Prenatal Risk Factors and Nomogram Prediction Model for Brain Injury in Preterm Infants of Twin Pregnancies

Objective: Brain injury remains a leading cause of long-term neurodevelopmental impairment in preterm infants, yet risk prediction models specifically designed for twin pregnancies remain scarce. This study aimed to establish and internally validate a prenatal nomogram for individualized risk prediction of brain injury in preterm twins.

Methods: We conducted a retrospective cohort study of 2,610 live preterm twin infants (28–36⁺⁶ weeks) and their mothers from a single tertiary center (2018–2022). Brain injury was diagnosed by cranial imaging (ultrasound/MRI), including intraventricular hemorrhage, periventricular leukomalacia, and other parenchymal brain lesions, and was further stratified into mild-to-moderate and severe subgroups according to lesion location and extent. Candidate prenatal factors were evaluated by univariate analysis and multivariate stepwise logistic regression. Model performance was quantified by AUC, calibration plots, and bootstrap internal validation.

Results: Six independent predictors were retained: earlier gestational age (aOR=0.674), hyperemesis gravidarum (aOR=1.877), gestational anemia (aOR=1.317), intrahepatic cholestasis of pregnancy (aOR=1.471), premature rupture of membranes (aOR=1.378), and polyhydramnios (aOR=2.703). The nomogram achieved an AUC of 0.712 (95%CI 0.688–0.736), with satisfactory calibration and no overfitting. Importantly, all predictors are readily available during routine antenatal care, enabling seamless integration into clinical workflows.

Conclusions: Our findings underscore the multifactorial etiology of brain injury in preterm twins and provide a practical, imaging-based nomogram that may assist in prenatal risk stratification. This tool holds promise for bridging obstetric and neonatal care, allowing early referral and tailored intervention for high-risk twin pregnancies. However, this nomogram was developed from a single-center cohort and only underwent internal validation; external multi-center validation is required before its routine clinical application.