This was a retrospective study, conducted in a tertiary universitary center in Paris, France.
We included all singleton living preterm breech deliveries between 28 and 36 + 6 weeks of gestation. We excluded situations with medical contraindication to vaginal delivery (abnormal placenta implantation, multiple scared uterus, maternal disease, cephalopelvic disproportion) and fetal malformations. We collected data from 2000 to 2018.
We compared patients in spontaneous labor versus patients with a medical indication for the child to be born with an unfavorable cervix and thereby, needing induction of labor.
The maternity protocol did not include a systematic X-ray pelvimetry for breech even if preterm.
Considering patients with an unfavorable cervix, if a delivery was indicated but not emergent and could be delayed in a few days, induction of labor was considered with a close fetal and maternal monitoring in hospitalization. Cervical ripening was performed according to the local protocol (appendix 1).
During labor whatever the onset of labor, under continuous FHR monitoring, oxytocin infusion could be used if indicated. After two hours of failure to progress with oxytocin infusion a caesarean delivery was considered.
2.1 Data collection
Data were collected retrospectively from patients’ electronic medical records and medical files. All patients were identified from the department's computer database. The entire medical file was computerized and was completed prospectively at each visit, at delivery, and during the postpartum period. All data were verified at the obstetrics staff meeting the day after the delivery. All maternal and pediatric records were also reviewed individually.
2.2 Outcomes
We defined as primary composite outcome a score of neonatal mortality and severe morbidity, including complications of prematurity and complications of breech delivery, adapted to composite criteria used in the Term Breech Trial and PREMODA study. (15) (16)This composite criterion included perinatal death (until 28 days of life), traumatic event during delivery (basal skull fracture, fracture of long bone or clavicle, brachial plexus injury), 5-min Apgar < 4, moderate or severe encephalopathy according Sarnat score, seizures within the first 24 hours, Intra Ventricular Hemorrhage grade 3 or 4, necrotizing enterocolitis grade 2 or 3.
Secondary neonatal less severe outcomes included arterial cord-blood pH at birth <7.10, intubation and ventilation > 24 hours during the first three days, and early-onset bacterial infection (proved: clinical symptoms with bacteria in blood and/or cerebrospinal fluid ; probable : clinical or biological abnormalities).
Maternal morbidity was evaluated by the following events: post-partum hemorrhage (blood loss ≥ 500cc) and its management, maternal need of secondary surgery and endometritis (defined by abdominal pain and fever, abnormal discharge, biological inflammatory syndrome and positive vaginal swab).
The following labor and delivery characteristics were also analyzed: modalities of induction, duration of labor, amniotic fluid characteristics, FHR anomalies during labor, use of oxytocin and mode of delivery.
2.3 Statistical analyses
First, we compared maternal, obstetric and neonatal characteristics between the two groups. Then we conducted univariate analyses to compare the two groups for the principal and secondary outcome measures. We used Student's t, Chi2, and Fisher's exact tests for the univariate analyses. To assess the association between the onset of labor and severe neonatal morbidity, we used a logistic regression model to estimate adjusted OR (OR) and 95% CI.
To further control for indication bias related to confounding factors that might influence both the choice of the onset of labor and severe neonatal morbidity, we used a propensity score approach. The propensity score was defined as the probability of a spontaneous labor based on covariates. This analysis followed a three-step process. First, a propensity score was estimated for all pregnancies by a logistic regression model with the onset of labor as the dependent variable in relation to the baseline maternal, obstetric, neonatal, characteristics (nullipara, previous cesarean delivery, previous hypertensive condition, hospitalization for premature labor, premature rupture of the membranes, prenatal corticosteroids, suspected fetal growth restriction, gestational age at delivery, weight at delivery). Then, the inverse probability of treatment weighting (IPTW) based on estimated propensity scores was used to obtain a synthetic population in which the onset of labor was independent of measured baseline covariates, as confirmed by balance standardized differences. Imbalance was checked by propensity score distribution and calculation of standardized mean differences before and after applying the IPTW. Finally, the association of the onset of labor with severe neonatal morbi-mortality was quantified by OR and 95% CIs estimated from a logistic regression model, without further adjustment.
The statistical analyses were performed using STATA software Version 13 (Stata Corp, College Station, TX, USA).
2.4 Ethics
Under French regulations, this study was exempt from Institutional Review Board approval because it was an observational study using anonymized data from medical records. Women were informed that their records could be used for the evaluation of medical practices and were allowed to opt out of these studies. The medical database was approved by the National Data Protection Authority (Commission Nationale de l’Informatique et des Libertés, CNIL no. 293683).