Dr KEA Hack, Department of Obstetrics, Wilhelmina Children’s Hospital, University Medical Centre Utrecht, K.J. 02.507.0, PO Box 85090, 3508 AB Utrecht, the Netherlands. Email email@example.com
Objective To evaluate mortality and morbidity in a large cohort of twin pregnancies according to chorionicity. We aimed to estimate the optimal time of delivery.
Design Historical cohort design.
Setting Two teaching hospitals.
Population Twin pregnancies delivered in the University Medical Centre, Utrecht, and the St Elisabeth Hospital, Tilburg (1995–2004), The Netherlands (n = 1407).
Methods Pregnancy outcomes were documented according to chorionicity. Mortality ≥32 weeks was reviewed carefully with special attention to antenatal fetal monitoring, autopsy and placental histopathology to find an explanation for adverse outcome.
Main outcome measures Perinatal mortality and morbidity in monochorionic (MC) and dichorionic (DC) twins.
Results Perinatal mortality was 11.6% in MC twin pregnancies and 5.0% in DC twin pregnancies. After 32 weeks, the risk of intrauterine death (IUD) was significantly higher in MC twins than in DC twins (hazard ratio 8.8, 95% CI 2.7–28.9). In most of these cases of IUD, no antenatal signs of impaired fetal condition had been present. Median gestational age was 1 week longer in DC twins than in MC twins, and the mean birthweight was 221 g higher. Severe birthweight discordancy (>20%) occurred more often in MC twins than in DC twins (OR 1.23, 95% CI 0.97–1.55). The incidence of necrotising enterocolitis (NEC) was higher in MC twins, after adjustment for age and weight at birth (OR 4.05, 95% CI 1.97–8.35). There was a trend towards higher neuromorbidity in MC twins.
Conclusions This is the largest cohort study of twin pregnancies evaluating outcome according to chorionicity thus far. MC twins are at increased risk for fetal death (even at term), NEC and neuromorbidity. Current antenatal care is insufficient to predict and prevent this excess perinatal mortality and morbidity. Planned delivery at or even before 37 weeks of gestation seems to be justified for MC twins.
Perinatal mortality and morbidity is 3–7 times higher in twin pregnancies compared with singleton pregnancies. High mortality and morbidity rates are due to a higher incidence of antepartum complications, preterm birth and uteroplacental insufficiency.1 Therefore, close fetal surveillance of twins is warranted, irrespective of chorionicity.1 Monochorionic (MC) twins are at increased risk for perinatal mortality and morbidity compared with dichorionic (DC) twins due to haemodynamic imbalance caused by vascular anastomoses between both sides of the placenta. Monochorionicity is associated with a higher perinatal mortality2–5 and with increased incidence of preterm birth,4,6 low birthweight4,5 and prolonged stay in the neonatal intensive care unit (NICU).6 Following single-twin death, perinatal mortality and morbidity for the surviving co-twin is higher in MC twins than in DC twins.7 Consequences to the surviving co-twin can be profound, including co-twin death, survival with cerebral impairment or preterm delivery with its sequelae.7,8
Despite fetal surveillance and the possibility of early intervention, mortality in uncomplicated MC twin pregnancies cannot easily be foreseen. Even at term, unexpected fetal demise occurs.9 This has led to the discussion as to how to manage the uncomplicated (MC) twin pregnancy at term.9–13 There is a controversy in the literature about intrauterine death (IUD) rates after 32 weeks of gestation with subsequent varying recommendations of optimal time of delivery.9,14,15 We therefore studied a large cohort of twin pregnancies to assess perinatal mortality and neonatal morbidity, according to chorionicity, with special emphasis towards adverse outcome after 32 weeks of gestation. Our aim was to identify risk factors for adverse outcome and to estimate the optimal time of delivery. To the best of our knowledge, this is the largest cohort of twin pregnancies that has ever been studied according to chorionicity.
Materials and methods
The medical records of all twin pregnancies delivered at the University Medical Centre, Utrecht, a tertiary referral centre, and the St Elisabeth Hospital, Tilburg, a large general teaching hospital, both in The Netherlands, were reviewed for the period January 1995 until December 2004. All twin pregnancies were retrieved from the electronical database of both departments. During this 10-year study period, 1407 sets of twins were delivered in the two participating hospitals. There were no missing data. Baseline characteristics, neonatal and maternal outcomes were documented according to chorionicity. Chorionicity was determined on the basis of first-trimester ultrasound assessment of the dividing membrane characteristics and/or postpartum pathological examination of placentas and intertwin membranes. Pregnancies referred to the University Medical Centre, Utrecht, before 32 weeks with a delivery in this hospital were included in analysis.
We excluded pregnancies with unknown chorionicity (n = 50), monoamnionicity (n = 18), selective fetal reduction to singleton pregnancies because of Down syndrome or anencephaly (n = 3), pregnancy loss <20 weeks (n = 14), first-trimester termination for congenital malformations or early twin-twin transfusion syndrome (TTTS) (n = 2) and major lethal chromosomal and/or congenital malformations (n = 15). Of the remaining 1305 twin pregnancies, 198 were MC and 1107 were DC.
One thousand and fifty-three of 1305 pregnancies delivered after 32 weeks of gestation. Pregnancies resulting in intrauterine demise or neonatal death were reviewed carefully with special attention to the adherence to the standard protocol for fetal monitoring, autopsy and placental histopathology to find an explanation for the adverse outcome. The standard protocol for the monitoring of twin pregnancies consisted of routine first-trimester ultrasound with determination of chorionicity, a detailed anomaly scan at 20 weeks of gestation for MC twin pregnancies and fortnightly scans for growth, amniotic fluid and Doppler assessments thereafter. In contrast to DC twin pregnancies in which spontaneous onset of delivery was awaited if the course of the pregnancy was uncomplicated, elective delivery of uncomplicated MC twin pregnancies was planned at around 37–38 weeks of gestation. This difference in management policy gradually evolved over the past 5 years based on recent literature about risks continuing into later pregnancy.
Perinatal mortality was defined as death of an infant >500 g within 1 week after delivery, including stillbirth.16 Stillbirth was defined as intrauterine demise of a fetus >500 g and/or ≥20 completed weeks of gestation. Gestational age at time of IUD was ascertained by ultrasound and was accounted for in the analysis. Early neonatal mortality was defined as death of an infant during the first 7 days of life and late neonatal death as death between 8 and 28 days after birth. Mortality rates were also stratified to week of gestation, that is the number of infants that died divided by total number of infants born that week, expressed as percentage.
Admission to the neonatal ward (medium care, high care or intensive care) was used as indicator for neonatal morbidity, as well as number of admission days. Neonatal morbidity was further defined as the occurrence of respiratory distress syndrome (RDS), bronchopulmonary dysplasia, intraventricular haemorrhage (IVH), periventricular leucomalacia (PVL), necrotising enterocolitis (NEC), neonatal icterus, a persistent ductus arteriosis requiring indomethacin therapy or surgical closure, sepsis and hypoglycaemia (defined as a random glucose <2.6 mmol/l). Asphyxia was defined as a 5-minute Apgar score ≤5 combined with an umbilical artery pH < 7.00. Small for gestational age (SGA) was defined as a birthweight below the 10th percentile,17 adjusted for gestational age, sex and parity.
Preterm delivery was defined as delivery before 37 completed weeks of gestation and very preterm delivery as delivery before 32 completed weeks.
The diagnosis of TTTS was made antenatally by standard prenatal ultrasound criteria: (1) monochorionicity established by absence of a ‘twin peak’ sign and presence of a thin dividing membrane early in pregnancy, (2) oligohydramnios (deepest vertical pocket ≤ 2 cm) in the twin sac of one fetus and (3) polyhydramnios (deepest vertical pocket ≥ 8 cm before 20 weeks of gestation or ≥ 10 cm after 20 weeks of gestation) in the twin sac of the other fetus. If appropriate, pregnant women were referred for laser coagulation of the connecting anastomoses. All cases of TTTS with delivery after 20 weeks were reported. Acute perinatal TTTS (an acute transfusion event during labour) was defined as a difference in haemoglobin (Hb) concentration of at least 5 g/dl at birth between both twins and the identification of unbalanced placental vascular anastomoses by the pathologist without any indications of chronic TTTS or other causes for such an Hb difference.
Statistical analysis was performed with the SPSS statistical package 12.0 (SPSS Inc., Chicago, IL, USA). Differences between categorical variables were analysed using chi-square test. Differences between continuous variables were tested using independent sample t tests or a Mann–Whitney U tests, where appropriate. Possible confounders were adjusted for by logistic and linear regressions. Logistic regression was also used to evaluate the relation between birthweight differences and chorionicity. Odds ratios were calculated to estimate the relation of chorionicity with neonatal morbidity. Cox proportional hazard analysis was used to estimate the relation of chorionicity with mortality. Kaplan–Meier analysis was used to estimate cumulative survival. P values of <0.05 (two-sided) were considered to indicate statistically significant differences. Prior to the combined analysis, we analysed the data from both hospitals separately to assess a possible selection bias.
Table 1 shows the characteristics of the study population according to chorionicity. MC twins did not substantially differ from DC twins regarding maternal age, parity or gender distribution. Median gestational age was 1 week longer in DC twin pregnancies, and the mean birthweight was 221 g higher (95% CI 139–304). After adjustment for gestational age, birthweight did not differ. Severe birthweight discordancy (>20%) occurred more often in MC twins than in DC twins (OR 1.23, 95% CI 0.97–1.55).
Table 1. Characteristics of twins born in St Elisabeth Hospital (Tilburg) and University Medical Centre (Utrecht) between 1995 and 2004
MC (n = 198)
DC (n = 1107)
Age (years), median (range)
Parity, median (range)
Nulliparity, n (%)
Multiparity, n (%)
Number of infants (%)
Gestational age (weeks), median (range)
Birthweight (g), mean ±SD
2060 ± 788
2282 ± 769
After exclusion of stillbirths
2151 ± 705
2309 ± 739
Intertwin birthweight difference of >20%, n (%)
Sex distribution, n (%)
Perinatal mortality (≥20 weeks of gestation) was 11.6% in MC twin pregnancies and 5.0% in DC twin pregnancies (P < 0.001). The hazard ratio (HR) for perinatal death in MC twins compared with DC twins was 2.44 (95% CI 1.73–3.44). The excess perinatal mortality in MC twins was mainly due to the high incidence of stillbirths: 7.6% in MC twins compared with 1.5% in DC twins (HR 5.21, 95% CI 3.18–8.51). The incidence of neonatal death did not differ substantially between MC and DC twins. Thirty-eight MC twin pregnancies were complicated by TTTS, of which 9 were treated by laser occlusion of vascular anastomoses. Twenty-two percent (17 out of 76) of MC twins with TTTS died.
Table 2 shows the incidence of stillbirth and neonatal death according to chorionicity and gestational age at delivery. At all gestational ages, perinatal mortality was higher in MC twins. The HR for perinatal death in MC twins compared with DC twins between 32 and 37 weeks was 2.71 (95% CI 0.50–14.79) and after 37 weeks was 15.79 (95% CI 3.74–66.71).
Table 2. Mortality in MC and DC twins born in Elisabeth Hospital (Tilburg) and University Hospital Utrecht between 1995 and 2004 according to gestational age
Number of infants
IUD, n (%)
NND <8 days, n (%)
NND 8–29 days, n (%)
Total, n (%)
Number of infants
IUD, n (%)
NND <8 days, n (%)
NND 8–29 days, n (%)
Total, n (%)
NND, neonatal death.
Total, n (%)
Eighty percent of MC twins and 66% of DC twins were admitted to the neonatal nursery. The proportion of twins admitted to the NICU was also higher for MC twins than for DC twins (29.4 and 19.5%, respectively, Table 3). Overall, neonatal morbidity was considerably higher in MC twins, but after adjustment for gestational age and birthweight, only NEC occurred significantly more often in MC twins than in DC twins (OR 4.05, 95% CI 1.97–8.35, Table 3). There was a trend towards higher neonatal neuromorbidity (PVL and/or IVH) in MC twins compared with DC twins. There were two cases of cystic PVL in MC twins; one caused by TTTS during pregnancy and the other following IUD of the co-twin. The incidence of IVH grade III and IV was 1.4% in MC twins and 0.8% in DC twins (P = 0.363).
Table 3. The association of chorionicity and adverse pregnancy outcomes in twins born between 1995 and 2004 in Tilburg and Utrecht (The Netherlands)
IUD rate in continuing pregnancies after 32 weeks of gestation was 2.1% in MC twin pregnancies and 0.3% in DC twin pregnancies (HR 8.75, 95% CI 2.65–28.88). Again, the incidence of neonatal mortality did not differ between both groups. Figure 1 shows the fetal survival curve of both MC and DC twins, starting at 32 weeks of gestation. From 37 weeks onwards, intrauterine survival of MC twins decreased, whereas the survival of DC twins remained high. Mortality rates according to chorionicity did not differ between both hospitals.
The pregnancies in which perinatal death occurred after 32 weeks of gestation were studied in more detail, with the exception of two late neonatal deaths: a DC twin, born at 32+2 weeks who died 11 days after delivery due to a severe NEC, and a DC twin, born at 35+2 weeks who died 1 month after delivery due to a late onset sepsis/meningitis. The remaining 11 IUDs and 3 early neonatal deaths occurring after 32 weeks (7 MC and 7 DC infants) are summarised in (Tables 4 and 5).
Table 4. Features of MC twin pregnancies with intrauterine demise or neonatal death ≥32 weeks of gestation
Cause of death: severe asphyxia (pH 6.74) due to pre-eclampsia with severe pulmonary oedema
Convulsions treated with phenobarbital. No brain damage
S: 20th centile
D: <5th centile
IUD of smaller twin at routine scan
True knot in umbilical cord of demised twin. Autopsy declined
S: 50th centile
D: <5th centile
D: no data
IUD at routine scan (33+2 weeks)
Small placenta (<p10) with macroinfarctions. Severe asymmetrical fetal growth restriction
S: 5th centile
S: high PI (>3 SD) and low MCA PI
D: 50th centile
IUD at routine scan (32+2 weeks)
No placental cause of death found. Autopsy declined
S: 50th centile
D: <5th centile
Single NND of smaller twin
Small placenta (<10th centile) with signs of chronic fetal hypoxia/ischaemia. Cardiac and cerebral abnormalities due to chronic ischaemia
S: 50th centile
D: 20th centile
IUD of smaller twin at routine scan (35+5 weeks)
Small placenta with one large infarction and autolysis of umbilical cord. Placental insufficiency
S: 20th centile
D: 30th centile
IUD of larger twin at routine scan (40 weeks)
Small placenta with macro-infarctions. Placenta-fetus ratio 1:10. Fetal growth restriction
S: 30th centile
In two of the seven perinatal deaths in MC pregnancies, signs of TTTS had been present antenatally (at 32 and 33 weeks, respectively, case 1 and 2; Table 4). One MC IUD (38 weeks, case 5) and one neonatal death (40 weeks, case 6) were caused by (acute) TTTS during labour, without any signs of it antenatally. The same is likely to have occurred in one of the deaths in the twin pair that died antenatally or during early labour (37+ weeks, case 4), with exsanguination of the second twin. The other death (case 3) remained unexplained.
In three of the seven deaths in DC pregnancies, it was known that the infant was SGA (case 2, 3 and 5; Table 5). In two of these cases (at 37 and 38 weeks, case 2 and 5) pulsatility indices of the umbilical arteries of both fetuses had been normal at antenatal Doppler ultrasound examination. Two more cases (at 35 and 40 weeks, case 6 and 7) were SGA at birth, although umbilical artery pulsatility indices and ultrasonographic fetal weight estimation had been normal. There was one perinatal death due to acute deterioration of the maternal condition (asphyxia caused by acute pulmonary oedema of the mother, case 1) and one death remained unexplained (case 4).
This is the largest cohort of twin pregnancies in which adverse outcome was studied according to chorionicity. Perinatal mortality was higher in MC twin pregnancies than in DC twin pregnancies. This was partly consequence of a higher incidence of (very) preterm delivery, again confirming earlier observations.4,6,18 However, mortality also appeared to be higher at all gestational ages, with an eight-fold increased intrauterine mortality rate after 32 weeks of gestation. Thus, the poorer outcome in MC twins is not only explained by preterm delivery, as has been suggested.6 Our data clearly show that increased risk of IUD persists throughout gestation and even increases with advancing gestation beyond 32 weeks. Gestational age was 1 week shorter in MC twin pregnancies than in DC twin pregnancies, and, consequently, birthweight was lower, which reconfirms earlier observations.4,5
Although this study is not entirely population-based, we believe our cohort forms a good representation of the general twin population in The Netherlands. The majority of the pregnancies under investigation were uncomplicated pregnancies treated either in a regional teaching hospital or in a University Medical Centre. The latter hospital also acts as a tertiary referral centre for complicated cases before 32 weeks. This subgroup is, however, small, and the majority of these women return to their own regional hospital if delivery had not been taken place before 32 weeks. In addition, analysis of the data after 32 weeks (the most important period studied in this paper) did not show any major differences in results indicative of selection bias between both hospitals.
In the second half of the period studied in this paper, elective delivery of MC twins at 37–38 weeks of gestation was introduced based on recent literature about risks continuing into later pregnancy. This policy was applied to 90 of the MC twin pregnancies in this series. This explains the rather high upper range of gestational age in MC twins (40 weeks, Table 1) originating from our policy in the late 1990s. Management of uncomplicated DC twin pregnancies did not change in this study period, and spontaneous delivery was awaited.
A recent study has reported 2.9% of unexplained IUD after 32 weeks of gestation in MC diamniotic twin pregnancies without TTTS or intrauterine growth restriction (n = 139) and despite intensive fetal surveillance,9 which is comparable with the 2% that we found. In a Portuguese study, only one intrauterine fetal death occurred in a group of 168 MC diamniotic twin pregnancies (uncomplicated cases as well as TTTS, growth discordance and/or malformations) after 32 weeks of gestation. The latter group was monitored more intensively (weekly instead of biweekly and cardiotocography combined with sonography instead of sonography alone).14 Infants were also delivered earlier, although exact comparison of the age of delivery could not be made. In another study, no fetal deaths occurred in 53 uncomplicated MC pregnancies, all delivered preterm between 34 and 37 weeks of gestation.15 Acosta-Rojas et al.19 described only one fetal loss in their Spanish series of 104 uncomplicated MC twin pregnancies.
In two of the seven cases of perinatal death in our MC subgroup, antenatal signs of TTTS had been present (at 32 and 33 weeks, respectively), but in the other five, all delivered after 37 weeks, no antenatal signs of impaired fetal condition had been present. So it appears that third-trimester MC pregnancies at risk of perinatal death are difficult to identify. This is in agreement with the paper by Barigye et al.9 In this study, most deaths occurred between 34 and 37 weeks of gestation, whereas most deaths in our study occurred >37 weeks. These differences may well be due to chance. Combining data from the studies by Barigye et al.,9 Simoes et al.14 and ourselves, IUD occurred in 4/144 fetuses (2.8%) at 32–34 weeks, in 7/429 (1.6%) delivered at 34–37 weeks and in 4/329 (1.2%) born ≥37 weeks. With this calculation, we have assumed that all cases described by Simoes et al. were indeed born before 37 weeks of gestation.
Previous studies have also shown that perinatal mortality in DC pregnancies (i.e. after 37 weeks of gestation) is relatively low at term (0.5–4.5%; in our study 0.3%) and that death in term MC twins is higher (2.0–6.8%; in our study 3.5%).2,20,21 Compared with a cohort of twin pregnancies in the early 1900s, when fetal surveillance and intervention options were limited, perinatal mortality at term has decreased considerably in DC pregnancies but not in MC twins. Apparently, modern surveillance and intervention are effective in most DC twins but not in MC twins.22
Besides unexpected and sometimes unpredictable mortality (even at term), MC twinning is also associated with increased neuromorbidity. Several investigators have shown that both donor and recipient survivors from pregnancies complicated by TTTS are at significant risk for antenatally acquired cerebral lesions.23–25 Single fetal demise is also a known risk factor for adverse neurological outcome or co-twin death.7,8 In our cohort, one case of cystic PVL was found after TTTS, and there were two cases of single fetal death where in one case, cystic PVL with mild hemiplegia was found in the survivor and in the other case, death of the co-twin due to exsanguination was found.
However, even in the absence of either TTTS or single intrauterine fetal death, neuromorbidity is increased in (preterm) MC twins compared with DC twins.26 Multiple vascular anastomoses may indeed cause transitory cardiovascular imbalance, severe enough to decrease brain perfusion and to cause cerebral lesions without resulting in an IUD or clinically evident TTTS.27 We did not assess long-term neurological outcome, but we found a trend towards higher neonatal neuromorbidity in MC twins compared with DC twins. Overall, perinatal morbidity was higher in the MC twins in our study. However, this was largely explained by differences in age and weight at delivery. After adjustment for these differences, only the occurrence of NEC remained significantly (four times) increased in MC twins compared with DC twins. To our knowledge, this has not been described before. Histopathological changes that occur in NEC, such as mucosa oedema, haemorrhage and necrosis, are most often found in the watershed areas between the superior and inferior mesenteric artery, at the level of ileum and proximal colon, which strongly suggests ischaemic injury.28 It is tempting to speculate that in MC twin pregnancies, this ischaemic injury is a consequence of episodes of hypovolaemia that may occur due to imbalanced vascular anastomoses in the placenta.
Our and previous findings raise the question as to when to deliver the MC twin. The increased mortality (even at term), the failure to predict and prevent all cases of excess IUD by current antenatal care and the increased risk of (neuro) morbidity warrant planned delivery in the early term period or even in the late preterm period. Our study suggests that the optimal time of delivery is 36 weeks of gestation. Barigye et al.9 suggested an even earlier delivery (34–35 weeks), given the earlier occurrence of IUDs in their study. On the basis of our study and that of Barigye et al.,9 it seems likely that about one-third of the intrauterine demise may be prevented when delivery takes place before 37 weeks. Such an early delivery is also suggested by the recent small population of Cordero et al.,15 who found no fetal deaths in deliveries between 34 and 37 weeks. However, such advantages of early delivery have to be weighed against the risks of iatrogenic prematurity. RDS, for instance, has an almost three times higher incidence in the late preterm period compared with the term period.29 Therefore, amniocentesis to determine fetal lung maturation and/or administration of corticosteroids may well precede such an early delivery. Earlier observational studies have suggested a more conservative approach, that is planned delivery at 37–39 weeks of gestation.10,30–35 These studies were smaller than ours and used various endpoints (from morbidity to IUD and neonatal mortality). We acknowledge that there is a need for randomised data to determine the best strategy with respect to the delivery of MC twins. A randomised trial should also address (intensity of) fetal surveillance and include sufficient sample size to use IUD and neurological injury as endpoints. Such a trial, however, will be very hard if not impossible to accomplish. There is currently only one small randomised trial comparing elective induction of labour at 37 weeks of gestation and continued expectant management in uncomplicated twin pregnancies. In that study, no statistical differences in outcome were found.36 However, the study was underpowered, and no distinction was made between MC and DC twins. Currently, a randomised trial assessing the optimal time for delivery of term twins is conducted in South Australia.12 As an alternative, IUD in MC twins might be prevented by very intensive monitoring, as was described by Simoes et al.14 Elective delivery of uncomplicated DC twin pregnancies at 36–37 weeks of gestation does not seem to improve the outcome of these infants.11
In conclusion, MC twins with their single placenta and vascular anastomoses still pose unique problems. Compared with DC twins, they are at increased risk for perinatal death, NEC and neurological injury. Current antenatal surveillance is insufficient to identify the cases at risk for adverse outcome. This severely hampers safe management of the MC twin in utero. The best strategy is unclear at present. Awaiting randomised data, we and others recommend planned delivery at 36 weeks of gestation.