3. Multivariate analysis of influencing factors of cervical insufficiency
Presence or absence of CI (yes = 1, no = 2) was used as the dependent variable, while significant and other relevant variables in univariate analysis (age, gestation, parity, BMI, hypothyroidism, uterine fibroids, pregnancy-induced hypertension, anemia, vaginitis, and hyperlipidemia) were included in the logistic regression model, where α in = 0.05 and α out = 0.1. After conducting a multivariate adjustment for significant factors and possible confounders on univariate analysis, the results demonstrated that patients with a history of multiple pregnancies had a 17.51 times higher risk of CI than negative patients(OR=17.51, 95%CI: 2.16-141.85, P=0.007); patients who were pregnant by IVF-ET/ovulation induction were found to have a 3.26 times higher risk of CI than negative patients(OR=3.26, 95%CI: 1.69-6.30, P<0.001); patients with gestational diabetes/pregnancy with diabetes were noted to have a 2.88 times higher risk of CI than negative patients(OR=2.88, 95%CI:1.87-4.44, P<0.001); patients with PCOS had an 8.72 times higher risk of CI than negative patients(OR=8.72, 95%CI: 2.33- 32.64,P=0.001); and patients with uterine malformations had a 4.00 times higher risk of CI than negative patients(OR=4.00, 95%CI:1.07-14.97, P=0.040), as shown in Table 3 . Notably, after adjusting for significant factors and possible confounders in the univariate analysis, the multivariate analysis showed that BMI was not associated with the development of CI.
The prediction probability was calculated using the model of logistic regression, in which a new variable and multi-index combination was generated. The ROC curve was then plotted (Figure 2). The larger the area under the ROC curve (AUC), the stronger the prediction ability of the index. The results showed that the area under the ROC curve: multi-index combination (0.728) > GDM or PGDM (0.614) > IVF-ET/ovulation induction (0.586) > PCOS (0.539) > previous multiple pregnancy (0.548) > uterine malformation (0.510). Evidently, the area under the curve following the combination of multiple indicators was 0.728, which was greater than 0.7, and was also significantly greater than the area under the curve of a single indicator, indicating that the model is fair. Accordingly, this once again suggests that previous multiple pregnancy, IVF-ET/ovulation induction conception, diabetes, uterine malformation, and PCOS are related to the occurrence of CI, and the combination of multiple indicators have a certain predictive ability for CI.
Discussion