1. Introduction
In recent years, the rate of twin pregnancies has continued to rise due to the growing utilization of assisted reproductive technology (ART) and late childbirth 1, 2. Available data suggest that ART accounts for a third of twin pregnancies 3, 4. It has been well documented that fetal growth of twin fetuses is slower than that of singletons, usually starting from 28 to 32 week of gestation5-8,owing to the limited uterine space9.
Up to date the clinical examination for the intrauterine growth of twins still largely relies on the growth standards of singletons, and it has been an increasing focus to develop twin-specific biometry chart to monitor fetal growth trajectory for twin pregnancy 10. In recent years, several ultrasonographic reference charts of twins have been established 5-8, 11-16, however, they were derived from small populations5, 11, 15 or did not rule out high risk pregnancies5, 12, 16. In addition, evidence suggested that, compared with dichorionic diamniotic (DCDA) twins, monochorionic diamniotic (MCDA) twins showed a slower growth rate6, 11, 16, and ART may affect the perinatal outcome of twin pregnancies17-20. Therefore, both chorionicity and conception mode should be taken into consideration when developing fetal biometric reference for twins. A newly published study from Italy established the first longitudinal growth charts for fetal ultrasound biometry customized for chorionicity, however, the data didn’t show statistical difference of fetal growth over gestation age between DCDA and MCDA twins6. To date, no study explored the differences of fetal intrauterine growth between ART and spontaneously conceived (SC) twin pregnancies. To fill the knowledge gap, our study would examine the growth difference of twins with varied chorionicity and conception mode, aiming to establish chorionicity- and conception mode-specific fetal biometric parameters reference.
Existing studies modeled the growth curve adopting linear mixed model5-7, 12, 13, multilevel linear models14, 15 or hierarchical Bayesian models8, in all of which the fitting precision and accuracy were somewhat weakened by the data’s deviation in skewness and kurtosis coefficient. Compared to linear model, the generalized additive model for location, scale and shape (GAMLSS) extends to model all the fourth-order variations, including median, standard deviation, skewness, and kurtosis, demonstrating a strong strength in improving the accuracy of fitting smoothed percentile curves21. Since 2006, WHO performed GAMLSS to establish child growth standards10.
The fetal growth can be differed by race or ethnicity22, 23. In 2015, a Chinese study initially established a standard for twin fetal weight growth24. However, the study was based on birth weight data but not ultrasonographic biometric parameters, while the birth weight data can be biased by preterm delivery since preterm delivery is usually associated with pregnancy complications and fetal growth abnormalities. Therefore, the standard established in this study somewhat sacrificed the sensitivity to identify the early onset of growth restriction and cannot convey the longitudinal pattern of fetal growth from early pregnancy 24.
The present study would step forward to fit GAMLSS model based longitudinal growth trajectories among Chinese pregnant women by using ultrasonographic biometry data. We are the first to develop fetal growth reference for Chinese twin pregnancies stratified by both chorionicity and mode of conception, and the reference would be tested through the comparison with data from singletons.