# Development and validation of a predictive model for cervical insufficiency incorporating AMH and androstenedione

## TL;DR

This study creates a model to predict cervical insufficiency in IVF-ET patients using factors like AMH and androstenedione levels.

## Contribution

The novel contribution is a predictive model for cervical insufficiency incorporating AMH and androstenedione in IVF-ET patients.

## Key findings

- A predictive model for cervical insufficiency achieved an AUC of 0.819 in IVF-ET patients.
- Risk factors include BMI > 22.83, testosterone > 0.74 nmol/L, and AMH > 3.50 ng/ml.
- Cervical length > 3.15 cm was identified as a protective factor.

## Abstract

This study aims to develop a predictive model for cervical insufficiency (CI) in women who undergo in vitro fertilization and embryo transfer (IVF-ET) based on relevant indicators measured prior to pregnancy. A total of 2,494 women who received IVF-ET at the Reproductive Medical Center of the Third Hospital of Peking University between 2016 and 2022 were included. All participants ultimately delivered at the same institution. 1,745 patients were assigned to the training cohort and 749 to the validation cohort. Both univariate logistic regression analysis and multiple logistic regression analysis were conducted to establish the CI prediction model. Among the 2,494 cases, the incidence rate of CI was 3.2%. Risk factors identified to be associated with CI included body mass index (BMI) > 22.83 kg/m2, testosterone (T) level > 0.74 nmol/L, androstenedione (A) level > 11.45 nmol/L, anti-Müllerian hormone (AMH) level > 3.50 ng/ml, frequency of hysteroscopic surgery, number of previous pregnancies (gravidity), and pre-pregnancy diabetes. Cervical length > 3.15 cm is a protective factor for cervical insufficiency. Conversely, factors such as the endometrial preparation regimen, occurrence of an intrauterine operation within six months before pregnancy, and the uterine length were not found to be significant risk factors for CI. The area under the curve (AUC) for this model achieved 0.819, with a 95% confidence interval of 0.758 to 0.881. Despite the lack of external validation from independent cohorts, this study successfully developed a comprehensive predictive model for CI in women undergoing IVF-ET, providing a preliminary exploration for early prediction and intervention strategies.

The online version contains supplementary material available at 10.1038/s41598-025-31678-8.