Assessment of growth is an indispensable component of health care of infants and children. Deviations of growth are often the earliest recognizable signs of medical, social–emotional, or nutritional problems. A large number of medical conditions can affect growth, and reliable recognition of growth deviation is essential for prompt diagnosis and treatment. Disturbances in parent–child relations are common but are often recognized only when growth failure is apparent. Growth during infancy and early childhood is particularly susceptible to nutritional deprivation, because during the early months of life sizable amounts of energy and nutrients are deposited in newly formed tissues. As growth slows down with increasing age, the need for growth-supporting nutrients diminishes, and susceptibility to nutritional deprivation lessens.
Assessment of growth comprises two components, measurement (anthropometry) and interpretation of measurements. Anthropometry, the physical measurement of the body and its component parts, is in principle simple, portable, inexpensive, and noninvasive and therefore universally applicable. Although measurement errors are common in spite of the simplicity of the methods, or perhaps because of it, the commonly performed anthropometric measurements are indispensable and immensely valuable tools in health assessment. Over the years, the World Health Organization (WHO) has sought to disseminate information on proper techniques (1) and to provide guidance on the appropriate use of anthropometry (2,3). Various indices that are derived by combining measurements are useful in describing medically relevant aspects of body composition. For example, the body mass index (BMI; kg/m2) is used widely for the purpose of describing leanness and adiposity (4).
Interpretation consists of comparison of patients' measurement data with normative (reference) data. The validity of the normative data is therefore crucial. The reference data of the National Center for Health Statistics (NCHS) (5) have been adopted by WHO (6) and have been used worldwide. However, the NCHS-WHO standards are seriously flawed by technical shortcomings. For children in the first 3 years of life, the NCHS-WHO standards are based on data derived from a longitudinal study with measurements taken only every 3 months. This paucity of data points is responsible for the fact that the standards for the first year of life do not accurately reflect the dynamic nature of infant growth. Also of significance is that during early life most of the subjects providing the data were fed formula. Finally, the joining of the two separate data sets which were used (i.e., one for birth to 3 years and one for 2–18 years), created a “disjunction” (nonsmoothness) in the NCHS-WHO standards at 24 months of age (7,8). One consequence of the shortcomings of the NCHS-WHO standards is that weight and length of normal infants regularly exceed the NCHS-WHO standards during the first 2 to 3 months of life, and infants later on give the false impression that their growth is faltering (8–10). During late infancy the discrepancy in weight for age is sufficiently large to lead health care workers to faulty conclusions regarding the adequacy of growth. The WHO Working Group on Infant Growth (9) has concluded that new and better growth standards are needed to enable adequate health management of infants and small children. It was emphasized that new standards would best be derived from populations that follow current health and feeding recommendations.
In 1988, the European Union initiated a concerted action in the field of nutrition and health (Euronut). In part because of the recognized shortcomings of the NCHS-WHO standards (7–10), the Steering Committee of Euronut launched the European Growth Study (Euro-Growth). The objectives of Euro-Growth were to establish European growth references and to study factors influencing growth—in particular, nutrition. Although it was appreciated that a population-based random sample would be optimal for the purpose, it was realized that creating and studying a random European sample would pose insurmountable obstacles under the prevailing circumstances. Therefore, a cluster sample of children available at collaborating sites was studied. A longitudinal study design was chosen because, among other advantages, it is suitable for linking growth data with observational data regarding diet and lifestyle.
The objectives of Euro-Growth, the multicenter study of infants and young children, thus were to record in longitudinal fashion the growth of contemporary European children who were presumably fed in accordance with prevailing feeding recommendations, and to construct growth and growth-velocity references; to assess dietary habits; to evaluate the influences of nutrition, genetic, socioeconomic, and lifestyle factors on growth; and to assess iron and iodine nutritional status.
WHO AND HOW
A Planning Group (F.H., M.A.v.H., J.G.A.D., and J.S.) was established in 1990. Its responsibilities were to select study sites, to develop a study protocol that specified which variables were to be measured and by which methods, and to set up data quality control procedures. An invitation to participate in this longitudinal study was accepted by 22 research centers from 11 European countries (Table 1). Although subjects were to be studied from 1 to 36 months of age, only 16 of the sites were able to monitor children to 36 months of age, whereas the remaining 6 sites were able to study infants to 12 or 24 months of age. A steering committee was established on which each of the 11 participating countries was represented. After some discussion, the final study protocol was established, and the Steering Committee assumed oversight and monitoring functions, for which it held regular meetings (initially twice a year and later yearly).
The protocol defined subject eligibility and acceptable methods of enrollment, established precise measurement ages, and prescribed strictly standardized methods of measurement. Mothers were approached about participation during the postpartum period while still in the maternity hospital or during a visit to pediatric offices (Table 1). Two sites (Dortmund, Germany, and Glasgow, UK) were able to draw population-based random samples. All prospective infants were screened according to the exclusion criteria listed in Table 2. Low-birth-weight infants (<2500 grams) with gestational age more than 36 weeks were followed but not included in this study analysis for the following reasons: It is well known that such infants are more frequently sick during the neonatal period and may need hospital treatment. It is therefore almost impossible to recruit a representative subsample as part of the overall cohort in a longitadinal study. Moreover, low-birth-weight infants with gestational age more than 36 weeks may show catch-up growth after intrauterine malnutrition or irregular growth related to genetic factors. As could be expected from the truncated birth weight distribution, approximately 60 low-birth-weight infants must be present of which we had only 24.
At the time of enrollment, parents were interviewed, and demographic, lifestyle, socioeconomic, and pregnancy-and birth-related data were collected. Reported height of both parents and prepregnancy weight of the mother, actual weight of the father, and reported weight, length, and head circumference of the infant at birth were recorded (Appendix 1).
Participating infants were enrolled before 30 days of age. To avoid possible seasonal effects, an enrollment period of at least 1 year was required from each center. Measurements were made at the following target ages: 1, 2, 3, 4, 5, 6, 9, 12, 18, 24, 30, and 36 months. Precisely, measurements were to be made within 5 days of 30, 61, 91, 122, 153, and 183 days of age; within 7 days of 274 and 365 days of age, and within 14 days of 547, 730, 912, and 1095 days of age. Measurements were performed in pediatric offices, outpatient clinics of pediatric departments, or in the home (Table 1). In addition to anthropometric measurements, during each visit the diet of the child was assessed by a semiquantitative dietary recall method (11), intercurrent illnesses were recorded (Appendix 1), and data on family demographic, socioeconomic, and lifestyle factors were obtained.
Measurements were obtained as described in the literature (1,12) using standardized procedures. Weight was measured to the nearest gram by electronic balances, with subjects in the nude. Recumbent length (1–36 months) was measured by two examinations to the nearest millimeter using a Harpenden (UK) stadiometer. One examiner held the child's head and applied gentle pressure to bring the top of the head into contact with the fixed headboard. The second examiner extended the child's legs and, applying gentle traction, brought the movable footboard to rest firmly against the child's heels with the toes pointing directly upward. During measurement of standing height, which was performed at some sites in addition to recumbent length between 18 and 36 months, a standing-height Harpenden stadiometer was used. The child stood with the bare heels together and back as straight as possible, with the heels, buttocks, shoulders, and head touching the vertical surface of the measuring device. For the measurement of head circumference, a flexible, narrow steel tape was used. The tape was applied firmly around the head above the supraorbital ridges, covering the most prominent part of the frontal bulge anteriorly, and over the part of the occiput that gives maximum circumference. Mid-upper arm circumference (MUAC), upper thigh circumferences, and maximum calf circumference were measured as described by Cameron (1).
Calculation of Sample Size
The overall projected sample size was determined by the need to obtain sufficient data for valid calculation of outlying percentile values (e.g., 3rd and 97th percentiles). However, there was also a perceived need to be able to detect possible differences between study sites. It was thought that the width of the 95% confidence intervals for site-specific mean values should not exceed 0.5 standard deviations (SD). This width of 0.5 SD corresponds to 4 SEM. Therefore 1 SEM is equal to 0.125 SD (=SD/8). Because SEM = SD/[square root]n, a required sample size of 64 per site was calculated. Therefore, taking into account an expected withdrawal rate of 15% per year, each site was required to enroll 100 infants.
In total, 2245 infants (1154 boys and 1091 girls) were enrolled before 30 days of age. Table 3 lists the number of enrolled subjects by study site and also shows the age when last seen for those who withdrew. However, subjects who missed only one (in rare cases, two) visits could continue to participate in the study. Similarly, subjects who missed only the last scheduled measurement (numbers shown in bold) were not considered to have withdrawn. By these criteria, 1746 subjects completed the study until 12 months of age, representing 78% of those enrolled. At 20 sites 1205 subjects (54% of enrollment) were observed to 24 months of age, and at 16 sites 1071 subjects (48% of enrollment) were observed to 36 months of age. In total, during 21,773 visits, complete sets of measurements were obtained, for an average of 9.94 (range, 2–12) sets of measurements per infant. Enrollment occurred from July 1990 through September 1993. Most of the sites spread enrollment over at least 1 year. The last measurement, at the age of 36 months, was performed in October 1996.
Sex, birth order, and social background of subjects are presented in Table 4. The educational level of the parents is listed in Table 5, and infant characteristics at birth and parental age and height are summarized in Table 6.
Implementation of the Study
After the study protocol was adopted by the entire Euro-Growth Study Group, including agreement on technical details of measurement, the training of national coordinators and local anthropometrists was initiated. A videotape, produced by Dr. Renate Bergmann (Berlin, Germany) demonstrated the anthropometric measurement methods to be used (1,12). One of the investigators (J.D., Glasgow, UK) assumed responsibility for quality management of anthropometric measurements. He performed on-site training sessions and, to check reproducibility, organized measurement and remeasurement sessions at the beginning and periodically during the study. Intra-and interexaminer reproducibilities were comparable to those recently published for anthropometric measurements of term infants (13).
Data Quality Control
Major concerns in a longitudinal study include the constancy of measurements, participation selectivity, and subject withdrawals (14,15). Prospectively applied measures to ensure high data quality included a precise and clear study protocol, uniform training of study personnel, standardized data entry, and ongoing quality checks performed centrally on all incoming data. Quality checks performed after completion of the study included cross-sectional and longitudinal consistency checks of the data. checks for selectivity in participation. checks of withdrawal selectivity, and evaluation of reproducibility of measurements over time.
Cross-Sectional and Longitudinal Consistency Checks
The data were entered by study site using EPI-Info (16). After submission of the EPI-info files to the central data bank, a program was run to look for data errors cross-sectionally. Data files containing suspected errors were returned to study sites for checks against original records. These checks were applied to range errors (e.g., gestational age outside 37–44 weeks) and logical errors (e.g., milk consumption was indicated, but the recorded volume of milk intake was 0 ml).
Next anthropometric measurements of each individual subject were examined on a longitudinal basis. Questionable values were returned to the study site for confirmation or correction. Some questionable values were excluded in a manner similar to that used by Karlberg (17) (i.e., on the assumption that transcription errors or misassignments of measurement value had occurred). Fortunately, as indicated in Table 7, only a small percentage of values were eliminated by this procedure.
Selectivity of Participation
It was appreciated from the outset that the chosen method of subject recruitment would not necessarily yield a sample that was representative of the background population and thus could introduce selection bias (14). Parents who commit to participate in a study that requires a number of measurements at precisely defined ages must be highly motivated and probably should reside close to the study site. At most sites, only parents who were thought to be willing and able to participate were invited to enroll their infants. At two centers a random sample from the background population was drawn and invited to participate. Despite this, participation may not have been representative of the general population. The reasons parents declined participation may, for example, have included distance from the study site, whereas a general interest in health problems, higher education, and/or health-related occupation may have increased the probability of participation.
To assess the representativeness of the study sample, a nonparticipation study was conducted wherein the characteristics of those who refused participation were examined in certain sites (design type A, see Appendix 2), whereas a separate sample was drawn from the same population as the participants (design type B). Appendix 2 shows the number of subjects at the 18 sites that participated in the nonparticipation study. Parameters compared between the participating and nonparticipating subjects were maternal age, weight, height, BMI, education, type of household, birth weight, and sex of the infant. Participants and nonparticipants were compared using analysis of variance and logistic regression analysis.
The results showed that overall, the mothers of participants were significantly older and had a higher educational level and that the family was more likely to live in an urban area and to include the infant's father (Table 8). On the other hand, selectivity was minor and nonsignificant with regard to maternal height, weight, and BMI and birth weight and sex of the infant. It was evident that selectivity in participation was not uniform across sites. Significant interactions between participation status and site were seen for the mother's age, weight, BMI, and education; family location (urban vs. rural); presence of the father; and infant birth weight. For example, as Table 8 shows, the mean birth weight of nonparticipating infants was between 280 g higher and 170 g lower than that of participating infants, depending on site. At some sites, predominantly rural families participated and at other sites mainly urban families. Concerning maternal education, Table 8 indicates that in certain sites the less educated mothers participated, while in others the more highly educated mothers participated.
The nonparticipation study thus showed that the sample was selective relative to the background population. This is not particularly surprising. For instance, in Umea (Sweden) the catchment area of the involved clinic was so large that in effect only families living in the city of Umea were likely to participate. The selectivity revealed by the nonparticipation study is superimposed on probable selectivity inherent in the selection of study sites. Other selectivity was found for educational level. In some centers less educated mothers and in other centers more highly educated mothers were overrepresented. This is probably because mothers working outside the home were not likely to participate. Depending on the country, working by the mother can be associated with either low or high educational level. This type of selectivity is a problem in most observational studies. For instance, in the Nijmegen Growth Study (18), higher socioeconomic levels were overrepresented, which is difficult to overcome. Knowing the factors influencing participation gives the opportunity to correct for selectivity in participation (16). This has not been applied in constructing the new references, but has been used in the evaluation of early nutrition on growth (19).
Altogether, the nonparticipation study showed that infants with more highly educated mothers who did not work outside the home and were interested in health problems were overrepresented in the Euro-Growth Study. Therefore, the study may be slightly biased toward an above-average socioeconomic and educational stratum of the European population. The nonparticipation study provides the opportunity to weight under-or overrepresented subgroups during further analysis (19). A proper geographic representation may be obtained by weighting the centers according to region—i.e., southern Europe (Portugal, Spain, Italy, Croatia, and Greece), central Europe (France, Germany, Austria and Hungary), or northern Europe (Ireland, Scotland and Sweden).
Selectivity in Withdrawal: Withdrawal Analysis
Although withdrawal of subjects from the study was generally due to external reasons (e.g., family moving), it could also have been prompted by concerns about the infant's health, including growth performance. Because of possible selectivity in withdrawal, an analysis of anthropometric measurements of the withdrawals was performed. If withdrawal occurred as a result of ill health, it was assumed that the last recorded measurement would have given some indication of the poor performance of the infant. Therefore, the withdrawal analysis focused on the last measurement completed and compared it with corresponding measurements of all other participants. This meant that measurements at different ages and from different sexes had to be combined into one analysis. It is common to use z-scores (20) in such a multiage–sex comparison. Mean z-scores for the last obtained measurement deviating from zero, or SD of the z-scores deviating from one would indicate selectivity in withdrawals.
Five hundred thirty-three infants (24%) missed the last two or more scheduled measurements and were considered withdrawals. The z-scores at their last measurement were calculated for weight, length, and the four circumferences, using means and SDs of all obtained measurements. Means and SDs of these z-scores are presented in Table 9. None of the considered anthropometric values showed z-scores significantly deviating from a mean of 0 or an SD of 1. This implies that withdrawal occurred as a random event and that the construction of growth velocity standards was not affected by withdrawal problems.
The impact of the six sites' providing measurements only to 12 or 24 months was assessed by determining the effect of exclusion of the data from those six sites on the z-scores. This analysis showed that there was a significant effect on length and head circumference. However, the effect was quantitatively small, amounting to between 0.01 and 0.05 SD units. This effect is on the same order of magnitude as the errors introduced by the smoothing process. Therefore, it was reasonable to include data from the six sites in the overall analysis.
Reproducibility Over Time and Reliability
Because anthropometric values change only gradually over time, interperiod correlations (IPCs) could be used for data quality control (18,21). Table 10 gives as an example the matrix of IPC coefficients for length. This symmetric matrix contains correlation coefficients between all measurement occasions. The main diagonal of the matrix contains the usual value of 1, which represents the correlation of a measurement with itself. The correlations close to the main diagonal are generally high, because they are most close in time. The farther from the main diagonal, the longer the intervals between measurements are, and the lower the correlations. This general picture means that there is a relationship between correlation coefficients and time intervals. Graphically, this relationship can be displayed as a descending regression line of correlation coefficients against time interval. The zero time intercept of this regression line may be interpreted as an estimate for the direct measurement–remeasurement correlation. This correlation is equivalent to the cross-sectional reliability coefficient (22) of anthropometric variables. Reliability coefficients express the power of a measurement technique to discriminate among individual subjects within the given context of its application (i.e., age, observer, and site). The cross-sectional reliability compares the error variance (e2) with the population variance (s2): cross-sectional reliability =s2/(s2+e2). In the IPC analysis the cross-sectional reliability may be estimated by the intercept of the descending regression line.
The longitudinal reliability represents the power to discriminate changes that occur with aging (increments) among subjects. The error variance in the increments (2 e2) is compared with the population variance in the true increments (g2) as: longitudinal reliability =g2/(g2 + 2 e2). The increment contains the measurement error twice, causing a large error variance in the increment (i.e., 2 e2), whereas the variability in the increments (g2) is small compared with the cross-sectional situation. Therefore, in general, the longitudinal reliability is much lower than the cross-sectional reliability. The IPC analysis also provides a method to estimate longitudinal reliabilities on the basis of repeated measurements.
For the longitudinal reliability the slope of the regression of correlation coefficients against time interval plays an important role. The slope represents the decay in correlation over time. A horizontal regression line indicates that the predictability is stable over a long period, but also that the increments do not vary within the sample (i.e., small g2). A steep descending regression line means that the increments show a large variation (large g2). It has been shown that: decay = -slope = ½g2/ s2(18). From intercept =s2/(s2 +e2) and decay = ½g2/ s2 it follows that: longitudinal reliability =g2/(g2 + 2 e2) = decay/[(decay + 1)/(intercept - 1)].
The residual SD of the correlations around the regression line is a second indicator of the quality of the collected data. A large residual SD (i.e., a wide spread around the regression line) indicates the existence of measurement occasions that correlate poorly, probably because of low reproducibility of the measurements on one or more measurement occasions. The SE of a correlation coefficient is known to be SE(r) = (1 -r2)/[square root](n - 2). In the Euro-Growth Study, per site n = 100 and r approximately 0.9 (weight, length, and head circumference) or r approximately 0.5 (arm, thigh, and calf circumferences) produced a residual SD of 0.04 or 0.09. Sites having a residual SD more than three times the expected value were eliminated, because their measurements had to be considered inconsistent.
Calculations of residual SD intervals and cross-sectional and longitudinal reliabilities were performed per site for weight, length, and the four circumferences. Because the number of observations was decreased with age, the underlying number of observations weighted the correlations in the regression analyses.
In the example of the IPC matrix for length shown in Table 10, the correlations along the main diagonal were high, and the correlations toward the upper right corner were low. Generally, the first two measurement occasions (age 1 and 2 months) produced low correlations. This may have been due to true variability of growth during the early months of life, or because in this young age group accurate measurements of length are particularly difficult to obtain (12). This means that the regression line concept should be applied only to measurements after 2 months of age. The first two occasions were therefore disregarded in further IPC analyses. The correlations and corresponding regression line for length and the other anthropometric variables are presented in Figure 1, which shows that the regression lines fit the observed correlations very well. Table 11 presents the cross-sectional and longitudinal reliabilities for the six anthropometric variables, as estimated from the interperiod correlation matrices per site. Reliabilities vary from site to site, reflecting differences in data quality. In general, the cross-sectional reliability was high for weight, length, and head circumferences, whereas the reliability for the other circumferences was only acceptable. The longitudinal reliabilities were low (see Table 11), indicating that increments on a monthly basis (the time unit in this study) were not detected reliably. The consequence of this is that growth velocity standards had to be specified for periods longer than 1 month (i.e., at least 2-month increments).
The different sites were screened for indicators of extremely low quality. One of the sites was an outlier for the anthropometric variables, weight and length. At this site the cross-sectional reliability for weight was only 0.89 (range for other sites, 0.91–0.98, Table 11), and the cross-sectional reliability for length was 0.78 (range of other sites, 0.81–0.96). In the Euro-Growth Study, for length and weight measurements, a residual SD of 0.04 was expected, as described. The residual SD for length of the site in question was 0.16, clearly beyond three times the expected value of 0.04. A closer inspection of data from this site showed an unusual variation in the increments between the ages of 1 and 2 years. For all these reasons, it was decided to exclude anthropometric data (but not other data) from this site from further analyses.
A biological spin-off from the IPC figures is that long-term predictability of anthropometric parameters may easily be read (Fig. 1). Head circumference has the highest predictability, with a correlation coefficient of 0.75 over the period of 3 to 36 months. Weight and length were equally predictable at a correlation of 0.5. The high predictability of head circumference was due to low variation in increments, because head circumference of all children increased at approximately the same rate (low g2, see IPC analysis). This caused low longitudinal reliability compared with weight and length (cf., Table 10). The circumferences of upper arm, thigh, and calf, mainly comprising soft tissues, were poorly predictable (r = 0.3) in the first 3 years of life.
The study complied with the proposed International Ethical Guidelines for Biomedical Research Involving Human Subjects in the Declaration of Helsinki (23). Ethical approval was obtained by the ethics committees of each center. Informed consent was obtained from the parents of infants enrolled in the study.
In spite of precautionary measures, the subject sample used in the Euro-Growth Study proved to be selective with regard to some educational and family characteristics compared with the background population. Nevertheless, we consider the sample of the Euro-Growth Study to be as representative of the population of Europe as is practically attainable at this time.
In contrast to minor questions about the external validity of the study, the internal validity proved to be good. The IPC analysis turned out to be a powerful tool in data quality control. The reproducibility of the anthropometric measurements was high in all sites except one. The withdrawal rate was low (24% in 3 years), and withdrawals occurred at random. The data are suitable for the construction of new growth references (24,25) and increments (26), for the evaluation of dietary practices (27), and for the assessment of the influence of early nutrition on growth (28). Preliminary results of the Euro-Growth Study have been published elsewhere (29–39).
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. Cameron N. The methods of auxological anthropometry. In: Falkner F, Tanner JM, eds. Human growth: A comprehensive treatise.
2nd ed. Vol 3. Methodology, ecological, genetic and nutritional effects on growth.
New York: Plenum Press, 1986: 3–36.
2. Gorstein J, Sullivan K, Yip R, et al. Issues in the assessment of nutritional status using anthropometry. Bull World Health Organ 1994; 72:273–83.
3. 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.
4. Rolland-Cachera MF, Cole TJ, Sempe M, Tichet T, Rosignol C, Charraud A. Body mass index variations: Centiles from birth to 87 years. Eur J Clin Nutr 1991; 45:13–21.
5. Hamill PVV, Drizd TA, Johnson CL, Reed RB, Roche AF, Moore WM. Physical growth: National Center for Health Statistics percentiles. Am J Clin Nutr 1979; 32:607–29.
6. WHO. Measuring change in nutritional status.
Geneva: World Health Organization, 1983.
7. 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 1998; 152:471–9.
8. Garza C, De Onis M. A new international growth reference for young children. Am J Clin Nutr 1999; 70(suppl):169S–72S.
9. 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
10. Dewey KG, Peerson JM, Brown KH, et al. and the World Health Organization Working Group on Infant Growth. Growth of breast-fed infants deviates from current reference data: A pooled analysis of US, Canadian, and European data sets. Pediatrics
11. Fomon SJ. Nutritional disorders of children: Prevention, screening, and followup.
Washington, DC: US Dept. Health, Education and Welfare; 1977. Publication (HSA) 77-5104.
12. Fomon SJ. Nutrition of normal infants.
St. Louis: Mosby, 1993.
13. Johnson TS, Engstrom JL, Gelhar DK. Intra-and interexaminer reliability of anthropometric measurements of term infants. J Pediatr Gastroenterol Nutr 1997; 24:497–505.
14. Fletcher H, Fletcher W, Wagner EH. Clinical epidemiology: The essentials.
3rd ed. London: Williams and Wilkins, 1996.
15. Van't Hof MA, Truin GJ, Kalsbeek H, Burgersdijk RCW, Visser RSH, Heling G. The problem of participation in the Dutch National Dental Survey. Community Dent Oral Epidemiol 1991; 19:57–60.
16. Dean AG, Dean JA. Epi-Info 6. Version 6.04b: A word processing database and statistics program for public health.
Atlanta, GA: Centers for Disease Control and Prevention; Geneva, Switzerland: World Health Organization, 1997.
17. Karlberg P, Taranger J, Engström T, et al. The somatic development of children in a Swedish urban community. I: Physical growth from birth to 16 years and longitudinal outcome of the study during the same age period. Acta Pedriatr Scand Suppl 1976: 258:1–25.
18. Prahl-Anderson B, Kowalski CJ, Heydendael P. A mixed longitudinal interdisciplinary study on growth and development.
New York: Academic Press, 1979.
19. Van't Hof MA, Hautvast JGAJ, Schroll M, Vlachonikolis IG, and the Euronut SENECA Study on nutrition and the elderly, design, methods and participation. Eur J Clin Nutr
20. Wright EM, Royston P. Simplified estimation of age-specific reference intervals for skewed data. Stat Med 1997; 16:2785–803.
21. Creemers MC, van't Hof MA, Franssen MJ, van de Putte LB, Gribnau FW, van Riel PL. Disease activity in ankylosing spondylitis: selection of a core set of variables and a first set in the development of a disease activity score. Br J Rheumatol 1996; 35:867–73.
22. Armitage P, Colton T, eds. Encyclopedia of biostatistics.
Chichester, UK: John Wiley & Sons, 1988;3584.
23. Council for International Organizations of Medical Sciences/World Health Organization. International ethical guidelines for biomedical research involving human subjects.
Geneva: World Health Organization, 1993.
24. Haschke F, van't Hof MA, and the Euro-Growth Study Group. Euro-Growth references
for length, weight, and body circumferences. J Pediatr Gastroenterol Nutr
25. 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
26. Van't Hof MA, Haschke F, Darvay S and the Euro-Growth Study Group. Euro-Growth references
on increments in length, weight, head-and arm circumference during the first 3 years of life. J Pediatr Gastroenterol Nutr
27. Freeman V, van't Hof MA, Haschke F and the Euro-Growth Study Group. Patterns of milk and food intake during the first 3 years of life. The Euro-Growth Study. J Pediatr Gastroenterol Nutr
28. 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
29. Van't Hof MA and the Euro-Growth Study Group. The Euro-Growth Study. Proceedings of the 5th international workshop on Maternal and Extrauterine Nutritional Factors: Their influence on fetal and infant growth, Salamanca, Spain, October 1995.
Madrid: Ergon SA, 1995:281–9.
30. Van't Hof MA, Haschke F, Euro-Growth Study Group, Limitations of growth charts derived from longitudinal studies: The Euro-Growth Study. Int J Sports Med
31. Haschke F, Male C. Iron nutritional status during early childhood: the importance of weaning foods to combat iron deficiency. In: Hallberg L, Asp Nils-Georg, eds. Iron nutrition in health and disease. London, UK: John Libbey, 1996; 325–29.
32. Haschke F, Golser A, Heil M, Krobath E, van't Hof M, and the Euro-Growth Study Group: Food intake, energy requirements, and growth of older infants and toddlers in Europe. Monatsschr Kinderheilkd
33. Haschke F. Anemia: A persistent problem. In: Tsang R, Zlotkin SH, Nichols BL, Hansen JW, eds. Nutrition during infancy. Cincinnati, OH: Digital Educational Publishing, 1997; 241–54.
34. Manz F, van't Hof M, Haschke F, and the Euro-Growth Study Group. The mother-infant relationship: Who controls breastfeeding frequency? Lancet
35. Persson LA, Lundstrom M, Lönnerdal B, Hernell O. Are weaning foods causing impaired iron and zink status in 1-year-old Swedish infants? A cohort study. Acta Paediatr 1998; 87:618–22.
36. Freeman VE, Mulder J, van't Hof MA, Hoey HMV, Gibney MJ. A longitudinal study of iron status in children at 12, 24 and 36 months. Public Health Nutr 1998; 1:93–100.
37. Reference not provided.
38. Reference not provided.
39. Reference not provided.