Abstract
Objective
Anthropometrics are commonly used indices of total and central adiposity. No study has compared anthropometric measurements to dual-energy X-ray absorptiometry (DXA) measurements as correlates of cardiovascular risks in a nationally representative sample of youth. We aimed to evaluate the validity of anthropometrics compared to DXA-assessed adiposity in relation to cardiovascular risks in youth aged 8-19 years.
Methods
Data were from the National Health and Nutrition Examination Survey 1999-2004 (n=7013). We examined the correlations between anthropometric and DXA measures of adiposity (i.e., body mass index (BMI) versus percent fat mass (%FM) and fat mass index, and waist circumference (WC) and waist-to-height ratio (WHtR) versus percent trunk fat mass (%TFM)) with nine cardiovascular risks, stratified by sex and age, or race-ethnicity.
Results
Anthropometric and DXA adiposity measures were significantly correlated with insulin (r: 0.48 to 0.66), C-reactive protein (r: 0.47 to 0.58), triglycerides (r: 0.15 to 0.41), high-density lipoprotein cholesterol (HDL-C, r: −0.44 to −0.22), systolic blood pressure (SBP, r: 0.10 to 0.31), low-density lipoprotein cholesterol (r: 0.09 to 0.30), total cholesterol (TC, r: 0.01 to 0.29) and glucose (r: 0.05 to 0.20). Only in all youth, BMI was more strongly correlated with SBP (0.22 vs. 0.12, P<0.0001) and HDL-C (−0.34 vs. −0.25, P<0.0001) than %FM; WC but not WHtR was more strongly correlated with HDL-C (−0.37 vs. −0.30, P<0.0001) but less strongly associated with TC (0.12 vs. 0.21, P<0.0001) than %TFM.
Conclusions
DXA adiposity measures do not produce stronger associations with cardiovascular risk factors in youth than BMI or WC.
Keywords: body mass index, waist circumference, waist-to-height ratio, trunk fat
Introduction
As the prevalence of childhood obesity has increased [1], concerns about the impact of excess adiposity on the early appearance of cardiovascular risk factors have deepened. Anthropometric indices such as body mass index (BMI) and waist circumference (WC) provide simple and inexpensive indicators of body fat and body fat distribution for these studies, however, these indices cannot distinguish fat mass (FM) from fat-free mass (FFM). Previous studies in youth have documented that BMI and DXA-assessed total body fat are strongly correlated [2-8] and that both indices are associated with cardiovascular risk factors [5-7, 9-11]. It remains unclear whether anthropometric indices are equivalent predictors of cardiovascular risk when compared with measures of adiposity by DXA.
Recently, Sun et al. [12] presented a study comparing the validity of DXA measurements and several anthropometric indices with respect to their correlations with obesity-related biologic factors using data from adults in the NHANES study. Previous studies had examined samples that differed in demographic and regional characteristics that might impact results [2-11, 13]. The work presented here extends the work done by Sun et al. in adults to children. The NHANES data supports a more comprehensive and generalizable analysis than has been previously presented in youth [14]. The aim of this study was to evaluate the validity of anthropometric indices of total and regional body fat, including BMI, WC, and waist-to-height ratio (WHtR), and their relation to cardiovascular risk factors, compared with DXA-assessed total and regional body fat, in a representative sample of American children and adolescents aged 8-19 years. We hypothesize that DXA-assessed percent fat mass (%FM), fat mass index (FMI) and percent trunk fat mass (%TFM) are not superior to BMI and WC in relation to cardiovascular risk in this population. We fill gaps in previous work by exploring the usefulness of FMI in children and examine whether WHtR serves as a better indicator of cardiovascular risk than WC.
Subjects and methods
Data were from the NHANES 1999-2004 [15]. The NHANES is a stratified, multistage probability sample that represents the US civilian non-institutionalized population. Non-Hispanic Blacks, Mexican Americans, low-income Whites (beginning in 2000), adolescents aged 12-19 years were oversampled to provide more reliable estimates for those groups. The protocol of the survey was approved by the National Center for Health Statistics Institutional Review Board. Parental consent and child assent were obtained for children aged 8 – 17 years. Written consent was obtained from youth aged 18 - 19 years.
Anthropometric measurements
Height, weight and WC were measured by a trained technician in a mobile examination center following standard procedures [16]. Standing height without shoes was measured with a stadiometer to the nearest 1 millimeter, and weight was measured in an examination gown and without shoes to the nearest 0.1 kilogram using a Toledo self-zeroing digital scale. WC was assessed with a measuring tape at the uppermost lateral border of the hip crest (ilium) to the nearest 0.1 cm. BMI was calculated as weight in kilograms divided by height in meters squared. The sex-specific percentile on the CDC’s 2000 BMI-for-age growth charts were used to define overweight (≥85th and < 95th) and obesity (≥ 95th) [17]. WHtR was calculated as WC in centimeters divided by height in centimeters.
DXA measurements
Total body fat mass and total body mass, and trunk fat mass and total trunk mass were determined by whole body DXA scans using a Hologic QDR 4500A fan-beam densitometer (Hologic, Inc., Bedford, Massachusetts) following the manufacturer’s acquisition procedures in the fast mode. Hologic DOS software (version 8.26:a3*; Hologic) and Hologic Discovery software (version 12.1; Hologic) were used to administer and analyze the scans, respectively. DXA scans were administered to eligible participants 8 years of age and older in the mobile examination centers.
Participants were excluded from the DXA examination if they were pregnant, reported taking tests with radiographic contrast material or participating in nuclear medicine studies in the past 72 hours, or their self-reported weight or height exceeded the DXA table limit (300 pounds or 6′5″). To resolve the problem of potential biases due to missing DXA data, five imputation datasets [18-20] were created by the National Center for Health Statistics. DXA data were not available for girls aged 8-17 years in 1999-2000 due to Institutional Review Board issues.
%FM was calculated as total body fat mass divided by total body mass times 100, and %TFM was calculated as trunk fat mass divided by total trunk mass times 100. FMI was calculated as total body fat mass in kilograms divided by the square of height in meters.
Assessment of cardiovascular risk factors
We examined nine well-established cardiovascular risk factors including systolic blood pressure (SBP), diastolic blood pressure (DBP), total cholesterol (TC), high density lipoprotein cholesterol (HDL-C), low density lipoprotein cholesterol (LDL-C), triglyceride (TG), glucose, insulin, and C-reactive protein (CRP).
Blood pressure was measured following the recommendations of the American Heart Association Human Blood Pressure Determination by sphygmo-manometers [21] using a mercury sphygmomanometer in a mobile examination center. Three blood pressure measurements were taken in 77.6% of participants aged 8-19 years. The average of the second and third measurements was used.
Blood samples were collected in the morning or afternoon examination sessions. TC, HDL-C and CRP were assessed in the morning or the afternoon examination sessions for participants aged 8-19 years, whereas TG, LDL-C, glucose and insulin were assessed only in the morning session among participants over 12-19 years who reported they had fasted at least 8.5 hours but less than 24 hours before the examination. TC and TG were measured enzymatically. Most HDL-C samples were measured by heparin manganese precipitation method and others were by direct HDL-C immunoassay method in 1999-2002, while in 2003-2004, all HDL-C samples were analyzed using direct HDL-C immunoassay method. The bias for the HDL-C method for 2003-2004 was acceptable (<4%) and the results were not corrected [22]. Friedewald’s equation [23] was used to calculate LDL-C levels for participants with TG ≤400 mg/dL. CRP was assessed by latex-enhanced nephelometry [24]. Glucose was assessed using enzyme hexokinase and insulin was assessed using two-site immunoenzymometric assay [25]. There were changes to the equipment and laboratory method in the assessment of insulin in 2003-2004 [26].
Analysis sample
The NHANES 1999-2004 sample consisted of 7862 participants aged 8-19 years who were interviewed and eligible for DXA assessments. Youth who did not have a fat mass estimate by DXA calculated or imputed (n = 215); were missing weight, height or WC (n = 137); were missing all cardiovascular risk factors (n = 211); or were using medication in the past month that may influence blood pressure, blood lipids, glucose or body weight (n = 286) were excluded from analyses. A total of 7013 participants were included. Of these participants, DXA data were imputed for 606 (8.6%) participants. Among the 7013 participants, SBP and DBP were available for 5597 and TC, HDL-C and CRP were available for 6549 children aged 8-19 years. Fasting TG, LDL-C, glucose and insulin were not measured in children less than 12 years of age, and data were available for 2227 adolescents aged 12-19 years.
Statistical analysis
All analyses were conducted using SAS (version 9.2; SAS institute, Cary, NC, USA). Sample characteristics were described using mean and standard error after adjusting for the complex sampling design. We examined the partial Pearson correlations between BMI and %FM and FMI. WC and WHtR are both proxies for central fat, therefore, we examined correlations between WC, WHtR and %TFM. In the partial correlation analyses of adiposity indices with cardiovascular risk factors, all variables except SBP and DBP were log-transformed to account for the skewed distribution. All correlation analyses were stratified by gender and age (8-11 years and 12-19 years), race-ethnicity, or in all youth after adjusting for sampling weight, age in years, gender and race-ethnicity whenever applicable. Analysis stratified by race-ethnicity was limited to Mexican Americans, non-Hispanic White and Black because of small sample size in other race-ethnic groups. The assessment methods for insulin were modified in 2003, and therefore survey year (1999-2002 or 2003-2004) was also adjusted in analyses of insulin. We initially adjusted associations with risk factors for physical activity and sedentary behavior (i.e., sitting and watching TV/videos, using computer), however, these adjustments did not appreciably alter our estimates. Therefore we did not include those variables in final models.
For each cardiovascular risk factor, comparisons of partial correlation coefficients were made between anthropometric and DXA measures of body fat (i.e., BMI compared to %FM, BMI compared to FMI, WC compared to %TFM, and WHtR compared to %TFM) within gender- and age-specific groups, within race/ethnicity-specific groups and within all youth. Comparisons were made by Wolfe’s method for comparing dependent correlation coefficients estimated in the same sample [27] after taking variance introduced by multiple imputations of DXA data into account following the methods used by Sun et al. [12] Because multiple correlation coefficients were compared, we show Bonferroni-corrected P values (P < 0.000195, corresponding to 0.05 divided by 256 comparisons) and uncorrected P < 0.05 significance levels.
Results
In our study population the average age was 13.6 years in boys and 13.3 years in girls (Table 1). The sample included Mexican Americans, non-Hispanic Whites, non-Hispanic Blacks and a combined category of other race-ethnicities. Around one-third of youth were overweight or obese. The mean BMI (22.0 kg/m2) was similar in boys and girls but boys had a lower mean %FM compared to girls (25.4% vs. 32.9%). Mean blood pressure, lipids, glucose and insulin levels did not differ substantially by gender.
Table 1.
Boys | Girls | |||
---|---|---|---|---|
nb | Mean (SE)c | nb | Mean (SE)c | |
Age (years) | 4209 | 13.6 (0.1) | 2804 | 13.3 (0.1) |
Gender (%) | 4209 | 52.4 (0.8) | 2804 | 47.6 (0.8) |
Race-ethnicity (%) | ||||
Mexican Americans | 1473 | 11.6 (1.3) | 952 | 12.1 (1.6) |
Non-Hispanic Whites | 1044 | 59.9 (2.2) | 735 | 60.4 (2.7) |
Non-Hispanic Blacks | 1373 | 15.1 (1.5) | 897 | 15.1 (1.6) |
Others | 319 | 13.4 (1.6) | 220 | 12.4 (1.7) |
Overweight (%) | 4209 | 16.4 (0.8) | 2804 | 16.8 (0.9) |
Obesity (%) | 4209 | 17.9 (1.0) | 2804 | 17.0 (1.2) |
%FM | 4209 | 25.4 (0.2) | 2804 | 32.9 (0.3) |
FMI (kg/m2) | 4209 | 5.9 (0.1) | 2804 | 7.6 (0.1) |
BMI (kg/m2) | 4209 | 21.9 (0.1) | 2804 | 22.0 (0.2) |
%TFMd | 4209 | 22.2 (0.3) | 2804 | 29.0 (0.3) |
WC (cm) | 4209 | 77.0 (0.4) | 2804 | 76.1 (0.5) |
WHtR | 4209 | 0.479 (0.002) | 2804 | 0.492 (0.003) |
Blood pressure | ||||
SBP (mmHg) | 3336 | 112.1 (0.5) | 2261 | 108.3 (0.8) |
DBP (mmHg) | 3336 | 62.3 (0.4) | 2261 | 63.3 (0.4) |
Fasting and nonfasting blood CVD risk factors | ||||
TC (mg/dL) | 3928 | 161.8 (0.9) | 2621 | 164.5 (0.9) |
HDL-C (mg/dL) | 3928 | 49.2 (0.4) | 2621 | 52.8 (0.3) |
CRP (mg/dL) | 3928 | 0.15 (0.01) | 2621 | 0.15 (0.01) |
Fasting blood CVD risk factors in 12-19 years | ||||
TG (mg/dL) | 1370 | 91.6 (2.3) | 857 | 82.8 (2.6) |
LDL-C (mg/dL) | 1370 | 92.6 (1.2) | 857 | 91.8 (1.2) |
Glucose (mg/dL) | 1370 | 93.3 (0.4) | 857 | 90.0 (0.4) |
Insulin (μU/mL) | 1370 | 11.8 (0.3) | 857 | 12.1 (0.4) |
Abbreviations: SE for standard error, %FM for percent fat mass, FMI for fat mass index, BMI for body mass index, %TFM for percent trunk fat mass, WC for waist circumference, WHtR for waist-to-height ratio, SBP for systolic blood pressure, DBP for diastolic blood pressure, TC for total cholesterol, TG for triglycerides, HDL-C for high-density lipoprotein cholesterol, LDL-C for low-density lipoprotein cholesterol, and CRP for C-reactive protein.
Analytic sample sizes (n) were unweighted. DXA data were not collected for girls aged 8-17 years in 1999-2000 due to unresolved Institutional Review Board issues concerning the reporting pregnancy test results in minors.
Mean and standard error were adjusted for complex sampling design.
%TFM was calculated as total trunk fat mass divided by total trunk mass times 100.
Table 2 shows partial Pearson correlation coefficients between DXA and anthropometric measures of body fat. Over all youth (8-19 year-olds), BMI was strongly correlated with %FM (r = 0.79) and FMI (r = 0.94). WC (r = 0.84) and WHtR (r = 0.90) were strongly correlated with %TFM. Similar patterns were seen after stratification by gender and age, or by race-ethnicity.
Table 2.
n d | BMI | WC | WHtR | ||
---|---|---|---|---|---|
%FM | FMI | %TFM | |||
All youth | 7013 | 0.79 | 0.94 | 0.84 | 0.90 |
Race-ethnicity | |||||
Mexican Americans | 2425 | 0.80 | 0.95 | 0.85 | 0.89 |
Non-Hispanic Whites | 1779 | 0.80 | 0.94 | 0.85 | 0.90 |
Non-Hispanic Blacks | 2270 | 0.82 | 0.96 | 0.87 | 0.90 |
Boys | |||||
8-19 years | 4209 | 0.81 | 0.94 | 0.88 | 0.91 |
8-11 years | 948 | 0.86 | 0.96 | 0.90 | 0.91 |
12-19 years | 3261 | 0.80 | 0.94 | 0.88 | 0.92 |
Girls | |||||
8-19 years | 2804 | 0.82 | 0.97 | 0.85 | 0.89 |
8-11 years | 651 | 0.85 | 0.97 | 0.87 | 0.89 |
12-19 years | 2153 | 0.83 | 0.97 | 0.86 | 0.89 |
Pearson correlation coefficients were adjusted for age in years, gender and race-ethnicity whenever applicable. Each correlation coefficient is significantly different from 0 (P < 0.0001).
%TFM was calculated as total trunk fat mass divided by total trunk mass times 100.
Abbreviations: BMI for body mass index, %FM for percent fat mass, FMI for fat mass index, WC for waist circumference, WHtR for waist-to-height ratio, %TFM for percent trunk fat mass.
Analytic sample sizes (n) were unweighted. DXA data were not collected for girls aged 8-17 years in 1999-2000 due to unresolved Institutional Review Board issues concerning the reporting pregnancy test results in minors.
Among all youth (Table 3), BMI was more strongly correlated with SBP (r: 0.22 vs. 0.12) and HDL-C (r: −0.34 vs. −0.25) than %FM. At the P < 0.05 level, BMI was more strongly correlated with fasting insulin level (r: 0.59 vs. 0.51) but less strongly correlated with TC (r: 0.13 vs. 0.19) than %FM. Compared to FMI, BMI was more strongly correlated with SBP (r: 0.22 vs. 0.17) and HDL-C (r: −0.34 vs. −0.31) but less strongly correlated with TC (r: 0.13 vs. 0.18). In the comparisons of correlation of central fat variables with cardiovascular risk factors (Table 4), WC was more strongly correlated with HDL-C (r: −0.37 vs. −0.30) but less strongly associated with TC (r: 0.12 vs. 0.21) than %TFM. At P < 0.05 level, WC was more strongly correlated with SBP (r: 0.22 vs. 0.15) and fasting insulin (r: 0.61 vs. 0.56) but less strongly correlated to LDL-C (r: 0.18 vs. 0.25) than %TFM. WHtR was more strongly correlated with SBP (r: 0.20 vs. 0.15) and HDL-C (r: −0.35 vs. −0.30) than %TFM.
Table 3.
8-19 years | 12-19 years | |||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|
Race-ethnicity and total body adiposity |
n | SBP (mmHg) |
DBP (mmHg) |
n | TC (mg/dL) |
HDL-C (mg/dL) |
CRP (mg/dL) |
n | TG (mg/dL) |
LDL-C (mg/dL) |
Glucose (mg/dL) |
Insulin (μU/mL) |
All youth | ||||||||||||
%FM | 5597 | 0.12† | 0.03 | 6549 | 0.19* | −0.25† | 0.48 | 2227 | 0.28 | 0.23 | 0.09 | 0.51* |
FMI (kg/m2) | 5597 | 0.17* | 0.02 | 6549 | 0.18* | −0.31* | 0.52 | 2227 | 0.30 | 0.22 | 0.11 | 0.58 |
BMI (kg/m2) | 5597 | 0.22 | 0.01 | 6549 | 0.13 | −0.34 | 0.50 | 2227 | 0.28 | 0.19 | 0.11 | 0.59 |
Mexican Americans | ||||||||||||
%FM | 1998 | 0.14* | 0.03 | 2281 | 0.23 | −0.25* | 0.48 | 768 | 0.35 | 0.25 | 0.15 | 0.55* |
FMI (kg/m2) | 1998 | 0.18 | 0.02 | 2281 | 0.22 | −0.31 | 0.51 | 768 | 0.39 | 0.27 | 0.16 | 0.62 |
BMI (kg/m2) | 1998 | 0.21 | 0.00 | 2281 | 0.19 | −0.34 | 0.47 | 768 | 0.38 | 0.25 | 0.16 | 0.62 |
Non-Hispanic Whites | ||||||||||||
%FM | 1454 | 0.13* | 0.02 | 1646 | 0.19* | −0.24* | 0.50 | 581 | 0.27 | 0.19 | 0.06 | 0.51 |
FMI (kg/m2) | 1454 | 0.18 | 0.02 | 1646 | 0.17 | −0.31 | 0.53 | 581 | 0.29 | 0.19 | 0.09 | 0.57 |
BMI (kg/m2) | 1454 | 0.22 | 0.00 | 1646 | 0.12 | −0.35 | 0.50 | 581 | 0.28 | 0.16 | 0.11 | 0.59 |
Non-Hispanic Blacks | ||||||||||||
%FM | 1711 | 0.11* | 0.04 | 2123 | 0.15* | −0.23* | 0.47 | 703 | 0.26 | 0.23 | 0.17 | 0.55 |
FMI (kg/m2) | 1711 | 0.16 | 0.04 | 2123 | 0.13 | −0.28 | 0.50 | 703 | 0.29 | 0.22 | 0.19 | 0.61 |
BMI (kg/m2) | 1711 | 0.20 | 0.02 | 2123 | 0.08 | −0.32 | 0.48 | 703 | 0.30 | 0.18 | 0.19 | 0.61 |
Abbreviations: %FM for percent fat mass, FMI for fat mass index calculated as total body fat in kg divided by the square of height in meters, BMI for body mass index calculated as body weight in kg divided by the square of height in meters, CVD for cardiovascular disease, SBP for systolic blood pressure, DBP for diastolic blood pressure, TC for total cholesterol, HDL-C for high-density lipoprotein cholesterol, CRP for C-reactive protein, TG for triglycerides, and LDL-C for low-density lipoprotein cholesterol.
All variables were log-transformed except SBP and DBP. For analysis by ethnicity, Pearson correlation coefficients were adjusted for age in years and gender. For analysis in all youth, race-ethnicity was additionally adjusted for. Survey year (1999-2002 and 2003-2004) was additionally adjusted for in the correlation analyses regarding fasting insulin.
Within each race-ethnicity group and CVD risk factor, correlation coefficients of BMI was compared to those of %FM and FMI, respectively. Comparisons were not made between correlation coefficients of %FM and FMI and between race-ethnicity groups.
Significantly different from the correlation coefficient between BMI and CVD risk factor of interest at α = 0.05 level.
Significantly different from the correlation coefficient between BMI and CVD risk factor of interest, P < 0.000195 (equivalent to P < 0.05 after Bonferroni-correction).
Table 4.
8-19 years | 12-19 years | |||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|
Race-ethnicity and regional body adiposity |
n | SBP (mmHg) |
DBP (mmHg) |
n | TC (mg/dL) |
HDL-C (mg/dL) |
CRP (mg/dL) |
n | TG (mg/dL) |
LDL-C (mg/dL) |
Glucose (mg/dL) |
Insulin (μU/mL) |
All youth | ||||||||||||
WC (cm) | 5597 | 0.22* | 0.03 | 6549 | 0.12† | −0.37† | 0.49 | 2227 | 0.33 | 0.18* | 0.13 | 0.61* |
WHtR | 5597 | 0.20* | 0.02 | 6549 | 0.18 | −0.35* | 0.52 | 2227 | 0.32 | 0.21 | 0.11 | 0.59 |
%TFMc | 5597 | 0.15 | 0.04 | 6549 | 0.21 | −0.30 | 0.51 | 2227 | 0.32 | 0.25 | 0.09 | 0.56 |
Mexican Americans | ||||||||||||
WC (cm) | 1998 | 0.23* | 0.03 | 2281 | 0.19 | −0.34 | 0.47 | 768 | 0.39 | 0.25 | 0.17 | 0.63 |
WHtR | 1998 | 0.20 | 0.01 | 2281 | 0.23 | −0.32 | 0.50 | 768 | 0.40 | 0.26 | 0.17 | 0.64 |
%TFMc | 1998 | 0.16 | 0.03 | 2281 | 0.24 | −0.29 | 0.49 | 768 | 0.39 | 0.28 | 0.16 | 0.58 |
Non-Hispanic Whites | ||||||||||||
WC (cm) | 1454 | 0.21 | 0.02 | 1646 | 0.11* | −0.38* | 0.50 | 581 | 0.33 | 0.15 | 0.10 | 0.61 |
WHtR | 1454 | 0.20 | 0.02 | 1646 | 0.17 | −0.36 | 0.53 | 581 | 0.31 | 0.18 | 0.10 | 0.58 |
%TFMc | 1454 | 0.16 | 0.05 | 1646 | 0.21 | −0.30 | 0.53 | 581 | 0.32 | 0.22 | 0.06 | 0.56 |
Non-Hispanic Blacks | ||||||||||||
WC (cm) | 1711 | 0.21* | 0.05 | 2123 | 0.09 | −0.34* | 0.48 | 703 | 0.33 | 0.21 | 0.20 | 0.64 |
WHtR | 1711 | 0.18 | 0.02 | 2123 | 0.13 | −0.31 | 0.51 | 703 | 0.31 | 0.22 | 0.19 | 0.62 |
%TFMc | 1711 | 0.13 | 0.05 | 2123 | 0.15 | −0.27 | 0.49 | 703 | 0.29 | 0.24 | 0.17 | 0.58 |
Abbreviations: WC for waist circumference, WHtR for waist-to-height ratio, %TFM for percent trunk fat mass, CVD for cardiovascular disease, SBP for systolic blood pressure, DBP for diastolic blood pressure, TC for total cholesterol, HDL-C for high-density lipoprotein cholesterol, CRP for C-reactive protein, TG for triglycerides, and LDL-C for low-density lipoprotein cholesterol.
All variables were log-transformed except SBP and DBP. For analysis by ethnicity, Pearson correlation coefficients were adjusted for age in years and gender. For analysis in all youth, race-ethnicity was additionally adjusted for. Survey year (1999-2002 and 2003-2004) was additionally adjusted for in the correlation analyses regarding fasting insulin.
Within each race-ethnicity group and CVD risk factor, correlation coefficients of WC and WHtR were respectively compared to that of %TFM. Comparisons were not made between correlation coefficients of WC and WHtR and between race-ethnicity groups.
Significantly different from the correlation coefficient between %TFM and CVD risk factor of interest at α = 0.05 level.
Significantly different from the correlation coefficient between %TFM and CVD risk factor of interest, P < 0.000195 (equivalent to P < 0.05 after Bonferroni-correction).
Fewer significant differences were observed in the race-ethnicity (i.e., Mexican American, non-Hispanic White and Black) specific analyses than in all youth combined analyses (Table 3 and Table 4). In the comparisons of correlation of total body fat variables with cardiovascular risk factors (Table 3), no significant differences were observed in each of the 3 race-ethnicities at the Bonferroni-corrected P level. At P < 0.05 level, BMI was more strongly correlated with SBP and HDL-C in each of the 3 race-ethnicities. Furthermore, BMI was more strongly correlated with fasting insulin level in Mexican Americans (r: 0.62 vs. 0.55) but less strongly correlated with TC in non-Hispanic Whites (r: 0.12 vs. 0.19) and Blacks (r: 0.08 vs. 0.15) than %FM. In the comparisons of correlation of central fat variables with cardiovascular risk factors (Table 4), compared to %TFM, WC was more strongly correlated with SBP in Mexican American (r: 0.23 vs. 0.16) and non-Hispanic Black (r: 0.21 vs. 0.13) youth, with HDL-C in non-Hispanic White (r: − 0.38 vs. −0.30) and Black (r: − 0.34 vs. – 0.27) youth but less strongly correlated with TC in non-Hispanic White youth (r: 0.11 vs. 0.21).
Table 5 compares partial Pearson correlation coefficients of total body fat variables (i.e., BMI vs. %FM, and BMI vs. FMI) with each of the nine cardiovascular risk factors within gender- and age-specific groups. No significant differences were observed at the Bonferroni-corrected P level. At P < 0.05 level, significant differences were observed only in boys aged 12-19 years. Specifically, compared to %FM, BMI was more strongly correlated with SBP (r: 0.20 vs. 0.10), HDL-C (r: −0.38 vs. −0.32) and fasting insulin (r: 0.64 vs. 0.56). Compared to FMI, BMI was more strongly correlated with SBP (r: 0.20 vs 0.15).
Table 5.
Blood pressure | Fasting and nonfasting blood CVD risk factors | Fasting blood CVD risk factors | ||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|
Gender, age group and total body fat |
n | SBP (mmHg) |
DBP (mmHg) |
n | TC (mg/dL) |
HDL-C (mg/dL) |
CRP (mg/dL) |
n | TG (mg/dL) |
LDL-C (mg/dL) |
Glucose (mg/dL) |
Insulin (μU/mL) |
Boys | ||||||||||||
8 - 11 years | ||||||||||||
%FM | 683 | 0.21 | 0.19 | 859 | 0.14 | −0.32 | 0.57 | - | - | - | - | - |
FMI (kg/m2) | 683 | 0.27 | 0.19 | 859 | 0.13 | −0.33 | 0.58 | - | - | - | - | - |
BMI (kg/m2) | 683 | 0.31 | 0.18 | 859 | 0.12 | −0.32 | 0.56 | - | - | - | - | - |
12 - 19 years | ||||||||||||
%FM | 2653 | 0.10* | −0.02 | 3069 | 0.27 | −0.32* | 0.47 | 1370 | 0.36 | 0.28 | 0.13 | 0.56* |
FMI (kg/m2) | 2653 | 0.15* | −0.03 | 3069 | 0.26 | −0.37 | 0.50 | 1370 | 0.38 | 0.28 | 0.14 | 0.62 |
BMI (kg/m2) | 2653 | 0.20 | −0.03 | 3069 | 0.22 | −0.38 | 0.48 | 1370 | 0.37 | 0.26 | 0.15 | 0.64 |
Girls | ||||||||||||
8 - 11 years | ||||||||||||
%FM | 479 | 0.16 | 0.03 | 599 | 0.08 | −0.37 | 0.56 | - | - | - | - | |
FMI (kg/m2) | 479 | 0.20 | 0.03 | 599 | 0.05 | −0.40 | 0.57 | - | - | - | - | |
BMI (kg/m2) | 479 | 0.23 | 0.02 | 599 | 0.02 | −0.40 | 0.53 | - | - | - | - | |
12 - 19 years | ||||||||||||
%FM | 1782 | 0.18 | −0.01 | 2022 | 0.14 | −0.22 | 0.47 | 857 | 0.17 | 0.14 | 0.05 | 0.48 |
FMI (kg/m2) | 1782 | 0.20 | −0.02 | 2022 | 0.11 | −0.25 | 0.52 | 857 | 0.17 | 0.13 | 0.06 | 0.54 |
BMI (kg/m2) | 1782 | 0.20 | −0.03 | 2022 | 0.08 | −0.26 | 0.52 | 857 | 0.15 | 0.11 | 0.06 | 0.54 |
Abbreviations: %FM for percent fat mass, FMI for fat mass index calculated as total body fat in kg divided by the square of height in meters, BMI for body mass index calculated as body weight in kg divided by the square of height in meters, CVD for cardiovascular disease, SBP for systolic blood pressure, DBP for diastolic blood pressure, TC for total cholesterol, HDL-C for high-density lipoprotein cholesterol, CRP for C-reactive protein, TG for triglycerides, and LDL-C for low-density lipoprotein cholesterol.
All variables were log-transformed except SBP and DBP. Pearson correlation coefficients were adjusted for age in years and race-ethnicity. Survey year (1999-2002 and 2003-2004) was additionally adjusted for in the correlation analyses regarding fasting insulin.
Within each gender-age group and CVD risk factor, correlation coefficient of BMI was compared to those of %FM and FMI, respectively. Comparisons were not made between correlation coefficients of %FM and FMI and between gender-age groups.
Significantly different from the correlation coefficient between BMI and CVD risk factor of interest at α = 0.05 level.
Significantly different from the correlation coefficient between BMI and CVD risk factor of interest, P < 0.000195 (equivalent to P < 0.05 after Bonferroni-correction).
Similar patterns were seen in the comparison of the correlations between central fat variables (i.e., WC vs. %TFM, and WHtR vs. %TFM) with cardiovascular risk factors except for TC (Table 6). At the P < 0.05 level, WC was more strongly correlated with SBP (r: 0.20 vs. 0.12), HDL-C (r: −0.41 vs. −0.35) and fasting insulin (r: 0.66 vs. 0.59) than %TFM in boys aged 12-19 years but less strongly correlated with TC in youth aged 12-19 years across genders (r: 0.23 vs. 0.29 for boys and 0.07 vs. 0.15 for girls). No significant differences were found between correlations of WHtR and %TFM to the nine cardiovascular risk factors.
Table 6.
Blood pressure | Fasting and nonfasting blood CVD risk factors | Fasting blood CVD risk factors | ||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|
Gender, age group and regional body fat |
n | SBP (mmHg) |
DBP (mmHg) |
n | TC (mg/dL) |
HDL-C (mg/dL) |
CRP (mg/dL) |
n | TG (mg/dL) |
LDL-C (mg/dL) |
Glucose (mg/dL) |
Insulin (μU/mL) |
Boys | ||||||||||||
8 - 11 years | ||||||||||||
WC (cm) | 683 | 0.29 | 0.17 | 859 | 0.08 | −0.37 | 0.56 | - | - | - | - | - |
WHtR | 683 | 0.26 | 0.14 | 859 | 0.12 | −0.38 | 0.56 | - | - | - | - | - |
%TFM | 683 | 0.24 | 0.22 | 859 | 0.14 | −0.34 | 0.58 | - | - | - | - | - |
12 - 19 years | ||||||||||||
WC (cm) | 2653 | 0.20* | 0.00 | 3069 | 0.23* | −0.41* | 0.50 | 1370 | 0.41 | 0.25 | 0.15 | 0.66* |
WHtR | 2653 | 0.16 | −0.01 | 3069 | 0.27 | −0.38 | 0.50 | 1370 | 0.41 | 0.28 | 0.14 | 0.63 |
%TFM | 2653 | 0.12 | 0.00 | 3069 | 0.29 | −0.35 | 0.49 | 1370 | 0.40 | 0.30 | 0.13 | 0.59 |
Girls | ||||||||||||
8 - 11 years | ||||||||||||
WC (cm) | 479 | 0.25 | 0.06 | 599 | 0.01 | −0.44 | 0.52 | - | - | - | - | |
WHtR | 479 | 0.24 | 0.05 | 599 | 0.08 | −0.44 | 0.51 | - | - | - | - | |
%TFM | 479 | 0.18 | 0.04 | 599 | 0.09 | −0.40 | 0.56 | - | - | - | - | |
12 - 19 years | ||||||||||||
WC (cm) | 1782 | 0.18 | −0.02 | 2022 | 0.07* | −0.27 | 0.49 | 857 | 0.20 | 0.09 | 0.08 | 0.54 |
WHtR | 1782 | 0.20 | −0.03 | 2022 | 0.11 | −0.29 | 0.51 | 857 | 0.19 | 0.11 | 0.07 | 0.55 |
%TFM | 1782 | 0.19 | 0.00 | 2022 | 0.15 | −0.25 | 0.49 | 857 | 0.20 | 0.17 | 0.05 | 0.52 |
Abbreviations: WC for waist circumference, WHtR for waist-to-height ratio, %TFM for percent trunk fat mass, CVD for cardiovascular disease, SBP for systolic blood pressure, DBP for diastolic blood pressure, TC for total cholesterol, HDL-C for high-density lipoprotein cholesterol, CRP for C-reactive protein, TG for triglycerides, and LDL-C for low-density lipoprotein cholesterol.
All variables were log-transformed except SBP and DBP. Pearson correlation coefficients were adjusted for age in years and race-ethnicity. Survey year (1999-2002 and 2003-2004) was additionally adjusted for in the correlation analyses regarding fasting insulin.
Within each gender-age group and CVD risk factor, correlation coefficients of WC and WHtR were respectively compared to that of %TFM. Comparisons were not made between correlation coefficients of WC and WHtR and between gender-age groups.
Significantly different from the correlation coefficient between %TFM and CVD risk factor of interest at α = 0.05 level.
Significantly different from the correlation coefficient between %TFM and CVD risk factor of interest, P < 0.000195 (equivalent to P < 0.05 after Bonferroni-correction).
Discussion
We found that anthropometric (BMI) and DXA (%FM and FMI) measures of total body fat were strongly correlated, and were comparably associated with cardiovascular risk factors. Similar patterns were observed between anthropometric (WC and WHtR) and DXA (%TFM) measures of central fat. These relationships were largely consistent across different age, gender and race-ethnicity groups.
Our study confirms the strong correlations between BMI and DXA-assessed %FM observed in previous, more limited studies of children and adolescents [2-8, 28], and adds to the growing evidence that BMI can accurately distinguish lean individuals from those with higher body fat at the population level. Our study is also consistent with previous studies that found that BMI was significantly correlated with cardiovascular risk factors in children and adolescents, and that these correlations were at least as strong as the correlations between DXA-assessed %FM and cardiovascular risk factors in this age group [5-7, 9, 11].
We found that BMI and FMI were highly correlated (r: 0.94-0.97), and that both BMI and FMI were similarly correlated with cardiovascular risk factors. This is consistent with a study in children in New Zealand which showed that addition of BMI into a base model including age, gender and ethnicity additionally explained similar proportions of variance in TC and HDL-C compared to addition of FMI into the model [29]. However, in this study body fat was assessed using bioimpedence analysis, which is associated with larger error than methods such as DXA [30].
Since DXA cannot distinguish visceral fat from abdominal subcutaneous fat, we could not examine the validity of WC and WHtR in assessing distribution of fat between these compartments. However, DXA-assessed trunk FM has been shown to be strongly correlated with visceral abdominal tissue (r = 0.87) and subcutaneous abdominal adiposity tissue (r = 0.96) assessed by CT in children [31]. The correlation between %TFM and WC in our study was similar to correlations between MRI- or CT-assessed visceral abdominal tissue and WC (r: 0.8 to 0.9) found in previous studies [31-33]. No study has evaluated the validity of WC and WHtR as an indicator of central fat in relation to cardiovascular risk factors in youth compared to MRI- or CT-measured central fat. With body fat measured by DXA, one study reported trunk FM, WC and WHtR were similarly correlated to blood cholesterols in Portuguese children 10 to 15 years old [9]. This is consistent with our study, where the comparable correlations between WC and WHtR with cardiovascular risk factors suggest that WHtR was not superior to WC in relation to cardiovascular risk factors. This indicates that adjustment for height contributes little to the ability of WC to discriminate individuals with different levels of cardiovascular risk [34], and argues for the use of WC, which is a simpler indicator of central fat than WHtR, in population-based studies.
The only other study that used NHANES participants to examine the validity of anthropometric indices of body fat against DXA-assessed body fat in relation to cardiovascular risk factors focused on adults [12]. Comparing their results with those shown here in youth, the magnitudes of correlations between body fat variables and cardiovascular risk factors are similar [12]. It is notable that correlations between body fat and cardiovascular risk factors were weak to modest (the maximum correlation coefficient is 0.66), suggesting that cardiovascular risk factors are determined by other factors in addition to body fat. For studies that aim to accurately predict levels of cardiovascular risk factors or to screen those with abnormal levels of cardiovascular risk factors, other factors influencing cardiovascular risk should be considered.
A strength of our analysis include the large size and the nationally representative nature of our sample. To our knowledge, this is the first study to evaluate the validity of anthropometric measures of body fat as indicators of body fat and their correlations with cardiovascular risk factors in a large-scale nationally representative sample of American children and adolescents. Nevertheless, the cross-sectional nature of this work is a weakness. A recent study of over 5,000 youth conducted in the United Kingdom reported that changes in z-scores of BMI, WC and FM (by DXA) from 9-12 years to 15-16 years were similarly associated with cardiovascular risk factors at 15-16 years of age [35]. Thus, that longitudinal work is consistent with the cross-sectional analyses presented here.
Another weakness of this work was that we assessed trunk fat rather than directly measured visceral fat. Also, DXA data were imputed using multiple imputation methods for 8.6% of participants. We could not examine waist-to-hip ratio because hip circumference was not collected in NHANES 1999-2004. Finally, we could not evaluate relations with metabolic syndrome because of the lack of a well-accepted definition of metabolic syndrome in children and adolescents.
The feasibility and reproducibility of measurement of height and weight are generally high. Although the accurate measurement of WC is more challenging [36] especially in those who are morbidly obese, careful quality control can reduce measurement error. This study supports the continued use of BMI and WC as surrogates for body fat in cross-sectional epidemiologic studies that aim to investigate the association of excess body fat with cardiovascular risk factors in children and adolescents.
Acknowledgement
The project described was supported by Award Number 1U01HL103561 from the National Heart, Lung, And Blood Institute, the Eunice Kennedy Shriver National Institute of Child Health and Development and the Office of Behavioral and Social Sciences Research. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Heart, Lung, and Blood Institute or the National Institutes of Health.
Footnotes
Conflict of interest – The authors have no disclosure to make.
References
- 1.Ogden CL, Carroll MD, Kit BK, et al. Prevalence of obesity and trends in body mass index among US children and adolescents, 1999-2010. JAMA. 2012;307:483–490. doi: 10.1001/jama.2012.40. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Pietrobelli A, Faith MS, Allison DB, et al. Body mass index as a measure of adiposity among children and adolescents: a validation study. J Pediatr. 1998;132:204–210. doi: 10.1016/s0022-3476(98)70433-0. [DOI] [PubMed] [Google Scholar]
- 3.Dencker M, Thorsson O, Linden C, et al. BMI and objectively measured body fat and body fat distribution in prepubertal children. Clin Physiol Funct Imaging. 2007;27:12–16. doi: 10.1111/j.1475-097X.2007.00709.x. [DOI] [PubMed] [Google Scholar]
- 4.Goran MI, Driscoll P, Johnson R, Nagy TR, Hunter G. Cross-calibration of body-composition techniques against dual-energy X-ray absorptiometry in young children. Am J Clin Nutr. 1996;63:299–305. doi: 10.1093/ajcn/63.3.299. [DOI] [PubMed] [Google Scholar]
- 5.Lindsay RS, Hanson RL, Roumain J, et al. Body mass index as a measure of adiposity in children and adolescents: relationship to adiposity by dual energy x-ray absorptiometry and to cardiovascular risk factors. J Clin Endocrinol Metab. 2001;86:4061–4067. doi: 10.1210/jcem.86.9.7760. [DOI] [PubMed] [Google Scholar]
- 6.Steinberger J, Jacobs DR, Raatz S, et al. Comparison of body fatness measurements by BMI and skinfolds vs dual energy X-ray absorptiometry and their relation to cardiovascular risk factors in adolescents. Int J Obes (Lond) 2005;29:1346–1352. doi: 10.1038/sj.ijo.0803026. [DOI] [PubMed] [Google Scholar]
- 7.Falaschetti E, Hingorani AD, Jones A, et al. Adiposity and cardiovascular risk factors in a large contemporary population of pre-pubertal children. Eur Heart J. 2010;31:3063–3072. doi: 10.1093/eurheartj/ehq355. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Goulding A, Gold E, Cannan R, et al. DEXA supports the use of BMI as a measure of fatness in young girls. Int J Obes Relat Metab Disord. 1996;20:1014–1021. [PubMed] [Google Scholar]
- 9.Teixeira PJ, Sardinha LB, Going SB, et al. Total and regional fat and serum cardiovascular disease risk factors in lean and obese children and adolescents. Obes Res. 2001;9:432–442. doi: 10.1038/oby.2001.57. [DOI] [PubMed] [Google Scholar]
- 10.Daniels SR, Morrison JA, Sprecher DL, et al. Association of body fat distribution and cardiovascular risk factors in children and adolescents. Circulation. 1999;99:541–545. doi: 10.1161/01.cir.99.4.541. [DOI] [PubMed] [Google Scholar]
- 11.Parrett AL, Valentine RJ, Arngrimsson SA, et al. Adiposity and aerobic fitness are associated with metabolic disease risk in children. Appl Physiol Nutr Metab. 2011;36:72–79. doi: 10.1139/H10-083. [DOI] [PubMed] [Google Scholar]
- 12.Sun Q, van Dam RM, Spiegelman D, et al. Comparison of dual-energy x-ray absorptiometric and anthropometric measures of adiposity in relation to adiposity-related biologic factors. Am J Epidemiol. 2010;172:1442–1454. doi: 10.1093/aje/kwq306. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Maynard LM, Wisemandle W, Roche AF, et al. Childhood body composition in relation to body mass index. Pediatrics. 2001;107:344–350. doi: 10.1542/peds.107.2.344. [DOI] [PubMed] [Google Scholar]
- 14.Lamb MM, Ogden CL, Carroll MD, et al. Association of body fat percentage with lipid concentrations in children and adolescents: United States, 1999-2004. Am J Clin Nutr. 2011;94:877–883. doi: 10.3945/ajcn.111.015776. [DOI] [PubMed] [Google Scholar]
- 15.National Center for Health Statistics [accessed 4 June 2012];National Health and Nutrition Examination Survey. Available from: http://www.cdc.gov/nchs/nhanes/nhanes_questionnaires.htm.
- 16.Lohman TG, Roche AF, Martorell R. Anthropometric standardization reference manual. Human Kinetics Books; Champaign, Illinois: 1988. [Google Scholar]
- 17.Centers for Disease Control and Prevention [accessed 4 June 2012];Percentile Data Files with LMS Values. Available from: http://www.cdc.gov/growthcharts/percentile_data_files.htm.
- 18.Rubin DB. Multiple imputation for nonresponse in surveys. Wiley Classics Library; Hoboken, NJ: 2004. [Google Scholar]
- 19.Raghunathan TE, Lepkowski JM, Van Hoewyk J, et al. A multivariate technique for multiply imputing missing values using a sequence of regression models. Surv Methodol. 2001;27:85–95. [Google Scholar]
- 20.National Center for Health Statistics [cited September 25, 2012];National Health and Nutrition Examination Survey: technical documentation for the 1999-2004 dual energy X-ray absorptiometry (DXA) multiple imputation data files. Available from http://www.cdc.gov/nchs/data/nhanes/dxa/dxa_techdoc.pdf.
- 21.Perloff D, Grim C, Flack J, et al. Human blood pressure determination by sphygmomanometry. Circulation. 1993;88:2460–2470. doi: 10.1161/01.cir.88.5.2460. [DOI] [PubMed] [Google Scholar]
- 22.National Center for Health Statistics [cited September 25, 2012];National Health and Nutrition Examination Survey. Available from http://www.cdc.gov/nchs/nhanes/nhanes2003-2004/l13_c.htm.
- 23.Friedewald WT, Levy RI, Fredrickson DS. Estimation of the concentration of low-density lipoprotein cholesterol in plasma, without use of the preparative ultra-centrifuge. Clin Chem. 1972;18:499–502. [PubMed] [Google Scholar]
- 24.National Center for Health Statistics [cited September 25, 2012];National Health and Nutrition Examination Survey: C-reactive protein. Available from http://www.cdc.gov/nchs/nhanes/nhanes1999-2000/LAB11.htm.
- 25.National Center for Health Statistics [accessed 25 September 2012];National Health and Nutrition Examination Survey. Available from http://www.cdc.gov/nchs/nhanes/nhanes2003-2004/lab_methods_03_04.htm.
- 26.National Center for Health Statistics [cited September 25, 2012];National Health and Nutrition Examination Survey. Available from http://www.cdc.gov/nchs/nhanes/nhanes2003-2004/L10AM_C.htm.
- 27.Rosner B. Regression and correlation methods. In: Rosner B, editor. Fundmentals of Biostatistics. 6th Duxbury Press; Pacific Grove, CA: 2006. pp. 464–556. [Google Scholar]
- 28.Daniels SR, Khoury PR, Morrison JA. The utility of body mass index as a measure of body fatness in children and adolescents: differences by race and gender. Pediatrics. 1997;99:804–807. doi: 10.1542/peds.99.6.804. [DOI] [PubMed] [Google Scholar]
- 29.Rush EC, Scragg R, Schaaf D, et al. Indices of fatness and relationships with age, ethnicity and lipids in New Zealand European, Maori and Pacific children. Eur J Clin Nutr. 2009;63:627–633. doi: 10.1038/ejcn.2008.15. [DOI] [PubMed] [Google Scholar]
- 30.Lee SY, Gallagher D. Assessment methods in human body composition. Curr Opin Clin Nutr Metab Care. 2008;11:566–572. doi: 10.1097/MCO.0b013e32830b5f23. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Goran MI, Gower BA, Treuth M, et al. Prediction of intra-abdominal and subcutaneous abdominal adipose tissue in healthy pre-pubertal children. Int J Obes Relat Metab Disord. 1998;22:549–558. doi: 10.1038/sj.ijo.0800624. [DOI] [PubMed] [Google Scholar]
- 32.Harrington DM, Staiano AE, Broyles ST, et al. Waist circumference measurement site does not affect relationships with visceral adiposity and cardiometabolic risk factors in children. Pediatr Obes. 2012;8:199–206. doi: 10.1111/j.2047-6310.2012.00106.x. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Brambilla P, Bedogni G, Moreno LA, et al. Crossvalidation of anthropometry against magnetic resonance imaging for the assessment of visceral and sub-cutaneous adipose tissue in children. Int J Obes (Lond) 2006;30:23–30. doi: 10.1038/sj.ijo.0803163. [DOI] [PubMed] [Google Scholar]
- 34.Kawada T. Waist-height ratio, indicator of central obesity, is strongly related to waist circumference but there is poor additional information regarding waist circumference. Circ J. 2010;74:2501. doi: 10.1253/circj.cj-10-0583. [DOI] [PubMed] [Google Scholar]
- 35.Lawlor DA, Benfield L, Logue J, et al. Association between general and central adiposity in childhood, and change in these, with cardiovascular risk factors in adolescence: prospective cohort study. BMJ. 2010;341:c6224. doi: 10.1136/bmj.c6224. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Ulijaszek SJ, Kerr DA. Anthropometric measurement error and the assessment of nutritional status. Br J Nutr. 1999;82:165–177. doi: 10.1017/s0007114599001348. [DOI] [PubMed] [Google Scholar]