Construction of China national newborn growth standards based on a large low-risk sample

Most published newborn growth references are based on conventional monitoring data that usually included both low- and high-risk pregnancies. We sought to develop a set of neonatal growth standards constructed from only a large sample of low-risk pregnancies. A total of 24,375 naturally conceived singleton live births with gestational ages of 24–42 weeks were collected in 69 hospitals in thirteen Chinese cities between 2015 and 2018. Unhealthy infants or those with high-risk mother were excluded. Smoothed percentile curves of six anthropometric indicators were established using the Generalized Additive Model for Location, Scale and Shape. The 3rd, 10th, 25th, 50th, 75th, 90th, and 97th percentile references for birth weight, length, head circumference, weight/length, body mass index, and ponderal index were calculated for neonates with gestational ages of 24–42 weeks. This set of neonatal growth standards with six anthropometric indicators can provide more tools for growth and nutrition assessment and body proportionality in neonatal clinical practice. These standards might also help to show the differences between growth curves based on low-risk and mixed low- and high-risk pregnancies.


Measurement.
Birth weight was measured within 12 h of birth with an electronic scale (maximum range 20 kg, accurate to 10 g). Birth recumbent length was measured within 24 h with infantometer (maximum range 65 cm, accurate to 0.1 cm) for term babies and preterm babies of large GA and new patent infantometer (Patent No. zl201520996396.X) (maximum range 45 cm, accurate to 0.1 cm) for preterm babies of small GA. Birth head circumference was measured within 24 h with a flexible non-stretchable plastic tape (0.7 cm wide, maximum range 100 cm, accurate to 0.1 cm). Birth weight, length, and head circumference were measured twice and recorded twice in a standardized measurement procedure 18 . Each measurement was collected independently by two trained doctors or nurses. If the difference between the two measurements exceeded the maximum allowable difference (weight 10 g, length 0.5 cm, head circumference 0.5 cm), a third measurement was taken, and then those two measurements not exceeding the allowable difference were recorded. Quality control. Uniform measuring tools were equipped for all sites, including infantometer and new patent infantometer for length measurement, non-stretchable plastic tape for head circumference measurement, standard weights (accuracy 10 g, 50 g, 100 g, 500 g) for calibration of electronic scale, and standard steel tape (accuracy 1 mm) for calibration of infantometer and plastic tape. The electronic scales in each site were used for investigation after evaluation and calibration of standard weights with a maximum allowable difference of 10 g. Calibration was taken every week with maximum allowable differences of weight 10 g, length 0.5 cm, and head circumference 0.5 cm. Questionnaires were completed by pairs of trained doctors or nurses, with one recording the answers and the other reviewing. The completed questionnaires in each city were sent to the Beijing Steering Committee for final check and data entry. EpiData 3.0 software was used for double entry and logic check of the questionnaires.
Statistical analysis. The mean of two measurements for birth weight, length and head circumference was used for data analysis and calculation of anthropometric ratio. Weight/length, BMI, and PI were calculated according to the following formula: [weight (kg)/length (m)], [weight (kg)/length (m) 2 ], and [weight (kg)/length (m) 3 ], respectively. During data cleaning, we excluded 2 missing weight values, 8 missing length values, and 16 missing head circumference values. Few measures not within ± 5 standard deviation (SD) of the mean of overall sex-and GA-specific values was also excluded (12 for weight, 17 for length, 10 for head circumference, 29 for weight/length, 49 for BMI, and 208 for PI). The final sample sizes contributing to the establishment of the growth curves for each indicator are listed below: weight (13,192 males and 11,169 females), length (13,183 and 11,167), head circumference (13,181 and 11,168), weight/length (13,176 and 11,159), BMI (13,162 and 11,153), and PI (13,075 and 11,081). During the establishment of the growth curves, normality test, and skewness and kurtosis analysis were assessed for each indicator. Data analyses used SAS v9.4 (SAS Institute Inc).
The Generalized Additive Model for Location, Scale and Shape (GAMLSS) which is a general framework for fitting regression models where the distribution of the response variable allows for highly skewed and kurtotic continuous distribution 19 20,21 . After comparative testing of alternative methods (ie, distribution transformation and smoothing function) used to generate the growth curves, birth weight, weight/length, BMI, and PI percentile curves were established using the GAMLSS with Box-Cox t (BCT) distribution with cubic splines, and birth length and head circumference using the GAMLSS with Box-Cox power exponential (BCPE) distribution with cubic spline. All these GAMLSS models did not need to be weighted because the difference of each indicator between non-weighting and equal proportional weighting were negligibly small 16 . The differences between fitted percentiles and empirical values at each week were examined for all the six anthropometric indicators.

Comparison of the China standards with the new US curves.
There are small differences for the percentile curves of birth weight, length, and head circumference at 24-36 weeks between China and US curves, but considerable differences at 37-41 weeks, especially at the upper centiles (e.g., 90th); and there is a large disparity for birth BMI from 24 to 41 weeks ( Fig. 3A to D).

Discussion
The reference sample of our standards derive from a population-based survey of newborns from economically developed urban areas with strict criteria for inclusion, such as single live birth, naturally conception, and health condition of both the mother and the newborn. The shapes of growth curves based on low-risk pregnancies with a normal outcome may differ from those of growth curves generated from more conventional data sets that include both low-and high-risk pregnancies. These differences mainly reflect the variations in the distance between the lowermost and uppermost centiles (e.g., 3rd and 97th). For example, the distance in our weight curves from low-risk individuals was shorter compared to other weight curves from Chinese routine monitoring data that included both low-and high-risk individuals 11,12 . Similar to our observations, new and improved Dutch birth weight percentile curves based on data from low-risk pregnancies displayed lower range/variation, which proved to be more effective in identifying clinically important risk SGA infants 2,22 . The health of the mother and the newborn is the foundation of sustainable development for individuals, families, and societies as it is closely linked with health throughout life 23 . Assessment of growth and nutrition of the neonates is essential for a positive outcome in later life. However, no single anthropometric measure fully reflects growth, development and health of newborn babies, so we established percentile curves of multiple anthropometric measures, with each measure revealing distinct relationships with specific health risks or diseases. Birth weight is typically used to define the classification of newborn size as small, appropriate, or large for a specified GA at birth 24 . Birth length is helpful in evaluating whether postnatal catch-up growth is appropriate 25,26 . Birth head circumference reflects intrauterine brain development and predicts the prognosis of nervous system development 27 . A consensus was reached on the definition of growth restriction as birth weight < 3rd percentile or at least 3 out of 5 of the following: birth weight < 10th percentile, length < 10th percentile, head circumference < 10th percentile, prenatal diagnosis of fetal growth restriction, and maternal pregnancy complications 28 . Traditional classification based on birth weight centiles for GA does not reflect body fat in both term and preterm newborns 29 , while weight/length greatly aids in predicting newborn fat mass and fat-free mass as well as body proportionality 15,30 . BMI is useful for measuring body proportionality for newborn infants [30][31][32] . PI is a customary measure to evaluate whether abnormalities in growth in preterm infants are symmetric or asymmetric [33][34][35] . In sum, our established reference values of six indicators can provide more tools for growth and nutrition assessment (e.g., frequently using weight, length and head circumference) and nutrition assessment/body proportionality (e.g., frequently using weight/length, BMI and PI) in neonatal clinical practice. www.nature.com/scientificreports/ Our study has several strengths. First, our standards were based on low-risk pregnancies with a normal outcome. Second, besides the commonly used weight, length, and head circumference, we also established percentile curves of weight/length, BMI, and PI that help assess whether abnormal babies in growth are symmetric or asymmetric. Under strict exclusion criteria, we included a relatively large sample size of preterm infants that guaranteed more reliable percentile curves; however, the screening efficacy of this set of new standards for SGA or LGA and body proportionality still needs to be further validated and evaluated.
Based on a contemporary, large-scaled, population-based cross-sectional nationally representative sample from low-risk pregnancies with a normal outcome that represents optimal intrauterine growth, we developed a set of neonatal growth standards for 24-42 weeks of gestation, including reference values of six anthropometric indicators that can provide more tools for growth and nutrition assessment and body proportionality in neonatal clinical practice. In addition, our study aids in better understanding the differences in the shapes of growth curves between based on data from low-risk pregnancies only or from mixed low-and high-risk pregnancies.