Success rate of repeated cycles of induction in labour in a UK clinical setting: A cohort study

Introduction

Births are increasingly started artificially using medications such as prostaglandin, oxytocin, or other mechanical methods. This process is known as induction of labour (IOL). IOL typically involves a combination of medications known as dinoprostones that are inserted into the vagina, causing artificial rupture of the membranes (‘releasing the waters’), and administration (via a drip) of the synthetic hormone oxytocin.

This retrospective cohort study focused on the success rate of IOL with a repeated cycle of cervical ripening with dinoprostone prostaglandin E2 as vaginal pessary 10mg (Propess) and dinoprostone prostaglandin E2 as vaginal tablet 3mg (Prostin E2).

Furthermore, the study looked for any associations between smoking, maternal characteristics, and cervical Bishop score and success rate of repeated cycles of IOL.

IOL aims to achieve cervical ripening and progressive uterine contractions leading to progressive dilatation of the cervix to facilitate childbirth. In the UK, between 2007 and 2017, IOL interventions increased by nearly 10% to a frequency of 25.5%. This increase has been driven by the increasing prevalence of medical complications during pregnancy leading to elective IOL prior to 42 weeks of gestation.¹

Studies which investigated repeated cycle of induction have been inconsistent in finding a common ground and defining clear guidelines for how to reduce failure of IOL. NICE guidelines² recommend that if IOL fails, a decision about further management should be made in accordance with the mother’s wishes and clinical circumstances: NICE² recommends offering a rest period, a further attempt to induce, or a C-section.

Successful IOL largely depends on the status of the cervix prior to onset of labour (Bishop score). Inducing labour when the Bishop score is low will lead to an increased rate of C-section. The incidence of adverse outcomes following IOL is highest in nulliparous women who have an unfavourable cervix. For a woman with an unfavourable cervix, IOL interventions such as oxytocin administration or artificial ruptured of membrane (ARM) are associated with reduced effectiveness and high failure rates. Prostaglandin medications are frequently used in the process of IOL. Understanding the history and research that supports prostaglandin use for IOL is crucial for safe practice.³ We attempted to look at IOL interventions to find out if there is an association between administration of prostaglandin for IOL and mode of delivery.

Smoking may influence metabolism of prostaglandin and consequently the response to prostaglandin during IOL.^4,5 The increased risk of preterm labor and delivery in smokers is related to increased contractile sensitivity and activity of the uterine myometrium⁴ upon exposure to PG F2α through the respective receptors. Although smoking during pregnancy is a major risk factor for preterm delivery, the underlying mechanism by which smoking stimulates uterine contractions is still poorly understood. In one study, Nakamoto et al.⁴ showed that cigarette smoke extract enhances oxytocin-induced rhythmic contractions of rat and human preterm myometrium. In this study, we explored whether smoking status was associated with outcome of IOL with prostaglandins.

This study aimed first to determine whether mode of delivery (normal vaginal delivery versus C-section and assisted vaginal delivery) varied between labours with and without a repeated cycle of IOL, with 10 mg dinoprostone prostaglandin E2 vaginal pessary and 3 mg dinoprostone prostaglandin E2 vaginal tablet. Secondly, potential associations between smoking and response to prostaglandin were explored. Further, any associations between mother’s age, ethnicity, BMI, Bishop score, gestational age and IOL outcome were explored, as well as between IOL intervention and delivery method. Finally, the data were analysed to identify the most common reason for IOL.

Methods

Results

Of the 252 women in the sample, 86.5% received at least one form of prostaglandin (n = 218; see Table 1). There was close to an even split of ARM (n = 137; 54.4%) versus SROM. The most common type was unassisted vaginal delivery (n = 121; 48.0%), followed by C-section (n = 105; 41.7%), while only 10.3% had an assisted vaginal delivery (n = 10.3%). A little over half the sample had at least one partner who had smoked at some point (n = 129; 51.2%). The sample population was predominantly White (n = 209; 82.9%). Reasons for IOL are presented in Table 2. Means and standard deviations are reported for mother’s age, BMI, Bishop score, and gestational age in Table 3.

A chi-square test was completed to explore associations between repeated IOL with prostaglandin and delivery outcome. There was a significant difference in delivery method between patients with repeated IOL cycles and those with either one cycle of prostaglandins or no prostaglandins (χ²(2) = 5.802, p = 0.055). Of those who had a repeated IOL cycle, 85.7% had a C-section (see Table 4). This was marginally significant, likely because of the small sample size (n = 7).

Chi-square tests showed no difference in delivery mode between one cycle of prostaglandin compared to none (χ²(2) = 1.367, p = 0.505; see Table 5), or between ARM and SROM (χ²(2) = 3.038, p = 0.219; see Table 6). Further, for those who had received at least one type of prostaglandin, there was no difference in delivery method between ARM and SROM (χ²(2) = 2.651, p = 0.266; see Table 7).

Chi-square tests showed a significant difference in delivery mode following administration of prostaglandin, with and without administration of oxytocin (χ²(2) = 11.221, p = 0.004). There was a higher number of unassisted vaginal deliveries without oxytocin administration compared to those who had received oxytocin (see Table 8).

However, for those who had received at least one type of prostaglandin and who had experienced SROM, there was no significant difference in delivery mode between cases with or without oxytocin (χ²(2) = 1.241, p = 0.538; see Table 9). Similarly, for those who had experienced SROM and no type of prostaglandin, there was no significant difference in delivery mode with or without oxytocin (χ²(2) = 0.699, p = 0.705; see Table 10).

There was no difference in prostaglandin administration by smoking status (χ²(1) = 0.071, p = 0.790; see Table 11), nor was there any association between smoking status and mode of delivery in IOL with prostaglandins.

A multinomial logistic regression was planned for the three delivery modes as the dependent variable, and prostaglandins, smoking status, and their interactions as the independent variables. However, there were insufficient assisted vaginal deliveries, both without prostaglandins treatment and with no smoking background (see Table 12). To conduct the moderation analysis, a binomial logistic regression was therefore completed. This model used two delivery methods (C-section and unassisted vaginal delivery) as the dependent variables, with the same set of independent variables. The Hosmer and Lemeshow goodness-of-fit test was used to assess how well the model fit the data. The test suggested that the model might not be the most appropriate fit for the data (p = 0.999). The model predicted 58.5% of deliveries correctly if all cases had the outcome of unassisted vaginal delivery (see Table 13). The model was not able to increase this classification percentage when adding prostaglandins, smoking, and their interaction as predictors. Collectively, the set of predictors only accounted for 1.2% of the variation in delivery method (Nagelkerke R² = 0.012). There was no predictive effect of prostaglandins, smoking, or their interaction for delivery mode (see Table 14). Therefore, there was no evidence for a moderating effect of smoking on the relationship between prostaglandins and delivery mode.

We conducted a multinomial logistic regression with method of delivery as the dependent variable, and maternal age, Bishop score, BMI, and gestational age as continuous predictors, and ethnicity as a categorical predictor. Unassisted vaginal delivery was used as the reference group. A Pearson Chi-square goodness-of-fit test suggested that the model was a good fit for the data (p = 0.465). Additionally, the model predicted delivery method significantly better than a model containing only the intercept (p = 0.003). Collectively, the set of predictors accounted for 13.7% of the variation in delivery method (Nagelkerke R² = 0.137). The model was able to correctly classify 53.3% of cases (see Table 15). There was an overall main effect of mothers’ age at delivery (χ²(2) = 10.672, p = 0.005), Bishop score (χ²(2) = 11.391, p = 0.003), and gestational age (χ²(2) = 6.207, p = 0.045; see Table 16). The effect of each predictor was examined by comparing each delivery method to unassisted vaginal delivery (see Table 17). For every year that a mother’s age increased, the odds of delivering by C-section rather than by unassisted vaginal delivery increased by 1.096 (p = 0.002). However, as the Bishop score increased, the likelihood of delivering by C-section decreased compared to unassisted vaginal delivery (p = 0.021). A similar effect was found where assisted vaginal delivery was less likely compared to unassisted vaginal score as the Bishop score increased (p = 0.004). For every one day increase in gestational age, there was an increase in the odds of having an assisted rather than unassisted vaginal delivery (p = 0.030). There were no effects of ethnicity or BMI in predicting mode of delivery.

A binomial one-sample test showed no difference between sample rate of C-sections (0.417; CI: 0.355-0.480) and the population rate of C-sections for the hospital (0.379; p = 0.121).

However, the binomial one-sample test showed that there was a marginally lower rate of standard vaginal deliveries in the sample (0.480; CI: 0.417-0.544) compared to the population rate of standard unassisted deliveries for the hospital (0.528; p = 0.072). Further, a binomial one-sample test showed no difference between the sample rate of assisted vaginal deliveries (0.103; CI: 0.069-0.148) and the population rate of assisted deliveries for the hospital (0.092; p = 0.307). Finally, a Chi-square one-sample test showed no difference in the sample compared to the hospital population between rates for all three types of delivery (χ²(2) = 2.375, p = 0.305; see Figure 1).

Figure 1. Observed and hypothesised delivery method frequencies.

Discussion

Conclusions

In the sample studied, a repeated IOL cycle did not enhance vaginal delivery and further, unassisted (normal) vaginal deliveries were more frequent where prostaglandins had been administered without oxytocin. Smoking was not shown to influence association failed induction of labour with prostaglandins, although a larger study is needed.

Author contributions

Anas Alojayli (AA) proposed the original idea for the study. AA and Rahim Haboob (RH) designed the retrospective cohort study. AA and RH performed the analyses. AA drafted the article, which was reviewed and revised by both authors.