High perinatal survival in monoamniotic twins managed by prophylactic sulindac, intensive ultrasound surveillance, and Cesarean delivery at 32 weeks' gestation
Abstract
Objectives
Increased perinatal mortality in monoamniotic twin pregnancies is attributed to cord accidents in utero and at delivery. We evaluated the following parameters in monoamniotic pregnancies: (1) the incidence of cord entanglement; (2) the effect of sulindac on amniotic fluid volume and stability of fetal lie; and (3) the perinatal outcome with our current management paradigm.
Methods
This is a retrospective review of monoamniotic pregnancies of ≥20 weeks' gestation managed with serial ultrasound surveillance, medical amnioreduction and elective Cesarean delivery at 32 weeks' gestation. Mean amniotic fluid index (AFI) and change in AFI in monoamniotic pregnancies managed with oral sulindac was compared with 40 gestation-matched monochorionic–diamniotic controls.
Results
Among 44 monoamniotic pregnancies, 20 with two live structurally normal twins at 20 weeks' gestation satisfied the inclusion criteria. All fetuses survived to 28 days postnatally despite early prenatal cord entanglement in all but one case. Whereas AFI remained stable throughout gestation in the controls, the AFI fell in those patients on sulindac from a mean value of 21.0 cm (95% CI, 18.5–23.6 cm) at 20 weeks to a mean of 12.4 cm (95% CI, 10.1–14.6 cm) at 32 weeks (ANOVA P across gestation = 0.001) but mainly remained within normal limits. Fetal lie was stabilized in 11/20 cases in the monoamniotic group compared with 13/40 in the control group (P < 0.0001).
Conclusions
Cord entanglement appears unpreventable, as it typically occurs in early pregnancy. Sulindac therapy reduces AFI, leads to more stable fetal lie, and may prevent intrauterine death by diminishing the risk of constricting cords that are already entangled. Perinatal survival in monoamniotic pregnancies managed by a regime of sulindac from 20 weeks' gestation, close ultrasound surveillance and elective abdominal delivery at 32 weeks' gestation seems empirically higher than that in the literature. Copyright © 2006 ISUOG. Published by John Wiley & Sons, Ltd.
INTRODUCTION
Monoamniotic (MA) twins have long been associated with a high perinatal mortality rate of up to 30–50%. The rate has fallen in most but not all the more recently reported series1, Roque et al. reporting a 23% perinatal mortality rate from their literature review of all cases published between 1990 and 20022. This applies particularly where monoamnionicity has been diagnosed antenatally, with figures of 13–15% in two large series comprising a total of 139 cases3, 4. The primary cause of fetal death is cord occlusion as a result of cord entanglement, implicated in over 50% of cases of intrauterine loss. Other contributions come from cord entanglement/prolapse at vaginal delivery and an increased incidence of congenital malformations. Despite an improved outcome associated with antenatal diagnosis, Allen et al. estimated that 25% of pregnancies still suffered at least one fetal loss after 20 weeks' gestation3.
Antenatal management is aimed at preventing fetal death, but how best to achieve this remains uncertain5. A number of strategies have been suggested to reduce the high perinatal mortality, for example delivery by Cesarean section. Another strategy is elective preterm delivery, although there is still controversy regarding exactly when this should occur between the limits of 32 and 35 weeks' gestation1, 3, 6, 7. To address this, Roque et al. recently estimated that perinatal mortality rises from around 1% per week before 32 weeks to 4% per week at 33–35 weeks and to 7% per week thereafter2.
The majority (60%) of deaths in structurally normal MA fetuses, however, occur before 32 weeks' gestation7. To prevent this, several authors have suggested that such cases should be treated by admission to hospital—after viability has been established—with cardiotocography several times per day1, 4, 7. Although this may prevent some fetal deaths1, 4, it is not surprising that it does not prevent all intrauterine death (IUD)8, as cord accidents can occur acutely without prior warning. Furthermore, intensive cardiotocographic monitoring over 6–10 weeks in hospital comes at considerable social and economic cost, and appears to increase the incidence of emergency delivery for fetal distress4.
Nearly a decade ago we suggested an outpatient strategy of medical amnioreduction prior to 32 weeks' gestation to reduce IUD secondary to cord entanglement in MA twins9. The rationale was to prevent the accumulation of a relative abundance of amniotic fluid that would otherwise allow unrestricted fetal movements and changes in fetal position, which contribute to the high IUD rate from cord entanglement, compression and/or tightening of cord knots. Sulindac, a non-selective prostaglandin synthase inhibitor used to treat preterm labor and polyhydramnios10, crosses the placenta along with its active metabolite in a dose-dependent manner, to decrease fetal urine output and therefore amniotic fluid volume9, 11. We chose this agent over indomethacin, as sulindac has better patient tolerance and does not produce significant ductal constriction prior to 32 weeks' gestation10, 12.
In our 1997 report, maternal administration of prophylactic sulindac in three uncontrolled MA cases was associated with induced borderline oligohydramnios and stabilization of fetal lie9. Our management of MA twins since then has comprised prophylactic maternal sulindac from 20–22 weeks' gestation, serial ultrasound surveillance, and elective Cesarean section at 32 weeks' gestation after the administration of steroids to ensure lung maturity. The aim of this study was to audit perinatal outcome in relation to the above management protocol, and in particular the effects of sulindac on cord entanglement, fetal lie and amniotic fluid volume.
METHODS
Patients
All cases of MA multiple pregnancies referred to our tertiary referral fetal medicine service between December 1994 and December 2005 were reviewed. Monoamnionicity was diagnosed based on established ultrasound criteria, comprising lack of a dividing membrane on at least two consecutive examinations in single-placenta same-sex fetal pairs with evidence of unrestricted fetal movement in the absence of features of ‘stuck twin’ syndrome13, 14. Monoamnionicity was confirmed postnatally by placental and membranous histology. Inclusion criteria were MA fetal pairs with two live fetuses at 20 weeks' gestation. Exclusion criteria were conjoined twins, higher order multiple gestations, MA twins discordant for major fetal abnormalities, and twin reversed arterial perfusion sequence.
All the women were counseled about the risk of IUD from cord complications and our clinical approach to reducing this risk, which comprises intensive ultrasound surveillance, maternal sulindac therapy and elective delivery at 32 weeks' gestation. Patients were advised that sulindac therapy is aimed at reducing amniotic fluid volume by reducing fetal urine production, and thus the risk of cord complications and IUD. They were also informed that it had been used at the same dose in this and other institutions as a tocolytic without significant adverse effects on mother or baby and that, like most interventions in fetal medicine and rare event conditions, it had not been tested in randomized trials.
Clinical management
Oral sulindac 200 mg b.d. was commenced in women with both MA twins alive at 20 weeks' gestation or at the first visit if referred after 20 weeks. This gestational age was chosen because fetal urine production starts at around this time, therefore changing fetal urine production should alter amniotic fluid volume15, 16.
Fetal growth was evaluated every 2 weeks until delivery, as standard in monochorionic twins. However because of the need to monitor the amniotic fluid index (AFI) in response to sulindac to avoid moderate to severe oligohydramnios, the amniotic fluid volume was checked weekly, as also was the umbilical artery pulsatility index, fetal position, and the presence of cord entanglement, using color flow Doppler ultrasonography. AFI was measured as the sum of the deepest pools in each of four quadrants. Sulindac dosage was adjusted (to 0, 100, 200 or 400 mg/day) where necessary to aim for an AFI between 5 and 8 cm. Maximum systolic velocities in the ductus arteriosus were measured once every 4 weeks for signs of ductal constriction, defined as a pulsatility index of < 1.917. Fetal echocardiography was performed by a perinatal cardiologist at 20–24 weeks' gestation. Mothers were administered prophylactic corticosteroids 2–7 days prior to delivery and elective Cesarean section was planned under regional block at 32 weeks' gestation in our center, with level III neonatal facilities and at least two intensive care cots available.
Audit
Following our initial report on this strategy9, it has been incorporated into the local management protocol for MA twins. Like most treatments in fetal medicine, it has not been tested in large cohort or randomized studies, and clinical management was thus necessarily based on individual risk–benefit assessment with the aim of preventing perinatal death in a high-risk situation. Accordingly, as with similar reports on MA twins5, this report takes the form of a retrospective audit exempt from research ethics approval, as it was not designed or undertaken as a research study. All data were acquired as part of routine clinical management, and retrieved from our computerized database and augmented by case record and neonatal discharge summary review. Perinatal survival was defined as survival to 28 days of postnatal age.
Data analysis
Parametric and non-parametric statistical methods were applied as appropriate. Data are presented as mean and 95% CIs or medians and ranges after testing for normality using Kolmogorov–Smirnov analysis. Parametric data were compared by the unpaired t-test or analysis of variance (ANOVA) with post-hoc t-testing for significance, while categorical data underwent Chi-square analysis using GraphPad Prism 4.0 (GraphPad Software Inc, Institute for Scientific Information, San Diego, USA). Data from the first three cases have been previously reported9. Mean AFI on treatment, mean change in AFI and change in fetal position were compared with a control group of 40 uncomplicated monochorionic–diamniotic (MCDA) twin pregnancies retrieved from our database who were monitored fortnightly as part of their routine clinical care over the same period. The control group was chosen as the next uncomplicated (no twin-to-twin transfusion syndrome or intrauterine growth restriction) MCDA pregnancy in the database following each index MA pregnancy, that had undergone serial ultrasound surveillance from 20 weeks' gestation in our center. The change in AFI (ΔAFI) for each 2-week block was calculated by subtracting the AFI at the beginning of the interval from the AFI at the end of the interval. A change in fetal lie was defined as a distinct change in fetal position from one week to the next, from cephalic to oblique, transverse or breech (or vice versa). The percentage estimated fetal weight difference (ΔEFW) was calculated as the difference in twin weights, divided by the larger twin's weight, multiplied by 100, with growth defined as concordant when ΔEFW was <25%18. Small for gestational age was defined as birth weight below the 10th centile for gender and gestational age19.
RESULTS
Forty-four pregnancies were identified as MA. Twenty-four were excluded: eight acardiac, six complicated by single or double IUD before 20 weeks' gestation, four discordant for major fetal anomalies, one triplet pregnancy and five with conjoined twins. Twenty MA pregnancies with two live fetuses at 20 weeks' gestation were identified for analysis. Median gestational age at presentation was 15 + 2 (range, 8 + 4 to 27 + 3) weeks. Approximately two-thirds of the fetuses were female (28/40) for a sex ratio (female/total) of 0.7, in accord with the known predominance of female babies among MA twins20.
Cord entanglement was diagnosed ultrasonically at the first fetal assessment in all but one case. This woman was referred late at 27 weeks' gestation, and cord entanglement was not diagnosed until 30 weeks.
Sulindac was started at a median gestational age of 21 (range, 20 + 0 to 27 + 3) weeks and continued for a median duration of 10 + 6 (range, 3 + 0 to 12 + 2) weeks. One patient stopped treatment after 3 weeks as a precautionary measure after she developed obstetric cholestasis, which was confirmed biochemically. Two had their dose reduced to 100 mg daily because of severe oligohydramnios (AFI < 3 cm), with subsequent return of AFI to >5 cm. Two further patients reported abdominal pain and diarrhea, which resolved after halving the dose to 200 mg per day for a week as a precaution; it did not recur on resuming the standard dose thereafter.
In the MA group, the AFI fell from 21.0 (95% CI, 18.5–23.6) cm at 20 weeks' gestation to 12.4 (95% CI, 10.1–14.6) cm at 32 weeks' gestation; ANOVA P across gestation = 0.001. In contrast, in the controls, the mean AFI remained stable over the same time period: 19.7 (95% CI, 18.0–21.3) cm at 20 weeks' gestation and 19.4 (95% CI, 10.0–20.9) at 32 weeks (P = 0.65; Figure 1). This analysis was supported by the area under the curve (AUC, cm–weeks) of AFI against gestational age, which at 146 (95% CI, 115–176) was lower overall in the MA group than in the control group (205 (95% CI, 181–229), P = 0.002).

Change in amniotic fluid index (AFI) over gestation in monoamniotic twin pregnancies treated with sulindac (●) and gestation-matched monochorionic–diamniotic control twin pregnancies (○). Data are expressed as mean and 95% confidence intervals.
Analysis of the change in AFI over each consecutive 2-week period after commencing sulindac showed that the main reduction in AFI occurred in the first 2 weeks of treatment (ANOVA P = 0.003, post hoc testing for week 2 of treatment compared with week 0, P < 0.001). Thereafter the AFI remained relatively stable (Figure 2).

Change in amniotic fluid index (delta AFI) over 2-weekly consecutive time periods from commencement of therapy in monoamniotic (●) compared to gestation-matched monochorionic–diamniotic control (○) twins. Data are expressed as mean and 95% confidence intervals.
However, although AFI fell significantly on treatment—confirming our earlier findings in MA twins9—the target AFI of 5–8 cm was achieved in only seven out of 20 pregnancies. The proportion of pregnancies in which the fetal lie remained stable (i.e. did not change during treatment) was greater in the MA group (11/20, 55%) than in the control group (13/40, 32.5%), (Chi-square P < 0.001).
ΔEFW of >25% was detected in two out of 20 MA sets antenatally, and comparable birth-weight discordance was confirmed postnatally. Ductal constriction did not occur in any case.
All the women delivered two live infants by planned Cesarean section at a median gestation of 32 + 1 (range, 31 + 0 to 33 + 1) weeks; one patient delivered at 31 + 0 weeks after administration of steroids because of spontaneous preterm labor. All the babies were born in good condition, with median Apgar scores of 9 (range, 5–9) and 9 (range, 6–10) at 1 and 5 minutes, respectively. The median birth weight was 1609 (range, 1020–2300) g and eight babies (20%) were small for gestational age at birth. All the neonates were discharged from the neonatal unit with a median stay of 28 (range, 3–64) days, and there were no perinatal deaths. Neonatal complications included hyperbilirubinemia requiring phototherapy (n = 8), mild pulmonary hypertension (n = 1), transient tachypnea of the newborn (n = 1), respiratory distress syndrome (RDS) (n = 11) and bronchopulmonary dysplasia (n = 1). The median number of days' ventilation in the RDS group was 2 (range, 0–7). There were no major neurological complications (>grade I intraventricular hemorrhage or cystic periventricular leukomalacia). Mild intracranial abnormalities on cranial ultrasonography (parenchymal flares, slight ventricular asymmetry, and minor echodensities) were present in 15 cases, but no evidence of cystic periventricular leukomalacia was present in any case at 4 weeks of age. One baby required neonatal surgery for an otherwise unexplained gastric perforation on day 2. One baby had an episode of abdominal distension and another one had a septic paralytic ileus on day 10, both of which were managed conservatively with intravenous antibiotics. One fetus with congenitally corrected transposition had pulmonary artery banding on day 18 and balloon atrial septostomy on day 20. Only two babies had malformations, both diagnosed prenatally: one corrected transposition of the great arteries and one small ventricular septal defect that closed postnatally. All were well at discharge.
DISCUSSION
Despite improvements in perinatal outcome as a result of early antenatal diagnosis of monoamnionicity and intensive fetal surveillance, several recent series have concluded that the resultant perinatal mortality rates of 13–32% remain high1, 4, 8. In contrast, our audit shows that a combined strategy of medical amnioreduction, weekly ultrasound surveillance and elective abdominal delivery at around 32 weeks' gestation following antenatal steroids was associated with a high survival rate, here 100% (40/40 fetuses) within the limits of small numbers.
The main challenge has been to prevent the majority of IUDs that occur prior to 32 weeks' gestation, as at this gestational age delivery can be fairly safely effected with reasonable neonatal outcome in pregnancies otherwise at high risk of IUD. This risk of IUD is largely attributed to umbilical cord entanglement and subsequent cord occlusion, the prevention of which has hitherto been challenging5.
Roque et al. in their literature review reported a rate of prenatal diagnosis of cord entanglement of only 23%, although the degree to which antenatal diagnosis was attempted is not clear. In our series, in which monoamnionicity was diagnosed early in pregnancy and cord entanglement carefully searched for, all pregnancies except one already had cord entanglement at the first scan. It seems intuitive to infer that cord entanglement is present in virtually all MA twins from the first trimester, as this is when fetal position changes most frequently as there is a relative abundance of amniotic fluid and thus little restriction to fetal movements. Accordingly, it seems that the majority of cases with cord entanglement do not suffer adverse outcome, but in those that do, cord compression/occlusion occurs many weeks after the cords have become entangled. We speculate that entanglement may be loose at least initially, but has the potential to tighten and compromise the fetal circulation later in the pregnancy. Therefore strategies for preventing IUD should be based not around preventing cord entanglement, but around reducing the risk of further changes in position leading to compression or occlusion of an already entangled cord.
As excessive fetal movements may increase tightening of an already entangled umbilical cord which could lead to occlusion, fetal distress and IUD, the rationale for using sulindac was thus not to prevent cord entanglement, but instead to reduce the amount of amniotic fluid and consequently fetal movements and cord accidents. In our series sulindac significantly reduced AFI compared with an untreated MCDA group with the same baseline AFI, with the effect predominantly in the first 4 weeks of therapy. This confirms our earlier case report in MA twins9 and more recent data from our group in singleton pregnancies21, the latter showing that a reduction in AFI was detectable within 48 h of initiating therapy at this dose. Other groups have used sulindac at the same dose to prevent preterm labor in singleton pregnancies; Kramer et al.10 showed that sulindac reduced the AFI compared with terbutaline, while Humphrey et al.22 demonstrated a reduction in AFI with a maximum effect after 2 weeks of treatment. However we found that sulindac did not consistently induce oligohydramnios, which was our original rationale for reducing fetal movements and thus cord accidents. Indeed oligohydramnios of at least mild severity with an AFI <8 cm was only achieved in half these patients, only two cases achieving an AFI <5 cm. The modest fall in AFI we observed on sulindac is supported by two studies published well after the introduction of this management strategy, both Kramer et al. and Humphrey et al. finding that none of 56 pregnancies had a fall in AFI to <5 cm on sulindac therapy10, 22.
Notwithstanding the above, significantly fewer babies in the MA group (45%) changed fetal lie during the course of treatment compared to those in the control group (67%). Therefore despite only achieving a minor reduction in AFI to a mean of 12 cm, sulindac appeared to stabilize fetal lie, and was associated with a 100% perinatal survival in this series. This is in contrast to a case report of two MA pregnancies treated with sulindac, in which perinatal survival was only 25%23. However, in that report, AFI was not quantitated in one case and there was polyhydramnios in the other.
The optimal method of fetal surveillance of MA twins remains unclear, with some authors advocating hospitalization and intensive cardiotocography (CTG). In our series the only method of surveillance was ultrasonography and Doppler scan. Our rate of intervention for abnormal monitoring was zero, in contrast to high rates of delivery for fetal distress (61.5%) reported by others7. This is not surprising, as CTG traces are difficult to interpret in the late mid-trimester, when variability is low and decelerations may occur as normal events. Moreover serial CTG monitoring has psychologically adverse effect on patients, with whom the lengthy hospital admission from viability to 32 weeks' gestation is unpopular. Long-term hospital admission for CTG monitoring also has huge cost implications compared with weekly outpatient ultrasound monitoring.
The final component of our strategy was delivery at 32 weeks' gestation following the administration of steroids, as the risks of continuing the pregnancy were then considered to outweigh those of preterm delivery. Although some case series suggest that IUD is uncommon after 30–32 weeks' gestation8, 24, 25, meta-analysis of 133 recent cases reported a substantial perinatal loss rate (10.2%) in pregnancies continuing beyond 32 weeks2. Moreover, recent data from our group have shown a 1 in 23 prospective risk of unexpected death in uncomplicated monochorionic twins even with diamnionicity26. Complicated MCDA twin pregnancies, such as those with twin-to-twin transfusion syndrome or abnormal umbilical artery Doppler scan, are already delivered in many centers at 32 weeks' gestation. At this stage neonatal survival is now comparable to that at term27, 28. Minor neonatal morbidity is still likely at 32 weeks, but the chance of major respiratory and neurological problems is reduced substantially by administration of antenatal steroids26. At 32 weeks' gestation the majority of pregnancies have mature lecithin/sphingomyelin ratios following prophylactic corticosteroids29, 30, and in our institution neonatal survival at 32 weeks is similar to that at term. The incidence of RDS at 32 weeks' gestation has been estimated to be 5% with surfactant use31 and mortality to be less than 1% for non-anomalous fetuses32. In keeping with this, there were no neonatal deaths in our study group, and only a minority developed RDS. We accept that elective delivery of all MA twins at 32 weeks' gestation carries an attendant neonatal morbidity, but this is outweighed by the 10% prenatal loss rate if the pregnancy is allowed to continue.
All deliveries were by Cesarean section. Notwithstanding some reports suggesting that vaginal delivery may be safe, we, like most centers, deliver all women with MA twins by Cesarean section because of the possibility of obstetric accidents (cord constriction/tightening during labor, cord prolapse for Twin 2 when Twin 1 is already delivered). This risk is emphasized by our data suggesting that cord entanglement is almost ubiquitous in MA twins.
In terms of potential adverse drug effects, maternal complications were minimal. Two women complained of gastrointestinal symptoms, which resolved following a reduction in dose, and did not recur on resumption of the standard dose. Another developed obstetric cholestasis, a complication seen at higher frequency in multiple pregnancies. Both were considered unlikely to be related to the use of sulindac. In terms of potential fetal side-effects, we found no significant constriction of the fetal ductus arteriosus on monthly clinical checks. More recently our group has reported that sulindac leads to a 27% fall in ductus arteriosus pulsatility index between 28 and 32 weeks' gestation within 48 h of starting the therapy, and that this resulted in ductal constriction in around half the cases, but it was always transient and reversible21. Cessation of therapy by 32 weeks' gestation may be important in this regard. In terms of potential neonatal drug effects, one baby in the treated group had mild persistent pulmonary hypertension and another had gastric perforation, both recognised fetal complications of non-steroidal anti-inflammatory drugs in pregnancy33, 34. However, both are also recognized complications of prematurity, and in addition have been associated with the use of antenatal steroids for fetal lung maturity as well as postnatal steroids35, 36. As all our pregnancies received antenatal steroids prior to delivery, it is not possible to determine whether these complications were due to sulindac, steroids, or prematurity, either alone or in combination.
In conclusion, our audit data suggest that cord entanglement does not appear to be preventable, as it usually occurs early in pregnancy. Sulindac therapy reduces AFI by over 40%, and is associated with more stable fetal lie and thus a probable reduction in the risk of tightening and occluding cords that are already entangled. Overall the high perinatal survival in our series of 20 MA pregnancies indicates that a regime of sulindac after 20 weeks' gestation, close ultrasound surveillance and elective abdominal delivery at 32 weeks' gestation following antenatal steroids seems promising, with relatively few maternal or fetal adverse effects and considerable cost savings compared with prolonged hospital admission for CTG monitoring. A randomized controlled trial would establish definitively whether sulindac therapy, intensive ultrasound surveillance and delivery at 32 weeks' gestation improves perinatal survival over conservative management, although testing any individual component like prophylactic sulindac would be logistically challenging even if carried out on a multicenter basis, owing to the rarity of MA twins.
Acknowledgements
We acknowledge long-term support from the Richard and Jack Wiseman Trust and the Institute of Obstetrics & Gynaecology Trust, and equipment support from the John Ellerman Foundation and Children's Nationwide Medical Research Foundation.