Several concerns have been raised regarding the adequacy of currently existing growth references, including those developed by the National Center for Health Statistics (NCHS) (1) and currently supported by the World Health Organization (WHO) (2). The following major problems have been detected with the NCHS-WHO reference: predominance of formula-fed babies, nonrepresentativeness and excess homogeneity of data between 0 and 2 years, unavailability of monthly measurements during early infancy, small sample sizes, and unavailability of repeated measurements on the same children, outdated curve-fitting procedures, and length–height disjunction at 24 months. (3–5). These concerns resulted in the WHO (3,4) recommendation to replace the NCHS-WHO reference data. As an interim step, the WHO Working Group on Infant Growth published references (“interim” WHO references for breast-fed infants) (3,6) that were based on data from infants who were fed according to the present WHO recommendations (7). Furthermore, WHO is collecting data to develop a new international growth reference (8). NCHS is going to revise the NCHS growth references but still will rely on cross-sectional data from predominantly formula-fed infants. This study presents the Euro-Growth references for recumbent length, weight, and selected circumferences for children between 1 and 36 months of age, which are based on a large and well-defined sample of the European population (9,10).
The Euro-Growth references were constructed in such a way that the calculations of z-scores (11) for the age range between 1 and 36 months was possible for the following anthropometric measurements: recumbent length; weight; head; and mid-upper arm (MUAC), thigh, and calf circumferences. A preview of the Euro-Growth references has already been published (9). This article describes the process of the construction of sex-specific growth references and compares the Euro-Growth references with the NCHS-WHO references (1,2), the recently published WHO interim references for breast-fed infants (3,6), references proposed by Fomon (12), and the Netherlands growth reference (13). The relations of growth with demographic variables such as region (study center), mid-parental height, birth order, birth length, and birth weight are also presented.
MATERIALS AND METHODS
The Euro-Growth Study was designed as a multicenter, longitudinal, observational study using standardized methodology. (9,10). Healthy term infants (gestational ages between 37 and 44 weeks) without signs of intrauterine growth pathology and who did not meet the other exclusion criteria (10) were enrolled before 30 days of age. The cohort consisted of 2245 subjects (1154 boys and 1091 girls) from 22 study centers in 12 European countries. Information regarding parental weight and height (reported), educational achievement, family demographics, and socioeconomic data, and infant characteristics at birth was obtained at enrollment. Information about the subject's nutrition and about family lifestyle was obtained periodically. Anthropometric measurements were made at precisely defined age intervals between 1 and 36 months of age (Table 1) using standardized methodology (14–16). From 18 through 36 months of age, 10 sites performed measurements of standing height (n = 2423) in addition to recumbent length.
Construction of Age-and Sex-Specific References
The number of observations available for the construction of the Euro-Growth references is shown in Appendix 1. This included observations made in children who subsequently withdrew from the study. For reasons detailed previously (10), the data from one center (at enrollment, n = 100; at 36 months of age, n = 67) were not used for the construction of growth references, leaving 2145 children, who were measured on 20,696 occasions. Not all measurements were recorded within the age ranges specified by the protocol. The age ranges from which measurements were used for the construction of the growth references are presented in Table 1. At various ages, between 96% and 98% of measurements occurred within these boundaries.
The data in Table 2 show the differences between recumbent length and standing height measurements in 10 sites, with recumbent length being approximately 5 mm larger. Eight sites provided consistent quality of measurements, which is indicated by low standard deviations (SD) of the differences (0.24–0.86 cm). The variability in mean differences (0.14–0.80 cm) may be related to the stretching force that was applied during measurements. The two sites with the smallest sample sizes showed less consistent quality of measurements. However, the length measurements of the two sites showed sufficient quality during the longitudinal and cross-sectional data checks (10). Standing height references were not constructed. If measured, standing height is to be interpreted on the basis of recumbent length references, 5 mm must be added to value for standing height.
Because all measurements were made within narrow ranges around the target ages, the method described by Wright and Royston (11) was used. We preferred to apply a uniform approach for the construction of the six age-and sex-specific references. The method applied (11) specifies that three decisions have to be made after answering the following questions:
• Is logarithmic transformation of data necessary in the calculation of the initial z-scores? Results of a pilot study indicated that log transformation was not beneficial.
• What time basis is to be used in the smoothing polynomials? The time basis is usually chosen to be age, ln(age), or sqrt(age). Results in a pilot study suggest that sqrt(age) is satisfactory for the time basis. For the readability of the polynomial regression coefficients and to avoid multicolinearity of age powers, in practice, sqrt(age in years) - 1 was used as the time basis in the smoothing procedures.
• What degree of smoothing polynomials is to be used? Five degrees was chosen.
It was not considered necessary to test the degree of the polynomials by formal hypothesis testing. The possibility of overparameterization was judged by visual inspection of the reference curves. These choices may not be optimal in a broad perspective, but they served the intention of the growth references, which was to fit close to the raw percentiles, while preserving the normal distribution, which is important for the proper interpretation of z-scores. According to the method of Wright and Royston (11), the further calculations included six steps applied separately for each gender:
• Mean, SD, and skewness for each target age were calculated.
• Fifth-degree polynomials for target ages were fitted for means (m), SD (s), and skewnesses (k).
• Per target age, the Manly transformation was applied to obtain normality, using the smoothed skewness (k) as transformation parameter [kappa].
• For the transformed values, corrected means and SD per target age were calculated.
• For the corrected means (g) and SD (d) fifth-degree polynomials were fitted.
• Those five fifth-degree polynomials (m, s, k, g, and d) defined the reference curves and allowed the calculation of z-scores.
The results of the calculations were checked against selected raw percentiles (P): P3, P5, P10, P25, P50, P75, P90, P95, and P97. The initially obtained curves did not precisely fit the extreme raw percentiles, because of a few strong outliers, which influenced skewness and SD. This problem was overcome by a slight degree of winsorization—that is, strong outliers were made less extreme, so that they no longer influenced the P3 or P97 values (17). The winsorization was used temporarily, only for the calculation of the polynomials (m, s, k, g, and d). This action may be seen as part of the smoothing procedure, which in the end does not provide z-scores that have a mean of exactly 0 and SD of exactly 1. Minor deviations have to be accepted in smoothing, which is necessary to enable the calculation of z-scores in ages between the target ages.
Although the construction of the references used a strictly cross-sectional approach, the data were not statistically independent, because of the longitudinal data collection. Compared with purely cross-sectional data, longitudinal data require less smoothing but have larger standard errors after smoothing. This aspect of smoothing is not well documented in the theoretical statistical literature.
Comparisons with published references (1,2,6,12,13) were made by using the z-score technique and/or by comparing selected percentiles. For comparison with the NCHS reference (1) the Centers for Disease Control (Atlanta, GA, U.S.A.) anthropometric software package (18) was used.
Influence of Sex, Site, and Genetic Factors
Sex differences were evaluated using Student's t-test. Regression analysis of means and SD of z-scores against age assessed the effects of sites on anthropometric variables. The average level of the z-score of a site was chosen to be the estimated value at the age of 12 months. The change in mean z-score with age (slope) would be close to zero if the rate of growth corresponded to the overall growth of the European cohort. If the slope were positive, growth would be faster, if the slope were negative, growth would be slower than that of the overall European cohort. Homogeneity within sites was evaluated by analyzing the behavior of the SD of the z-scores in a manner similar to that applied to the mean z-scores. An average level of SD equal to 1 implies a homogeneity that corresponds to that of the overall European cohort, whereas a level of SD less than 1 indicates greater homogeneity, and a level of SD more than 1 indicates lower homogeneity. The slope of the SD indicates change in homogeneity with age.
Correlation and regression analyses were used to study the influence of parental height on infant length. The influence of mid-parental height, length at birth, and site were assessed for selected ages using multiple regression, analysis of variance, and analysis of covariance of the z-scores. Pearson's correlation coefficients provided information on the predictive power of length and weight at birth, 1 month, and 3 months of age on the respective values at 12, 24, and 36 months of age. The effect of birth order was studied by analysis of covariance of the z-scores for length, correcting for site influences and mid-parental height.
Characteristics of the references
Selected sex-specific raw percentile values, means, and SD for length (Table 3), weight (Table 4), and body circumferences (Appendix 2: head circumference; Appendix 3: MUAC; Appendix 4: thigh circumference; and Appendix 5: calf circumference) are presented. Of specific interest are the 3rd and 5th age-and sex-specific percentiles for length, weight, and MUAC, because those indicators are used as cutoff values to define suspected failure to thrive.
Influence of Sex
Mean values for weight, length, and head circumference were significantly larger in boys than in girls between 1 and 36 months of age (Table 5). The difference in mean weight increased until 12 months of age and decreased thereafter, whereas the difference in length increased until 6 months of age and decreased thereafter. Differences in mean MUAC and calf circumferences between sexes were small and no longer significant at 30 and 36 months of age. Girls had significantly higher mean thigh circumferences between 24 and 36 months of age.
The smoothed Euro-Growth percentile (P) curves (P3, P5, P10, P25, P50, P75, P95, and P97 values) for all anthropometric indexes are presented in Figures 1 through 9. Figure 10 shows the smoothed percentiles for length with the raw percentile values superimposed and indicates a close fit of the smoothed to the raw percentiles. This figure shows absence of over-or underparameterization by the fifth-degree polynomials used in constructing the smoothed percentiles. A numerical impression of the fit of the smooth polynomials to the raw percentiles is presented in Appendix 6. The deviations were smallest for head circumference (<0.8%) and length (<1.2%). Because of smoothing and winsorization, the z-scores did not equal a mean of 0 and SD of 1. Table 6 shows that the mean z-scores and SD z-scores were very close to their intended values (0 and 1, respectively).
Comparison With Published References
The Euro-Growth references for length and weight were also compared with the NCHS-WHO references (1,2) by expressing five selected Euro-Growth z-score values (-2, -1, 0, +1, +2) as NCHS z-scores (18). Because the NCHS references apply to standing height between 24 and 36 months of age, 5 mm was subtracted from the Euro-Growth length references for this comparison. As Table 7 shows, the NCHS and Euro-Growth 0 z-scores for length were similar, but in Euro-Growth the range between -2 and +2 z-scores was narrower. The differences in the references for height between 24 and 36 months of age are obvious and were related to the length–height disjunction of the NCHS references (3–5). As Table 8 shows, the NCHS references for weight assign markedly different z-scores at essentially all ages. For example, the -2 z-scores of boys and girls, which are widely used as cutoff values for failure to thrive, correspond to NCHS z-scores between -1.49 and -1.03 during the first 6 months of age. Part of the differences between the two sets of references during early infancy may be related to enrollment and exclusion criteria (10), because the length and weight distributions of the Euro-Growth references were somehow narrower at birth.
The Euro-Growth references for length and weight were also compared with the WHO interim references for breast-fed infants (6) by expressing the Euro-Growth z-score values as WHO interim z-scores (Appendix 7). Data on birth length were not available for the WHO interim references. The birth–weight distribution of boys and girls of the WHO interim reference was narrower (6) than the distribution of the Euro-Growth cohorts. The z-scores for length of both sexes indicated that the Euro-Growth infants were shorter between 1 and 4 months of age and taller between 5 and 12 months of age than the infants of the WHO interim reference. The z-scores for weight indicated that the Euro-Growth infants were lighter between 1 and 4 months of age and heavier between 5 and 12 months of age.
Appendix 8 presents a comparison (z-scores) of the length and weight references with the data from the United States collected by Fomon and Nelson (12), which were derived from term infants in Iowa (0 to 6 months of age), and data collected in the Fels longitudinal study (1,12). The Euro-Growth z-scores were expressed as (calculated) “Fomon”z-scores. At birth, the boys and girls of the Euro-Growth cohort were 1.4 cm and 0.4 cm, respectively, smaller than the infants in Fomon and Nelson. Means for length of boys were similar until 12 months of age, but those for girls were higher than those of the data of Fomon and Nelson. Weight of both sexes was slightly lower between 1 and 4 months of age and was higher thereafter than indicated by the Foman and Nelson data (12).
Appendix 9 presents a comparison (z-scores) of the length and weight references with the recent Dutch references (13). Dutch boys were consistently taller, and comparison of weight indicated that until 9 months of age Dutch boys and girls were lighter. They tended to be heavier between 12 and 36 months of age.
Influence of Site
Appendix 10 summarizes the results (length, weight) of the regression analysis of z-scores with age for site-specific effects. It is evident that there were substantial differences between sites in z-score levels. Homogeneity was reasonably uniform across sites. Differences between sites reflect, among other influences, the effects of genetic and cultural factors.
Genetic Influences and Effect of Size at Birth and Birth Order
Parental height and weight can be seen as markers of genetic influences. Table 9 shows correlations of z-scores for length and weight with maternal, paternal, and mid-parental height, and weight. Maternal height was a better predictor of child length and weight than paternal height. As expected, mid-parental height was the best predictor of child length. The predictive power increased until 2 years of age. Ten percent of the variance in length was explained by mid-parental height at 1 month of age, and 21% of the variance in length at 24 months of age. Table 10 presents the regression coefficients of mid-parental height on childhood length at selected ages. At 1 month of age, each centimeter of mid-parental height added, on average, between 1.0 and 1.1 mm to the infant's length. With age, the effect of mid-parental height increased gradually, so that at 36 months of age, each centimeter of mid-parental height added, on average, between 2.7 and 2.8 mm to the child's length.
The power of length and weight at birth and at 1 and 3 months of age to predict anthropometric indexes at 12, 24, and 36 months of age is indicated in Table 11. The predictive power of birth length for length between 12 and 36 months of age was higher (r = 0.32–0.44) than of birth weight for weight between 12 and 36 months of age (r = 0.21–0.35). Length and weight at 3 months of age had the highest predictive power for subsequent anthropometric indices.
Table 12 indicates the relative strength (explained variance) of the different influences on the variance of z-scores for length. As expected, the influence of birth length decreased with age. At 1 month of age, birth length was the single best predictor and explained 47% of the variance in length. The addition of study site and mid-parental height increased the predictive value to 57%. At 36 months of age mid-parental height was the single best predictor (19% of variance explained). Analysis of covariance (Appendix 11) indicated that the site-specific differences in length (raw data) tended to become smaller when correction for mid-parental height was applied (adjusted data). However, site-specific differences remained substantial, indicating that site-specific factors other than those expressed by mid-parental height (e.g., nutrition) play an important role.
Until 24 months of age, birth order had no significant effect on z-scores for length. However, at 30 and 36 months of age, the first-born children were significantly taller than the second-and third-born children, even after correcting for mid-parental height and site (Table 13). Although the effect of birth order was statistically significant, it was considered small enough to be omitted in further analyses, such as that of the influence of early nutrition on growth (19).
There are several choices for construction of anthropometric growth references (20). Those methods differ in the transformation to normality, the smoothing procedure, and the way the data are clustered into homogeneous age groups. The least mean square method (LMS) developed by Cole (21) is an advanced, flexible technique that has found many applications in anthropometry. For this study we decided to apply the method described by Wright and Royston (11). The reason was that this method did not require special software and that the Euro-Growth design, based on a narrow age band around the target ages, was easy to analyze with respect to skewness, without the need to create artificial age groups. The Manly [kappa] transformation is identical with the Box–Cox transformation after an initial log transformation; therefore both methods (11,21) are very similar. In our study the calculated polynomials represented the raw percentiles very well. The smoothed references for length, weight, and head circumference did not show signs of over-or undersmoothing. The smoothed references for MUAC and thigh and calf circumferences showed minimal flaring. We therefore calculated lower order or fractional polynomials (22), but this did not eliminate all flaring.
In 1977 a WHO working group on the use of anthropometric indicators (23) pointed out that a reference population should meet the following criteria: the measurements should be drawn from a well-nourished population, the sample should include at least 200 individuals in each age and sex group, the sampling procedures should be defined and reproducible, the measurements should be carefully taken and recorded by observers trained in anthropometric techniques and using equipment of well-tested design calibrated at frequent intervals, the measurements should include all the anthropometric variables necessary for the evaluation of nutritional status, and the sample should be cross-sectional, because it will be used for cross-sectional comparisons. Between 1995 and 1999, WHO Expert Committees (3,4,6,8) modified those criteria to include several countries from different geographical regions, to make the raw data available, and to take into account longitudinal rather than cross-sectional data collected during the first 2 years with frequent assessment of feeding practices. The Euro-Growth references fulfill all WHO criteria and can be compared with existing (1,2,6,12,13) and upcoming references (8).
The existing NCHS references (1,2) for children between 0 and 23 months of age were based on a group of children in the Ohio Fels Research Institute longitudinal study from 1929 through 1975. This reference reflects the growth of children who were fed primarily with (outdated) infant formulas, faced different (infectious) diseases that are rare today, and were of restricted genetic, geographic, and socioeconomic backgrounds. Measurements during early infancy were not taken at monthly intervals. It is therefore not surprising that the Euro-Growth references indicate that the infants and children are now taller than the NCHS references. The -2 z-scores for length and weight for age are of particular interest. If the NCHS -2 z-score values had been used as a cutoff (anthropometric indicator), a substantial part of infants and small children in this study with failure to thrive would have been classified as normal. Differences in birth length and birth weight distribution, feeding practices (3,6,19,24), health status, and the interpolation of the NCHS data between 0 and 6 months may be accountable in part for the differences observed during the first months. The length–height-for-age disjunction of the NCHS references at 24 months has been described to correspond to 0.5 z-scores (1,3,5,6,8). The disjunction was at 0.56 and 0.52 z-scores when the NCHS references for boys and girls were compared with the respective Euro-Growth references, which were corrected for the differences between recumbent length and standing height.
Comparison of z-scores with the existing WHO interim reference for breast-fed infants (3,6), which describes growth of infants fed according to the present WHO feeding recommendations (7), indicated lower length and weight during the first 4 months of life. The biggest difference in z-scores for length was found at 1 month of age, at which the WHO interim reference was approximately 0.3 z-scores higher. However, length gain between 1 and 12 months of age seemed to be higher than those of the WHO interim reference. The biggest difference in weight was also found at 1 month of age, at which the WHO interim reference was 0.5 to 0.6 higher in z-scores. This indicates that the boys and girls of the WHO interim reference must have gained weight more rapidly during the first month than the infants in this study. The birth weight distribution of the WHO interim reference was narrower, and mean birth weights were slightly higher (3,6,24), which could explain part of the higher weight gain. Weight gain of boys and girls during the first month in this study corresponded well to gain in well-controlled studies (12). Between 1 and 12 months of age, weight gain was higher than of the WHO interim reference. Comparison of the two references between 1 and 12 months of age therefore indicated that both length and weight gain was higher. The +2 z-scores for length and weight of the Euro-Growth reference at 12 months of age correspond to +2.6 and +3.3 WHO interim reference z-scores, respectively. Comparison of the WHO interim references with the NCHS references (24) indicated that the infants who formed the WHO interim cohort gained length and weight more rapidly at least during the first 2 to 3 months of age. It seems therefore that growth during infancy is influenced by early nutrition, in particular by duration of breast-feeding and the time of introduction of solids (3,19,24), but other factors could also play an important role. The long-term effects of early nutrition on growth can only be evaluated if large cohorts of infants who are fed according to the WHO recommendations (7) are compared with infants living under similar conditions who receive formula and solids early in life (19).
Fomon and Nelson (12) published reference data on length and weight between birth and 14 months of age by combining the NCHS (1) and Iowa data sets (12). Combining the data sets allowed them to include many more infants during the early months of life. Birth weights and lengths were comparable with those in the current study. Between 1 and 12 months of age, the length of boys was similar, but the girls in the Euro-Growth Study were taller. Between 5 and 12 months of age, the infants in the Euro-Growth Study were also heavier. Differences during later infancy can be explained by the fact that beyond 6 months of age, the Fomon references rely exclusively on NCHS data (1,2).
Comparison with the recent Dutch references (13) was also of interest, because the Dutch study cohort is a pure random sample of the population. The -2 z-score values for length were similar, but the mean indicated that the Dutch population tended to be taller during early childhood.
For many years, MUAC has been used as an alternative indicator of nutritional status if the collection of height and weight measurements was difficult, such as during emergencies, famines, or refugee crisis. The operational advantages of MUAC include that a single cutoff value (generally 12.5 or 13 cm) is used for children less than 5 years of age. The concept of a sex-and age-independent cutoff value has been challenged, and reference values that were based on the NCHS sample (1) have recently been published (25). The data from the current study confirm that MUAC between 1 and 36 months is sex and age dependent. Means and SD were close to those of the NCHS-based cohort (25). It is of interest that in infants less than 4 months of age the 5th percentile was below the cutoff value of 12.5 cm. Above 9 months of age, the 5th percentile was above 13 cm. It should also be mentioned that the -2 SD values of the current study and the NCHS-based cohort (25) were similar between 18 and 36 months of age, whereas the -2 z-scores of weight and length (height) for age and the weight–height index (26) showed marked differences.
Regional differences in growth within Europe are well described in the literature (11,27,28), but the Euro-Growth reference is the first attempt to compare growth applying standardized methods at corresponding ages. This study confirms marked differences between the regions, which do not coincide precisely with the north–south division of Europe.
The parent–child relation in body size turned out to be most important in the long term. Tanner and Israelsohn (29) have calculated parent–child correlations for length and height and found low correlations at 1 month of age that increased considerably up to 2 years and continued to increase more gradually up to 7 years. The longitudinal Euro-Growth cohort confirmed the increase in parent–child length correlation between 1 month and 2 years of age but not thereafter. Himes et al. (30) reported a difference of 4 cm in infant length at 9 to 12 months of age, for a difference of 22 cm in mid-parental height (1.8 mm/cm), which compares well with the data from the current study.
The long-term predictability of length and weight on the basis of early infancy values seems to be higher for length than for weight. This may be because of the complex nature of body weight, which is a combination of bones and soft tissues, while length mainly represents bone dimensions. It is logical that 3-month values are more predictive than 1-month values of the measurements at 1, 2, and 3 years of age, but the difference in predictability is more pronounced, as expected, only from the 3-month difference in age (9). This may be explained by assuming that it takes approximately 3 months for an infant to find its (genetic) growth track.
EURO-GROWTH STUDY GROUP
Austria (A): C. Male, A. Golser, C. Huemer, B. Pietschnig
Croatia (HR): I. Svel, G. Armano
France (F): J. Schmitz, J. L. Muns, J. Beley, B. Digeon, J. Panis, G. Degy
Germany (D): F. Manz, E. Jekov, M. Radke
Greece (G): T. Zachou, S. Egglezou, J. Sofatzis
Hungary (H): E. Barko, S. Darvay
Italy (I): M. Salerno
Ireland (IRL): V. Freeman, H. Hoey, M. Gibney
Portugal (P): N. Teixeira Santos, A. Guerra, C. Rego, D. Silva
Spain (E): M. Hernandes, J. Molina, C. Ruiz, R. Tojo, E. Sanches, I. Rica, J. Argmeni, J. Rivera, C. Garcia-Caballero, M. Monleon, M. Manrique
Sweden (S): L. Persson, M. Lundstrom
United Kingdom (GB): J. Durnin, J. Reilly, S. Savage
1. Hamill PVV, Drizd TA, Johnson CL, Reed RB, Roche AF. NCHS growth curves for children birth–18 years. United States vital and health statistics.
Washington, DC: Government Printing Office; 1977. Department of Health, Education and Welfare Publication PHS 78-1650, series 11 74.
2. World Health Organization. Measuring change in nutritional status: Reference data for the weight and height of children.
Geneva: World Health Organization, 1983.
3. World Health Organization Working Group on Infant Growth. An evaluation of infant growth: the use and interpretation of anthropometry in infants. Bull World Health Organ
4. De Onis M, Habicht JP. Anthropometric reference data for international use: Recommendations from a World Health Organization Expert Committee. Am J Clin Nutr 1996; 64:650–8.
5. Cole TJ. A critique of the NCHS weight
for height standard. Hum Biol 1985; 57:183–96.
6. World Health Organization Working Group on Infant Growth. An evaluation of infant growth. Geneva: Nutrition Unit, World Health Organization, 1994.
7. World Health Assembly resolution WHA 33.32. Thirty-third World Health Assembly, May 1980.
8. Garza C, De Onis M. A new international growth reference for young children. Am J Clin Nutr 1999; 70(suppl):169S–72S.
9. Van't Hof MA, Haschke F, and the Euro-Growth Study Group. Limitations of growth charts derived from longitudinal studies: The Euro-Growth Study. Int J Sports Med
10. Van't Hof MA, Haschke F, and the Euro-Growth Study Group. The Euro-Growth Study: Why, who, and how. J Pediatr Gastroenterol Nutr
11. Wright EM, Royston P. Simplified estimation of age-specific reference intervals for skewed data. Stat Med 1997; 16:2785–803.
12. Fomon SJ, Nelson SE. Size and growth. In:Nutrition of normal infants.
St. Louis, Mosby, 1993:38–84.
13. Fredriks AM, van Buuren S, Burgmeijer RJF, et al. Continuing positive secular growth change in the Netherlands 1955–1997. Pediatr Res 2000; 47:316–323.
14. Fomon SJ. Nutritional disorders of children: Prevention, screening, and follow-up.
Washington, DC: Dept. Health, Education, and Welfare; 1977. Publication (HSA) 77-5104.
15. Cameron N. The methods of auxological anthropometry. In: Falkner F, Tanner FM, eds. Human growth: A comprehensive treatise.
2nd ed. New York: Plenum Press, 1986: 3–45.
16. Lohman TG, Roche AF, Martorell R. Anthropometric standardization reference manual.
Champaign IL: Human Kinetic Books, 1991.
17. Armitage P, Colton T, eds. Encyclopedia of biostatistics.
Chichester, UK: John Wiley & Sons, 1998:6:4590.
18. Jordan MD. The Centers for Disease Control (CDC) Anthropometric software package, version 3.0
. Atlanta, GA: Centers for Disease Control and Prevention; 1986.
19. Haschke F, van't Hof MA, and the Euro-Growth-Study Group. Euro-Growth references for breast-fed boys and girls: The influence of breast-feeding and solids on growth until 36 months of age. J Pediatr Gastroenterol Nutr
20. Harris EK, Boyd JC. Statistical bases of reference values in laboratory medicine. Statistics: Textbooks and monographs.
Vol. 146. New York: Marcel Dekker, 1995.
21. Cole TJ, Green PJ. Smoothing reference centile curves: The LMS method and penalized likelihood. Stat Med 1992; 11:1305–19.
22. Royston P, Altman DG. Regression using fractional polynomials of continuous covariates: Parsimonious parametric modelling. Appl Stat J R Stat Soc C 1994; 43:429–67.
23. Waterlow JC, Buzina R, Keller W, Lane JM, Nichaman MZ, Tanner JM. The presentation and use of height and weight
data for comparing nutritional status of groups of children under the age of 10 years. Bull World Health Organ 1977, 55:489–98.
24. Dewey KG, Peerson JM, Brown KH, et al. Growth of breast-fed infants deviates from current reference data: A pooled analysis of US, Canadian and European data sets. Pediatrics 1995, 96:495–503.
25. de Onis M, Yip R, Mei Z. The development of MUAC-for age reference data recommended by a World Health Organization Expert Committee. Bull World Health Organ 1997; 75:11–18.
26. Van't Hof MA, Haschke F, and the Euro-Growth Study Group. Euro-Growth references for body mass index and weight
for length. J Pediatr Gastroenterol Nutr
27. Hernandez M, Castellet J, Narvacia JL, et al. Curvas y tablas crecimiento.
Obregosso Bilbao, Bilbao, Spain: Fundacion F, 1988.
28. Iben OG, Barabas A Panto E. The Hungarian national Growth Study. Budapest, Hungary:Humanbiologia,
29. Tanner JM, Israelsohn WJ. Parent-child correlations for body measurements of children between the ages one month and seven years. Ann Hum Genet 1963, 26:245–59.
30. Himes JH, Roche AF, Thissen D. Parent-specific adjustment for evaluation of recumbent length and stature of children. Pediatrics 1985; 75:304–13.