Does antenatal ultrasound labeling predict birth order in twin pregnancies?


Correspondence to: Prof. B. Thilaganathan, Fetal Maternal Medicine Unit, St George's University of London, London SW17 0RE, UK (e-mail:



It is often assumed by obstetricians, neonatologists and parents that the prenatal nomenclature used to identify twins on ultrasound is consistent with twin labeling after their birth. The aim of this study was to use a large regional database of twin ultrasound scans to validate the effectiveness of a scan before delivery in predicting twin birth-order.


A large regional database of twin ultrasound scans with data from nine hospitals over a 10-year period was used to identify all ultrasound examinations carried out just before birth. The discordance in twin order between the last scan and birth was evaluated by observing discrepancies in fetal sex and weight.


In total, 2103 twin pregnancies with ultrasound estimated fetal weights (EFWs) and birth weights were assessed. Of these, fetal sex was recorded in 149 different-sex pregnancies. Discrepancy between antenatal labeling and the anticipated birth order was noted in 37.6% (56/149) of cases when judged by sex discordance and in 36% (757/2103) of cases when judged by weight discordance. Multiple logistic regression analyses demonstrated that weight discordance, but not chorionicity, scan-to-delivery interval, gestation at scan or gestation at delivery, significantly influenced the change in birth order (P < 0.001).


Antenatal ultrasound labeling does not predict twin birth-order in a significant proportion of twin deliveries. This finding should be borne in mind not only by parents, but also by physicians when delivering twins discordant for anomalies that are not evident on external examination. Copyright © 2012 ISUOG. Published by John Wiley & Sons, Ltd.


Twin pregnancies are at increased risk of perinatal morbidity and mortality compared with singleton pregnancies, mainly as a consequence of fetal anomalies, growth discordance, preterm delivery and complications of monochorionic placentation[1-6]. Serial antenatal ultrasound monitoring is used in the management of these complications; in the past this has been hampered by the lack of a consistent or reproducible system of labeling twins on antenatal scans[7-9]. A variety of twin-labeling protocols, such as fetal presentation, sac position and placental site, have been used by different units, and even by individual operators. The lack of consistency in labeling predisposes to errors in twin identification, especially when monitoring fetal growth or wellbeing on serial scans. This is particularly important when undertaking antenatal screening for Down syndrome, so that invasive prenatal diagnosis can be carried out on the correct twin[7].

The lack of a standardized way to label twins correctly and consistently on antenatal ultrasound has also led to uncertainty regarding the accuracy of predicting the birth order on scans. Twin order and presentation are relevant information for obstetricians planning timing and mode of delivery. Birth order is also especially important in twin pregnancies discordant for fetal abnormalities that are not immediately evident on external examination at birth, for example, if one twin is affected by a duct-dependent cardiac lesion. A recent small study of antenatal twin labeling proposed a consistent method of twin identification throughout pregnancy. This study also suggested that twins may apparently switch label or nomenclature at the time of birth[8]. The aim of this study was to use a large regional database of twin ultrasound scans to assess the relationship between scan findings before delivery and the final birth order.


This was a retrospective study of 3166 twin pregnancies booked for antenatal care and delivered in nine hospitals in the Southwest Thames region of London over a period of 10 years from 2000. The database of twin ultrasound scans from the Southwest Thames Obstetric Research Collaborative (STORK) was used to identify all examinations carried out within 5 weeks of birth. Gestational age was determined by the crown–rump length of the larger twin at the 11–14-week scan or by head circumference (HC) if seen after 14 weeks' gestation[10, 11]. Chorionicity was determined, by ultrasound evaluation, according to the number of placentae and the presence of the lambda/T-sign, and was confirmed by placental histopathology in all cases[12]. Only pregnancies in which the presenting or lower twin was documented were used to determine the discordance in twin order between the scan and at birth, rather than presuming that Twin A or Twin 1 was the presenting fetus. Monochorionic monoamniotic and higher-order multiple gestations were not included in the analysis. The estimated fetal weight (EFW) on ultrasound was calculated using the Hadlock formula based on the HC, abdominal circumference and femur length[13].

The likelihood of a perinatal switch was estimated by matching discrepancies in ultrasound EFW with birth-weight discordance for all twins. Thus, if the presenting twin was larger on ultrasound but the first-born twin was smaller, a perinatal switch was considered to have occurred. Likelihood of a switch was also calculated in different-sex twins from discrepancies between ultrasound and birth order. The null hypothesis that there is no difference in the proportion of change in the birth order judged by two methods was also tested. Multiple logistic regression analysis was used to assess the influence of chorionicity, scan-to-delivery interval, gestation at scan, gestation at delivery and weight discordance on the discordance between ultrasound- and birth-order. Intergroup comparisons were tested using Fisher's exact test, and statistical significance was set at P < 0.05. All calculations were performed using SPSS version 15.0 (SPSS Inc., Chicago, IL, USA).


There were 2103 twin pregnancies (4206 fetuses) identified during the study period from the nine hospitals in the STORK database for which ultrasound EFW and birth weights were available for analysis. Of these pregnancies, 260 (12.4%) were monochorionic. The mean maternal age was 33.0 (range, 16–48) years, and the mean gestational age at the last ultrasound scan was 34.0 (range, 25–39) weeks and at delivery was 35.8 (range, 27–41) weeks.

The discrepancy between antenatal labeling and twin birth-order estimated from fetal sex and fetal-weight discordance was 37.6% (56/149) and 36.0% (757/2103), respectively, and there was no significant difference between the two methods (P = 0.72; Table 1). Multiple logistic regression analysis demonstrated that birth-weight discordance, but not chorionicity, scan-to-delivery interval or gestational age at scan or at delivery, was significantly associated with the occurrence of a discrepancy between ultrasound prediction of twin order and that observed at birth (as defined by comparison of EFW and birth weight) (P < 0.001, Table S1). Larger birth-weight discordance was associated with a lower probability of a discrepancy.

Table 1. Change in twin birth-order from that predicted by last ultrasound scan, as evaluated by fetal sex or weight discordance, for various scan-to-delivery intervals
Scan-to-delivery intervalChange in birth order (n (%)) according to:P
Birth weight discordanceFetal sex discordance
Within 1 week108/366 (29.5)10/30 (33.3)0.68
1–2 weeks200/592 (33.8)13/33 (39.4)0.57
2–3 weeks212/552 (38.4)18/43 (41.9)0.74
3–4 weeks194/491 (39.5)10/30 (33.3)0.56
4–5 weeks43/102 (42.1)5/13 (38.5)1
Total757/2103 (36.0)56/149 (37.6)0.72
 (95% CI, 33.9–38.0)(95% CI, 29.8–45.4) 


It is often assumed by obstetricians, neonatologists and parents that the prenatal nomenclature used to identify twins will also apply to the birth order and postnatal nomenclature. The data of this study demonstrate that antenatal ultrasound labeling does not predict twin birth-order in a significant proportion of twin deliveries. In the present study, a large, multicenter twin database was used to validate the effectiveness of a scan just before delivery in predicting twin order at birth. In about one-third of twin pregnancies, there is discordance between the prenatal ultrasound order and the eventual birth order of the twins. The latter estimate may be affected by the limitation of prenatal ultrasound in accurately predicting fetal weight discordance[14-16]. Any inaccuracies in predictions of fetal weight by ultrasound should be random rather than systematic and equally likely to overestimate one twin or the other. As such, artifactual discrepancies in twin order may result from such errors, particularly when the fetuses are of a similar size as suggested by the multiple logistic regression analysis. However, the likelihood of a difference in birth order calculated from discordant fetal sex and weight was not significantly different, supporting the use of fetal weight for accurately estimating differences between antenatal ultrasound and postnatal twin order.

Multiple regression analysis demonstrated that weight discordance, but not chorionicity, scan-to-delivery interval, gestation at scan or gestation at delivery significantly influenced the ultrasound prediction of twin order at birth. The latter may be explained by the increased elective Cesarean section rate with large weight discordances, which is associated with a higher inaccuracy in predicting the twin order[8]. Despite the medical and cultural importance of birth order, there are only three previous, relatively small, studies assessing the correlation between twin orders on prenatal ultrasound and at birth[8, 17, 18]. Two of the studies compared the sex-discordant twin positions on early second-trimester ultrasound scans with subsequent twin order at delivery several months later[17, 18]. Both studies reported that in 90% of cases, the twin designated as occupying the lower amniotic sac in the second trimester, was subsequently first born. In one study, only vaginal deliveries were included[17] and in both there was a considerable time interval from second-trimester assessment to delivery, limiting the clinical utility of the data. The third study reported the relationship between antenatal ultrasound just before delivery and twin birth-order in a cohort of 108 sex-discordant twins[8]. In this study, the birth order changed in approximately 16% (95% CI, 10–25%) of cases and was affected by mode of delivery, with the rate of perinatal switch being higher with Cesarean section (20%) than with vaginal delivery (6%).

The poor association between prenatal and postnatal order of twins may occur for a number of different reasons. It may be the result of a true switch in twin order occurring before or during labor. It could be speculated that the relationship between the amniotic sac and the cervix, which presumably remains relatively constant throughout gestation, may change near term or during labor, leading to a change in the anticipated twin order[17]. Alternatively, the relationship of the twins to the birth canal may also change following rupture of the membranes surrounding one or both twins. The birth order may also be influenced by folding of the intertwin membrane, possibly because of discordance in fetal size or amniotic fluid volumes. Cesarean section affords access to the lower uterine segment and ignores the relationship between the amniotic sac and the maternal cervix, giving preferential access to Twin 2. The physiological changes in the lower segment with active labor may alter the rate at which the twin labeled as first is delivered second, and vice versa[8]. Other factors not assessed in this study, such as fetal presentation, uterine morphology and active management of labor, may also play a role or influence the likelihood of a discordance between antenatal labeling and twin birth-order. Lastly, the discrepancy observed between the antenatal prediction and the actual twin order at birth may be also linked to the variable methods being used for the antenatal labeling of twins on ultrasound.

Knowledge of the anticipated birth order of twins is routinely used by obstetricians when deciding the timing and mode of delivery and is of importance for the wellbeing of twins discordant for some fetal anomalies. Even in uncomplicated twin pregnancies, the neonatal morbidity of the second-born twin is higher[19-24]. Hence, obstetricians need to be aware of the poor association between antenatal ultrasound labeling and twin birth-order when planning delivery of twins discordant for fetal growth or for non-cephalic presentation of the leading twin. Obstetricians and neonatologists also have to be aware of the likelihood of a discrepancy between prenatal and postnatal twin order when delivering babies with discordant anomalies where immediate postnatal intervention is required at birth. For example, when delivering twins discordant for cardiac abnormalities or diaphragmatic hernia, dedicated perinatal assessment is required in order to identify the affected twin quickly after delivery. Correct and consistent identification of antenatal and postnatal twins is also important in studies on perinatal outcomes. Finally, prospective parents should also be informed of the potential discordance between prenatal and postnatal twin order so that this event is not perceived as an error on the part of their carers. An error in the seemingly basic task of labeling twins is likely to result in parents losing trust in their physician for getting the order ‘wrong’.

The most significant limitation is the retrospective nature of this study and the associated confounding biases. We limited such biases by using all twin deliveries from the regional database rather than a highly selected cohort of cases. Furthermore, the use of weight discordance to ascertain the discordance in twin order between ultrasound and birth is limited by the accuracy of ultrasound in accurately predicting fetal weight. However, a secondary check on the perinatal switch rate using fetal sex discordance demonstrated that birth-weight discordance produced similar estimates of the perinatal switch rate. Data regarding the mode of delivery were not available for the majority of deliveries, making it impossible to assess the influence of mode of delivery on the likelihood of change in birth order. It is therefore possible that a high rate of Cesarean section delivery in this cohort may have contributed to the findings of the study. However, the fact that the change in twin birth-order may be higher in Cesarean section deliveries is physiologically very interesting, but clinically irrelevant. Although it would be ideal to compare the difference in twin birth-order between Cesarean section and vaginal deliveries, it would not help physicians or parents to know if a switch had occurred at a particular delivery. Our findings clearly indicate that pediatricians are mandated to confirm the identity of the twin before instigating any treatment, irrespective of the gestation or mode of delivery.

In conclusion, we have demonstrated the limitations of a scan before birth in predicting twin order at birth. If an accurate method of twin labeling ensures a proper longitudinal evaluation of the twins during the gestation and is crucial in the prenatal management of twin pregnancies, the high likelihood for a perinatal switch should be borne in mind in order to manage twins correctly at birth, especially those discordant for fetal anomalies. If determination of chorionicity is the mainstay in the antenatal care of twin pregnancies allowing for stratification of perinatal risk, then understanding the limitations of predicting twin birth-order is important for the correct perinatal management of multiple pregnancies.

STORK contributors

Arash Bahamie, St Peter's Hospital

Amar Bhide, St George's University of London

Andrew Breeze, Kingston Hospital

Anne Deans, Frimley Park Hospital

Michael Egbor, St Helier's Hospital

Cheryl Ellis, Epsom General Hospital

Hina Gandhi, East Surrey Hospital

Rosol Hamid, Mayday University Hospital

Renata Hutt, Royal Surrey Hospital

Adetunji Matiluko, St Helier's Hospital

Faz Pakarian, Worthing Hospital

Aris Papageorghiou, St George's University of London

Elisabeth Peregrine, Kingston Hospital


The following supporting information may be found in the online version of this article:

Table S1 Result of logistic regression analysis of risk factors for discordance between antenatal ultrasound labeling and birth order in twin pregnancies