Using graph learning to understand adverse pregnancy outcomes and stress pathways

Area of Science:

• Reproductive Health
• Maternal-Fetal Medicine
• Psychoneuroendocrinology

Background:

• Maternal stress is a known risk factor for adverse pregnancy outcomes.
• Existing research has identified some stress-related pathways, but a comprehensive understanding is lacking.
• The Measurement of Maternal Stress (MOMS) study provides a rich dataset for exploring these complex relationships.

Purpose of the Study:

• To identify and visualize complex pathways between maternal stress indicators and adverse pregnancy outcomes.
• To validate existing knowledge and discover novel associations using advanced statistical methods.
• To provide a framework for understanding the multifactorial nature of adverse pregnancy outcomes.

Main Methods:

• Application of a nonparametric graph-learning algorithm, PC-KCI, to observational prospective cohort data.
• Analysis of data from 744 women in the Measurement of Maternal Stress (MOMS) study.
• Examination of infant outcomes (prematurity, hospital days, small for gestational age) and maternal outcomes (pre-eclampsia, gestational diabetes, gestational hypertension).

Main Results:

• PC-KCI successfully replicated known pathways, such as gestational weeks and preterm premature rupture of membranes.
• Novel pathways identified include hair cortisol levels linked to pre-eclampsia.
• Race-specific pathways to preterm birth were observed, alongside links between childhood trauma, adult stress, and low birth weight.

Conclusions:

• The study confirmed established stress-outcome relationships and revealed previously unrecognized pathways.
• The findings highlight the importance of considering race, cortisol levels, and cumulative stress in pregnancy.
• The PC-KCI method offers a powerful tool for global assessment and causal pathway exploration in clinical problems.