The predictive value of cervical shear wave elastography in the outcome of labor induction

Abstract Introduction Bishop score, the traditional method to assess cervical condition, is not a promising predictive tool of the outcome of labor induction. As an objective assessment tool, many cervical ultrasound measurements have been proposed to represent the individual components of the Bishop score, but none of them can measure the cervical stiffness. Cervical shear wave elastography is a novel tool to assess the cervical stiffness quantitatively. Material and methods A total of 475 women who required labor induction were studied prospectively. Prior to routine digital assessment of the Bishop score, transvaginal sonographic measurement of cervical length, posterior cervical angle, angle of progression and shear wave elastography was performed. Shear wave elastography measurement was made at the inner, middle and outer regions of the cervix to assess homogeneity. Association of labor induction outcomes including the overall cesarean section and subgroups of cesarean section for failure to enter active phase, with cervical sonographic parameters and the Bishop score, were assessed using multivariate regression analyses. The predictive accuracy of the outcomes using models based on ultrasound measurement and the Bishop score was compared using the area under the receiver‐operating characteristics curves. Results Among 475 women, 82 (17.3%) required cesarean section. Shear wave elasticity was significantly higher in the inner cervical region than in other regions, indicating a greater stiffness (P < 0.001). Both inner cervical shear wave elasticity and cervical length were independent predictors of overall cesarean section (respective adjusted odds ratio [95% CI] 1.338 [1.001‐1.598] and 1.717 [1.077‐1.663]) and cesarean section for failure to enter active phase (respective adjusted odds ratio [95% CI] 1.689 [1.234‐2.311] and 2.556 [1.462‐4.467]), after adjusting for other covariates. Outcome prediction models using inner cervical shear wave elasticity and cervical length, had increased area under curve compared with models using the Bishop score (0.888 vs 0.819, P = 0.009). Conclusions The cervix is not a homogenous structure, with the inner cervix having the highest stiffness, which is an independent predictor of overall cesarean section, and specifically for those indicated because of failure to enter active phase. Models based on shear wave elastography and cervical length had higher predictive accuracy than models based on the Bishop score.


| INTRODUC TI ON
Approximately one in five inductions of labor (IOL) results in an emergency cesarean section (CS) due to failure to reach active phase or labor, failure to progress beyond the active phase or fetal distress. 1 The expected outcome and management of IOL has traditionally been based on vaginal digital assessment of the cervix to assess the Bishop Score (BS). 2 However, the BS has been shown to be subjective 3 and has relatively low predictive performance. 4 Recent research has therefore focused on the use of ultrasound for more objective assessment of individual components of BS. Besides cervical length, posterior cervical angle (PCA) 5 and angle of progression (AOP) 6 can be measured sonographically to reflect the cervical position and fetal head descent, respectively. Although some studies have shown that these ultrasonic measurements are superior to digital assessment, their predictive values remain suboptimal for clinical use. 7,8 One main reason for this is the inability to measure cervical consistency, a major component of BS, using conventional ultrasound technology. Hence strain-based sonoelastography has also been investigated for measuring uterine cervical stiffness. However, this requires human movements on the ultrasound probe to generate a 'stress' on the target tissue. It is limited to measuring relative strain of the target tissue in comparison with its adjacent tissues, but the cervix lacks good surrounding tissues to act as reference. 9 Several small-cohort studies have conflicting results regarding the predictive value of strain-based elastography. [10][11][12][13] In contrast, shear wave elastrography (SWE) uses ultrasound pulses to generate shear waves across a target tissue, and the shear wave velocity ('v') correlates to the tissue stiffness. Young's modulus (E) in kPa can be estimated using the formula E ≅ 3ρv 2 , where ρ is the density of the tissue (kg/m 3 ), which is assumed to be constant. 14 SWE assessment of the cervix has recently been reported to have good intra-and interobserver reproducibility. 15 Cervical stiffness using SWE was also shown to decrease with gestational age. 15 Yet its value in predicting IOL outcome is unknown. Therefore the aim of the present study is to evaluate whether cervical SWE can improve the predictive performance of the outcome of IOL when combined with other ultrasound-based assessments as compared with BS alone.

| Study design
This was a prospective observational study conducted between September 2015 and November 2017. Women admitted for IOL were invited to participate. The inclusion criteria were: (1) Chinese women carrying singleton pregnancies; (2) ≥37 gestational weeks; (3) vertex presentation; (4) normal fetal well-being on cardiotocography. Women who had a history of cervical surgery or any contraindication for vaginal delivery were excluded.

| Measurement of cervical SWE
Participants were assessed on admission for IOL. Before the digital assessment for the BS, women were asked to empty the bladder and in modified lithotomy, a transvaginal scan of the cervix was performed using an SE 12-3 probe (3-12 MHz) of the SuperSonic Imagine ultrasound system (Aixplorer supersonic imagine, Aix-en-Province, France). The probe was inserted gently without any pressure being exerted on the cervix, and the mid-sagittal view of the cervix was identified by clear visualization of internal os, canal and external os. The cervical image was magnified to occupy at least 75% of the screen. Once an optimal image of the cervix was obtained, a sampling box was put over the anterior lip and then the posterior lip of the cervix. To optimize the quality of the elastogram color image, each time the size of the sampling box was adjusted to just fit either the anterior or the posterior cervix. Each of the cervical lips was divided into three equal parts along its longitudinal axis: the inner part (proximal one-third), the middle part, and the outer part (distal one-third). Then the SWE value of each region of Conclusions: The cervix is not a homogenous structure, with the inner cervix having the highest stiffness, which is an independent predictor of overall cesarean section, and specifically for those indicated because of failure to enter active phase. Models based on shear wave elastography and cervical length had higher predictive accuracy than models based on the Bishop score. interest (ROI) was measured with a 5-mm-diameter circle placed at the center of each ROI. The SWE value was automatically displayed in pressure units (kPa) on the screen (Figure 1). The sampling method was repeated two more times from each cervical lip so as to obtain three independent SWE measurements of each ROI. The average of these three measurements was used for analysis.
The scan were performed by trained ultrasonographers. Their intra-and interobserver reproducibility was assessed among the first consecutive 30 women, who were assessed twice by the same operator, and then reassessed by the second operator. All the measurements were made on the independent images selected from the saved images.
Both operators were blinded to the measurements made by each other.

| Other ultrasonic assessment
The cervical length was measured as the linear distance between the internal os and the external os. 16 The PCA was defined as the in-

| Management of IOL
Subsequently, an independent obstetrician, blinded to the ultrasound findings, performed the per-vaginal digital examination to determine the BS. The decision on the method of IOL was based on the BS. The standard practice of the studying unit was that, when the BS was ≥6, the cervix was regarded as ripened or favorable, and the IOL proceeded with amniotomy and/or syntocinon infusion; when the BS was <6, vaginal prostaglandin E2 (PGE2) gel or dinoprostone pessary was used. All clinical staff were blinded to the ultrasound findings. Failure to enter active phase was defined as failure of the cervix to efface and dilate to 3 cm in 12 hours after amniotomy or initiation of syntocinon infusion, or remaining unfavorable (BS <6) in 24 hours after a single pessary of 10 mg dinoprostone or three doses of 3 mg PGE2 gel. Failure to progress in the active phase was defined as cervical dilation slower than 1 cm/h for 4 hours during the active phase of labor. Fetal distress was defined as the presence of pathological cardiotocography which required immediate delivery. 20

| Sample size
Our previous model to predict outcome of induction based on BS alone gave an AUC of 0.65. 7 To detect a change of 0.1 in AUC with a new prediction model would require a minimum sample size of 425 for a type 1 error of 0.05, 80% power and assuming a CS for failed induction of 20% and a correlation between AUC of 0.5. Planned sample size was increased by a further 10% to 468 to allow for up to 10% failure rate to measure one or more ultrasound markers.

| Statistical analyses
The primary outcome of IOL was successful vaginal delivery vs CS. A two-tailed P value <0.05 was considered statistically significant.

| RE SULTS
A total of 500 pregnant women were recruited, of which 25 were excluded because 12 cases had early sign of spontaneous onset of labor, 4 cases had CS due to suspected macrosomia and 9 cases declined IOL, leaving 475 cases for IOL. The demographic characteristics of the studied population are shown in to failure to progress, and three due to other reasons.
The ICCs of intra-and interobserver reproducibility were >0.85 in each ROI (Table S1, Figures S1 and S2). Table 2 gives the SWE values at each ROI and shows that an elastic gradient exists along the longitudinal axis, with the inner part being significantly stiffer than the middle part, and the middle part being significantly stiffer than the outer part, along both the anterior (5.4 kPa vs 4.8 kPa vs 3.8 kPa; all P < 0.001) and posterior lips (5.0 kPa vs 4.7 kPa vs 3.9 kPa; all P < 0.001). The SWE values at different ROI are also significantly intercorrelated with each other (Spearman coefficients are shown in Table 3) (all P < 0.001). Hence, for subsequent comparison we used the inner cervical SWE (the mean SWE of the inner anterior and inner posterior cervix), the stiffest region.
Comparison of maternal characteristics, the BS, fetal and cervical sonographic measurements between the vaginal delivery group  Note: Data are given as median (range) or n (%). Abbreviation: BMI, body mass index. and the whole group of CS is illustrated in Table 4.  (Figure 4). Body mass index (BMI) ≥0 kg/m 2 , Bishop score and AOP were not independent predictors. an AUC of 0.888 (95% CI 0.853-0.916; Figure 5). If the two cervical ultrasound measurements were replaced by the Bishop score, the AUC significantly dropped to 0.819 (95% CI 0.778-0.855).
The difference between two AUCs was 0.0687 with a 95% CI of 0.0175-0.12 (DeLong test: z = 2.631, P = 0.009). As the multi-parity is the most significant predictor for success of IOL, we further focused on the nulliparous subgroup, and found that sonographic prediction was even stronger than BS among nulliparous women  Table 8.
In the subgroup of CS indicated for failure to progress in the active phase, EFW was significantly higher and the Bishop score lower than in the vaginal delivery group. The BMI was also higher and maternal height less, and there were fewer multiparous women. In the subgroup of CS for fetal distress, the MCA PI were significantly lower, the proportion with EFW below 10th percentile was higher, the mothers were shorter and there were fewer multiparous women; the BS was also lower, all based on univariate analysis (Table 6).

| D ISCUSS I ON
This is the first study using SWE to predict the outcome of IOL, Our finding of decreasing stiffness from the inner to the outer part of the cervix is concordant with several studies that have shown the spatial heterogeneity in the stiffness within the cervix using SWE. 15 This has been hypothesized to be attributable to the cervical collagen fiber orientation. 23 The collagen cross-link around the internal os is significantly more heterogeneous than that around the external os, and therefore the stroma around the internal os functions distinctively from the external os. 23 Hernandez-Andrade et al found that the stiffness of the inner cervix is more predictive of spontaneous preterm delivery. 24 They showed that a hard internal os at 16-24 weeks is 80% less likely to have spontaneous preterm delivery compared with a soft internal os. 24 In a small cohort study, a hard internal os was associated with the failure of IOL. 10 Therefore, the inner cervical SWE was selected in the regression analysis. Besides the objective measurement of the cervical stiffness, SWE also has a potential advantage over manual examination, as the latter cannot easily access the innermost part of the cervix.   We also found that PCA and AOP, which are respectively the proxies of cervical position and fetal head station in the BS, are no longer independent predictors when SWE is included. This result provides further evidence of intercorrelation between the different components of the BS. 29 As shown in our comparison of the regression models ( Figure 5, Receiver-operating characteristics curves were constructed and the Youden index was used to determine the optimal cutoff. Abbreviations: AUC, area under curve; CL, cervical length; inner cervical SWE, mean of shear wave elasticity of anterior and posterior inner cervix; LR, likelihood ratio; NPV, negative predictive value; PPV, positive predictive value. but rather restricted fetal growth and fetal compromise, whereas failure to progress is related more to large fetal size, as reflected from our results ( Table 6). Our findings also indicate that it is not straightforward to create a prediction model for all CS. Whereas a large fetal weight increases the chance of CS for slow progress, a small fetus is associated with CS for fetal distress. The effect of fetal weight may be masked when overall CS is the primary outcome.
The major strengths of our study are that, by measuring elasticity in different regions of the cervix, we demonstrated that the inner part of the cervix is the most useful predictor of different regions of the cervix. The large sample size from a homogeneous ethnic group is another advantage of our study. However, the overall number of CS of 80 can only allow a maximum of eight variables for multivariate analysis.
Therefore we could only select the eight strongest variables based on univariate analysis. 21 Nonetheless, our study has tested multiple clinical and ultrasonic variables, of which the combination has significantly improved the prediction compared with using clinical variables alone. 31 The choice of the method of IOL was based on the BS alone. It is worth investigating in future research whether SWE may provide a better guide of IOL method and improve the chances of success.

| CON CLUS ION
Shear wave elastography is a useful tool in pre-IOL assessment of the stiffness of the cervix, which is an independent predictor of overall CS, and specifically CS indicated for the failure to enter active phase.
PCA, AOP and the Bishop score were not independent predictors of CS. The combination of sonographic cervical length and shear-wave elastography is superior to the Bishop score in predicting failure of IOL.

ACK N OWLED G M ENTS
We thank Ms WM Yuan and Ms HW Cheung for their contribution in the case recruitment.

CO N FLI C T O F I NTE R E S T
The authors have stated explicitly that there are no conflicts of interest in connection with this article.