Gestational age at cervical length measurement and preterm birth in twins

Authors

  • R. M. Ehsanipoor,

    Corresponding author
    1. Division of Maternal–Fetal Medicine, Department of Gynecology and Obstetrics, Johns Hopkins University, Baltimore, Maryland, USA
    2. Division of Maternal–Fetal Medicine, Department of Obstetrics and Gynecology, University of California, Irvine, Orange, California, USA
    • Division of Maternal-Fetal Medicine, Department of Gynecology and Obstetrics, 600 North Wolfe Street, Phipps 228, Baltimore, MD, USA 21287

    Search for more papers by this author
  • M. L. Haydon,

    1. Department of Obstetrics and Gynecology, Hoag Memorial Hospital Presbyterian, Newport Beach, California, USA
    Search for more papers by this author
  • C. Lyons Gaffaney,

    1. Obstetrix Medical Group, San Jose, Campbell, California, USA
    Search for more papers by this author
  • J. A. Jolley,

    1. Division of Maternal–Fetal Medicine, Department of Obstetrics and Gynecology, University of California, Irvine, Orange, California, USA
    Search for more papers by this author
  • R. Petersen,

    1. Center for Statistical Consulting, University of California, Irvine, California, USA
    Search for more papers by this author
  • D. C. Lagrew,

    1. Department of Obstetrics and Gynecology, Saddleback Memorial Medical Center, Laguna Hills, California, USA
    Search for more papers by this author
  • D. A. Wing

    1. Division of Maternal–Fetal Medicine, Department of Obstetrics and Gynecology, University of California, Irvine, Orange, California, USA
    Search for more papers by this author

Abstract

Objectives

To estimate the risk of preterm delivery of twin pregnancies based upon sonographic cervical length measurement and gestational age at measurement.

Methods

Twin pregnancies that delivered between 1999 and 2005 and that underwent sonographic measurement of cervical length between 13 and 34 + 6 weeks' gestation were identified and a retrospective review performed. Women with anomalous pregnancies, multifetal reduction, cerclage placement or medically indicated deliveries before 35 weeks were excluded. Logistic regression analysis was used to estimate the risk of preterm delivery before 35 weeks.

Results

A total of 561 women underwent 2975 sonographic cervical length measurements during the study period. The rate of preterm delivery before 35 weeks was 19.4%. The risk of delivery before 35 weeks decreased by approximately 5% for each additional mm of cervical length (odds ratio (OR) 0.95 (95% CI, 0.93–0.97); P < 0.001) and by approximately 6% for each additional week at which the cervical length was measured (OR 0.94 (95% CI, 0.92–0.96); P < 0.001).

Conclusion

The gestational age at which cervical length is measured is an important consideration when estimating the risk of spontaneous preterm birth in twins. The risk of preterm delivery is increased at earlier gestational ages and as cervical length decreases. Copyright © 2012 ISUOG. Published by John Wiley & Sons, Ltd.

Introduction

Preterm delivery is the leading cause of neonatal mortality in the USA and a leading cause of neonatal morbidity and neurologic deficits. Despite extensive research, the rate of preterm delivery continues to rise in the USA, and in 2007 12.7% of live births were preterm. The rise in multiple gestations in the past two decades has in part contributed to the rise in preterm delivery. In 2007, 3.2% of live births in the USA were twins, and these infants had a three-fold increased risk of death compared to singletons1. Of additional concern is the fact that the rate of preterm delivery among twins has increased by 22% from 1981/1982 to 19972.

Sonographic measurement of the cervix in the second trimester is useful in predicting preterm delivery3. When risk factors for preterm delivery were studied in twins, a sonographic cervical length of less than 25 mm at 24 weeks' gestation was noted to be the strongest predictor of preterm delivery4. The correlation between shortened cervix in the late second or early third trimester and preterm delivery in twin gestations has been corroborated by multiple studies, with cut-offs ranging from 20 to 35 mm4–8. However, the literature evaluating cervical length as a function of gestational age is limited. Berghella et al.9 showed that, in singleton pregnancies, the gestational age at measurement is an important factor in estimating the risk of preterm delivery.

The objective of this study was to estimate the risk of preterm delivery prior to 35 weeks' gestation in twin pregnancies based on cervical length and gestational age at measurement.

Methods

After obtaining institutional review board approval, we reviewed charts of twin gestations delivered between January 1999 and December 2005 at two community hospitals in California. This time frame was chosen because at the time routine evaluation of cervical length was performed at these institutions. Subjects included had undergone transvaginal sonographic cervical length surveillance at least once between 13 and 34 + 6 weeks' gestation. Data were obtained and recorded by reviewing maternal prenatal records, inpatient hospital charts and ultrasound reports. Only measurements obtained in asymptomatic women were included. Pregnancies with major fetal anomalies or intrauterine death, multifetal pregnancy reduction, medically indicated delivery before 35 weeks, twin–twin transfusion syndrome and cerclage placement were excluded. Gestational age was estimated from the date of the last menstrual period unless revision of the estimated date of confinement was indicated by ultrasonography.

Experienced sonographers using a standard technique measured cervical length with transvaginal sonography (as is the protocol in our units)3. With the maternal bladder empty, the transducer was introduced into the anterior fornix of the vagina and a sagittal view of the cervix obtained. At least three measurements were made from the internal os to the external os. Following this either fundal pressure was applied or the patient was asked to perform a Valsalva maneuver to check for the presence of funneling and another measurement was taken. The shortest measurement recorded was used as the cervical length. If more than one measurement was obtained in a week, the shortest measurement was recorded and used in the analysis. The practitioners were not blinded to the cervical length measurements and there was no standard management protocol in place for cases of cervical shortening. However, physicians would often suggest closer follow-up and modifications in activity for patients with cervical shortening. Inpatient management with tocolysis was routine if a patient exhibited evidence of uterine activity. Neither vaginal nor intramuscular progesterone was routinely administered, even with the finding of a shortened cervix.

The primary outcome was spontaneous preterm birth before 35 weeks' gestation, with cervical length—as measured by transvaginal sonography—and week of measurement as predictor variables. Logistic regression analysis was used to estimate the risk of preterm delivery before 35 weeks as a function of cervical length and the gestational age at which the measurement was obtained. Additionally, we calculated adjusted odds ratios (ORs) and CIs for each of these two variables, using the other as a control variable. Because the data are longitudinal in nature, correlations between observations on a given subject may exist, and needed to be accounted for. Therefore the method of generalized estimating equations (GEE) was used to model the data. The GEE model is an extension of the generalized linear model method to correlated data, so that valid standard errors of the parameter estimates can be drawn by the use of an empirical (sandwich/robust) variance estimator10. The GEE model was also used to calculate the probability of spontaneous preterm birth before 32 weeks' gestation. For this analysis only indicated deliveries prior to 32 weeks were excluded as opposed to indicated preterm deliveries prior to 35 weeks. An adjusted analysis was performed to control for the potential confounders of parity and conception with assisted reproductive technology. Analysis was done with SAS 9.2 and JMP 7.0.2 (SAS Institute, Cary, NC, USA).

Results

A total of 561 women meeting the inclusion criteria underwent 2975 transvaginal sonographic evaluations of the cervix during the study period. A total of 188 subjects were excluded from the study (58 because of cerclage, 59 because of multifetal reductions, 18 because of fetal anomalies or death, 22 because of twin–twin transfusion syndrome and 31 because of medically indicated delivery prior to 35 weeks). The population was composed primarily of Caucasians, and more than 50% of the women were nulliparous and more than 50% conceived with infertility treatment. The characteristics of the study population are given in Table 1.

Table 1. Characteristics of the study population
CharacteristicValue
  1. Data are given as n (%) except where indicated.

Number of pregnancies561
Maternal age (years, mean ± SD)34.6 ± 5.1
Ethnicity 
 Caucasian460 (82.0)
 Asian52 (9.3)
 Hispanic38 (6.8)
 African–American3 (0.5)
 Other8 (1.4)
Nulliparous314 (56.0)
Conceived after infertility treatment308 (54.9)
Prior procedure for cervical dysplasia80 (14.3)
Tobacco use during pregnancy6 (1.1)
Prior preterm birth (< 37 weeks)16 (2.9)
Dichorionic pregnancy473 (84.3)

The mean number of measurements per pregnancy was 5.3 ± 3.2 (range, 1–19). Of the 494 (88.1%) subjects that had more than one measurement, the median times for the first and last cervical length measurements were the 19th and 31st weeks, respectively, and the mean interval between cervical length measurements was 2.4 ± 1.6 (range, 1–18) weeks. Figure 1 shows cervical length measurements obtained according to the gestational age at which they were measured. The number of cervical length measurements obtained at each given gestational age between 16 and 31 weeks ranged from 90 to 213; the number of measurements obtained at the extreme earlier and later gestational ages were significantly less (26 at 13 weeks and 37 at 34 weeks).

Figure 1.

Box-and-whisker plots of cervical length against gestational age in twin pregnancies (n = 2975), showing median values and interquartile range (75th and 25th percentiles, boxes). Whiskers represent values within 1.5 × interquartile range and square dots represent outliers.

Cervical length generally remained stable until approximately 20–22 weeks' gestation and then steadily decreased. The mean cervical length was 3.96 cm at 15 weeks and 4.14 cm at 20 weeks, and it had decreased to 3.50 cm at 25 weeks and to 3.16 cm at 30 weeks. The overall rate of preterm delivery was 19.4% prior to 35 weeks, 3.9% prior to 32 weeks and 0.9% prior to 28 weeks.

Weekly data were first qualitatively assessed with spline smoothing curves to evaluate the association between cervical length measurement and preterm delivery rates. Similar shapes were found between weeks 16 and 31; however, the measurements obtained prior to week 16 and after week 31 did not generate consistent patterns. Therefore, the analyses for preterm birth prior to 35 weeks were based on the 2586 measurements obtained from weeks 16 to 31.

The final model included only the main effects for cervical length and gestational age at measurement. The interaction term was not included (P = 0.92 for the interaction between cervical length and gestational age at measurement). Hospital site was also initially included in the model to determine any possible effect but it was not found to be significant (P = 0.49). Both cervical length and gestational age at measurement were significant predictors of preterm birth before 35 weeks in the GEE regression analysis. The risk of preterm birth decreased by approximately 5% for each additional mm of cervical length (odds ratio (OR) 0.95 (95% CI, 0.93–0.97); P < 0.001) and by approximately 6% for each additional week of pregnancy (OR 0.94 (95% CI, 0.92–0.96); P < 0.001). The adjusted ORs and CIs controlling for parity and conception with assisted reproductive technology revealed the same values. Figure 2 is a plot of the estimated probability of spontaneous preterm birth before 35 weeks' gestation, by cervical length measured at 16, 20, 24 and 28 weeks and Table 2 gives detailed estimates of the risk of spontaneous preterm birth prior to 35 weeks, by cervical length and gestational age at measurement.

Figure 2.

Predicted probability of delivery of twin pregnancies before 35 weeks' gestation based on cervical length and gestational age (GA) at time of measurement. equation image, GA 16 weeks; equation image, GA 20 weeks; equation image, GA 24 weeks; equation image, GA 28 weeks.

Table 2. Predicted probability of preterm delivery before 35 weeks' gestation according to cervical length and time of measurement
 Predicted probability (%) at gestational week:
Cervical length161718192021222324252627282930
0 mm77.476.375.274.072.871.570.268.967.666.264.863.461.960.458.9
5 mm72.271.069.768.367.065.664.262.761.259.858.356.755.253.752.1
10 mm66.364.963.562.160.659.157.656.154.553.051.449.948.346.845.2
15 mm59.958.456.955.453.852.350.749.247.646.144.543.041.540.038.5
20 mm53.151.650.048.546.945.443.842.340.839.337.836.435.033.632.2
25 mm46.244.743.241.640.138.737.235.834.332.931.630.329.027.726.5
30 mm39.538.036.535.133.732.331.029.728.427.225.924.823.622.521.5
35 mm33.131.730.429.127.826.625.424.223.122.021.020.019.018.117.2
40 mm27.326.124.923.722.621.620.519.518.617.716.815.915.114.313.6
45 mm22.221.120.119.118.217.316.415.514.814.013.312.611.911.310.6
50 mm17.816.916.015.214.413.712.912.311.611.010.49.89.38.88.3
55 mm14.113.312.612.011.310.710.19.69.18.68.17.67.26.86.4

A secondary analysis was performed to estimate the risk of preterm birth before 32 weeks' gestation. Nineteen women who underwent 200 measurements had an indicated delivery between 32 and 35 weeks' gestation and were excluded from the primary analysis of those with delivery before 35 weeks. However, they were included in the model to estimate the probability of spontaneous preterm birth prior to 32 weeks, which thus included 580 women with 3175 measurements. Again, both cervical length and time of measurement were significant predictors of preterm birth before 32 weeks. The risk of preterm birth decreased by approximately 8% for each additional mm of cervical length (OR 0.92 (95% CI, 0.89–0.95); P < 0.001) and by approximately 14% for each additional week of pregnancy (OR 0.86 (95%, CI 0.82–0.90); P < 0.001). The adjusted ORs and CIs after controlling for parity and conception with assisted reproductive technology revealed the same values. Figure 3 is a plot of the estimated probability of spontaneous preterm birth before 32 weeks' gestation, by cervical length measured at 16, 20, 24 and 28 weeks, and Table 3 provides detailed estimates of the risk of spontaneous preterm birth prior to 32 weeks, by cervical length and gestational age at measurement.

Figure 3.

Predicted probability of delivery of twin pregnancies before 32 weeks' gestation based on cervical length and gestational age (GA) at time of measurement. equation image, GA 16 weeks; equation image, GA 20 weeks; equation image, GA 24 weeks; equation image, GA 28 weeks.

Table 3. Predicted probability of preterm delivery before 32 weeks according to cervical length and time of measurement
 Predicted probability (%) at gestational week:
Cervical length161718192021222324252627282930
0 mm68.064.661.057.353.649.846.042.338.635.131.728.525.522.820.2
5 mm58.154.350.646.843.039.335.832.429.126.123.320.718.316.114.2
10 mm47.543.840.136.533.029.826.723.821.218.816.614.612.811.29.8
15 mm37.233.730.427.324.421.719.217.014.913.111.510.08.77.66.6
20 mm27.924.922.219.717.415.313.511.810.39.07.86.85.95.14.4
25 mm20.217.815.713.812.110.69.28.07.06.15.24.53.93.42.9
30 mm14.212.410.99.58.37.26.25.44.74.03.53.02.62.21.9
35 mm9.78.57.46.45.64.84.23.63.12.72.32.01.71.51.3
40 mm6.65.74.94.33.73.22.82.42.01.81.51.31.11.00.8
45 mm4.43.83.32.82.42.11.81.61.31.21.00.90.70.60.5
50 mm2.92.52.21.91.61.41.21.00.90.80.70.60.50.40.4
55 mm1.91.71.41.21.10.90.80.70.60.50.40.40.30.30.2

Discussion

Our study provides detailed estimates of the risk of preterm delivery before 35 weeks' gestation for twins based on cervical length measured at various gestational ages, which could be useful to the obstetrician in counseling women regarding such risk. The large sample size and number of cervical length measurements obtained across a variety of gestational ages are strengths of the study, which enabled us to evaluate different cervical lengths across gestational age with a robust sample size.

The population studied is unique and this can be considered both a strength and limitation of the study. The subjects consisted primarily of Caucasian subjects with a mean age of 34.6 years. Additionally, the majority were nulliparous and conceived with infertility treatment. This may in some ways limit generalization of the data; however, this is a growing patient demographic, as more women are delaying childbearing and an increasing number of multiple pregnancies are conceived with assisted reproductive technologies11. The exclusion of subjects who underwent cerclage placement or fetal reduction, in addition to the low number of patients with a prior spontaneous preterm birth, make this a low-risk population, thus the results of these data are most appropriate for the low-risk patient with twins. This is probably the explanation for the relatively low rate of delivery before 32 weeks' gestation (3.9%).

The retrospective nature of this study introduces certain limitations. Cervical lengths were not measured as part of a formal study protocol and there was a potential for variable clinical practices, as physicians were not blinded to the results. There was no standard intervention for patients with cervical shortening. Another limitation is that measurements were obtained in a variable fashion and this could introduce bias into the study. The GEE modeling method assumes some form of randomness in the pattern of missing values, which is unlikely in a retrospective study such as this. Specifically, we did not have a large number of measurements in the early second trimester, so these estimates may be less reliable.

Numerous studies have concluded that cervical length is predictive of preterm delivery in twins and this was succinctly highlighted in a recent meta-analysis that concluded that a cervical length measurement made between 20 and 24 weeks' gestation is a good predictor of spontaneous preterm birth8. The majority of studies included primarily evaluated cervical length in the latter half of the second trimester; however, we were able to evaluate cervical length across a broad range of gestational ages in substantial numbers. Our results suggest that cervical length data obtained between 16 and 31 weeks' gestation correlate with the risk of preterm delivery. This is similar to the finding of Berghella et al.9, that in singleton pregnancies cervical length data obtained between 15 and 28 weeks correlate with risk of preterm delivery.

The role of routine cervical length assessment in twins is unclear. Gyamfi et al.12 evaluated outcomes in twins with routine cervical length measurements vs. those without and did not demonstrate an improvement in outcome. Arguably, the best use for cervical length assessment is to identify those at highest risk for preterm delivery, thus permitting appropriately designed trials to evaluate potential interventions. Such trials have been conducted in singleton pregnancies supporting the use of progesterone13, 14 and cerclage15.

The optimal management of twin pregnancies with cervical shortening in the second trimester remains unclear and controversial. To date, progesterone has not been shown to be beneficial in unselected twin pregnancies16–19. Data on twins with cervical shortening are limited to secondary analyses of randomized controlled trials20, 21. These studies failed to show a benefit of progesterone but lack sufficient power. Cerclage appears to be beneficial in singleton pregnancies with cervical shortening, but limited data in twins suggest potential harm. A meta-analysis of trials evaluating cerclage placement for a short cervix included only 49 twin pregnancies and found that those undergoing cerclage placement were more likely to deliver before 35 weeks' gestation (relative risk 2.15 (95% CI, 1.15–4.01))22. A Cochrane review evaluating inpatient management with bed rest in twins did not demonstrate any difference with regard to the risk of preterm delivery23. Only one study evaluated subjects with evidence of cervical change and it also did not demonstrate a significant difference in outcomes, including preterm delivery24. No trials evaluating activity restrictions or prophylactic tocolysis in twin pregnancies with a shortened cervix were identified.

Further research evaluating the use of sonographic cervical length measurement in the management of twin pregnancies is warranted. It still remains unclear whether routinely measuring cervical length in such pregnancies improves outcomes or potentially results in more interventions without improving outcomes. However, without early identification of twin pregnancies destined to deliver prematurely we cannot expect to improve outcomes, therefore defining the role of cervical length is important. Appropriately designed interventional studies to improve outcomes in twin pregnancies with cervical shortening in the second trimester are warranted to optimize management in this high-risk population.

Acknowledgements

This research study was made possible through a grant from the Memorial Medical Center Foundation, Long Beach, California, USA.

Ancillary