Background
Induction of labor is one of the most common interventions in modern obstetrics, and its frequency is expected to continue to increase. There is inconsistency as to how failed induction of labor is defined; however, the majority of studies define success as the achievement of vaginal delivery. Induction of labor in nulliparous women poses an additional challenge with a 15% to 20% incidence of failure, ending in emergency operative deliveries. The Bishop score has been traditionally used before decisions for induction of labor. Nonetheless, it is subjective and prone to marked interobserver variation. Several studies have been conducted to find alternative predictors, yet a reliable, objective method still remains to be introduced and validated. Hence, there is still a need for the development of new predictive tools to facilitate informed decision making, optimization of resources, and minimization of potential risks of failure. Furthermore, a peripartum transperineal ultrasound scan has been proven to provide objective, noninvasive assessment of labor.
Objective
This study aimed to assess the feasibility of developing and validating an objective and reproducible model for the prediction of cesarean delivery for failure to progress as an outcome of labor induction in term singleton pregnancies.
Study Design
This was a prospective observational cohort study conducted in Cairo University Hospitals and University of Bologna Hospitals between November 2018 and November 2019. We recruited 382 primigravidae with singleton term pregnancies in cephalic presentation. All patients had baseline Bishop scoring together with various transabdominal and transperineal ultrasound assessments of the fetus, maternal cervix, and pelvic floor. The managing obstetricians were blinded to the ultrasound scan findings. The method and indication of induction of labor, the total duration of stages of labor, mode of birth, and neonatal outcomes were all recorded. Women who had operative delivery for fetal distress or indications other than failure to progress in labor were excluded from the final analysis, leaving a total of 344 participants who were randomly divided into 243 and 101 pregnancies that constituted the model development and cross-validation groups, respectively.
Results
It was possible to perform transabdominal and transperineal scans and assess all the required parameters on all study participants. Univariate and multivariate analyses were used for selection of potential predictors and model fitting. The independent predictive variables for cesarean delivery included maternal age (odds ratio, 1.12; P =.003), cervical length (odds ratio, 1.08; P =.04), angle of progression at rest (odds ratio, 0.9; P =.001), and occiput posterior position (odds ratio, 5.7; P =.006). We tested the performance of the prediction model on our cross-validation group. The calculated areas under the curve for the ability of the model to predict cesarean delivery were 0.7969 (95% confidence interval, 0.71–0.87) and 0.88 (95% confidence interval, 0.79–0.97) for the developed and validated models, respectively.
Conclusion
Maternal age and sonographic fetal occiput position, angle of progression at rest, and cervical length before labor induction are very good predictors of induction outcome in nulliparous women at term.
Introduction
Induction of labor (IOL) is one of the most common exercised and studied interventions in obstetrics. Its frequency has been increasing, with reports of 1 in 5 pregnant women undergoing IOL, , and is expected to continue to rise given the increase in the evidence-based, recommended indications for IOL, whether for obstetrical, fetal, maternal, or medical reasons. There is inconsistency in defining failed IOL; some authors define failure of IOL based on the duration of the latent phase, using 15 hours as a cutoff value, and others consider an inability to achieve cervical dilatation >4 cm within 12 hours of oxytocin administration as an indicator of failed IOL. Another study suggested that the simple achievement of active labor should be considered a measure of successful IOL. Nonetheless, the majority of authors find it pertinent to consider the outcome, rather than the process, and propose vaginal delivery as the main IOL outcome. After all, for the expectant woman, when embarking on IOL, the outcome sought is vaginal delivery; otherwise she would opt for cesarean delivery from the start. IOL in nulliparous women at term does not always lead to a normal spontaneous vaginal delivery; some cases, especially primigravidae of advanced age, need assistance with an instrumental delivery or require cesarean delivery. , It is estimated that 15% to 20% of IOLs fail to result in vaginal birth, ending in intrapartum operative deliveries.
Why was this study conducted?
This study aimed to develop a reliable model for prediction of cesarean delivery for failure to progress as an outcome of labor induction in term singleton pregnancies.
Key findings
A predictive model composed of maternal age, cervical length, angle of progression at rest, and fetal occiput posterior position provided accurate prediction of successful induction of labor (area under the receiver operating characteristic curve [AUC], 0.79; 95% confidence interval, 0.71–0.87). There was also a good performance in validation of the model (AUC, 0.88; 95% confidence interval, 0.79–0.97).
What does this add to what is known?
This study provides a model for prediction of the success of induction of labor, focusing on objective, accessible, and acceptable predictors.
Numerous investigators have evaluated several clinical and ultrasonographic parameters as predictors of IOL outcome and reported varying results. The Bishop score has traditionally been used as the standard test before IOL determination. Nonetheless, it is a subjective assessment associated with poor predictive value, reproducibility, and high degrees of inter- and intraobserver disagreement. Moreover, studies that compared the predictive value of ultrasonographic indices to the Bishop score have generated contradictory results. The negative impacts of failed IOL include the stress of enduring a futile, prolonged trial of labor; an increased economic burden and misuse of healthcare resources owing to a prolonged hospital stay; excessive use of medications; vigilant maternal and fetal monitoring; and an increased rate of interventions to the increased prevalence of maternal, fetal, and neonatal complications of an emergency cesarean delivery. Therefore, to enable obstetricians to individualize the care offered to patients, it is important to identify women at high risk of IOL failure, improve clinical outcomes, and optimize the cost-effectiveness of healthcare interventions. In an attempt to identify methods of assessment more objective than digital examination, ultrasound has been shown to be suitable to assess labor progression. Transabdominal and transperineal ultrasound have been shown to provide reproducible, objective, and noninvasive assessment of labor progression. , Nevertheless, a reliable, objective method to predict the likelihood of vaginal delivery still remains to be introduced and validated. This calls for the development of new predictive tools for the success of IOL to allow for informed decision making, optimization of resources, and minimization of potential risks of failure. The objective of this study was to assess the feasibility of developing and validating an objective and reproducible model for the prediction of cesarean delivery for failure to progress as an outcome of labor induction in term singleton pregnancies.
Methods
Design and setting
This was a prospective observational cohort study conducted between November 2018 and November 2019 in 2 tertiary-level university-affiliated maternity units: Kasr Al Ainy University Hospital, Cairo University, Egypt, and Sant’Orsola Malpighi University Hospital, University of Bologna, Bologna, Italy. The local research ethics committees of both participating units approved the study protocol before study commencement (Kasr Al Ainy University Hospital, reference number O18005 and Sant’Orsola Malpighi University Hospital, reference number 139/2016/U/Oss). All study participants provided written informed consent before enrollment.
Participants
Women were considered eligible for inclusion in this study if they met the following requirements: ≥18 years of age, nulliparous, singleton, term pregnancy (37–42 weeks of gestation) planned for IOL for any indication, and a fetus in a cephalic presentation. Women presenting in labor or with a history of uterine surgery or scarring were excluded from the study. Recruitment into the study was nonconsecutive, depending on the availability of a member of the study team trained to undertake the a priori set of ultrasound parameters under consideration.
A total of 382 nulliparous women were enrolled into the study, including 268 of a total of 1440 (18.6%) pregnancies during the study period at Kasr Al Ainy University Hospital and 114 of a total of 983 (11.6%) at Sant’Orsola Malpighi University Hospital. All participants had a baseline clinical cervical assessment using the modified Bishop score ; the attending obstetricians managed the labor in line with the unit’s protocol and were blinded to the ultrasound scan findings ( Supplemental Material ). In addition to demographic details, data were collected as follows: the method and indication of IOL, the total duration of labor (onset of induction to delivery), duration of first and second stages including length of the pushing phase, mode of birth, and neonatal outcomes. Because the aim of our study was to develop and validate a prediction model for successful IOL, women who had a cesarean delivery for fetal distress or indications other than failure to progress in labor were excluded from the final analysis.
Ultrasound parameters
Once enrolled, study participants underwent a transabdominal scan to evaluate fetal biometry and fetal occiput position, and a transperineal ultrasound examination was conducted to measure the cervical length, angle of progression (AoP), anteroposterior diameter of the levator hiatus, head-to-perineum distance, and head-to-symphysis distance; the last 4 parameters were assessed both at rest and at maximum Valsalva ( Figures 1 and 2 ). Scans were performed using a convex 3.5 to 5 MHz transducer (Voluson 730 Expert, Voluson P8, or Voluson E10, GE Medical Systems, Zipf, Austria) by 1 of 2 operators with more than 3 years of experience in obstetrical and transperineal ultrasound (R.A.K. and A.Y.) who were blind to clinical examination findings. Fetal biometry was conducted in accordance with published International Society of Ultrasound in Obstetrics and Gynecology guidelines. Occiput position determination was made by transabdominal ultrasound as previously published. This was performed by looking for the following landmarks: the fetal occiput, the fetal orbits, the midline of the fetal brain, and the cerebellum. According to theses landmarks, the fetal occiput position was described in relation to a clockface. Occiput position was described as anterior if the occiput was between 09:30 and 02:30 hours, occiput transverse (OT) if between 02:30 and 03:30 hours or 08:30 and 09:30 hours, and occiput posterior (OP) if between 03:30 and 08:30 hours.
For transperineal ultrasound examination, the transducer was covered with a sterile surgical glove. The transducer was placed between the labia majora in a midsagittal plane, aligning the acquisition plane with the long axis of the pubic symphysis. Cervical length was measured along the length of the endocervical canal with simultaneous visualization of the internal os and external os, using a straight line drawn between internal os and external os for the measurement. Transvaginal ultrasound was used in cases of nonoptimal visualization with care not to compress and distort the cervix by the probe. The anteroposterior diameter of the levator hiatus was measured in midsagittal view as the distance between the inferior border of the symphysis pubis to the anterior border of the puborectalis muscle. AoP was measured as the angle between a line running along the long axis of the pubic symphysis and another line extending from the most inferior portion of the pubic symphysis tangentially to the fetal skull contour. Head-to-symphysis distance is the distance along the infrapubic line between the caudal end of the pubic symphysis and the fetal skull. For head-to-perineum distance, the transducer was rotated into a transperineal transverse plane at the level of the posterior commissure and pressed against the pubic rami. Head-to-perineum distance is defined as the shortest distance between the perineum and the outermost part of the bony skull.
Statistical analysis
Simulation studies examining predictor variables for inclusion in logistic regression models suggest that 5 to 10 events are necessary for each candidate predictor to avoid overfitting. Based on 7 events per predictor and the assumption that we will examine 10 candidate predictors, it was estimated that 70 women with the primary outcome of interest (cesarean delivery following IOL owing to failure to progress) would be required. Based on a cesarean delivery rate of 22% following IOL, a sample size of 318 women would be required. Applying the methodology proposed by Riley et al, a global shrinkage factor and adjusted R 2 (R 2 adjust ) are required to estimate the minimum number of events per predictor. In view of the absence of any information regarding these 2 parameters, we assumed that R 2 adjust and shrinkage factor values would be 0.25 and 0.9, respectively. To develop our logistic regression model based on up to 10 predictors and assuming a cesarean delivery rate of 22%, a sample size of 307 would be needed and the events per predictor would be 7 per predictor ( Supplemental Material ).
The study sample (n=344) was randomly divided into 243 and 101 pregnancies that constituted the model development and cross-validation groups, respectively. For model development, the differences of the maternal and ultrasonographic data between the vaginal delivery and cesarean delivery groups were calculated by a Student’s t test (for continuous variables) and the chi-square test (for categorical variables). All variables in the bivariate analysis with P <.2 were evaluated further using multiple logistic regression analysis by computing odds ratios (ORs) and their 95% confidence intervals (CIs). Variables with a P value >.2 were removed from the model. The reduced model was then successively refitted, and the model with the lowest Akaike’s information criteria value was considered the best. Akaike’s information criteria represents the ratio between the number of parameters in the numerator and log likelihood in the denominator ( Supplemental Material ). Akaike’s information criteria score of the model will increase in proportion to the growth in the value of the numerator, which contains the number of parameters in the model (ie, a measure of model complexity). The Akaike’s information criteria score will decrease in proportion to the growth in the denominator, which contains the maximized log likelihood. Thus, a lower value of Akaike’s information criteria suggests a better model.
Only substantial objective variables that predicted the risk of cesarean delivery after IOL were included in the final model. We constructed a receiver operating characteristic (ROC) curve to assess the prognostic accuracy of the devised model. The predicted probability of cesarean delivery was used as the predictive variable with the actual occurrence of cesarean delivery as the tested outcome. The area under the ROC curve (AUC), expressing the prognostic performance of the model, was calculated and compared for statistically significant differences.
We applied bootstrap resampling methodology of AUC as previously described. This method was used to implement 10-fold cross-validation for the AUC for a dependent variable after fitting a logistic regression model and provides the cross-validated fitted probabilities for the dependent variable. Then, bootstrap resampling for AUCs and 95% CIs were generated. Bootstrap resampling methodology was done using Stata Statistical Software, release 13 (StataCorp LP, College Station, TX) with the command of CVAUROC.
The final model was then applied to the cross-validation group by using the holdout sample validation method and an ROC curve was constructed to assess the accuracy of the cross-validated model.
We conducted all data analyses by using statistical software programs (MedCalc, version 12.1.4.0 [MedCalc Software Ltd, Mariakerke, Belgium] and Statistical Product and Service Solutions for Windows, version 21.0 [International Business Machines Corporation, Armonk, NY]).
Results
A total of 382 women who fulfilled the inclusion criteria were enrolled in the study. Of these participants, 38 women underwent a cesarean delivery for unpredictable indications (eg, fetal distress) and were excluded from the study herein, leaving a total of 344 pregnancies contributing to the analysis ( Figure 3 ). It was possible to perform ultrasound scans and assess all the required parameters on all study participants who found it quite acceptable. The characteristics of the study population are shown in Table 1 .
| Variable | Vaginal delivery (n=172) | Cesarean delivery (n=71) | P value |
|---|---|---|---|
| Age (y) | 26.6 (6) | 28.5 (6.4) | .02 |
| Body mass index (kg/m 2 ) | 29 (4) | 31 (5.8) | .001 |
| Gestational age (wk) | 39 (1.5) | 39 (1.5) | .80 |
| Tobacco use | 1 (0.5) | 3 (4) | .04 |
| Fetal sex: male | 85 (49) | 38 (53) | .30 |
| Epidural | 30 (18) | 15 (23) | .30 |
| Prepidil dinoprostone gel | 17 (9.9) | 10 (14) | .416 |
| Propess dinoprostone vaginal insert | 27 (15.7) | 14 (20) | |
| Misoprostol | 128 (74.4) | 47 (66) | |
| Occiput anterior | 48 (28) | 15 (21) | .30 |
| Occiput transverse | 93 (54) | 39 (55) | |
| Occiput posterior | 31 (18) | 17 (24) | |
| Head circumference (mm) | 333 (15) | 334(15) | .40 |
| Biparietal diameter (mm) | 92 (4) | 93 (4) | .27 |
| Femur length (mm) | 72 (4) | 72 (4) | .34 |
| Abdominal circumference (mm) | 337 (21) | 344 (22) | .017 |
| Estimated fetal weight (gm) | 3244 (447) | 3405 (503) | .01 |
| AoP at rest (degrees) | 92.7 (10.8) | 86 (10.7) | <.0001 |
| AoP at Valsalva (degrees) | 100.8 (12.2) | 95.6 (11.4) | .002 |
| Head-to-symphysis distance at rest (mm) | 46.3 (9.8) | 50.6 (11) | .015 |
| Head-to-symphysis distance at Valsalva (mm) | 38.4 (9.8) | 43.2 (11.9) | .006 |
| Head-to-perineum distance at rest (mm) | 51.1 (8.5) | 55.7 (10.6) | .02 |
| Head-to-perineum distance at Valsalva (mm) | 45.3 (7.9) | 49.8 (9.5) | .0003 |
| Anteroposterior diameter of the levator hiatus at rest (mm) | 53.8 (8.7) | 54.9 (8.7) | .39 |
| Anteroposterior diameter of the levator hiatus at Valsalva (mm) | 59.5 (10.4) | 59.6 (11) | .90 |
| Cervical length (mm) | 27.7 (5) | 29.9 (6.8) | .016 |
| Bishop score | 3.6 (1.7) | 3.4 (1.4) | .25 |
We aimed to study variables that are objective, easily assessed, and reproducible to minimize inter- and intraobserver variability and to establish a reliable model. Multivariate logistic regression analysis ( Table 2 ) revealed the independent predictive variables for cesarean delivery to be maternal age (OR, 1.12; 95% CI, 1.03–1.2; P =.003), cervical length (OR, 1.08; 95% CI, 1.002–1.17; P =.04), AoP at rest (OR, 0.9; 95% CI, 0.85–0.96; P =.001), OP position, where occiput anterior (OA) is the reference position, (OR, 5.7; 95% CI, 1.6–19; P =.006).
| Variable | Odds ratio | 95% CI | P value |
|---|---|---|---|
| Age (y) | 1.120 | 1.030–1.200 | .003 |
| Cervical length (mm) | 1.080 | 1.002–1.170 | .040 |
| Angle of progression at rest (degrees) | 0.900 | 0.850–0.960 | .001 |
| Head-to-symphysis distance at Valsalva (mm) | 1.009 | 0.960–1.050 | .600 |
| Occiput position | |||
| Occiput anterior (reference) | — | — | — |
| Occiput transverse | 0.700 | 0.200–2.000 | .600 |
| Occiput posterior | 5.700 | 1.600–19.000 | .006 |
The following equation can calculate the probability of cesarean delivery:
P(CS)=e1.62+0.11Xage+0.08X cervical length−0.09XAOPrest+0.009XHSD−val+(0[OA]|−0.28[OT]|+1.75[OP])1+e1.62+0.11Xage+0.08X cervical length−0.09XAOPrest+0.009XHSD−val+(0[OA]|−0.28[OT]|1.75[OP])
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
