Prediction of Preterm Birth by Maternal Characteristics and Medical History in the Brazilian Population

Abstract EN

Objectives.

The aim of this study was to assess the performance of a previously published algorithm for first-trimester prediction of spontaneous preterm birth (PTB) in a cohort of Brazilian women.

Methods.

This was a retrospective cohort study of women undergoing routine antenatal care.

Maternal characteristics and medical history were obtained.

The data were inserted in the Fetal Medicine Foundation (FMF) online calculator to estimate the individual risk of PTB.

Univariate and multivariate logistic regression analyses were performed to determine the effects of maternal characteristics on the occurrence of PTB.

A receiver-operating characteristics (ROC) curve was used to determine the detection rates and false-positive rates of the FMF algorithm in predicting PTB <34 weeks of gestation in our population.

Results.

In total, 1,323 women were included.

Of those, 23 (1.7%) had a spontaneous PTB before 34 weeks of gestation, 87 (6.6%) had a preterm birth between 34 and 37 weeks, and 1,197 (91.7%) had a term delivery.

Smoking and a previous history of recurrent PTB between 16 and 30 weeks of gestation without prior term pregnancy were significantly more common among women who delivered before 34 weeks of gestation compared to those who delivered at term were (39.1% vs.

12.0%, p=0.001 and 8.7% vs.

0%, p<0.001, respectively).

Smoking and history of spontaneous PTB remained significantly associated with spontaneous PTB in the multivariate logistic regression analysis.

Significant prediction of PTB <34 weeks of gestation was provided by the FMF algorithm (area under the ROC curve 0.67, 95% CI 0.56–0.78, p=0.005), but the detection rates for fixed false-positive rates of 10% and 20% were poor (26.1% and 34.8%, respectively).

Conclusions.

Maternal characteristics and history in the first trimester can significantly predict the occurrence of spontaneous delivery before 34 weeks of gestation.

Although the predictive algorithm performed similarly to previously published data, the detection rates are poor and research on new biomarkers to improve its performance is needed.