Objective
Universal transvaginal cervical length screening can be associated with a significant logistical burden. We hypothesized that there is a threshold cervical length measured by transabdominal ultrasound above which risk for short transvaginal cervical length is extremely low.
Study Design
This prospective cohort study evaluated a consecutive series of women offered universal transvaginal cervical length screening during anatomy ultrasound. Transabdominal measurement of the cervix—obtained before and after voiding for each patient—was performed before transvaginal ultrasound. The study was powered to detect a transabdominal cervical length cutoff with 95% sensitivity (95% confidence interval, 90–99%) for transvaginal cervical length of ≤25 mm.
Results
One thousand two hundred seventeen women were included in the analysis. Prevoid transabdominal cervical length ≤36 mm detects 96% of transvaginal cervical lengths ≤25 mm with 39% specificity. A prevoid transabdominal cervical length ≤35 mm detects 100% of transvaginal cervical lengths ≤20 mm with 41% specificity. Transabdominal images of the cervix could not be obtained in 6.2% of women prevoid and 17.9% of women postvoid.
Conclusion
Transabdominal cervical length screening successfully identifies women at very low risk for short transvaginal cervical length. Transabdominal screening may significantly reduce the burden of universal cervical length screening by allowing approximately 40% of women to avoid transvaginal ultrasound. To ensure high sensitivity of transabdominal screening, approximately 60% of patients will still require a transvaginal study.
Two randomized trials demonstrated that vaginal progesterone reduces the risk of spontaneous preterm birth in women with short midtrimester cervical length diagnosed by transvaginal (TV) ultrasound. However, the prevalence of a short cervix in a general obstetric population is low. Although recent decision analyses found universal TV ultrasound and treatment with vaginal progesterone to be cost effective, universal screening requires a significant dedication of resources that may not be available at all centers. Universal screening would entail performing millions of additional transvaginal studies annually in the United States.
For Editors’ Commentary, see Contents
Some research studies have suggested that assessment of the cervix by transabdominal (TA) ultrasound may be a useful initial screening test to detect short cervix diagnosed by TV ultrasound thereby more efficiently identifying candidates for vaginal progesterone. However, other studies have found TA ultrasound screening to be a poor test for detecting short cervix ( Table 1 ). Disagreement in the literature regarding the sensitivity of TA screening may be secondary to the relatively small numbers of patients with short cervix included in each study because of the rare occurrence of this condition in an unselected population.
| Study | Gestational age | Number of patients | Bladder status during TA assessment | Mean TA cervical length | Mean TV cervical length | Main findings |
|---|---|---|---|---|---|---|
| Hernadez-Andrade et al | Mean: 24.4 wks Range: 6.3–39 wks | 220 | Prevoid | 34.6 mm | 34.8 mm | TA ≤25 mm 43% sensitive for TV ≤25 mm. TA ≤30 mm 57% sensitive for TV ≤25 mm. |
| Stone et al | Range: 18–20 wks | 203 | Postvoid | 36.6 mm | 39.1 mm | In discrepant cases, TA measurements ≤33 mm were shorter than TV measurements in 97% of cases. |
| Saul et al | Range: 14–34 wks Mean: 22.2 wks | 191 | Postvoid | 35.7 mm | 36.1 mm | TA ≤30 mm 100% sensitive for TV ≤25 mm. TA ≤33 mm 100% sensitive for TV ≤30 mm. |
| To et al | Range: 22–24 wks Mean: 23 weeks | 149 | Bladder volume calculated | 34 mm | 37 mm | Cervix visualized transabdominally in 49% of cases overall. Visualization more likely with increased bladder volume. |
We hypothesized that there is a threshold cervical length measured by TA ultrasound above which the risk of short cervical length (≤25 mm) measured by TV ultrasound is extremely low. To test this hypothesis we determined (1) the proportion of women in whom maternal cervical length could be imaged adequately by TA ultrasound at 18 to 24 weeks’ gestation; and (2) the test characteristics for TA ultrasound as a screening test to detect women with short cervix on TV ultrasound within a universal screening protocol. The identification of a TA cervical length threshold above which TV ultrasound does not have to be performed has the potential to reduce the clinical burden posed to prenatal ultrasound units by universal TV cervical length screening.
Materials and Methods
This prospective cohort study evaluated consecutive patients who underwent anatomy ultrasound in the maternal-fetal medicine division at the Hospital of the University of Pennsylvania between January 2012 and June 2012. Approval for this study was granted by the institutional review board of the University of Pennsylvania (protocol no. 815974). Cervical length measurement data were prospectively collected as part of a quality assurance initiative. Demographic and medical data were collected through review of electronic medical records.
At the study center TV assessment of the cervix was offered as routine care during anatomy ultrasound. Patients were included if they were between 18 weeks’ 0 days’ and 23 weeks’ 6 days’ gestational age and had a singleton gestation at the time of examination. Patients beyond 23 weeks’ 6 days’ gestational age were excluded because prior studies have not evaluated the benefit of initiating vaginal progesterone beyond this gestational age. Two other categories of women were excluded: (1) women with cerclage already in place and (2) women with a prior spontaneous preterm birth already receiving 17-alpha hydroxyprogesterone caproate (17P) who were either not eligible or would not elect for cerclage if found to have a short cervix (based on the findings of Owen et al ). Ultrasound is not offered to these patients because detection of a short cervix would not alter preterm birth prevention strategies. For women with a prior preterm birth who would undergo cerclage if found to have a short cervix, only the TV ultrasound performed during fetal anatomic survey at 18-24 weeks was included in this analysis.
For each patient included in the study, prevoid and postvoid TA cervical length measurements were performed before TV ultrasound. Patients were given instructions during prenatal care to refrain from voiding in the 2 hours before their study. Pre- and postvoid measurement were obtained because some experts have expressed concern that bladder status may affect TA measurement. Sonographers were instructed that if TA cervical length measurements were not readily attainable, the examination time of the study should not be extended to allow bladder filling and TA views should be noted as unobtainable.
Before including patients in the study, sonographers received an in-service training program, which reviewed landmarks for assessing the cervix transabdominally based on the study by Saul et al. Sonographers were instructed to obtain TA images in the midsagittal plane obtaining the following landmarks: the cervical/vaginal interface, the internal cervical os, the external cervical os, the outline of the cervical corpus, and the full length of the cervical canal. TA assessment of the cervix was initiated 1 month before the start of the protocol, and each sonographer was required to submit images from 5 patients demonstrating all landmarks. With the use of these landmarks as a guide, the sonographers were instructed during the study period to obtain the best possible image for TA cervical length assessment regardless of which landmarks were obtained. The TA cervical length was measured with a single linear measurement and rounded to the nearest millimeter. If a sonographer was unable to obtain an acceptable image, they were instructed to designate the TA cervical length “unobtainable.” TV ultrasound was performed according to standard practices. Figures 1 and 2 demonstrate TA prevoid and TA postvoid ultrasound images respectively for a single patient.
Demographic and medical information for each patient was collected from electronic medical records and included body mass index (BMI) calculated from first prenatal visit, gravidity and parity, major medical comorbidities, age, race, and whether the patient had undergone prior cesarean section and prior cervical surgery.
The study was powered to allow for detection of a screen positive rate at which the sensitivity would be 95% (95% confidence interval [CI], 90–99%) for detecting women with short TV cervical length ≤25 mm. The anticipated prevalence for TV cervical length ≤25 mm was 6%. This estimate of short cervical length prevalence was chosen based on findings from our institution accounting for exclusion of women having already received 17P and/or cerclage. The number of patients needed for the analysis based on these assumptions was 1217. A high sensitivity was chosen because false positive cases (performing unnecessary TV ultrasound studies) were preferred over false negative cases (missing women with short TV cervical length). The cutoff of 25 mm was used for our primary outcome because cervical length up to 25 mm is associated with increased preterm birth risk and women with cervical length ≤25 mm may benefit from interventions such as cerclage if they have certain risk factors. Because vaginal progesterone has only been shown to be beneficial up to a cervical length of 20 mm test statistics associated with this cutpoint are reported as secondary outcomes.
Secondary analyses also included generating receiver operating characteristic (ROC) curves to assess detection of women with TV cervical length of ≤25 mm and ≤20 mm. In addition, regression analysis was performed to determine whether demographic variables such as BMI at first prenatal visit, race, parity, and prior cesarean section were associated with increased absolute difference between TA and TV cervical length. Prevoid and postvoid TA cervical length measurements were compared to determine whether there was a significant difference in measurement based on bladder status. For all analyses, patients with unobtainable TA images were considered screen positive with a short measurement and would need to undergo TV as part of a screening protocol.
All analyses were performed separately for both prevoid and postvoid measurements. However, primary and secondary outcomes are reported based on prevoid TA measurements; researchers have expressed concern that TA ultrasound with a full bladder may cause the cervix to appear falsely elongated. By using TA images with a full bladder, the study design is biased toward overestimation of TA cervical length and TA screening failing as a good screening test.
STATA 10.0 was used to perform statistical analysis (StataCorp, College Station, TX). Categorical variables were studied using χ 2 tests or Fisher exact as appropriate. Means of normally distributed continuous variables were studied using Student t tests or nonparametic statistics as appropriate. Paired data were accounted for using appropriate statistical methods. A P value of < .05 was considered significant for all analyses.
Results
There were 1349 patients with singleton pregnancy who met inclusion criteria and were entered into the study. A total of 132 women (9.8%) were excluded because they declined TV ultrasound; consequently, 1217 women underwent TA and TV cervical length screening and were included in our final analyses ( Figure 3 ). In this cohort 6.2% of patients had TV cervical length ≤25 mm (n = 76) and 2.6% of patients (n = 32) had TV cervical length ≤20 ( Table 2 ). The mean TV cervical length was 36.1 mm (SD 8.3 mm; 95% CI, 35.6–36.6 mm), the mean prevoid TA cervical length was 34.6 mm (SD 8.4 mm; 95% CI, 34.1–35.1 mm), and the mean postvoid TA cervical length was 33.5 mm (SD 8.4 mm; 95% CI, 33.0–34.1 mm). Both prevoid and postvoid mean TA cervical length were significantly shorter than mean TV cervical length ( P < .01). A scatterplot demonstrates the relationship between TV and prevoid TA cervical length in Figure 4 .
| Characteristic | Value |
|---|---|
| Age (mean, SD) | 28.2 (6.3) |
| Race, % | |
| Black | 63.8 |
| White | 23.0 |
| Asian | 7.6 |
| Other | 4.3 |
| Unknown | 1.3 |
| Ethnicity, % | |
| Hispanic | 2.7 |
| Not Hispanic | 97.3 |
| Gestational age (mean, SD) | 20.5 (1.0) |
| Nulliparous, % | 43.1 |
| BMI (mean, SD) | 27.8 (7.8) |
| Medicaid, % | 58.7 |
| Medical comorbidity, % | |
| Hypertension | 5.1 |
| Pregestational diabetes | 2.0 |
| Renal disease | 0.4 |
| Other major medical comorbidity | 3.1 |
| Prior spontaneous preterm birth, % | 5.9 |
| History of cervical surgery, % | 3.0 |
| Uterine anomaly, % | 0.7 |
| TV CL ≤20 mm, n (%) | 32 (2.6) |
| TV CL ≤25 mm, n (%) | 76 (6.2) |
Prevoid TA ultrasound was 96.1% sensitive at a cutoff of ≤36 mm for detecting short cervix on TV ultrasound of ≤25 mm (95.0% CI, 90.0–99.2%) with a specificity of 39.4% (95% CI, 36.7–42.2%) ( Table 3 ). The area under the ROC for TV cervical length ≤25 mm was 0.76 ( Figure 5 ). Prevoid TA ultrasound was 100% sensitive at a cutoff of ≤35 mm for detecting short cervix on TV ultrasound ≤20 mm (95% CI, 89.1–100.0%). Specificity was 40.8% (95% CI, 38.0–43.7%) at this cutoff ( Table 4 ). The area under the ROC for TV cervical length ≤20 mm was 0.86 ( Figure 6 ).
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
