NOMOGRAM TO PREDICT CERVICAL INSUFFICIENCY INCIDENCE IN PATIENTS UNDERGOING IVF/ICSI TREATMENT

Abstract

This study aimed to identify the risk factors for Cervical insufficiency (CI) occurrence after in vitro fertilization (IVF) or intracytoplasmic sperm injection (ICSI) treatment, and set up a predictive model providing personalized and clinically specific information related to CI incident. A retrospective observational cohort study. This retrospective study included 4710 women pregnant after IVF/ICSI treatment from Jan 2011 to Dec 2018 at The Six Affiliated Hospital of Sun Yat-sen University. CI was first diagnosed during pregnancy based on cervical dilation and/or cervical shortening. Univariate and multivariate logistic regression were used to analysis the correlation of pre-pregnancy clinical covariates with CI occurrence. Nomogram model for CI incidence rate prediction was built from a training cohort of 3108 patients and tested on an independent validation cohort of 1602 patients. A bootstrapping technique was used for external validation. A total of 109 patients (2.31%) were diagnosed with CI among all the enrolled patients. In multivariable analysis of the training cohort, CI occurrence was significantly related to serum Testosterone (odds ratio [OR], 7.103; 95% confidence interval [CI], 3.451 - 8.987; p <0.001), BMI (OR, 2.382; 95% CI, 1.186 - 5.786; p = 0.009) , uterine length (OR, 0.261; 95% CI, 0.132 - 0.525; p = 0.005) and gravidity (P = 0.031). Testosterone over 0.7ng/ml, uterine shorter than 45mm, increased BMI and gravidity were highly risky for CI occurrence . The nomogram model was built based on BMI, serum T, gravidity and uterine length, the area under the curve (AUC) was 0.839 (95% CI: 0.763 - 0.900) for the training cohort, the AUC for the validation cohort was 0.708 (95% CI: 0.685 - 0.832). The model showed a satisfactory goodness-of-fit and discrimination ability. No significant difference was detected between the predicted probability obtained from the bootstrap correction and the actual probabilities of CI occurrence (P = 0.261), which means the model is well calibrated. The user-friendly nomogram graphically presented with pre-pregnancy predicted tool for CI incidence in patients undergoing IVF/ICSI treatment. It is of great importance to monitor highly risk women have polycystic ovary syndrome and take preventive measures prior to IVF/ICSI treatment and subsequent pregnancy as well.