Prematurity remains the most important risk for morbidity and mortality for the unborn child. Transvaginal ultrasonography of the cervix is a safe1 and well-accepted2 technique to assess the risk of preterm birth.3 Cervical length and funneling have both been used to predict preterm birth, but the value of one finding with or without the other has been uncertain.4–6
Our three objectives were to estimate: 1) the natural history of funneling in the second trimester by serial transvaginal ultrasonograms, 2) whether the presence of funneling is associated with an increased risk of earlier spontaneous birth after controlling for cervical length, and 3) whether funneling increased the risk of earlier spontaneous birth in the presence of a normal or shortened cervical length.
MATERIALS AND METHODS
We performed a secondary analysis of a blinded observational study of cervical ultrasonography in women with a prior spontaneous preterm birth between 16.0 and 31.9 weeks with singleton gestations. The primary study was a multicenter prospective study performed between 1997 and 1999.7 Women were excluded if they had chronic medical or obstetric problems that might result in an indicated preterm birth (such as hypertension or alloimmunization), a history of substance abuse, a uterine anomaly, or a history-indicated cerclage. At or before 18 weeks gestation, eligible women were offered enrollment in the study, which was approved by the institution review board in each center. Consenting asymptomatic women were scheduled to have a transvaginal ultrasound examination of the cervix screening at 16 to 18 weeks 6 days and then every 2 weeks until 23 weeks 6 days, for a maximum of four ultrasound examinations.
The transvaginal ultrasound technique has been described in detail.7 All ultrasound ultrasonologists received specific training before the study and submitted at least 10 transvaginal ultrasound images of the cervix before they were approved to perform study scans. The training and ultrasound images were assessed centrally, by the same person. After emptying her bladder, each woman inserted or had the ultrasonologist insert the sterile, covered probe. Avoiding excessive pressure, the adequate image for cervical measurements was defined as visualization of internal and external os with complete view of the endocervical canal. The association of spontaneous preterm birth with cervical length and other measurements has been previously described and reported for this study.5,7
We defined a cervical length less than 25 mm as short, representing the lower tenth percentile in our study population. Funneling was defined as prolapse of the membranes through a dilated endocervical canal. Funnel depth was measured from the functional internal os or tip of the funnel to its “shoulder,” ie, the presumptive original internal os more cephalad. Funnel depth had to be at least 5 mm to be defined as funneling. Funneling percent was defined as the ratio of funnel depth divided by funnel depth plus cervical length.8 Funnel shape (V or U) was recorded.9
After measurements were performed, transfundal pressure was applied for 15 seconds, and if the cervical length shortened or a funnel developed (or enlarged), the cervical measurements were repeated. Ultrasound evaluations also lasted a minimum of 5 minutes to detect any spontaneously occurring dynamic shortening after fundal pressure was released. By protocol, all scans were videotaped, and the results were not available to the managing obstetricians except in cases of previa or fetal death. All cases of funneling were later reviewed by three of the investigators (J.O., V.B., and C.M.) for quality assurance purposes.
To evaluate the natural history of funneling, we planned to examine the progression of the presence or shape of funnel from initial to final scan. Funneling was also classified and analyzed as: 1) Progressive: At least two scans with appreciable increase in funnel size of 3 mm or more (either V or U, including a change in shape) over time; 2) Persistent: At least two consecutive scans with a funnel (either type) whose size remained relatively constant over time; 3) Isolated: Only observed on one (or occasionally two) scan(s); if observed twice, there must have been a funnel-free examination in the interim. Importantly, this was also associated with a stable cervical length on the serial evaluations; 4) Shortening in progress: Funnels that were intermittent over time, but were associated with appreciable cervical shortening of 3 mm or more on serial evaluations; 5) Last only: A funnel was only observed on the last recorded scan, in patients with at least two scans.
The primary outcome for this secondary analysis was gestational age at delivery. We focused our analyses on shortest observed cervical length and on funneling observed on any serial ultrasonograms after dynamic changes. Data were analyzed with SAS 8.2 (SAS Institute Inc, Cary, NC). Baseline comparisons between patients with and without a funnel were performed using the Wilcoxon rank sum test (for continuous variables) and the χ2 test (for categorical variables). The Spearman correlation coefficient was used to assess the relationship between funnel index and gestational age at delivery. Both parametric (t test, linear regression) and nonparametric (Wilcoxon rank sum, Kruskal-Wallis) analyses were used to analyze the relationship between gestational age at delivery and funnel variables, with and without considering the effects of cervical length. P values less than .05 were considered statistically significant. No adjustments were made for multiple comparisons.
A total of 590 transvaginal ultrasound evaluations of the cervix were performed on 183 high-risk women. Twenty-six percent of these women experienced a recurrent spontaneous preterm birth at less than 35 weeks of gestation in the current pregnancy, with a mean gestational age at birth of 35.2 (±6.3) weeks of gestation.
With serial evaluations, and including all observed dynamic changes, 60 (33%) women had funneling of any of the categories during at least one ultrasonogram. These 60 women differed significantly from the 123 high-risk women who never developed funneling (Table 1). Women ultrasonographically exhibiting a funnel were more likely to be African-American, had government-assistance insurance and experienced an earlier gestational age at their prior preterm birth. Their shortest observed cervical lengths were smaller, and they experienced an earlier gestational age at delivery than women with no funneling. Preterm birth occurred more commonly by preterm labor with intact membranes than premature rupture of membranes in women who displayed funneling.
The natural history of these 60 women with funneling can be seen in Table 2. In most of these high-risk women, funneling was not present initially, but developed on subsequent evaluations. The progression over time of internal os cervical anatomy from a “T” to a “V” to a “U” shape was associated with earlier gestational age at delivery; development of a “U” shaped funnel was associated with early birth, usually in the second trimester. In the uncommon occurrence of resolution on subsequent ultrasounds of a “V” shape funnel, women delivered at term. Table 3 shows gestational age at birth of different categories of funneling progression. None of the categories proved to be significantly different from the others in its prediction of gestational age at delivery (P=.14). Largest observed funneling percent, considering dynamic changes and serial examinations, correlated with the gestational age at delivery (Fig. 1). As the funneling percent increased, the mean gestational age at delivery decreased (Spearman correlation coefficient=−0.3, P=.049).
Seventy-seven percent (46 of 60) of women with funneling on serial ultrasonograms after dynamic changes also had a shortened cervical length less than 25 mm (Table 4). In the presence of a shortened cervical length, funneling did not alter outcome, because both groups, with and without funneling, had similar mean preterm gestational ages at birth. In the presence of a cervical length of 25 mm or more, the presence of funneling was associated with more than a 2-week earlier gestational age at delivery compared with no funneling, but this difference was not statistically significant (P=.22). Among the 60 women with funneling, regression analysis showed an inverse association between funneling percent and cervical length (P=.009). Regression analysis controlling for gestational age at first funnel, maximal funnel depth, and shortest cervical length at largest funnel depth, for the dependent variable gestational age at delivery, revealed only shortest cervical length to be significant. Largest funnel percent observed at any time did not seem to be a significant independent risk factor.
We report a blinded study on the natural history of second trimester funneling in women at high risk for spontaneous preterm birth. High-risk status was defined as a history of a prior early spontaneous preterm birth. The results demonstrate that funneling occurs in a third of these women, and therefore is a common finding. Funneling has a significant association with earlier gestational age at delivery. Progression of funneling from no funneling (“T”) to “V” to “U” and increasing funneling percent were both associated with earlier gestational age at delivery. Nonetheless, as we have observed, funneling was associated with considerable variability.
These findings confirm prior smaller or nonblinded studies.8,10 If funneling is present, it can be recorded in different ways. This study, by design, defined funneling as prolapse of the membranes through a dilated endocervical canal with a funnel depth of at least 5 mm measured from the “shoulder” of the presumptive original internal os to the tip of the funnel at the closed endocervical canal. Funneling has also been recorded in using 3 mm instead of 5 mm for membrane prolapse,6 measuring the funnel depth from the middle and not the lateral border of the original internal os,10 measuring the dilatation of the internal os with 5 mm or more funnel width as significant,11 measuring a cervical index (funnel length+1/closed cervical length),12 measuring a cervical score (cervical length minus cervical dilatation),13 and measuring funneling percent—or percentile (also called funnel index in one reference)14 as the ratio of funnel depth to total cervical length (funnel depth plus remaining closed cervical length).4,8 Some studies of funneling did not give an objective definition.15 Although it is unclear which of these methods of assessing funneling is best, all have been reported to predict preterm birth.16 Reporting funneling as a ratio allows for quantification of funneling. For example, minimal (less than 25%) funneling, a very common finding, may be physiologic, and should not raise suspicion for later preterm birth or trigger interventions, whereas funneling progressively worse after 25% is linearly associated with the risk of preterm birth.8 Like cervical length assessment, it is important to report the more significant (worse) funneling seen during the whole ultrasound session (ie, before and after dynamic spontaneous or after transfundal pressure changes) to obtain the best prediction. Unlike cervical length, funneling is associated with substantial variations in appearance and technique of measurement that may complicate clinical interpretation of this finding.
If funneling is present, the shape can be recorded. Zilianti et al9 described a process of initial long and closed cervix (T-shaped), then initial Y- and then V-shaped funneling, later evolving into U-shape funneling in women at the beginning of term labor. Our study reports the natural progression of funneling shape in asymptomatic preterm cervical changes. Indeed the progression to U-shape funnel is associated with the earliest gestational age at delivery (Table 2). The study might have lacked the power to show significant differences between funneling categories (Table 3). We remind the readers that funneling shape can be somewhat subjective.
Uterine segment contractions can mimic funneling. The appearance of a rounded myometrium above the internal os and of different echogenicity between the contracted lower uterine segment and the actual cervix are ultrasonographic signs that help distinguish between the two entities. Usually the presence of a normal (ie, one with a length of 25 mm or more) cervix below the possible funneling, also termed poorly developed lower uterine segment,7 has been reported as a reassuring finding, one that possibly rules out any significant risk.11 Our data, though, show that funneling at initial examination in the presence of a normal cervix is associated with more than a 2-week earlier gestational age at delivery compared with no funneling. This difference is not statistically significant (P=.22), and this could be secondary to either lack of association or lack of power. Further research is needed to clarify this issue. Some cervices are associated with preterm birth through a process of funneling and shortening, some just through shortening. However, because our observation windows were relatively small (5 minutes every 2 weeks), it is possible that some women with progressive cervical shortening experienced unobserved funneling.
Although in this study the natural history of second trimester funneling in high-risk women suggests that it has a significant association with earlier gestational age at birth, as an independent finding we found that it does not add appreciably to the risk of spontaneous preterm birth associated with the finding of a shortened second-trimester cervical length less than 25 mm. This finding confirms some prior studies of high-risk women,4,11,14 but is at odds with others.6,8,15 The association between funneling and preterm birth was stronger than the one between cervical length and preterm birth in the first report on funneling.8 In that study, predictive accuracy of funneling was similar to that of cervical length, and at times better, especially for specificity and positive predictive value. However, that study was limited to only women with funneling. When all high-risk women from this center were followed up prospectively, predictive values were very similar for cervical length less than 25 mm or funneling more than 25%, with sensitivity increased from 61% to 74% and relative risk from 3.3 to 4.8 if funneling was added to cervical length as a predictive variable.3 The largest study of cervical sonography to compare prediction for preterm birth by either funneling or cervical length found the two had very similar predictive accuracy, with funneling remaining a significant predictor even after controlling for cervical length.6 A short cervical length less than 25 mm in the absence of funneling has been correlated with a very low risk of preterm birth (mean gestational age at delivery 36.2 weeks) in a recent study.17 Our data show instead a mean gestational age at delivery of 31.9 weeks in a similar group of women (Table 4). This might be due to the fact that all our women also had a prior preterm birth, whereas only about a third of women in the prior study had such a risk factor.17 In other words, a cervical length less than 25 mm at less than 24 weeks in the absence of funneling may be associated with a high risk of preterm birth, especially in women with a prior preterm birth.
After careful review of these new analyses and prior literature, we conclude that funneling is indeed predictive of early gestational age at delivery, but does not add significantly to the prediction obtained by an accurate cervical length measurement, at least in high-risk women. Funneling can be falsely alarming in at least two situations: if indeed a lower uterine contraction is present and there is not true funneling, and if it is only minimal (less than 5 mm deep or less than 25%), which is a normal finding. Funneling is also more subjective, and therefore less reproducible. Funneling may be useful in the minority (about 5%) of women with initially a normal cervical length but significant funneling, because this funneling may be the first marker of later shortening and eventual preterm birth or could be an artifact from a poorly-developed uterine segment. Therefore, cervical length, recorded as the shortest best after dynamic changes are taken into account, should be regarded as the standard measurement for evaluation of the cervix for prediction of preterm birth in the second trimester. While several other measurements can be obtained, none is more predictive or has better reproducibility. Further research should focus on interventions to prevent preterm birth once a short cervical length is detected.
1. Krebs-Jimenez J, Neubert AG. The microbiological effects of endovaginal sonographic assessment of cervical length. J Ultrasound Med 2002;21:727–9.
2. Dutta RL, Economides DL. Patient acceptance of transvaginal sonography in the early pregnancy unit setting. Ultrasound Obstet Gynecol 2003;22:503–7.
3. Berghella V, Bega G, Tolosa JE, Berghella M. Ultrasound assessment of the cervix. Clinical Obstet Gynecol 2003;46:947–62.
4. Guzman ER, Walters C, Ananth CV, O’Reilly-Green C, Benito CW, Palermo A, et al. A comparison of sonographic cervical parameters in predicting spontaneous preterm birth in high-risk singleton gestations. Ultrasound Obstet Gynecol 2001;18:204–10.
5. Yost NP, Owen J, Berghella V, MacPherson C, Swain M, Dildy GA 3rd, et al. Second-trimester cervical sonography: Features other than cervical length to predict spontaneous preterm birth. Obstet Gynecol 2004:103;457–62.
6. 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.
7. Owen J, Yost N, Berghella V, Thom E, Swain M, Dildy GA, et al. Mid-trimester endovaginal sonography in women at high risk for spontaneous preterm birth. JAMA 2001;286:1340–8.
8. Berghella V, Kuhlman K, Weiner S, Texeira L, Wapner RJ. Cervical funneling: sonographic criteria predictive of preterm delivery. Ultrasound Obstet Gynecol 1997;10:161–6.
9. Zilianti M, Azuaga A, Calderon F, Pages G, Mendoza G. Monitoring the effacement of the uterine cervix by transperineal sonography: a new perspective. J Ultrasound Med 1995;14:719–24.
10. Berghella V, Tolosa JE, Kuhlman K, Weiner S, Bolognese RJ, Wapner RJ. Cervical ultrasonography compared with manual examination as a predictor of preterm delivery. Am J Obstet Gynecol 1997;177:723–30.
11. To MS, Skentou C, Liao AW, Cacho A, Nicolaides KH. Cervical length and funneling at 23 weeks of gestation in the prediction of spontaneous early preterm delivery. Ultrasound Obstet Gynecol 2001;18:200–3.
12. Gomez R, Galasso M, Romero R, Mazor M, Sorokin Y, Goncalves L, et al. Ultrasonographic examination of the cervix is better than cervical digital examination as a predictor of the likelihood of preterm delivery in patients with premature labor and intact membranes. Am J Obstet Gynecol 1994;171:956–64.
13. Hartmann K, Thorp JM, McDonald TL, Savitz DA, Granados JL. Cervical dimensions and risk of preterm birth: a prospective cohort study. Obstet Gynecol 1999;93:504–9.
14. Hasegawa I, Tanaka K, Takahashi K, Tanaka T, Aoki K, Torii Y, et al. Transvaginal ultrasonographic cervical assessment for the prediction of preterm delivery. J Mater Fetal Med 1996;5:305–9.
15. Andrews WW, Copper R, Hauth JC, Goldenberg RL, Neely C, Dubard M. Second-trimester cervical ultrasound: associations with increased risk for recurrent early spontaneous delivery. Obstet Gynecol 2000;95:222–6.
16. Honest H, Bachmann LM, Coomarasamy A, Gupta JK, Kleijnen J, Khan KS. Accuracy of cervical transvaginal sonography in predicting preterm birth: a systematic review. Ultrasound Obstet Gynecol 2003;22:305–22.
17. Rust OA, Atlas RO, Kimmel S, Roberts WE, Hess LW. Does the presence of a funnel increase the risk of adverse perinatal outcome in a patient with a short cervix? Am J Obstet Gynecol 2005;192:1060–6.
In addition to the authors, other members of the National Institute of Child Health and Human Development Maternal–Fetal Medicine Units Network are as follows:
University of Alabama at Birmingham—J. Hauth, A. Northen, C. Neely, and D. Thom.
University of Chicago—A. H. Moawad and P. Jones.
University of Cincinnati—N. Elder, T. Haskins, and D. Shultz.
The George Washington University Biostatistics Center—E. Thom and L. Leuchtenburg.
National Institute of Child Health and Human Development—D. McNellis and C. Spong.
University of Tennessee—B. M. Mercer, R. Ramsey, J. Fricke, and M. Peterson.
University of Texas at San Antonio—S. Barker and C. Leija.
University of Texas Southwestern Medical Center—K. Leveno, J. McCampbell, and R. Benezue.
Thomas Jefferson University—R. Wapner, M. DiVito, and G. Bega.
University of Utah—M. Varner, E. Taggart, and R. Zollinger.
Wake Forest University—P. Meis and A. Henshaw.