Abstract
OBJECTIVE: To develop twin-specific birth weight percentiles with corresponding singleton percentiles from the same population. METHODS: Using 5 years (2019-2023) of twin and singleton birth data from the U.S. National Vital Statistics System, we applied two established methods for generating centile curves: quantile regression and the lambda-mu-sigma (LMS) method. We created birth weight curves by gestational age and corresponding reference values for twins and singletons separately for girls and boys. We compared percentage classified under specific percentiles of birth weight between our curves and an existing reference previously developed for twins (Alexander), which had errors in last menstrual period gestational dating among fetuses born more than 30 years ago, and a separate reference developed for preterm neonates (Fenton). RESULTS: The quantile regression and LMS methods produced similar twin and singleton growth curves, classifying an almost identical proportion of neonates as less than the 3rd, 5th, or 10th percentile and more than the 90th, 95th, or 97th percentile. The Alexander twin reference classified fewer twin neonates than expected with birth weight below the 10th percentile (6.4%) or above the 90th percentile (6.6%). In contrast, the Fenton reference developed among primarily singletons for preterm neonates classified more twin neonates than expected with birth weight below the 10th percentile (18.7%) and fewer than expected for above the 90th percentile (1.4%). CONCLUSION: In a modern cohort of neonates, our twin reference provides improved classification of birth weight percentiles over the existing Alexander reference and over the Fenton reference for preterm neonates. Our singleton reference is a useful addition for studies that examine birth weight centiles among both twins and singletons because it is derived from the same underlying population as the twin reference.