| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Original Research Communication |
1 From the National Center for Health Statistics, Centers for Disease Control and Prevention, Hyattsville, MD.
2 KM Flegal, National Center for Health Statistics, 6525 Belcrest Road, Room 900, Hyattsville, MD 20782. E-mail: kflegal{at}cdc.gov.
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
Objective: The objective of this study was to compare BMI-for-age with weight-for-stature.
Design: We used data for 4348 children (aged 2–5 y) from the third National Health and Nutrition Examination Survey. Weight-for-stature and BMI-for-age percentiles were calculated for each child. The 10th and 85th percentiles of weight-for-stature at selected ages were also reexpressed as BMI-for-age percentiles.
Results: More than 63% of children had lower weight-for-stature than BMI-for-age percentiles. Children were more likely to be classified as 
10th percentile by weight-for-stature than by BMI-for-age, but less likely to be classified as 
85th percentile. Differences in classification by the 2 measures varied with age and stature and were greater for shorter children. The 10th and 85th percentiles of weight-for-stature corresponded to BMI-for-age percentiles from the 3rd to the 21st percentile and from the 74th to the 92nd percentile, respectively, depending on age and stature.
Conclusions: Weight-for-stature is easier to use than BMI-for-age. However, BMI-for-age captures changes in the weight-height relation with age and can be used continuously up to the age of 20 y. BMI-for-age is recommended in most situations. BMI-for-age and weight-for-stature will not give identical results and are not interchangeable.
Key Words: Body weight • body mass index • children • growth charts • health surveys • overweight • underweight • weight-for-stature • third National Health and Nutrition Examination Survey
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
BMI, calculated as weight (kg)/height (m) squared, is a weight-height index that in effect is a method of adjusting weight for height (stature) (3). BMI is a convenient index to use for adults because it allows comparisons of weight between adults of different statures. BMI standards for adults are not age specific. For children, however, the distribution of BMI varies by age. As a consequence, BMI reference data for children are age specific (2). BMI does not increase monotonically with age at younger ages. Beginning at the age of 2 y, BMI tends to first fall and then rise again (4). The BMI-for-age reference data from the CDC growth charts can be used to compare a child's BMI with the BMI distribution of a reference group of children of the same age but not necessarily of the same stature.
Weight-for-stature is another way of expressing weight relative to height. In the CDC growth charts, weight-for-stature reference data are not age specific. However, the limited range of statures means that for practical purposes the use of weight-for-stature reference data is limited to younger children. The weight-for-stature reference data from the CDC growth charts can be used to compare a child's weight with the weight distribution of a reference group of children of the same stature but not necessarily of the same age.
Thus, either BMI-for-age or weight-for-stature growth charts can be used to compare a child's weight with a reference population of children. The objective of this study was to compare the results of applying both sets of reference data to the same group of children with the use of data from the third National Health and Nutrition Examination Survey (NHANES III) (5).
|
METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
The revised growth charts include not only selected percentiles but also 3 parameters for each month of age or centimeter of stature, a skewness parameter (L), the mean (M), and the SD (S), from which any percentile can be calculated exactly. To calculate a z score for a given measurement X by using the LMS parameters, the following equation is used:
 | (1) |
This provides a z score or an SD score that can be referred to a standard normal distribution with a mean of zero and a variance of 1 to ascertain the corresponding percentile.
NHANES data set
NHANES III was conducted from 1988 to 1994 by the NCHS of the CDC. A nationally representative sample of the US civilian noninstitutionalized population was selected by using a complex, stratified, multistage probability cluster sampling design. A home interview was conducted followed by a physical examination performed in a mobile examination center. A description of the plan and operation of the survey was published previously (5).
Age was expressed as age in months at the time of examination, with age in months truncated to the last full month. Weight and stature were measured by using standardized techniques and equipment (6). BMI was calculated as weight (kg)/stature (m) squared. For the present study, we used data from 4348 children aged 2 y, 0 mo, through 5 y, 11 mo, for whom measured weight and stature data were available. All percentiles were calculated by using the nearest month of age. To calculate summary statistics, children were grouped according to their age at their last birthday.
Statistical analysis
Statistical analyses were conducted by using SAS software (7). All analyses of NHANES III data incorporated sampling weights that took into account differential probabilities of selection and nonresponse. A weight-for-stature percentile and a BMI-for-age percentile were calculated for each child in the NHANES III data set. The difference between these 2 percentiles for each child and the distribution of these differences were calculated. Children were also classified according to the 10th and 85th percentiles of weight-for-stature and BMI-for-age, and the classifications by the 2 methods were compared. These 10th and 85th percentile cutoffs were chosen because they correspond to standards used to identify children at potential nutritional risk in the Special Supplemental Nutrition Program for Women, Infants, and Children (WIC) (8).
Direct comparisons between the BMI-for-age and weight-for-stature reference values were also conducted. The 85th percentile weight-for-stature values were converted into the equivalent BMI-for-age percentiles for ages 24, 36, 48, and 60 mo. This was done for a given stature H in the following way. First, the 85th percentile of weight W for stature H was found. A BMI value corresponding to these weight and stature values was calculated as W/H2. Equation 1
was used with the LMS parameters for BMI-for-age at each specified age to find a value of z corresponding to this BMI value. Finally, the z score was converted to the corresponding BMI-for-age percentile for the selected ages.
For example (with some rounding to simplify), at a stature of 100.5 cm, the 85th percentile of weight-for-stature corresponds to a weight of 17.27 kg for boys. These values of weight (17.27 kg) and stature (100.5 cm) can be used to calculate a BMI of 17.1. For boys aged 36 mo (36.0–36.9 mo), if equation 1 is applied with the tabulated age-specific BMI-for-age L, M, and S parameters (L = –1.42, M = 16.0, and S = 0.0726), with X equal to the BMI value of 17.1, the resulting z score is 0.874. This z score corresponds to the 81st percentile of a standard normal distribution. Thus, the 85th percentile of weight-for-stature at a stature of 100.5 cm is equivalent to the 81st percentile of BMI-for-age for boys aged 36 mo.
This process was repeated for each 1-cm increment of stature and for the ages of 24, 36, 48, and 60 mo. For each age, the stature range used was from the 3rd percentile to the 97th percentile of stature for that age. Because BMI varies with age, identical values of W and H will represent different BMI-for-age percentiles at different ages. A similar process was conducted for the 10th percentile of weight-for-stature.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
. Because the weight-for-stature charts are available for only a selected range of statures, a weight-for-stature percentile cannot be calculated for children who are taller or shorter than the selected range. The number and proportion of children at each age for whom a weight-for-stature percentile could not be calculated are also shown in Table 1. Approximately 5% of children aged 5 y were too tall to have a weight-for-stature percentile calculated.
|
0.98). However, the correlation coefficient gives only the degree of linear relation between 2 variables, not the degree of identity, and thus cannot by itself establish the interchangeability of the 2 charts.
There was some shifting between percentiles of BMI-for-age and weight-for-stature. We calculated the differences between the weight-for-stature percentile and the BMI-for-age percentile. For example, if a child was at the 75th percentile of weight-for-stature and at the 80th percentile of BMI-for-age, the difference would be -5.0 percentage points. The distribution of differences between the weight-for-stature and BMI-for-age percentiles is shown in Table 2. Overall, the weight-for-stature and BMI-for-age percentiles agreed to within 2.5 percentage points for approximately one-third (31.2%) of children. A small proportion of children (5.3%) had weight-for-stature percentiles >2.5 percentage points higher than their BMI-for-age percentiles. However, a high proportion of children (63.4%) had weight-for-stature percentiles >2.5 percentage points lower than their BMI-for-age percentiles. Thus, weight-for-stature percentiles tended to be lower than BMI-for-age percentiles for the same children. This asymmetrical distribution of differences was apparent for each age group, but was more pronounced among children aged 4 and 5 y.
|
The results for boys are shown in Figure 1 for the 10th percentile and in Figure 2 for the 85th percentile. The results for girls (not shown) were similar to those for boys. For boys, the 10th percentile of weight-for-stature corresponded to BMI-for-age percentiles ranging from the 3rd percentile to the 21st percentile, depending on age and stature. The 85th percentile of weight-for-stature corresponded to BMI-for-age percentiles ranging from the 74th percentile to the 92nd percentile, depending on age and stature.
|
|
). For example, for children aged 24 mo with a stature of 79.5 cm, the 10th percentile of weight-for-stature corresponded to the 22nd percentile of BMI-for-age, but for children aged 24 mo with a stature of 93.5 cm, the 10th percentile of weight-for-stature corresponded to the 4th percentile of BMI-for-age. At shorter statures, the 10th percentile of weight-for-stature was equivalent to a higher percentile of BMI-for-age and thus would be expected to classify more children as underweight than would the 10th percentile of BMI-for-age. For a given stature, the BMI-for-age percentile equivalent to the 10th percentile of weight-for-stature was strongly dependent on age. For example, at a stature of 89.5 cm, the 10th percentile of weight-for-stature was equivalent to the 7th percentile of BMI-for-age for a child aged 24 mo but to the 14th percentile for a child aged 36 mo.
The BMI-for-age percentile that was equivalent to the 85th percentile of weight-for-stature varied with both age and stature in a similar way for children aged 24, 36, and 48 mo. The associations with weight and stature were stronger at younger ages (Figure 2). At shorter statures, the 85th percentile of weight-for-stature was equivalent to a higher percentile of BMI-for-age and thus would be expected to classify fewer children as overweight than would the 85th percentile of BMI-for-age. For children aged 48 and 60 mo, the 85th percentile of weight-for-stature was always above the 85th percentile of BMI-for-age; thus, for these ages, the 85th percentile of weight-for-stature would be expected to classify fewer children as overweight than would the 85th percentile of BMI-for-age.
Classification by weight-for-stature and BMI-for-age
The 10th and the 85th percentiles of BMI-for-age are used as cutoffs for classification of some degree of potential nutritional risk. It was suggested that children be classified as either at risk of overweight or overweight if their BMI value is at or above the 85th percentile of BMI-for-age (9,10). Thus, it was of interest to examine the properties of classification by the 10th and 85th percentiles of weight-for-stature relative to the classification by the 10th and 85th percentiles of BMI-for-age. Because the CDC recommends the use of BMI-for-age, we considered BMI-for-age as the reference method and weight-for-stature as the test method. We considered children to be classified correctly by weight-for-stature if the classification matched their classification according to BMI-for-age. We excluded children for whom weight-for-stature percentiles could not be calculated.
Relative to BMI-for-age 10th percentile, both the sensitivity and specificity of weight-for-stature 10th percentile were high, >94% in almost all sex-age groups except for 3-y-old boys, for whom the sensitivity was 76%. As expected, because weight-for-stature percentiles tend to be lower than the corresponding BMI-for-age percentiles of the same children (Table 2), specificity was high at the 85th percentile. Ninety-seven to 100% of children below the 85th percentile of BMI-for-age were also below the 85th percentile of weight-for-stature, both overall and within each age group. The sensitivity (proportion of children 85th percentile of BMI-for-age who were also 85th percentile of weight-for-stature) was lower, ranging from 70% for boys aged 4 y to 88% for boys aged 2 y. For example, this sensitivity indicates that, of 4-y-old boys who were classified as being 85th percentile of BMI-for-age, 30% would be classified as <the 85th percentile of weight-for-stature.
Prevalence of BMI-for-age or weight-for-stature 10th percentile or 85th percentile
The prevalence of the categories 10th percentile and 85th percentile from the 2 different charts by sex and year of age, along with the differences between the 2 estimates, is shown in Table 3. The findings were consistent with the relatively lower values of weight-for-stature percentiles than of BMI-for-age percentiles shown in Table 2. Except for 3-y-old boys, the proportion of children 10th percentile of weight-for-stature was greater than the proportion 10th percentile of BMI-for-age. The differences were more pronounced for girls than for boys and were largest for 2-y-old children, both boys and girls. Of 2-y-old girls, 10.1% were 10th percentile of weight-for-stature compared with 5.1% who were 10th percentile of BMI-for-age, a difference of 5.0 percentage points.
|
85th percentile of BMI-for-age than were 85th percentile of weight-for-stature. The differences were largest for children, both boys and girls, aged 4 y (5.4 percentage points for boys and 6.0 percentage points for girls). Thus, for 4-y-old boys, BMI-for-age categorized 18.2% of children as 85th percentile, but weight-for-stature categorized only 12.8% of the children as 85th percentile, a difference of 5.4 percentage points.
Because the relation of weight-for-stature percentile to BMI-for-age percentile differed by stature, as shown in Figures 1 and 2, we repeated the analyses shown in Table 3 further by dividing children into those 50th percentile of stature-for-age (shorter) and <50th percentile of stature-for-age (taller). The differences between weight-for-stature and BMI-for-age varied by stature, particularly in younger children (data not shown). Shorter children were considerably more likely to be classified as underweight (10th percentile) and less likely to be classified as overweight (85th percentile) by weight-for-stature than by BMI-for-age, but the differences were small for taller children. For example, of shorter 2-y-old boys, 6.4% more were underweight by weight-for-stature (14.0%) than by BMI-for-age (7.6%), but for taller 2-y-old boys, the difference between the 2 indicators was only 1.0%. Of shorter 2-y-old boys, 3.1% more were overweight by BMI-for-age (11.6%) than by weight-for-stature (7.5%), but for taller 2-y-old boys the difference was -0.3%.
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
Both weight-for-stature and BMI-for-age are intended to show a child's weight relative to similar individuals. Both these indicators are expressions of weight, not adiposity. They should correlate with body fat to the extent that weight correlates with body fat (keeping sex, age, and stature constant). However, because weight is not identical to adiposity, neither indicator would be expected to have a perfect correlation with adiposity. At the high end it is likely, although not inevitable, that a child with a high BMI-for-age or weight-for-stature has a high level of adiposity; at the low end, a child with an extremely low BMI-for-age or weight-for-stature is likely to have both low adiposity and low muscle and lean tissue mass. Because both weight-for-stature and BMI-for-age express weight relative to a reference population, it is appropriate to compare them with each other.
Both weight-for-stature and BMI-for-age may be imperfect indicators of a child's relative weight status. The most exact approach would be to compare a child's weight with a reference population of the same sex, age, and stature. However, to have enough children of the same age and stature to create reference percentiles for weight for each sex-age-stature combination would require a much larger sample than any presently available.
Our results show that weight-for-stature and BMI-for-age are not interchangeable and do not produce equivalent results. Weight-for-stature percentiles tend to be lower than BMI-for-age percentiles. As a result, children are less likely to be classified as at risk of overweight or overweight on the basis of weight-for-stature than they are on the basis of BMI-for-age but more likely to be classified as underweight or at risk of underweight. The overall agreement between the 2 is poorer at the ages of 4 and 5 y than at the ages of 2 and 3 y. Particularly at the ages of 2 and 3 y, the agreement between the 2 also depends on stature, with weight-for-stature being equivalent to a higher BMI-for-age percentile at shorter statures but a lower BMI-for-age percentile at taller statures.
Weight-for-stature has been used in the past and is easier to understand and to calculate than is BMI. The 1977 NCHS growth charts included weight-for-stature but not BMI-for-age. However, because of the increasing use and acceptance of BMI as an appropriate indicator for children (9–11), it was decided to include BMI-for-age charts in the revised version. BMI-for-age is recommended as the appropriate indicator for use in most settings. Weight-for-stature charts for younger children were subsequently also included to provide additional flexibility for a smoother transition to the new charts, particularly in situations such as the WIC program for which the previous weight-for-stature charts had already been used extensively and thus were more familiar to users. Currently, for purposes of the WIC program, BMI-for-age from the 2000 CDC growth charts can be used as an indicator of nutritional risk, with weight-for-stature as a possible alternative.
Weight-for-stature has, however, been criticized for some reasons similar to those discussed in the present study (12,13). As reported by Cole (12,13), weight-for-stature is not independent of age. At present, the use of BMI-for-age rather than weight-for-stature seems generally more advisable. BMI-for-age was recommended for use in identifying children as either at risk of overweight or overweight (9,10) and at present there are no similar recommendations for weight-for-stature. Compared with weight-for-stature, BMI-for-age has the advantages of being able to capture the change in the weight-height relation with age and of providing a measure that can be used continuously up to the age of 20 y with the 2000 CDC growth charts. In any case, BMI-for-age and weight-for-stature will not give identical results and the 2 methods are not interchangeable.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  A. Nash, D. Secker, M. Corey, M. Dunn, and D. L. O'Connor Field Testing of the 2006 World Health Organization Growth Charts From Birth to 2 Years: Assessment of Hospital Undernutrition and Overnutrition Rates and the Usefulness of BMI JPEN J Parenter Enteral Nutr, March 1, 2008; 32(2): 145 - 153. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 T. J Cole, K. M Flegal, D. Nicholls, and A. A Jackson Body mass index cut offs to define thinness in children and adolescents: international survey BMJ, July 28, 2007; 335(7612): 194 - 194. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 S. Evers, R. Arnold, T. Hamilton, and C. Midgett Persistence of Overweight among Young Children Living in Low Income Communities in Ontario J. Am. Coll. Nutr., June 1, 2007; 26(3): 219 - 224. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 S. Phillips, A. Edlbeck, M. Kirby, and P. Goday Ideal Body Weight in Children Nutr Clin Pract, April 1, 2007; 22(2): 240 - 245. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 E. W. Demerath, C. M. Schubert, L. M. Maynard, S. S. Sun, W. C. Chumlea, A. Pickoff, S. A. Czerwinski, B. Towne, and R. M. Siervogel Do Changes in Body Mass Index Percentile Reflect Changes in Body Composition in Children? Data From the Fels Longitudinal Study Pediatrics, March 1, 2006; 117(3): e487 - e495. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 S. T. Iannaccone, H. Owens, J. Scott, and B. Teitell Postoperative Malnutrition in Duchenne Muscular Dystrophy J Child Neurol, January 1, 2003; 18(1): 17 - 20. [Abstract] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
