O2074: Development of Postnatal Growth Reference Charts for Very Preterm Infants

Background: Commonly used growth reference charts for assessing postnatal growth of very preterm infants are based on cross-sectional anthropometric measurement obtained at birth. These cross-sectional growth charts are thought to represent normal intrauterine growth and are to be used to guide nutrition management, with the goal of attempting to have postnatal growth approximate intrauterine growth. However, we recently showed that postnatal growth followed a distinct pattern different from the cross-sectional growth charts. As a result, growth comparison among preterm infants using cross-sectional charts is not adequate. This project aims to develop postnatal growth trajectory curves for clinical use.

Methods: Anthropometric data from the Pediatrix® Clinical Data Warehouse (CDW) between 2010 and 2020 were used to develop additive mixed models of growth trajectories. All infants who met gestational age (GA) criteria and were admitted to one of the Pediatrix® neonatal intensive care units by 7 days of life were included. Those infants without any measurement data in the CDW or without sex assignment were excluded. Infants were grouped by sex and nearest complete week of gestation. The model contained a smooth function of day-of-life. Random intercept and slope at the infant level were allowed. Raw data from 70% of randomly selected infants were first used for model training, with the remaining 30% for model validation to examine overfitting. All measurement data were then used to develop the final models, from which standard deviations were derived to construct percentile lines. An open-source web application based on the R programming language was developed to promote access to the postnatal growth trajectory curves.

Results: A total of 89,218 infants were included in the study (Table 1). Sex-specific weight, length, and head circumference models for eight (23- to 30-week) GA groups were developed. No model overfitting was observed. The web application requires GA and sex information to render the appropriate anthropometric trajectory curves. When plotting individual anthropometric data points on the postnatal growth trajectory curves, the z-score/percentile can be calculated to allow growth assessment. When measurements of an infant over time are plotted as a series, the data points can be further summarized by a representative percentile line following a grid search for the percentile that gives the smallest root-mean-square-error (Figure 1). After identifying the representative percentile for the data series, the corresponding reference value of the same percentile at Age 0 on the World Health Organization Child Growth Standards can be determined to guide the transition of growth reference.

Conclusion: Postnatal growth trajectory curves offer a complementary tool for the assessment of physical growth of very preterm infants, allowing peer-to-peer comparisons and identification of abnormal growth. The postnatal growth trajectory curves may also serve as a novel tool for growth assessment in future nutrition and growth research.