Development and validation of a model for individualized prediction of cervical insufficiency risks in patients undergoing IVF/ICSI treatment

Cervical insufficiency (CI) is a clinical diagnosis used to describe second trimester painless pregnancy loss in the absence of a precipitating cause. The incidence rate of CI is around 1–2% in all pregnancies, however, up to 15% of pregnancy losses during 16–28 weeks could be ascribed to CI [1, 2]. As one of the important factors in preterm birth (PTB), an increased incidence of CI has become a serious obstacle to healthy delivery and complication-free postpartum period for both mother and their fetus [3].

The cervix remains closed during pregnancy and undergoes a simultaneously progressive physiological remodeling with serially changing competence to prepare for labor. Cervical ripening is a crucial process in cervical remodeling [4]. In cases of CI, the cervix dilates without painful uterine contraction, leading to inability of the cervix to remain closed until a term pregnancy. Although cervical weakness may be associated with a variety of events including cervical ablation, cervical excision or cervical hypoplasia after diethylstilbestrol, [5] most women diagnosed with CI actually have normal cervical integrity [6]. On the other hand, premature cervical ripening which can also be caused by subclinical infection, local inflammation, hormonal effects or genetic factors, is now the generally accepted cause of CI [4, 5]. The initial signs in patients with CI are usually present before the experience of contractions or any other clinical symptom of miscarriage or preterm labor, so the opportunity of providing interventional treatment is limited. Emergency cervical cerclage is sometimes performed in patients with cervical dilation, but infections and postoperative contractions with subsequent miscarriage may also happened. There is still a lack of a well-defined population for whom the procedure is clearly beneficial [7].

In 2011, routine cervical ripening recording was recommended by the Global Alliance to Prevent Prematurity and Stillbirth, [8] where a cervix shorter than 25 mm was regarded as the best predictor of PTB due to the inverse relationship between mid-trimester cervical length and gestational age at delivery [9, 10]. However, in the preterm prediction study, among all the women with a short cervix, only 27% of them delivered before 37 weeks, and less than 18% women delivered before 35 weeks [10]. Extremely short cervix less than 15 mm only conferred a 50% chance of delivering prior to 33 weeks, [11] so cervical length showed limited predictive power as single indicator. There were also studies which revealed a crucial role of androgens in regulating cervical function and PTB, but failed to provide clinical evidence of consistent relationship between androgen excess and CI [12, 13].

In vitro fertilization (IVF) and intracytoplasmic sperm injection (ICSI) have been extensively applied as important assisted reproductive technologies in the past 30 years. However, IVF/ICSI treatment has been reported to be accompanied with increased pregnancy-related complications besides adverse perinatal outcomes [14, 15]. Minimal amounts of published data are available regarding the rate of CI or its risk factors in women conceived by IVF/ICSI procedure, although CI is a significant cause of miscarriage/ preterm birth. In this aspect, exploration of the potential risk factors contributing to CI prior to conception and further setting up of a predictive calculation model with a combination of all the risk factors will be of great importance. The results will be beneficial for medical staff to take preventive measures to minimize fetal loss after IVF/ICSI treatment besides aiding the decision-making process.

Therefore, the aims of this study were to investigate the risk factors of CI in patients who had their first intrauterine pregnancy after IVF/ICSI treatment, and to set up a nomogram model based on retrospective data analysis to predict the incidence of CI.

To the best of our knowledge, the nomogram we present here is the first to predict individual CI occurrence rate. Data analysis of 4710 infertile patients who have undergone IVF/ICSI procedure resulted in an original nomogram which could predict the likelihood of the occurrence of CI based on data with remarkable clinical significance. The nomogram was developed in a training cohort which included 3108 patients and was tested on an external independent validation cohort which included 1602 patients. Performance was evaluated using both calibration and discrimination. The value of the model lies in the combination of covariate data that was easily accessible which originated from clinical, biological and imaging characteristics, including: pre-pregnancy BMI, serum T, uterine length and gravidity. The results of the derived nomogram provide a graphically simple and straightforward calculator which is of particular interest for clinicians to make an informed decision on the number of embryos to transfer or preventive measures to be taken in early pregnancy.

In our study, we found that basal serum T was a significant risk factor to predict CI incidence. In the research of Dawood et al., a notable increase in serum T was observed at the third trimester of normal pregnancy [17]. Previous studies have found that androgens were vital for cervical remodeling and for the promotion of cervical ripening by altering the collagenase activity and thus decreasing fibrillar collagen organization [13]. Thus, a circulating increase of androgens throughout the course of pregnancy probably resulted in improper remodeling and advanced timing of cervical ripening, which gave rise to CI with adverse consequences of preterm birth and miscarriage. The most common hyperandrogenic conditions are known as ‘non-tumour ovarian hyperandrogenism’ which includes Polycystic Ovarian Syndrome (PCOS) and hyperreactioluteinalis (HL) [18]. HL, which predominantly occurs during the second or third trimester, is highly associated with adverse pregnancy outcomes such as preterm birth [19, 20]. Patients with PCOS with hyperandrogenism were reported to have a higher prevalence of CI and an approximately 6% higher risk of preterm delivery compared to women without PCOS [21, 22]. Though some studies emphasized that increased androgen levels might be crucial for the maintenance of pregnancy, the mechanism by which androgens impacted pregnancy outcomes has not been fully understood. Further investigation of the underlying mechanisms of androgen action on cervical remodeling is still needed.

Another risk factor of the occurrence of CI is uterine length which was less than 45 mm as revealed by MRL. Routine checks for uterine size before Artificial Reproductive Therapy (ART) is beneficial for detecting patients at an increased risk. In this retrospective study, the length of corpus uteri was recorded as the distance from the internal cervical os to the uterine fundus. Previous studies have supported that abnormal development of the corpus uteri impacted the development of cervix uteri [23]. A prospective study by Hawkins et al. revealed that women with uterine lengths (defined as the distance from the external cervical os to the uterine fundus) shorter than 6 cm were more likely to experience spontaneous abortions [24]. Collectively, the use of uterine length as a significant determinant factor in pre-pregnancy examinations should never be ignored. In addition, the increased weight may increase pressure on the cervix. Frezza et al. measured opening abdominal pressures in patients with varying BMI values by connecting a Verress needle to a pressure monitor, where the results proved that every increase of 1 kg/m² in BMI was accompanied with a 0.07 mmHg increase of average abdominal pressure [25]. Thus, the increased abdominal pressure may be transmitted to the cervix and facilitated the occurrence of CI. Consistently, BMI was a strong correlative factor in our model. Proper weight management would be beneficial to women with BMIs higher than 23.9 kg/m² before IVF/ICSI treatment.

Our data revealed the CI incidence rate of 2.31% in the chosen population, which was higher than the reported CI rate in the general obstetric population [1]. Some studies also reinforced that an increased relative risk of CI had been associated with IVF/ICSI procedure [26, 27]. This was probably because these patients usually experienced more intrauterine surgical intervention during IVF/ICSI procedure, which were potential risk factors for cervical injury. Besides, our study showed that increased gravidity was a significant risk of CI occurrence. It is known that labor, especially which experienced precipitous deliveries and difficult deliveries, impacts cervical competence in varying degrees. On the other hand, our study was limited by the relatively small number of patients with twin pregnancies, where the correlation with CI was insignificant. Nonetheless, single embryo transfer is encouraged for fear of the adverse outcomes of multiple pregnancy. Combined with previous studies and our results, women with PCOS who are known to be associated with hyperandrogenism, overweight, insulin resistance and subfertility, may be at a higher risk of CI and should garner the focused attention of clinicians.

Still, some limitations of the present study have to be underlined. First, a lack of consensus criteria and advances in molecular and imaging technology have contributed to the challenges in the consistency of the diagnosis of CI in this study. Second, as a retrospective study, we lose the records of the cervical length before pregnancy and the circulating T concentration along with the other types of androgens in mid-trimester pregnancy which could provide more clinical details in the association between androgens and CI. Third, the retrospective nature of the study cannot exclude all biases.

Despite these limitations, our results provide the clinical evidence of the correlation between androgen excess and CI occurrence. Our nomogram model to predict the probability of CI occurrence could be a useful tool in aiding physicians and patients undergoing the IVF/ICSI procedure to decide on embryo-transfer option and to pay special attentions during prenatal visits. Additionally, we highlighted that the treatment of hyperandrogenism before pregnancy would be beneficial for a healthy newborn.

In conclusion, our analysis presents a user-friendly model to predict the incidence rate of CI prior to IVF/ICSI treatment. Relative risk assessment could be performed during infertility consultation and appropriate measures could be carried out in advance to minimize the probability of fetal loss. Furthermore, our results support the concept that androgen excess negatively impacts cervical competence and draws attention to certain populations of infertile women (especially women with PCOS) with risk factors.