Preterm birth is the main cause of perinatal morbidity and mortality. A short cervical length on transvaginal ultrasonography has been shown to be one of the best predictors of spontaneous preterm birth.1 Screening for a short cervical length has been studied in several populations, including asymptomatic women with singleton gestations at either low2 or high risk3 for preterm birth, multiple gestations,4 and symptomatic women with either preterm labor5 or preterm premature rupture of the membranes.6 Cervical ultrasonography has yet no appreciable clinical usefulness in low-risk populations because of its poor positive predictive value and the absence of preventive therapy.1,2 Studies of asymptomatic singletons with risk factors for spontaneous preterm birth have been published on women with a prior preterm birth,3 prior cone biopsy,7 uterine anomalies,8 or prior dilation and evacuation (D&E) (Visintine J, Berghella V, Henning D, Baxter J. Cervical length for prediction of preterm birth in women with multiple prior induced abortions. Ultrasound Obstet Gynecol 2007. In press). In women with both historical risk factors and a short cervical length, interventions (eg, cerclage) may be beneficial in preventing spontaneous preterm birth.9 Earlier evaluations of transvaginal ultrasound cervical length at less than 23 weeks are more predictive than those performed later,1 but so far there have been no data published on the risk of preterm birth by cervical length stratified by gestational age. This information is important not only for patient counseling regarding risk, but also for management decisions. Our objective was to estimate the risk of spontaneous preterm birth before 35, 32, and 28 weeks in high-risk women with a singleton gestation based on cervical length and the gestational age at which the cervical length is measured.
MATERIALS AND METHODS
Women with a singleton gestation and historical risk factors for spontaneous preterm birth such as prior spontaneous preterm birth at 14 0/7–34 6/7 weeks, prior excisional cervical biopsy (either cold knife, loop electrosurgical excision procedure, or laser conization), müllerian anomaly, or 2 or more D&Es secondary to spontaneous or induced abortion, were examined with transvaginal ultrasound cervical length serially between approximately 12 0/7 and 31 6/7 weeks. Gestational age was determined by menstrual history and confirmed for all women by ultrasonography before 20 weeks. Before 13 6/7 weeks, if last menstrual period and ultrasound-based dating differed by 7 days or more, preference was given to the ultrasound-based date. Between 14 and 19 6/7 weeks, if last menstrual period and ultrasound-based dating differed by 10 days or more, preference was given to the ultrasound-based date. Exclusion criteria were multiple gestations, cerclage (history-indicated, ultrasound-indicated, or physical examination-indicated), iatrogenic preterm birth, or major fetal anomaly.
All women were examined between July 1995 and June 2005 at Thomas Jefferson University Hospital, in Philadelphia, Pennsylvania. In 1995, the institutional review board of Thomas Jefferson University approved prospective general data collection for all women at high risk for preterm birth. In 2005, the institutional review board separately approved the retrospective protocol for evaluating the association between cervical length and preterm birth, using the database accumulated over time. During the study period, there were no local policies regarding management, except for the fact that women with a short cervical length (less than 25 mm between 14 0/7 and 23 6/7 weeks) were offered randomization for cerclage.3
All cervical length measurements were performed by experienced ultrasonographers using standard technique.1,2 Women had an empty bladder and were in a dorsal lithotomy position. After visualization of the cervix, only the minimum pressure necessary was used to obtain a clear image of the cervical canal. The shortest best length of the closed cervix (cervical length) was recorded and used for analysis. Short cervical length was defined as less than 25 mm.1,2 Cervical length results were not blinded to managing obstetricians. Analysis was based on the shortest best cervical length obtained as well as each measurement along the pregnancy.
The primary outcome was incidence of spontaneous preterm birth less than 35 weeks. Spontaneous preterm birth less than 32 weeks and less than 28 weeks were also analyzed. In preliminary analyses, we used the nonparametric Lowess smoother10 to investigate the shape of the association between cervical length and preterm delivery, separately for each time of measurement (week of pregnancy). For the main analyses, we used logistic regression to model the probability of preterm delivery as a function of cervical length and the time of measurement. Because each woman's multiple measurements over time are correlated, the usual model-based standard errors are invalid. Therefore, we used the generalized estimating equations approach with the robust (“sandwich”) variance estimator to compute odds ratios, confidence intervals, and P values.11 In addition to odds ratios, confidence intervals, and P values, we also computed the estimated spontaneous preterm birth risk as a function of the cervical lengths and the gestational age (week of pregnancy). Statistical analyses were conducted in Stata 8 (StataCorp, College Station, TX).
Seven hundred five singleton gestations with risk factors for spontaneous preterm birth and transvaginal ultrasound cervical length data between 12 0/7 to 31 6/7 weeks of gestation were identified and analyzed in the study period. Population characteristics and risk factors are shown in Table 1. Some women had more than one risk factor. A total of 2,601 cervical length measurements were obtained (range 1 to 11 measurements per pregnancy, mean 3.7, standard deviation 2.3). The median times for the first and last cervical length measurement were the 17th week and the 25th week, respectively, and the average interval between cervical length measurements ranged between 1 and 7 weeks (mean 2.1, standard deviation 1.0). Figure 1 is a graphic summary of the cervical length measurements according to the time they were obtained (12 to 32 weeks of pregnancy). This number of cervical length measurements ranged between 104 and 203 for weeks 14 through 28, although it was lower for the earlier and later weeks (60 in week 13 to 11 in week 32). Cervical length before 14 weeks was very rarely below 25 mm. Cervical length remained stable between approximately 14 weeks and 28 weeks, with average cervical length of 36.7 mm at 15 weeks, 35.7 mm at 20 weeks, and 33.8 mm at 25 weeks. After 28 weeks, even women who deliver at term began to have physiologic cervical shortening. The overall incidences (95% confidence interval; CI) of spontaneous preterm birth were 40.1% (36.5–43.8%) at less than 37, 17.7% (14.9–20.6%) at less than 35, 10.6% (8.4–12.9%) at less than 32, and 6.7% (4.8–8.5%) at less than 28 weeks.
In analyses based on the Lowess smoother, the curve describing the association between the risk of preterm delivery and cervical length measurements had similar shape across the various time points between weeks 15 and 28 and was quite consistent. However, cervical length measurements obtained either before week 15 or after week 28 showed variable patterns. Therefore, the main analyses of preterm birth before week 35 and before week 32 were based only on the 2,257 observations obtained between the 15th and 28th week of pregnancy. For similar reasons, the analyses of preterm delivery before week 28 were restricted to the 1,881 observations obtained between the 15th and 25th week of pregnancy.
For the outcome of preterm birth before week 35, the main logistic regression analyses showed that the shape of the association between cervical length and preterm birth was similar across pregnancy weeks 15 to 28 (P=.50 for the interaction between cervical length and time of measurement). Therefore, the final model included only main effects for cervical length and time of measurement, but not their interaction. These two variables were both significant predictors of preterm birth. The risk of preterm birth decreased by approximately 6% for each additional mm of cervical length (odds ratio [OR] 0.94, 95% CI 0.92–0.95, P=.001) and by approximately 5% for each additional week of pregnancy (OR 0.95, 95% CI 0.92–0.98, P=.004). Figure 2 is a plot of the estimated risk of spontaneous preterm birth less than 35 weeks, by cervical length measured at 16, 20, 24, and 28 weeks of pregnancy. Table 2 is a more detailed table of the model-based estimate of the spontaneous preterm birth risk, by cervical length and time of measurement (week of pregnancy).
For the outcome of preterm birth before week 32, the model for delivery before week 32 again included only main effects for cervical length (OR 0.92, 95% CI 0.90–0.94, P=.001) and week of pregnancy (OR 0.87, 95% CI 0.82–0.91, P=.001), because the interaction of the two variables was not significant (P=.98). Figure 3 and Table 3 present the model-based estimate of the risk of spontaneous preterm birth less than 32 weeks, by cervical length and week of pregnancy.
For the outcome of preterm birth before week 28, the model for delivery before week 28 again included only main effects for cervical length (OR 0.90, 95% CI 0.88–0.93, P=.001) and week of pregnancy (OR 0.81, 95% CI 0.74–0.88, P=.001), because the interaction of the two variables was not significant (P=.46). Figure 4 and Table 4 present the model-based estimate of the risk of spontaneous preterm birth less than 28 weeks, by cervical length and week of pregnancy.
Our study reports detailed results regarding the correlation between both transvaginal ultrasound cervical length and the gestational age at which the transvaginal ultrasound cervical length was obtained with preterm birth outcomes at less than 35, less than 32, and less than 28 weeks. Such data should be of major use for the obstetric clinician.
Transvaginal ultrasound cervical length for prediction of preterm birth is often difficult for the clinician to interpret unless it is assessed in the clinical scenario of the individual patient. The other important variables predictive of preterm birth that can be correlated with the transvaginal ultrasound cervical length measurement without doing another test are asymptomatic compared with symptomatic patient, singleton compared with multiple gestation, prior obstetric or gynecologic history, and gestational age at which the transvaginal ultrasound cervical length is obtained.1 These variables significantly affect the predictive values of transvaginal ultrasound cervical length. For example, it has been previously reported that a transvaginal ultrasound cervical length in the mid trimester (at 22–24 weeks) of less than 25 mm is associated with a 18% risk of preterm birth less than 35 weeks in low-risk singleton gestations,2 and a greater than 50% risk of preterm birth less than 35 weeks in a woman with a prior preterm birth.3 It has also been previously reported that a transvaginal ultrasound cervical length less than 25 mm at 14–24 weeks has a positive predictive value of 40% for preterm birth less than 35 weeks if detected at 20 weeks, but a 70% positive predictive value for preterm birth less than 35 weeks if detected at 16 weeks.12
These observations are not very useful to the clinician unless most of the important variables are considered together for an overall assessment of risk. This concept is similar to first trimester screening. Pregnancy-associated plasma protein A, beta human chorionic gonadotrophin, and nuchal translucency are all individually associated with an increased risk of Trisomy 21, but it is only when they are combined that we get a clinically useful assessment of risk that allows for meaningful individualized counseling for each woman. These measures, in common with cervical length, also change with gestation, and this change must be taken into consideration to allow robust prediction. Similarly, cervical length changes with gestation, so that the gestation at screening is crucial, as is the absolute cervical length.
We decided to study asymptomatic woman with prior risk factors for preterm birth (more than 60% of them with a prior preterm birth) carrying a singleton gestation. This is because this is the only patient population in which there is information that an intervention, based on the additional risk of a short cervical length, may prevent the predicted outcome, ie, preterm birth. In a meta-analysis of the trials published so far, cerclage based on a short cervical length before 24 weeks significantly reduced preterm birth less than 35 weeks by 39% in women with a prior preterm birth.9
The significance of gestational age at which cervical length is observed cannot be overstated. For example, in our population, a transvaginal ultrasound cervical length of 20 mm is associated with risks of 39% compared with 24% for spontaneous preterm birth less than 35 weeks if detected at 15 weeks or 28 weeks, respectively. Perhaps even more impressively, a transvaginal ultrasound cervical length of 15 mm is associated with risks of 48% compared with 13% for spontaneous preterm birth less than 32 weeks if detected at 15 or 28 weeks, respectively. These charts should be used as practical guides for management. Additionally, they can be used for further research. For example, infection is associated with earlier spontaneous preterm birth, and could be assessed in relation to these data. In addition, other risk factors that we did not study, such as fetal fibronectin and inflammation markers, may further improve the assessment of the risk of delivering preterm.
Transvaginal ultrasound cervical length has been shown to be predictive of preterm birth in all populations studied, including asymptomatic women with singleton gestations at low risk for preterm birth,1 asymptomatic women with singleton gestations at high risk for preterm birth (because of a prior preterm birth,3 multiple D&Es (Visintine J, Berghella V, Henning D, Baxter J. Cervical length for prediction of preterm birth in women with multiple prior induced abortions. Ultrasound Obstet Gynecol 2007. In press), müllerian anomalies,8 cone biopsy7, asymptomatic multiple gestations (eg, twins,4 triplets13), and symptomatic women with either preterm labor5 or preterm premature rupture of the membranes.6 In the future, tables and figures that correlate transvaginal ultrasound cervical length at each gestational age should be created for these specific populations as well.
The main limitation of our study is that it was not blinded. The only proven intervention in asymptomatic patients based on transvaginal ultrasound cervical length that may affect outcome is cerclage.9 We did exclude all types of cerclages for this reason in the methods. No other intervention (eg, bed rest, hospitalization, monitoring, tocolysis, etc) has been shown to affect the different outcomes of preterm birth reported in our study. None of the patients received progesterone for a short cervical length, because there had been no published study yet on its effectiveness. Knowledge of the cervical length may have led clinicians to perform other interventions that could have influenced the chance of preterm birth. The fact that no other interventions have been proven to prevent preterm delivery does not mean that an unproven effect does not exist.
It is not surprising that our analysis revealed that measurements before 15 weeks and after 28 weeks were not as consistent as measurements obtained between 15 and 28 weeks. Before 15 weeks, cervical length is usually normal (eg, 25 mm or more), and in fact at times artificially long because the lower uterine segment cannot be distinguished from the cervical canal itself. Only 5% of high risk women, those with either multiple prior second trimester losses or large cold knife cones, have a short cervical length this early in pregnancy.14 On the other end of the gestational age continuum, it is well known that transvaginal ultrasound cervical length begins to shorten physiologically after 28 weeks even in women destined to deliver at term, and therefore cervical shortening in the third trimester usually represents the beginning of the normal ripening process that eventually manifests itself with term labor.15 We also cannot rule out the possibility of substantial selection bias in the measurement of cervical length very early or very late in the pregnancy.
Cervical assessment with ultrasonography has been widely investigated as a predictor of spontaneous preterm birth in asymptomatic and symptomatic women with low or high risk, but none of them have correlated the transvaginal ultrasound cervical length week by week with the incidence of spontaneous preterm birth. Our study provides a more formalized assessment of the risk of preterm birth, based not only on number of fetuses (singleton), historical high risk factors, asymptomatic status, and cervical length, but also on the gestational age at which the cervical length was obtained. Although our findings do not apply to patients who would have been excluded from this study (such as women with multiple gestations, presence of symptoms, or different historical risk factors), they confirm and render clinically easy to use the fact that the incidence of spontaneous preterm birth is inversely associated with cervical length. Our results can be used to quantify the predicted risk of spontaneous preterm birth as a function of the cervical length and of the gestational age at measurement. These risks of spontaneous preterm birth are important for counseling and management for women with different degrees of short cervical length at different gestational ages.
1. Berghella V, Bega G, Tolosa JE, Berghella M. Ultrasound assessment of the cervix. Clin Obstet Gynecol 2003;46:947–62.
2. Iams JD, Goldenberg RL, Meis PJ, Mercer BM, Moawad A, Das A, et al. The length of the cervix and the risk of spontaneous premature delivery. National Institute of Child Health and Human Development Maternal Fetal Medicine Unit Network. N Engl J Med 1996;334:567–72.
3. Owen J, Yost N, Berghella V, Thom E, Swain M, Dildy GA 3rd, et al. Mid-trimester endovaginal sonography in women at high risk for spontaneous preterm birth. JAMA 2001;286:1340–8.
4. Goldenberg RL, Iams J, Miodovnik M, Van Dorsten JP, Thurnau G, Bottoms S, et al. The preterm prediction study: risk factors in twin gestation. National Institute of Child Health and Human Development Maternal-Fetal Medicine Units Network. Am J Obstet Gynecol 1996;175:1047–53.
5. Vendittelli F, Mamelle N, Munoz F, Janky E. Transvaginal ultrasonography of the uterine cervix in hospitalized women with preterm labor. Internat J Gynecol Obstet 2001;72:117–25.
6. Carlan SJ, Richmond LB, O'Brien WF. Randomized trial of endovaginal ultrasound in preterm premature rupture of membranes. Obstet Gynecol 1997;89:458–61.
7. Berghella V, Pereira L, Gariepy A, Simonazzi G. Prior cone biopsy: Prediction of preterm birth by cervical ultrasound. Am J Obstet Gynecol 2004;191:1393–7.
8. Airoldi J, Berghella V, Sehdev H, Ludmir J. Transvaginal ultrasonography of the cervix to predict preterm birth in women with uterine anomalies. Obstet Gynecol 2005;106:553–6.
9. Berghella V, Odibo AO, To MS, Rust OA, Althuisius SM. Cerclage for short cervix on ultrasound: meta-analysis of trials using individual patient-level data. Obstet Gynecol 2005;106:181–9.
10. Cleveland WS. The elements of graphing data. Summit (NJ): Hobart Press; 1994.
11. Liang KY, Zeger SL. Longitudinal data analysis using generalized linear models. Biometrika 1986;73:13–22.
12. Berghella V, Daly SF, Tolosa JE, DiVito MM, Chalmers R, Garg N, et al. Prediction of preterm delivery with transvaginal ultrasonography of the cervix in patients with high-risk pregnancies: does cerclage prevent prematurity? Am J Obstet Gynecol 1999;181:809–15.
13. Guzman ER, Walters C, O'Reilly-Green C, Meirowitz NB, Gipson K, Nigam J, et al. Use of cervical ultrasonography in prediction of spontaneous preterm birth in triplet gestations. Am J Obstet Gynecol 2000;183:1108–13.
14. Berghella V, Talucci M, Desai A. Does transvaginal sonographic measurement of cervical length before 14 weeks predict preterm delivery in high-risk pregnancies? Ultrasound Obstet Gynecol 2003;21:140–4.
15. Smith CV, Anderson JC, Matamoros A, Rayburn WF. Transvaginal sonography of cervical width and length during pregnancy. J Ultrasound Med 1992;11:465–7.