Background

The assessment of child growth is important in detecting under- and over-growth, which can provide information for timely intervention. The first 1000 days of life (from conception to 2 years of age) is a period of rapid growth and development, and vulnerable to nutritional and environmental influences [1]. Identifying normal child growth patterns is of fundamental importance in growth assessment.

Both the World Health Organization (WHO) growth standards [2] and the China growth references [3] are being applied in China. The WHO growth standards for children aged 0 to 24 months were constructed based on longitudinal data of children (n = 882) by using selection criteria of having socioeconomic conditions favorable to growth and having access to breastfeeding support (for qualifying as “standard”) from the WHO Multicenter Growth Reference Study (MGRS) conducted in six countries from 1997 to 2003 (without a site in China). The China growth charts were constructed from a large (n = 44,250) cross-sectional study based on stratified random sampling of children in nine cities of China, which was conducted from May to October in 2005 [3]. Comparison of the growth curves over the restricted range of ages from 0 to 2 years indicated the reference for China was significant higher for BMI for boys and girls. However, the comparisons were complicated by differences in inclusion/exclusion criteria (for the WHO sample, strict criteria about known constraints on growth and cooperation with feeding recommendations, which led to over 80% of mother-infant pairs being ineligible; for the China sample, multistage stratified cluster sampling was used based on urban/suburban areas, districts, and community, with several exclusion criteria), as well as by differences in the design of the studies (longitudinal for the WHO study and cross-sectional for the study in China). The difference between China growth references and WHO growth standards could have been an artifact, so confirmation study is warranted.

Historically, in some circumstances, secular trends of height have occurred from one generation to the next generation [4]. China has a diverse population, environment, dietary habits and tradition, and it is going through rapid modernization and urbanization. Early child growth has drawn much attention since these factors may affect growth. China started the 1st National Survey on the Physical Growth and Development of Children (NSPGDC) in the nine cities of China in 1975, and conducted the survey every 10 years from 1975 to 2005 to address possible secular trends, with the most recent data (from 2005) providing the current references for growth in China [3] (but in need of a 10-year update in 2015). Longitudinal data from a sample with stricter inclusion/exclusion criteria would provide a better comparison to the WHO standards. A small cohort [5] recruited in 2007 (n = 1531 retained up to 1 year of age) with strict WHO criteria applied showed significant differences (heavier in weight, longer in length, and bigger in head circumference) compared to WHO standards, as well as compared to the current cross-sectional references, which showed similar differences (except for the 97th percentiles that were lower rather than higher).

Long-term follow-up data has enormous value in evaluating the optimal individual growth trajectory, which may not be captured by cross-sectional data [3, 6]. Between 2012 and 2014, six longitudinal birth cohort studies were launched in China. A number of common exposures shared by all cohorts were collected and common outcomes were observed, which formed the foundation of China Birth Cohort Consortium (CBCC). This collaboration provided, for the first time in China, longitudinal growth data from birth cohorts from various regions of the country, but it still is a convenience sample from an efficient combination of cohorts.

This report examines growth patterns from birth to 24 months in Chinese children by pooling the individual level anthropometric follow-up measures from CBCC. The growth references from the 2015 CBCC will be used for comparison to the 2006 WHO longitudinal growth standards and the 2005 China cross-sectional growth references to provide an update on how healthy infants are growing in modern China.

Methods

Study population and data collection

This study used data from six birth cohorts of CBCC which were located at Shanghai (2 cohorts), Anhui, Guangdong, Hubei, and Jiangsu Provinces and were initiated between 2012 and 2014 (Additional file 1: Table S1_1 and S1_2). Additional file 1: Table S1_2 presents the study objective of each of the 6 cohorts. The original aims of these prospective cohorts were to study the environmental, genetic and behavioral factors during pregnancy and in early childhood, and their effects on pregnancy outcomes, fetal and child growth and development, and risks of diseases. Pregnant women were recruited at hospitals when they came for their routine prenatal care visits.

Weight, length, head circumference, and gestational age at birth were obtained from obstetrical medical records. Child anthropometric measurements including weight, length, and head circumference were conducted by trained study staff or trained pediatric nurses in maternal and child health care centers according to the WHO protocol at 7 targeted ages (42 days, 3, 6, 9, 12, 18 and 24 months; http://www.who.int/childgrowth/training/en/). Recumbent length on infants was measured with infant head position in the Frankfort Vertical Plane, and the soles of the feet flat on the moveable footboard. The cohort staffs were trained by group-watching WHO training video course on weight, length, and head circumference. The pediatric nurse measurements were made as routine care was provided. Infant age was calculated by date at measurement minus date of birth. Feeding type in the first 6 months was classified into three types: exclusive breastfeeding, mixed feeding (i.e., combination of breastfeeding and formula feeding), and exclusive/only formula feeding [7]. Infant passive smoking exposure was defined by the mother or father smoking, or for anyone else living in the home smoking. The diagnosis of gestational diabetes mellitus (GDM) in pregnant women was based on the recommendations of International Association of Diabetes and Pregnancy Study Groups (IADPSG) [8].

For this project, we requested each of the six birth cohort studies to contribute longitudinal child growth data of 1000 singleton children from birth to 2 years of age, or maximum number available at the time of our data request in July, 2016. Two cohorts contributed child follow-up measurements up to 12 months due to later starting date (2014) or child follow-up schedule (Additional file 1: Table S1). The inclusion criteria included singleton live births. The exclusion criteria included: (1) infants born with congenital malformations; (2) pregnancy conceived by assisted reproductive technologies (ART); (3) women with medical complication of sexually transmitted diseases (syphilis, HIV infection, and AIDS); (4) women with pre-existed diabetes. There were 5152 mother-child pairs, which provided a sample almost 6 times greater than the WHO longitudinal cohort from 2003 and over 3 times greater than the previous China longitudinal cohort from 2007. While birth cohort studies used better trained personnel for the growth assessments, more observations can also offset “imprecise observations”.

Among the 5152 mothers, 672 had GDM, 213 had preterm deliveries (gestational age < 37 weeks), and 71 had hypertensive disorders in pregnancy. Among the remaining 4258, 7 had missing data on infant sex. To generate the growth references, we used data from 4251 normal term-born children and excluded children of mothers with GDM, hypertensive disorders in pregnancy (e.g., chronic hypertension, gestational hypertension, preeclampsia and eclampsia),children born preterm to avoid the potential influences of known prenatal risk factors [10,11,12],and children with missing data on sex.

Statistical analysis

We used the LMS method to fit smooth z-score curves for length, weight, head circumference and BMI according to age, and for weight according to length respectively in normal term-born healthy children, stratified by infant sex. [13] The three curves of median (M), coefficient of variation (S) and skewness (L, which is expressed as a Box-Cox power) across age/or length were fitted as cubic splines by using maximum penalized likelihood [13]. The z-score of child growth measures y (length, weight, head circumference and BMI) at time t (or length t, for weight-for-length) was calculated from the smooth curve L(t), M(t), and S(t) by the formula:

$$ z=\frac{{\left[y/M(t)\right]}^{L(t)}-1}{L(t)S(t)},\mathrm{if}\kern0.5em L\left(\mathrm{t}\right)\ne 0;\kern0.5em z=\frac{\log \left[y/M\left(\mathrm{t}\right)\right]}{S\left(\mathrm{t}\right)},\mathrm{if}\kern0.5em L\left(\mathrm{t}\right)=0 $$

By using the maximum penalized likelihood and LMS method, all available data of infants from birth to 27 months, including those followed up to 12 months were able to be used to estimate the smoothing parameters and generate the smoothed curves [9, 13]. The age-based references were truncated at 24 completed months to avoid the right-edge effect [14]. We compared z-scores of 0, ±2, and ± 3 for the growth measures in this study with the WHO standards (http://www.who.int/childgrowth/standards/en/), and the China 2005 references for children aged 0 to 2 years [3], both of which were constructed using similar LMS methods for smoothing procedures [3, 14]. The two-sided t-test was used to test statistical significance of the difference at a p < 0.05. The growth curves were constructed by using LMSchartmaker Pro version 2.54 software (Medical Research Council, UK).

We also calculated the 3rd, 10th, 50th, 90th and 97th percentiles of all growth measures in both boys and girls by age with subgroup sample size > 100 observations to summarize our data (without using smoothing technique), and compared these percentiles with WHO standards to show the differences. The analyses were conducted by using SAS 9.4 software (SAS Institute, Cary, North Carolina).

Results

This report presented the z-score curves of 4251 children who were born at term to mothers without gestational or preexisting diabetes, chronic hypertension, preeclampsia, or eclampsia. A total of 28,298 anthropometric measures were obtained from ages 0 to 27 months (Additional file 1: Tables S2 and S3). All were urban children. 51.1% were boys and 54.0% were delivered via C-section. The mean maternal and paternal height was 161.4 (SD 4.9) cm and 174.4 (SD 5.3) cm, respectively. Mean (pre-pregnancy) BMI was 20.6 (SD 2.8) kg/m2 for mothers and 23.9 (SD 3.3) kg/m2 for fathers. As expected, boy infants had greater birthweight, length and head circumference than girl infants (Table 1). Most (80.3%) of mother had college education or more and 98.3% of mother were Han ethnicity. During the first 6 months, most (77.6%) of infants were mixed fed, and 13.4% had exclusive breast-feeding. In the first 2 years, 27.9% of children were exposed to passive-smoking. There was no sex difference for these factors (Table 1). Over the follow-up assessments (see Fig. 1), the children aged 0 to 2 years in this cohort were taller, heavier, and had greater head circumference than the children in the WHO cohort.

Table 1 Characteristics of 4251 mothers, fathers and children by child sex
Full size table
Fig. 1
figure 1

Comparison of growth-for-age z-score curves with WHO standards in boys and girls

Full size image

Length-for-age

Table 2 presents the growth references of length-for-age at 0, ±1, ±2, and ± 3 SD in our study. In comparison with the corresponding WHO growth standard from 0 to 24 months of age, the median length-for-age was on average 0.9 cm (range 0.2–1.3 cm) higher in Chinese boys, and 1.3 cm (range 0.5–1.9 cm) higher in Chinese girls (Fig. 1). Similarly, for z-score of − 2 (i.e. the cutoffs for defining stunting), child length was on average 1.1 cm taller (range 0.8–1.8 cm) in Chinese boys and 1.6 (range 1.1–2.0) cm taller in Chinese girls than the corresponding sex-specific WHO curves. Likewise, for z-score of − 3 was higher in Chinese boys and girls across age.

Table 2 Length (cm)-for-age z-score curves at 0, ±1, ±2, and ± 3 SD for Chinese boys and girls from birth to 24 months
Full size table

Compared to the China growth reference (2005 data), the median length-for-age in our study (2015 data) was on average 0.3 cm higher in boys, and 0.5 cm higher in girls across age (Fig. 2). This might be evidence of a small secular trend. The comparisons to the 2005 China references were more similar than that for the comparisons to the WHO standards (Figs. 1 and 2).

Fig. 2
figure 2

Comparison of growth-for-age z-score curves from China 2015 data (the present study) with those from China 2005 data in boys and girls

Full size image

Weight-for-age

Table 3 presents the growth reference of weight-for-age at 0, ±1, ±2, and ± 3 SD in our study. For weight-for-age z-score of − 2 (cutoff point for defining underweight), weight was on average 0.60 (range 0.13–0.94) kg heavier in Chinese boys and 0.80 (range 0.19–1.10) kg heavier in Chinese girls than those of WHO standards across age (Fig. 1).

Table 3 Weight (kg)-for-age z-score curves at 0, ±1, ±2, and ± 3 SD for Chinese boys and girls from birth to 24 months
Full size table

Compared to China reference from 2005 data, the weight-for-age median in our study (China 2015 data) was on average 0.25 kg higher (range 0.07–0.33 kg) in boys, and 0.34 kg higher (range 0.09–0.42 kg) in girls across age (Fig. 2).

Head circumference-for-age

Table 4 presents the growth reference of head circumference-for-age at 0, ±1, ±2, and ± 3 SD in our study. At the z-score of − 2, head circumference was 0.36 cm greater (range 0.08 to 0.86 cm) in Chinese boys, and 0.76 cm greater (range 0.54 to 1.04 cm) in Chinese girls, than the corresponding WHO standards (Fig. 1).

Table 4 Head circumference (cm)-for-age z-score curves at 0, ±1, ±2, and ± 3 SD for Chinese boys and girls from birth to 24 months
Full size table

Compared to cross-sectional 2005 norms for China, the median head circumference-for-age in our study was similar in boys, but on average 0.3 cm greater (range 0.1–0.7 cm) in girls across age (Fig. 2).

BMI-for-age

Table 5 presents the growth reference of BMI-for-age at 0, ±1, ±2, and ± 3 SD in our study. As shown in Fig. 1, median BMI-for-age was on average 0.70 kg/m2 (range 0.01 to 0.92 kg/m2) higher in Chinese boys, and 0.7 (range 0.0 to 1.0) kg/m2 higher in Chinese girls than the corresponding WHO standards across the age of 0–24 months. For z-score of 2, BMI on average ~ 0.70 kg/m2 higher in Chinese boys and girls than the WHO standards.

Table 5 BMI-for-age z-score at 0, ±1, ±2, and ± 3 SD for Chinese boys and girls from birth to 24 months
Full size table

Compared to the China corresponding growth references from 2005 data, the median BMI-for-age in our study was on average 0.3 kg/m2 higher in boys and 0.4 kg/m2 higher in girls across age (Fig. 2).

Weight-for-length

Table 6 presents the growth references of weight-for-length at 0, ±1, ±2, and ± 3 SD in our study. Median weight-for-length was on average 0.43 kg greater (range 0.01 to 1.07 kg) than WHO standards in boys, and 0.42 kg greater (range 0.00 to 0.64 kg) in Chinese girls from body length > 50 cm (Fig. 3), but lighter weight at the very short length in Chinese girls (< 52 cm).

Table 6 Weight (kg)- for-length (cm) z-score curves at 0, ±1, ±2, and ± 3 SD for Chinese boys and girls from birth to 24 months
Full size table
Fig. 3
figure 3

Comparison of weight-for-length z-score curves from China 2015 data with the WHO standards and Chinese references from 2005 data in boys and girls

Full size image

For z-score of − 2 (cutoff for wasting definition) in boys, weight was ~ 0.29 kg higher (range 0.003–0.94 kg) than the WHO standard at length > 64 cm; between length 45–63 cm, it was 0.08 kg lower (ranged 0.02 to − 0.17) (Fig. 3). In Chinese girls, the weight-for-length values at z-score of − 2 were on average 0.44 kg heavier (ranging 0.001 to 0.85 kg) than the WHO standards for length > 49 cm. For z-score of 2 (cutoff for overweight definition), compared to the WHO standards, weight was on average 0.39 kg higher (range 0.04 to 0.75 kg) in Chinese boys, and 0.34 kg higher (range 0.06 to 0.64 kg) in Chinese girls for the length > 50 cm. Similarly, for z-score of 3, weight-for-length was on average 0.16 kg higher (range − 0.11to 0.36 kg) in Chinese boys, and was 0.30 kg higher (range 0.00 to 0.64 kg) at most length (49 cm to 95 cm) in Chinese girls than the WHO standards.

Compared to cross-sectional 2005 growth references for China, the median weight-for-length was on average 0.31 kg-cm higher (range 0.03–1.00 kg-cm) in boys and 0.28 (range 0.02–0.56) kg-cm higher in girls across length in this study (Fig. 3).

The difference between our raw data and WHO standards

The numbers of anthropometric measurements used for generating smoothed growth curves was shown in Additional file 1: Tables S2 and S3. This study measured the children at 7 targeted ages (42 days, 3, 6, 9, 12, 18 and 24 months), but in fact provided adequate monthly numbers in the first 12 months (Additional file 1: Tables S2 and S3). In addition to above comparison of the LMS-method-fitted smoothing curves, we also presented the 3rd, 10th, 50th, 90th and 97th percentiles of growth measures by age in both boys (Additional file 1: Table S4) and girls (Additional file 1: Table S5). Compared to the corresponding 2006 WHO percentile standards, the 3rd, 10th, 50th, 90th and 97th percentiles (across the ages evaluated in this study from 0 to 2 years) for length, weight, and BMI (Additional file 1: Table S4 for boys and Additional file 1: Table S5 for girls) were consistently higher in healthy Chinese boys (Additional file 1: Table S6) and girls (Additional file 1: Table S7) in 2015. For example, the median lengths from 0 to 2 years were 50.0–89.5 cm in boys (Additional file 1: Table S4), which were 0.1–3.1 cm taller than the WHO percentile standards (Additional file 1: Table S6). The differences compared to WHO standards also were present for weight by length in both boys and girls (Additional file 1: Tables S8 and S9). This indicates the robust of our results.

Discussion

This report of growth measures is based on a large cohort of children (n = 4251) from six recent birth cohorts from China. Growth references from this study represent normal growth of today’s Chinese children from birth to 24 months by using the multicenter data collected recently (from 2012 to 2015). Compared with the WHO standards (collected more than 10 years ago from mid-1997 to end of 2003) and the current China references (collected 10-years ago in late 2005), the median values of length-, weight-, and BMI-for-age reported here were all higher across the ages from 0 to 2 years, and also for median head circumference-for-age except for boys in our study compared to the 2005 references for China. The weight-for-length in our study was also slightly higher at most times in both boys and girls. The magnitude of differences between the WHO standards and the current large cohort (assessed in 2015) was larger than the magnitude of differences previously reported compared to the outdated 2005 references for China. Our report provides improved references for evaluating growth of children aged 0–24 months in modern China.

The height- and weight-for-age values were higher in our longitudinal cohort assessed in five cities of China (Shanghai, Ma’anshan Anhui, Wuhan, Jiangsu, and Guangzhou) than in the cohort based on a cross-sectional study in nine cities of China (Beijing, Shanghai, Harbin, Xi’an, Nanjing, Wuhan, Guangzhou, Fuzhou, and Kunming) [3]. This could be a secular trend. The CBCC cohorts recruited pregnant women in provincial or large tertiary maternity and child hospitals. Most mothers had high education (college or higher), maternal smoking was rare, and the living standard were relatively high. Thus, the growth data in this study may reflect infant growth patterns under near-optimal circumstances. Since our data were acquired recently (10 years since 2005), the higher length and weight may also reflect an ongoing secular trend [4]. The WHO data suggest that secular trend may depend on where the cohort was acquired: the predicted adult height from the child’s length at 2 years suggested there would be no parent-offspring difference in Norway and the United States (i.e., no increase due to a secular trend), but the predicted adult height was much larger than mid-parental height for the other four countries (Brazil, Ghana, India and Oman). [15] Based on the taller height reported here for ages 0 to 24 months than the 2005 China data, we expect a secular trend (i.e., we predict that average adulthood the height of the children in China will exceed the average height of their parents). While China has undergone dramatic progress in economic and social development, the differences still exist between urban and rural areas, different ethnics, and different social economic. The growth pattern observed in this study may reflect infant growth patterns under more optimal circumstances.

Some studies have found that some child population might have their own growth pattern [16], and our study confirmed that Chinese children may be one of them [3, 17, 18]. The difference in values for height-, weight- and BMI-for-age, weight-for-length, and head circumference in this report in comparison to the WHO standards suggests an interesting country difference, and adds to previous comparison that have been summarized in a recent review [19]. Based on studies from both longitudinal and cross-sectional designs, this review concluded that the WHO standards for height and weight “… endorsed slenderness in the midst of an obesity epidemic” and for head circumference were underestimates (and “… would put many children at risk for misdiagnosis of macrocephaly and microcephaly”). Healthy children in some countries are classified (perhaps inappropriately) as “stunted” [16]. In opposite of findings from some countries (overestimating stunting) [16], overall, our study confirmed that the values of growth measures were higher for the key z-score cutoffs in Chinese children in comparison with WHO growth standards [3, 5].

Our references provide the potential cutoffs for evaluating child growth in a population (like in modern China), where children are the center of attention in the family and are growing under favorable environments. Length has been widely used in early detection of stunting, while weight is commonly used as a measure responsive to short-term influences [20]. Head circumference is then the next most-used measure in clinical settings. To reflect the growth centile (position) of a Chinese child in local population, conditioned on age and sex, the Chinese growth standards need to be considered. It may help identify the infants who suffer from poor and modifiable conditions, and thus target those who may benefit most from intervention. In this study, while another term was considered (“growth pattern”), the term “growth reference” was used to maintain consistency with the term used in other publications about Chinese cohorts and to contrast to the term “growth standard” used for the WHO cohort.

One characteristic of this study (the large-scale multicenter prospective birth cohort design) allows us to obtain data on pre- and perinatal risk factors including GDM, chronic hypertension, pre-eclampsia and preterm status. Based on this strength, we could exclude affected mother-infant pairs cases at risk for abnormal patterns of child growth. In this study, the difference of mean paternal age among the three groups of children (mothers with GDM, born preterm, and healthy children) is interesting. Older fathers have more de novo mutations in DNA, and this probably contributes to growth in some cases [21]. Another strength of this study is the longitudinal rather than cross-sectional design. Additional longitudinal analysis [22, 23] of these longitudinal data could better capture and describe the tempo of growth, but due to space limitations will be presented elsewhere. Also, in this sample the educational level of mothers was high, and few of the mothers smoked, so the children lived in advantaged condition, and approach the criteria used for establishing the WHO standards (reflecting how children should grow). Therefore, the data here may reflect growth in near-optimal conditions in China, and provide a growth pattern for contemporary Chinese children.

On the other hand, one limitation of this study is that in some cases head circumference at birth was not measured, and some of children were just followed up to 12 months, which reduced the sample size for this measurement. However, our sample size is still larger than the sample sizes in similar longitudinal birth cohort studies conducted in other countries. We have also performed sensitivity analysis to summary the 3rd, 10th, 50th, 90th and 97th percentiles of all growth measures in infant who had all observations up to 24 months (i.e., without missing observations) and the results were similar to those from all observations (data not shown). Thus, the missing data should be “at random” [9] Also, the birth measures obtained from medical records may not be ideal despite of the high number of the participating hospitals (which were all provincial or large tertiary maternity and child hospitals). Thirdly, this was a convenience sample without specific entry criteria as in the WHO study.

Conclusions

The growth curves in this study represent the growth pattern of today’s normal Chinese children, and may provide references for evaluation of the individual growth status of children growing up in modern China.