About 13% of twins are born before 34 weeks and 7% before 32 weeks. The prediction of preterm birth in twins is based on the same tests as in singleton pregnancies. In twin pregnancies, the cut-off for short cervix at the second trimester scan is less than 25 mm (compared with 15 mm in singletons); length less than 20 mm is associated with 42% risk for birth before 32 weeks and cervical length less than 25 mm is associated with 28% risk for birth before 28 weeks. The measurement of cervical length in pregnancies with symptoms of preterm labour may have limited accuracy in predicting preterm birth. In asymptomatic women, a positive fetal fibronectin test seems to be associated with 35% risk for birth before 32 weeks and 40% risk for birth less than 34 weeks, whereas a negative test decreases the risk to 6% and 17%, respectively. The differences in the predictive value of tests between twins and singletons reflect the diverse pathophysiology of preterm birth between the two groups.
Preterm birth in twins
Prematurity is one of the leading causes of perinatal mortality, morbidity, and long-term neurodevelopmental impairment. A child born at 25 weeks has 50% risk for death within the first few months of life, and about 50% of the survivors are expected to have moderate-to-severe handicap as infants . Both survival and neurological prognosis improve with advancing gestational age, and children born at 32 weeks have a 97% survival rate , with only 4.4% risk for cerebral palsy at the age of 2 years and 8% risk for neurodevelopmental delay .
Although up to 12% of children may be born before 37 weeks , the rate of birth before 35 weeks in singleton pregnancies is about 4% , and the rates of spontaneous births before 34 and before 32 weeks are 1.1% and 0.6%, respectively . The risk of prematurity is much higher in twins. In a prospective study of 800 dichorionic and 200 monochorionic twin pregnancies, 29% of dichorionic pregnancies developed a maternal or fetal condition, leading to birth before 36 weeks, whereas 34% of the monochorionic twins were born before 34 weeks . In a recent meta-analysis, the pooled rates of birth before 37, 34 and 32 weeks for twins were 41%, 13% and 7%, respectively .
During the past 15 years, significant progress has been achieved in the prediction and prevention of preterm birth in singleton pregnancies. This process has been slower in twins, apparently reflecting the diverse pathophysiology of preterm birth in multiple pregnancy.
Pathophysiology and demographics
Preterm birth is a syndrome because of its multifactorial aetiology and diverse clinical manifestations, and this has implications in its understanding, prediction, and prevention. The most widely accepted contributing factors for prematurity include (1) intrauterine infection or inflammation; (2) uterine ischaemia; (3) uterine overdistension; (4) abnormal allograft reaction; (5) allergy; (6) cervical insufficiency; and (7) hormonal disorders (related to progesterone and corticotrophin-releasing factors) . Although over-distension would be the obvious candidate mechanism in the case of twins, it seems that multiple gestations are also a heterogeneous group with different pathophysiological pathways , which is also supported by the decreased effectiveness of interventions that are commonly successful in singletons. It has been suggested that all mechanisms that participate in the initiation of term labour and preterm birth in singletons, may be present in a ‘supraphysiological’ degree in multiple gestations .
In singleton pregnancies, the risk of prematurity is increased in women who had previous late miscarriages or preterm births, those of Afro–Caribbean origin, teenagers, those will low body mass index, and cigarette smokers . None of these risk factors was found to be significant in twins , and obstetric history became a significant predictor only after cervical length was excluded . Moreover, the use of assisted reproduction techniques was found to be associated with preterm birth only in overweight or obese women .
The effect of chorionicity
Chorionicity is a unique factor in twins, which also contributes to the risk of prematurity. Monochorionic diamniotic (MCDA) twins are at higher risk compared with dichorionic (DCDA) twins for preterm birth and prenatal complications. In a historical cohort of 1407 twin pregnancies , the rate of birth before 28 weeks was 11% for MCDA compared with 7% for DCDA twins, and the rates for birth before 32 weeks were 26% compared with 18%, respectively. Monochorionic diamniotic twins were also at increased risk for fetal death (even at term), necrotising enterocolitis, and neuromorbidity. A significant contributor of prematurity and morbidity in MCDA twins is twin-to-twin transfusion syndrome (TTTS). In a recent multicentre study of about 10,000 twin pregnancies , the rate of birth before 34 weeks was 24% in the MCDA group compared with 16% in the DCDA twin group. The rates of preterm premature rupture of membranes and pregnancy-induced hypertension, however, were similar between the two types of twins, and most of the difference in the rates of prematurity could be attributed to TTTS in MCDA twins, which was 7.3% before 37 weeks.
Early markers increasing the risk for development of TTTS later in pregnancy include inter-twin discrepancy in the nuchal translucency test of 20% or more (30% risk for TTTS) and presence of at least one abnormal blood flow waveform in the ductus venosus .
When TTTS is excluded, there seems to be no difference in median cervical length between monochorionic and dichorionic twins , and the risk for early preterm birth does not significantly differ between the two groups .
Prediction of preterm birth in asymptomatic twin pregnancies
The two most commonly used screening tests for the prediction of preterm birth in singletons are cervical length measurement and testing for fetal fibronectin in the cervicovaginal fluid at the time of the anomaly scan (21–24 weeks), and the same tests have been tried in twins.
Cervical length
The rationale for measuring cervical length is that the mechanism of normal human labour requires cervical remodelling, in which the cervix of the uterus gradually ripens, shortens (effaces), and finally dilates. Normally, cervical length shows a continuous linear reduction between 10 and 40 weeks. For example, the median cervical length in singleton pregnancies is 36 mm at 23 weeks, which then decreases to 33 mm at 28 weeks, and then to 29 mm at 34 weeks; the corresponding values for the 5th centile are 20 mm, 17 mm and 10 mm, respectively . In twin pregnancies, the median cervical length at the time of the anomaly scan (22–24 weeks) is similar to that of singletons, but a higher proportion of twins have cervical length less than 25 mm (12.9% v 8.4% in singletons) and less than 15 mm (4.5% v 1.5%) .
Evidence shows that the initiators of cervical remodelling may differ between preterm and term birth, with a distinct role for complement activation and macrophages in the former case; however, the final steps of the pathway are common, involving release of matrix metalloproteinases, collagen degradation, and increased cervical distensibility .
In singleton pregnancies, screening with cervical length and demographic factors at the time of the second-trimester scan can predict 69% of cases that will deliver before 32 weeks for a false–positive rate of 10% . Cervical length less than 15 mm (which corresponds to the 1.5th centile in singletons) predicts 58% of the cases who will deliver before 32 weeks, and, conversely, it is associated with 4% risk for birth before 32 weeks .
In contrast to singleton pregnancies, where the risk for preterm birth increases exponentially when the cervix is less than 15 mm, the cut-off for high risk in twin pregnancies is 25 mm, apparently because a longer cervix is needed to counteract the increased uterine activity during pregnancy . Cervical length of 25 mm or less at 22–24 weeks was reported in 100% of women delivering before 28 weeks , 80% of those delivering before 30 weeks , and about 50% of those delivering before 28 weeks . Cervical length less than 25 mm is associated with 10% and cervical length less than 15 mm is associated with 30% risk for birth before 33 weeks .
In the largest single study on asymptomatic twins ( n = 1163) , cervical length, previous obstetric history, cervical surgery, and cigarette smoking were all found to be associated with preterm birth; however, when the independent contribution of these factors was tested in a multiple logistic regression model, only cervical length remained a significant predictor.
A recent meta-analysis summarised data from 16 studies ( n = 3213) on asymptomatic women with twin pregnancy at the second trimester . It was found that the most accurate cut-off for the prediction of birth before 28 weeks was cervical length 25 mm or less (area under the curve 0.86), whereas cervical length 20 mm or less and cervical length 25 mm or less were the most accurate predictors for birth before 32 weeks (area under curve 0.80 for both), and cervical length 20 mm or less and 35 mm or less were the best predictors for birth before 34 weeks. In asymptomatic women, a cervical length of 20 mm or less at 20–24 weeks was associated with 42% risk for birth before 32 weeks and 62% risk for birth before 34 weeks. A cervical length of 25 mm or less was associated with 26% risk for birth before 28 weeks; in contrast, a longer cervix was associated with only 1.4% risk for birth before 28 weeks and about 65% chance for term birth .
From a qualitative point of view, the accuracy of cervical length measurement may be slightly better for birth before 30 weeks than after 30 weeks, and overall accuracy of cervical length measured after 24 weeks may be slightly better than measurements before 24 weeks . The latter result, however, has not been replicated in the meta-analysis of Conde-Agudelo et al. , and it is unlikely that these differences have actual clinical significance. Both meta-analyses highlight the profound heterogeneity of data for many of their analyses. The pooled risk for preterm birth according to mid-trimester cervical length in asymptomatic women is shown in Table 1 .
| Pooled risk % (95% CI) | ||||
|---|---|---|---|---|
| Birth before 28 weeks | Birth before 30 weeks | Birth before 32 weeks | Birth before 34 weeks | |
| Cervical length less than 15 mm | 44.4 (18.9 to 73.3) | 50.0 (35.2 to 64.8) | 72.5 (57.2 to 83.9) | 75.6 (60.7 to 86.2) |
| Cervical length less than 20 mm | 34.4 (20.4 to 51.7) | 24.8 (17.9 to 33.3) | 19.9 (16.0 to 24.4) | 55.6 (46.5 to 64.4) |
| Cervical length less than 25 mm | 12.6 (7.4 to 20.8) | 20.7 (15.4 to 27.1) | 27.7 (22.8 to 33.3) | 51.2 (43.7 to 58.6) |
| Cervical length less than 30 mm | 3.9 (1.5 to 9.7) | 10.1 (7.3 to 13.8) | 26.5 (21.8 to 31.7) | 66.3 (58.6 to 73.1) |
Fewer studies have assessed the potential value of follow-up cervical length measurements. In a study of 209 asymptomatic women, cervical length shortening by more than 25% over 3–5 weeks was associated with an over seven-fold increase in the risk of birth before 32 weeks; in absolute rates, the rate of birth less than 32 weeks in these women was 29% compared with 4% in women with less shortening . When examining only women with initially long cervix (>25 mm), another study reported that shortening of over 13% 4–5 weeks after the initial scan, or cervical length less than 30 mm at the repeat scan, were statistically significant predictors of birth at 32 weeks or earlier; however, the actual odds ratios were only about 1.1 .
Cervical length in the first trimester
Recently, attention has shifted towards the potential role of first-trimester cervical length measurement in the prediction of preterm birth. Studies in singleton pregnancies have shown that the cervix changes minimally in most women between 11 and 24 weeks, but shortening tends to be more prominent in women with a history of cervical surgery or preterm delivery, and the median cervical length in the first trimester is significantly lower in those who will deliver preterm . The results on the potential application of first-trimester cervical screening for the early prediction of preterm birth in an unselected population of singletons are conflicting , and no studies have yet been conducted in twins.
Biochemical markers: fetal fibronectin
Fetal fibronectin (fFN) is a glycoprotein that is produced by the extravillous trophoblasts, and acts much like an adhesive substance at the interface between the chorion and the decidua . Although it can normally test positive during the first half of pregnancy, its detection in the cervicovaginal fluid between 22 and 34 weeks may indicate mechanical separation or inflammatory-mediated extracellular matter distortion at the choriodecidual interface , and has been associated with increased risk for preterm birth.
In asymptomatic women with a singleton pregnancy, a positive fFN increases the risk for birth before 34 weeks by 7.6 times, whereas a negative result decreases this risk by a factor of 0.8 .
Cervicovaginal fFN can be measured in twins in the same way as in singletons. In one of the first studies on twins, a positive fFN result was found in about 5% of women.
A meta-analysis conducted in 2010 pooled data from 11 studies on asymptomatic twins, in which fFN was tested (mostly) during the second trimester. In general, fFN showed limited accuracy in the prediction of preterm birth, as the pooled likelihood ratios did not significantly alter the pre-test probabilities, especially for outcomes other than birth less than 32 weeks. The pooled sensitivity for birth before 32 weeks and before 34 weeks was 35% and 45%, respectively, for 6% and 19% false positive rates, respectively. On the basis of the data of the particular populations described in this meta-analysis, a positive fFN test was associated with 34% risk for birth before 32 weeks and 42% risk for birth before 34 weeks, whereas a negative test decreased the risk to 6% and 17%, respectively. The investigators, however, highlight the high heterogeneity across studies .
Few studies have directly compared cervical length and fFN in twins. In a retrospective study of 155 asymptomatic twin pregnancies in which simultaneous cervical length and fFN screening was offered between 22 and 32 weeks , the combination of cervical length less than 20 mm and positive fFN had significantly higher positive predictive value than either of the tests alone, and was associated with about 55% risk for birth before 32 weeks. The prospective Preterm Prediction Study evaluated multiple potential risk factors for preterm birth and concluded that, for birth before 32 weeks, cervical length 25 mm or less was the best predictor at 24 weeks, whereas positive fFN was the best predictor at 28 weeks .
Other biochemical markers
Less frequently studied biochemical markers in singleton pregnancies include cervical phosphorylated IGFBP-1, cytokines in the cervicovaginal fluid, adhesion molecules, PAPP-A, beta-human chorionic gonadotropin, and proteases. The data on twins are even more limited. A small retrospective study of 70 twin pregnancies reported a five-fold risk for birth before 32 weeks when first-trimester free beta-human chorionic gonadotropin was below the 25th centile . Recently, a secondary analysis of a randomised-control trial on progesterone for the prevention of preterm birth in twins reported that, in placebo-treated women, interleukin-8 levels were increased in women who delivered before 34 weeks, and the risk for preterm birth increased with a large weekly increase in interleukin-8 .
Prediction of preterm birth in asymptomatic twin pregnancies
The two most commonly used screening tests for the prediction of preterm birth in singletons are cervical length measurement and testing for fetal fibronectin in the cervicovaginal fluid at the time of the anomaly scan (21–24 weeks), and the same tests have been tried in twins.
Cervical length
The rationale for measuring cervical length is that the mechanism of normal human labour requires cervical remodelling, in which the cervix of the uterus gradually ripens, shortens (effaces), and finally dilates. Normally, cervical length shows a continuous linear reduction between 10 and 40 weeks. For example, the median cervical length in singleton pregnancies is 36 mm at 23 weeks, which then decreases to 33 mm at 28 weeks, and then to 29 mm at 34 weeks; the corresponding values for the 5th centile are 20 mm, 17 mm and 10 mm, respectively . In twin pregnancies, the median cervical length at the time of the anomaly scan (22–24 weeks) is similar to that of singletons, but a higher proportion of twins have cervical length less than 25 mm (12.9% v 8.4% in singletons) and less than 15 mm (4.5% v 1.5%) .
Evidence shows that the initiators of cervical remodelling may differ between preterm and term birth, with a distinct role for complement activation and macrophages in the former case; however, the final steps of the pathway are common, involving release of matrix metalloproteinases, collagen degradation, and increased cervical distensibility .
In singleton pregnancies, screening with cervical length and demographic factors at the time of the second-trimester scan can predict 69% of cases that will deliver before 32 weeks for a false–positive rate of 10% . Cervical length less than 15 mm (which corresponds to the 1.5th centile in singletons) predicts 58% of the cases who will deliver before 32 weeks, and, conversely, it is associated with 4% risk for birth before 32 weeks .
In contrast to singleton pregnancies, where the risk for preterm birth increases exponentially when the cervix is less than 15 mm, the cut-off for high risk in twin pregnancies is 25 mm, apparently because a longer cervix is needed to counteract the increased uterine activity during pregnancy . Cervical length of 25 mm or less at 22–24 weeks was reported in 100% of women delivering before 28 weeks , 80% of those delivering before 30 weeks , and about 50% of those delivering before 28 weeks . Cervical length less than 25 mm is associated with 10% and cervical length less than 15 mm is associated with 30% risk for birth before 33 weeks .
In the largest single study on asymptomatic twins ( n = 1163) , cervical length, previous obstetric history, cervical surgery, and cigarette smoking were all found to be associated with preterm birth; however, when the independent contribution of these factors was tested in a multiple logistic regression model, only cervical length remained a significant predictor.
A recent meta-analysis summarised data from 16 studies ( n = 3213) on asymptomatic women with twin pregnancy at the second trimester . It was found that the most accurate cut-off for the prediction of birth before 28 weeks was cervical length 25 mm or less (area under the curve 0.86), whereas cervical length 20 mm or less and cervical length 25 mm or less were the most accurate predictors for birth before 32 weeks (area under curve 0.80 for both), and cervical length 20 mm or less and 35 mm or less were the best predictors for birth before 34 weeks. In asymptomatic women, a cervical length of 20 mm or less at 20–24 weeks was associated with 42% risk for birth before 32 weeks and 62% risk for birth before 34 weeks. A cervical length of 25 mm or less was associated with 26% risk for birth before 28 weeks; in contrast, a longer cervix was associated with only 1.4% risk for birth before 28 weeks and about 65% chance for term birth .
From a qualitative point of view, the accuracy of cervical length measurement may be slightly better for birth before 30 weeks than after 30 weeks, and overall accuracy of cervical length measured after 24 weeks may be slightly better than measurements before 24 weeks . The latter result, however, has not been replicated in the meta-analysis of Conde-Agudelo et al. , and it is unlikely that these differences have actual clinical significance. Both meta-analyses highlight the profound heterogeneity of data for many of their analyses. The pooled risk for preterm birth according to mid-trimester cervical length in asymptomatic women is shown in Table 1 .
