Pre-delivery prediction of vaginal delivery in late-onset foetal growth restriction: development and internal validation of a model with decision curve analysis and bedside calculator

Abstract

Purpose

Foetal growth restriction (FGR) complicates 5–10% of pregnancies and is a major contributor to perinatal morbidity and mortality. In late-onset FGR, induction of labour is recommended; however, intrapartum caesarean rates remain high (20–40%). Accurate counselling regarding the likelihood of vaginal delivery is essential, yet no validated prediction model exists for this population. We aim to develop and internally validate a pre-delivery prediction model for vaginal delivery in patients with late-onset FGR undergoing a trial of labour, and to assess its clinical utility.

Methods

This retrospective cohort study included singleton pregnancies with late-onset FGR managed at a tertiary centre (2017–2024). Eligible patients had no prior caesarean delivery, underwent attempted labour, and had complete Doppler assessment within 7 days of delivery. Logistic regression models were constructed using clinical and Doppler variables. Internal validation was performed using bootstrap resampling, and clinical utility was assessed with decision curve analysis.

Results

Among 477 patients, 378 (79.2%) achieved vaginal delivery. Nulliparity (OR 0.22, 95% CI 0.12–0.38; p < 0.001), pre-pregnancy BMI (OR 0.93 per kg/m², 95% CI 0.90–0.96; p < 0.001), abnormal cerebroplacental ratio (OR 0.47, 95% CI 0.28–0.78; p = 0.004), and pre-gestational diabetes (OR 0.28, 95% CI 0.10–0.78; p = 0.015) were independently associated with lower odds of vaginal delivery. The model demonstrated moderate discrimination (AUC 0.74, 95% CI 0.70–0.80) and good calibration, with net benefit above a 55% threshold.

Conclusion

This four-variable model provides individualised prediction of vaginal delivery in late-onset FGR. External validation is required before clinical implementation.