点击显示 收起
1 From the Laboratory for Human Nutrition, Institute for Food Science and Nutrition (MBZ), and the Institute for Pharmaceutical Science (CG and CP), Swiss Federal Institute of Technology Zürich, Switzerland; and the Department of Growth and Development, University Childrens’ Hospital, Zürich, Switzerland (LM).
2 Supported by the Swiss Foundation for Nutrition Research, Zürich, Switzerland, and the Swiss Federal Institute of Technology, Zürich, Switzerland. 3 Reprints not available. Address correspondence to MB Zimmermann, Laboratory for Human Nutrition, Swiss Federal Institute of Technology Zürich, PO Box 474, Seestrasse 72, CH-8803 Rüschlikon, Switzerland. E-mail: michael.zimmermann{at}ilw.agrl.ethz.ch.
ABSTRACT
Background: For defining overweight in children, reference values for body mass index (BMI) are available from the US Centers for Disease Control and Prevention (CDC) and the International Obesity Task Force (IOTF). However, these 2 sets of reference criteria differ, and their accuracy in classifying adiposity has not yet been validated in most countries.
Objective: We compared BMI criteria from the IOTF and the CDC with percentage of body fat (%BF) from multisite skinfold thicknesses (SFTs) for identification of overweight in 6–12-y-old Swiss children.
Design: In a representative sample (n = 2431), weight, height, and 4 SFTs were measured. Regression and receiver operating characteristic (ROC) curves were used to evaluate BMI as an indicator of adiposity.
Results: BMI and %BF were well correlated (r2 = 0.74), and the areas under the ROC curves for overweight and obesity were 0.956–0.992. The sensitivity and specificity of the IOTF and CDC overweight criteria and of the CDC obesity criteria were high. The sensitivity of the IOTF obesity criteria was only 48% and 62% in boys and girls, respectively. Overall, the performance of the CDC criteria was superior. With the use of the CDC criteria, the prevalence of overweight in girls and boys was 19.1% and 20.3%, respectively.
Conclusions: BMI is an excellent proxy measure of adiposity in 6–12-y-old children. In Swiss children, both BMI criteria accurately predict overweight, but the sensitivity of the IOTF obesity criteria is poor. They failed to detect one-half of the children identified as obese on the basis of %BF from SFTs.
Key Words: Body mass index • skinfold thickness • anthropometry • percentage of body fat • sensitivity • specificity • children • Switzerland
INTRODUCTION
Measurement of body mass index (BMI; in kg/m2) is a practical and reproducible method for classifying overweight in adults (1, 2) and is increasingly recommended for screening overweight in children and adolescents (3-5). New growth charts from the US Centers for Disease Control and Prevention (CDC) include age- and sex-specific BMI reference values for children and adolescents aged 2–20 y (6). The International Obesity Task Force (IOTF) has also published age- and sex-specific BMI criteria for children and has proposed them as international reference values (5). Because these 2 sets of reference criteria differ, they may produce different estimates of overweight and obesity (7-9). Moreover, BMI is an expression of weight, not adiposity, and the accuracy of these reference values in classifying adiposity in children has not yet been validated in most countries (10).
Many methods available to measure body fatness, including dual-energy X-ray absorptiometry (DXA), underwater weighing, and total body potassium, are limited by their complexity and cost to research settings (11-14). In clinical and public health settings, body fatness has traditionally been estimated from skinfold thicknesses (SFTs) (1, 15, 16). Although single SFT measurements have only limited precision (17, 18), reproducibility is improved by using multisite measurements integrated into validated prediction equations (18, 19). Schaefer et al (18) reported an intraobserver CV of 2%, which corresponded to 0.4% of fractional fat mass, with the use of multisite SFTs in children. SFT measurements can accurately predict percentage of body fat (%BF) in childhood (13, 18, 20). In the present study, we compared the new CDC and IOTF sex-specific BMI-for-age reference values to %BF values estimated from multisite SFTs in screening for overweight and obesity in a nationally representative sample of 6–12-y-old Swiss schoolchildren.
SUBJECTS AND METHODS
A probability-proportionate-to-size cluster sampling based on current census data was used to obtain a representative national sample of 2600 Swiss children aged 6–12 y. This represents 1 in 250 children in this age group in Switzerland (21). Sixty communities and schools across Switzerland were identified by stratified random selection. Three or 4 classrooms were then randomly selected from each school, and all students from the selected classrooms were invited to participate. The average sample size at each school was 45 students, and the number varied according to the size of the classrooms. Ethical approval for the study was obtained from the Swiss Federal Institute of Technology, Zürich, Switzerland. Written informed consent was obtained from the school physician, the teachers, and the parents of the children.
For the measurements, the subjects removed their shoes, emptied their pockets, and wore light indoor clothing. Height and weight were measured by using standard anthropometric techniques (1). Body weight was measured to the nearest 0.1 kg by using a Tanita digital scale (HD-313; Tanita, Tokyo) calibrated with standard weights. Height was measured to the nearest 0.1 cm by using a pull-down, metal measuring tape (person-check REF 44 444, Medizintechnik KaWe; Kirchner & Wilhelm, Asperg, Germany). SFTs were measured by 2 trained examiners (CG and CP) using a Harpenden Skinfold Caliper (HSK-BI; British Indicators, West Sussex, United Kingdom) with a constant spring pressure of 10 g/mm2 and a resolution of 0.2 mm. SFTs were measured at the triceps, biceps, subscapular, and suprailiac sites (22). For the triceps, the midpoint of the back of the upper arm between the tips of the olecranal and acromial processes was determined by measuring with the arm flexed at 90°. With the arm hanging freely at the side, the caliper was applied vertically above the olecranon at the marked level. Over the biceps, the SFT was measured at the same level as the triceps, with the arm hanging freely and the palm facing outwards. At the subscapular site, the SFT was picked up just below the inferior angle of the scapula at 45° to the vertical along the natural cleavage lines of the skin. The suprailiac SFT was measured above the iliac crest, just posterior to the midaxillary line and parallel to the cleavage lines of the skin, with the arm lightly held forward. All sites were measured on the right site of the body in duplicate. For each site, 10% of the SFT measurements were repeated by a second examiner to calculate interobserver variation.
With the use of mean values from repeated SFT measurements, body density (D) and %BF were calculated according to the following equations from Deurenberg et al (23):
RESULTS
At the schools, 3413 children were invited to participate, and 2672 accepted. Of these, 64 were absent on the day of measurement. The overall response rate was 76.4%. Two percent of the subjects participated in the weight and height measurement but declined the SFT measurements. After removing subjects with incomplete data and a small number of subjects aged =" BORDER="0">13 y, a sample of 2431 subjects (1235 girls and 1196 boys) remained. The descriptive characteristics of the sample are shown in Table 1. The interobserver and intraobserver CVs for measurement of SFTs were 3.1% and 1.8%, respectively.
View this table:
TABLE 1. Descriptive characteristics of the national sample of 6–12-y-old Swiss children1
The prevalence of overweight and obesity in the sample by age and sex according to the IOTF and CDC BMI reference values is shown in Table 2. There were no significant differences between the sexes in the prevalence of overweight and obesity, although the prevalence of overweight, as assessed on the basis of the IOTF cutoffs, was higher in the girls than in the boys for all age groups. There was no significant effect of age on the prevalence of either overweight or obesity.
View this table:
TABLE 2. The prevalence of overweight and obesity in a national sample of 6–12-y-old Swiss children by age and sex according to BMI criteria from the International Obesity Task Force (IOTF) and the US Centers for Disease Control and Prevention (CDC)1
The 85th and 95th percentiles for %BF by age from the Deurenberg equation, as calculated by quantile regression for boys and girls, are shown in Figure 1. The regression of BMI on %BF for boys and for girls is shown in Table 3. The boys and the girls differed significantly in the slope of the regression (P < 0.001, ANCOVA). In the boys, age was not a significant predictor of BMI after %BF was controlled for (P = 0.6, ANCOVA). In the girls, age was a significant predictor of BMI (P = 0.001, ANCOVA), although it enhanced the multiple correlation only minimally, from 0.742 to 0.744.
FIGURE 1.. The 85th and 95th percentiles (P85 and P95, respectively) for percentage of body fat (%BF) from the Deurenberg equation, as calculated by quantile regression in a national sample of 6–12-y-old Swiss boys and girls (n = 2431).
View this table:
TABLE 3. Regression of log (BMI – 11) on percentage of body fat for boys and for girls in a national sample of 6–12-y-old Swiss children1
The ROC curves of BMI-SDS for prediction of overweight in the boys and the girls on the basis of the 85th percentile for %BF, as well as the positions on the curves of the CDC and IOTF BMI reference values for overweight, are shown in Figure 2. The ROC curves of BMI-SDS for prediction of obesity in the boys and the girls on the basis of the 95th percentile for %BF, as well as the position on the curves of the CDC and IOTF BMI reference values for obesity, are shown in Figure 3. The areas under the ROC curves of BMI-for-age for prediction of overweight and obesity in the boys and the girls on the basis of the 85th and 95th percentiles of %BF, respectively, are shown in Table 4. The sensitivity and specificity of the IOTF and CDC reference cutoffs for overweight and obesity in the boys and the girls are shown in Table 5.
FIGURE 2.. The receiver operating characteristic (ROC) curves of BMI SD scores for prediction of overweight in boys and girls on the basis of the 85th percentile for percentage of body fat calculated from skinfold thicknesses in a national sample of 6–12-y-old Swiss children (n = 2431). The position of the BMI reference values for overweight from the US Centers for Disease Control and Prevention (CDC) and the International Obesity Task Force (IOTF) on the ROC curves is shown, and the area under the curve (AUC; ±SE) is indicated for the boys and the girls.
FIGURE 3.. The receiver operating characteristic (ROC) curves of BMI SD scores for prediction of obesity in boys and girls on the basis of the 95th percentile of percentage of body fat calculated from skinfold thicknesses in a national sample of 6–12-y-old Swiss children (n = 2431). The position of the BMI reference values for obesity from the US Centers for Disease Control and Prevention (CDC) and the International Obesity Task Force (IOTF) on the ROC curves is shown, and the area under the curve (AUC; ±SE) is indicated for the boys and the girls.
View this table:
TABLE 4. Areas under the receiver operating characteristic curves of BMI SD scores for prediction of overweight and obesity in boys and girls on the basis of the 85th and 95th percentiles of percentage of body fat (%BF), respectively, in a national sample of 6–12-y-old Swiss children1
View this table:
TABLE 5. Sensitivity and specificity of the age- and sex-specific BMI reference values for overweight and obesity from the International Obesity Task Force (IOTF) and the US Centers for Disease Control and Prevention (CDC) in a national sample of 6–12-y-old Swiss children1
DISCUSSION
Although measurement of BMI is practical and reproducible, the correlation coefficient between BMI and %BF by DXA or densitometry in children varies between 0.4 and 0.9 according to age, ethnicity, and sex (27, 28). We measured adiposity by using multisite SFTs to judge the performance of BMI as an indicator of overweight in our sample. Studies have shown that %BF values calculated from SFTs have high reproducibility (18) and correlate well with %BF values measured by DXA in children (13, 20). Using ROC curve analysis to compare the accuracy of SFTs and BMI with that of DXA in 10–15-y-old children, Sardinha et al (20) reported that the AUC for SFTs was equal to or greater than the AUC for BMI.
We found a strong and age-independent association between BMI and %BF calculated from SFTs. By regression, 74% of the variability in %BF was explained by BMI in both the boys and the girls. The boys and the girls differed significantly in the slope of the regression (P < 0.001). However, this difference appeared to be physiologically irrelevant because the use of a common slope of 0.374 for the boys and the girls in the regression equation left the residual SE practically unchanged at 0.20 (Table 3). The areas under the ROC curves for the girls and the boys were 0.956 and 0.967, respectively, for overweight (NS) and 0.970 and 0.992, respectively, for obesity (P < 0.001; Table 4). This suggests that the accuracy of BMI in predicting adiposity was greater in the boys than in the girls. In 6–11-y-old US children in the third National Health and Nutrition Examination Survey (NHANES III), the correlation coefficients between BMI-for-age and the average of the triceps and subscapular SFTs in boys and girls were 0.88 and 0.85, respectively (29). Mei et al (29) determined the performance of area under the ROC curve of BMI-for-age as defined by the average of the triceps and subscapular SFTs at the cutoffs for overweight (>85th percentile) from the NHANES III. For children aged 6–11 y, the mean AUC was 0.973, which is similar to the value obtained in the present study.
In our sample, the CDC and IOTF reference values for overweight showed fairly high sensitivity and high specificity in both sexes (Table 5). The CDC and IOTF cutoffs for the boys and the girls were close together and were well placed on the bend of the ROC curve (Figure 2). The CDC reference value for obesity had a higher sensitivity and specificity than did the IOTF reference. The sensitivity of the IOTF reference value for obesity was poor, and the false negative rate was 38% for the boys and 52% for the girls. This was reflected in the better position of the CDC reference values on the ROC curve for obesity (Figure 3). Reilly et al (7) compared the sensitivity and specificity of the 1990 UK reference values with those of the IOTF reference values for detecting adiposity (>95th percentile for %BF) measured by bioelectrical impedance in 7-y-old children in the United Kingdom. The sensitivity of the IOTF reference values was low and differed significantly between boys (46%) and girls (72%). Flegal et al (8) used the new CDC and IOTF criteria to compare the prevalence of overweight and obesity in 6–11-y-old US children in the NHANES III (1988–1994). Compared with the CDC criteria, the IOTF criteria gave lower prevalence estimates for overweight and obesity in boys and for obesity in girls. The differences in prevalence were not systematic, and some were large, up to 10% for overweight and up to 50% for obesity. Kain et al (9) reported that in 6-y-old Chilean children, the CDC and IOTF criteria generated comparable prevalence estimates for overweight, but the IOTF reference value for obesity generated an 50% lower prevalence estimate than did the CDC reference value.
The CDC and IOTF BMI criteria were generated by using different data sets and smoothing methods, and their approaches to setting cutoffs were different (8). The CDC criteria were based on the BMI distribution of representative samples of US children (6). The IOTF criteria, on the other hand, were not related to a reference population distribution; they were instead extrapolated from adult BMI cutoffs for overweight and obesity and are based on the assumption that children with those BMI values have inherent health risks (5). For ages 6–12 y, the IOTF BMI cutoffs are generally higher than are the CDC reference values. For boys, the mean differences between the CDC and IOTF reference values for overweight and obesity are 0.5 and 1.5–2.0 BMI units, respectively. For girls, the 2 sets of reference values are similar for overweight, but the IOTF cutoffs are 1.0 BMI unit higher (8). These differences explain both the lower prevalence estimate that was obtained in our sample with the IOTF reference values for overweight in boys than with the CDC reference values and the sharply lower prevalence estimate for obesity in boys and girls that was obtained with the IOTF reference values (Table 2).
Our data indicate that BMI is an excellent proxy measure of adiposity in 6–12-y-old Swiss children. Although both the IOTF and CDC age- and sex-specific BMI criteria accurately predict overweight, the IOTF criteria for obesity are insensitive and failed to identify 40–50% of obese children in our sample. Overall, the performance of the CDC reference values was superior, and they provided more accurate estimates of adiposity. Although the IOTF reference values have been proposed for international use, before they are widely adopted to detect childhood adiposity, their validity should first be tested in other countries around the world.
ACKNOWLEDGMENTS
We thank the teachers and children at the participating schools for their cooperation and P Ballmer (Canton Hospital Winterthur, Switzerland) for technical advice.
Each of the authors made substantial contributions to the study design, data collection, data analyses, and the writing or editing of the manuscript. None of the authors had any personal or financial interests, including advisory board affiliations, in the companies or organizations sponsoring this research.
REFERENCES