Prognosis for the co-twin following single-twin death: a systematic review

Abstract

Background  Following single-twin death, the perinatal mortality and morbidity for the surviving co-twin is increased but difficult to quantify. We present data on prognosis from a systematic review.

Objectives  We aimed to determine the incidence of a) co-twin death, b) neurological abnormality and c) preterm delivery for the surviving co-twin following single-twin death after 14 weeks of gestation.

Search strategy  Literature was identified by searching two bibliographical databases and specialist journals between 1990 and 2005.

Selection criteria  The selected studies of ≥5 cases reported on perinatal death and/or neurodevelopmental delay of the surviving co-twin.

Data collection and analysis  Studies were assessed for quality and data extracted to allow computation of rates. The data were inspected for heterogeneity using a Forrest plot and examined statistically using the chi-square test. Data from individual studies were pooled within subgroups defined by prognosis.

Main results  The search strategy yielded 632 potentially relevant citations. Full manuscripts were retrieved for 54 citations and 28 studies were finally included in the review. Following the death of one twin, the risk of monochorionic and dichorionic co-twin demise was 12% (95% CI 7–11) and 4% (95% CI 2–7), respectively. The risk of neurological abnormality in the surviving monochorionic and dichorionic co-twin was 18% (95% CI 11–26) and 1% (95% CI 0–7), respectively. The risk of preterm delivery was 68% (95% CI 56–78) and 57% (95% CI 34–77), respectively. Where there was comparative data within studies, the odds of monochorionic co-twin intrauterine death was six times that of dichorionic twins (OR 6.04 [95% CI 1.84–19.87]). Neurological abnormality was also higher in monochorionic compared with dichorionic pregnancies (OR 4.07 [95% CI 1.32–12.51]).

Author’s conclusions  More prospective research is required to inform decision making on this subject, especially with data that allow stratification based upon chorionicity.

Introduction

Intrauterine death of one fetus in a twin pregnancy is uncommon in the second or third trimester. However, the consequences to the surviving co-twin can be profound, especially in monochorionic twins. These may include co-twin death, survival with cerebral impairment or preterm labour with its sequelae.¹ Recent changes in technology and prenatal ultrasound scan have meant that clinicians can diagnose this condition and have the option to intervene. Decision making should be informed by existing prognostic evidence, but individual studies including data from case reports, follow up of cohorts and twin registries tend to have imprecise results as the event is uncommon. More precise information may be obtained by pooling these results statistically in a meta-analysis. A comprehensive systematic review is required to capture the literature scattered across many journals. Thus far, existing reviews^2,3 have not been systematic and this may be because the methodology of meta-analysis involving observational studies is not widely disseminated. We conducted such a review to provide a reliable estimate of prognosis for the co-twin following single-twin death after the first trimester of pregnancy.

Methods

The project was based on a prospective protocol developed using widely recommended methods for systematic review of observational studies.^4,5

Sources and study selection

Literature was identified by searching the bibliographic databases MEDLINE and EMBASE between 1990 and 2005 inclusive. The search term combination captured citations with the relevant population (e.g. single-twin death; surviving twin) using a combination of MeSH and text words. A hand search of relevant literature was also conducted from specialist journals. There was no language restriction in search or selection.

Our study selection criteria were:

• • 
  Population: Twin pregnancies with death of one twin.
• • 
  Outcome: Perinatal death and/or neurodevelopmental delay of the surviving twin.
• • 
  Study design: Series with follow up of ≥5 cases of twin pregnancies with death of one twin.

From the electronic searches, full manuscripts of all the potentially relevant citations were obtained. The final inclusion/exclusion decisions were made after evaluation of these by two reviewers (S.O. and M.K.). In cases of data duplication (i.e. the same data published in two or more reports), only the most recent and complete report was included.

We followed a deliberate policy of excluding studies of twin pregnancies in which a ‘disappearing sac’ was suspected in the first trimester of pregnancy.

Quality assessment and data extraction

One reviewer (S.O.) extracted data from all articles meeting the selection criteria including data on features of methodological quality. This was double checked by another reviewer (M.K.). The studies were assessed for quality by the following criteria derived from existing check-lists:⁶

• • 
  Data collection: Prospective collection of data was considered ideal, retrospective collection was considered second best.
• • 
  Description of population: A well-defined sample at a uniform early stage with clear documentation of chorionicity and gestational age of single-twin death was considered ideal.
• • 
  Prognostic factors considered: Clear documentation of consideration of prognostic co-factors, including chorionicity, gestation of single-twin death, presence of twin–twin transfusion syndrome, discordant congenital anomalies and concurrent maternal illness, was considered ideal.
• • 
  Length of follow up: Follow up beyond 4 weeks following delivery was considered ideal.
• • 
  Objective outcome: Outcomes including co-twin death and validated measures of neurological abnormality of co-twin survivors were considered ideal.
• • 
  Outcome ascertainment: Greater than 90% follow up of the original study population was considered ideal, less than 90% was considered second best.

Data extraction sought information regarding co-twin intrauterine (or neonatal) death, neurological abnormality by 4 weeks following delivery in co-twin survivors and preterm delivery before 34 weeks of gestation. Data were extracted to allow computation of rates, stratifying by chorionicity and other prognostic factors wherever possible.

Data synthesis

The extracted data were tabulated to allow qualitative inspection for clinical and methodological heterogeneity. Heterogeneity of rates was explored graphically using Forrest plot and examined statistically using the chi-square test.⁷ Data from individual studies were pooled within subgroups defined by prognosis and outcome.⁸

Results

Identification of the literature

The electronic search yielded a total of 632 citations, of which 54 were considered potentially relevant. Examination of the full manuscripts revealed that 26 did not meet the selection criteria. This was mostly because the case series reported less than five cases. Thus, a total of 28 primary studies were selected for review^1,2,9–34 (Figure 1, Table 1). One study was derived from data obtained from three separate institutions.¹² All other studies published data either from follow up of case series (cohort studies) or from registries.

Figure 1. Flow diagram of study selection for systematic review. Outcome of surviving co-twin following single-twin death after 14 weeks of gestation.

Table 1.  Studies selected for systematic review and availability of data for extraction and analysis

Author	Number of cases	Data extractable
		In utero death	Neurological abnormality	Preterm delivery
N, no; Y, yes.
Case series
Abdal-Khalig and Sobande^33	35	Y	Y	N
Aslan et al.^11	25	N	Y	Y
Axt et al.^10	7	Y	Y	Y
Eglowstein and D’Alton^30	20	Y	Y	Y
Fusi and Gordon^21	16	Y	Y	Y
Gaucherand et al.^26	9	Y	Y	Y
Ishimatsu et al.^32	15	Y	Y	Y
Jou et al.^29	12	Y	Y	Y
Kilby et al.^15	20	Y	Y	N
Malinowski et al.^23	15	N	N	N
Malinowski et al.^9	11	Y	N	N
Lin et al.^24	17	N	Y	Y
Liu et al.^34	3	N	N	N
Petersen and Nyholm^16	12	Y	Y	Y
Prompeler et al.^17	43	N	N	N
Saito et al.^18	30	N	Y	Y
Santema et al.^19	29	Y	Y	N
Sonneveld and Correy^31	25	Y	N	Y
van Heteren et al.^20	11	N	Y	Y
Wang et al.^25	9	N	Y	Y
Woo et al.^2	7	Y	Y	Y
Zorlu et al.^22	9	Y	N	N
Birth register
Rydhstrom and Ingemarsson (1993)^27	206	Y	N	N
Rydhstrom (1994)^28	326	Y	N	N
Pharoah and Adi^1	597	Y	N	N
Glinianaia et al.^13	164	Y	N	N
Johnson and Zhang^14	3599	Y	N	N
Other
Bajoria et al.^12	92	Y	N	N

Study characteristics and quality

Study characteristics and population are shown in Table 1. Data from birth registries were extractable for in utero death of the co-twin but not for neurological abnormality or preterm delivery. Quality of studies summarised in Figure 2 showed that data collection was retrospective in all studies; chorionicity and gestational age of single-twin death was reported in 82% (23/28) and 64% (18/28) of studies, respectively; and presence or absence of twin–twin transfusion syndrome, discordant congenital anomalies and concurrent maternal illness was reported in 46% (13/28), 54% (15/28) and 33% (9/28) of studies, respectively.

Figure 2. Quality assessment of studies in systematic review.

The risk of co-twin death in utero

Preliminary analysis to assess the risk of co-twin death involved 20 studies (4503 twin pregnancies). The preliminary analysis was highly heterogeneous (P= 0.00) for dichorionic twins, with the heterogeneity contributed by a single study. This study by Johnson and Zhang¹⁴ involved data from a birth register of unlike-sex twins. In order to compute numbers of co-twin death, we had to convert percentages to numbers without the benefit of the availability of raw data. Analysis revealed that this study contributed to more than one-quarter of heterogeneity (I²= 96.5% decreasing to I²= 71.6% after exclusion of this study). Accordingly, we excluded the data from Johnson and Zhang.¹⁴

In the final analysis, to assess the risk of co-twin death, data were derived from 19 studies (904 twin pregnancies) (Figure 3). The risk of monochorionic and dichorionic co-twin death was 12% (95% CI 7–18) (heterogeneity P= 0.02) and 4% (95% CI 2–7) (heterogeneity P= 0.74), respectively. Where there was comparative data within studies, the odds of monochorionic twin death following single-twin death after 20 weeks of gestation was six times higher compared with dichorionic twins (five studies; OR 6.04 [95% CI 1.84–19.87]; heterogeneity P= 0.56) (Figure 4).

Figure 3. Forrest plots of outcomes for the surviving co-twin following single-twin death according to chorionicity.

Figure 4. Meta-analysis of studies providing information of co-twin death, neurological abnormality and preterm delivery following single-twin death after 14 weeks of gestation. A comparison of odds between monochorionic and dichorionic pregnancies.

The risk of neurological sequelae

To assess the risk of neurological abnormality, data were derived from 17 studies (267 twin pregnancies) (Figure 3). The risk of neurological abnormality in the surviving monochorionic and dichorionic co-twin was 18% (95% CI 11–26) (heterogeneity P= 0.45) and 1% (95% CI 0–7) (heterogeneity P= 0.81), respectively. Where there was comparative data within studies, the odds of neurological abnormality in monochorionic survivors following single-twin death after 20 weeks of gestation was four times higher compared with dichorionic survivors (eight studies; OR 4.07 [95% CI 1.32–12.51]; heterogeneity P= 1.00) (Figure 4).

Preterm delivery

To assess the risk of preterm delivery, data were derived from 11 studies (100 twin pregnancies) (Figure 3). The risk of preterm delivery in monochorionic and dichorionic pregnancies was 68% (95% CI 56–78) (heterogeneity P= 0.52) and 57% (95% CI 34–77) (heterogeneity P= 0.40), respectively. These figures for preterm delivery include both iatrogenic and spontaneous preterm delivery. Preterm delivery before 34 weeks of gestation appeared to be marginally higher in monochorionic pregnancies compared with dichorionic pregnancies although this was not statistically significant (six studies; OR 1.91 [95% CI 0.70–5.21] heterogeneity P= 0.20) (Figure 4).

Management strategy

In the majority of studies, data were not available concerning the type of management strategy employed. Where the pregnancy was monochorionic, a conservative approach following single-twin death resulted in a 20% (12/60) intrauterine or neonatal loss by rate by 4 weeks after delivery. Where the pregnancy was dichorionic, this figure was 13% (5/38). There was insufficient data concerning immediate delivery following single-twin death.

Discussion and conclusion

Single fetal death in a twin pregnancy is known to be a serious complication of pregnancy. It is a relatively rare complication of multiple pregnancies (5% of all twin pregnancies)¹⁵ but may carry with it an increased risk of perinatal morbidity and mortality. The main findings in this systematic review are that following the death of one twin after the first trimester, the odds of intrauterine death of the co-twin and neurological abnormality among survivors was six and four times higher in monochorionic compared with dichorionic pregnancies.

The main strength of this review is that it employed an exhaustive research strategy. In this way, we were able to assemble evidence for a condition that is rare and is imprecisely assessed in individual studies. In addition, the quality of these studies was assessed together with stratification of results according to chorionicity. However, the number of studies where chorionicity has been indicated is relatively small. It was also not possible to identify the relationship of chorionicity to zygosity in the multiple pregnancy cohorts described.

This systematic review includes more monochorionic than dichorionic twins in the analysis of comparative data within studies. This is a potential source of ascertainment bias given that in a general population, two-thirds of twin pregnancies are dichorionic. Furthermore, many of the studies were small case series. Publication bias is an issue where individuals with a case series of successful outcomes would be more likely to report than other individuals with a similar case series of poor outcomes.

These data do though provide clinicians and patients with contemporary and reliable estimates of outcomes regarding prognosis following single fetal death in a twin pregnancy. However, due to poverty of reporting in individual studies, these data cannot reliably guide clinicians regarding management, particularly issues concerning the timing of delivery remain unexplored.

There are two theories that have been advanced to explain multicystic encephalomalacia and co-twin death in monochorionic pregnancies. The first is that there is passage of thrombotic material from the dead to healthy twin following derangement in coagulation due to the death of one twin.³⁵ The second theory is the ‘haemodynamic imbalance theory’. This states that the placental anastomoses (frequently present in monochorionic placentas) allow transfer of blood from the surviving twin to the dead co-twin giving rise to periods of hypoperfusion, hypotension and acute fetal anaemia, resulting in neurological damage.³⁶ In this systematic review, where there was comparative data within studies, there were only two deaths and one case of neurological injury in known dichorionic twins.

It is clear from this review that data has thus far been poorly reported and that this area needs further robust research. There is a need for population-based data relating to this obstetric complication with pregnancies being classified prospectively by chorionicity. In the UK, there are a growing number of regional congenital anomaly databases that report the incidence and prevalence of congenital anomalies and relate this to denominator data for the background number of maternities. If multiple pregnancies were reported and chorionicity was also recorded, robust population data would be available by which to assess further these complications.