Second-trimester discordance and adverse perinatal outcome in twins: the STORK multiple pregnancy cohort

Authors

  • F D'Antonio,

    1. Fetal Medicine Unit, Division of Developmental Sciences, St George's, University of London, London, UK
    Search for more papers by this author
  • A Khalil,

    Corresponding author
    1. Fetal Medicine Unit, Division of Developmental Sciences, St George's, University of London, London, UK
    • Correspondence: Dr A Khalil, Fetal Maternal Medicine Unit, St George's, University of London, London SW17 0RE, UK. Email asma.khalil@stgeorges.nhs.uk

    Search for more papers by this author
  • B Thilaganathan,

    1. Fetal Medicine Unit, Division of Developmental Sciences, St George's, University of London, London, UK
    Search for more papers by this author
  • on behalf of the Southwest Thames Obstetric Research Collaborative (STORK)


Abstract

Objective

The aim of this study was to determine the association between biometry discordance at the time of the anomaly scan and adverse perinatal outcomes in twin pregnancies.

Design

Retrospective cohort study.

Setting

Nine hospitals in the Southwest Thames Region of London Obstetric Research Collaborative (STORK).

Population or sample

Population multicentre retrospective study of all twin pregnancies booked for antenatal care in nine hospitals over a period of 10 years. Methods Logistic regression and receiver operating characteristic (ROC) curve analyses were performed to evaluate the association between abdominal circumference (AC) and estimated fetal weight (EFW) discordance, recorded between 20 and 22 weeks of gestation, and adverse pregnancy outcomes.

Main outcome measures

Stillbirth, neonatal mortality, preterm birth (PTB) at <34 weeks of gestation, and birthweight (BW) discordance ≥25%.

Results

A total of 2399 twin pregnancies [457 monochorionic (MC) and 1942 dichorionic (DC)] were included in the study. The predictive accuracy of the EFW discordance was poor for fetal loss after 22 weeks of gestation (area under the curve, AUC 0.54, 95% CI 0.46–0.64), fetal loss beyond 28 weeks of gestation (AUC 0.42, 95% CI 0.31–0.52), perinatal loss (AUC 0.51, 95% CI 0.44–0.57), BW discordance (AUC 0.63, 95% CI 0.56–0.65), and PTB before 34 weeks of gestation (AUC 0.52, 95% CI 0.49–0.55). There was no significant difference in the prediction of these outcomes when using EFW discordance or AC discordance.

Conclusions

Once structural malformations, chromosomal abnormalities, and twin-to-twin transfusion syndrome have been excluded, second-trimester EFW and AC discordance have poor predictive value for adverse perinatal outcomes in twin pregnancy.

Introduction

Twin pregnancies are at higher risk of perinatal mortality than singletons, mainly from fetal growth restriction and preterm birth (PTB).[1] A priority in the management of twin pregnancy is to identify pregnancies at risk of adverse outcome, so as to allow the appropriate delivery of care, to optimise pregnancy outcome, and to tailor parental expectations. The early identification of at-risk twin pregnancies has relied significantly on the assessment of inter-twin crown–rump length (CRL) discordance because of the association between CRL discordance and adverse twin pregnancy outcomes.[2-17] More recently, however, systematic large-scale studies have shown that despite the latter association, early pregnancy CRL discordance is of limited value in predicting poor pregnancy outcomes.[16, 17] The main limitation of CRL discordance appears to be that it is only weakly associated with adverse outcome, and is biased towards identifying twin pregnancy losses that occur before 20 weeks of gestation, usually as a consequence of fetal abnormality.[2, 4-6, 14, 16]

Studies in singleton pregnancies have reported that second-trimester fetal biometry is associated with adverse obstetric and neonatal outcomes, including perinatal death, small-for-gestational-age neonates, and PTB.[18-20] Furthermore, fetal weight discordance estimated by ultrasound in the third trimester has been found to reliably predict birthweight discordance and perinatal loss in twins.[21] However, the association between second-trimester ultrasound biometry and adverse pregnancy outcome in twins has not been fully investigated in large cohort studies.[22-28] The aim of this study was to ascertain the performance of second-trimester ultrasound biometry in the prediction of adverse perinatal outcomes in twin pregnancies.

Methods

This was a retrospective study of all twin pregnancies booked for antenatal care in nine hospitals in the Southwest Thames Region of London Obstetric Research Collaborative (STORK), over a period of 10 years, from 2000. All women registering for routine antenatal care by 11 weeks of gestation were considered suitable for the analysis. Scan data were obtained by a computerised search of the obstetric ultrasound database at each hospital, and the outcome details were obtained from computerised maternity records. These two databases were cross-checked to ensure full data capture of all twin pregnancies during the study period. Ultrasound and outcome data from these databases were matched to the mandatory national register for stillbirth and neonatal losses provided by the former Centre for Maternal and Child Enquires (CMACE). In accordance with CMACE regulations, patient identifiers such as name, hospital number, and date of birth were not made available to the researchers. All data included in the analysis were collected prospectively but were analysed retrospectively. Ethics approval for this retrospective study was obtained from the local research ethics committee. Terminations of pregnancy, structural or chromosomal abnormalities, pregnancies of unknown chorionicity, monochorionic, monoamniotic and higher order multiple gestations were not included in the analysis.

Gestational age was determined by the CRL of the larger twin at the 11–14 week scan, or by head circumference (HC) if assessed after 14 weeks of gestation.[29, 30] Chorionicity was determined by ultrasound evaluation according to the number of placentas and the presence of the lambda or T-signs, and was confirmed after birth.[31] A routine fetal structural survey was carried out at 20–22 weeks of gestation, and all monochorionic twins had two additional scans at around 17 and 19 weeks of gestation, specifically to identify early features of twin-to-twin transfusion syndrome (TTTS).[32] If TTTS was suspected, women were referred to the local tertiary centre for assessment for fetoscopic laser ablation of the placental interconnecting vessels. All scans were performed by qualified experienced sonographers.

Ultrasound estimated fetal weight (EFW) was calculated using the Hadlock formula, based on the head circumference (HC), abdominal circumference (AC) and femur length (FL).[33] Ultrasound EFW discordance was calculated as 100 × (larger EFW – smaller EFW)/larger EFW, whereas actual birthweight (BW) discordance was calculated as 100 × (larger BW – smaller BW)/larger BW, and discordance in abdominal circumference (AC) was calculated as 100 × (larger AC – smaller AC)/larger AC. Only the ultrasound examinations at the time of the routine anomaly scan, between 20 and 22 weeks of gestation, were considered for this analysis. Using EFW and AC discordance, the likelihood of the following outcomes was assessed using receiver operating characteristic (ROC) curve analysis: fetal loss beyond 22 weeks of gestation, fetal loss beyond 28 weeks of gestation, perinatal loss (defined as stillbirth, and early and late neonatal death, up to 28 days of life), BW discordance ≥25%, and PTB as birth between 24 and 34 weeks of gestation. The choice of a cut-off of 25% for BW discordance was justified by the fact that this threshold has been suggested to have the best predictive accuracy in detecting perinatal death during the third trimester of pregnancy.[21]

Logistic regression analysis was carried out in order to assess the association between pregnancy characteristics, such as chorionicity, gestational age at delivery, EFW discordance, AC discordance, EFW < fifth centile in at least one twin,[34] CRL discrepancy, and the adverse pregnancy outcomes analysed in this cohort study. A search of PubMed and EMBASE was made to identify previous citations in the English literature reporting the association between discordance, either in EFW or in AC, in the second trimester and adverse perinatal outcomes in twin pregnancies. We used medical subject heading (MeSH) terms and keywords related to second-trimester ultrasound biometry and twin pregnancies, including ‘EFW discordance’, ‘EFW discrepancy’, ‘AC discordance’, ‘AC discrepancy’, ‘early pregnancy growth’, ‘second trimester growth’, ‘second trimester ultrasound’, ‘twin’, ‘multiple pregnancy’, ‘monochorionic’, ‘dichorionic’, ‘chorionicity’, ‘adverse outcome’, ‘stillbirth’, ‘perinatal mortality’, ‘fetal loss’, ‘intrauterine death’, ‘neonatal death’, ‘birthweight discordance’, ‘growth restriction’, ‘preterm birth’, ‘preterm delivery, and ‘preterm labour’.

The statistical analysis was performed using spss 15.0 (SPSS Inc., Chicago, IL, USA) and prism 5.04 for windows (GraphPad Software, La Jolla, CA USA). Statistical significance was set at P < 0.05. All P values were two-tailed.

Results

A total of 2399 twin pregnancies were included in the study, of which 457 were monochorionic (MC) and 1942 were dichorionic (DC). The fetal loss rate beyond 22 weeks of gestation and total perinatal loss rate were 2.1% (n = 61) and 3.4% (n = 82), respectively. The rate of PTB at <34 weeks of gestation was 16.1% (n = 387), and the rate of BW discordance of ≥25% was 12.1% (n = 290). The study population characteristics and outcomes for MC and DC pregnancies are shown in Table 1.

Table 1. Study population characteristics and outcomes according to chorionicity
Pregnancy characteristics and outcomesDC twins (n = 1942)MC twins (n = 457)P value
  1. Values are given as median (IQR) or number (%).

EFW discordance (%)10.69 (4.89–20.41)10.71 (4.86–21.05)0.002
AC discordance (%)5.27 (2.35–10.05)5.29 (2.36–10.06)0.006
Fetal loss41 (2.1)20 (4.4)0.009
Perinatal loss58 (3.0)24 (5.3)0.024
Birthweight discordance ≥25%210 (10.8)80 (17.5)<0.001
Preterm birth <34 weeks of gestation290 (14.9)97 (21.2)0.001

The predictive accuracy of EFW discordance was poor for fetal loss beyond 22 weeks of gestation, fetal loss beyond 28 weeks of gestation, perinatal loss, BW discordance of ≥25%, and PTB at <34 weeks of gestation (Table 2). No clear threshold could be identified for any of the outcomes analysed. The use of AC discordance compared with EFW discordance (Figure 1; Table 2) did not result in any significant difference in the predictive accuracy for all adverse outcomes. When the area under the curve (AUC) curve analysis was restricted to only MC twins, the predictive accuracy using EFW or AC discordance remained poor for all adverse outcomes (Table 2).

Table 2. Receiver operating characteristic (ROC) curves for the prediction of adverse pregnancy outcomes, using estimated fetal weight (EFW) and abdominal circumference (AC)
OutcomeEFW discordance (all twins)AC discordance (all twins)EFW discordance (MC only)AC discordance (MC only)
  1. The table shows the area under the curve (AUC) values and their 95% confidence intervals (95% CIs).

Fetal loss >22 weeks0.54 (0.46–0.64)0.54 (0.46–0.54)0.50 (0.37–0.62)0.54 (0.47–0.60)
Fetal loss >28 weeks0.42 (031–0.52)0.51 (0.31–0.54)0.41 (0.26–0.62)0.51 (0.40–0.62)
Perinatal loss0.51 (0.44–0.57)0.54 (0.44–0.57)0.42 (0.29–0.54)0.53 (0.47–0.60)
BW discordance0.63 (0.56–0.65)0.61 (0.58–0.63)0.61 (0.50–0.71)0.61 (0.58–0.63)
PTB <34 weeks0.52 (0.49–0.55)0.53 (0.49–0.56)0.50 (0.44–0.56)0.53 (0.49–0.56)
Figure 1.

Receiver operating characteristic (ROC) curves for the prediction of adverse pregnancy outcomes, including fetal loss beyond 22 weeks of gestation, fetal loss beyond 28 weeks of gestation, perinatal loss, birthweight (BW) discordance ≥25%, and preterm birth (PTB) before 34 weeks of gestation in twin pregnancies, using estimated fetal weight (EFW) and abdominal circumference (AC) discordance (EFW discordance: dotted line).

Multiple regression analysis demonstrated that for fetal loss, a significant independent contribution was provided by chorionicity. In the case of perinatal loss, a significant independent contribution was provided by gestational age at delivery. EFW discordance was independently associated with BW discordance. Finally, chorionicity was significantly associated with PTB at <34 weeks of gestation (Tables 3 and S1).

Table 3. Results of the multivariate regression analysis for the adverse perinatal outcomes explored in the study
VariableOR (95% CI)P value
Fetal loss beyond 22 weeks of gestation
Chorionicity2.14 (1.27–3.62)0.005
Perinatal loss
GA at delivery0.79 (0.75–0.82)<0.0001
Birthweight discordance ≥25%
EFW discordance at second trimester1.04 (1.02–1.06)<0.0001
Preterm birth before 34 weeks of gestation
Chorionicity1.53 (1.18–1.98)0.001

The review included seven studies that examined the role of fetal EFW or AC discordance in the second trimester and adverse perinatal outcomes in twin pregnancies (Table 4).[22-28] The studies included have different thresholds of EFW or AC discordance. The largest study included 503 cases, and reported that AC discordance has a predictive value for BW discordance (78% sensitivity and 76% specificity in MC twins; 40% sensitivity and 86% specificity in DC twins).[25]

Table 4. Studies reporting data on second-trimester estimated fetal weight (EFW) or abdominal circumference (AC) discordance at the time of the anomaly scan and adverse pregnancy outcomes
AuthorChorionicityCasesDiscordance cut-offGestational age (week)Deaths n (%)Fetal lossPerinatal lossBW discordance/sIUGRPTB
  1. BW, birthweight; sIUGR, selective intrauterine growth restriction; PPV, positive predictive value; PTB, preterm birth.

  2. a

    Different gestational age periods analysed in the study.

  3. b

    Results reported as correlation between continuous variables or comparison between means, medians, and/or proportions.

  4. c

    Different means provided for difference in biparietal diameter, head circumference, head area, transverse abdominal diameter, AC, abdominal area, and femur length, according to chorionicity and zygosity.

  5. d

    A significant association was found in the difference of all intrapair measurements, except AC in patients with BW discordance ≥15%, compared with those with BW discordance ≤15%.

  6. e

    Calculated from the original cohort of 781 twin pregnancies.

  7. f

    Corresponding to the 90th centile in the population analysed.

Rodis et al.[22]MC and DC8514–24aAssociated with BW discordance >20% and sIUGRb
Grobman and Parilla[23]MC and DC4420% (EFW discordance)20–24aPredictive of BW discordance ≥20% (PPV 69%) and IUGR (PPV 77%)
Charleimaine et al.[24]MC and DC75c18, 23aNot associatedd
Klam et al.[25]MC and DC5037% (AC discordance) 25% (EFW discordance)16–23a31 (4.0)eAssociatedbPredictive of BW discordance (78% sens, 76% spec in MC twins, 40% sens, 86% spec in DC twins)Associatedd
Van Mieghem et al.[26]MC only6025% (EFW discordance)16, 20aPredictive of BW discordance >25% (AUC 0.87)
Hoopman et al.[27]MC and DC19618–25Not associated
Fox et al.[28]MC and DC30615% (EFW discordance)f18–242 (0.7)Not associatedAssociated with BW discordance >20% in MC and DC twinsbNot associated
Current study (2012)MC and DC239920–2282 (3.4)Not associatedNot associatedNot predictiveNot associated

Discussion

Main findings

Our data demonstrate that second-trimester ultrasound discordance in twins has a poor predictive value for adverse perinatal outcomes in twin pregnancies, irrespective of the chorionicity.

Strengths and limitations

The STORK cohort examined here is the largest series to date assessing the value of second-trimester ultrasound discordance in predicting adverse outcomes in twin pregnancies.

A limitation of our study is the lack of data on maternal characteristics, which are potential risk factors for adverse pregnancy outcome; however, maternal factors are not known to influence the parameter of interest in this study, which is the discordance in fetal biometry in the second trimester. We therefore feel it unlikely that such maternal factors would be potential confounding variables in the relationship between discordance in fetal biometry and adverse pregnancy outcomes.

Interpretation

Birthweight (BW) and EFW discordance are independent predictors of perinatal outcome in twin pregnancies.[21]

Weight discordance is unique and invariably presents in all twin pregnancies, and may represent a physiological condition or an adaptive measure to the intrauterine environment. On the other hand it may be the result of pathological conditions involving the fetus or the placenta.[35, 36] Severe growth discordance in twins is associated with an increased risk of perinatal loss and morbidity.[35, 37-44] Previous studies have reported different thresholds of EFW discordance for the prediction of fetal loss and adverse perinatal outcomes (Table 4).

Our findings differ from those reported in singleton pregnancies, which suggest that the presence of an EFW of <25th centile at the second-trimester ultrasound scan was associated with poor obstetric and neonatal outcomes in singleton pregnancies.[20]

The pathophysiology of fetal growth restriction in multiple gestations differs from that of singletons. The uterine milieu is usually able to supply the metabolic demands of both twins during the second and early third trimester, until approximately 28–32 weeks of gestation, after which twin growth usually diverges from that of singletons.[36, 45] Moreover, even in cases of severely discordant growth, the uteroplacental unit supply is 50–75% more than the average for gestation in singletons, and it has been hypothesised that growth discordance may serve as an adaptive measure to promote maturity by reducing uterine over-distension. This suggests that multiple pregnancies can be considered as growth promoted, as the uterine environment produces a greater fetal mass compared with singleton pregnancies.[36, 45] The current study findings suggest that discordant growth at the time of the anomaly scan is unlikely to occur as the result of dysfunction in the uteroplacental unit.

In our previous study, where we examined the role of CRL discordance in predicting adverse pregnancy outcomes, we reported a poor predictive value.[17] The perinatal loss rate reported in the latter article was lower compared with the current study [1.1% (n = 23) versus 3.4% (n = 82), respectively]. This discrepancy is likely to result from the fact that the current study included women at higher risk who booked late for pregnancy care and had their first scan after 14 weeks of gestation, when second-trimester fetal biometry was undertaken instead of CRL measurement. Stratification of the risk of adverse pregnancy outcome in twin pregnancies, in order to plan for appropriate care, relies on the findings of the first-trimester scan, and emphasises the importance of the current study findings, in particular for women who miss their first-trimester scan.

Only a few studies have looked at the role of biometry discordance, in either EFW or AC, at the time of the routine anomaly scan and adverse outcomes in twin pregnancies (Table 4).[22-28] The major limitation of these studies has been the very small numbers of twin pregnancies studied and the subsequent inevitably small numbers of perinatal losses observed. Hence, most authors have used EFW to predict actual BW discordance as a proxy for perinatal mortality and morbidity (Table 4).

Similar to CRL discordance, the rationale behind the use of second-trimester EFW discordance as a predictor of adverse outcomes is that discordant growth, which carries an increased risk of pregnancy loss, may begin in the first half of pregnancy. Monochorionicity was independently correlated with fetal loss beyond 22 weeks of gestation; however, even after adjusting for chorionicity, the predictive performance using either EFW or AC discordance did not improve. TTTS, which is the major contributor to mortality in MC twins, is a state of haemodynamic imbalance rather than placental insufficiency, even if a certain degree of discordance in size is always present. Our analysis included data at a gestational age period that represents the tail of the overall incidence of TTTS in MC twins. The relatively small number of cases complicated by TTTS in the remaining time window would theoretically reduce the impact of TTTS-related growth discordance on mortality. Second-trimester EFW discordance was found to be independently associated with BW discordance ≥25%; however, the predictive accuracy is poor, thus precluding its use in clinical practice.

According to the UK National Institute for Health and Care Excellence (NICE) guideline on the antenatal management of multiple pregnancies, there is no strong evidence that any composite screening strategy predicts intrauterine growth impairment in multiple pregnancies.[46] Therefore, the guidance is to scan twin pregnancies every 4 weeks to monitor fetal growth from 20 weeks of gestation, and to refer to a tertiary centre if the discordance in EFW is ≥25%. However, this recommendation is derived from evidence of low or moderate quality. The results of this current study, in addition to earlier studies, suggest that until a more effective screening strategy emerges, regular frequent ultrasound monitoring of twin pregnancies is necessary to identify those that would develop adverse pregnancy outcomes.[47]

Conclusion

In conclusion, this is the largest series to date to assess the value of second-trimester EFW and AC discordance in predicting adverse outcomes in twin pregnancies. Once structural malformations, chromosomal abnormalities, and TTTS were excluded, second-trimester ultrasound discordance was not predictive of subsequent adverse outcomes after 22 weeks of gestation in either MC or DC twins.

Disclosure of interests

None of the STORK authors have any conflicts of interest to declare.

Contribution to authorship

Study design: FDA, AK, and BT Study data collection and analysis: FDA and AK Writing the article: all authors.

Details of ethics approval

No ethical approval required.

Funding

Francesco D'Antonio is funded by University of Chieti, Italy, for a PhD in biomedical, clinical and experimental sciences.

Acknowledgements

STORK collaborative authors and contributors: Rosol Hamid, Mayday University Hospital; Hina Gandhi, East Surrey Hospital; Cheryl Ellis, Epsom General Hospital; Anne Deans, Frimley Park Hospital; Liz Peregrine and Andrew Breeze, Kingston Hospital; Renata Hutt, Royal Surrey Hospital; Amar Bhide and Aris T Papageorghiou, St George's Hospital; Adetunji Matiluko and Michael Egbor, St Helier's Hospital; Arash Bahamie, St Peter's Hospital; Faz Pakarian, Worthing.

Commentary on ‘Second-trimester discordance and adverse perinatal outcome in twins: the STORK multiple pregnancy cohort’

 The rate of multiple births has increased over the last 30 years, largely as a result of assisted reproductive technology and the increasing number of births to mothers of advanced maternal age (Black et al. Semin Fetal Neonatal Med 2010;15:306–12). Growth restriction has been identified as one of the attributable risk factors for increasing perinatal mortality in twin pregnancies (Chauhan et al. Am J Obstet Gynecol 2010;203:305–5).

Reports associating first-trimester intertwin crown–rump length discordance with adverse pregnancy outcomes (Harper et al. Ultrasound Obstet Gynecol 2012;41:627–31) have prompted several investigators, including those of the current study, to evaluate the association between discordance in second-trimester biometry and estimated fetal weight and pregnancy complications in twins.

The STORK multiple pregnancy cohort is a multicentre retrospective study using a superb internally validated database to evaluate the association between routine second-trimester biometric measurements and estimated fetal weight and fetal loss, perinatal loss, and preterm birth. The finding that these parameters are poor predictors of adverse outcome suggest that the association between growth restriction and perinatal outcomes in twins may result from insults occurring later in pregnancy, resulting in the disruption of the accelerated phase of fetal growth rather than at the time of the anatomy survey.

The findings of the study are different from those of smaller studies suggesting an association between discordant second-trimester estimated fetal weight or abdominal circumference and adverse pregnancy outcomes (Table 4). Interestingly, the present study found that chorionicity is an independent predictor of fetal loss and preterm birth, despite the overall poor performance of biometry as an adverse pregnancy predictor. The latter finding is consistent with a recent report focused on growth discordance in appropriately grown twins that found monochorionic discordant pairs to be at significantly higher risk for preterm birth and stillbirth (Harper et al. Am J Obstet Gynecol 2013; 208: 393.e1–5). The latter study, however, evaluated growth discordance at birth.

Some of the limitations of the present study must be considered when interpreting the findings. The authors acknowledge the lack of data on maternal characteristics. This limitation may be non-differential epidemiologically, given that if maternal characteristics affect fetal biometry, they should affect both twins equally, and therefore should not be a significant confounding factor in the primary outcome of the study.

The optimal definition of birthweight discordance in twin gestations remains unknown. The current study used a 25% difference, whereas several studies in the past have used a threshold of 20% (Table 4). The authors justify the chosen cut-off based on their previous study showing this to be a better threshold for predicting perinatal death (D'Antonio et al. Ultrasound Obstet Gynecol 2013;41:643–8). Theoretically, choosing a higher threshold for defining discordance could explain the lack of association between second-trimester biometry and birthweight discordance, but should not affect the association with the more important perinatal outcomes studied, including perinatal loss and preterm birth. Finally, despite the conclusion, it was not clear in reading the article whether cases with twin-to-twin transfusion syndrome were excluded from the analysis.

Disclosure of interests

I have no interests to disclose.

  • A Odibo

  • Division of Ultrasound and Genetics, Washington University School of Medicine, St Louis, MO, USA

Ancillary