Objective
To determine whether the predictive accuracy of sonographic cervical length (CL) for preterm delivery (PTD) in women with threatened preterm labor (PTL) is related to gestational age (GA) at presentation.
Study Design
A retrospective cohort study of all women with singleton pregnancies who presented with PTL at less than 34 + 0 weeks and underwent sonographic measurement of CL in a tertiary medical center between 2007 and 2012. The predictive accuracy of CL for PTD was stratified by GA at presentation.
Results
Overall, 1077 women who presented with PTL have had sonographic measurement of CL and met the study inclusion criteria. Of those, 223 (20.7%) presented at 24 + 0−26 + 6 weeks (group 1), 274 (25.4%) at 27 + 0−29 + 6 weeks (group 2), 283 (26.3%) at 30 + 0−31 + 6 weeks (group 3), and 297 (27.6%) at 32 + 0−33 + 6 weeks (group 4). The overall performance CL as a predictive test for PTD was similar in the 4 GA groups, as reflected by the similar degree of correlation between CL with the examination to delivery interval (r = 0.27, r = 0.26, r = 0.28, and r = 0.29, respectively, P = .8), the similar area under the receiver-operator characteristic curve (0.641-0.690, 0.631-0.698, 0.643-0.654, and 0.678-0.698, respectively, P = .7), and a similar decrease in the risk of PTD of 5-10% for each additional millimeter of CL. The optimal cutoff of CL, however, was affected by GA at presentation, so that a higher cutoff of CL was needed to achieve a target negative predictive value for delivery within 14 days from presentation for women who presented later in pregnancy. The optimal thresholds to maximize the negative predictive value for delivery within 14 days were 36 mm, 32.5 mm, 24 mm and 20.5 mm for women who presented at 32 + 0 to 33 + 6 weeks, 30 + 0 to 31 + 6 weeks, 27 + 0 to 29 + 6 weeks and 24 + 0 to 26 + 6, respectively.
Conclusion
CL has modest predictive accuracy in women with threatened PTL, regardless of GA at presentation. However, the optimal cutoff of CL for the purpose of clinical decision making in women with PTL needs to be adjusted based on GA at presentation.
See related editorial, page 443
Prematurity is a major cause for neonatal mortality and morbidity. Different modalities have been used to predict preterm delivery (PTD) in women presenting with preterm labor (PTL) including digital examination, fetal fibronectin, and several biomarkers, but individually these were all found to have only limited predictive ability. Thus, the distinction between true and false PTL is often challenging as only fewer than 15% of women presenting with threatened PTL will actually deliver prematurely.
Cervical length (CL) has been previously suggested to be associated with the risk of PTD in women with PTL. However, it is reasonable to assume that the predictive accuracy of CL in women with PTL would be affected by factors such as the definition of PTL and the copresence of other risk factors for PTD. Another potential significant factor that may affect the predictive accuracy of CL in these cases is gestational age (GA) at the time of presentation with PTL. Considering the fact that the cervical canal undergoes a physiologic gradual shortening as pregnancy advances, it is reasonable to assume that the predictive accuracy of cervical length for PTD as well as the optimal cervical length cutoff for the prediction of PTD would be related to gestational age at the time of presentation with PTL. Nevertheless, data in support of this assumption and on the use of GA-specific cutoff values are limited.
3Thus, our aim was to determine whether the predictive accuracy of sonographic CL in women with threatened PTL is related to GA at presentation and to calculate GA-specific optimal cutoff values of CL for the prediction of PTD.
Materials and Methods
Study population
This was a retrospective cohort study of women who presented with threatened PTL in the presence of intact membranes and underwent sonographic measurement of CL according to a standardized protocol in a tertiary referral medical center between January 2007 and December 2012. Only women with singleton pregnancies who presented at GA of 24 + 0 to 33 + 6 weeks were included. Women with multiple gestations, cervical cerclage, cervical dilatation >3 cm at presentation, pregnancies complicated by placental abruption, chorioamnionitis, stillbirth or major fetal anomalies, and women who underwent indicated delivery before 37 + 0 weeks of gestation or did not deliver in our medical center were excluded. The study was approved by the local institutional review board.
Data collection
Women were identified using the comprehensive database of sonographic examinations in our ultrasound unit. Initially, all women who underwent sonographic measurement of CL using transvaginal ultrasound were identified and the information regarding CL, presence of funneling, change in CL during the examination (either spontaneously or in response to external uterine fundal pressure or Valsalva maneuver) was extracted. The medical charts of these women were then thoroughly reviewed to identify only those women for whom the indication for sonographic measurement of CL was threatened PTL. For these women, the medical charts were then reviewed for the following information: demographic, medical and obstetric history, complications during current pregnancy, validation of GA by first trimester ultrasound, GA at presentation, frequency and regularity of uterine contractions, results of vaginal examination, interventions for threatened PTL, GA at delivery, and type of delivery (ie, spontaneous vs indicated). For women with repeated measurements of CL at different times along gestation only the first measurement was included in the analysis.
Definitions
Threatened PTL was defined as the presence of at least 3 regular and painful uterine contractions within 30-minute period. This definition was used for the purpose of concordance with previous studies addressing this issue. Tocolysis with either nifedipine, indomethacin or tractocile, and betamethasone for fetal lung maturation were administered based on the decision of the attending physician.
The measurement of CL was performed transvaginally only after the patients have emptied their bladder and according to the standard technique. Briefly, the measurement of CL is performed in the sagittal plane, visualizing the full length of the endocervial canal from the internal to the external cervical os although exerting as little pressure with the transducer as possible. At least 3 measurements are obtained and the shortest measurement is recorded. The presence of cervical funneling or change in CL, either spontaneously or in response to external uterine fundal pressure or Valsalva maneuver are routinely documented as well.
Statistical analysis
Data analysis was performed with the SPSS v21.0 software. The relationship between CL at presentation and the risk of spontaneous PTD (at <37 and 35 weeks and within 14 and 7 days from presentation) was analyzed for 4 prespecified groups of patients based on GA at the time of presentation: 24 + 0 to 26 + 6 weeks (group 1), 27 + 0 to 29 + 6 weeks (group 2), 30 + 0 to 31 + 6 weeks (group 3), and 32 + 0 to 33 + 6 weeks (group 4).
The characteristics of the different groups were compared using the χ 2 test and 1-way analysis of variance for categorical and continuous variables, respectively.
The Spearman’s correlation coefficient was used to assess the correlation between CL at presentation and examination to delivery interval and the correlation coefficients for the 4 groups were compared using Fisher Z-transformation. Multivariable logistic regression analysis was used to assess the association between CL (as a continuous measure) and the risk of PTD within each of the 4 gestational age groups whereas adjusting for potential confounders. Receiver-operator characteristic (ROC) analysis was used to determine the area under the ROC curve (AUC) as an overall measure of the discriminative ability of CL. The AUC for the 4 groups were compared using the method of Hanley and McNeil.
The effect of the cutoff of CL on the predictive accuracy of CL for PTD was calculated for each of the GA groups. The following measures of predictive accuracy were calculated: sensitivity, specificity, and positive and negative predictive values (PPV and NPV, respectively).
Results were considered significant when P value was less than .05.
Results
Characteristics of the study groups
Of a total of 2680 sonographic examinations of cervical length performed in our ultrasound unit during the study period, 1766 were performed in women who presented with threatened PTL, and 1077 women were found to be eligible for the study ( Figure 1 ). Of those, 223 (20.7%) presented between 24 + 0 to 26 + 6 weeks (group 1), 274 (25.4%) presented between 27 + 0 to 29 + 6 weeks (group 2), 283 (26.3%) presented between 30 + 0 to 31 + 6 weeks (group 3), and 297 (27.6%) presented between 32 + 0 to 33 + 6 weeks (group 4) ( Figure 1 ).
The characteristics and pregnancy outcome of the women in the 4 study groups are presented in Table 1 . The study groups differed with respect to maternal age and the proportion of women with a history of previous PTD, so that women who presented before 30 weeks of gestation (groups 1 and 2) were older and were more likely to have a history of PTD ( Table 1 ). Women who presented earlier in pregnancy had a higher mean CL at presentation. In addition, women who presented earlier in pregnancy had a lower mean GA at delivery and were more likely to experience PTD at less than 35 or 32 weeks, but had a higher examination to delivery interval and were less likely to deliver within 14 days or 7 days from presentation compared with women who presented with PTL later in pregnancy ( Table 1 ).
| Characteristic | Group 1 24+0-26+6 wks, n = 223 | Group 2 27+0-29+6 wks, n = 274 | Group 3 30+0-31+6 wks, n = 283 | Group 4 32+0-33+6 wks, n = 297 | P value |
|---|---|---|---|---|---|
| Maternal age, y | 33.7 ± 5.8 | 33.6 ± 5.9 | 33.0 ± 5.5 | 32.3 ± 5.0 | .01 |
| Age >35 y | 81 (36.3) | 100 (36.5) | 93 (32.9) | 73 (24.6) | .007 |
| Nulliparity | 90 (40.4) | 121 (44.2) | 133 (47.0) | 153 (51.5) | .07 |
| Past preterm delivery | 34 (15.2) | 54 (19.7) | 29 (10.2) | 35 (11.8) | .007 |
| Cervical length measurement | |||||
| GA at examination, wks | 25.6 ± 0.8 | 28.5 ± 0.8 | 31.0 ± 0.6 | 32.9 ± 0.6 | < .001 |
| Cervical length, mm | 32.4 ± 11.2 | 32.0 ± 9.8 | 30.6 ± 10.0 | 29.0 ± 9.6 | < .001 |
| <10 mm | 7 (3.1) | 6 (2.2) | 8 (2.8) | 5 (1.7) | .7 |
| <15 mm | 21 (9.4) | 13 (4.7) | 23 (8.1) | 22 (7.4) | .2 |
| <20 mm | 32 (14.3) | 33 (12.0) | 34 (12.0) | 57 (19.2) | .045 |
| <25 mm | 46 (20.6) | 58 (21.2) | 71 (25.1) | 93 (31.3) | .01 |
| <30 mm | 71 (31.8) | 93 (33.9) | 117 (41.3) | 141 (47.5) | .001 |
| Delivery outcome | |||||
| GA at delivery, wks | 36.8 ± 3.9 | 37.4 ± 2.6 | 37.6 ± 2.1 | 37.7 ± 2.0 | < .001 |
| <37 + 0 wks | 66 (29.6) | 69 (25.2) | 67 (23.7) | 72 (24.2) | .4 |
| <35 + 0 wks | 47 (21.1) | 26 (9.5) | 30 (10.6) | 25 (8.4) | < .001 |
| Examination to delivery interval, d | 71.6 ± 22.1 | 64.4 ± 18.6 | 49.2 ± 14.9 | 37.1 ± 15.0 | < .001 |
| Within 7 days of examination | 4 (1.8) | 5 (1.8) | 8 (2.8) | 11 (3.7) | .2 |
| Within 14 days of examination | 10 (4.5) | 11 (4.0) | 19 (6.7) | 31 (10.4) | .008 |
| Birthweight, g | 2727 ± 768 | 2890 ± 626 | 2991 ± 555 | 2985 ± 552 | < .001 |
| Male fetus | 117 (52.5) | 152 (55.5) | 164 (58.0) | 151 (50.8) | .3 |
Risk of spontaneous preterm delivery by CL at presentation–the effect of GA at presentation
The rate of PTD at less than 37, 35, and within 14 and 7 days from presentation as a function of CL at presentation is presented in Figure 2 . Overall, the expected inverse relationship between CL at presentation and the rate of PTD at less than 37 and 35 weeks and within 14 days from presentation was observed in all of the 4 GA groups and was thus unaffected by GA at presentation ( Figure 2 ).
However, the rate of PTD at less than 37 or 35 weeks of gestation for any given CL at presentation was higher for women who presented before 27 weeks of gestation (group 1) compared with women who presented later in pregnancy ( Figure 2 ). In contrast, the rate of PTD within 14 and 7 days from presentation for any given CL at presentation was higher for women who presented after 32 + 0 weeks of gestation (group 4) compared with women who presented earlier in pregnancy ( Figure 2 ).
The effect of GA at presentation on the performance of CL as a predictive test for PTD
We next assessed the performance of CL as a predictive test for PTD in each of the GA groups. First, the correlation between CL and the examination to delivery interval was found to be similar for all of the 4 GA groups (r = 0.27, r = 0.26, r = 0.28, and r = 0.29, respectively; P = .8) ( Figure 3 ).
We also used multivariable logistic regression analysis to assess the association between CL (as a continuous variable) and risk of PTD in each of the GA groups whereas controlling for potential confounders including parity, age, history of PTD, and GA at presentation ( Table 2 ). The degree of association between CL and the risk of each of the PTD outcome variables was similar for all of the GA groups, as reflected by the overlapping confidence intervals ( Table 2 ). Thus, the risk of PTD decreased by 5-10% for each additional millimeter of CL at presentation irrespective of GA at presentation ( Table 2 ). This was also confirmed by the fact that the interaction term [CL*GA at presentation] was not significant and was thus not included in the regression model, implying that GA at presentation did not affect the association between CL and the risk of PTD.
Gestational age at presentation | Association of cervical length (as a continuous variable) with risk of preterm delivery | |||
|---|---|---|---|---|
| Delivery <37 wks | Delivery <35 wks | Delivery within 7 days | Delivery within 14 days | |
| 24+0 and 26+6 wks | 0.95 (0.92–0.98) | 0.94 (0.91–0.96) | 0.92 (0.87–0.98) | 0.93 (0.88–0.98) |
| 27+0 and 29+6 wks | 0.94 (0.92–0.97) | 0.94 (0.90–0.98) | 0.96 (0.88–1.05) | 0.92 (0.86–0.97) |
| 30+0 and 31+6 wks | 0.94 (0.92–0.97) | 0.95 (0.91–0.98) | 1.01 (0.91–1.12) | 0.94 (0.87–0.99) |
| 32+0 and 33+6 wks | 0.93 (0.90–0.96) | 0.90 (0.86–0.95) | 0.96 (0.90–1.02) | 0.93 (0.89–0.97) |
In addition, the ability of CL to discriminate between women who will or will not deliver prematurely following presentation with PTL, as reflected by the area under the ROC curve (AUC), was similar between different GA groups (0.641-0.690, 0.631-0.698, 0.643-0.654, and 0.678-0.698, respectively; P = .7-.8) ( Table 3 ).
| Gestational age at presentation | Area under the ROC curve | ||
|---|---|---|---|
| Delivery <37 wks | Delivery <35 wks | Delivery within 14 days | |
| 24+0 to 26+6 wks | 0.661 | 0.690 | 0.641 |
| 27+0 to 29+6 wks | 0.631 | 0.643 | 0.698 |
| 30+0 to 31+6 wks | 0.654 | 0.643 | 0.646 |
| 32+0 to 33+6 wks | 0.678 | 0.698 | 0.693 |
| P value a | 0.8 | 0.7 | 0.8 |
a Comparisons of the AUC were made using the method of Hanley and McNeil.
The effect of GA at presentation with threatened PTL on the optimal cutoff of CL
Although the performance of CL as a predictive test for PTD was not affected by GA at presentation, the optimal cutoff of CL that should guide clinical decision making (eg, admission, administration of corticosteroids) may be related to GA at presentation, considering the differences in the rate of PTD in each of the groups and the physiologic gradual shortening of CL as pregnancy advances. For that reason, we next analyzed the effect of GA at presentation on the relationship between the cutoff of CL and the predictive accuracy of CL for PTD ( Figures 4 and 5 ).
