Dr A Bhide, Fetal Medicine Unit, 4th Floor, Lanesborough Wing, St George’s, University of London, Cranmer Terrace, London, SW17 0RE, UK. Email email@example.com
Please cite this paper as: Dias T, Mahsud-Dornan S, Thilaganathan B, Papageorghiou A, Bhide A. First-trimester ultrasound dating of twin pregnancy: are singleton charts reliable? BJOG 2010;117:979–984.
Objective The aim of this study was to assess the performance of validated singleton crown–rump length (CRL) formulae in dating twin pregnancies at 11–14 weeks of gestation.
Design Retrospective cohort study.
Setting Fetal medicine unit of a London teaching hospital.
Sample Three hundred and eighty-four pregnancies with known dates of conception.
Methods Retrospective analysis of 266 singletons and 118 twin pregnancies conceived by in vitro fertilisation (IVF), with a known date of conception. The gestation calculated from the date of conception was compared with the expected gestation from fetal size using a number of different CRL formulae.
Main outcome measures Difference in gestational age computed from fetal size (CRL) of the bigger and smaller fetus in twin pregnancies and singleton pregnancies using three formulae.
Results Two of the three studied CRL formulae systematically underestimated the mean gestational age and size of singleton IVF pregnancies (Robinson formula: gestation = 1.4 days, size = 2.7 mm). Twin CRL measurements straddled those of singletons, regardless of the CRL formula used (Robinson formula: larger twin gestation = 2.4 days, size = 4.7 mm; smaller twin gestation = 0.8 days, size = 1.7 mm). These underestimates in gestation and size for IVF singleton and twin pregnancies are well within the known limits of accuracy of first = trimester ultrasound measurements, and are of limited clinical significance.
Conclusions Currently available CRL charts underestimate both the age and size of IVF singleton pregnancies by a clinically insignificant amount. This difference is similar for twin pregnancies, suggesting that singleton CRL charts can be used to date twin pregnancies accurately.
Routine dating of pregnancy from a first-trimester crown–rump length (CRL) is superior to the use of menstrual dates.1–3 Most countries now recommend routine ultrasound dating of pregnancy.4,5 There are several formulae available for calculating the gestational age from the CRL.6–14 The use of various CRL formulae introduces systematic differences, which are usually of limited clinical significance in the dating of singleton pregnancies.15 The particular formula chosen for routine use in many settings is often decided on the basis of consensus, rather than evidence of accuracy or reproducibility.16
Twin pregnancies are at increased risk of perinatal morbidity and mortality, compared with singletons, mainly as a consequence of both preterm delivery and fetal growth restriction. Accurate dating of twin pregnancies is therefore vital, but there are two particular areas of concern when trying to achieve this. Firstly, there has been no systematic evaluation of whether CRL charts derived from singleton pregnancies can be used to accurately date twin pregnancies. Secondly, there is a lack of consensus about whether the pregnancy should be dated on measurements taken from the larger twin, smaller twin or on the mean measurement of the twins.17–19 The aim of this study was to determine the accuracy of singleton CRL formulae in dating twin pregnancies from the smaller, larger or mean twin CRL.
This was a retrospective case–control study of pregnancies conceived by in vitro fertilisation (IVF) or intracytoplasmic sperm injection (ICSI) seen in a routine obstetric setting between June 1997 and October 2009. Only dichorionic twin and singleton pregnancies that had undergone first-trimester ultrasound assessment between 11 and 14 weeks of gestation were included in the study. IVF/ICSI pregnancies were used because the date of conception was known, and the gestational age at scan could be calculated. In order to correct for any variation in dating and/or early fetal growth that might occur in pregnancies achieved by assisted reproduction techniques, IVF/ICSI singleton pregnancies were used as controls.
Ultrasound examinations were only carried out by trained sonographers who were certified as competent in first-trimester ultrasound assessment by the Fetal Medicine Foundation (http://www.fetalmedicine.com). The majority of women chose to have nuchal translucency assessment as part of routine screening for trisomy 21. Even when the latter was declined, the CRL was routinely measured to confirm normal fetal growth in the first trimester. The measurements of the CRL were carried out by sonographers according to local protocol. The local protocol required that a single CRL measurement was taken with the fetus in a neutral position.
Calculation of gestational age
The IVF/ICSI gestational (expected) age was calculated by using the embryo transfer date as a proxy for the date of conception (day 14). The ultrasound gestational (observed) age was calculated using three ultrasound dating formulae, based on CRL measurements (Table 1).7,10,11
Table 1. The mean difference between observed (from ultrasound) and expected (from IVF history) measurements expressed either in days or mm, with 95% confidence interval in parentheses. These differences were calculated for the 266 singleton, 110 bigger twin and 110 smaller twin IVF fetuses for various CRL formulae
The mean difference and standard deviation between gestational ages calculated from the IVF history versus ultrasound measurement were expressed in terms of both gestation (days) and size (mm). To determine if there was any systematic over- or under-estimation of gestational age, the 95% confidence interval (CI) of the mean difference was calculated for singleton versus bigger, smaller and mean twin CRLs. To show the CRL frequency distribution, z-scores were calculated as the number of standard deviations an observed CRL measurement deviated from the mean for gestation, with a negative value being smaller. Mean z-scores were compared between singleton, bigger and smaller twins using analysis of variance (ANOVA). Bonferroni’s post-hoc test was used to explore differences between individual groups.
There were 384 IVF pregnancies seen during the study period that had undergone first-trimester ultrasound assessment between 11 and 14 weeks of gestation. This group comprised 266 singleton, 110 dichorionic and eight monochorionic twin pregnancies. The eight monochorionic pregnancies were excluded from the analysis, as the limited number of cases precludes confidence in the statistical analysis of mean differences in size. The mean gestational age (standard deviation) at inclusion was 87.8 (4.5) days for singleton and 87.4 (4.1) days for twin pregnancies (P = 0.4). The proportion of ICSI pregnancies was not significantly different between the two groups (singleton 13.2%, twins 15.8%; P = 0.9).
The mean differences in gestation and size between that observed from ultrasound measurement versus that expected from the IVF history are shown in Table 1. Regardless of the ultrasound formula used, the observed ultrasound gestational age from CRL measurements in singleton pregnancies was longer than expected from IVF dating. The mean CRL of singleton fetuses conceived by IVF was larger by 2.7 mm (1.4 days) when plotted on the Robinson CRL chart. The equivalent mean differences for the bigger, smaller and mean twin sizes were +4.7, +1.7 and +2.8 mm, respectively. The correlations between observed and expected CRL measurements in the smaller and bigger twins are shown in Figure 1.
The frequency distribution of CRL measurement z-scores in singletons and both twins are shown in Figure 2. The Kolmogorov–Smirnoff test demonstrated that all three CRL z-scores were normally distributed. The ANOVA showed significant difference between CRL z-score groups (P < 0.005), and the post-hoc test results for comparisons of the mean z-scores are shown in Table 2.
Table 2. Differences in mean CRL z-scores compared using a post-hoc test (Bonferroni). Numbers represent probabilities
Mean twin size
Bold values indicate significance at 0.05 level.
This is the largest study to evaluate the accuracy of using singleton pregnancy-based CRL charts to date twin pregnancies. The findings of the study demonstrate that the variation in fetal CRL size between singleton and twin pregnancies at 11–14 weeks of gestation is unlikely to be of clinical significance. The maximum difference in size was 2 mm (between the singleton and the larger twin), which is well within the established margin of error for measurement of CRL,20 and is equivalent to a difference of 1 day in gestation. Additionally, the smaller and bigger twin mean CRL measurements ‘straddle’ that of singletons, consistent with the expectation that twins are similar in size to singletons in the first trimester, and the difference in their size represents normal physiological variation. The finding that the smaller twin was on average only 1 mm smaller (equivalent to 0.5 days) than a singleton pregnancy supports the hypothesis that normal twin pregnancies do not exhibit significant growth restriction in the first trimester.
Three previous studies attempted to assess twin pregnancy dating in the first trimester. Martins et al.21 examined ten twin pregnancies between 7 and 10 weeks of gestation, and concluded that there were no significant differences between respective CRLs. The latter study is limited by the small number of patients, very early gestation at ultrasound and the lack of distinction between the larger and smaller twin CRL measurements. In contrast, Wisser et al.22 studied 21 multiple pregnancies (including five higher-order multiples), and demonstrated a maximum discrepancy of 1.6 days in multiple pregnancies. Although this discrepancy is unlikely to have been clinically significant, the authors went on to recommend that a different formula should be used to date twin pregnancies. One further study, evaluating 104 singleton and 81 twin pregnancies, suggested that dating was more accurate in the first rather than second trimester.23 The authors showed that dating by the Hadlock formula over-estimated both singleton and twin pregnancies by a similar duration of 1–2 days. However, their analysis did not differentiate or distinguish the effect of smaller and larger twins.
The current study finding that singleton IVF fetuses were slightly bigger than expected, by 2.7 mm (1.4 days equivalent, CRL z-score of 0.85) using the Robinson formula,7 is in concordance with another study assessing IVF pregnancies and CRL formulae.24 However, the finding that the CRL z-score distribution is normal rather than skewed suggests that either IVF fetuses are uniformly bigger or that dating from the embryo transfer date systematically underestimates the gestational age. When comparing the z-score distributions, singleton pregnancies were significantly different compared with the bigger, but not compared with either the smaller or the mean, twin size. The z-scores for the bigger and smaller twins were normally distributed, indicating that these fetuses were either uniformly bigger or smaller, respectively. The lack of a skewed distribution suggests that these populations were not affected by a cohort of pregnancies exhibiting either growth restriction or macrosomia.
The difference in expected and observed CRLs may also be influenced by our use of embryo transfer date (which is typically on day 2), as opposed to oocyte retrieval date, as the date of conception. Some of the pregnancies were achieved with frozen embryos from a previous IVF cycle. The oocyte retrieval date would have been several months or years earlier, and clearly not applicable. In any case, we used embryo transfer date for both singleton and twin pregnancies. So the conclusion that singleton charts can be used for dating twin pregnancies should remain unchanged.
The findings of the current study may be limited by prior knowledge of the date of conception by the sonographer, thereby potentially introducing bias. However, the extent of bias should be the same for both singleton and multiple pregnancies, and is equally applicable to studies in spontaneous pregnancies where the last menstrual period is known. Although there is conflicting data about early pregnancy growth in IVF and naturally conceived fetuses,24,25 the purpose of the current study was not to assess the accuracy of the dating, but to explore systematic differences in size between singleton and twin pregnancies. Furthermore, this study did not evaluate monochorionic twin pregnancies, but existing data suggests that discrepancy in twin CRLs is independent of chorionicity.18,19
The issue of whether the bigger or smaller twin CRL should be used to date spontaneously conceived pregnancies is an unresolved one. Multiple pregnancies are at increased risk of fetal growth restriction compared with singletons, and the purpose of pregnancy dating is primarily to allow accurate serial assessment of growth in subsequent scans. Estimating the due date is a secondary goal in this instance, as most multiple pregnancies deliver before 40 weeks of gestation, either spontaneously or on medical advice. Lack of a significant difference between singleton and both the smaller and mean twin size infer that the latter two measurements could be used to date the pregnancy with accuracy equal to a singleton IVF pregnancy. Indeed, this finding has been used previously to support the proposal to date twins on the CRL of the smaller fetus.18 Although the CRL z-scores were significantly different in the bigger, compared with the smaller, twin, the clinical relevance of this finding is very limited. The mean CRL difference between singletons and the bigger twin is only 2.0 mm, which is equivalent to a difference of 1 day, and is well within the measurement error reported for CRL assessment.
A policy of dating by the smaller twin will minimise parental anxiety about apparent ‘poor’ growth in the first trimester. However, it is not possible to be confident whether the small twin is ‘normally’ or ‘pathologically’ small at this stage of pregnancy. In contrast, it is relatively infrequent for a twin to be pathologically large, and a policy of dating by the larger twin might be more effective. However, the latter policy would increase the likelihood of unnecessary parental anxiety, because it has the effect of exaggerating the smaller twin’s apparent lack of growth in the pregnancy. It is evident from the current data that dating by the mean twin size would be the most accurate policy. Furthermore, the latter policy would also have the advantages of minimising parental anxiety and making clinicians place less emphasis on CRL discrepancy, which is known to be a poor discriminator of subsequent, clinically significant twin growth discordance.18,19
Disclosure of interests
None to disclose.
Contribution to authorship
TD collected data and wrote the manuscript. SM-D wrote the manuscript. BT conceived the idea and wrote the manuscript. AP wrote the manuscript. AB conceived the idea, analysed the data and wrote the manuscript.
Details of ethics approval
The retrospective nature of this observational study did not require ethics approval.
There was no funding for this study.
We would like to acknowledge all of the staff members of the Fetal Medicine Unit, St George’s, University of London, for their cooperation.