Cervical length at 18–22 weeks of gestation for prediction of spontaneous preterm delivery in Hong Kong Chinese women

Authors

  • T. N. Leung,

    Corresponding author
    1. Fetal Medicine Unit, Department of Obstetrics and Gynaecology, The Chinese University of Hong Kong, The Prince of Wales Hospital, Hong Kong
    • Fetal Medicine Unit, Department of Obstetrics and Gynaecology, The Chinese University of Hong Kong, The Prince of Wales Hospital, Hong Kong
    Search for more papers by this author
  • M. W. Pang,

    1. Fetal Medicine Unit, Department of Obstetrics and Gynaecology, The Chinese University of Hong Kong, The Prince of Wales Hospital, Hong Kong
    Search for more papers by this author
  • T. Y. Leung,

    1. Fetal Medicine Unit, Department of Obstetrics and Gynaecology, The Chinese University of Hong Kong, The Prince of Wales Hospital, Hong Kong
    Search for more papers by this author
  • C. F. Poon,

    1. Fetal Medicine Unit, Department of Obstetrics and Gynaecology, The Chinese University of Hong Kong, The Prince of Wales Hospital, Hong Kong
    Search for more papers by this author
  • S. M. Wong,

    1. Fetal Medicine Unit, Department of Obstetrics and Gynaecology, The Chinese University of Hong Kong, The Prince of Wales Hospital, Hong Kong
    Search for more papers by this author
  • T. K. Lau

    1. Fetal Medicine Unit, Department of Obstetrics and Gynaecology, The Chinese University of Hong Kong, The Prince of Wales Hospital, Hong Kong
    Search for more papers by this author

Abstract

Objective

To assess the value of a single cervical length measurement by transvaginal sonography (TVS) at the time of mid-trimester anomaly scan for predicting spontaneous preterm delivery (SPD) among Chinese women.

Methods

A prospective observational study was carried out involving 2880 subjects with singleton pregnancies and confirmed gestational age. Cervical length was measured at 18–22 weeks of gestation.

Results

The incidence of SPD < 34 weeks and < 37 weeks were 0.7% and 3.7%, respectively. Women with SPD < 34 weeks and SPD < 37 weeks had shorter median cervical lengths (32.6 mm and 36.2 mm, respectively) than those with term deliveries (37.6 mm) (P = 0.006 and 0.025, respectively). The predictive performance of cervical length was better for SPD < 34 weeks compared with < 37 weeks. A cervical length ≤ 27 mm, which corresponded to the 4th centile, occurred in 36.8%, 62.5% and 100% of those with SPD < 34, < 30 and < 26 weeks, respectively. The positive likelihood ratio (LR) of a cervical length ≤ 27 mm in predicting SPD < 34 weeks was 9.8. Using logistic regression, both short cervix and funneling were independent predictors for SPD < 34 weeks of gestation. The coexistence of funneling and a cervical length ≤ 27 mm gave a positive predictive value (PPV) and LR of SPD < 34 weeks of 14.7% and 26.0, respectively.

Conclusions

Mid-trimester cervical length is predictive of SPD in Chinese women. However, given the low PPV of a short cervical length, its clinical utility is still limited in low-risk populations. Copyright © 2005 ISUOG. Published by John Wiley & Sons, Ltd.

Introduction

It is well known that women with a short cervix in the mid-trimester carry a higher risk for spontaneous preterm delivery (SPD)1–3. Transvaginal ultrasonographic assessment was shown to be more sensitive and reproducible than digital examination4 or transabdominal ultrasonography5 in predicting the risk of preterm birth. The technique of cervical length measurement by transvaginal ultrasound has been well standardized and its reproducibility confirmed6. However, most of these studies were performed in Caucasian populations. Ethnic differences in cervical length have been reported; for instance, women of Afro-Caribbean origin were shown to have significantly shorter cervices compared with Caucasian women7. Previously in our unit, cervical length of Chinese women measured in the third trimester was also found to be shorter than that reported in the Caucasian population8. SPD can be the final outcome of various pathologies and functional incompetence of the cervix might be just one of them9. The significance of a short cervix for preterm labor among a Chinese population remains to be elucidated.

The objective of this study was to examine the predictive value of cervical length and funneling for subsequent SPD assessed by mid-trimester transvaginal ultrasonography in our Chinese population.

Methods

Study population

This was a prospective observational study conducted in a tertiary obstetric unit in Hong Kong between March 2000 and July 2002. This unit catered for both low- and high-risk obstetric populations. As a unit policy, all women were offered a routine ultrasound examination for fetal biometry and assessment of morphology at around 20 weeks of gestation if booked before this date. All ethnically Chinese women attending this routine ultrasound session were invited to participate in the study. Ethical approval was obtained from the institutional human research ethics committee concerned and all women gave their informed consent to the study. Only singleton pregnancies with ultrasound measurement at 18–22 completed weeks of gestation were included. Pregnancies complicated by major fetal anomalies detected during the ultrasound session were excluded. The women's gestational ages were confirmed by ultrasound examination.

Sonographic assessment of cervix

After fetal biometry and morphology assessment, the cervix was examined using real-time transvaginal ultrasonography with a 5–8-MHz transducer (Acuson 128XP/10 (Acuson, Mountain View, CA, USA) or Apogee 800PLUS (Apogee ATL, Bothell, WA, USA)). The cervical length was measured as described in the literature1, 6. While the woman's bladder was empty, the transvaginal ultrasound probe was placed in the anterior fornix of the vagina. The appropriate sagittal view was identified by the location of the triangular area of echodensity at the external os, a V-shaped notch at the internal os and a faint line of echodensity or echolucency between the two. Undue pressure on the cervix that might artificially increase its apparent length was avoided. Care was taken to ensure that the whole length of the cervical canal could be observed and the distance from the surface of the posterior lip to the cervical canal was equal to the distance from the anterior lip to the cervical canal. The cervical length was measured in this fashion three times along the interface of the mucosal surface. The shortest measurement was recorded as the cervical length. The presence or absence of a funnel (defined as the protrusion of the amniotic membranes 5 mm or more into the cervical canal measured along the lateral border of the funnel) was also recorded. When funneling was present, the width and depth was measured. Each scan lasted at least 5 min to observe any funneling. Ultrasound images of all cervical measurements were recorded for audit purposes.

All subjects and the clinical staff were blinded to the cervical measurements. The only exception according to the study protocol was that the clinician in charge would be informed of the results if there was a cervical dilatation ≥ 2 cm and effacement of the cervix. However, this did not happen in any of the subjects. Hence, the clinical management of all recruited women was not affected by the cervical measurements. None of the recruited subjects had cerclage.

Obstetric outcome

All subjects were followed until delivery. The gestational age at delivery, mode of onset of labor, obstetric complications (if any) and neonatal outcomes were recorded. For those subjects who subsequently delivered in another hospital, the obstetric information was obtained by telephone from the subject. Preterm delivery was defined as delivery < 37 weeks of gestation. Analysis was also performed for those who delivered before 34 weeks.

Statistics

Statistical analyses were performed using the Statistical Package for the Social Sciences Version 11.0 (SPSS Inc., Chicago IL, USA). Percentiles for cervical length were constructed. The cervical lengths between those with SPD and term deliveries were compared using a Mann–Whitney U-test. A value of P < 0.05 was considered statistically significant. Predictive values of cervical length (continuous variables) and/or the presence of funneling (dichotomous variable) for SPD were then assessed. The sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV) and likelihood ratio (LR) together with the 95% CI for each cut-off were calculated.

The target sample size was 3000. Assuming a 15% drop-out rate and a 5% chance of SPD < 37 weeks, there should be over 100 cases of preterm births in the study cohort for prediction analysis using a receiver–operating characteristics (ROC) curve. Assuming the ultimate estimated area under the curve (AUC) is between 0.6 to 0.8, the standard error (SE) of the estimate AUC will be between 0.027 (for AUC of 0.6) and 0.024 (for AUC of 0.8). The size of this SE is reasonably small, and therefore provides a reasonably precise estimate of the true AUC.

Results

A total of 2952 subjects were recruited and all had cervical length measured on transvaginal ultrasound examination. Twenty-six subjects were later excluded: 23 cases, at the time of scanning, who had a gestational age outside the specified gestation of the study; two had a missed miscarriage within 1 week of scanning; and one woman was not of Chinese ethnic origin. Forty-six (1.6%) additional subjects were lost to follow-up, and hence could not be included due to a lack of information on their obstetric outcome. The cervical length of these 46 subjects was not significantly different from those known to deliver at term (P = 0.23, Mann–Witney U-test). Analysis was performed for the remaining 2880 (97.6%) pregnant women.

The mean age of the study population was 29.6 years (SD 4.9) and 1572 (54.6%) subjects were nulliparous. The mean gestational age at cervical assessment was 20.1 (range, 18.0–22.9; SD 1.0) weeks, and the mean gestational age at delivery was 39.3 (range, 21.3–43.1; SD 1.8) weeks. Two hundred (6.9%) subjects had preterm delivery before 37 weeks of gestation, of whom 106 (3.7%) had SPD and 94 (3.2%) had preterm deliveries induced due to obstetric indications. Of those who had SPD, 19 (0.7%) delivered before 34 weeks and 87 (3.0%) delivered between 34 and 37 weeks. Only one subject had delivery before 24 weeks of gestation. She underwent ultrasound examination at 18.5 weeks of gestation, at which time she was asymptomatic. The cervical length was 10.1 mm, with evidence of funneling. She subsequently presented at 21 weeks of gestation with cervical dilatation and spontaneous rupture of membranes, resulting in spontaneous delivery of an abortus weighing 395 g at 21.3 weeks. This subject was included for analysis.

The mean cervical length was 37.9 (range, 6.2–59.7; SD 6.5) mm. The distribution of the cervical length measurement of all 2880 subjects is shown in Figure 1. The 2.5th, 5th and 10th centiles were 25.7 mm, 27.6 mm and 30.0 mm, respectively. As the number of the SPD group was small and the data were non-Gaussian in distribution, non-parametric statistics were used. The median cervical length of subjects with SPD below 37 weeks was 36.2 (interquartile range (IQR), 31.6–40.8) mm. This was significantly shorter compared with those who delivered at term (median 37.6 mm; IQR, 33.6–42.0) (Mann–Whitney U-test, P = 0.025). The median cervical length of those with SPD below 34 weeks was 32.6 (IQR, 22.8–40.4) mm, which was also significantly shorter compared with the term group (Mann–Whitney U-test, P = 0.006).

Figure 1.

Distribution of cervical length between 18–22 weeks in 2880 singleton pregnancies. Mean 37.9 mm, SD 6.5 mm.

Using ROC curve analysis, cervical length for prediction of SPD < 34 weeks or < 37 weeks was assessed. The AUC for < 34 weeks and < 37 weeks were 0.68 (95% CI: 0.53–0.83, P = 0.007) and 0.56 (95% CI: 0.51–0.62, P = 0.025), respectively. Table 1 shows the sensitivities, specificities, PPV and NPV, and the positive LR of the various cut-offs of cervical length in predicting SPD < 34 weeks. The cut-off of ≤ 27 mm, which corresponded to the 4th centile, gave a false positive rate close to 5%. The sensitivity and specificity were 36.8% and 96.2%, respectively. A cervical length of ≤ 27 mm occurred in 62.5% and 100% of those with SPD < 30 and < 26 weeks, respectively. In the literature, cut-offs of ≤ 25 mm9 and ≤ 15 mm3 have been proposed. In our cohort, a cervical length of ≤ 25 mm occurred in 15.8%, 37.5% and 100% of those with SPD < 34, < 30 and < 26 weeks, respectively. Its predictive performance is summarized in Table 1. Only 2 (0.1%) subjects had a cervical length ≤ 15 mm and both were delivered < 34 weeks.

Table 1. Cervical length at 18–22 weeks of gestation in predicting spontaneous preterm delivery < 34 weeks of gestation
Cervical length (mm)PercentileSensitivity % (95% CI)Specificity % (95% CI)PPV % (95% CI)NPV % (95% CI)Positive LR % (95% CI)
  1. Background prevalence of spontaneous preterm delivery < 34 weeks was 0.7% in the study cohort. LR, likelihood ratio; NPV, negative predictive value; PPV, positive predictive value.

≤ 200.210.5 (0.0–24.3)99.9 (99.8–100.0)40.0 (0.0–82.9)99.4 (99.1–99.7)100.4 (17.7–567.0)
≤ 251.826.3 (6.5–46.1)98.3 (97.9–98.8)9.4 (1.6–17.3)99.5 (99.3–99.8)15.7 (7.0–35.0)
≤ 27436.8 (15.2–58.5)96.2 (95.5–96.9)6.1 (1.7–10.5)99.6 (99.3–99.8)9.8 (5.3–18.1)
≤ 301036.8 (15.2–58.5)90.1 (89.0–91.2)2.4 (0.7–4.2)99.5 (99.3–99.8)3.7 (2.0–6.8)
≤ 353563.2 (41.5–84.9)65.5 (63.8–67.3)1.2 (0.5–1.9)99.6 (99.4–99.9)1.8 (1.3–2.6)

Funneling of the cervix was present in 181 (6.3%) subjects. The medians of the width and depth of the funneling were 6.4 (IQR, 5.2–9.5) mm and 4.3 (IQR, 3.2–5.9) mm, respectively. The median cervical length for this group was 33.1 (IQR, 28.9–37.0) mm, which was significantly shorter compared with those without funneling (Mann–Whitney U-test, P < 0.001). Ten out of these 181 (5.5%) subjects had SPD < 37 weeks of gestation, of whom 6 (3.3%) delivered < 34 weeks of gestation. Subjects with funneling were more likely to have SPD < 34 weeks (Chi-square test, P < 0.001) but not for < 37 weeks (Chi-square test, P = 0.173). Funneling occurred in 31.6%, 50.0% and 66.7% of those with SPD < 34, < 30 and < 26 weeks, respectively.

There was an inverse relationship between the gestational age at delivery and the funnel width (Pearson correlation, P = 0.002), the funnel depth (Pearson correlation, P < 0.001) and the proportion of funneling, defined as (funnel depth/(funnel depth + cervical length) × 100%)9 (Pearson correlation, P < 0.001). In those individuals with funneling, the predictive values of funnel width, depth and the proportion of the funneling were assessed by ROC curves. The proportion of funneling gave the best AUC (0.81, 95% CI: 0.59–1.0, P = 0.01). A cut-off of funneling of > 25% gave a sensitivity of 33.3% detecting SPD < 34 weeks at a specificity of 93.7% within this group.

Using logistic regression, both a short cervix and the presence of funneling were found to be independent predictors for SPD < 34 weeks of gestation. Table 2 shows the sensitivities, specificities, PPV and NPV, and the positive LR of using funneling alone or in combination with cervical length in predicting SPD < 34 weeks of gestation.

Table 2. Combined use of cervical length and funneling at 18–22 weeks for prediction of spontaneous preterm delivery before 34 weeks of gestation
Cervical length/funnelingSensitivity % (95% CI)Specificity % (95% CI)PPV % (95% CI)NPV % (95% CI)Positive LR % (95% CI)
  1. Background prevalence of spontaneous preterm delivery < 34 weeks was 0.7% in the study cohort. LR, likelihood ratio; NPV, negative predictive value; PPV, positive predictive value.

Funneling only31.6 (10.7–52.5)93.9 (93.0–94.8)3.3 (0.7–5.9)99.5 (99.3–99.8)5.2 (2.6–10.2)
> 25% funneling10.5 (0.0–24.3)99.6 (99.4–99.8)15.4 (0.0–35.0)99.4 (99.1–99.7)27.5 (6.5–116.0)
≤ 27 mm + funneling26.3 (6.5–46.1)99.0 (98.6–99.4)14.7 (2.8–26.6)99.5 (99.3–99.8)26.0 (11.3–59.8)
≤ 27 mm or funneling42.1 (19.9–64.3)91.1 (90.1–92.2)3.1 (1.0–5.1)99.6 (99.3–99.8)4.7 (2.8–8.1)
≤ 25 mm + funneling15.8 (0.0–3.2)99.4 (99.1–99.7)15.0 (0.0–30.7)99.5 (99.2–99.7)26.9 (8.6–84.4)
≤ 25 mm or funneling42.1 (19.9–64.3)92.8 (91.9–93.8)3.7 (1.2–6.3)99.6 (99.3–99.8)5.8 (3.4–10.1)

Discussion

Despite the lack of effective measures to prevent preterm labor, identification of individuals at high-risk for SPD remains important for gaining an understanding of the various pathophysiological pathways and for assessment of therapeutic efficacy9. Although preterm labor can be the result of various causes, cervical shortening has consistently been shown to occur prior to the onset of preterm labor. The evaluation of cervical length has potential value particularly among nulliparous women who do not have any obstetric history upon which risk assessment can be based. It can also be used as an additional tool for those with a poor obstetric history where a long cervix and the absence of funneling can provide some reassurance (or vice versa)10, 11.

Our study provides a large dataset on mid-trimester cervical length in an ethnically Chinese population. Although it has been suggested that Chinese women might have a shorter cervix in the third trimester compared to Caucasian women8, our mean cervical length of 37.9 mm appeared to be in accordance with the mean values described in the Caucasian population for the mid trimester, where analysis of two unselected pregnant populations reported mean values of 35–36 mm at 22–24 weeks of gestation1, 3. Progressive shortening of cervical length from 18 weeks of gestation has been reported12, so our data, which were collected mainly at 20 weeks of gestation, may not be directly comparable. A more comparable study involving 3694 Finnish women between 18–22 weeks reported a mean cervical length of 40.7 mm for the term group2. Their 9th centile of 31 mm was very similar to our 10th centile of 30 mm.

Our results confirm previous observations that short cervical length and funneling are predictors of SPD. In particular, they are better predictors of early rather than late preterm birth, which is of greater clinical relevance. A cervical length ≤ 27 mm, which corresponded to the 4th centile in our population, occurred in 36.8%, 62.5% and 100% of those with SPD < 34, < 30 and < 26 weeks, respectively. The positive LR for SPD < 34 weeks was 9.8. The low prevalence of early SPD in our cohort has limited the performance of this screening test, as reflected by the low PPV and large 95% CI. Cervical length measurement at the time of the routine anomaly scan will be of limited clinical value in our situation. However, the high positive LR implies that the test might be useful among the high-risk population. Several recent publications have shown improved performance of mid-trimester cervical length assessment among women with previous spontaneous preterm birth10, 11, 13.

There is some controversy in the literature regarding the significance of funneling as an independent predictor of preterm delivery. In a large cohort of 6819 women, To et al. concluded that funneling did not provide additional contribution to cervical length as a predictor3. However, other groups have reported that the additional feature of funneling has significantly increased the overall risk for preterm birth2, 11. In a recent publication involving 1958 women, the risk of SPD ≤ 34 weeks was 7% for cervical length ≤ 20 mm but it increased to 34% if funneling was present11. The discrepancy might be related to the different cut-offs of cervical length used for evaluation. To et al.3 used a cervical length of ≤ 15 mm, which gives a higher PPV for SPD. It could well be that when the cervix has shortened to ≤ 15 mm, the chance of SPD is so high that the presence or absence of funneling provides much less additional value.

Our data confirmed an association between funneling and a short cervix (using 25 mm or 27 mm as cut-off). Further logistic regression analysis has shown that both variables are independent predictors for SPD < 34 weeks. Among the group with funneling, the proportion of funneling, which is a measure of both the funnel depth and the cervical length, appeared to be the best predictor compared with funnel width or depth alone. It is true that precise measurement of funnel depth can be difficult as the shoulders of the funnel might not be apparent in some cases14. This has certainly limited the clinical application of the ‘proportion of funneling’. However, conceptually, our data have confirmed that the combination of a bigger funnel and a shorter cervix gives rise to a higher PPV for early SPD.

The coexistence of cervical length ≤ 27 mm and funneling gave a 15% PPV for SPD before 34 weeks, with a positive LR of 26. Although the current data suggest that the value of cervical length assessment in our low-risk population is limited, the information on predictive performance of variable cervical length remains useful for designing a study protocol for a high-risk population or intervention trials involving an ethnically Chinese population.

Acknowledgements

We thank our midwifery team in the Fetal Medicine Unit, the Department of Obstetrics and Gynaecology, Prince of Wales Hospital, for their assistance in subject recruitment and data entry, and Dr Jon Hyett, Consultant Obstetrician, King's College Hospital, London, UK for helpful discussion.

Ancillary