. Participants 8 to 17 years of age provided written assent, and parents or guardians provided written consent. Participants aged 18 years and over provided written consent.

Weight and height were measured using standardized techniques and equipment, and BMI was calculated as weight (kg) ÷ height (m)2. Although obesity in children is frequently defined as a BMI ≥95th percentile of the CDC growth charts,2,19 we refer to these values as high BMI to avoid confusion with high adiposity determined by DXA.

We acknowledge that race and ethnicity are primarily social constructs rather than genetic or biological categories but provide this information for descriptive purposes. This information was reported separately by participants or proxy respondents (such as parents). We do not examine race and ethnicity differences in the relation of BMI to adiposity because this was the focus of a recent article.12 

Whole-body DXA scans were acquired using a Hologic Discovery A densitometer (Hologic Inc, Bedford, Massachusetts).20 Staff in the Department of Radiology at the University of California, San Francisco, and the University of Hawaii Cancer Center reviewed and analyzed each scan. Hologic Discovery APEX version 4.0 software was used with the NHANES body composition analysis option.21 DXA scans were performed in 2 sets of NHANES cycles: 1999 to 2000 through 2007 to 2008 and from 2011 to 2012 through 2017 to 2018. The current analyses focus on the recent (2011–2018) data.

About 15.6% of the 8202 participants with weight and height data did not have a DXA-calculated value of fat mass. Beyond planned exclusions, such as pregnancy, about 10% of eligible participants did not have a DXA scan for various reasons, such as a lack of time or participant refusal (a code of 6 for the variable “DXAEXSTS”).21 Further, about 5% of the participants with a DXA scan were determined to have invalid data in 1 or more body regions and were excluded from the analyses. The small difference in mean BMIs between the 1279 participants without valid DXA data and the 6923 participants with complete DXA data were not statistically significant. We did not, therefore, use multiple imputations to estimate the body composition values for those with missing data.

Although %fat (fat mass ÷ weight) has been used to quantify adiposity, its interpretation is complicated because an elevated value could be because of a high fat mass or a low lean mass. Therefore, our primary outcome for adiposity was the fat mass index (FMI), calculated as fat mass ÷ height (m)2.22,23 FMI has been frequently used in recent studies3,4,12,24 because it is independent of lean mass and may be more strongly related to obesity-related diseases than %fat.25,26 We also examine the relation of BMI to lean BMI (LMI), calculated as (lean mass − bone mineral content) ÷ height (m)2.

Analyses were performed using R27,28 and accounted for the complex sample design of NHANES. We present descriptive characteristics among boys and girls. The relation of age to BMI, LMI, and adiposity is illustrated using locally weighted scatterplot smoothing. This technique makes no assumptions concerning the shape of the relation.29 We used the function “loessFit” in the limma package30 for these analyses because it also accounts for the sample weights.

We then assessed the multiple R2s of linear regression models predicting FMI, LMI, and %fat from age and BMI in each sex. These predictors were modeled as main effects using natural splines with 3 knots.31 (Adding interaction terms between age and BMI resulted in only minimal R2 increases.) We illustrate the usefulness of BMI in assessing FMI among 16-year-old girls using mean–difference plots.32 

We also examined the classification of high levels of LMI and adiposity by a high BMI. In these analyses, we selected cut points for LMI, FMI, and %fat so that the prevalence of a high value would be very similar (within 1%) to the prevalence of high BMI in each sex and year of age. For example, the prevalence of BMI ≥ CDC 95th percentile among 8-year-old boys was 23%, with 77% having a lower BMI. Therefore, we considered an FMI ≥77th percentile high for this sex and age. These cut points assess whether children and adolescents with the highest BMI levels for their sex and age also have the highest FMI, %fat, and LMI values. We used percentile cut points because an American Medical Association expert committee33 concluded that no single cut point for adiposity distinguishes health from disease risk. We present several screening characteristics, such as positive predictive value, sensitivity, relative risk (RR), and the kappa statistic, an overall measure of chance-corrected agreement.34 

Although other cut points have been used to define high adiposity,35 these alternative cut points could influence estimates of sensitivity or positive predictive value. For example, if the prevalence of high adiposity was 50% greater than that of high BMI, the estimated sensitivity of BMI could not be greater than 67%.

Among participants with a BMI ≥ CDC 95th percentile, we also examined the relation of extended BMIz and percentiles17 to the probability of a high value of FMI, LMI, and %fat. This extended scale was developed to address the limitations of the CDC growth charts in characterizing BMIs above the 97th percentile. Logistic regression models provided the predicted probabilities of high FMI, %fat, and LMI values.

Table 1 shows descriptive characteristics of the 6923 8- to 19-year-olds with valid DXA scans. The mean BMIz was 0.53 (boys) and 0.61 (girls), and the prevalence of high BMI was 20.6% (boys) and 19.5% (girls). Although these sex differences were statistically significant (P < .05), the proportional differences in FMI and %fat were much larger, with girls having mean levels about 30% higher than boys. In contrast, the mean LMI was 10% higher among boys. Sex differences in FMI, %fat, and LMI were statistically significant (P < .001).

TABLE 1

Descriptive Characteristic, by Sex

Boys (n = 3584)Girls (n = 3339)
Age (y) 13.4 ± 0.1b 13.4 ± 0.1 
BMI (kg/m222.1 ± 0.1 22.6 ± 0.2 
BMI-for-age z-score (SDs) 0.53 ± 0.03 0.61 ± 0.3 
BMI ≥95th CDC percentile (%) 20.6 ± 1.0 19.5 ± 1.0 
FMI (kg/m26.1 ± 0.1 8.0 ± 0.1 
%fat 26.3 ± 0.2 33.7 ± 0.2 
LMI (kg/m215.5 ± 0.1 14.1 ± 0.1 
Race and ethnicity (%) 
 Non-Hispanic white 52.9% 54.3% 
 Non-Hispanic Black 13.7% 12.8% 
 Hispanica 23.7% 23.1% 
 Non-Hispanic Asian American 4.4% 4.6% 
 Other 5.3% 5.2% 
Boys (n = 3584)Girls (n = 3339)
Age (y) 13.4 ± 0.1b 13.4 ± 0.1 
BMI (kg/m222.1 ± 0.1 22.6 ± 0.2 
BMI-for-age z-score (SDs) 0.53 ± 0.03 0.61 ± 0.3 
BMI ≥95th CDC percentile (%) 20.6 ± 1.0 19.5 ± 1.0 
FMI (kg/m26.1 ± 0.1 8.0 ± 0.1 
%fat 26.3 ± 0.2 33.7 ± 0.2 
LMI (kg/m215.5 ± 0.1 14.1 ± 0.1 
Race and ethnicity (%) 
 Non-Hispanic white 52.9% 54.3% 
 Non-Hispanic Black 13.7% 12.8% 
 Hispanica 23.7% 23.1% 
 Non-Hispanic Asian American 4.4% 4.6% 
 Other 5.3% 5.2% 

Except for age, BMI ≥95th percentile, and race–ethnicity, all sex differences were statistically significant at the .05 level.

a

The Hispanic category includes Mexican Americans and “other Hispanics.” The other category includes multiracial children.

b

Values are mean ± SE for continuous variables or percentages for categorical variables.

Figure 1 shows the smoothed relation of age to BMI, FMI, %fat, and LMI. BMI increased fairly linearly with age in both sexes, whereas the relation of age to the other metrics differed markedly by sex. Levels of FMI showed a more linear increase with age among girls than boys, whereas the opposite was true for LMI. Levels of %fat generally increased after age 12 years among girls but decreased among boys, reaching a minimum at about 17 years.

FIGURE 1

Relation of BMI, FMI, %fat, and LMI to age. Note that the y axes differ across panels. Lines were smoothed with locally weighted scatterplot smoother, using a span of 0.3, and accounted for the sample weights (see Methods).

FIGURE 1

Relation of BMI, FMI, %fat, and LMI to age. Note that the y axes differ across panels. Lines were smoothed with locally weighted scatterplot smoother, using a span of 0.3, and accounted for the sample weights (see Methods).

Close modal

Table 2 shows the multiple R2 values for models within each sex, predicting FMI, %fat, and LMI from age and BMI. Age accounted for little (1%–11%) of the variability in either FMI or %fat, but accounted for 47% (boys) and 26% (girls) of the variability in LMI. The addition of BMI to these models substantially increased their explained variance, with multiple R2s of 0.90 to 0.94 for LMI and FMI. Age and BMI accounted for less variability in %fat, with R2s of 0.68 and 0.71.

TABLE 2

Multiple R2 Values of Regression Models Predicting Lean Mass Index, Fat Mass Index, and %Fat

PredictorsFMI%FatLMI
Boys Age 0.01 0.11 0.47 
 Age, BMI 0.90 0.71 0.90 
     
Girls Age 0.10 0.03 0.26 
 Age, BMI 0.94 0.68 0.90 
PredictorsFMI%FatLMI
Boys Age 0.01 0.11 0.47 
 Age, BMI 0.90 0.71 0.90 
     
Girls Age 0.10 0.03 0.26 
 Age, BMI 0.94 0.68 0.90 

Based on cross-sectional regression analyses that allowed for curvilinearity in the relation of both age and BMI to the 3 outcomes.

We illustrate the improvement in explained variance for FMI when both age and BMI are accounted for in Fig 2, which focuses on 16-year-old girls (n = 300). The left panel shows the difference between DXA-measured FMI and the mean FMI in this age group for each individual. Differences ranged from ∼−5 kg to 19 kg. Of note, 1 girl (orange point) had an FMI of 28.3, which was 19.3 kg/m2 higher than the overall mean. (This girl’s BMI was 55.3.) The right panel illustrates the differences (y axis) between DXA-measured FMI and BMI-predicted FMI. The reduced spread around the y = 0 line, with differences ranging from −2 to +2 kg/m2, demonstrates the ability of BMI to estimate adiposity. The +19.3 kg/m2 difference (orange circle) was reduced to −0.2 kg/m2 by accounting for the child’s BMI.

FIGURE 2

Prediction of FMI by BMI among 300 16-year-old girls. The left panel shows differences in FMI from the mean FMI for this sex/age. The right panel is a Bland-Altman plot showing the differences between DXA-calculated FMI and BMI-predicted FMI. The orange circle represents a girl with an FMI of 28.3 kg/m2 and a BMI of 55.3.

FIGURE 2

Prediction of FMI by BMI among 300 16-year-old girls. The left panel shows differences in FMI from the mean FMI for this sex/age. The right panel is a Bland-Altman plot showing the differences between DXA-calculated FMI and BMI-predicted FMI. The orange circle represents a girl with an FMI of 28.3 kg/m2 and a BMI of 55.3.

Close modal

We then examined the ability of a BMI ≥ CDC 95th percentile to accurately characterize high levels of FMI, %fat, and FMI. We defined high levels of these 3 variables so that the number of participants with a high level of LMI, FMI, and %fat would be very similar to those with a high (≥95th percentile) BMI within each sex and year of age. This classification assesses whether the participants with the highest BMIs for their sex and age also have the highest FMI, %fat, and LMI levels. The resulting cross-classifications are shown in the upper part of Table 3, with the sample weights rescaled to sum to 6923.

TABLE 3

Screening Characteristics of a High BMI for High Levels of Lean Mass Index, Fat Mass Index, and %Fat

High FMIHigh %FatHigh LMI
High BMIYesNoTotalYesNoTotalYesNoTotal
 Yes 1217b 169 1385 1039 347 1385 1092 294 1385 
No 165 5372 5538 344 5194 5538 298 5240 5538 
Total 1382 5541  1383 5541  1390 5533  
Screening characteristicsa 
 Positive predictive value 0.88 (0.86–0.90) 0.75 (0.72–0.78) 0.79 (0.75–0.82) 
 Sensitivity 0.88 (0.86–0.09) 0.75 (0.72–0.78) 0.79 (0.75–0.82) 
 Specificity 0.97 (0.96–0.97) 0.94 (0.93–0.95) 0.95 (0.94–0.96) 
 RR 29 (24–36) 12 (11–14) 15 (13–17) 
 Kappa statistic 0.85 (0.83–0.86) 0.69 (0.66–0.71) 0.73 (0.71–0.76) 
High FMIHigh %FatHigh LMI
High BMIYesNoTotalYesNoTotalYesNoTotal
 Yes 1217b 169 1385 1039 347 1385 1092 294 1385 
No 165 5372 5538 344 5194 5538 298 5240 5538 
Total 1382 5541  1383 5541  1390 5533  
Screening characteristicsa 
 Positive predictive value 0.88 (0.86–0.90) 0.75 (0.72–0.78) 0.79 (0.75–0.82) 
 Sensitivity 0.88 (0.86–0.09) 0.75 (0.72–0.78) 0.79 (0.75–0.82) 
 Specificity 0.97 (0.96–0.97) 0.94 (0.93–0.95) 0.95 (0.94–0.96) 
 RR 29 (24–36) 12 (11–14) 15 (13–17) 
 Kappa statistic 0.85 (0.83–0.86) 0.69 (0.66–0.71) 0.73 (0.71–0.76) 

High levels of LMI, FMI, and %fat were categorized so that the weighted prevalence of a high level would be very similar to that of a high BMI within each sex and year of age. For example, there were 1385 participants with a high BMI and 1382 with a high FMI (left 2 × 2 table).

a

Values are estimates and 95% confidence intervals.

b

Values are weighted Ns.

As indicated by the kappa statistic, RR, and other screening characteristics (lower part of Table 3), a high BMI was a very good predictor of a high FMI and a good predictor of high levels of LMI and %fat. For high FMI, the kappa statistic was 0.85, and both the positive predictive value (probability of a child with a high BMI having a high FMI) and the sensitivity (probability that a child with a high FMI had a high BMI) were 88%. Participants with a high BMI were 29 times more likely to have a high FMI than those with lower BMIs. A high BMI was less predictive of high levels of %fat and LMI, with RRs of 12 to 15 and kappa statistics of about 0.70.

Figure 3 shows the relation of BMIz to the probability of high sex- and age-specific levels of FMI, %fat, and LMI fat among participants with a BMI ≥ CDC 95th percentile. For those with a BMI at the 95th percentile, the probability of a high FMI or %fat was about 60%. This probability increased rapidly with increasing BMIz, with the likelihood of a high FMI approaching 100% at about the 98th percentile (boys) or 97th percentile (girls). In contrast, the probability of high %fat did not approach 100% until BMI was ≥99.5th percentile. The proportion of participants with a high LMI also increased with increasing BMIz, so almost all participants with a very high BMI had high levels of both FMI and LMI. The only statistically significant sex interaction was for FMI. As illustrated by the difference in slopes, the probability of a high FMI increased more rapidly with BMIz among girls than boys (P < .01 for difference).

FIGURE 3

Probability of a high value of FMI, LMI, and %fat by extended BMI-for-age z-score among participants with a high BMI. Logistic regression was used to model the relation of the 3 metrics to BMIz within each sex and to obtain the predicted probabilities. BMI percentiles corresponding to the extended z-scores are shown at the bottom of the figure.

FIGURE 3

Probability of a high value of FMI, LMI, and %fat by extended BMI-for-age z-score among participants with a high BMI. Logistic regression was used to model the relation of the 3 metrics to BMIz within each sex and to obtain the predicted probabilities. BMI percentiles corresponding to the extended z-scores are shown at the bottom of the figure.

Close modal

Additional analyses indicated among participants with a BMI between the 89.5th and 90.5th percentiles, only 10% of boys and 1% of girls had a high FMI. In contrast, about 20% of these participants had a high LMI.

Our results indicate that, within each sex, the combination of BMI and age accounted for 90% (boys) to 94% (girls) of the variability in FMI. A BMI ≥95th CDC percentile was also a very good screening tool for high FMI, with a positive predictive value and sensitivity of 88% and a kappa statistic of 0.85. Although the interpretation of kappa is somewhat arbitrary, it has been suggested that values above 0.75 or 0.80 indicate excellent or almost perfect agreement.36 Further, participants with a BMI ≥ CDC 95th percentile were 29 times more likely to have a high FMI than participants with lower BMIs. Although BMI was less strongly related to %fat, the multiple R2s were high (about 0.70), and 75% of participants with a high BMI had a high %fat. As illustrated in Fig 3, as BMI-for-age increased above the 95th percentile, the probabilities of high levels of FMI and %fat also increased.

Because BMI cannot distinguish between fat and lean mass, it is not a perfect indicator of adiposity. We found, for example, that BMI and age explained almost as much of the variability in LMI (90%) as in FMI. Further, a participant with a BMI ≥ CDC 95th percentile was 15 times more likely to have a high LMI than those with lower BMIs. Children and adolescents with obesity have higher levels of both fat and lean mass than those with a healthy weight,37 and fat mass and lean mass change in the same direction with weight change.38 However, because a high BMI is more strongly related to high levels of FMI than LMI, it is a good screening tool for high adiposity.

Although our analyses included both FMI and %fat, there are several reasons why FMI may be a better adiposity metric.22 %Fat is relatively well preserved under semistarvation conditions, so it is not a good indicator of nutritional status or energy reserves.23 Further, a high %fat could be the result of a high value of fat mass or a low lean mass, making it difficult to interpret this proportion. It has also been suggested that it is more appropriate to adjust fat mass for height than weight when characterizing a child’s adiposity.22 We found that BMI showed a slightly weaker association with %fat than FMI, possibly because some individuals with a normal %fat had high levels of both fat and lean mass.4 Additional analyses in the current study confirmed this possibility: 94% of the participants with high FMI and normal %fat had a high LMI (data not shown). Because almost all children with very high BMI levels have high levels of both FMI and LMI (Fig 3), %fat is unlikely to be the best metric for assessing increased adiposity, because its denominator includes both fat mass and lean mass.

The current analyses did not include race and ethnicity comparisons. Previous studies have found that, as compared with levels of %fat among white individuals, Black individuals and adults typically have lower levels of adiposity.4,11,12,39,40 However, a recent analysis of the 8- to 19-year-olds in the current study suggests that race–ethnicity differences are most evident at BMIs below the 95th percentile.12 The underlying mechanisms for these race and ethnicity differences in adiposity are unknown, but environmental, social, behavioral, and nutritional factors may contribute to these group differences.40 

There are several limitations in the current study. About 15.6% of participants with weight and height data did not have valid DXA information. Further, there can be substantial differences in DXA estimates of adiposity compared with more complex body composition methods.41–45 For example, compared with various criterion methods, the Hologic QDR 4500A fan-beam densitometer overestimated lean body mass among adults by 5%.45,46 Therefore, the National Center for Health Statistics reduced the lean mass estimate by 5% and added an equivalent kilogram weight to fat mass.46 Although these differences would influence comparisons of levels across studies, they likely had little impact on our analyses concerning the relation of BMI to FMI, %fat, and LMI. The distribution of BMI levels also would influence the screening ability of a high BMI. In the current study, the median BMI percentile among participants with a high BMI was about 97, but if it had been closer to the 95th or 99th percentiles, the observed screening characteristic would differ.

Further, because BMI cannot characterize body fat distribution, several alternatives have been proposed, and many focus on waist circumference (WC).47,48 However, (1) it may be challenging to standardize WC measurement across health care providers, (2) it is difficult to measure WC among individuals with high BMIs, and (3) the optimal cut points for a high waist are uncertain.48 The neck circumference may be a more convenient measure of fat distribution,49 but there is little longitudinal data on its relation to adverse outcomes.

BMI is strongly related to high levels of adiposity and LMI among 8- to 19-year-olds, with multiple R2s of 0.90 to 0.94 for FMI and LMI. Although the prediction was not perfect, a BMI ≥ CDC 95th percentile was a very good indicator of a high FMI and a good indicator of high %fat. Participants with a high BMI were 29 times more likely to have a high FMI than those with lower BMIs. Our findings provide further evidence of the utility of BMI in research and clinical care.

Dr Freedman conceptualized and designed the study, performed the analyses, drafted the initial manuscript, and revised the manuscript; Drs Dietz, Zemel, and Daymont helped interpret the results and critically reviewed the manuscript for important intellectual content; and all authors approved the final manuscript as submitted and agreed to be accountable for all aspects of the work.

COMPANION PAPER: A companion to this article can be found online at www.pediatrics.org/cgi/doi/10.1542/peds.2024-066370.

FUNDING: Supported by grants R01 HD100406 and UL1 RR-026314.

CONFLICT OF INTEREST DISCLOSURES: The authors have indicated they have no conflicts of interest relevant to this article to disclose.

%fat

percentage of body fat

BMIz

BMI z-score

CDC

Centers for Disease Control and Prevention

DXA

dual-energy x-ray absorptiometry

FMI

fat mass index

LMI

lean mass index

NHANES

National Health and Nutrition Survey

RR

relative risk

WC

waist circumference

1
Nuttall
FQ
.
Body mass index: obesity, BMI, and health: a critical review
.
Nutr Today
.
2015
;
50
(
3
):
117
128
2
Kuczmarski
RJ
,
Ogden
CL
,
Guo
SS
, et al
.
2000 CDC growth charts for the United States: methods and development
.
Vital Health Stat 11
.
2002
;
11
(
246
):
1
190
3
Freedman
DS
,
Ogden
CL
,
Berenson
GS
,
Horlick
M
.
Body mass index and body fatness in childhood
.
Curr Opin Clin Nutr Metab Care
.
2005
;
8
(
6
):
618
623
4
Weber
DR
,
Moore
RH
,
Leonard
MB
,
Zemel
BS
.
Fat and lean BMI reference curves in children and adolescents and their utility in identifying excess adiposity compared with BMI and percentage body fat
.
Am J Clin Nutr
.
2013
;
98
(
1
):
49
56
5
Flegal
KM
,
Ogden
CL
,
Yanovski
JA
, et al
.
High adiposity and high body mass index-for-age in US children and adolescents overall and by race–ethnic group
.
Am J Clin Nutr
.
2010
;
91
(
4
):
1020
1026
6
Berenson
GS
,
Srinivasan
SR
,
Bao
W
,
Newman
WP
III
,
Tracy
RE
,
Wattigney
WA
.
Association between multiple cardiovascular risk factors and atherosclerosis in children and young adults. The Bogalusa Heart Study
.
N Engl J Med
.
1998
;
338
(
23
):
1650
1656
7
Owen
CG
,
Whincup
PH
,
Orfei
L
, et al
.
Is body mass index before middle age related to coronary heart disease risk in later life? Evidence from observational studies
.
Int J Obes
.
2009
;
33
(
8
):
866
877
8
Bray
GA
,
DeLany
JP
,
Volaufova
J
,
Harsha
DW
,
Champagne
C
.
Prediction of body fat in 12-year-old African American and white children: evaluation of methods
.
Am J Clin Nutr
.
2002
;
76
(
5
):
980
990
9
Prentice
AM
,
Jebb
SA
.
Beyond body mass index
.
Obes Rev
.
2001
;
2
(
3
):
141
147
10
Hampl
SE
,
Hassink
SG
,
Skinner
AC
, et al
.
Clinical practice guideline for the evaluation and treatment of children and adolescents with obesity
.
Pediatrics
.
2023
;
151
(
2
):
e2022060640
11
Freedman
DS
,
Wang
J
,
Thornton
JC
, et al
.
Racial/ethnic differences in body fatness among children and adolescents
.
Obesity (Silver Spring)
.
2008
;
16
(
5
):
1105
1111
12
Martin
CB
,
Stierman
B
,
Yanovski
JA
,
Hales
CM
,
Sarafrazi
N
,
Ogden
CL
.
Body fat differences among US youth aged 8–19 by race and Hispanic origin, 2011–2018
.
Pediatr Obes
.
2022
;
17
(
7
):
e12898
13
Austin
SB
,
Richmond
TK
.
It’s time to retire BMI as a clinical metric
. Available at: https://www.medpagetoday.com/opinion/second-opinions/101296. Accessed November 8, 2023
14
Rubino
F
,
Batterham
RL
,
Koch
M
, et al
.
Lancet Diabetes & Endocrinology Commission on the definition and diagnosis of clinical obesity
.
Lancet Diabetes Endocrinol
.
2023
;
11
(
4
):
226
228
15
American Medical Association
.
AMA: use of BMI alone is an imperfect clinical measure
. Available at: https://www.ama-assn.org/delivering-care/public-health/ama-use-bmi-alone-imperfect-clinical-measure. Accessed January 24, 2024
16
Freedman
DS
,
Wang
J
,
Thornton
JC
, et al
.
Classification of body fatness by body mass index-for-age categories among children
.
Arch Pediatr Adolesc Med
.
2009
;
163
(
9
):
805
811
17
Hales
CM
,
Freedman
DS
,
Akinbami
L
,
Wei
R
,
Ogden
CL
.
Evaluation of alternative body mass index (BMI) metrics to monitor weight status in children and adolescents with extremely high BMI using CDC BMI-for-age growth charts
.
Vital Health Stat 1
.
2022
;(
197
):
1
42
18
CDC National Center for Health Statistics
.
NHANES questionnaires, data sets, and related documentation
. Available at: https://wwwn.cdc.gov/nchs/nhanes/Default.aspx. Accessed June 23, 2020
19
Ogden
CL
,
Flegal
KM
.
Changes in terminology for childhood overweight and obesity
.
Natl Health Stat Rep
.
2010
;
25
(
25
):
1
5
20
National Health and Nutrition Examination Survey
.
Body composition procedures manual
. Available at: https://wwwn.cdc.gov/nchs/data/nhanes/2017-2018/manuals/Body_Composition_Procedures_Manual_2018.pdf. Accessed October 9, 2023
21
National Health and Nutrition Examination Survey
.
2017–2018 data documentation, codebook, and frequencies
. Available at: https://wwwn.cdc.gov/Nchs/Nhanes/2017-2018/DXX_J.htm. Accessed October 9, 2023
22
Wells
JCK
,
Coward
WA
,
Cole
TJ
,
Davies
PSW
.
The contribution of fat and fat-free tissue to body mass index in contemporary children and the reference child
.
Int J Obes Relat Metab Disord
.
2002
;
26
(
10
):
1323
1328
23
VanItallie
TB
,
Yang
MU
,
Heymsfield
SB
,
Funk
RC
,
Boileau
RA
.
Height-normalized indices of the body’s fat-free mass and fat mass: potentially useful indicators of nutritional status
.
Am J Clin Nutr
.
1990
;
52
(
6
):
953
959
24
Shepherd
JA
,
Ng
BK
,
Sommer
MJ
,
Heymsfield
SB
.
Body composition by DXA
.
Bone
.
2017
;
104
:
101
105
25
Liu
P
,
Ma
F
,
Lou
H
,
Liu
Y
.
The utility of fat mass index versus body mass index and percentage of body fat in the screening of metabolic syndrome
.
BMC Public Health
.
2013
;
13
(
1
):
629
26
Zhang
S
,
Wang
L
,
Yu
M
,
Guan
W
,
Yuan
J
.
Fat mass index as a screening tool for the assessment of non-alcoholic fatty liver disease
.
Sci Rep
.
2022
;
12
(
1
):
20219
27
The R Foundation
.
The R Project for statistical computing
. Available at: https://www.r-project.org/. Accessed October 8, 2023
28
Lumley
T
.
Analysis of Complex Survey Samples
.
John Wiley & Sons
;
2010
29
Cleveland
WS
.
LOWESS: a program for smoothing scatterplots by robust locally weighted regression
.
Am Stat
.
1981
;
35
(
1
)
30
Smyth
G
,
Hu
Y
,
Ritchie
M
, et al
. limma: linear models for microarray data.
2023
. Available at: https://www.bioconductor.org/packages/limma/. Accessed May 5, 2024
31
Harrell
FE
.
Regression Modeling Strategies: With Applications to Linear Models, Logistic Regression, and Survival Analysis
. Springer-Verlag, NY;
2010
32
Bland
JM
,
Altman
DG
.
Statistical methods for assessing agreement between 2 methods of clinical measurement
.
Lancet
.
1986
;
1
(
8476
):
307
310
33
Barlow
SE
.
Expert Committee
.
Expert committee recommendations regarding the prevention, assessment, and treatment of child and adolescent overweight and obesity: summary report
.
Pediatrics
.
2007
;
120
(
Suppl 4
):
S164
S192
34
Bloch
DA
,
Kraemer
HC
.
2 x 2 kappa coefficients: measures of agreement or association
.
Biometrics
.
1989
;
45
(
1
):
269
287
35
Javed
A
,
Jumean
M
,
Murad
MH
, et al
.
Diagnostic performance of body mass index to identify obesity as defined by body adiposity in children and adolescents: a systematic review and meta-analysis
.
Pediatr Obes
.
2015
;
10
(
3
):
234
244
36
Wikipedia
.
Cohen’s kappa
. Available at: https://en.wikipedia.org/w/index.php?title=Cohen%27s_kappa&oldid=1197848362. Accessed January 24, 2024
37
Wells
JC
,
Fewtrell
MS
,
Williams
JE
,
Haroun
D
,
Lawson
MS
,
Cole
TJ
.
Body composition in normal weight, overweight and obese children: matched case-control analyses of total and regional tissue masses, and body composition trends in relation to relative weight
.
Int J Obes
.
2006
;
30
(
10
):
1506
1513
38
Forbes
GB
. The companionship of fat and lean mass. In:
Human Body Composition. In Vivo Methods, Models, and Assessment, Volume 60
.
Springer
;
1993
:
1
14
39
Wagner
DR
,
Heyward
VH
.
Measures of body composition in Blacks and whites: a comparative review
.
Am J Clin Nutr
.
2000
;
71
(
6
):
1392
1402
40
Zemel
BS
,
Shepherd
JA
,
Grant
SFA
, et al
.
Reference ranges for body composition indices by dual energy X-ray absorptiometry from the Bone Mineral Density in Childhood Study Cohort
.
Am J Clin Nutr
.
2023
;
118
(
4
):
792
803
41
Van Der Ploeg
GE
,
Withers
RT
,
Laforgia
J
.
Percentage body fat via DEXA: comparison with a 4-compartment model
.
J Appl Physiol (1985)
.
2003
;
94
(
2
):
499
506
42
Withers
RT
,
Laforgia
J
,
Heymsfield
SB
.
Critical appraisal of the estimation of body composition via 2-, 3-, and 4-compartment models
.
Am J Hum Biol
.
1999
;
11
(
2
):
175
185
43
Wells
JC
,
Fuller
NJ
,
Dewit
O
,
Fewtrell
MS
,
Elia
M
,
Cole
TJ
.
Four-component model of body composition in children: density and hydration of fat-free mass and comparison with simpler models
.
Am J Clin Nutr
.
1999
;
69
(
5
):
904
912
44
Wong
WW
,
Hergenroeder
AC
,
Stuff
JE
,
Butte
NF
,
Smith
EOB
,
Ellis
KJ
.
Evaluating body fat in girls and female adolescents: advantages and disadvantages of dual-energy X-ray absorptiometry
.
Am J Clin Nutr
.
2002
;
76
(
2
):
384
389
45
Schoeller
DA
,
Tylavsky
FA
,
Baer
DJ
, 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
(
5
):
1018
1025
46
National Health and Nutrition Examination Survey
.
2011–2012 data documentation, codebook, and frequencies
. Available at: https://wwwn.cdc.gov/Nchs/Nhanes/2011-2012/DXX_G.htm. Accessed January 26, 2021
47
Stevens
J
,
McClain
JE
,
Truesdale
KP
.
Selection of measures in epidemiologic studies of the consequences of obesity
.
Int J Obes
.
2008
;
32
(
Suppl 3
):
S60
S66
48
Klein
S
,
Allison
DB
,
Heymsfield
SB
, et al
.
Waist circumference and cardiometabolic risk: a consensus statement from Shaping America’s Health: Association for Weight Management and Obesity Prevention; NAASO, the Obesity Society; the American Society for Nutrition; and the American Diabetes Association
.
Obesity (Silver Spring)
.
2007
;
15
(
5
):
1061
1067
49
Joshipura
K
,
Muñoz-Torres
F
,
Vergara
J
,
Palacios
C
,
Pérez
CM
.
Neck circumference may be a better alternative to standard anthropometric measures
.
J Diabetes Res
.
2016
;
2016
:
6058916