Abstract Background: Twin pregnancies, increasingly prevalent due to assisted reproductive technology, carry a 7-10-fold higher preterm birth (PTB) risk than singletons, with >50% delivering before 37 weeks. Existing prediction models predominantly rely on obstetric history and ultrasound parameters, lacking integration of routinely available serological markers. While inflammation and hematological dysregulation are implicated in PTB pathogenesis, the predictive value of complete blood count (CBC) and metabolic biomarkers remains underexplored in twins. Moreover, few models undergo prospective validation, limiting clinical adoption. This study aimed to develop and rigorously validate a clinically implementable PTB prediction tool by synthesizing serological indicators with established risk factors. Study Design: A single-center study comprising a retrospective training cohort (n=1,270 twin pregnancies, 2019–2021) and a prospective validation cohort (n=227). Multivariable logistic regression identified independent predictors from maternal characteristics, prenatal serology (complete blood count, lipids), pregnancy complications, and fetal factors. Model performance was assessed for discrimination (area under the receiver operating characteristic curve, AUC), calibration (Hosmer-Lemeshow test, bootstrap-corrected calibration curves), and clinical utility (decision curve analysis). The relative weights of predictors influencing neonatal outcomes were additionally explored. Results: 13 variables were independently associated with preterm birth (all PConclusion: This validated nomogram integrates routine serological markers with clinical predictors to accurately stratify preterm birth risk in twin pregnancies (AUC >0.78). It demonstrates immediate clinical utility for targeted monitoring and requires external validation in diverse populations.
Zhang et al. (Fri,) studied this question.