The predictive value of first-trimester ultrasound and biochemical markers in threatened miscarriage

Abstract

Threatened miscarriage (TM) is a common complication in early pregnancy, marked by vaginal bleeding before 20 weeks’ gestation with a closed cervix and detectable fetal cardiac activity. This study investigated the prevalence, risk factors, and predictive markers of TM in a national state hospital using a structured, three-part research design. A retrospective cohort analysis of 711 women presenting with first-trimester bleeding showed that 33.9% of TM cases progressed to live births beyond 22 weeks. Advanced maternal age (≥35 years) significantly increased miscarriage risk. Among neonates born following TM, both birthweight and gestational age were lower compared to those without TM: 3008g vs 3239g for female neonates and 3085g vs 3346g for males (P < 0.001); 38.1 weeks vs 39.0 weeks for females and 38.0 weeks vs 38.6 weeks for males (P < 0.05). These findings reinforce the clinical relevance of TM and the need for structured early pregnancy support. A scoping review of 128 studies assessed ultrasound and biochemical markers for miscarriage risk. Intrauterine haematoma, yolk sac abnormalities, and fetal heart rate deviations were strong ultrasound predictors, particularly when combined with biochemical markers such as beta-human chorionic gonadotropin. Low levels of progesterone and beta-human chorionic gonadotropin were consistently linked with adverse outcomes. However, heterogeneity in study design, thresholds, and follow-up periods limited generalisability. These insights informed a prospective case-control study involving 118 women with TM and 59 controls. Data included ultrasound, biochemical, clinical, and demographic variables. Statistical modelling using logistic regression and random forest analysis identified key predictors including progesterone, fetal heart rate, crown-rump length, mean amniotic sac diameter, trophoblast thickness, mean gestational sac diameter, cervical length, soluble fms-like tyrosine kinase-1 to placental growth factor ratio and maternal age. Model performance was strong (area under the receiver-operating curve of 0.85 and 0.968; accuracies 82.7% and 93.1%), underscoring the potential of integrated predictive tools to enhance early miscarriage risk assessment and care delivery.