Background
There is a lack of consensus on the optimal transvaginal cervical length for determining risk for spontaneous preterm birth in twin pregnancies. Change in transvaginal cervical length over time may reflect early activation of the parturition process, as has been demonstrated in singleton pregnancies. The association between change in transvaginal cervical length and the risk for spontaneous preterm birth has not yet been described in the population of women with diamniotic twin pregnancies.
Objective
Our primary objective is to determine whether rate of change in transvaginal cervical length in the midtrimester is associated with spontaneous preterm birth in twin gestations. Our secondary objective is to describe parameters for identifying patients at increased risk for spontaneous preterm birth based on change in transvaginal cervical length over time.
Study Design
This is a retrospective cohort of serial transvaginal cervical length performed for twin pregnancies at a single institution from 2008 through 2015. Women with diamniotic twin pregnancies who had transvaginal cervical length measurements at 18 and 22 weeks’ gestation and outcome data available were included. Logistic regression was used to determine the relationship between the rate of change in transvaginal cervical length and the risk for the primary outcome of spontaneous preterm birth <35 weeks as well as spontaneous preterm birth <32 weeks.
Results
In all, 527 subjects met inclusion criteria for this study. The average rate of change in transvaginal cervical length for patients with spontaneous preterm birth <35 weeks was –0.21 cm/wk (SD 0.27) vs –0.10 cm/wk (SD 0.24) for patients who delivered ≥35 weeks ( P < .01). The rate of change in transvaginal cervical length was associated with spontaneous preterm birth <35 weeks when controlling for initial transvaginal cervical length and other important risk factors for spontaneous preterm birth. Results for spontaneous preterm birth <32 weeks were similar. This association remained significant when the rate of weekly change was treated as a dichotomous variable based on an apparent inflection point in the risk for spontaneous preterm birth: women with rapid change in transvaginal cervical length, ≥–0.2 cm/wk, had 3.45 times the odds of spontaneous preterm birth as those with less rapid change (95% confidence interval, 2.15–5.52) when controlling for initial transvaginal cervical length.
Conclusion
Change in transvaginal cervical length in the midtrimester is associated with spontaneous preterm birth, and therefore protocols for serial transvaginal cervical length measurement can provide the clinician with information to identify at-risk patients. A decrease of ≥0.2 cm/wk of transvaginal cervical length identifies patients at increased risk for spontaneous preterm birth <35 weeks.
Introduction
Preterm delivery is the predominant complication of multigestational pregnancies, however the risk factors for spontaneous preterm birth (SPTB) remain poorly defined. Ultrasound for measurement of transvaginal cervical length (TVCL) has proven to be the most useful tool available to date for assessing the risk for SPTB in singleton pregnancies. Risk-reducing interventions, such as vaginal progesterone and cerclage, have therefore targeted women with a sonographically short cervix. Serial assessment of TVCL has also been considered in singleton pregnancies with a short cervix (<2.5 cm). Change in TVCL over time has been associated with increased risk of SPTB for women with a short cervix, although this association does not exist in women without a short cervix. Whether a physiologic degree of cervical shortening occurs in twin pregnancies or whether decrease in TVCL is associated with an increased risk for SPTB has not been determined.
Prior research has examined the relationship between TVCL cut-offs at a range of gestational ages and the risk for SPTB in pregnancies with multiple gestation. These studies have provided conflicting information: proposed cut-offs for TVCL range from 20-35 mm and optimal gestational age for measurement ranges from 18-28 weeks’ gestation. Protocols for serial TVCL measurement have not been shown to change gestational age at delivery or neonatal outcomes. Thus, while TVCL is a promising tool for risk assessment in twin pregnancies, definitions of physiologic vs pathologic parameters remain ill defined.
The present analysis focuses on the change in TVCL between 18 weeks, when a patient would be seen for an anatomic survey, and a second assessment around 22 weeks, prior to viability and at a gestational age when TVCL may be a better predictor of SPTB. The purpose of the present study is first to determine whether the rate of change in TVCL on serial assessments in the midtrimester is associated with SPTB in women with diamniotic twin pregnancies, and secondly to describe parameters for identifying patients at increased risk for SPTB based on change in TVCL over time.
Materials and Methods
This is a retrospective cohort study of women with diamniotic twin pregnancies who underwent serial TVCL measurements at Columbia University Medical Center from 2008 through 2015. This clinical investigation was approved by the Columbia University Institutional Review Board (protocol IRB-AAAP3710).
Women in the present cohort underwent serial sonographic cervical length measurements between 16-28 weeks’ gestation at 2-week intervals as part of our institution’s routine antenatal surveillance protocol for women with multiple gestations. More frequent measurements were performed as determined by individual obstetric providers based on clinical circumstances. Management decisions (eg, initiation of vaginal progesterone, inpatient admission) were made at the discretion of individual providers. Women were included if they had TVCL measurements at both 18 and 22 weeks and available delivery data. Women were excluded for the following reasons: an indicated preterm delivery, a diagnosis of fetal hydrops, twin-twin transfusion syndrome, a lethal anomaly, fetal demise, and monoamniotic or conjoined twin pregnancy.
Electronic ultrasound records from OBServer (version 7, Vision Chips, Inc., Laguna Hills, CA) were queried to identify women with twin pregnancies who had TVCL measurements during the study period.
TVCL measurements were performed by certified sonographers at 1 of several ultrasound units at our tertiary referral center for perinatal medicine. During the study period, 96% of sonographers were certified either through the Cervical Length Education and Review Program or through a research study protocol from the Maternal-Fetal Medicine Units Network of the Eunice Kennedy Shriver National Institute of Child Health and Human Development. Three TVCL measurements were captured at each visit, with and without fundal pressure. The shortest measurement on a high-quality image was recorded. When a cervical funnel was present, the remaining closed cervical length below the funnel was measured. The electronic medical record was queried for patient demographic, clinical, and outcome data.
The independent variable of interest is the rate of change in cervical length at 18 and 22 weeks’ gestation. Because serial measurements were performed at a spectrum of gestational ages for each woman, measurements taken between 17-18 weeks and 6 days were classified as the 18-week cervical length measurement. Measurements taken between 21-22 weeks and 6 days were classified as the 22-week cervical length measurement. The rate of change was calculated by dividing the difference in TVCL between 22-18 weeks by the number of days between TVCL measurements and multiplying by 7. The units of change are therefore reported in cm/wk. The primary outcome studied is SPTB, defined as <35 weeks of completed gestation at the time of delivery in the absence of a medical indication for delivery. We also investigated the secondary outcome of SPTB at <32 weeks.
We planned this study in a cohort of 527 subjects who met inclusion criteria. Our analysis plan included fitting a logistic regression model comparing the primary outcome to the independent variable, rate of change in TVCL. The incidence of SPTB was then examined by increasing levels of rate of change in TVCL to determine whether a threshold effect on the outcome was present. The effect of change in TVCL on SPTB was then tested as a binary variable: rapid change in TVCL was defined as a weekly decrease of at least 0.2 cm and compared with smaller weekly decreases in TVCL. With an incidence of SPTB of 16% among those 352 subjects without rapid change in TVCL, we calculated a minimum detectable odds ratio (OR) for SPTB of 1.88 under the assumptions of a 2-sided type I error of 5% and a power of 80%.
Demographic data were evaluated using descriptive statistics. We used χ 2 statistics for comparison of categorical variables and the Student t test for comparison of normally distributed continuous variables. Logistic regression was performed to evaluate the association between independent variables and the occurrence of SPTB. Statistical tests were 2-sided with a significance level of .05. All analyses were conducted in software (SAS, Version 9.4; SAS Institute, Cary, NC).
Materials and Methods
This is a retrospective cohort study of women with diamniotic twin pregnancies who underwent serial TVCL measurements at Columbia University Medical Center from 2008 through 2015. This clinical investigation was approved by the Columbia University Institutional Review Board (protocol IRB-AAAP3710).
Women in the present cohort underwent serial sonographic cervical length measurements between 16-28 weeks’ gestation at 2-week intervals as part of our institution’s routine antenatal surveillance protocol for women with multiple gestations. More frequent measurements were performed as determined by individual obstetric providers based on clinical circumstances. Management decisions (eg, initiation of vaginal progesterone, inpatient admission) were made at the discretion of individual providers. Women were included if they had TVCL measurements at both 18 and 22 weeks and available delivery data. Women were excluded for the following reasons: an indicated preterm delivery, a diagnosis of fetal hydrops, twin-twin transfusion syndrome, a lethal anomaly, fetal demise, and monoamniotic or conjoined twin pregnancy.
Electronic ultrasound records from OBServer (version 7, Vision Chips, Inc., Laguna Hills, CA) were queried to identify women with twin pregnancies who had TVCL measurements during the study period.
TVCL measurements were performed by certified sonographers at 1 of several ultrasound units at our tertiary referral center for perinatal medicine. During the study period, 96% of sonographers were certified either through the Cervical Length Education and Review Program or through a research study protocol from the Maternal-Fetal Medicine Units Network of the Eunice Kennedy Shriver National Institute of Child Health and Human Development. Three TVCL measurements were captured at each visit, with and without fundal pressure. The shortest measurement on a high-quality image was recorded. When a cervical funnel was present, the remaining closed cervical length below the funnel was measured. The electronic medical record was queried for patient demographic, clinical, and outcome data.
The independent variable of interest is the rate of change in cervical length at 18 and 22 weeks’ gestation. Because serial measurements were performed at a spectrum of gestational ages for each woman, measurements taken between 17-18 weeks and 6 days were classified as the 18-week cervical length measurement. Measurements taken between 21-22 weeks and 6 days were classified as the 22-week cervical length measurement. The rate of change was calculated by dividing the difference in TVCL between 22-18 weeks by the number of days between TVCL measurements and multiplying by 7. The units of change are therefore reported in cm/wk. The primary outcome studied is SPTB, defined as <35 weeks of completed gestation at the time of delivery in the absence of a medical indication for delivery. We also investigated the secondary outcome of SPTB at <32 weeks.
We planned this study in a cohort of 527 subjects who met inclusion criteria. Our analysis plan included fitting a logistic regression model comparing the primary outcome to the independent variable, rate of change in TVCL. The incidence of SPTB was then examined by increasing levels of rate of change in TVCL to determine whether a threshold effect on the outcome was present. The effect of change in TVCL on SPTB was then tested as a binary variable: rapid change in TVCL was defined as a weekly decrease of at least 0.2 cm and compared with smaller weekly decreases in TVCL. With an incidence of SPTB of 16% among those 352 subjects without rapid change in TVCL, we calculated a minimum detectable odds ratio (OR) for SPTB of 1.88 under the assumptions of a 2-sided type I error of 5% and a power of 80%.
Demographic data were evaluated using descriptive statistics. We used χ 2 statistics for comparison of categorical variables and the Student t test for comparison of normally distributed continuous variables. Logistic regression was performed to evaluate the association between independent variables and the occurrence of SPTB. Statistical tests were 2-sided with a significance level of .05. All analyses were conducted in software (SAS, Version 9.4; SAS Institute, Cary, NC).
Results
Of 1691 twin gestations that delivered at this institution, 1526 met inclusion criteria for this study. Reasons for exclusion included monoamnionicity (24 patients), pregnancies, lethal anomaly (12 patients), reduction to singleton gestation (68 patients), and twin-to-twin transfusion syndrome (61 patients). An additional 916 patients were missing TVCL measurements or outcome data, leaving 610 patients available for analysis. Eighteen patients had indicated deliveries <32 weeks, and 80 patients had indicated deliveries <35 weeks. Therefore, 595 and 527 patients were included in the cohorts to evaluate SPTB <32 and 35 weeks, respectively. The demographic characteristics of the cohort used to evaluate SPTB <35 weeks are listed in Table 1 . Approximately one third of subjects belonged to a racial or ethnic minority group. The majority of the cohort was nulliparous (66.8%), and 6.5% of patients had a history of preterm birth <37 weeks. In all, 14.4% of the cohort had monochorionic diamniotic pregnancies.
| N (%) | |
|---|---|
| Total | 527 (100%) |
| Race | |
| White | 297 (56.4%) |
| Asian | 59 (11.2%) |
| Black | 44 (8.4%) |
| Hispanic | 52 (9.9%) |
| Other/unknown | 75 (14.2%) |
| Age, y, mean (SD) | 35.1 (±6.3) |
| Nulliparous | 352 (66.8%) |
| Prior preterm birth | 34 (6.5%) |
| Chronic medical disease | 189 (35.9%) |
| Tobacco use | 2 (0.4%) |
| IVF | 299 (56.7%) |
| Monochorionic diamniotic | 76 (14.4%) |
| Interpregnancy interval ≤1 y | 48 (9.1%) |
| Prior cervical surgery | 36 (6.8%) |
| Cervical cerclage | 15 (2.9%) |
| Vaginal progesterone | 59 (11.2%) |
| 17-Hydroxyprogesterone caproate | 25 (4.8%) |
| TVCL 18 wk, cm, mean (SD) | 4.42 (±0.9) |
| TVCL 22 wk, cm, mean (SD) | 3.92 (±1.1) |
| Decrease in TVCL 18–22 wk, cm, mean (SD) | 0.50 (±1.0) |
| Rate of decrease in TVCL, cm/wk, mean (SD) | 0.13 (±0.2) |
| SPTB <35 wk | 112 (21.3%) |
TVCL was normally distributed at 18 and 22 weeks’ gestation. The mean TVCL at 18 weeks was 4.42 cm (SD 0.9), and the mean TVCL at 22 weeks was 3.92 cm (SD 1.1). Change in TVCL between visits was also normally distributed: the mean change was 0.50 cm (SD 1.0). Expressed in cm/wk of decrease, the mean was 0.13 cm/wk (SD 0.2).
In all, 112 patients (21.3%) delivered spontaneously <35 weeks’ gestation, whereas 53 (8.9%) patients delivered spontaneously <32 weeks. The percentage of patients who delivered <35 weeks was similar regardless of when TVCL was measured: the rates of preterm birth at each 2-week interval between 18-24 weeks were 21.3%, 22.1%, 23.2%, and 21.9%. Of those patients with SPTB <35 and 32 weeks, 19 (17%) and 11 (20.8%), respectively, had a history of SPTB. Of patients with a history of SPTB, the rate of recurrent SPTB <35 weeks was 55.9%. Fifteen patients (2.8%) received a cerclage, 60 (11.3%) used vaginal progesterone, and 25 (4.7%) received 17-hydroxyprogesterone injections. The average rate of change in TVCL for patients with SPTB <35 weeks was –0.21 cm/wk (SD 0.27) vs –0.10 cm/wk (SD 0.24) for patients who delivered ≥35 weeks ( P < .01). The average rate of change was –0.31 cm/wk (SD 0.27) for patients with SPTB <32 weeks vs –0.11 cm/wk (SD 0.24) for patients who delivered ≥32 weeks ( P < .01).
Examination of the initial 18-week TVCL revealed a 10th percentile of 3.34 cm. However, construction of a receiver operator curve (ROC) showed 4.25 cm to be the optimal TVCL for prediction of SPTB <35 weeks. Using the 10th percentile, 3.34 cm, to define a short TVCL, the sensitivity is 20.8% and the specificity is 93.0%, whereas using the cut-off based on the ROC, 4.25 cm, the sensitivity and specificity are 27.2% and 83.6%, respectively. The areas under the ROC for 18-week TVCL and change in TVCL between 18-24 weeks are similar (0.59 vs 0.63).
When controlling for initial TVCL at 18 weeks’ gestation: the odds of SPTB <35 weeks increased by 12.96 for every 1 cm/wk decrease in TVCL between ultrasounds (95% confidence interval [CI], 4.97–33.79) ( Table 2 ). The analogous OR for SPTB <32 weeks was 44.48 (95% CI, 13.58–145.63). Considering only patients with a short cervix as defined by the ROC (TVCL <4.25 cm), the OR for change in cervical length were similar to that of the whole cohort, however the CI are larger and overlap considerably with those of the whole cohort ( Table 2 ). These associations remained after controlling for other important risk factors for SPTB. The adjusted OR for SPTB <35 and 32 weeks were 12.69 (95% CI, 4.66–34.58) and 57.88 (95% CI, 16.11–207.94), respectively ( Table 3 ). The only other significant variable was history of SPTB, which increased the odds of SPTB <35 and 32 weeks ( Table 3 ). When the model was restricted to patients with a history of SPTB, neither TVCL at 18 weeks nor rate of change of TVCL was associated with SPTB <35 or 32 weeks; however only 34 patients had such a history. The relationship between change in TVCL and the risk for SPTB is shown in Figure 1 . For every 1 cm/wk decrease in TVCL between measurements, the odds of SPTB increase by 5.75 (95% CI, 2.45–13.48).
