Study design
The prospective, blinded observational study was conducted between April 2020 and April 2021 at Women Hospital, Zhejiang University, school of medicine. Nulliparous women were eligible if they had a viable singleton cephalic presentation pregnancy at full term (≥ 37 completed weeks) with clinical suspicion of CPD (either short maternal stature, unengaged fetal presentation or suspected macrosomia). Exclusion criteria were congenital malformations, maternal contraindications to vaginal delivery and contraindications to the use of MRI such as severe claustrophobia and metal implants.
The enrolled women had spontaneous onset of labor or performed induction of labor. Failure to labor progress was defined according to national guidelines11. Cesarean delivery was recommended if failure to generate cervical change after at least 24 hours of oxytocin administration during the latent phase or the rate of cervical dilation <1cm in 4 hours when the cervix was > 6cm dilated11. The women were also excluded if they required cesarean deliveries with other indications, such as fetal distress, placental abruption and preeclampsia. Managing practitioners were blinded to the results of MRI findings. Vaginal delivery included spontaneous and assisted operative deliveries using forceps.
The study was approved by the local ethics and research committees (IRB-20200044-R on April 04, 2020). In addition, written consent was obtained from all women who agreed to participate.
MRI pelvimetry and fetalbiometry
MRI measurements were performed within fourteen days before labor. All antenatal MRI images were obtained using a 1.5-T unit system (GE Signal HDxt; GE Healthcare, USA) and an eight-element phased-array body coil. The enrolled women were placed in a supine position without sedation. After a localizing gradient echo sequence, the imaging protocol consisted of axial T1-weighted fast spin-echo and sagittal T1 fast spin-echo sequences (repetition time 520 ms, echo time 7.8 ms, 5mm slice thickness, 0.5 mm gap, a field of view 36-40 cm). The scan extended down to the level of the lower margin of the pubic symphysis. Parameters were measured using the institution’s picture archiving and communication system (Zhejiang Greenlander I.T. Co., Ltd., Hangzhou, China).
On transverse sections, bilateral femoral head distance, interspinous distance, intertuberous distance and subpubic angle were visualized and measured (Figure S1). The midsagittal section measurements included the following parameters: obstetric conjugate, pelvic width, sacral outlet diameter, outlet diameter of the pelvis, sacrum length and pelvic inclination (Figure S2). Fetal biometry, including fetal biparietal diameter, head circumstance and abdominal circumstance, were obtained from the 4-mm acquisition. All parameters were measured independently by two radiologists with at least five years of experience in gynecological MRI for inter-observer and intra-observer reliability.
Maternal characteristics including age, gestational age at delivery, height, weight, weight gain during pregnancy, induced or spontaneous labor and the final mode of delivery were recorded. Data on the neonatal outcomes were also collected: sex, birth weight, Apgar scores after 1 and 5 minutes and the admission of the newborn to the neonatal unit. Gestational age was determined by the first day of the last menstrual period and confirmed by the first-trimester ultrasound measurement of crown-rump length12. Body mass index (BMI) was calculated according to the standard formula.
Statistical analysis
The baseline characteristics were described using means (standard deviations [SDs]) for continuous variables and numbers (proportions) for categorical variables. The candidate variables associated with the risk of cesarean delivery were selected a priori based on the clinical feature, common sense and predictors assessed in the previously published literature13-15. Univariable logistics regression analyses were performed to estimate odds ratios (OR) and 95% confidence intervals (CIs). We excluded maternal height and fetal biparietal diameter as the candidate predictors because of the multicollinearity with other variables. A backward stepwise elimination approach was applied to select independent variables for the multivariable logistics regression model with the Akaike Information Criterion16.
Nomograms were constructed by selected variables to predict the probability of cesarean section using statistical software (rms in R; http:// www.r-project.org). For model performance, we assessed the discrimination (the ability to differentiate between the prediction and outcome) and calibration (the discrepancy between predicted and observed outcomes). To quantify the model discrimination, we calculated the concordance (C-) index as the area under the receiver operating characteristic curve (AUC). The calibration was evaluated by calibration plots, accompanied by the Hosmer-Lemeshow goodness-of-fit test. The model internal validation was accessed by bootstrapped resampling to quantify overoptimism17. The optimal cutoff value was calculated by maximizing the Youden index using receiver operating characteristic curve (ROC) analysis (i.e., sensitivity + specificity − 1), and then the sensitivity and specificity were estimated. Finally, decision curve analyses were also applied to evaluate the net benefit of the prediction model (rmda in R; http:// www.r-project.org)18. All tests were 2-sided, P value less than 0.05 was considered statistically significant. The analyses were performed using R statistical software, version 4.0.3.