Development and validation of a predictive tool for spontaneous preterm birth incorporating cervical length and quantitative fetal fibronectin in asymptomatic high-risk women

OBJECTIVE: Preterm birth rates are rising but clinical prediction models are limited. Accurate prediction enables interventions to be targeted to at risk women. Single threshold (50ng/ml) fetal fibronectin (fFN) and cervical length (CL) have been the best predictors of spontaneous preterm birth (sPTB) but quantification of fFN improves prediction and has not yet been included in a prediction model with other clinical factors in asymptomatic women.

METHODS: Blinded prospective data from 1249 high-risk women attending preterm surveillance clinics were analysed. Parametric survival models, with time-updated covariates for sPTB, were developed and the best selected using the Akaike and Bayesian Information Criteria. The model was developed on the first 624 women and validated on the remaining 625. Fractional polynomials accommodated possible non-linear effects of quantitative fFN (qfFN) and CL. Probability of delivery before 30, 34 or 37 weeks' gestation and within 2 or 4 weeks of test were compared to actual event rates. Predictive statistics were calculated to compare training and validation sets.

RESULTS: The final model used a lognormal survival curve with CL, √fFN and previous sPTB/ PPROM as predictors. Predictive statistics were similar for training and validation sets. Areas under the ROC curves ranged from 0.77-0.99 indicating accurate prediction across all 5 outcomes.

CONCLUSIONS: sPTB in high-risk asymptomatic women can be accurately predicted using a model combining qfFN and CL which supersedes the single threshold fFN test, demographic information and obstetric history. This algorithm has been incorporated into an app for widespread use.