Customized versus Population-Based Birth Weight References for Predicting Fetal and Neonatal Undernutrition

Abstract

Objectives: The aim of our study was to construct a model of customized birth weight curves based on a Spanish population and to compare the ability of this customized model to our population-based chart to predict a neonatal ponderal index (PI) <10th percentile. Methods: We developed a model that can predict the 10th percentile for a fetus according to gestational age and gender as well as maternal weight, height, and age. We compared the ability of this customized model to that of our own population-based model to predict a neonatal PI <10th percentile. Data from a large database were used (32,854 live newborns, from 1993 through 2012). Only singleton pregnancies with a gestational age at delivery of 32-42 weeks were included. Results: In the entire pregnant population, the customized method was superior to the population-based method for detecting newborns with a PI <10th percentile (sensitivity: 55 vs. 40.96%; specificity: 99.6 vs. 91.23%; positive predictive value: 11.49 vs. 9.55%, and negative predictive value: 98.84 vs. 98.55%, respectively). In pregnant women with a BMI >90th percentile, the sensitivity was 75%, compared to 50% in the population-based method. In pregnant women with a height >90th percentile, the sensitivity was almost as high as in the population-based method (61.53 vs. 33.33%). Conclusion: The customized birth weight curve is superior to the population-based method for the detection of newborns with a PI <10th percentile. This is especially the case in women in the higher scales of height and weight as well as in preterm babies.