Application of multilevel linear spline models for analysis of growth trajectories in a cohort with repeat antenatal and postnatal measures of growth: a prospective cohort study

IntroductionAntenatal and postnatal growth are important indicators of fetal and child health and development. Studies frequently have repeat antenatal and postnatal measures of growth available and require approaches that can maximise the use of these measures to examine growth trajectories. We demonstrate the use of multilevel linear spline modelling to model growth trajectories with repeated antenatal and postnatal measures of growth from 20 weeks gestation to five years in a cohort at high risk of macrosomia. MethodsProspective follow-up data from 720-759 mother-child pairs from the ROLO study (initially a randomized controlled trial of a low glycemic index diet in pregnancy to prevent recurrence of macrosomia [birthweight > 4K]) were analysed. Fetal measurements were obtained from ultrasound scans performed on mothers at 20-and 34-weeks gestation, including abdominal circumference (AC) and head circumference (HC). An estimated fetal weight was obtained at 20-and 34-weeks gestation, calculated using the Hadlock 4-parameter formula. At delivery, AC, HC, weight and length were recorded. Follow-up anthropometry assessments (AC, HC, weight and length/height) were also obtained in childhood at six months, two years and five years. Linear spline multilevel models were used to examine trajectories of AC, HC and weight from 20 weeks gestation to five years and length/height from birth to five years. Results754, 756 and 759 participants were included in analyses of AC, HC and weight respectively, while 720 participants were included in analysis of length/height. Over 50% of women had 3rd level education and over 90% were of White ethnicity. Women were a mean (SD) age of 32 (4.2) at recruitment. Following exploration of a series of different models for each growth measure, the best fitting model for AC, HC and weight included a model with knots at each measurement occasion giving rise to five linear spline periods from: 20 weeks to 34 weeks gestation, 34 weeks gestation to birth, birth to six months, six months to two years and two years to five years. The best fitting models for length/height included a model with three linear spline periods from birth to six months, six months to two years and two years to five years. Comparison of observed and predicted values for each model demonstrated good model fit. For all growth measures, fetal growth rates were generally fastest in pregnancy or immediately postpartum (for length/height), with rates of growth slowing after birth and becoming slower still as infancy and childhood progressed. We found little difference in growth trajectories between the intervention and control group. There was some evidence of slightly lower HC, weight and length among females compared with males at birth which appeared to widen by age five years due to slower postnatal growth rates among females. ConclusionWe demonstrate the application of multilevel linear spline models for examining growth trajectories when both antenatal and postnatal measures of growth are available. The approach may be useful for cohort studies or randomised controlled trials with repeat prospective assessments of growth spanning pregnancy and childhood.