Trial of labour after caesarean
Background: Induction of labour and caesarean delivery are common interventions in obstetric care and over the last decades both have been steadily increasing in frequency worldwide. The two interventions are concatenated, since many of the indications for either intervention often are the same, and approximately 20-40% of all inductions in first time mothers ends with a caesarean delivery. Women with a caesarean delivery are in their next pregnancy and delivery at risk for both maternal and neonatal adverse outcomes. The aim of this thesis was to study the woman’s chances of a vaginal birth after a first caesarean delivery, and her risk of having a repeat caesarean, in the light of the previous reason for the first caesarean. And also study the risk of a negative birth experience depending on delivery mode after a trial of labour after caesarean. Predicting a woman’s probabilities of a vaginal birth could facilitate the antenatal decisions. Having a previous vaginal birth is one of the strongest predictors for a vaginal birth after caesarean. Delivery mode in women with only a caesarean delivery is more unpredictable. Therefore we aimed to develop a prediction model to predict vaginal birth in women with only a previous caesarean delivery. A further aim was to study the differences in time-to-delivery, caesarean delivery rate, and other maternal and neonatal outcomes between different induction methods in nulliparous women with an unripe cervix.
Material and methods: In these population-based studies we used two different data cohorts based on pregnant women’s antenatal, delivery and postnatal electronic medical records. The Stockholm-Gotland Obstetric Cohort includes the whole population of women delivering in the region and includes approximately 25% of all births in Sweden. The study period was over 7 years, between 2008 and 2014 (Study I, II, IV). The Swedish Pregnancy Register is a new register that has been in use since 2013, today covering 98.5% of all deliveries in Sweden. In our cohort the women studied had two following deliveries between 2014 and 2017 (Study III). In all four studies all the pregnancies and deliveries were at or beyond term, with a singleton infant, in cephalic presentation and live born. The induced women in Study I were nulliparous and in studies II, III and IV the women had one previous caesarean delivery. By using different regression analyses (linear, logistic and Poisson) we calculated time-to-delivery, adverse outcomes, risk of repeat caesarean, mean birth experience and risk of negative birth experience in study I, II and III. In study IV we used both regression and machine learning methods (conditional inference tree and random forest, lasso binary regression) to develop prediction models for predicting vaginal birth after caesarean.
Results: When labour was induced in first time mothers, compared to dinoprostone, an association of a 7 hour shorter mean time-to-delivery with balloon catheter was found, and 1.5 hour shorter mean time-to-delivery with misoprostol. The caesarean delivery rates were high, but the different induction methods showed no significant difference with regard to adverse outcomes. Of all women undergoing a trial of labour after caesarean, 69% had a vaginal delivery. Women with a first unplanned caesarean had increased risk of repeat caesarean compared to women with elective first caesarean (risk ratio 1.64, 95% confidence interval 1.43-1.89). With a previous labour dystocia the risk of repeat caesarean in the second labour was almost twofold. In women with a history of labour dystocia the risk for repeat caesarean decreased with increasing cervical dilation at first delivery. Mean birth experience was rated high for all women, but having an unplanned repeat caesarean was associated with an increased risk of negative birth experience. Machine learning and classical regression models had an area under the receiver-operating curve ranging between 0.61 to 0.69, with a high sensitivity and a low specificity in predicting vaginal birth in women with one previous birth, a caesarean delivery.
Conclusions: To be induced with a balloon catheter is associated with a shorter time-to-delivery than prostaglandins. Induced women have high caesarean rates. Almost 70% of all eligible women deliver vaginally after a trial of labour after caesarean, even women with a history of labour dystocia have a good chance. Most women with a first caesarean score their next birth experience as positive, irrespective of the mode of delivery. However, having a repeat unplanned caesarean is associated with the risk of a negative birth experience. To predict vaginal birth after caesarean is difficult. All the models misclassified unplanned repeat CDs, the majority of individuals with an unplanned repeat CD in the second delivery, had a predicted probability of more than 60% chance of giving birth vaginally.
List of scientific papers
I. Time-to-delivery and delivery outcomes comparing three methods of labor induction in 7551 nulliparous women: A population-based cohort study. Charlotte Lindblad Wollmann, Mia Ahlberg, Sissel Saltvedt and Olof Stephansson. Journal of Perinatology. (2017), 1-7.
https://doi.org/10.1038/jp.2017.122
II. Risk of repeat cesarean delivery in women undergoing trial of labor: A population-based cohort study. Charlotte Lindblad Wollmann, Mia Ahlberg, Sissel Saltvedt, Kari Johansson, Charlotte Elvander and Olof Stephansson. AOGS. 2018, 1-6.
https://doi.org/10.1111/aogs.13447
III. Risk of negative birth experience in women with a previous caesarean delivery: A population-based cohort study. Charlotte Lindblad Wollmann, Can Liu, Sissel Saltvedt, Charlotte Elvander, Mia Ahlberg and Olof Stephansson. [Submitted]
IV. Predicting vaginal birth after cesarean using machine learning models. Charlotte Lindblad Wollmann, Kyle Hart, Can Liu, Aaron B. Caughey, Olof Stephansson and Jonatan M. Snowden. [Submitted]
History
Defence date
2019-11-08Department
- Department of Medicine, Solna
Publisher/Institution
Karolinska InstitutetMain supervisor
Stephansson, OlofCo-supervisors
Ahlberg, Mia; Saltvedt, SisselPublication year
2019Thesis type
- Doctoral thesis
ISBN
978-91-7831-511-6Number of supporting papers
4Language
- eng