Effectiveness of dating twin and triplet pregnancies using measurements and charts for singleton pregnancies

Six studies (reported in seven publications) were identified for inclusion in relation to effectiveness of measurements and charts used for dating singletons when applied to twins or triplets.^30–36

The first study used data collected in the UK and compared biparietal diameter between twins and singletons, although details of the charts used were not reported.³⁰ This study used the day of fertilisation (or frozen embryo replacement) for dating pregnancies.

The second study (reported in two separate publications) was conducted in Brazil and prospectively compared crown–rump length between twins and singletons using published charts, although again details of the charts used were not provided.^31;32 Pregnancies were dated by day of oocyte retrieval, although embryo transfer was performed 2–3 days later.

The third study was conducted in the UK and used a retrospective cohort design.³³ Mean differences between the true gestational age and that estimated from first-trimester crown–rump length measurements were derived for singletons and twins and compared using three different formulae. In all pregnancies, gestational age was calculated using the date of embryo transfer.

The fourth study was also conducted in the UK and used a retrospective case–control design.³⁴ This study investigated whether there was a significant difference between second-trimester measurements of head circumference and femur length in twins when compared with measurements in singletons. In all pregnancies, gestational age was calculated using the date of embryo transfer. It is likely that this study involved the same population as the third study.

The fifth study used data collected in the USA to derive a prediction equation for gestational age in singleton pregnancies (using head circumference, femur length and abdominal circumference) and applied it to twins and triplets.³⁵ A ‘best-fit’ model for estimating gestational age in singletons was derived using the fetal biometric indices and then used to examine the accuracy of gestational age prediction in twin and triplet pregnancies (by comparing systematic and random errors). Data for this study came from birth records of women whose pregnancies were dated by day of oocyte retrieval and fertilisation.

The sixth study, conducted in Sweden, used a prediction equation for gestational age (using biparietal diameter with or without femur length) derived from maternity and ultrasound records of healthy women, and compared results between twins and singletons.³⁶ All pregnancies in this study were dated by day of oocyte retrieval and frozen–thawed embryos were transferred 2 days later.

With the exception of the sixth study, which involved Swedish women,³⁶ none of the studies provided information about ethnicity of the participants. The third and fourth studies excluded women with monochorionic twin pregnancies. None of the other studies provided information about chorionicity.^33;34

Choosing which fetus to use to date twin and triplet pregnancies

Three studies were identified for inclusion to address the question of which fetus should be used to establish gestational age in twin and triplet pregnancies.^33;35;37

The first study was a small prospective study, conducted in France, that compared gestational age predictions using crown–rump length measurements in twin pregnancies evaluated at 11–14 weeks of gestation.³⁷ The charts used in the study were not referenced and the method of dating the pregnancies was not reported.

The second study, which was conducted in the USA, was larger, although retrospective in design.³⁵ The gestational age range studied was later (second trimester) than in the first study. This study derived a ‘best-fit’ model for estimating gestational age in singletons using fetal biometric indices, which was then used to examine the accuracy of gestational age prediction using individual fetuses in twin and triplet pregnancies.

The third study was a retrospective cohort study conducted in the UK.³³ Crown–rump length measurements conducted routinely in the first trimester (at 11–14 weeks of gestation) were compared using charts attributed to Robinson, Rossavik and Von Kaisenberg. In all pregnancies, gestational age was calculated from the date of embryo transfer.

Chorionicity was reported in the first and third studies,^33;37 but not the second study.³⁵ Ethnicity was not reported in any study.

Published health economic evidence

No published health economic evidence was identified and this question was not prioritised for health economic analysis.

Evidence statement

Evidence was identified for all fetal ultrasound parameters prioritised for consideration in terms of determining gestational age in twin and triplet pregnancies. All evidence came from observational studies which constitute low (or very low) quality evidence.

With regard to whether the measurements and charts used in singletons were accurate when applied to twins and triplets, no statistically significant differences in size were found between twin and singleton pregnancies using crown–rump length (very low quality evidence) or biparietal diameter (low quality evidence). Significant differences were reported in the head circumference of larger and smaller twins compared with singletons, although this difference did not remain significant when an average of each set of twins was used (very low quality evidence). There was a significant difference between smaller twins and singletons in femur length, but the difference was not significant when comparing the larger twin or the average of each set of twins with singletons (very low quality evidence). Gestational age estimation in twins was not statistically significantly different from singletons when dating was carried out by a formula based on femur length, head circumference and abdominal circumference (very low quality evidence), but the same formula systematically underestimated gestational age in triplets by 1 day (very low quality evidence). There was no statistically significant difference in dating by day of oocyte retrieval between twin and singleton pregnancies (low quality evidence).

Similarly, there was no evidence to suggest that any specific fetal measurement in multiple pregnancies was more effective than another in gestational age estimation.

The majority of the studies appeared to use date of oocyte retrieval to determine the true gestational age. However, the studies were limited, with bias from small sample sizes, operator bias and studies being retrospective. The impact of the use of the timing of oocyte retrieval versus the timing of embryo transfer on dating could not be evaluated from the searches conducted for the guideline (no additional searches for evidence relating to singleton pregnancies could be conducted within the timescale for developing the guideline).

With regard to which fetus should be used for estimating gestational age in twin and triplet pregnancies, the GDG was of the view that there was a possibility that in the first half of pregnancy, when gestational age is determined, the smaller twin could be pathologically undergrown in some cases. That would mean that use of the measurements from the smaller fetus could lead to an underestimate of gestational age. No evidence was available for prediction of fetal growth restriction as an outcome and whether use of the smaller fetus in twin pregnancies with impaired growth potential leads to this error in practice. Evidence was, however, available for growth discordance between twins, that resulted in an average discrepancy of 3.4 mm in crown–rump length between the larger and the smaller twin (very low quality evidence). No evidence was available for prediction of other twin complications or congenital anomalies. One study suggested that dating of twin pregnancies was more accurate when the smaller twin, rather than the larger twin, was used (very low quality evidence). However, two other studies showed evidence supporting the use of the average fetal size to determine gestational age in twins and triplets (very low quality evidence).

Relative value placed on the outcomes considered

There is a need to determine which fetus should be used as the reference for the dating process in twin and triplet pregnancies. Accurate estimation of gestational age in such pregnancies is important because it forms the basis for predicting, assessing and managing the potential complications of the pregnancy. All outcomes specific in the review protocol were considered critical in terms of informing recommendations for clinical practice.

Trade-off between clinical benefits and harms

‘Antenatal care’ (NICE clinical guideline 62)¹⁴ already addresses estimation of gestational age using ultrasound and no additional benefits or harms were identified in relation to twin and triplet pregnancies. With regard to which fetus to use, the ultrasound measurements of all fetuses will be taken in the pregnancy in any case. The only issue is which measurement should be used to ‘date’ the pregnancy. Evidence shows limited differences between smallest, largest and mean measurements to predict gestational age. However, clinically it is counterintuitive to date the pregnancy by the smallest fetus, which is more likely to be affected by early growth pathology and/or may result in unnecessary early delivery. The GDG therefore considered it more appropriate to date the pregnancy using the largest fetus.

Trade-off between net health benefits and resource use

The review question (including its subsidiary questions) was not identified as being of high priority for health economic evaluation. Only one ultrasound scan is needed to estimate gestational age, and such a scan is a standard requirement of routine antenatal care as recommended in ‘Antenatal care’ (NICE clinical guideline 62).¹⁴ The GDG acknowledged that more time would be needed for scanning in twin and triplet pregnancies; however, the cost impact and opportunity costs of the additional time needed were thought to be negligible.

Quality of evidence

The available evidence was limited in quantity and quality. No randomised controlled trials (RCTs) were identified and most of the included studies were retrospective in design, using a variety of different methodologies (for example, categorical versus continuous representation of gestational age, smaller and larger twins analysed independently or combined, size of fetus used to date pregnancy, head circumference versus crown–rump length). The quality of evidence for differences in fetal size in twin and triplet pregnancies versus singleton pregnancies was mainly very low. The quality of evidence for differences in dating of twin and triplet pregnancies versus singleton pregnancies was also mainly very low, as was the quality of evidence for prediction of growth discordance and accuracy of dating.

Other considerations

The majority of the studies did not report chorionicity or ethnicity. Only one study considered triplets, with the other studies concentrating on twins. This review question addressed whether there are differences in dating or the size of singleton versus twin or triplet pregnancies that should be taken into account when calculating gestational age in clinical practice. In view of the limitations of the evidence, the GDG based its recommendation on consensus within the group and highlighted the need for further research in this area. The GDG was of the view that estimating gestational age by ultrasound using crown–rump length (between 10 weeks 0 days and 14 weeks 1 day) or head circumference (from 14 weeks 0 days) as recommended for singleton pregnancies in ‘Antenatal care’ (NICE clinical guideline 62),¹⁴ and incorporating recent changes to the gestational age ranges appropriate for use of crown–rump length and head circumference (see NHS Fetal Anomaly Screening Programme [FASP] programme statement 2010/02^*) would be appropriate in twin and triplet pregnancies.

Screening for Down’s syndrome is best undertaken when crown–rump length is between 45 mm and 84 mm (11 weeks 2 days and 14 weeks 1 day; see the FASP programme statement and Section 6.1). From a practical point of view, if Down’s syndrome screening is requested by the woman, it makes sense to perform it at the same first-trimester ultrasound scan as the estimation of gestational age and determination of chorionicity. The best interval for performing all three tests together is, therefore, when crown–rump length is between 45 mm and 84 mm (at approximately 11 weeks 0 days to 13 weeks 6 days). In practice, it may not be possible to schedule all three tests at the same appointment, and in such circumstances more than one appointment in a short period may be needed. Furthermore, it is important that adequate time is given to allow for the additional counselling required regarding Down’s syndrome screening once a multiple pregnancy has been identified. Also, some women may have their first scan as early as 10 weeks 0 days (in accordance with ‘Antenatal care’ NICE clinical guideline 62),¹⁴ in which case they would need a separate appointment for Down’s syndrome screening, if requested. However, if the woman is known in advance to have a twin or triplet pregnancy (for example, if such a pregnancy results from IVF treatment) it may be possible to plan to schedule all three tests in a single appointment. The GDG emphasised the importance of ensuring timely referral to maternity services in the first trimester, so that women with twin and triplet pregnancies have the opportunity to access first-trimester screening for Down’s syndrome (which is strongly preferred to second-trimester screening for Down’s syndrome; see Sections 5.4 and 6.1).

Evidence suggests that the mean twin measurement best reflects gestational age, both in the first and second trimester, whether using crown–rump length in the first trimester or head circumference in the second trimester. The GDG recommends using the larger twin measurement to determine gestational age (in the first half of pregnancy) because using the mean twin measurement would lead to an underestimate of gestational age if the smaller twin were pathologically undergrown. Similarly, the largest triplet measurement should be used to date triplet pregnancies.

Published health economic evidence

No published health economic evidence was identified and this question was not prioritised for health economic analysis.

Evidence statement

Evidence was identified for a variety of methods used to determine chorionicity from ultrasound scans in twin and triplet pregnancies.

The sensitivity and specificity of the methods used to determine chorionicity from ultrasound scans is generally high. Over half of the reported methods achieved both a sensitivity and specificity over 90%.

At a mean or median gestational age of 11–14 weeks at the time of scan, diagnostic accuracy statistics were reported for membrane thickness (low and moderate quality evidence), the number of placental masses and lambda/T-sign (very low quality evidence), and two different composite methods (low quality evidence). The strongest likelihood ratios were reported for a composite method involving lambda/T-sign and number of placental masses with or without concordant/discordant fetal sex. The sensitivity for this test was also high.

For a mean or median gestational age of more than 14 weeks at the time of scan, results were reported for the use of membrane thickness (very low quality evidence), the number of placental sites (moderate quality evidence) and two different composite methods (very low and moderate quality evidence). Composite methods (number of placental masses and lambda/T-sign, and concordant/discordant fetal sex with or without membrane thickness) showed the strongest likelihood ratios. The highest sensitivity was reported when membrane thickness was included in the composite method.

Some studies reported findings for a gestational age of less than 11 weeks or over a wide range of gestational ages with no mean age reported. Results were reported for membrane thickness (very low to moderate quality evidence), number of membrane layers (moderate quality evidence), the number of placental masses and lambda/t-sign (low quality evidence), and composite methods (low to moderate quality evidence). The composite methods showed the strongest likelihood ratios and high sensitivity. These methods used membrane thickness and number of placental masses, with or without lambda/T-sign, number of gestational sacs and number of fetal poles.

The GDG is aware that the evidence presented may be biased due to analysis after the study concluded for patterns that were not specified before the study, particularly in studies that examined individual methods such as membrane thickness. In these studies, it is not clear how a clinician determining chorionicity on one measure alone (such as subjectively thin or thick membrane) would not be influenced by other aspects of the ultrasound scan (such as the number of gestational sacs).

Relative value placed on the outcomes considered

Sensitivity is the percentage of pregnancies found to be monochorionic at placental examination that were predicted to be monochorionic at scan (true positive). One hundred minus sensitivity (100 − sensitivity) is the percentage of pregnancies found to be monochorionic at placental examination that were predicted to be dichorionic at scan (false negative).

Specificity is the percentage of pregnancies found to be dichorionic at placental examination that were predicted to be dichorionic at scan (true negative). One hundred minus specificity (100 − specificity) is the percentage of pregnancies found to be dichorionic at placental examination that were predicted to be monochorionic at scan (false positive).

PPV is the percentage of pregnancies predicted to be monochorionic by the scan that were confirmed at placental examination to be monochorionic. One hundred minus PPV (100 − PPV) is the percentage of pregnancies predicted to be monochorionic by the scan result that were confirmed at placental examination to be dichorionic.

NPV is the percentage of pregnancies predicted to be dichorionic by the scan that were confirmed at placental examination to be dichorionic. One hundred minus NPV (100 − NPV) is the percentage of pregnancies predicted to be dichorionic by the scan that were confirmed at placental examination to be monochorionic.

The positive likelihood ratio (LR⁺) shows how much the odds of a pregnancy being monochorionic increase when a scan predicts monochorionicity. The negative likelihood ratio (LR⁻) shows how much the odds of a pregnancy being monochorionic decrease when a scan predicts dichorionicity.

The GDG prioritised likelihood ratios and sensitivity when considering the evidence for different methods of predicting chorionicity. They considered a sensitivity of less than 75% to be an imprecise test, and this is reflected in the GRADE profiles for this review question.

Trade-off between clinical benefits and harms

Determination of chorionicity is required to correctly stratify perinatal risk according to the type of twin or triplet pregnancy. Since pregnancy risks, clinical management and subsequent outcomes are very different for monochorionic and dichorionic twin pregnancies (and monochorionic, dichorionic and trichorionic triplet pregnancies), accurately determining chorionicity is very important.

Monochorionic twin pregnancies have a higher risk of developing complications, including feto-fetal transfusion syndrome (FFTS), fetal growth problems, structural abnormalities and overall perinatal loss compared with dichorionic twin pregnancies. The assessment of chorionicity is easier in the first trimester than in later pregnancy and so it is important to assess and document chorionicity clearly at this gestational age. There is benefit in identifying true positives as women with monochorionic pregnancies will require additional fetal surveillance. Women can make decisions fully informed of risks and appropriate management of monochorionicity can be implemented.

Identification of true negatives (women with dichorionic pregnancies) will result in a saving of time and money by avoiding unnecessary additional interventions. False positives will result in additional and unnecessary monitoring, anxiety and cost in women with dichorionic pregnancies.

False negatives have the least desirable outcome, as monochorionic pregnancies will be monitored less, increasing the likelihood of missing serious complications. Furthermore women with false negative test results will not be informed about these potential risks and the consequences.

The trade-off between clinical benefits and harms is unaffected by the choice of methods for determining chorionicity since any measurements would be taken during a single ultrasound scan appointment.

Trade-off between net health benefits and resource use

There is no cost difference between the methods themselves (except that composite methods might take more time for measurements to be conducted) as they can be done at the same ultrasound scan. A method that is more accurate will be more cost effective than less accurate methods if it means fewer women with dichorionic pregnancies receive unnecessary extra monitoring. The GDG emphasised that these scans will tie in to the existing NICE guidance for dating pregnancy and screening, and so the extra costs will be minimal.

Quality of evidence

The quality of evidence was summarised separately for scans done at different times.

For scans at 11–14 weeks:

• membrane thickness: quality ranged from low to moderate and was mainly moderate
• number of placental masses and lambda or T-sign: quality was very low
• composite measures: quality was low.

For scans at more than 14 weeks:

• membrane thickness: quality was very low
• number of placental sites: quality was moderate
• composite methods: quality was very low and moderate.

For scans at less than 11 weeks or at a wide range of gestational ages:

• membrane thickness: quality was very low to moderate
• number of membrane layers: quality was moderate
• number of placental masses and lambda or T-sign: quality was low
• composite measures: quality was moderate to low.

Other considerations

Only one study reported on diagnosing chorionicity in triplet pregnancies and this study evaluated only one method. The GDG assumed that the diagnostic accuracy of methods for determining chorionicity were similar for twin and triplet pregnancies. The GDG is aware that current practice for determining chorionicity involves a composite of methods and there are differences across England and Wales in timing of ultrasound scans. If a twin or triplet pregnancy is diagnosed before 11 weeks of gestation, determining chorionicity immediately using a composite of the number of placental masses, the presence of a lambda or T-sign and membrane thickness is as effective as waiting for the 11 weeks 0 days to 13 weeks 6 days scan. There is no evidence that the use of three-dimensional scans improves the accuracy of chorionicity determination. From a practical point of view it makes sense to perform estimation of gestational age, chorionicity and fetal trisomy screening at the same first-trimester ultrasound scan and the best interval for all three is 11 weeks 0 days to 13 weeks 6 days.

The GDG recognised the importance of assigning nomenclature to fetuses (for example upper and lower, or left and right) and documenting this clearly to ensure consistency throughout pregnancy.

The GDG also recognised the importance of training and support from senior colleagues to ensure that ultrasonographers can identify the presence of a lambda or T-sign accurately and confidently. In view of the potential consequences of failure to determine chorionicity at the time of diagnosis of the twin or triplet pregnancy (especially failure to identify monochorionic pregnancies correctly) the GDG’s recommendations include the possibility of seeking advice from a senior colleague or referral for specialist advice (from a healthcare professional who is competent in determining chorionicity by ultrasound scan).

The GDG’s discussions highlighted that many women with twin and triplet pregnancies are told that the risks associated with such pregnancies depend on zygosity whereas in fact the risks are dependent on chorionicity, and so the GDG identified this as a specific issue to be covered in training.

The GDG also recognised the importance of maternity networks (proposed in the NHS White Paper ‘Equity and excellence: liberating the NHS’^**) in establishing appropriate care pathways for all twin and triplet pregnancies, regardless of chorionicity. Since maternity networks are not yet in place throughout England and Wales, the GDG has used the term ‘networks’ in its recommendations, in accordance with the Department of Health guidance.^† The GDG considered that special consideration should be given to monochorionic monoamniotic pregnancies (see Chapter 9 for further details).