The long term neurologic outcome of children from pregnancies complicated by twin-to-twin transfusion syndrome

Authors


Dr J. Dickinson, School of Women's and Infants' Health, The University of Western Australia, 374 Bagot Road, Subiaco, Western Australia, 6008, Australia.

Abstract

Objective  To assess long term outcomes of children from pregnancies complicated by twin-to-twin transfusion syndrome.

Design  Comparison of children from pregnancies with twin-to-twin transfusion syndrome in Western Australia with a contemporaneous regional comparison cohort of preterm and term infants studied using an identical assessment procedure.

Population and setting  All infants aged ≥18 months were identified from a geographically based longitudinal cohort of monochorionic twin pregnancies with an antenatal diagnosis of twin-to-twin transfusion syndrome conducted prospectively since 1992.

Methods  Children were evaluated using age-specific developmental and behavioural assessments. Cerebral palsy was diagnosed clinically and ascertainment confirmed through the Western Australian Cerebral Palsy Register.

Main outcome measures  Intellectual disability, cerebral palsy, behavioural and cognitive function.

Results  Fifty-two children were identified as eligible for study and assessments were performed on 49 (94%). Three surviving children had a diagnosis of cerebral palsy (5.8%). The mean IQ score was 8 points lower in twin-to-twin transfusion syndrome children compared with the comparison cohort although this was mainly due to a decrement of 13 points in those born before 33 weeks of gestation. There was no difference between the donor and the recipient twin in terms of IQ scores (median difference −3, 95% CI −9 to 6). There was no relationship of IQ score to the worst stage of the twin-to-twin transfusion syndrome. Child Behavior Check List and Vineland Adaptive Behavior Scale scores did not differ between twin-to-twin transfusion syndrome children and the comparison group.

Conclusions  Twin-to-twin transfusion syndrome is associated with a significant reduction in IQ score in very preterm survivors. There seems to be no increase in the prevalence of cerebral palsy. Overall behaviour and adaptive behaviour scale scores are similar to a comparison group.

INTRODUCTION

Over the past two decades short term outcomes for pregnancies complicated by twin-to-twin transfusion syndrome have improved from almost universal death to survival rates of 50–70%.1 These improvements in perinatal survival have been secondary to obstetric interventions such as serial amnioreduction2–4 and placental laser ablation therapy,5–7 combined with advances in neonatal intensive care. Although perinatal survival rates have improved, outcomes for twin-to-twin transfusion syndrome are confounded by the impact of preterm birth and abnormalities in fetal growth.8–12 The most common antenatal intervention for twin-to-twin transfusion syndrome is serial amnioreduction, for which cases the median gestation at delivery is 29–30 weeks.13,14 Clearly, such very preterm birth is associated with significant neonatal complications and the possibility of long term disability. The incidence of major neonatal cranial ultrasound abnormality has been reported as high as 20.9–24.3% with an incidence of periventricular leucomalacia in perinatal survivors of 5.1–7.5%.13,14 Given the high morbidity rates in the perinatal period, systematic study of the longer term outcomes is required.

In 1992, a prospective longitudinal database of pregnancies complicated with twin-to-twin transfusion syndrome in Western Australia was established. Due to the existence of a single tertiary referral centre for perinatal medicine in Western Australia, the unique opportunity to monitor and review all cases of prenatally diagnosed twin-to-twin transfusion syndrome is now possible. The aims of this study were to (i) assess the long term outcomes for all cases of monochorionic twin pregnancies complicated by twin-to-twin transfusion syndrome, independent of therapeutic intervention and (ii) contrast these outcomes with a contemporaneous regional comparison cohort.

METHODS

In 1992, a prospective cohort encompassing all cases of prenatally diagnosed twin-to-twin transfusion syndrome pregnancies was established at the Women and Infants Research Foundation, Perth, Western Australia. The study population for this manuscript includes all pregnancies from this cohort during the period from May 1992 through May 1999.

Obstetric and neonatal outcome data for all cases of twin-to-twin transfusion syndrome have been collected prospectively and maintained on the twin-to-twin transfusion syndrome register since 1992. Diagnostic criteria, patient characteristics and perinatal outcomes for this population have been reported previously.15 This study design involved the identification of all surviving children aged at least 18 months from the twin-to-twin transfusion syndrome cohort. Confirmation of survival was checked through the Western Australian Registrar General's Office for notification of death. The residence of all surviving children was identified using the statewide outpatient appointment system database and the responsible physician contacted the parents to ascertain willingness to participate in the study. The Institutional Ethics Committee of King Edward Memorial Hospital for Women provided permission for the conduct of this study.

Children were assessed using questionnaires and standardised psychological assessments. In addition, the children born very preterm (<33 weeks) underwent neurologic examination as did more mature infants where questionnaires or psychological assessments suggested possible neurologic deficits. Questionnaires used were (i) general health/demographic questionnaires, (ii) Vineland Adaptive Behavior Scales (VABS)16 and (iii) The Child Behavior Check List (CBCL).17 This VABS measures behaviour in several domains relevant to the child's function in their home environment (i.e. communication, daily living skills, socialisation and motor function) as well as an overall composite. The VABS was completed during an interview with the child's primary caregiver and standardised scores were compared. The CBCL is a validated questionnaire for parents to rate a wide range of their child's behaviours with responses being grouped in six subscales. This study reports the total standardised behavioural score (the sum of all six scales) and the externalising score (the sum of aggressive and destructive subscales). Problem behaviours were defined as those where the total or aggressive/destructive scores were greater than the 98th centile values for the Western Australian population.

Psychological assessments used were the Bayley Scales of Infant Development (BSID)18 for children below three years of age, or the Stanford–Binet Intelligence Scale, Fourth Edition, Australian Adaptation.19 Intellectual disability was defined as either a Mental Development Index (MDI) on the Bayley test or a Stanford–Binet Intelligence Score more than two standard deviations below the Western Australian comparison groups.

Cerebral palsy was defined clinically, based on findings of increased muscle tone in the presence of functional motor deficit. Completeness of cerebral palsy ascertainment was confirmed by checking for cases against the West Australian Cerebral Palsy Register. This statewide register has multiple sources of data input from all clinical services likely to provide services to children with disabilities, and has been shown to have comprehensive cerebral palsy ascertainment on a regional basis.

Infants from the twin-to-twin transfusion syndrome group were compared with a large contemporaneous regional comparison cohort of infants on whom similar assessments had been made at age three and six years of age. The majority of these were infants born very preterm or of very low birthweight, and represented all such infants delivered in West Australia from 1990 to mid 1992.20 The comparison cohort also included 200 infants who were born at term and were part of the Western Australian Pregnancy Cohort (Raine) Study.21

Comparisons were performed in two phases. The first phase was a multivariate analysis of variance to determine any effect of being born as part of a pregnancy affected by twin-to-twin transfusion syndrome, controlling for sex, gestational age at birth, birthweight ratio, being part of a multiple pregnancy and social class. In the second phase, twin-to-twin transfusion syndrome children were compared in a case–control manner with children from this comparison cohort. Cases were matched on sex, gestational age, birthweight ratio and social class. Two controls were sought for each twin-to-twin transfusion syndrome child, the first being part of a same-sex twin pair (twin control) and the second being a singleton delivery (singleton control). Due to difficulties in matching, only 27 twin controls were available and these were at the very preterm end (<33 weeks) of the gestational age range.

The data analyses were conducted using the SAS statistical package version 8 (SAS Systems, Cary, North Carolina, USA). Normally distributed continuous data were expressed as mean and standard deviation, and otherwise as median and interquartile range. Categorical variables were expressed as n (%). Comparisons for the IQ variable between the twin-to-twin transfusion syndrome group and the regional cohort were made using analysis of variance controlling for gestational age, sex and social class. For the non-parametric data (CBCL, VABS) comparisons were made using the Wilcoxon two-sample test. Outcomes for twin-to-twin transfusion syndrome groups and matched controls were compared using the paired Student's t test (IQ) and Wilcoxon matched pairs signed rank test (CBCL, VABS). Adjusted differences with 95% confidence intervals have been quoted for IQ measures, and median differences for behavioural measures. For all tests, a P value of 0.05 was considered to be statistically significant.

RESULTS

A total of 64 pregnancies complicated by twin-to-twin transfusion syndrome occurred during the study period. In 13 pregnancies both fetuses were stillborn. In 44 pregnancies, both fetuses were alive at birth and in seven pregnancies only one fetus was born alive. A total of 81 neonates survived to hospital discharge and one infant subsequently died at six months of age from neurologic complications. Twenty-eight twins were aged less than 18 months at the time of the review, providing an eligible study population of 52 children (21 pairs of twins and 10 single surviving twin) from 31 pregnancies.

The perinatal outcomes for the 52 surviving children are shown in Table 1. The majority of surviving children were delivered preterm, with 52% delivered very preterm (<33 weeks of gestation). All pregnancies with at least one survivor were >25 weeks of gestation at delivery. The median duration of neonatal nursery admission was 37 days (22, 54 days). Grade III or IV intraventricular haemorrhage was present in four cases (7.7%) and cavitary periventricular leucomalacia present in two cases (3.8%).

Table 1.  Antenatal and delivery characteristics of surviving children. Data shown as median [interquartile range] or n (%) as appropriate.
  • #

    Data are presented per child (52 surviving children).

No. of pregnancies31
Gestation at diagnosis (weeks)21.4 [20.1, 26.1]
Critically abnormal umbilical flow patterns#8 (15.4)
Fetal hydrops of recipient at diagnosis1 (3.2)
Use of amnioreduction22 (71.0)
Amnioreduction per pregnancy (n= 22)3 [2, 4]
Diagnosis to delivery interval (weeks)9.0 [2.6, 14.7]
Antenatal corticosteroids27 (87.1)
Gestation at delivery (weeks)32 [29.1, 36.3]
Caesarean section17 (54.8)
 
Preterm birth
<33 weeks of gestation16 (51.6)
<37 weeks of gestation25 (80.7)
 
Birthweight (g)#
Recipient (n= 27)1730 [1265, 2460]
Donor (n= 25)1450 [940, 1700]

Forty-four of the eligible children were reviewed according to the complete study protocol (85%) with five others (9%) having partial assessments. Three children (6%) had no outcome data due to loss of contact with two families and one family declined participation (all single survivor families). The median age of assessment for twin-to-twin transfusion syndrome children was five years (3.3, 6.3 years), and for the comparison group was four years (3.2, 6.2 years). Eleven twin-to-twin transfusion syndrome children were tested with the Bayley MDI and the remaining 33 were assessed with the Stanford–Binet Intelligence Score, as were the entire comparison group. Clinical details of the comparison cohort and the selected controls are shown in Table 2.

Table 2.  Clinical details. Values are presented as median (interquartile range).
 Twin-to-twin casesComparison groups
Regional cohortMatched controls#
SingletonsTwins
  • #

    Controls matched by sex, gestational age, birthweight ratio and social class from the regional cohort.

n5211054427
Gestational age (weeks)32 (30, 36)32 (30, 39)32 (30, 36)31 (29, 31)
Birthweight ratio0.81 (0.68, 0.98)0.99 (0.88, 1.1)0.88 (0.74, 1.0)0.91 (0.81, 1.0)
Social class2.4 (1.5, 3.0)2.4 (1.6, 3.0)2.6 (1.4, 3.2)2.2 (1.4, 3.2)
Male sex (%)5353  

Cerebral palsy was present in three children (5.8% overall) all of whom were independently ambulant. These were manifest as diplegia (two cases) and hemiplegia (one case). Thirty-two children were delivered very preterm at <33 weeks of gestation, two of whom had cerebral palsy (6.3%, 95% CI 0.8 to 20.8). In the comparison cohort, the prevalence of cerebral palsy in very preterm non-twin-to-twin transfusion syndrome twins was 9/184 (4.9%, 95% CI 2.3 to 9.1) and in very preterm singletons 17/510 (3.3%, 95% CI 2.0 to 5.3). These rates were not statistically significantly different.

Twin-to-twin transfusion syndrome children had a lower score on cognitive tests with a significant interaction between twin-to-twin status and gestational age (Table 3). Very preterm twins in the twin-to-twin transfusion syndrome group had a 13-point difference in IQ scores, whereas those at higher gestational ages had only a three-point difference (Table 4). These differences were confirmed in the case–control analyses. Six (14%) of the twin-to-twin transfusion syndrome children had IQ scores more than two standard deviations below the Western Australian mean. There was no significant difference in IQ scores between putative donor and recipient twins where both had survived (median difference −3, 95% CI −9 to 6). There was no relationship to the worst stage of the twin-to-twin transfusion syndrome; stages 1–2: median 96 (90, 104); stage 3: 94 (87, 103); stages 4–5: 101 (79, 105).

Table 3.  Cognitive outcomes. Values are presented as mean [standard deviation] unless otherwise indicated.
 Twin-to- twin casesComparison groups
Regional cohortMatched controls
SingletonsTwins
  • a

    Twin-to-twin transfusion subset matched with singleton controls.

  • b

    Twin-to-twin transfusion subset matched with twin controls.

  • #

    P value determined from linear regression adjusted for gestational age, sex and social class (regional cohort) and paired t test (matched controls).

n44a27b11054427
IQ score95 [15]91 [14]103 [11]104 [10]104 [11]
IQ difference (95% CI)  8 (4 to 11)9 (4 to 14)13 (6 to 20)
P#  0.00010.00210.0008
Table 4.  Cognitive outcomes by gestational age group. Values are presented as mean [standard deviation] unless otherwise indicated.
 ≤32 weeks≥33 weeks
Twin-to- twin casesRegional cohortTwin-to- twin casesRegional cohort
n2481420291
IQ score89 [16]102 [11]102 [9]105 [11]
IQ difference (95% CI) 13 (9 to 18) 3 (−2 to 8)
P 0.0001 0.515

There were no differences in any subscale score on the CBCL or on any subscale of the VABS on either analysis (Table 5). Five (12%) of twin-to-twin children had total or externalising behaviour scores on the CBCL above the 98th centile indicative of a clinical behaviour problem. Overall, disabilities were present in seven (14.3%) of twin-to-twin transfusion syndrome children, with five having intellectual disability, two having cerebral palsy only and one with both. There were no cases of blindness or deafness.

Table 5.  Behavioural outcomes. Values are presented as median (interquartile range).
 Twin-to-twin casesComparison groups
Regional cohortMatched controls
SingletonsTwins
  • CBCL = Child Behavior Check List Standardised scores; VABS = Vineland Adaptive Behavior Scales Standardised scores.

  • a

    Twin-to-twin transfusion subset matched with singleton controls.

  • b

    Twin-to-twin transfusion subset matched with twin controls.

  • #

    P value determined from Wilcoxon two-sample test (regional cohort) and Wilcoxon matched pairs signed ranks test (matched controls).

CBCL
n43a26b11914326
Externalising43 (39, 53)44 (38, 53)47 (40, 55)51 (41, 57)50 (41, 62)
Externalising difference  2 (−1, 6)3 (−3, 8)1 (−7, 8)
P  0.2220.3520.834
Total44 (38, 53)45 (37, 54)49 (43, 57)49 (41, 58)52 (37, 60)
Total difference  3 (−1, 6)3 (−3, 9)1 (−6, 7)
P  0.2160.3120.816
 
VABS
n432710454327
Communication100 (90, 107)93 (53, 124)96 (83, 106)95 (80, 106)91 (66, 107)
Communication difference   −4 (−23, 12)0 (−42, 23)
P#  0.8030.1350.252
Daily living skills89 (77, 98)95 (75, 118)84 (72, 92)81 (75, 90)88 (81, 96)
Daily living skills difference   −6 (−1, 5)3 (−2, 19)
P#  0.4210.0960.686
Socialisation93 (84, 100)97 (75, 123)94 (86, 100)94 (89, 101)94 (85, 101)
Socialisation difference   1 (−9, 15)0 (−8, 19)
P#  0.8760.8100.804

There was an antenatal demise of one twin with a single surviving twin in six of the twin-to-twin transfusion syndrome cases. Within this group one child, delivered at 26 weeks, was lost to follow up. A further child, delivered at 33 weeks, has cerebral palsy with the remaining four single surviving twins having normal IQ scores [median IQ 105 (101, 109.5)] and behavioural assessments. In all cases, the surviving twin was designated antenatally as the recipient fetus. The four normal single surviving twins all were stage 1 at diagnosis, with the cerebral palsy single surviving twin being stage 3 at diagnosis (the donor with absent end diastolic flow in the umbilical artery).

DISCUSSION

This is the first study to report a broad spectrum of long term outcomes in childhood survivors of twin-to-twin transfusion syndrome. It is also the first to contrast these outcomes in a rigorous manner with a contemporaneous regional comparison group. Children born very preterm in the twin-to-twin transfusion syndrome group had significantly lower IQ scores than non-twin-to-twin transfusion syndrome twins or singletons who were also born very preterm. There were no differences in children born at more mature gestational ages and no differences in overall behaviour scores or in adaptive behaviour scores at any gestational age. There was no increase in the prevalence of cerebral palsy in twin-to-twin transfusion syndrome children.

Previous studies of long term childhood outcome have reported variable rates for cerebral palsy (4.7–18%)8–12 but have not had a comparison group to ascertain if their reported rate is higher than expected at equivalent gestational ages in their own community. Our results show a trend to higher cerebral palsy rates in twin-to-twin transfusion syndrome children and non-twin-to-twin transfusion syndrome twins than in singletons at similar gestations. This would be consistent with other evidence showing higher rates in twin pregnancies.22,23 To demonstrate a statistically significant increase in cerebral palsy in twin-to-twin transfusion syndrome children compared with other twins would require a very large population of twin-to-twin transfusion syndrome children. It is likely that this would only be available in national registries of twin-to-twin transfusion syndrome pregnancies.

Assessments of developmental or cognitive outcome have previously been based on use of the Griffith's test8–11 with no formal assessment of intelligence and no appropriate comparison group. Thus, most of these studies have concluded that childhood survivors of twin-to-twin transfusion syndrome have a high rate of adverse long term outcomes. They have however been unable to assess whether this is due to their early gestational age at delivery or is a function of the pathological process involved in twin-to-twin transfusion syndrome and is hence an additional risk in these survivors. Our study would suggest that there are additional risks to development from being a member of a twin-to-twin transfusion syndrome pregnancy, particularly with the decrement in IQ score that we have observed in those born at very preterm gestations. Combining a cohort comparison and a case–control cohort makes us confident that we are describing a true difference in IQ scores. The estimate of this IQ difference would indicate that it would be of practical significance in these children in later life.

Necrosis of cerebral white matter has been described in preterm monochorionic twins and is related to vascular connections between their placental circulations.24 It was more common in those infants with functioning placental intravascular anastomoses. The infants in that study were all less than 36 weeks of gestational age but no breakdown of gestation by white matter lesion was given. This may be the underlying pathological process that leads to the adverse cognitive outcome in our very preterm infants. Thus, our results would suggest that, although the insult is antenatal, delivery at a very preterm gestation makes the brain more susceptible to damage.

The occurrence of a single intrauterine death in monochorionic twin pregnancies has been associated with a risk of neurologic sequelae in the survivor of 17–46.2%.25–27 Bajoria et al.28 reported the risk of adverse sequelae in twin-to-twin transfusion syndrome pregnancies to be related to which twin died first. Intrauterine demise of the recipient was associated with an increased frequency of subsequent co-twin death, severe anaemia and intracranial lesions, while demise of the donor was associated with a normal outcome in the recipient. Mari et al.12 did not observe this phenomenon with only one of three recipient survivors assessed as neurologically normal. One of our five single surviving assessed twins has neurologic handicap and it is likely that the placental vascular anastomoses and secondary haemodynamic alterations following single demise contribute to these outcomes, independent of fetal recipient/donor status.

There were no differences in behavioural scores between twin-to-twin transfusion syndrome children and the comparison group either in the multivariate analysis or in the case–control analysis. This is the first report of behavioural outcomes in twin-to-twin transfusion syndrome children and is important, as it suggests no increase in behaviours that might relate to later problems with attention or hyperactivity. The scores on the VABS also suggest that these children function appropriately in their home environment. This positive finding however relates only to the early years of life as few of these children have reached school age. It will be important that they are reviewed later to examine for deleterious effects on learning and performance at school.

We found no differences between donor and recipients in cognitive or behavioural outcomes and this is consistent with results in the published literature.10,11 Similarly, we found no relationship with twin-to-twin transfusion syndrome stage for any outcome.

CONCLUSION

Most of the children from this cohort of twin-to-twin transfusion syndrome pregnancies are functioning within the normal range on their cognitive and behavioural assessments. Of the seven children with disability, three were considered to have severe disability based on IQ, with one of these children also having mild cerebral palsy. This is despite the major adverse events they have experienced antenatally and perinatally. This knowledge is critical to parents when they are being counselled and advised at the time of their initial presentation during pregnancy with twin-to-twin transfusion syndrome.

Acknowledgements

This study was supported by a research grant from the Women and Infants Research Foundation, Perth, Western Australia.

Accepted 11 May 2004

Ancillary