# Uterine electromyography as a new predictor of extremely preterm birth: a multifactorial model integrating clinical and bioelectrical parameters

## TL;DR

Uterine electromyography (uEMG) combined with clinical factors improves prediction of extremely preterm birth, offering a non-invasive tool for risk assessment.

## Contribution

This study introduces a novel predictive model for extremely preterm birth that integrates uEMG parameters with traditional clinical risk factors.

## Key findings

- uEMG parameters like contraction frequency and duration are independent predictors of extremely preterm birth.
- A model combining uEMG with ART history, prior preterm deliveries, and cervical length shows better discrimination than traditional models.
- High-risk patients identified by the model had significantly higher preterm birth rates in both training and validation sets.

## Abstract

Extremely preterm birth (EPB), defined as delivery before 28 weeks of gestation, is a major contributor to neonatal morbidity and mortality. Accurate prediction of EPB is crucial for enabling timely interventions to improve neonatal outcomes and optimize resource allocation. Uterine electromyography (uEMG) is a non-invasive method that quantifies uterine electrical activity, offering potential for early EPB risk stratification. This study investigates the predictive value of uEMG parameters combined with traditional clinical risk factors for EPB.

In this retrospective study, 276 singleton pregnant women underwent uEMG monitoring between 20+0 and 27+6 weeks of gestation at Sun Yat-sen Memorial Hospital (Guangzhou, China) from January 2018 to May 2025 were collected. The association of uEMG parameters (contraction frequency, average peak contraction intensity, and average contraction duration) with EPB were analyzed using logistic regression.Two predictive models were developed: a traditional model, including: assisted reproductive technology (ART), prior deliveries between 12 and 28 weeks, and transvaginal cervical length (TVCL); an enhanced model incorporating uEMG parameters (contraction frequency, average contraction duration) and clinical risk factor. The area under the receiver operating characteristic curve (AUC-ROC), precision-recall curve, calibration curve and decision curve analysis were used to assess predictive performance.

In total, 37 of 276 women (13.4%) experienced EPB, corresponding to 1 of 103 women in the no uterine contraction subgroup and 36 of 173 women in the uterine contraction subgroup. For model development, we restricted the analysis to the 173 women with detectable uterine contractions. Compared with the non-EPB group, the EPB group showed significantly higher contraction frequency, average peak contraction intensity, and average contraction duration. In multivariable analysis, higher contraction frequency and longer average contraction duration, ART, prior deliveries between 12 and 28 weeks, and shorter TVCL were independently associated with EPB. The uEMG model showed better discrimination than the traditional model (AUC-ROC 0.859, 95% CI 0.798–0.920 vs. 0.716, 95% CI 0.606–0.827; P < 0.05, DeLong test). In the derived nomogram, high-risk patients (score > 76) had markedly higher EPB rates than low-risk patients (training set: 42.9% vs. 10.0%; validation set: 55.6% vs. 0%; P < 0.001).

uEMG parameters, particularly contraction frequency and average contraction duration, are independent predictors of EPB. A prediction model integrating these parameters with ART history, prior deliveries between 12 and 28 weeks, and TVCL provides good discrimination and clinical utility for EPB risk stratification. As a non-invasive and dynamic monitoring tool, uEMG may complement traditional assessment; however, our findings are derived from a single-center retrospective cohort and should be validated in larger, multicenter prospective studies before routine clinical implementation.

The online version contains supplementary material available at 10.1186/s12884-025-08539-3.