SEARCH

SEARCH BY CITATION

Keywords:

  • BMI;
  • ethnic groups;
  • childhood obesity;
  • epidemiology;
  • gender

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Research Methods and Procedures
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References

Objective: The objective was to assess the predictive value of weight-for-age to identify overweight children and adolescents in the unusual research or public health situations where height is not available to calculate BMI.

Research Methods and Procedures: Data from the National Health and Nutrition Examination Survey 1999 to 2004 were used to calculate the sensitivity, specificity, and positive and negative predictive values of selected weight-for-age cut-off points to identify overweight children and adolescents (as defined by BMI ≥95th percentile). Positive and negative predictive values are dependent on prevalence and are reported here for this study population only.

Results: The 50th and 75th weight-for-age percentiles had good sensitivity (100% and 99.6%, respectively), but poor positive predictive value (23.7% and 37.0%, respectively), while the 95th and 97th percentiles had reasonable positive predictive value (80.3% and 91.5%, respectively), but limited sensitivity (82.0% and 66.7%, respectively) to identify overweight subjects. The properties of weight-for-age percentiles to identify overweight subjects differed between sex, age, and race/ethnicity but remain within a relatively narrow range.

Discussion: No single weight-for-age cut-off point was found to identify overweight children and adolescents with acceptable values for all properties and, therefore, cannot be used in the clinical setting. Furthermore, the positive predictive values reported here may be lower in populations with a lower prevalence of obesity. However, in unusual research or public health situations where height is not available, such as existing databases, weight-for-age percentiles may be useful to target limited resources to groups more likely to include overweight children and adolescents than the general population.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Research Methods and Procedures
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References

The definitions of obesity and overweight in children and adolescents are complex and have resulted in much confusion and controversy (1)(2). Ideally, these definitions should be based on their ability to predict morbidity and mortality (3). However, in the cases of children and adolescents, this approach is impractical, as most obesity comorbidities develop later in life. Even though the long-term health consequences of obesity among adolescents have been well documented (4)(5), most definitions are not based on the predictive value of childhood weight status. The next best option would be to use measurements of adiposity to classify children and adolescents as obese, but these measurements either require extensive training (e.g., skinfold thickness) or expensive tools that are sometimes difficult to transport to the field (e.g., DXA, air displacement plethysmography, bioimpedance); thus, although reference data are available, these are difficult measurements to obtain accurately and are not commonly used. Furthermore, no clear age- and sex-specific adiposity cut-off points are widely recognized to define pediatric obesity. Therefore, BMI is widely recognized as an appropriate tool to screen obese children and adolescents and to define overweight status, as a state of excessive weight relative to height, regardless of body composition (1)(2)(6)(7). Because BMI does not measure body composition and adiposity, it only is a screening tool to identify obesity, as defined by excessive adiposity (8). BMI cut-off points for sex and age have been defined either using a reference population (U.S. Centers for Disease Control and Prevention definitions) (9) or by extrapolation of adult BMI values associated with comorbidities (International Obesity Task Force definitions) (10).

BMI is easy to obtain, as it requires only measurement of weight and height, and BMI-for-age percentile curves are widely available on most current national growth charts. However, in some unusual circumstances, height is not routinely measured and, therefore, BMI cannot be calculated. For example, in most emergency departments, weight is measured but not height. Similarly, some clinical or survey databases only include weight. In telephone surveys, children's weight can easily be measured by parents, but height measurements are more difficult to obtain. These situations are problematic for researchers or public health officials who need to identify obese subjects in large populations. Furthermore, reliable height measurement requires more training (3) than weight measurement and may not be easily obtained under circumstances where resources are limited and when the goal is only to identify large numbers of obese children and adolescents for research or public health purposes. In these circumstances where height is not available, the question arises of whether weight-for-age cut-off points can be used to predict the likelihood of identifying overweight pediatric subjects.

Low weight-for-age is a reasonable assessment of under-nutrition, as children with either acute (low weight-for-height) or chronic (low height-for-age) under-nutrition are likely to have a low weight-for-age. However, acute and chronic under-nutrition cannot be differentiated by weight-for-age alone and assessments of height-for-age and weight-for-height are needed to distinguish these two conditions. Lower percentile weight-for-age cut-off points have been recommended by the World Health Organization to identify under-nutrition in some situations (7). However, less is known about the predictive value of weight-for-age at the upper end of the spectrum. Many, but not all, overweight children are tall for their age because of advanced maturity and puberty; and in countries in nutritional transition, overweight is actually associated with short stature for age (11).

The primary aim of this study was, therefore, to assess the predictive value of weight-for-age to identify population subgroups with large numbers of overweight children and adolescents (as defined by a BMI ≥95th percentile) (9) using a representative sample of U.S. children and adolescents. Sensitivity and specificity, as well as positive and negative predictive values of selected weight-for-age percentile cut-off points, were determined. The secondary aim was to assess if the predictive value of these cut-off points differed between sex, age, and race/ethnicity. The aim of this study was not to assess if weight-for-age cut-off points could replace BMI-for-age as a screening tool for obesity, as it is widely recognized that BMI-for-age would perform better than weight-for-age to assess adiposity, but rather to provide researchers and public health professionals with screening tools to identify subpopulations likely to contain a high proportion of overweight children and adolescents (with high BMI), in situations where height measurement is not available. Therefore, we chose to use the Centers for Disease Control definition of overweight, based on BMI, rather than a definition of obesity based on measurements of adiposity for primary analyses. In secondary analyses, however, the predictive values of weight-for-age cut-off points for excess subcutaneous adiposity were also calculated.

Research Methods and Procedures

  1. Top of page
  2. Abstract
  3. Introduction
  4. Research Methods and Procedures
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References

Cross-sectional data from the most recent National Health and Nutrition Examination Survey (NHANES)1 from 1999 to 2004 were used (12). Briefly, NHANES is a continuous survey that uses a complex sampling process to generate data representative of the entire U.S. non-institutionalized population. Children and adolescents, as well as black and Mexican-American subjects, are over-sampled to enable analyses among these population subgroups.

From 1999 to 2004, weight, height, triceps skinfold, subscapular skinfold, age, sex, and race/ethnicity were assessed in 12,382 children and adolescents between the ages of 2 and 19 years using standardized protocols and equipment. For the present study, the numbers of subjects by sex, age groups, and racial/ethnic groups are reported in Table 1. Weight was measured using a Toledo digital scale (Mettler-Toledo, Inc., Columbus, OH) with subjects wearing only underwear, disposable paper gowns, and foam slippers. Standing height was measured with a fixed Seca electronic stadiometer (Seca GmbH & Co., Hamburg, Germany) with a vertical backboard and a moveable headboard (12). Skinfold thickness was measured using a Holtain skinfold caliper (Holtain, Crymych, U.K.). Self-identification of ethnicity and race resulted in 4 categories: Mexican-American, non-Hispanic white, non-Hispanic black, and other. Mexican-American, non-Hispanic white, and non-Hispanic black constituted the three racial/ethnic groups large enough to conduct subgroup analyses.

Table 1. . Number of subjects and prevalence of overweight status (BMI at or above the 95th percentile) by sex, age groups, and racial/ethnic groups
 Number of subjectsPrevalence of overweight status (%)
 MaleFemaleTotalMaleFemaleTotal
  1. —, survey not designed to estimate reliable prevalence in these subgroups.

Age group (yrs)      
 2 to 511611177233811.811.011.4
 6 to 1115871608319517.815.816.8
 12 to 1934563393684317.316.116.7
Racial/ethnic group      
 Non-Hispanic white16381650328814.713.213.9
 Non-Hispanic black19831899388216.121.818.9
 Mexican-American20652081414622.716.719.7
 Other/multiple5185481066
Total6204617812,38216.314.915.7

All analyses were conducted with Stata 8.2 (StataCorp LP., College Station, TX) using the survey commands (svy) to account for non-response and the complex sampling process used for NHANES. A p value <0.05 was considered statistically significant. BMI was calculated as weight in kilograms divided by height in meters squared. Subjects with a BMI at or above the sex- and age-specific 95th percentile of the Centers for Disease Control growth charts were considered overweight (9). The sum of triceps and subscapular skinfolds was calculated, and overfat status was defined as a sum of skinfolds at or above the sex- and age-specific 95th percentile of a U.S. representative population measured before the obesity epidemic (13), similar to the population used to derive the Centers for Disease Control BMI growth charts. Height-for-age, weight-for-age, and BMI-for-age z-scores were also calculated using Stata “zanthro” command and averaged. The following sex- and age-specific weight-for-age cut-off points were considered, as they represent the percentiles seen on various versions of the Centers for Disease Control growth charts: 3rd, 5th, 10th, 25th, 50th, 75th, 90th, 95th, and 97th percentiles of the reference population used to derive these growth charts (9). For each weight-for-age cut-off point, the following characteristics were calculated, with 95% confidence intervals: prevalence of subjects at or above this cut-off percentile, based on the NHANES 1999 to 2004 survey; sensitivity, as the probability of a weight at or above the weight-for-age cut-off point among overweight subjects; specificity, as the probability of a weight below the weight-for-age cut-off point among non-overweight subjects; positive predictive value (PPV), as the probability of being overweight among subjects with a weight at or above the weight-for-age cut-off point; and negative predictive value (NPV), as the probability of being non-overweight among subjects with a weight below the weight-for-age cut-off point (14). It should be noted that sensitivity and specificity are independent of the prevalence of overweight in the population studied, while prevalence of subjects at or above the weight-for-age cut-off percentile, PPV, and NPV are dependent on the prevalence of overweight in the study population and are reported here for the NHANES 1999 to 2004 study sample. The PPV and NPV, however, can easily be calculated in a different population with known prevalence of overweight using Bayes’ rule (14). For selected weight-for-age cut-off points, prevalence at or above the cut-off point, sensitivity, specificity, PPV, and NPV were examined for homogeneity between sexes, and among age groups, and racial/ethnic groups, using the χ2 test. The analyses were repeated using overfat status instead of overweight status as the criterion method.

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Research Methods and Procedures
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References

The prevalence of overweight status is reported in Table 1. Overall, 15.7% of the subjects had a BMI at or at or above the 95th percentile (overweight), while only 7.2% had a sum of skinfold thickness at or above the 95th percentile (overfat). On average (95% confidence interval) in this sample, height-for-age z-score was 0.18 (0.15 to 0.21), weight-for-age z-score was 0.46 (0.42 to 0.50), and BMI-for-age z-score was 0.44 (0.40 to 0.48). For each weight-for-age cut-off point, prevalence at or above the cut-off point, sensitivity, specificity, PPV, and NPV for overweight or overfat status are reported in Table 2 for the entire sample of 12,382 subjects from the NHANES 1999 to 2004 survey. No single weight-for-age cut-off point was found to identify overweight or overfat children and adolescents with acceptable values for all indicators of accuracy that we examined. However, based on these results, and depending on the goal (high sensitivity, high specificity, high PPV, or high NPV), the weight-for-age percentile cut-offs most likely to be used to identify population subgroups with large numbers of overweight children and adolescents are the 90th, 95th, and 97th percentiles. For these percentile cut-offs, comparisons of prevalence at or above the cut-off point, sensitivity, specificity, PPV, and NPV between sexes, age groups, and racial/ethnic groups are reported in Table 3.

Table 2. . For each weight-for-age percentile cut-off point, prevalence of subjects at or above the cut-off, sensitivity, specificity, and PPV and NPV, with 95% CIs, for overweight (BMI at or above the 95th percentile) or overfat (triceps plus subscapular skinfolds at or above the 95th percentile) status in a representative sample of U.S. children and adolescents from 1999 to 2004
Weight-for-age percentile cut-off pointPrevalence at or above percentile cut-off point (95% CI)Overweight statusOverfat status
  Sensitivity (95% CI)Specificity (95% CI)PPV (95% CI)NPV (95% CI)Sensitivity (95% CI)Specificity (95% CI)PPV (95% CI)NPV (95% CI)
  1. PPV, positive predictive value; NPV, negative predictive value; CI, confidence interval. Sensitivity is calculated as the probability of a weight at or above the weight-for-age cut-off point among overweight or overfat subjects; specificity, the probability of a weight below the weight-for-age cut-off point among non-overweight or non-overfat subjects; positive predictive value, the probability of being overweight or overfat among subjects with a weight at or above the weight-for-age cut-off point; negative predictive value, the probability of being non-overweight or non-overfat among subjects with a weight below the weight-for-age cut-off point.

397.7 (97.3 to 98.2)100 (NA)2.7 (2.4 to 3.2)16.0 (14.8 to 17.3)100 (NA)100 (NA)2.6 (2.2 to 3.0)7.4 (6.5 to 8.0)100 (NA)
596.5 (95.9 to 97.0)100 (NA)4.2 (3.6 to 4.8)16.2 (15.0 to 17.6)100 (NA)100 (NA)4.0 (3.4 to 4.6)7.5 (6.8 to 8.3)100 (NA)
1093.6 (92.8 to 94.4)100 (NA)7.5 (6.7 to 8.5)16.7 (15.5 to 18.1)100 (NA)100 (NA)7.3 (6.4 to 8.2)7.8 (7.0 to 8.6)100 (NA)
2584.2 (83.1 to 85.3)100 (NA)18.7 (17.5 to 20.0)18.6 (17.2 to 20.0)100 (NA)100 (NA)17.9 (16.7 to 19.2)8.7 (7.9 to 9.6)100 (NA)
5066.1 (64.6 to 67.6)100 (99.8 to 100)40.2 (38.7 to 41.7)23.7 (22.1 to 25.4)100 (99.9 to 100)99.7 (98.7 to 100)38.1 (36.5 to 39.6)11.2 (10.1 to 12.3)100 (99.7 to 100)
7542.1 (40.5 to 43.8)99.6 (99.1 to 99.8)68.5 (67.1 to 69.9)37.0 (35.0 to 39.1)99.9 (99.8 to 100)97.9 (95.9 to 99.0)64.7 (63.1 to 66.3)17.8 (16.2 to 19.5)99.8 (99.5 to 99.9)
9023.9 (22.3 to 25.5)95.5 (94.2 to 96.6)89.4 (88.3 to 90.4)62.6 (60.1 to 65.1)99.1 (98.8 to 99.3)90.4 (86.9 to 93.0)84.2 (82.9 to 85.4)30.8 (28.4 to 33.3)99.1 (98.7 to 99.4)
9516.0 (14.9 to 17.2)82.0 (79.7 to 84.1)96.3 (95.7 to 96.8)80.3 (77.7 to 82.7)96.6 (96.1 to 97.2)78.2 (73.6 to 82.1)91.7 (90.9 to 92.4)42.2 (39.4 to 45.1)98.2 (97.6 to 98.6)
9711.4 (10.4 to 12.6)66.7 (63.2 to 70.0)98.8 (98.4 to 99.1)91.5 (88.9 to 93.5)94.1 (93.3 to 94.8)65.6 (60.2 to 70.7)95.4 (94.7 to 96.1)52.8 (49.1 to 56.5)97.3 (96.6 to 97.8)
Table 3. . For selected weight-for-age percentile cut-off points, comparison of prevalence of subjects at or above the cut-off, sensitivity, specificity, and PPV and NPV for overweight (BMI at or above the 95th percentile) or overfat (triceps plus subscapular skinfolds at or above the 95th percentile) status among sexes, age groups, and racial/ethnic groups, based on a representative sample of U.S children and adolescents in 1999 to 2004
 Prevalence at or above percentile cut-off pointOverweight statusOverfat status
  SensitivitySpecificityPPVNPVSensitivitySpecificityPPVNPV
  1. PPV, positive predictive value; NPV, negative predictive value. Sensitivity is calculated as the probability of a weight at or above the weight-for-age cut-off point among overweight or overfat subjects; specificity, the probability of a weight below the weight-for-age cut-off point among non-overweight or non-overfat subjects; positive predictive value, the probability of being overweight or overfat among subjects with a weight at or above the weight-for-age cut-off point; negative predictive value, the probability of being non-overweight or non-overfat among subjects with a weight below the weight-for-age cut-off point.

Weight-for-age 90th percentile cut-off         
 Sex         
  Male25.294.688.361.398.889.483.435.198.7
  Female22.496.590.564.199.392.285.025.199.5
  p0.010.130.020.20.060.40.12<0.0010.02
 Age groups (yrs)         
  2 to 518.088.490.955.598.484.987.735.698.6
  6 to 1125.496.188.963.699.192.783.035.999.1
  12 to 1925.397.389.164.199.490.783.424.399.4
  p<0.001<0.0010.170.040.040.20.003<0.0010.3
 Race/ethnicity         
  Non-Hispanic white22.897.389.259.399.590.384.328.699.2
  Non-Hispanic black28.496.487.464.099.095.681.128.599.6
  Mexican-American25.892.890.771.098.187.783.736.898.4
  p0.0010.010.0090.0020.0030.020.030.010.02
Weight-for-age 95th percentile cut-off         
 Sex         
  Male17.080.995.577.896.278.291.347.597.7
  Female14.983.297.183.397.178.092.034.998.7
  p0.030.290.0070.030.081.00.3<0.0010.01
 Age groups (yrs)         
  2 to 510.267.297.174.895.970.394.650.997.6
  6 to 1117.282.495.980.396.480.290.847.397.8
  12 to 1917.686.196.181.797.280.391.034.398.7
  p<0.001<0.0010.270.200.100.2<0.001<0.0010.04
 Race/ethnicity         
  Non-Hispanic white15.183.195.976.697.278.291.639.498.4
  Non-Hispanic black20.087.895.782.597.189.289.840.799.1
  Mexican-American17.075.597.387.294.272.791.748.596.9
  p<0.001<0.0010.050.01<0.001<0.0010.010.02<0.001
Weight-for-age 97th percentile cut-off         
 Sex         
  Male12.367.898.690.294.066.795.459.296.6
  Female10.565.599.192.994.263.595.543.598.0
  p0.010.450.090.230.740.60.9<0.0010.007
 Age groups (yrs)         
  2 to 57.354.398.784.494.454.899.656.396.4
  6 to 1112.466.998.690.793.769.395.058.996.8
  12 to 1912.570.399.193.894.367.495.245.198.0
  p<0.0010.020.340.020.550.160.040.010.03
 Race/ethnicity         
  Non-Hispanic White10.566.898.789.394.863.395.549.797.4
  Non-Hispanic Black15.074.098.692.794.276.193.949.798.0
  Mexican-American12.359.899.496.091.061.695.559.495.8
  p<0.001<0.0010.0050.009<0.0010.0090.0040.020.003

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Research Methods and Procedures
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References

In this study, no single weight-for-age percentile cut-off point had acceptable values to identify clinically overweight children and adolescents for all predictive values. This is not surprising, considering that tall children can have an elevated weight-for-age without being overweight and that short children can have a low weight-for-age while still being overweight. This finding reinforces the recommendation to measure height systematically to calculate, plot, and track BMI in the routine pediatric healthcare setting, as recommended by the Institute of Medicine and the Centers for Disease Control (2)(9). Outside of the clinical setting, in particular in the research, epidemiological, and public health setting, some weight-for-age percentile cut-off points may, however, be useful to identify groups of children with a high prevalence of overweight. The implications of our findings for emergency care, where height is typically not measured, or for telephone surveys are more difficult to define.

This study shows that for situations, mostly outside of the clinical setting, in which height cannot be readily measured, eliminating the possibility of calculating BMI, some of the cut-off points of weight-for-age can be used as initial screening tools to identify subgroups of children who are likely to be overweight and who should be evaluated through formal assessment of BMI. For example, if the goal is to ensure that most overweight children are identified (high sensitivity), even if the specificity is low and many non-overweight children are included, the 90th weight-for-age percentile may provide a useful cut-off point. In contrast, if the goal is to ensure that children who are identified using weight-for-age are very likely to be overweight, i.e., high PPV, even if the sensitivity is low and many overweight children are not included, the 95th and 97th percentiles may provide useful cut-off points. However, as PPV and NPV are dependent on the prevalence of overweight in the population under study, these conclusions may not be valid in every population.

Even within the U.S. population of children and adolescents, the predictive value of various weight-for-age percentile cut-off points may differ significantly. Therefore, the predictive values of the 3 selected percentile cut-offs were compared between subgroups of the U.S. pediatric population. The weight-for-age percentile cut-offs exhibited similar characteristics between the subgroups. Even if some of the subgroup comparisons were statistically significant, due to the large sample size, the between-group differences in sensitivity, specificity, PPV, and NPV were relatively small within each weight-for-age percentile cut-off, in general <10% between extreme values. Furthermore, with few exceptions in the extreme values, the distribution of sensitivity, specificity, PPV, and NPV between subgroups within a weight-for-age percentile cut-off point did not overlap with the subgroup distribution within the next percentile cut-off. Therefore, the choice of weight-for-age percentile cut-offs for various uses is unlikely to differ between the population subgroups investigated here. For example, the PPV of the 95th weight-for-age percentile cut-offs was above 74%, and the PPV of the 97th percentile above 84% in every subgroup studied. As the largest differences in predictive values were observed between racial/ethnic groups, our findings may not apply to populations with different racial/ethnic composition from the U.S. population. As expected in this contemporary sample of U.S. children and adolescents, average height-for-age, weight-for-age, and BMI-for-age z-scores were above 0, reflecting the secular trend in height, early maturation, and the obesity epidemic.

In general, when using overfat status instead of overweight status as criterion variable to assess the predictive value of weight-for-age percentiles, the sensitivity and specificity were lower, but in a similar range, with a maximum difference of 7% between the two analyses. As expected, because the prevalence of overfat status was about half the prevalence of overweight status, the PPV of weight-for-age percentiles was much lower and the NPV much higher for overfat status than for overweight status.

This study had some limitations. For example, other racial/ethnic population subgroups were not studied. The sample size for other racial/ethnic groups is too small in NHANES to provide meaningful results. Analyses by subgroup of height status, i.e., comparing the predictive values for short and tall children, were not performed, as the premise of this study was to examine use of weight-for-age in the absence of height measurement. Although BMI is a good tool to identify overweight children and adolescents, definitions based on BMI alone clearly are not gold standards because, in the pediatric age, body composition can vary greatly at similar BMI levels (15). Using the sum of triceps and subscapular skinfolds at or above the 95th percentile of a reference population in secondary analyses provided similar results as using the Centers for Disease Control definition of overweight. However, this definition of overfat status also has important limitations, as it accounts only for subcutaneous fat, as skinfold thickness can be difficult to measure in obese individuals, and as no cut-off points are widely recognized. Furthermore, unlike the Centers for Disease Control BMI growth charts, the available skinfold thickness percentiles that were used in the present study have not been statistically smoothed (13), which may have resulted in misclassification. These limitations highlight the need for a clear and widely recognized definition of excess adiposity in the pediatric age, against which other assessment methods can be compared. This study also had unique strengths, such as the large nationally representative sample, rigorous measurement methods, including measured rather than self-reported weight and height, and the possibility of comparing various subgroups of the U.S. pediatric population.

In summary, weight-for-age percentiles cannot be used as surrogates for BMI-for-age percentiles to screen for overweight among children and adolescents. In the circumstances where height is not available, such as existing databases, some of the weight-for-age percentile cut-off points can be used as screening tools to identify population subgroups more likely to be overweight. These cut-off points may be useful in research or public health projects to identify large numbers of overweight subjects but should not be used clinically in individual patients.

Acknowledgments

  1. Top of page
  2. Abstract
  3. Introduction
  4. Research Methods and Procedures
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References

This study was supported in part by NIH Grant K23RR16073 and the Department of Pediatrics of the University of Pennsylvania, Children's Hospital of Philadelphia (to N.S.).

Footnotes
  • 1

    Nonstandard abbreviations: NHANES, National Health and Nutrition Examination Survey; PPV, positive predictive value; NPV, negative predictive value.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Research Methods and Procedures
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References
  • 1
    World Health Organization (2000) WHO Consultation on Obesity (1999). Obesity: Preventing and Managing the Global Epidemic. Report of a WHO consultation World Health Organization Geneva, Switzerland. [WHO Technical Report Series 894].
  • 2
    Koplan, J., Liverman, C. T., Kraak, VI. (2005) Preventing Childhood Obesity: Health in the Balance Committee on Prevention of Obesity in Children and Youth, National Academy Press, U.S. Institute of Medicine Washington, DC.
  • 3
    Cole, T., Rolland-Cachera, MF. (2002) Measurement and definition Burniat, W. eds.. Child and Adolescent Obesity: Causes and Consequences, Prevention and Management 327. Cambridge University Press Cambridge U.K..
  • 4
    Freedman, D., Serdula, M., Khan, L. (2002) The adult health consequences of childhood obesity Chen, C. Dietz, WH. eds.. Obesity in Childhood and Adolescence 296 Lippincott Williams & Wilkins Philadelphia, PA.
  • 5
    Must, A., Jacques, P. F., Dallal, G. E., Bajema, C. J., Dietz, WH. (1992) Long-term morbidity and mortality of overweight adolescents: a follow-up of the Harvard Growth Study of 1922 to 1935. N Engl J Med. 327: 13501355.
  • 6
    Lobstein, T., Baur, L., Uauy, R. (2004) Obesity in children and young people: a crisis in public health. Obes Rev. 5: (Suppl 1) 485.
  • 7
    WHO Expert Committee on Physical Status (1995) Physical Status: The Use and Interpretation of Anthropometry. Report of a WHO Expert Committee World Health Organization Geneva, Switzerland.
  • 8
    Mei, Z., Grummer-Strawn, L. M., Pietrobelli, A., Goulding, A., Goran, M. I., Dietz, WH. (2002) Validity of body mass index compared with other body-composition screening indexes for the assessment of body fatness in children and adolescents. Am J Clin Nutr. 75: 978985.
  • 9
    Kuczmarski, R. J., Ogden, C. L., Grummer-Strawn, L. M., et al. (2000) CDC growth charts: United States. Advance Data 314: 127.
  • 10
    Cole, T. J., Bellizzi, M. C., Flegal, K. M., Dietz, WH. (2000) Establishing a standard definition for child overweight and obesity worldwide: international survey. BMJ 320: 12401243.
  • 11
    Popkin, B. M., Richards, M. K., Montiero, CA. (1996) Stunting is associated with overweight in children of four nations that are undergoing the nutrition transition. J Nutr. 126: 30093016.
  • 12
    National Health and Nutrition Examination Survey Data Sets and Related Documentation http:www.cdc.govnchsaboutmajornhanesdatalink.htm (Accessed November 28, 2006).
  • 13
    Frisancho, AR. (1990) Anthropometric Standards for the Assessment of Growth and Nutritional Status University of Michigan Press Ann Arbor, MI.
  • 14
    Rosner, B. (2006) Fundamentals of Biostatistics 6th ed. Thomson-Brooks/Cole Belmont, CA.
  • 15
    Ellis, K. J., Abrams, S. A., Wong, WW. (1999) Monitoring childhood obesity: assessment of the weight/height index. Am J Epidemiol. 150: 939946.