Cervical length twins.pdf

INTRODUCTION

Despite the large sum of money -estimated at 1.8 million pounds sterling by James in 2014 1 -invested in research into preterm birth (PTB), PTB remains a major problem in obstetrics, representing the second largest direct cause of child death during the first 5 years. It is also associated with a high morbidity rate and lifelong sequelae 2 . Of the spectrum of long-term neurological problems attributable to PTB, 50% occur in infants born before 32 weeks' gestation 2,3 .

The rate of PTB has increased in recent decades, partly owing to the lowering of the gestational age at which a neonate is considered viable, and partly because of the increased proportion of twin births, the latter being a consequence of the increase in both maternal age at conception and access to reproductive technology and treatment 4,5 . According to data from the National Center for Health Statistics, the number of twin births in the USA doubled from the 1980s to 2009, reaching 137 000 births per year, accounting for 3.3% of all births 6 and having a tremendous impact on the burden of prematurity. According to epidemiological studies, 40% of multiple gestations will deliver at or before 37 weeks' gestation, and twins are between four and five times more likely to deliver before 32 weeks than are singletons 2,3,7 .

To date, different strategies have been proposed for the prediction of PTB in twin gestations, and short cervical length (CL) in mid-gestation is considered predictive for spontaneous PTB. A recent meta-analysis has suggested a CL of 20 mm as the cut-off for the prediction of PTB < 32 weeks in twin gestations 8 . The aim of this study was to assess the predictive value of CL for spontaneous PTB < 32 weeks' gestation in twin pregnancies and to calculate the cut-off point that has the best sensitivity and specificity.

METHODS

This was a retrospective cohort study on twin pregnancies followed from the first trimester in a tertiary fetal medicine unit between June 2001 and December 2013.

The ultrasound electronic database was searched in order to identify twin pregnancies with two live fetuses, with a first scan performed before 16 + 0 weeks' gestation. Results were matched to the computerized maternity records; pregnancy notes were reviewed in cases of missing data. Exclusion criteria were: at least one fetus with structural or chromosomal abnormalities; higher order pregnancy; monoamniotic pregnancy; pregnancy referred after 16 weeks; cases complicated by twin-to-twin transfusion syndrome (TTTS) or those that required intrauterine therapy; pregnancies with indicated PTB; or follow-up data unavailable. As this was a retrospective audit of clinical data presented in anonymized form, no ethics committee approval was necessary, according to national regulations.

Pregnancies were dated according to the crown-rump length of the larger twin in the first trimester or by the date of the last menstrual period in women with regular cycles without antecedent oral contraceptive use, unless there was a discrepancy of more than 7 days between menstrual dating and ultrasonographic assessment 9 .

Chorionicity was determined using lambda and T signs, as appropriate 10 . Monochorionic-diamniotic (MCDA) and dichorionic-diamniotic (DCDA) twin pregnancies were followed up every 2 and 4 weeks, respectively; if any complication occurred, the frequency of examinations was increased as necessary. According to the local protocol, CL was measured on transvaginal ultrasound between 18 + 0 and 23 + 0 weeks' gestation using a standardized technique 11 -13 ; the patient had an empty bladder, which was confirmed after Valsalva maneuver. When the cervix was dilated, and no closed endocervical canal was present, the length of the remaining cervix surrounding the protruding membranes was measured between the internal and external ora. Cervical cerclage or Arabin's pessary placement were offered after individualized counseling in case of a CL ≤ 20 mm or cervical dilatation (at any stage) with membranes at or beyond the external cervical os on clinical examination 12 . According to the local protocol, uncomplicated MCDA and DCDA twin pregnancies were delivered electively at 35-36 and 37-38 weeks' gestation, respectively. For the purpose of this study, PTB was defined as spontaneous onset of labor and subsequent delivery of at least one live fetus between 24 and 32 weeks' gestation, regardless of the mode of delivery.

Statistical analysis

Data distribution was assessed according to the Kolmogorov-Smirnov test of normality. Data were expressed as mean (± SD), or median and interquartile range, for normally and non-normally distributed data, respectively. Categorical variables were described as n (%). Pearson's chi-square test was used to analyze categorical variables. The independent t-test and Mann-Whitney U-test were used to compare continuous variables, as appropriate.

Receiver-operating characteristics (ROC) curves were used to assess the CL value predictive for PTB < 32 weeks' gestation. The optimal cut-off point and the distribution of CL measurements in patients with PTB were calculated. Sensitivity, specificity, negative predictive value (NPV), positive predictive value (PPV) and accuracy in predicting PTB < 32 weeks' gestation were calculated for the best cut-off point.

Logistic regression analysis was performed to assess the association between pregnancy characteristics (maternal age, parity, smoking, chorionicity, CL, cervical procedures) and PTB < 32 weeks' gestation. All P-values were two-tailed and P < 0.05 was considered statistically significant. Statistical analysis was performed using SPSS 20.0 (SPSS Inc., Chicago, IL, USA) and GraphPad Prism (GraphPad Software, San Diego, CA, USA) statistical software.

RESULTS

During the study period, 940 twin pregnancies met the inclusion criteria; their demographic and pregnancy characteristics are presented in Table 1. CL showed a non-normal distribution. Forty-one cervical procedures were performed in these pregnancies, of which cervical cerclage was performed in 37 (3.9%) and Arabin's pessary was placed in four (0.4%). There were 78 (8.3%) pregnancies with PTB < 32 weeks' gestation. The demographic and pregnancy characteristics of women who delivered before and those who delivered at or after 32 weeks are also given in Table 1. A box plot comparing CL in these two groups is provided in Figure 1. Compared with those who delivered at or after 32 weeks, cases with PTB < 32 weeks were more likely to be nulliparous, had shorter CL in mid-gestation, earlier gestational age at delivery and gave birth to infants with a lower birth weight. The prevalence of PTB < 32 weeks was significantly higher for patients with cervical cerclage or Arabin's pessary placement than in untreated cases (11/41 vs 67/899; P < 0.01). A comparison between cases with and without cervical cerclage or Arabin's pessary placement is provided in Table S1. However, when included in logistic regression analysis, cervical procedures (cerclage and Arabin's pessary) were not related per se to PTB < 32 weeks ( Table 2). The results of logistic regression analysis are provided in Table 2. When adjusted for confounders, only CL (adjusted odds ratio (aOR), 0.94 (95% CI, 0.90-0.99); P = 0.03) was found to Table 1 Demographic and pregnancy characteristics of 940 women with twin pregnancy and cervical length (CL) assessment in mid-gestation, according to whether they delivered < 32 weeks or ≥ 32 weeks Data are expressed as median (interquartile range) or n (%). *Comparison between women who delivered before and women who delivered at or after 32 weeks. BMI, body mass index; GA, gestational age. Box-and-whisker plots of cervical length measurements in mid-gestation in women with twin pregnancy, according to whether they delivered < 32 weeks or ≥ 32 weeks. Boxes and internal line indicate the median and lower and upper quartiles, whiskers show the range with 1.5× the interquartile range (IQR) below and above the lower and upper quartiles, respectively, and outliers more than 1.5×IQR ( ) or 3.0×IQR ( ) below and above the lower and upper quartiles are also shown. On ROC curve analysis, CL showed an area under the curve of 0.65 (95% CI, 0.58-0.71) for the prediction of PTB < 32 weeks' gestation ( Figure 2). The optimal cut-off value was 36 mm (sensitivity, 64.1%; specificity, 62.8%; PPV, 13.5%; NPV, 95.1%; accuracy, 62.9%; Table S2). The relative risk of PTB for a CL ≤ 36 mm, compared to a CL > 36 mm, was 2.35 (95% CI, 1.53-3.60); (P < 0.001). Figure 3 shows the distribution of CL values in cases complicated by PTB < 32 weeks. Considering 36 mm as the cut-off, 35.9% (28/78) of cases of PTB < 32 weeks had a normal CL in mid-gestation. Considering a 20 mm cut-off, as reported in the literature 8 , 85% of women who subsequently delivered < 32 weeks' gestation had a normal CL in mid-gestation.

DISCUSSION

Main findings

We have shown that CL measured in mid-gestation was significantly lower in women who subsequently delivered < 32 weeks' gestation. However, CL values overlapped between cases that delivered before and those that delivered after 32 weeks. On ROC curve analysis, the ability of CL to predict PTB < 32 weeks was poor. The calculated optimal cut-off point of 36 mm showed poor sensitivity and specificity, low PPV, high NPV and low accuracy for this prediction. More than one-third (36%) of PTB cases delivering < 32 weeks had a normal CL in mid-gestation. After adjusting for confounders on logistic regression analysis, only CL showed a significant, although weak, independent association with PTB < 32 weeks.

Strengths and limitations of the study

The main strengths of this study are the large number of pregnancies, standardized transvaginal assessment and exclusion of cases with TTTS, intrauterine therapy (e.g. for twin anemia-polycythemia sequence) and indicated PTB. Another strength was adjusting the logistic regression analysis for confounders.

The main limitation of this study is the retrospective design; however the consistent and comprehensive data gathering mitigates this. Another limitation is represented by the lack of data on previous history of cervical excision; nevertheless, it has been shown that the increased risk for PTB among patients with previous cervical excision is reflected in the measurement of CL in mid-gestation 14 . It can be argued that the results could have been biased by the inclusion of cases that underwent cervical procedures. However, cervical procedures were not significantly correlated per se to PTB < 32 weeks on logistic regression analysis. Moreover, given that the local protocol offers a cervical procedure only to cases with a CL < 20 mm at mid-gestation or with cervical dilatation, exclusion of these cases would have excluded more severe cases, underestimating the predictive value of CL.

Interpretation

To date, there is evidence that both term and preterm labor are linked to an inflammatory process, and over the last few decades several biomarkers, including inflammatory cytokines, have been investigated for the prediction of PTB 1,15 -21 . Recently, Abbot et al. 19 have shown that inflammation anticipates cervical shortening, thus questioning the utility of CL screening in asymptomatic women.

Currently, CL and cervicovaginal fluid fetal fibronectin are used in clinical practice to assess the risk of spontaneous PTB; however their usefulness is limited by their weak positive predictive power 16,22 . Previous studies investigating the predictive ability of CL at mid-gestation to predict PTB < 32 weeks' gestation in asymptomatic twin pregnancies are reported in Table S3 23 -30 . A recent meta-analysis by Conde-Agudelo et al. 8 investigated the predictive role of CL for PTB in twin gestations; this review compared different cut-offs for the prediction of PTB at different gestational ages. When PTB was defined as birth < 32 weeks, the authors found that 20 mm was the best cut-off. However, this meta-analysis had some limitations, mainly owing to the small number of studies included and to the studies' heterogeneity; only seven studies considered the predictive ability of CL for PTB < 32 weeks, among these only five assessed the predictive value of 20 mm and two assessed CL between 22 and 32 weeks' gestation. As the authors stated, 'a normal CL was less accurate in predicting the absence of PTB' 8 .

The findings of the present study are in keeping with this and raise a question about the utility of CL screening in asymptomatic women. By definition, the purpose of a screening test is to identify a disease at preclinical stages, thus a high sensitivity is advisable. The CL values observed in the present study are in keeping with those given in the medical literature at comparable gestational ages 30 . Although significantly different, CL values overlapped in cases that delivered before and those that delivered after 32 weeks; consequently the predictive ability of CL was poor, and the optimal cut-off of 36 mm calculated from the ROC curve is usually considered physiological in clinical practice. Even when considering the cut-offs suggested by the literature the detection rate was poor. This is not dissimilar to the low sensitivity of CL screening in singleton pregnancies; using the 15-mm CL cut-off that is often adopted, approximately 40% of cases of singleton PTB < 32 weeks will not be detected 31,32 .

Conclusions

This study shows that, despite the weak independent association observed, CL assessed in mid-gestation is a poor predictor of PTB < 32 weeks' gestation in asymptomatic twin pregnancies. Even if interventions such as vaginal progesterone or Arabin's pessary prove able to improve perinatal outcomes in women with a short cervix 33,34 , it must be remembered that a significant proportion of PTB will never be detected by such screen-and-treat approaches.

SUPPORTING INFORMATION ON THE INTERNET

The following supporting information may be found in the online version of this article:

Table S1

Demographic and pregnancy characteristics in 940 women with twin pregnancy undergoing cervical-length assessment in mid-gestation, grouped according to placement of cervical cerclage or pessary Table S2 Two-by-two table showing data from which the ability of cervical length ≤ 36 mm to detect preterm delivery before 32 weeks' gestation was calculated Table S3 Previous studies investigating ability of cervical length in mid-gestation to predict preterm birth < 32 weeks' gestation in asymptomatic twin pregnancies