This report provides detailed information on how the 2000 Centers for Disease Control and Prevention (CDC) growth charts for the United States were developed, expanding upon the report that accompanied the initial release of the charts in 2000.
The growth charts were developed with data from five national health examination surveys and limited supplemental data. Smoothed percentile curves were developed in two stages. In the first stage, selected empirical percentiles were smoothed with a variety of parametric and nonparametric procedures. In the second stage, parameters were created to obtain the final curves, additional percentiles and z-scores. The revised charts were evaluated using statistical and graphical measures.
The 1977 National Center for Health Statistics (NCHS) growth charts were revised for infants (birth to 36 months) and older children (2 to 20 years). New body mass index-for-age (BMI-for-age) charts were created. Use of national data improved the transition from the infant charts to those for older children. The evaluation of the charts found no large or systematic differences between the smoothed percentiles and the empirical data.
The 2000 CDC growth charts were developed with improved data and statistical procedures. Health care providers now have an instrument for growth screening that better represents the racial-ethnic diversity and combination of breast- and formula- feeding in the United States. It is recommended that these charts replace the 1977 NCHS charts when assessing the size and growth patterns of infants, children, and adolescents.
Keywords: growth charts c height c length c weight c body mass index c head circumference c NHANES
The 2000 CDC Growth Charts consist of a set of charts for infants, birth to 36 months of age and a set of charts for children and adolescents from ages 2 to 20 years. The charts for infants include sex-specific smoothed percentile curves for weight-for-age, recumbent length-for-age, head circumference-for-age, and weight-for recumbent length; the charts for children and adolescents include weight-for-age,
Anthropometric data are valuable objective indicators of attained size and physical growth in children. A variety of growth references were developed and used in the United States since the early 1900s. Most of these earlier references have considerable limitations, including lack of coverage for infants and preschool children and limited
representation of ethnic, genetic, socioeconomic, environmental, and geographic variability (2). From 1946–1976, the Stuart/Meredith Growth
Table A. Charts included in the 1977 NCHS Growth Charts and in the 2000 CDC Growth Charts
1977 NCHS Growth Charts* 2000 CDC Growth Charts*
Charts were widely used. These charts were derived from stature and weight measurements taken on white children living near Iowa City, Iowa, or in Boston, Massachusetts, from 1930 to 1945 (3). The sample sizes were relatively small and the smoothed percentile lines were reportedly based on mathematical approximations of curves smoothed by hand (4). Thus, the data did not represent the diversity of children residing in the United States and statistical curve-fitting procedures were not used.
The impetus for the development of the 1977 NCHS Growth Charts began in 1971, when a study group sponsored by
the American Academy of Pediatrics and
Birth to 36 months 2 to 18 years
Birth to 36 months
Weight-for-length Birth to 36 months
Boys (49 to 103 cm)
Girls (49 to 101 cm)
Head circumference-for-age Birth to 36 months
Stature-for-age 2 to 18 years
Weight-for-stature** Boys (90 to 145 cm)
Girls (90 to 137 cm)
*All charts are sex-specific.
Birth to 36 months 2 to 20 years
Birth to 36 months
Weight-for-length Birth to 36 months
Boys (45 to 103 cm)
Girls (45 to 103 cm)
Head circumference-for-age Birth to 36 months
Stature-for-age 2 to 20 years
Weight-for-stature** Boys (77 to 121 cm)
Girls (77 to 121 cm)
2 to 20 years
the Maternal and Child Health Program of the Bureau of Community Health Services, U.S. Public Health Service,
**The 1977 charts are applicable to boys with stature from 90 to 145 cm and age under 11.5 years, and to girls with stature from 90 to 137 cm and under 10.0 years of age. They are not applicable for any child showing the earliest signs of pubescence. The revised charts have no similar age or pubescence restrictions. Although the revised charts were developed for children 2 to 5 years of age, in practice they may accommodate some shorter children with chronologic ages 5.0 years and over.
recommended new growth charts, based on data from the NCHS Health Examination Surveys, for the clinical assessment of infants and children (5). Subsequently, in 1974, the Food and Nutrition Board of the National Academy of Sciences made similar recommendations, emphasizing the need for new growth charts for infants and children based on nationally representative survey data, to be supplemented with data collected from infants in the Fels Longitudinal Growth Study (6). These recommendations were further supported in 1975 by a study group sponsored by the National Institute of Child Health and Human Development, National Institutes of Health (7).
Growth charts for the United States were developed by NCHS when nationally representative cross-sectional survey data became available for most of the pediatric age range (8). Data from the National Health Examination Survey (NHES) II (1963–65) for ages 6–11
years, NHES III (1966–70) for ages 12–17 years, and the first National Health and Nutrition Examination Survey (NHANES) I (1971–74) for ages 1–17 years were used to develop these charts. National survey data were not available for the period from birth to 1
year. Therefore, the national survey data were supplemented with data from the longitudinal growth study of the Fels Research Institute in Yellow Springs, Ohio. To avoid pooling multiple diverse data sets in the 1977 NCHS charts for birth to 3 years, the infant charts were based solely on the Fels data. The 14 sex-specific growth charts developed for infants birth to age 36 months and older children from ages 2 to 18 years are generally referred to as the 1977 NCHS Growth Charts (2,8,9).
In 1978 CDC produced a normalized version of the NCHS curves (10,11). The World Health Organization (WHO) subsequently recommended these normalized charts for international applications (12–14). These normalized versions of the 1977 charts are thus sometimes referred to as the NCHS/WHO, CDC/WHO, or NCHS/CDC/WHO growth charts.
Following the recommendation of Waterlow, et al., this version enabled the expression of body measurements in terms of standard deviations from the median or z-scores (15). Z-scores facilitate comparisons across ages and allow the mean and standard deviation to be calculated for a group of measures. Another advantage of the
normalized charts was the ability to describe the relative status of children at extremes of the distributions.
Although the normalized charts allowed users a means to better quantify growth at the extremes, normalization does not refer to an improved instrument to qualify growth as
‘‘normal’’ or as a ‘‘standard’’ for growth. Both the 1977 smoothed percentiles and the 1978 normalized growth curves are growth references. They allow the growth status of a child, or a group of children, to be compared with that of a reference population. The 1978 normalized curves are similar to, but not identical with the 1977 smoothed percentiles.
The 1977 charts were revised in part because more recent and comprehensive national data on body measurements in U.S. children were available. After the 1977 NCHS Growth Charts were developed, additional national survey data became available from the NHANES II (1976–80)
beginning at 6 months of age, and from the NHANES III (1988–94) beginning at 2 months of age (24,25). During the planning of NHANES III, increasing awareness of the concerns surrounding the 1977 NCHS Growth Charts influenced the decision to oversample children ages 2 months to 6 years.
The objective of the revision process was to use improved statistical smoothing procedures, in conjunction with more comprehensive national survey data, to provide a better instrument for health care providers who evaluate the growth status of children in the United States (16). The Advance Data report summarizing the development of the revised growth charts was issued in 2000 (1). The
present report is designed to expand on the content of the initial report and provide more detailed information on the development of the revised CDC growth charts for the United States, including the process, the data sources, the statistical procedures that were followed, and the results of statistical evaluations for the revised charts.
height (cm) range Primary data sources1 Supplemental data sources
Birth to 36 months
National surveys 3–52
National birth certificate data from United States Vital Statistics2
Birth to 36 months
National surveys 3–52,3
Birth certificate data from Wisconsin and Missouri State vital statistics2,4; CDC Pediatric Nutrition Surveillance System data for 0.5, 1.5, 2.5, 3.5, and
Birth to 36 months
National surveys 3–52
Fels Longitudinal Study data2
45 to 103 cm
National surveys 3–52,5
Birth certificate data from Wisconsin and Missouri State vital statistics2
77 to 121 cm
National surveys 3–55
24 to 240 months
National surveys 1–55
24 to 240 months
National surveys 1–5
24 to 240 months
National surveys 1–55
1Survey 1 = NHES II, Survey 2 = NHES III, Survey 3 = NHANES I, Survey 4 = NHANES II, and Survey 5 = NHANES III.
2Excludes birthweight <1,500 gm.
3Excludes data from NHANES III for ages <3.5 months.
4Wisconsin and Missouri were the only states with available data from birth certificates.
5Excludes data from NHANES III for ages 72 months.
Page 4 [ Series 11, No. 246
II (1976–80), and III (1988–94). The
survey designs are based on stratified, multistage probability samples of the civilian, noninstitutionalized population in the 48 contiguous States (NHES II, NHES III, NHANES I) or all 50 States (NHANES II, NHANES III). All
surveys consisted of a home interview and a standardized physical examination conducted in a mobile examination center. Age was calculated as age at the time of examination when the anthropometric data were recorded.
Children ages 6–11 years from NHES II, 12–17 years from NHES III, 1–19 years from NHANES I, six months–19 years from NHANES II, and 2 months–19 years from NHANES III were included in the revision. The small
Anthropometric procedures developed for NHANES III are documented on videotape
URL: http://www.cdc.gov and in the NHANES III procedures manual (30). The procedures were consistent with published recommendations for standardized anthropometric techniques (31). The NHANES III measurement techniques for weight, recumbent length, stature, and head circumference were based on procedures used in the previous NHES and NHANES surveys, and either the same or comparable measuring equipment was used across the surveys.
1Data from outside the 2 to 20-year range for the child/adolescent charts were used to improve estimates at the upper and lower age boundaries. Subject ages, shown for growth chart variables, reflect the endpoints of age ranges for data actually used to construct the smoothed percentile curves.
2W = weight-for-age; S = stature-for-age; BMI = body mass index-for-age; L = length-for-age; HC = head circumference-for-age; WL = weight-for-length; WS = weight-for-stature.
3Lengths at ages 18.0–23.99 months, and stature at all other ages.
weight-for-age curves included national birthweight distributions taken from birth certificates for more than 83 million infants born in the United States between 1968–80 and 1985–94, corresponding to years in which infants in the national surveys were born. Birth length data were only available from two States, Wisconsin and Missouri.
The majority of infant data in the PedNSS system come from clinical records of the Special Supplemental Nutrition Program for Women, Infants and Children (WIC) (20). Because the PedNSS data are not nationally representative, a subset of the PedNSS data collected from 1975 to 1995 was used. Clinics were selected for inclusion based on three specific conditions. First, the mean lengths and weights had to be within ±0.5 cm and ±0.5 kg of the mean from NHANES II and III combined for each single month of age from age 3 to 11 months. Second, the clinics had to have a SD within ±0.2 cm and ±0.2 kg of the SDs in the combined NHANES II and III. Finally, the skewness in weight distribution of the selected clinics’ population had to be within ±0.3 kg of the skewness of weight in the combined NHANES II and III. A total of 213 PedNSS clinics were selected, resulting in a sample of 14,846 observations at
0.5 months, 8,825 at 1.5 months, 5,240 at 2.5 months, 1,640 at 3.5 months, and 2,258 at 4.5 months. All the matching procedures excluded subjects with birthweights <1,500 grams.
Head circumference at birth was not available from the national surveys or from birth certificates. Consequently, the head circumference-for-age curves included data for head circumference at birth from 362 infants in the Fels
Longitudinal Study who were born between the years 1960 and 1994, corresponding to the national surveys. Most of these data were recorded in hospitals by the Fels staff. The Fels Longitudinal Study began in 1929 with a major goal of following physical growth and development in a cohort of 1,000 people from birth throughout their entire life cycle (32). Data at birth, including head circumference at birth, obtained since 1975, come from relatives of the early cohorts.
Several exclusions were made prior to curve smoothing. First, data for all very low birthweight (VLBW) infants (<1,500 grams) were excluded from the infant growth charts, primarily because the growth of VLBW infants is known to be markedly different from that of higher birthweight, full-term infants. For NHANES I and II, if a reported birthweight from the interview questionnaire was not available, an exclusion was not applied. As part of the NHANES III protocol, birth certificates were obtained from the States for children born in the United States. For NHANES III, if a reported
birthweight was missing from the survey interview data, birthweight from the child’s birth certificate was used to determine possible exclusion, and, if neither were available, an exclusion was not made. Second, data from NHANES III for children greater than or equal to 6 years of age were excluded from the charts for weight-for-age, weight-for stature, and BMI-for-age. Inclusion of these data would have led to the underclassification of overweight, because overweight cutoff criteria based on weight- and BMI-for-age percentiles would have been shifted upward. Third, 11 infants whose recumbent length and stature differed by greater than or equal to 5 cm were excluded from the
length-for-age, weight-for-length, stature-for-age, weight-for-stature, and BMI-for-age charts. Fourth, two outlier
values, one for head circumference of an infant girl, and one for recumbent length of an infant boy, were excluded because the measurement values and the sampling weights were extreme.
Data from the national surveys were pooled because no single survey in the NHANES series has enough observations to construct growth charts. Sample sizes from 400 to 500 are required to achieve precision of the empirical percentiles at the specific ages selected for the curve fitting (33). This is particularly important for outlying percentiles that are used in research and clinical practice. Pooling enhances the number of subjects at each age, thereby increasing the stability of the outlying percentile estimates.
Statistical sample weights were available for each national survey. These sample weights take into account the unequal probabilities of selection resulting from the complex sampling cluster design, planned oversampling of selected subgroups, nonresponse, and noncoverage. These survey-specific sample weights were applied to the national survey data resulting in each survey representing the U.S. population at the time the survey was conducted.
All statistics were calculated with the original survey sample weights.
Sampling weights were not available for the supplemental data.
Statistical procedures were applied to the observed data in two stages, first to generate initial smoothed curves for selected major percentiles and second to generate the parameters that were used to construct the final smoothed curves and additional percentiles. The first stage is referred to as the curve smoothing stage, and the second stage as the transformation stage. In the first stage, selected empirical percentiles were smoothed with a variety of parametric and nonparametric regression procedures. In the transformation stage, the smoothed curves were approximated using a modified LMS estimation procedure to provide the transformation parameters, lambda, mu, and sigma (LMS). This resulted in final percentile curves that closely matched the percentile curves smoothed in the first stage and allowed computation of additional percentiles and z-scores. The procedures for each stage are described
For birth to 36 months, a 3-parameter linear model fit to empirical percentile points for weight at midpoints of age intervals, and anchored (i.e., forced) at birth and at 2.75 years. Averages of weighted empirical percentiles at 2.25 and 2.75 years (based on data used for birth to 36 months and 2 to 20 years) substituted for data points in the 24 to 36 months range.
For 2 to 20 years, locally weighted regression (LWR) based on 15-point smoothing for boys and 17-point smoothing for girls. Fit to empirical percentile points for weight at midpoints of age intervals, combined averages of weighted empirical percentiles at 2.25 and
2.75 years (based on data used for birth to 36 months and 2 to 20 years), and predicted values for 2 to 3 years at 0.1 year intervals from the 3-parameter linear model used for birth to 36 months.
All results from the 3-parameter linear model and LWR were combined and refit at midpoints of age intervals from birth to 20 years using a 10-parameter polynomial regression model for boys and 9-parameter polynomial regression model for girls.
Length-for-age and Stature-for-age
For length-for-age, birth to 36 months, a 3-parameter linear model was fit to empirical percentile points for length at midpoints of age intervals and to birth data.
For stature-for-age, 2 to 20 years, a 10-parameter nonlinear model was fit to empirical points for stature at midpoints of age intervals.
Results from the 3-parameter linear model used for length-for-age were adjusted by subtracting 0.8 cm from length to make length continuous with stature in the overlapping age interval of 24 to 36 months. Percentiles in the overlap period were averaged by assigning weights of 1, 11/12, ..., 1/12, 0 at 24, 25, ..., 35, 36 months, respectively, to the adjusted length-for-age. Opposite weights of 0, 1/12, ..., 11/12, 12/12 at 24, 25, ..., 35, 36 months, respectively, were assigned to smoothed stature-for-age percentiles. The final length-for-age and stature-for-age curves were created by adding back 0.8 cm to the smoothed length.
Head circumference-for age
For birth to 36 months, a 3-parameter linear model was fit to empirical percentile points for head circumference at midpoints of age intervals and to birth data.
Weight-for-length and Weight-for-stature
Empirical weight-for-length data were adjusted by subtracting 0.8 cm from length to make length continuous with stature in the overlapping age interval of 24 to 36 months. The combined adjusted weight-for-length and weight-for-stature data were smoothed with a 5-parameter polynomial regression model, fit to empirical percentile points for weight at midpoints of 2 cm intervals. After applying LMS, separate weight-for-length curves were created by adding 0.8 cm back to length.
For 2 to 20 years, LWR model was based on a 5-point smoothing at midpoints of age intervals for ages 2 to 12.5 years, and a 25-point smoothing for boys and a 27-point smoothing for girls for ages 13 to 20 years. The curves were further smoothed with a 4-parameter polynomial regression model fit to smoothed percentile points for BMI at midpoints of age intervals.
Before smoothing, data were grouped by single month of age for the development of charts relating size to age. Each month of age was truncated to the nearest completed month, for example, 1 month (1.0–1.9 months), 11 months
(11.0–11.9 months), 23 months (23.0–23.9 months), and so forth. From birth to 12 months, the empirical percentile estimates were made at 1-month intervals; from 12 to 24 months, the empirical percentile estimates were made at 3-month intervals; and at 24 months and beyond, empirical percentile estimates were made at 6-month intervals. All ages were labeled as the midpoint of the defined age groups. For the infant charts, infants were grouped as follows: 0.5 months = 0.1–0.9 months (birth was not included),
1.5 months = 1.0–1.9 months, ...,
11.5 months = 11.0–11.9 months;
13.5 months = 12.0–14.9 months,
16.5 months = 15.0–17.9 months,
19.5 months = 18.0–20.9 months,
22.5 months = 21.0–23.9 months;
27.0 months = 24.0–29.9 months, and
33.0 months = 30.0–35.9 months.
For the charts for older children and adolescents, subjects were separated into 36 half-year age groups. Each age group was categorized by the midpoint of an age range. For example, age 2.25 years included ages from 2.0 years to 2.5 years of age. This pattern continued up to the 19.75 year age interval in which the age range is from 19.5 to 20.0 years of age.
Before smoothing, the length and stature data were grouped by 2-cm intervals. For example, 46 cm = 45–46.9 cm.
For each growth chart, the initial smoothing methods were applied to nine empirical percentiles (3rd, 5th, 10th, 25th, 50th, 75th, 90th, 95th, and 97th). In addition, the 85th percentile was included in the BMI-for-age charts because the 85th percentile of BMI has been recommended as a cutoff to identify children and adolescents at risk for overweight (34,35).
The weighted empirical percentile estimates were obtained by applying the
survey-specific sample weights to calculate weighted empirical percentile points at the midpoint of each age group (or the midpoint of each 2-cm interval for length or stature). The irregular plots of empirical percentile values had to be smoothed to produce clinically useful percentile curves.
Several different approaches were used in the smoothing stage. The empirical percentiles for infant weight, length, and head circumference were smoothed using a family of three- parameter linear models that have been used previously to describe age-related changes in growth from birth to age 36 months (36–38). Stature-for-age was smoothed using a nonlinear model whereas weight-for-stature and
weight-for-length were smoothed using a 5-degree polynomial regression model. Two-step smoothing was applied to weight-for-age in infants and older children and BMI-for-age. In the first step of smoothing weight-for-age and BMI-for-age for older children, locally weighted regression (LWR) was used.
Polynomial regression was used in the second step of smoothing weight-for-age and BMI-for-age. The smoothing stage resulted in every chart having a
Series 11, No. 246 [ Page 7
parametric form with estimated parameters specific for each selected major percentile.
The parameters of the linear regressions were estimated using the SAS procedure REG, and the parameters in the nonlinear regression were estimated using the SAS procedure NLIN (39). The fit of the models was evaluated using root mean square error (RMSE), R2, and CV (40).
In order to estimate any percentile and allow calculation of standard deviation units and z-scores, a modified LMS statistical procedure was applied to the smoothed percentile curves. The LMS method does not change the distribution of percentile curves in a growth chart; rather it provides a way to estimate percentiles in a continuous manner.
The distribution of some anthropometric data used in the growth charts are skewed. To remove skewness, a power transformation can be used to stretch one tail of the distribution while the other tail is shrunk. A Box-Cox transformation can make the distribution nearly normal (41). The assumption is that, after the appropriate power transformation, the data are closely approximated by a normal
distribution (42). The transformation does not adjust for kurtosis, which is a less important contributor to nonnormality than skewness (43).
In the LMS technique, three parameters are estimated: the median (M), the generalized coefficient of variation (S), and the power in the Box-Cox transformation (L). The L reflects the degree of skewness. The LMS transformation equation is:
X = M (1 + LSZ)1/LL 0
X = M exp(SZ) L =0
where X is the physical measurement and Z is the z-score that corresponds to the percentile.
The key task of the transformation was to estimate parameters L, M, and S. With estimates of L, M, and S, values of X are connected to the values of Z through the above equation. The
percentile is obtained from a normal distribution table where the z-score corresponds to the percentile of interest. For example, a z-score of 0.2019 corresponds to the 58th percentile. In
the case of growth charts, with the L, M, and S parameters, it is possible to evaluate any single measure in a population as an exact z-score or percentile.
To generate age-specific estimates of L, M, and S, Cole (42,44) has recommended applying a penalized likelihood estimation procedure to the raw data. In this approach smoothed curves of L, M, and S are generated first, and then smoothed percentile curves, or an individual standardized score, can be obtained from the values of L, M, and S.
In contrast to the original LMS procedure, a modified LMS estimation procedure was created and used to generate the 2000 CDC Growth Charts. In the modified LMS approach, empirical percentile curves were initially smoothed and parametric models were generated, as described above. Then, at each age or length/stature interval, a group of 9 equations (10 for BMI charts) was generated by specifying the LMS transformation equations for the previously smoothed major percentiles. A simultaneous solution for the three parameters of L, M, and S from the group of specified equations was generated using the SAS procedure NLIN (39). By minimizing the sum of squared errors, the set of L, M, and S parameters was obtained as the best solution to a system of equations rather than as likelihood-based estimates from empirical data. This approach is similar to the method used by Cole to estimate LMS parameters from published percentile curves (45–47).
This modified LMS procedure produced final curves that are extremely close to the smoothed percentile curves obtained from the first stage of smoothing. The net result is that the close fit of the smoothed curves from the first stage of smoothing to the empirical data is retained. In addition, the modified LMS method allows
z-scores to be obtained in a continuous manner. The LMS values were calculated by solving equations that
used the values for percentiles ranging from the 3rd to the 97th. Percentiles less than the 3rd or greater than the 97th are beyond the range of the data from
which the LMS parameters were calculated. As in any statistical procedure, extrapolation beyond the range of the data should be done with caution.
The final set of percentile curves for the CDC growth charts presented in this report was produced using this modified LMS estimation procedure. In the transformation stage, percentiles were developed at 1-month or
1-centimeter intervals in the infant and child charts. Estimates of L, M, and S parameters in these intervals were calculated to provide the necessary tools for determining additional percentiles.
Generally, 1-month or 1-centimeter intervals will be adequate for estimation or evaluation. To obtain percentiles at finer intervals, the L, M, and S values could be interpolated.
Combining infant and child/adolescent weight-for-age—After the infant weight-for-age and child/adolescent weight-for-age curves were smoothed using a 3-parameter linear model and LWR, the results were combined and refit from birth to age 20 years using a single regression model for each sex.
The smoothed weight-for-age curves for infants and for older children were combined to obtain a seamless transition between the curves. Ultimately the combined weight-for-age curve was separated into infant and child/ adolescent curves to facilitate use in clinical settings.
In order to combine the infant and
child/adolescent weight-for-age curves, weighted averages of overlapping empirical percentiles from infant and child charts at 2.25 years (24.0–29.9
months) and 2.75 years (30.0–35.9 months) were calculated using the combined infant and child/adolescent data. The empirical percentiles were not identical at ages 24–36.9 months because VLBW infants (<1,500 grams) were excluded from the infant
Page 8 [ Series 11, No. 246
percentiles, but included in the older child percentiles where the effect of VLBW is diminished. The averaged percentiles were substituted for the empirical values in smoothing the weight-for-age for infants and
weight-for-age for children/adolescents. These combined percentile curves were fit by a 10-degree polynomial for boys and by a 9-degree polynomial for girls. A set of nine polynomial regression equations, one for each of the major smoothed percentiles, was solved simultaneously for infants and for children to estimate the L, M, and S parameters for boys and for girls separately.
Weight-for-age, birth to 36 months— The infant weight-for-age curves for ages birth to 36 months were smoothed initially with a 3-parameter linear model:
f(t)= a+b* ln(t+0.5) + c(t+0.5)0.75
where f(t) is body weight in kg, t is age in months (calculated as midpoint of the age range), and a, b, and c are
percentiles for infants and older children at 2.25 and 2.75 years, predicted infant values for weight from 2 to 3 years at 0.1-year intervals from the 3-parameter linear model and the empirical percentiles from 3 to 20 years. This approach used all available information for infants after excluding data for VLBW infants and for older children where data for VLBW infants were not excluded. Some empirical data less than 2 years of age and greater than or equal to 20 years were included, as described below. This compound data set was smoothed with a locally weighted regression (LWR) procedure.
The LWR does not give any
parameter estimates, but provides an intermediate smoothed curve for further parametric smoothing. Locally weighted regression applies a weight function to data in the neighborhood of the value to be estimated (48). Ages at measurements that are near that of the value to be estimated received larger weights than those farther away from the specific age. A weight function
regression on the 15 neighborhood points adjacent to the estimated value. For girls, a 17-point smoothing procedure was chosen. Seven data points for boys and eight data points for girls at ages less than 2 years were necessary for smoothing the value at age 2 years. The weight value at age 1.75 years in boys was repeated 7 times from age
1.75 down to 1.15 years by 0.1-year intervals, and in girls was repeated 8 times from age 1.75 down to 1.05 years by 0.1-year intervals. Similarly, when smoothing the value at age 20 years, seven and eight data points at ages over 20 years were necessary for boys and girls, respectively. For boys, the empirical percentiles from 20.25 to
23.25 years by 0.5-year intervals were used as seven additional data points for smoothing. For girls, the empirical percentiles after age 18 years were very irregular due to the limited number of subjects and considerable variation in body weight. Therefore, instead of using empirical percentiles from 20 to 25 years, the value at age 20.25 years was
repeated 8 times from age 20.25 to
parameters to be estimated. The
Weight = 1–
23.75 years by 0.5-year intervals to
estimated values of f(t) are the smoothed values. At birth, t = 0. Estimates of a, b, and c are specific for each of nine percentiles being smoothed. An exponent of 0.75 was selected based on the evaluation of RMSE after several models were tried, beginning with the exponent of 0.3.
After excluding birthweights
<1,500 gm, the model was fit to the national data with anchor points at birth and at 2.75 years. In other words, when the regression line was fit, no error term was allowed at the two anchor points.
The curves were anchored to the national distribution of birthweights (years 1968–80 and 1985–94) from NCHS natality files (that is, forced through the sex-specific summary values provided for the major centile lines at birth) after excluding birthweights
<1,500 gm, and forced through the weighted average percentile values at
Weight-for-age, 2 to 20 years—Initial smoothing of weight-for-age for older children involved combining the weighted average of the empirical
? ΔX ?
applies larger weights to data points
near the value to be estimated than to those further away, where X0 is the central age at which the value is smoothed, Xi is the i th age from the central age and the value to which the weight is being assigned, and ΔX is the age range covered by the width of the moving regression window. The mechanics of LWR require the selection of a series of points on either side of the value to be estimated. This series of points constitutes the regression window. Windows for each of the values to be estimated overlap and are referred to as moving regression windows. The weighted least squares regression is applied to the values in each moving regression window to provide the smoothed estimate at X0. The resulting estimates from each regression window form a smoothed curve.
The width of the LWR moving window for weight-for-age was chosen after several trials to balance the degree of smoothness and fidelity to the data. For boys, each smoothed value was estimated by a weighted linear
achieve reasonably smoothed values near age 20 years.
Length-for-Age and Stature-for-Age
Combining length-for-age and stature-for-age—A smooth transition from the length-for-age to stature-for
age chart was required. Analyses of data from NHANES II and III children indicated that recumbent length is, on average, 0.8 cm longer than standing height (stature). Thus, it is desirable to have parallel smoothed percentile curves for length-for-age and stature-for-age with a difference of 0.8 cm between length and stature in the overlapping ages from 24 to 36 months.
To obtain these parallel curves,
percentiles for each month of age were predicted by the 3-parameter model in the length-for-age portion and by the 10-parameter nonlinear model in the stature-for-age portion. Length-for-age percentiles were adjusted by subtracting
0.8 cm from length to make length commensurate with stature. Averages of the percentiles in the two charts in the overlapping ages were calculated by assigning weights of 1, 11/12, 10/12, ...,
Series 11, No. 246 [ Page 9
1/12, 0 to values of length-for-age at 24, 25, 26, ..., 35, 36 months, respectively, and assigning opposite weights of 0, 1/12,
..., 11/12, 1 to values of stature-for-age at 24, 25, ..., 35, 36 months, respectively.
These weighted average percentile values in the overlapping section were shared by both charts. Length-for-age was readjusted by adding back 0.8 cm to length, producing separate length-for-age and stature-for-age curves. This resulted in the two sets of parallel percentiles with a consistent separation of 0.8 cm in the overlapping section.
Two sets of nine equations (one set for length-for-age and one set for stature-for-age), for each of the major smoothed percentiles, were solved
simultaneously to estimate the L, M, and S parameters for boys and girls separately. The nine final percentile curves for infants and children were predicted using the estimated L, M, and S values.
Length-for-age, birth to 36 months—The infant recumbent length-for-age curves for ages birth to 36 months were smoothed with a 3-parameter linear model:
f(t)= a+b * ln (t+1) + c(t+1)0.5
where f(t) is length in cm, t is age in months (calculated as midpoint of the age range), and a, b, and c are parameters to be estimated. The estimated values of f(t) are the smoothed values. At birth, t = 0. Estimates of a, b, and c are specific for each of nine percentiles being smoothed.
After excluding birthweights
<1,500 gm, the 3-parameter model was fit to the national data and birth length distributions from Wisconsin and Missouri. In contrast to data from the NCHS surveys that were pooled prior to curve smoothing, the Wisconsin and Missouri data were not pooled, but were treated as two separate sets of data points in the curve smoothing. Because the fitted values did not match other data sets well in the first few months of life, and because the smoothed point at
2.5 months was unstable due to insufficient sample size, data from a subset of the PedNSS at ages 0.5, 1.5, and 2.5 months were also included. Inclusion of these additional data at younger ages provided a better fit of the
3-parameter model in the first several months of age. To gradually merge the PedNSS data with the national survey data, PedNSS data were averaged with the national data at ages 3.5 and 4.5 months, after which the national data were used exclusively.
Stature-for-age—The stature-for-age curves for ages from 2 to 20 years were smoothed with a nonlinear model that ensured a monotonic increase in stature:
f(t)= aq/(1+e–b1(t–c1))+ ap/(1+e–b2(t–c2)
+ (f–a) / (1+e–b3(t–c3))
where f (t) is stature in cm, t is age in years (calculated as midpoint of the age range), and a, b1, b2, b3, c1, c2, c3, p, q, and f are parameters to be estimated
(p = 1–q). The estimated values of f (t) are the smoothed values. Estimates of a, b1, b2, b3, c1, c2, c3, p, q, and f are specific for each of nine percentiles being smoothed.
The empirical percentile curves derived from the national survey data were irregular partly due to sampling variations and the small number of subjects older than 18 years. To aid in smoothing the irregular empirical percentile curves at older ages, the empirical data were extended to age 25 years in 0.5-year intervals. For girls, empirical data for the 6-month age group from 1.5 to 2.0 years (midpoint 1.75) were also included to provide smoothed estimates closer to the empirical percentiles than the estimates obtained without inclusion of such data. Unadjusted recumbent length data were used as a proxy for stature in the age group from 1.5 to 2.0 years.
The infant head circumference-for age curves for ages birth to 36 months were smoothed with a 3-parameter linear model:
f (t)= a+b* ln (t+2) + c (t+2)0.5 where f(t) is head circumference in cm, t
is age in months (calculated as midpoint of the age range), and a, b, and c are parameters to be estimated. The estimated values of f(t) are the smoothed values. At birth, t = 0. Estimates of a, b, and c are specific for each of nine
percentiles being smoothed. After excluding birthweight <1,500 gm, the model was fit to the national data and head circumference at birth from the Fels Longitudinal Study.
A set of nine 3-parameter linear equations, one for each of the major smoothed percentiles, was solved simultaneously to estimate the L, M, and S parameters for boys and girls separately. The nine final percentile curves for infants were predicted using the estimated L, M, and S values.
Weight-for-Length and Weight-for- Stature
Similar to length-for-age and stature-for-age, weight-for-length and weight-for-stature required a smooth transition. Consequently weight-for length and weight-for-stature were developed together using a single regression model. Before fitting the model to the data, 0.8 cm was subtracted from each empirical
recumbent length data point. The shifted empirical weight-for-length data were smoothed simultaneously with the empirical weight-for-stature data using a 5-degree polynomial regression model.
The 5-degree polynomial regression model was fit to the national data and to weight-for-length data at birth from Wisconsin and Missouri. The birth length intervals from Wisconsin and Missouri began at 45 cm. However, there was a discontinuity between the State data and the national data perhaps due to lack of data between birth and 2 months. To avoid lowering the estimates for weight-for-length, data from 53 to 65 cm from the State or national data were omitted in the smoothing procedure.
The weight-for-length and weight-for-stature charts are age independent. Empirical percentiles of
weight were based on recumbent length or stature in 2-cm intervals. Although the weight-for-length chart was developed for use with infants from birth to 36 months, and weight-for stature was developed primarily for use with children from 24 to 60 months of age to maximize sample sizes, an upper age limit was not used in the development of either chart. Available weight-for-length data ranged from 45
Page 10 [ Series 11, No. 246
to 109 cm and available weight-for stature data ranged from 77 to 127 cm for calculating empirical percentiles. After the final curve smoothing, weight-for-length was truncated to 103 cm and weight-for-stature was
truncated to 121 cm on the final charts.
A set of nine 5-degree polynomial regression equations, one for each of the major smoothed percentiles for the combined weight-for-length and
weight-for-stature data, was solved simultaneously to estimate the L, M, and S parameters for boys and girls separately. The nine final percentile curves for infants were predicted using the estimated L, M, and S values. To obtain separate weight-for-length and weight-for-stature curves, 0.8 cm was added back to all the smoothed length values. This was accomplished by shifting back the corresponding length values for the length-specific weight, L,
M, and S, by 0.8 cm to form the weight-for-length measure from the weight-for-stature scale. The 85th percentile of weight-for-stature was calculated subsequently from the L, M, and S values.
Body Mass Index-for-Age
Empirical percentiles from the national data were smoothed with LWR. Ten empirical percentiles were calculated for the BMI-for-age charts because the additional 85th percentile was required for boys and girls to identify children and adolescents at risk for overweight. Each smoothed value was estimated by weighted linear regression on the five-neighborhood points adjacent to the value to be estimated from ages 2 to 12.5 years.
From 13 to 20 years, a 25-point smoothing procedure was used for boys and a 27-point smoothing procedure was used for girls. At the lower end (that is, age 2 years), two additional points were needed in the smoothing window, so a neighborhood point of 1.75 years was used for BMI. This was calculated using unadjusted recumbent length, repeated at
1.75 and at 1.71 years for both sexes. At the upper end (that is, age 20 years), the maximum BMI values in each empirical percentile from age 19.75 through 25.25 years were chosen and repeated in
0.5-year intervals from 20.25 through
25.75 years for boys or from 20.25 through 26.25 years for girls. Taking maximum values as additional data in smoothing the windows ensured that the BMI curves did not increase beyond the maxima at the upper ends of the age ranges. The smoothed percentile curves obtained through LWR were then fit by a 4-degree polynomial regression to achieve parametric percentiles. (See section on weight-for-age, 2 to 20 years, above, for further description of LWR.)
A set of 10, 4-degree polynomial regression equations, 1 for each of the major smoothed percentiles, was solved simultaneously to estimate the L, M, and S parameters for boys and girls separately. The 10 final percentile curves for infants were predicted using the estimated L, M, and S values.
The observed percentile distributions, along with means and standard deviations, are shown by sex and age in detailed tables 9–16. Selected smoothed percentiles (3rd, 5th, 10th, 25th, 50th, 75th, 90th, 95th, and 97th for all charts, and 85th for BMI-for-age and weight-for-stature) are shown in detailed tables 17–24 for the 16 age- and sex-specific growth charts. Also presented in these tables are the L, M, and S parameters that were used to create the final charts and are needed to generate additional percentiles and z-scores.
After the initial smoothing and transformation stages were completed, the percentile curves were evaluated. Transitions between the infant and older child charts were reviewed and each major percentile curve (3rd, 5th, 10th, 25th, 50th, 75th, 90th, 95th, 97th, and 85th for BMI-for-age) was compared graphically with the corresponding empirical data. In addition, the percent of empirical data below the smoothed percentiles was calculated and compared with the expected values. The objective of these procedures was to search for any large or systematic differences between the smoothed percentiles and the empirical data.
Evaluation of the transitions between the charts from birth to 36 months to the charts from 2 to 20 years indicated that disjunctions that were present in the 1977 NCHS charts were not present in the 2000 CDC charts. The final charts for length-for-age and stature-for-age were constructed simultaneously with an imposed parallel separation of 0.8 cm, based on the average difference between measured length and stature. Consequently, the smoothed percentile curves for
(figures 59–62), the empirical data, especially for values greater than or equal to the 75th percentile and during the adolescent years, showed a high degree of variability around the smoothed percentile lines.
Thus, comparisons were not possible for the new BMI-for-age charts, for the weight-for-stature charts above 120 cm of stature, for ages over 17 years on any chart, or for the 3rd and 97th percentiles on any chart.
Overall, the 1977 and the 2000 growth curves are quite similar. The differences that exist are more apparent in charts for infants where national survey data were previously lacking than in the charts for older children. For infants, the major differences were seen
in the head circumference-for-age charts. In the other infant charts, differences generally occurred at the outer (lower and upper) percentiles. For older children, there were almost no differences in the stature-for-age charts and only minor differences in the
When the 2000 charts are used there will be some differences in clinical
classifications relative to the 1977 charts. For example, with the 2000 charts, infants will be more often classified as underweight, less often classified as high weight-for-age, less often classified as having short length-for-age, less often classified as having small head circumference particularly after 6 months, and more often classified as having large head circumference.
The 1977 NCHS Growth Charts have been widely used in pediatric practice and public health for more than 20 years. Although some concerns about these growth charts arose, the charts had not undergone a major formal revision since they were developed. There was, however, an interim adjustment to the length-for-age and stature-for-age, and weight-for-length and weight-for-stature curves, designed for use in limited applications for the analysis of population-based survey and surveillance data (22,49). In 1985, when planning the sample design for NHANES III, NCHS began the growth
Page 12 [ Series 11, No. 246
chart revision process with the decision to oversample infants and young children for the purpose of revising the 1977 growth charts.
In the 2000 CDC charts, nationally representative survey data, supplemented with a limited amount of data from other sources, replaced the Fels data used in the 1977 NCHS infant growth charts from birth to 36 months.
Body mass index-for-age charts New sex-specific BMI-for-age charts for 2 to 20 years were developed to replace the 1977 NCHS weight-for-stature charts that were applicable only at prepubescent ages for statures for boys ranging from 90 to 145 cm and for girls ranging from 90 to 137 cm. Revised weight-for-stature charts are available for optional use from ages 2 to 5 years. Either BMI-for-age or weight-for-stature
charts may be used to assess risk of overweight from ages 2 to 5 years.
Extended age range
The revised charts were extended by 2 years beyond the 1977 NCHS charts to include children and adolescents from 2 to 20 years of age. This change was made to accommodate adolescents who continue to be seen by pediatricians through their later teenage years.
The 3rd and 97th percentiles were added to each chart to facilitate plotting data for children at extremes of the distribution. The major percentiles included in the 1977 NCHS charts (5th, 10th, 25th, 50th, 75th, 90th, and 95th) were retained in the revised charts. The 85th percentile was added to weight-for-stature and BMI-for-age charts, enhancing their use as screening tools to identify children and adolescents who may be
overweight or at risk of overweight.
Corresponding percentiles and z-scores
The 2000 CDC Growth Charts can be used to obtain percentiles and
z-scores. To meet the needs of researchers, for analyses of surveillance data, and to monitor changes in growth indicators for individuals, z-scores can be obtained and exact percentiles can be calculated.
Smooth junction between length and stature
The revised charts largely correct discontinuities that existed in the 1977 charts for infants and older children from 24 to 36 months. These disjunctions were the result of using data from different sources.
The 5th and 95th percentiles of the 1977 NCHS Growth Charts have been
used for many years as screening indicators, particularly for infants. These percentile cutoff values are used to help identify infants and children who are at increased nutritional and overall health risk.
Where there are differences between the 2000 and the 1977 curves, the percentile ranking of a child may differ.
In general, when the upper percentiles of the revised curves are higher than the corresponding percentiles of the 1977 curves, this will result in less frequent classification as high risk. When the lower percentiles of the revised curves are higher than the corresponding percentiles of the 1977 curves, classification as high risk will be more frequent. For example, the 90th and 95th infant weight-for-age percentiles of the revised curves are higher than the corresponding percentiles of the 1977 curves from approximately 6 to 36 months for boys, and from 12 to 36 months for girls. This shift would be expected to result in lower estimates for the prevalence of overweight when using the revised charts. In contrast, after approximately 6 months, all percentiles in the revised head circumference charts are lower than the corresponding percentiles in the 1977 charts. The expected impact for clinical screening or for population-based prevalence estimates, using the 2000 charts compared with the 1977 charts, is that children may be classified less frequently as having small head circumferences and may be more frequently classified as having large head circumferences. Other examples are evident in the detailed figures in this report that compare the 2000 percentile curves with the 1977 percentile curves.
Children of all major racial-ethnic groups appear to have similar growth potential. Studies have demonstrated that genetic effects on growth are small compared with the effects of the environment, nutrition, and health.
Regardless of racial-ethnic status, children provided with good nutrition, access to health care, and good social and general living conditions have similar growth patterns (13,50–55).
The 1977 NCHS Growth Charts for infants were based entirely on data from the Fels Research Institute collected 1929–75. In this study population, as in the general population during this period, infants were mainly formula-fed, although a small percent were exclusively breast-fed for 3 months. In addition, it is possible that infants in the Fels Longitudinal Study were supplemented early with solid foods, which could have an impact on growth that would be reflected in growth charts constructed from these data (56).
The 2000 CDC Growth Charts include data for formula-fed and breast-fed infants, proportional to the distribution of breast- and formula-fed infants in the population. During the past two decades, approximately
one-half of all infants in the United States received some breast milk and approximately one-third were breast-fed for 3 months or more (57). By replacing data from the Fels Research Institute with national survey data collected from
1970 to 1994, the combined growth patterns of breast-fed and formula-fed infants in the U.S. population are represented.
Breast-feeding is recommended as the feeding method of choice for most full-term infants during the first 6 months after birth and should be continued with the addition of solid foods through the first 12 months (58). The association between breast-feeding and size and growth in infants has been the topic of many investigations. In general, breast-fed infants tend to gain weight more rapidly in the first 2–3 months. From 6 to 12 months breast-fed infants tend to weigh less than
formula-fed infants (18,59). These observations have led to recommendations for the development of new infant growth references based on healthy breast-fed infants (18,19,60,61). A study sponsored by the WHO is currently collecting data at study centers in six countries. These data will be used to develop a new international growth reference.
Body Mass Index-for-Age Growth Charts
Body mass index (BMI; kg/m2) has been recommended for use in children, adolescents, and adults to assess weight, adjusted for stature (34,35,62). BMI tracks over time and the tracking tends to increase at older adolescent
ages (64,65). By age 8 years, however, most children are in the percentile range they will follow until the end of
The pattern of the BMI curves indicates that BMI increases rapidly from birth to approximately 8 months of age, then decreases until approximately age 6 years, when it reaches its nadir before rebounding, or increasing once again (67). This has been termed the adiposity rebound, because it is believed to be the age when body fatness begins to increase after reaching a minimum.
This pattern is evident in the revised growth charts. The younger the age at which a child’s BMI curve is at its nadir, the greater the likelihood that the level of adiposity will be high in adolescence and early adulthood (66,68,69). Children at higher percentiles for BMI tend to achieve their adiposity rebound at
younger ages. These children have a greater likelihood to track at a higher BMI percentile with increasing age.
0.3 units are preferred to intersecting line nomograms because they provide greater precision. In clinical assessment, a transition from the weight-for-length to the BMI-for-age charts from 24 to 36 months of age becomes possible when children can stand unassisted and adequately follow directions to assume the correct posture for a stature measurement.
The revised growth charts for the United States include data on low birthweight infants but do not include data on very low birthweight infants (VLBW; <1,500 gm). Alternate charts are available to assess the growth of VLBW infants. Perhaps the most recent are those developed from data collected in the National Institute of Child Health and Human Development Neonatal Research Network Centers (70). This recent reference was developed from data on 478 infants who were appropriate-for-gestational age and survived to discharge. Prospective growth curves are plotted by 200 gm birthweight intervals for birthweights ranging from 501 to 1,500 gm.
However, these charts only extend to about 120 days uncorrected postnatal age or until a body weight of 2000 gm. These charts have a very limited range of use. Perhaps the best general reference available for VLBW infants is the Infant Health and Development
A variety of health conditions such as Down syndrome, cerebral palsy, Turner syndrome, and others affect growth status. There are specialized charts that may be considered for use with children affected by these conditions (74–76). These specialized growth charts provide useful growth references, but may have some limitations. Generally, they are developed from relatively small samples. One option is to plot the growth patterns of these children on the specialized charts and the CDC growth charts. This will allow comparisons of growth to the general population of children and to the references for children identified with a given condition. In most cases, BMI charts are not available for special conditions, and the CDC charts would provide a useful reference to monitor weight in relation to stature.
The distinction between a growth reference and a growth standard is relevant to the development and application of growth charts. The WHO working groups (61,77) have defined a reference as a tool for providing a common basis for purposes of comparison, and a standard as embodying a concept of a norm or target, that is, a value judgment. In simple terms, a reference describes ‘‘what is,’’ whereas a standard prescribes ‘‘what should be.’’ In practice, however, reference values are often used as a standard (78). Growth references are intended to be used to screen and monitor growth in individuals and populations. They are not intended to be the sole independent diagnostic instruments upon which clinical decisions are made.
In the development of the 1977 NCHS Growth Charts, the task force adhered strictly to a policy of no data adjustments (8). In the development of the 2000 CDC Growth Charts, two notable data exclusions were made. The first exclusion was of data for all VLBW infants and the second was of body weight observations for NHANES III subjects ages 6 years or over. The exclusion of VLBW infants was based on the fact that alternative growth references are available and are appropriate to monitor the unique growth patterns of VLBW infants.
The exclusion of NHANES III body weight data for children and adolescents was made after consultation with numerous experts. Although this meant straying from a pure reference, the advantages of excluding these data were judged to outweigh the disadvantages.
Had these data been retained, the increase in body weight among NHANES III subjects would have shifted the weight- and BMI-for-age percentile curves upward. This would have led to identifying fewer children
and adolescents as overweight or at risk of overweight. The revised U.S. growth charts are still intended to be growth references.
In addition to data exclusions, there were some minor data adjustments and inclusions of data that were not nationally representative. The length data were adjusted to integrate them with stature data in the smoothing process. State data were used for length at birth and weight-for-length at birth, Fels data were used for head circumference at birth, and PedNSS data were used for length-for-age from 0.5 to
5.0 months. These approaches were taken to provide more complete and accurate charts, and do not diminish their use as growth references based on well-defined data sources.
There is a difference between growth (or size) charts and growth velocity charts (8). The 2000 CDC Growth Charts for the United States are based primarily on cross-sectional national survey data that were statistically smoothed to create percentile curves. The curves were superimposed on grids that permit interpretation of an individual’s attained body size (weight, length or stature, and head circumference) at given ages, or weight at given lengths or statures, in comparison with the body sizes of children in the reference population.
Therefore, these charts more appropriately may be considered size charts. When serial values for an individual are plotted, assessments can be made of that individual’s growth progress over time.
Growth velocity charts are constructed from incremental data obtained from longitudinal observations. Growth velocity charts are more sensitive indicators of small changes in growth status than the size-attained charts, and are more useful when evaluating changes in growth rates that are important in selected growth disorders and therapies. Incremental growth charts and growth tables for
Series 11, No. 246 [ Page 15
6-month increments have been developed using weight, recumbent length, stature, and head circumference data from the Fels Longitudinal
z-scores are used to evaluate the growth of children. Percentiles are the most commonly used clinical indicator to assess the size and growth patterns of individual children in the United States. Percentiles rank the position of an individual by indicating what percent of the reference population the individual would equal or exceed. For example, on the weight-for-age growth charts, a
5-year-old girl whose weight is at the 25th percentile, weighs the same or more than 25 percent of the reference population of 5-year-old girls, and weighs less than 75 percent of the
Which one is used is based primarily on convention or preference. In certain population-based applications, such as research settings and surveillance systems, the mean and standard deviation are often calculated for a group of z-scores (61). In selected clinical situations where growth monitoring is an important evaluation tool and greater measurement precision is necessary, z-scores or exact percentiles may be preferred by clinicians.
To track growth below the 5th percentile, z-scores achieved widespread use. A z-score of –2 SD is accepted as a standard statistical cutoff point to determine the need for nutritional intervention and corresponds approximately to the 3rd percentile
(z-score at 3rd percentile = 1.88) (77). Because the LMS values were calculated by solving equations that used the values for percentiles ranging
from the 3rd to the 97th, the outermost percentiles less than the 3rd or greater than the 97th are beyond the range of the data. The sample sizes for the data used to calculate the growth charts were not adequate to calculate percentiles below 3 and above 97, corresponding to z-scores of –1.88 and 1.88.
Extrapolation beyond this range of the data should be done with caution. In addition, a solution may not be obtained when both the Z and L are high and have opposite signs. Moreover, some
core values may have extreme corresponding measurement values. For these reasons, any use of the LMS values to calculate z-scores below
he purpose of this revision to the 1977 NCHS Growth Charts was to provide better estimates of size
and growth, using more comprehensive national survey data and improved statistical smoothing procedures. To construct clinically useful growth charts, it is necessary to have access to a reference population, to obtain anthropometric data, and to statistically smooth the observed data. In developing such reference data for children, it is not feasible to conduct a study that would obtain measures on the total U.S. population of children and adolescents. Instead, the reference population for the revised U.S. growth charts was primarily based on statistically representative samples of the U.S. pediatric population, measured in a series of cross-sectional surveys from 1963 to 1994, that were supplemented with limited data from other sources.
Replacing the Fels data in the 1977
NCHS infant charts with national survey data addressed one of the major concerns with the 1977 charts.
National survey data, in combination with improved statistical smoothing procedures, produced curves with better transitions between the infant and older child curves from 24 to 36 months of age than were found in the 1977 curves. The statistical procedures used to generate the 2000 growth charts will allow users to calculate percentiles and z-scores that are completely interchangeable. This is an improvement over the 1977 NCHS charts where they often did not agree.
A major shortcoming in the
1977 NCHS charts was the inability to calculate a weight-for-stature value
for post-pubertal adolescents. Overweight in children and adolescents is becoming more common in the United States (81– 83), and screening tools are needed to identify individuals and population groups at risk. Expert committees have recommended use of the BMI to identify children who are overweight or at risk of becoming overweight (34,35). Development of the new BMI-for-age growth chart for boys and girls from 2 to 20 years of age provides a screening instrument to monitor weight adjusted for stature.
The 2000 CDC Growth Charts for the United States are intended to serve as a reference to evaluate physical size and growth for the majority of the pediatric population. For preterm, VLBW infants and children with specific conditions that may affect size and growth, various charts are available for alternative use, although most of these were developed from restricted samples that were not based on nationally representative survey data.
Differences between the 1977 and the 2000 growth charts have been illustrated in this report. In clinical practice most of these differences will not affect the way in which current evaluations are made. Substitution of the 2000 CDC charts for the 1977 NCHS charts in research, surveillance, food assistance, and other population-based activities may affect the classification of some individuals.
The revised growth charts should continue to meet the many uses that
CDC has made available camera-ready negatives to State health offices, and copies are also available through some infant formula manufacturers and professional organizations such as the American Academy of Pediatrics.
Various proprietary software programs are available that incorporate the 2000 CDC Growth Charts in electronic formats, and CDC has included the normalized 1977 and the 2000 growth charts in the NutStat module of Epi Info 2000.
As the 2000 CDC Growth Charts are used in various applications, they will continue to be evaluated. Through this process, additional knowledge will be gained to give direction for future improvements in pediatric growth charts for the United States.
Waterlow JC, Buzina R, Keller W, et al. The presentation and use of height and weight data for comparing the nutritional status of groups of children under the age of 10 years. WHO Bulletin 55:489–98. 1977.
Roche AF, Johnson CL, Kuczmarski RJ, Briefel RR. Revision of the U.S. National Center for Health Statistics growth charts. In: Bodzsar BE, Susanne C, eds. Studies in Human Biology. Budapest: Eotvos University Press, 105–12. 1996.
Victora CG, Morris SS, Barros FC, et al. The NCHS reference and the growth of breast- and bottle-fed infants. J Nutr 128:1134–8. 1998.
WHO Working Group on the Growth Reference Protocol. A growth curve for the 21st century: The WHO multicentre
growth reference study (protocol). Geneva: World Health Organization. 1998.
Mei Z, Yip R, Grummer-Strawn LM, Trowbridge FL. Development of a research child growth reference and its comparison with the current international growth reference. Arch Pediatr Adolesc Med 152:471–9. 1998.
Binns HJ, SenturiaYD, LeBailly S, et al. Growth of Chicago-area infants, 1985 through 1987. Arch Pediatr Adolesc Med. 150:842–9. 1996.
de Onis M, Yip R. The WHO growth chart: Historical considerations and current scientific issues. Bibliotheca Nutritio et Dieta 53:74–89. 1996.
Garza C, de Onis M. A new international growth reference for young children. Am J Clin Nutr 70(supplement): 169s–72s. 1999.
McDowell A, Engel A, Massey JT, Maurer K. Plan and operation of the second National Health and Nutrition Examination Survey, 1976–80. Vital Health Stat 1(15). 1981.
National Center for Health Statistics. Plan and operation of the third National Health and Nutrition Examination Survey, 1988–94. National Center for Health Statistics. Vital Health Stat 1(32). 1994.
National Center for Health Statistics. Plan, operation, and response results of a program of children’s examinations. National Center for Health Statistics. Vital Health Stat 1(5). 1967.
National Center for Health Statistics. Plan and operation of a health examination survey of U.S. youths 12–17 years of age. National Center for Health Statistics. Vital Health Stat 1(8). 1969.
National Center for Health Statistics. Plan and operation of the Health and Nutrition Examination Survey, United States 1971–73. National Center for Health Statistics. Vital Health Stat 1(10). 1973.
National Center for Health Statistics. Plan and operation of the second National Health and Nutrition Examination Survey, 1976–80. National Center for Health Statistics. Vital Health Stat 1(15). 1981.
U.S. Department of Health and Human Services, PHS. National Center for Health Statistics. NHANES III Reference Manuals and Reports
(CD-ROM). Hyattsville, Maryland: Centers for Disease Control and Prevention. 1996.
Series 11, No. 246 [ Page 17
Lohman TG, Roche AF, Martorell R, eds. Anthropometric standardization Reference manual. Champaign, IL: Human Kinetics Books. 1988.
Roche AF. Growth, maturation, and body composition: The Fels Longitudinal Study 1929–1991. Cambridge: University Press. 1992.
Guo SS, Roche AF, Chumlea WC, et al. Statistical effects of varying sample sizes on the precision of percentile estimates. Am J Hum Biol 12:64–74. 2000.
Roche AF, Guo S, Woteki C, Trowbridge FL. Methods for the revision of the NHCS/CDC growth charts: Birth to 36 months. 1989 proceedings of the section on survey research methods, American Statistical Association, Alexandria, VA. 1989.
Habicht JP, Martorell R, Yarbrough C, et al. Height and weight standards for preschool children: How relevant are ethnic differences in growth potential? Lancet 1:611–5. 1974.
Mendoza FS, Castillo RO. Growth abnormalities in Mexican-American children in the United States: The National Health and Nutrition Examination Survey I study. Nutr Res 6:1247–57. 1986.
Malina RM, Martorell R, Mendoza F. Growth status of Mexican American children and youths: Historical trends and contemporary issues. Yearbook of Physical Anthropology 29:45–79. 1986.
Martorell R, Mendoza F, Castillo R. Poverty and stature in children. In: Linear growth retardation in less developed countries. Waterlow JD, ed. NewYork: Raven Press Ltd. 1988.
Martorell R, Mendoza F, Castillo R. Genetic and environmental determinants of growth in Mexican Americans. Pediatrics 84:864–71. 1989.
Yip R, Scanlon K, Trowbridge F. Improving growth status of Asian refugee children in the United States. JAMA 267:937–40. 1992.
Ahn CH, MacLean WC. Growth of the exclusively breast-fed infant. Am J Clin Nutr 33:183–92. 1980.
National Center for Health Statistics. Health, United States, 2000 With Adolescent Health Chartbook. Hyattsville, Maryland. 2000.
Committee on Nutrition, American Academy of Pediatrics. Breastfeeding. In: Pediatric nutrition handbook. 4th ed. Elk Grove Village, IL: American Academy of Pediatrics. 1998.
Dewey KG. Cross-cultural patterns of growth and nutritional status of breast-fed infants. Am J Clin Nutr 67:10–7. 1998.
de Onis M, Garza C, Habicht JP. Time for a new growth reference. Pediatrics 100(5):e8. 1997.
Ehrenkranz RA, Younes N, Lemons JA, et al. Longitudinal growth of hospitalized very low birthweight infants. Pediatrics 104:280–9. 1999.
Guo SS, Wholihan K, Roche AF, et al. Weight-for-length reference data for preterm, low birth weight infants. Arch Pediatr Adolesc Med 150:964–70. 1996.
Guo SS, Roche AF, Chumlea WC, et al. Growth in weight, recumbent length, and head circumference for preterm
low-birthweight infants during the first
Page 18 [ Series 11, No. 246
three years of life using gestation- adjusted ages. Early Hum Dev 47:305–25. 1997.
Roche AF, Guo SS, Wholihan K, Casey PH. Reference data for head circumference-for-length in preterm
low-birthweight infants. Arch Pediatr Adolesc Med 151:50–7. 1997.
Scott BJ, Artman H, Hill LA. Monitoring growth in children with special health care needs. Top Clin Nutr 13:33–52. 1997.
Agwu JC, Shaw NJ, Chapman S, et al. Growth in Sotos syndrome. Arch Dis Child 80:339–42. 1999.
Collins CE, MacDonald-Wicks L, Rowe S, et al. Normal growth in cystic fibrosis associated with a specialized centre. Arch Dis Child 81:241–6. 1999.
WHO Working Group. Use and interpretation of anthropometric indicators of nutritional status. Bull World Health Org 64:929–41. 1986.
de Onis M, Habicht JP. Anthropometric reference data for international use: Recommendations from a World Health Organization expert committee. Am J Clin Nutr 64:650–8. 1996.
Roche AF, Himes JH. Incremental growth charts. Am J Clin Nutr 33: 2041–52. 1980.
Baumgartner RN, Roche AF, Himes JH. Incremental growth tables: Supplementary to previously published charts. Am J Clin Nutr 43:711–22. 1986.
Flegal KM, Ogden CL, Wei R, et al. Prevalence of overweight in U.S. children: Comparison of U.S. growth charts from the Centers for Disease Control and Prevention with other reference values for body mass index. Am J Clin Nutr 73:1086–93. 2001.
National Center for Health Statistics. Health, United States, 2001 with Urban and Rural Health Chartbook. Hyattsville, MD 2001. URL: http://www.cdc.gov/nchs/
Casey PH, Kraemer HC, Bernbaum J, et al. Growth status and growth rates of a varied sample of low birthweight, preterm infants: A longitudinal cohort from birth to three years of age. J. Pediatr 119:599–605. 1991.
Troiano RP, Flegal KM, Kuczmarski RJ, et al. Overweight prevalence and trends for children and adolescents: The National Health and Nutrition Examination Surveys, 1963 to 1991. Arch Pediatr Adolesc Med 149:1085– 91. 1995.