- Top of page
Asymptomatic singleton pregnancies at high risk for preterm birth were followed prospectively with one transvaginal ultrasound examination of the cervix from 10 + 0 to 13 + 6 weeks, and then every 2 to 4 weeks up to 23 + 6 weeks, over a 7-year period ending in June 2002. Results from some of these patients examined between 14 and 23 weeks have been previously reported3. Patients with one or more of the following historical criteria were eligible for inclusion: one or more spontaneous preterm births between 14 and 34 weeks; two or more dilatation and curettage procedures because of voluntary abortion, Müllerian anomaly, cone biopsy, or diethylstilbestrol exposure. Exclusion criteria were placenta previa, current drug abuse, and severe fetal anomalies. Since the primary outcome was spontaneous preterm birth, pregnancies with induced preterm births were also excluded. Gestational age was verified by a fetal sonographic examination before 14 weeks in all patients.
Transvaginal ultrasound examination of the cervix was performed after the patient had emptied her bladder. After insertion of the probe and visualization of the cervix, the probe was withdrawn until the image blurred and was reapplied with only enough pressure to restore the image, as previously described6, 7. When the internal os was closed, cervical length was measured between the internal os and the external os; when the internal os was open, cervical length was measured between the tip of the funnel and the external os1. The internal os was defined as the sonographic point at which the most caudal tip of continuous contact between the two lips of the endocervical canal/lower uterine segment and the amniotic cavity met. This is also recognized as ‘a dimple against the hypoechogenic background of amniotic fluid’8. If a poorly developed lower uterine segment was initially noticed8, care was taken to wait a few minutes, to move the probe to try to obtain the best assessment of the true endocervical canal length, and to measure the shortest distance between the external os and the sonographic internal os (described above). Funneling refers to the percentage of the cervix which is open, and is calculated by dividing the funnel length by the total cervical length1. Transfundal pressure was applied as previously described9. The shortest best cervical length of at least three measurements was used for analysis.
All sonographic images were reviewed by a single investigator who was blinded to pregnancy risks and outcomes. A cervical length of < 25 mm was considered a short cervix because this sonographic cut-off has been previously shown to be associated with a higher risk of preterm birth in both high-risk1, 3, 10, 11 and general population7 pregnancies. Results of the ultrasound examinations were available to the obstetrician. All transvaginal ultrasound scans before 14 weeks were performed before any intervention, and were not used for clinical care. Prophylactic cerclages were performed at 12–14 weeks, after the ultrasound examinations of the cervix had been carried out, and they were performed based on prior obstetric history only.
The primary outcome was spontaneous preterm birth < 35 weeks. Data entered prospectively in an Excel file (Microsoft Corp, Redmond, WA, USA) were retrospectively reviewed and analyzed using Datadesk statistical program version 5 (Ithaca, NY, USA). Statistical analysis included t-test for continuous variables and the chi-square or Fisher's exact test for categorical variables.
- Top of page
One hundred and ninety-three high-risk patients had transvaginal ultrasound examination of the cervix between 10 and 14 weeks, of which 10 were excluded due to drug abuse (n = 3), voluntary abortion (n = 3), or induced delivery for fetal indications (n = 4). The incidence of preterm birth was 20% (36/183). Ten (5%) of the patients had a cervical length < 25 mm between 10 and 14 weeks, whereas 173 (95%) had cervical length ≥ 25 mm. Transfundal pressure did not modify any of these initial observations. The reviewer changed the length of the cervix only twice, where a poorly developed segment was seen, but these changes did not convert any examinations from a normal to a short cervix, or vice versa. These two measurements were the two longest in the study: 51 and 67 mm.
Demographic characteristics are presented in Table 1. There was no difference in age, race, parity, health care provider, or smoking between the patients with and those without a cervical length < 25 mm. The distribution of patients by risk factor is presented in Table 2. Sixty-nine percent of all patients had a history of preterm birth, with 52% having had a prior spontaneous second-trimester loss. Prior spontaneous second-trimester loss and cone biopsy were the only risk factors found to be significantly associated with a cervical length < 25 mm before 14 weeks.
Table 1. Demographic characteristics of the study population
|Characteristic||All patients (n = 183)||Total population|
|Short cervix*(n = 10)||Normal cervix†(n = 173)||Statistical significance‡|
|Age (years, mean ± SD)||31.4 ± 5.6||31.5 ± 4.5||31.4 ± 5.7||P = 0.9|
|Race (n (%))|
| White||88 (48)||4 (40)||84 (48)|| |
| African-American||81 (44)||4 (40)||77 (45)||P = 0.3|
| Other||14 (8)||2 (20)||12 (7)|| |
|Parity (n (%))|
| Primiparous||45 (25)||2 (20)||43 (25)|| |
| Multiparous||138 (75)||8 (80)||130 (75)||P = 0.7|
|Health care provider (n (%))|
| Private||152 (83)||9 (90)||143 (83)|| |
| Clinic||31 (17)||1 (10)||30 (17)||P = 0.5|
|Smoker (n (%))||18 (10)||1 (11)||17 (10)||P = 0.6|
Table 2. Distribution of patients by risk factor
|Risk factor||All patients (n = 183) ( n (%))||Total population|
|Short cervix*(n = 10) ( n (%))||Normal cervix†(n = 173) ( n (%))||Statistical significance‡|
|≥ 1 spontaneous 14–24-week loss||96 (52)||9 (90)||87 (50)||P = 0.01|
|≥ 1 painless spontaneous 14–24-week loss||50 (27)||5 (50)||45 (26)||P = 0.2|
|≥ 1 preterm birth at 14–34 weeks||126 (69)||9 (90)||117 (68)||P = 0.3|
|≥ 2 D&C||53 (29)||4 (40)||49 (28)||P = 0.4|
|Müllerian anomaly||16 (9)||0 (0)||16 (9)||P = 0.3|
|Cone biopsy||28 (15)||4 (40)||24 (14)||P = 0.03|
|Diethylstilbestrol exposure||20 (11)||2 (20)||18 (10)||P = 0.7|
|Cerclage in a prior pregnancy||58 (32)||4 (40)||54 (31)||P = 0.6|
|Failed cerclage in a prior pregnancy||34 (19)||3 (30)||31 (18)||P = 0.3|
Five (50%) of the 10 patients with an early cervical length < 25 mm delivered preterm (5/9 with prior spontaneous second-trimester loss and 1/4 with prior cone biopsy), while 31 (18%) of the 173 patients with a cervical length ≥ 25 mm delivered preterm (P = 0.04). The sensitivity, specificity and positive and negative predictive values of a cervical length < 25 mm between 10 and 14 weeks for predicting spontaneous preterm birth at < 35 weeks were 14%, 97%, 50%, and 82%, respectively (relative risk, 2.8; 95% confidence interval (CI), 1.4–5.6). The mean cervical length at 10–14 weeks of patients who delivered preterm was 33.7 ± 6.9 mm, versus 35.0 ± 6.8 mm in patients who delivered at term (P = 0.3). The distribution of cervical length at 10 to 13 + 6 weeks and related incidence of preterm delivery is shown in Figure 1.
Figure 1. Distribution of total number of patients (total bar heights) and number of patients with preterm delivery (bottom shaded areas) by cervical length at 10 to 13 + 6 weeks (n = 183). On top of each bar, the ratio of the number of patients with preterm delivery/total number of patients (for each cervical-length group) is given.
Download figure to PowerPoint
Follow-up transvaginal ultrasound from 14 to 24 weeks revealed that six of the 10 (60%) patients with a cervical length < 25 mm before 14 weeks continued to have a cervical length < 25 mm and four of these six (67%) delivered preterm; the other four did not and delivered ≥ 35 weeks. Of the 173 patients with a cervical length ≥ 25 mm before 14 weeks, 55 (32%) had developed a cervical length < 25 mm at 14–24 weeks, with 14 of 55 (25%) delivering preterm. The sensitivity, specificity and positive and negative predictive values of a cervical length < 25 mm between 14 and 24 weeks in our total population of 183 patients were 69%, 76%, 41%, and 89%, respectively (relative risk, 4.5; 95% CI, 2.4–8.6). The average gestational age between 10 and 24 weeks when patients had a cervical length < 25 mm was 18.7 ± 2.9 weeks, with 10 weeks being the earliest gestational age at which shortening occurred. The distribution of cervical length at 14 to 23 + 6 weeks and related incidence of preterm delivery is shown in Figure 2.
Figure 2. Distribution of total number of patients (total bar heights), and number of patients with preterm delivery (bottom shaded areas) by cervical length at 14 to 23 + 6 weeks. On top of each bar, the ratio of the number of patients with preterm delivery/total number of patients (for each cervical-length group) is given.
Download figure to PowerPoint
Eight of the 10 (80%) patients with a cervical length < 25 mm before 14 weeks had a prophylactic cerclage (the two who did not receive cerclage delivered at 19 and 33 weeks), while 86 (50%) of the 173 patients with an early cervical length ≥ 25 mm had cerclage (P = 0.12).
- Top of page
This study confirms that most patients destined to deliver preterm, whether they are at low or high risk of doing so, develop a cervical length < 25 mm between 16 and 22 weeks, as shown in a prior report2, and not earlier. While prior reports have evaluated the cervix by transvaginal ultrasound after 18 or more weeks7, 12–14, we have recently reported that a cervical length < 25 mm can also be detected between 14 and 18 weeks1, 3, and is indeed very predictive of preterm birth when it occurs (positive predictive value 70% for cervical length < 25 mm and/or funneling > 25%)3. We began to notice that some patients had a cervical length < 25 mm at their first ultrasound examination, raising the possibility that a short cervix might be present even before this gestational age and may even be congenital. Guzman et al.2 reported two patients, both with a history of prior second-trimester loss, who had a short cervix on their first transvaginal ultrasound examination, performed at 15 weeks. Transvaginal ultrasound earlier than 14 weeks has been evaluated only in two prior studies of low-risk patients. Rosenberg et al.4 reported a cross-sectional study of first-trimester (exact weeks not specified) transvaginal ultrasound cervical length in 280 patients, of which 261 had singleton gestations. The first-trimester cervical length of patients who eventually delivered at term did not differ significantly from that of patients who had losses at < 25 weeks (37 vs 36 mm). Zalar5 reported a cross-sectional study of transvaginal ultrasound cervical length performed at 11 weeks in 373 singleton pregnancies. Median cervical length was 47 mm. Gestational age at delivery of patients who had a first-trimester cervical length ≤ 40 mm (10th percentile) was not significantly different from that of patients who had a first-trimester cervical length > 40 mm (39.6 vs 39.7 weeks). Only 12 (3%) patients were found to have a cervical length ≤ 35 mm, which may not even represent all abnormal cervices, since the 10th percentile of cervical length at 24 weeks in population studies is 25 mm7.
The cervix of high-risk patients has never before been evaluated by transvaginal ultrasound before 14 weeks. We tried to study a very high-risk group, and indeed 52% of the patients had a prior second-trimester spontaneous loss. Only 10 (5%) of these 183 patients had a cervical length < 25 mm between 10 and 14 weeks. While cervical changes later in pregnancy could at least in part be attributed to uterine contractions, these uncommon early changes could be due either to congenital or iatrogenically acquired factors. Six of these 10 patients with early cervical changes had prior cervical surgery or manipulation (dilatation and curettages), so that it is unclear if any of them truly had a congenitally short cervix or if the early short cervix was acquired due to iatrogenic causes. A cervix congenitally abnormal because of altered histology, collagen content or function cannot be excluded in these patients, but our data do not support the notion that congenitally short cervices are a major etiology of preterm births. A congenitally short cervix may be difficult to identify in pregnancy since on ultrasound it appears that in the first trimester the lower uterine segment is not yet distended with the rest of the corpus by the growing gestational sac, and therefore can still appear as part of the cervix, with the true internal os not easily identifiable on ultrasound5. This may be why first-trimester ultrasound, especially in low-risk patients as those studied by Rosenberg et al.4 and Zalar,5 can result in cervical lengths which are longer than those in the second trimester. The vesicocervical fold15 or a change in echogenicity may be useful markers to distinguish the boundary between cervix and lower uterine segment. We would like to emphasize that to measure accurately cervical length on transvaginal ultrasound, a standard technique with attention to landmarks should be adhered to strictly7, 8, 16. It is unclear what the true normal in-vivo premenopausal non-pregnant cervical length is, but hysterectomy specimens from such patients revealed an average length of 38.5 mm17. Why, in the first trimester, the cervix does not often appear to be short is unclear. Mechanisms that may lead to a short cervix such as gravity, uterine distention, and inflammatory cytokines are not likely at this early gestational age18, 19.
As we have alluded to in a prior report1, in patients at high risk for preterm birth such as those included in this study (prior preterm birth, diethylstilbestrol exposure, Müllerian anomaly, cone biopsy, or more than one dilatation and curettage procedure), a baseline transvaginal sonographic evaluation of the cervix should be performed at 14–17 weeks, and then repeated at the time when most changes are detected, 18–22 weeks (the gestational age of the anatomy ultrasound). Only patients with a prior cone biopsy had an incidence of short cervix before 14 weeks of more than 10% (14%), and only these patients and those with a prior second-trimester loss had a significant association with short cervix before 14 weeks (Table 2). Given the poor sensitivity of transvaginal ultrasound of the cervix before 14 weeks in the other high-risk populations, only patients with a prior cone biopsy or second-trimester loss should possibly have this screening test before 14 weeks. Overall, while 5% of high-risk patients had a short cervix before 14 weeks, 33% of these same patients had a short cervix at 14–24 weeks. The relative risk for preterm birth < 35 weeks of a cervical length < 25 mm before 14 weeks was significant (relative risk, 2.8), but was much less than that when this finding was detected at 14–24 weeks (relative risk, 4.5).
Theoretically, if transvaginal ultrasound could reliably show early predictive changes, like cervical length < 25 mm, even before 14 weeks, the need for a prophylactic cerclage could be assessed by the ultrasound scan. Since too few patients who are destined to deliver preterm show a short cervix on transvaginal ultrasound before 14 weeks, the need for prophylactic cerclage cannot be assessed by this examination alone. The need for a cerclage must be assessed only by a clear history of cervical incompetence (more than three painless second-trimester losses or preterm births)20. The capacity of cerclage performed after cervical changes are detected on transvaginal ultrasound to prevent preterm birth is unclear2, 3, 13, but is currently being investigated in a randomized fashion at several centers. Optimal cervical length for cerclage placement and type of cerclage will be clearer when such studies are published.
Because the operators in this study were not prospectively blinded to the transvaginal ultrasound results, predictive accuracy may have been influenced by different interventions, especially cerclage. This influence is unlikely given the fact that cerclage in this clinical setting has not been shown to alter outcome in our population3, but we cannot rule out that early cerclages in patients with a cervix < 25 mm before 14 weeks could have affected the primary outcome, and therefore the value of this study in reference to our second objective, prediction of preterm birth. The main objective of the study was to determine if a cervical length < 25 mm develops before 14 weeks, and all cerclages were performed after the 10–14-week ultrasound examinations. The fact that four of the 10 patients who had a short cervix between 10 and 14 weeks later had a longer cervix may have been due to the effect of cerclage, since this procedure can have a lengthening effect on the cervix21, 22.
In summary, even in high-risk patients, transvaginal ultrasound of the cervix is not a very sensitive predictor of preterm birth before 14 weeks, since a cervical length < 25 mm rarely develops before this gestational age except in patients with a prior second-trimester loss or a prior cone biopsy. In most high-risk patients, cervical changes predictive of preterm birth such as a cervical length < 25 mm occur between 16 and 22 weeks.