A predictive scoring model for perinatal loss in monochorionic diamniotic twins with selective fetal growth
Shelly Soni1,2, Leny Mathew1, Nahla Khalek1,2
1Richard D. Wood Jr. Center for Fetal Diagnosis and Treatment Children's Hospital of Philadelphia Philadelphia Pennsylvania USA.
Objective:
To establish a clinically applicable prenatal scoring system to predict perinatal loss in monochorionic diamniotic (MCDA) twins with selective fetal growth restriction (sFGR).
Methods:
We performed a retrospective cohort study of MCDA twins with sFGR and managed expectantly at a single fetal therapy center between 2010 and 2024. Pregnancies with incomplete documentation, missing delivery records, spontaneous preterm birth, or well-documented cervical insufficiency were excluded. Perinatal loss was defined as the loss of one or both fetuses or neonates within a single pregnancy. Demographic, sonographic, and outcome variables were compared between pregnancies with and without perinatal loss. Multivariable logistic regression with a priori selected variables was used to identify predictors of perinatal loss. Backward elimination was performed using nested likelihood ratio testing, and variance inflation factors were assessed for multicollinearity. Odds ratios from the final model were used to construct an additive risk score. Receiver operating characteristic (ROC) curve analysis with maximization of the Youden index was used to identify optimal score thresholds.
Results:
Among 222 pregnancies with sFGR, 181 met inclusion criteria, including 132 Type I, 23 Type II, and 26 Type III cases. Perinatal loss occurred in 10%, 26%, and 15% of these subtypes, respectively, with an overall loss rate of 12.7% (23/181). Pregnancies with perinatal loss had greater intertwin estimated fetal weight (EFW) discordance (36% vs. 29%, p = 0.001), a higher prevalence of abnormal middle cerebral artery (MCA) Dopplers (39.1% vs. 10.7%, p < 0.0001) in the growth-restricted fetus (GRF), and earlier gestational age at delivery (27.4 vs. 32.4 weeks, p < 0.0001). Two predictive models with comparable likelihood ratios were identified. Model 1 included intertwin EFW discordance ≥33%, abnormal umbilical artery (UA) and MCA Dopplers, EFW < 3rd percentile, and fetal female sex (protective), achieving an area under the curve (AUC) of 0.80 with 82% sensitivity and 70% specificity at a score threshold of 10. Model 2 excluded fetal sex and yielded an AUC of 0.75 with 61% sensitivity and 83% specificity at a threshold of 9.
Conclusion:
A clinically applicable prenatal scoring system incorporating EFW discordance, UA and MCA Doppler abnormalities in the GRF, EFW < 3rd percentile, and fetal sex can assist in predicting perinatal loss in MCDA twins with sFGR. This model may improve risk stratification and support counseling and shared decision-making. The limited number of Type II and Type III cases restricts conclusions regarding performance across all sFGR subtypes.
