Analyses of child cardiometabolic phenotype following assisted reproductive technologies using a pragmatic trial emulation approach

Abstract

Huang and colleagues used machine-learning estimators to analyse a broad range of parameters in a prospective cohort consisting ART and spontaneously conceived children. Small differences in stature and growth could not be explained by parental or perinatal environment factors, nor differences in fetal DNA methylation. No strong differences in metabolic parameters were seen. Assisted reproductive technologies (ART) are increasingly used, however little is known about the long-term health of ART-conceived offspring. Weak selection of comparison groups and poorly characterized mechanisms impede current understanding. In a prospective cohort (Growing Up in Singapore Towards healthy Outcomes; GUSTO; Clinical Trials ID: NCT01174875) including 83 ART-conceived and 1095 spontaneously-conceived singletons, we estimate effects of ART on anthropometry, blood pressure, serum metabolic biomarkers, and cord tissue DNA methylation by emulating a pragmatic trial supported by machine learning-based estimators. We find ART-conceived children to be shorter (-0.5 SD [95% CI: -0.7, -0.2]), lighter (-0.6 SD [-0.9, -0.3]) and have lower skinfold thicknesses (e.g. -14% [-24%, -3%] suprailiac), and blood pressure (-3 mmHg [-6, -0.5] systolic) at 6-6.5 years, with no strong differences in metabolic biomarkers. Differences are not explained by parental anthropometry or comorbidities, polygenic risk score, breastfeeding, or illnesses. Our simulations demonstrate ART is strongly associated with lower NECAB3 DNA methylation, with negative control analyses suggesting these estimates are unbiased. However, methylation changes do not appear to mediate observed differences in child phenotype.