Mathematical modelling of fetal biometry and kidney morphometry for the prediction of birth weight and intrauterine growth restriction (IUGR)

Abstract

The aims of this study were (i) to develop a predictive model for birth weight using various fetal biometric indices measured by ultrasound scans at 28 and 32 weeks gestation, (ii) to determine whether the inclusion of kidney measurements in the model enhanced the accuracy of prediction and (iii) to develop a model to diagnose IUGR before delivery from these measurements. Serial fetal biometric and kidney morphometric measurements were undertaken in 100 singleton pregnancies between 24 and 36–38 weeks’ gestation. Longitudinal analysis was carried out using multilevel models to investigate the profiles of the variables and to obtain an initial assessment of which variables were useful for the prediction of birth weight. Forward selection was used to build a linear model for the logarithm of birth weight based on a subset of 51 babies with a normal growth trajectory. A model for the diagnosis of IUGR was then developed using forward selection logistic regression with a subset of 61 patients, 21 of whom were growth restricted. The most accurate model for predicting birth weight included abdominal circumference (AC) at 32 weeks, anterior–posterior diameter (A–P) of the fetal kidneys at 28 weeks’ gestation, changes in A–P measurements between 28 and 32 weeks’ gestation and changes in A–P². This model predicted birth weight with an average error of 265 g or 9% with 59% of cases predicted to within 10% and 91% to within 20% of the eventual birth weight. The best model for the diagnosis of IUGR included AC at 32 weeks, fetal kidney A–P at 28 weeks and the change in AC. This model, based on indices obtained at 28 and 32 weeks, forwardly correctly predicted 90% of IUGR fetuses delivered after 36 weeks’ gestation.