Intergenerational Effects of Maternal Health on Pregnancy and Neonatal Outcomes in Nova Scotian Children

Abstract

Maternal pre-pregnancy body mass index (BMI) is associated with first-generation health outcomes. The literature suggests an increased risk of low birthweight in infants born to mothers who are underweight, while infants born to mothers with obesity have increased risk of high birthweight, and of becoming obese themselves. Moderate associations between grandparental factors and child birthweight have been reported, but several studies have limitations affecting validity and precision and seldom examined mediation by first-generation factors. Two objectives of this research were to 1) examine the association between grandmaternal (G0) pre-pregnancy BMI and child (G2) birthweight, with investigation of mediation by maternal (G1) pre-pregnancy BMI, and 2) develop a prediction model for G2 fetal growth abnormalities using G0 risk factors, G1 birth characteristics, and G1 pregnancy characteristics in nulliparous G0s and G1s. These objectives were addressed using a subset of the Nova Scotia Atlee Perinatal Database (NSAPD) created by linking women’s birth information with their pregnancy information in adulthood. The clustering structure of the NSAPD, where delivery-level data is nested within women, creates challenges when imputing missing data. The third objective was to assess the use of a recently proposed tree-based method, mixed-effects random forest (MERF), which incorporates clustering in the prediction procedure to impute BMI. This study found imputation using MERF was moderately biased when BMI was missing at random but severely biased when missing not at random, and imputation using standard random forest was least biased and most efficient. In analyses of 20822 G1-G2 dyads, estimates of the total effect of G0 pre-pregnancy BMI on G2 birthweight z-score and the mediator-specific effect via G1 pre-pregnancy BMI, assuming G0s had a BMI of 22 kg/m$^2$ as compared to the `natural course' scenario, were small. G0 factors and G1 birth characteristics, together with G1 characteristics, modestly improved the prediction of fetal growth abnormalities as compared to models based solely on G1 characteristics in a sample of 9068 G2s. Key predictors included G1 gestational weight gain, pre-pregnancy BMI and birthweight z-score. These findings suggest negligible intergenerational effects of G0 pre-pregnancy BMI on G2 birthweight, but moderate predictive ability of G1 size at birth.