What will be the role of first-trimester ultrasound if cell-free DNA screening for aneuploidy becomes routine?

Until recently, most developed countries either had adopted the first-trimester combined test as the standard approach to antenatal screening for Down syndrome or were in the process of doing so. Some had even introduced sequential protocols such as the contingent test, which builds on the combined-test markers and offers second-trimester serum markers and/or further ultrasound examination to a large subgroup of women, selected according to their estimated risk. Others have extended the combined test by incorporating routinely simultaneous determination of additional ultrasound markers. However, the situation has changed radically with the introduction of non-invasive prenatal testing using maternal plasma cell-free (cf) DNA. In response to the encouraging published results of early studies, professional bodies cautiously recommended the use of cfDNA only as a secondary test in those patients already regarded as being at high risk. This approach restricted testing to women with positive conventional screening tests and, in some localities, to women with advanced maternal age, a family history of aneuploidy or ultrasound markers of aneuploidy. Since then, further confirmatory studies have been published, particularly in unselected populations, and now there can be no doubt that primary cfDNA screening could achieve vastly superior performance for the detection of Down syndrome compared with any of the conventional protocols or secondary cfDNA screening. However, this approach is currently impractical in most public-health settings because of the high cost of cfDNA testing and the fact that, in many localities, the tests are available only by shipping samples to commercial laboratories in the USA and China. To overcome the problem of cost, a compromise has been suggested: the contingent cfDNA test1. This approach proposes that initially all women, including those of advanced maternal age, have a conventional screening test; those estimated to have very high risk are then offered invasive testing and only those with borderline risk are offered cfDNA testing. In most localities, the combined test would be used for the initial risk assessment; its performance, and hence that of the contingent cfDNA test, could be further enhanced by including additional maternal serum markers, such as placental growth factor (PlGF) and maternal serum alpha-fetoprotein (AFP)2. For those with sufficient training, additional ultrasound markers, such as absent or hypoplastic nasal bone, and abnormal tricuspid valve and ductus venosus (DV) Doppler waveforms, could also be included. It is clear that cfDNA testing is here to stay and the only question remaining is how it is best implemented. As with conventional screening, a variety of strategies will be needed for different communities around the world. Furthermore, these strategies will change over time. Compromise protocols like the contingent cfDNA test are likely to be an interim measure, as it is expected that, in the near future, the cost of cfDNA testing will fall sufficiently such that routine screening will be feasible and testing will become available in local laboratories. Although we include here a brief assessment of the cost-effectiveness of different cfDNA approaches in relation to detection of Down syndrome, our focus is on the broader, unintended and potentially deleterious implications of the widespread use of this new technology. In particular, implementation of routine cfDNA screening could result in abandonment of the first-trimester ultrasound examination, simply because of a perception that 'nuchal translucency (NT) measurement will no longer be needed'. In our view, however, a strong argument can be made for retaining the first-trimester scan, its many benefits including: clarification of cfDNA results; early detection of fetal structural abnormalities, such as serious cardiac defects and open spina bifida (OSB); accurate determination of gestational age; assessment of multiple gestations; comprehensive evaluation of the uterus and adnexa; and prediction and prevention of adverse maternal–fetal complications of pregnancy. The cost-effectiveness of the first-trimester scan for purposes other than Down-syndrome screening is difficult to quantify. However, we hope to demonstrate that the clinical benefits of retaining it would be so great that the potential financial consequences will be self-evident. For the conventional screening protocols, performance is best derived from statistical modeling, since prospective screening studies are subject to 'viability bias', which necessarily overestimates Down-syndrome detection rates. We used model parameters derived from a published meta-analysis3 in order to reflect average practice. Performance varied according to a number of variables: the gestational age at which NT was measured; the inclusion of specific additional ultrasound markers, if any; the gestational age at which serum was obtained; the isoform of human chorionic gonadotropin used; the inclusion of additional serum markers, if any; the risk cut-off used; and the maternal age distribution. For the contingent test, there was also the variable risk cut-off used to trigger an evaluation of additional markers. For routine cfDNA screening, performance was estimated by meta-analysis of individual performances from the 13 published studies carried out on plasma samples obtained prior to invasive prenatal diagnosis (so-called 'high-risk' studies)4-16. In addition, there have been seven studies published in which plasma samples were collected as part of existing screening programs17-23 and an additional one with a mixture of both study designs24. The latter results were consistent with those of the high-risk studies but are less reliable due to both incomplete follow-up and viability bias. The cfDNA studies differed in technique (whole-genome shotgun sequencing, targeted sequencing, or quantitative assessment of single nucleotide polymorphism (SNP) differences between mother and fetus) and in the bioinformatics used to classify results as 'positive', 'negative' or 'no-call'. There were also differences in the gestational age distributions. For all techniques, performance was critically dependent on the fetal fraction (the proportion of cfDNA that was fetoplacental in origin), which varied with gestational age. Therefore, only a simple meta-analysis could be performed since these studies failed to provide sufficiently detailed information to enable a more sophisticated analysis25. For the contingent cfDNA test, performance was estimated by modeling using: marker parameters from the same meta-analysis3 that we used for the conventional screening protocols; PlGF parameters from a large published series26; and our results from the cfDNA meta-analysis described above. It was assumed that the classification of the cfDNA result was unaltered by the combined-test risk. Slightly improved performance would have been achieved by taking into account this risk, but more detailed information would be needed for this to be done. Our cost–benefit analysis adopted the method of Cuckle et al.27. The average cost per Down-syndrome birth avoided was computed for each protocol, as was the 'marginal' cost per additional birth avoided over and above those that would be avoided using the combined test. These values were highly dependent on the unit cost of the screening test, the unit cost of invasive prenatal diagnosis and the uptake rates used in the analysis. To illustrate, Table 1 shows the predicted screening performance and costs for three existing protocols and for four based on cfDNA testing, under a set of particular conditions, unit costs and uptake, the details of which are given in the footnote. In this example, with routine cfDNA screening, the average cost of preventing a Down-syndrome birth was increased between three- and four-fold, and the marginal cost was 15-fold higher than the average cost for the combined test. In contrast, with a contingent cfDNA test, the average cost was increased less than two-fold and the marginal cost was less than four-fold greater, whilst much of the improved performance of routine cfDNA was retained. Including additional serum markers did not increase the cost, and the screening performance further approached that of routine cfDNA. Selecting older women for primary cfDNA testing was more expensive and added very little to the overall screening performance. Routine cfDNA screening is estimated to have a positive-predictive value for viable Down syndrome of 54% (Table 1). This means that among pregnancies with positive results that are destined to survive to term, for each Down-syndrome case, there will be an unaffected case requiring invasive prenatal diagnosis. All commercially available cfDNA tests now also screen for trisomies 18 and 13 and monosomy X. However, the predictive value for these conditions is much lower. On the basis of published cfDNA studies7-10, 12-16, 28 and prevalence at birth29, 30, we estimate that it would be about 30-fold lower. Since those with trisomies 18 and 13 do not generally survive long-term and monosomy X has a mild phenotype, this is problematic. The predictive value could be improved considerably in this situation by the use of ultrasound to assess NT, nasal bone, fetal heart rate31, soft markers and fetal anomalies known to be associated with the different types of aneuploidy. This may also apply to some of the microdeletion and microduplication syndromes that are starting to be included in the cfDNA test; because the predictive values are relatively low, the phenotypes are variable and ultrasound markers are available for some of the syndromes. It is logical that the overlap in the distribution of cfDNA metrics between aneuploid and euploid pregnancies is greater when the fetal fraction is lower. This manifests in practice as increased false-positive rates in pregnancies with relatively low fetal fraction. Other reported false-positives have been due mainly to confined placental mosaicism (CPM) and vanishing twins. CPM can cause false-positive cfDNA results as fetal cfDNA results mainly from apoptotic placental cells. If chorionic villus sampling (CVS) is used to make the prenatal diagnosis, such a false-positive result may appear erroneously to be a confirmed case of aneuploidy. A prudent clinician would request a confirmatory amniocentesis instead of, or in addition to, CVS, although this would potentially lead to an extended period of anxiety. However, detailed ultrasound examination of the fetus may either help to resolve the situation or at least provide helpful additional information. Between 1% and 8% of cfDNA tests fail, mostly due to low fetal fraction. Retesting the sample will not always resolve the problem, with up to a third again failing. Moreover, a low fetal fraction is associated with an increased risk of aneuploidy, probably caused by small placental volume, characteristic in cases of trisomy 13 and monosomy X. In this situation, one option is to carry out a combined test, ideally enhanced with additional markers. In all three situations (cfDNA positive for aneuploidy other than Down syndrome, discordance between cfDNA and CVS results and cfDNA test failure), which are expected to affect only a small proportion of pregnancies, first-trimester ultrasound is likely to provide important additional information. Therefore, there is a need to maintain the skills required to perform an informative first-trimester ultrasound examination, including ongoing external quality assessment of NT measurements. It is well documented that a large NT is associated with increased risk of structural abnormalities and genetic syndromes, even in the absence of aneuploidy. For example, in a study of 4697 euploid fetuses with NT > 95th centile, 7.0% had a major abnormality32, 33. The risk increases considerably with NT > 3.5 mm; in two large studies combined, incidence of abnormality was 46% in fetuses with NT ≥ 6.5 mm34. A more comprehensive first-trimester anomaly scan can detect a large proportion of major abnormalities. In a recently published meta-analysis of 19 studies, including a total of 78 002 fetuses, of which 996 had a structural abnormality, Rossi and Prefumo35 estimated that the overall detection rate of fetal abnormalities was 51%. The detection rate was higher (62%) when combining transabdominal and transvaginal approaches, and in those at high risk (65%) due to increased NT measurement. Structural cardiac defects are relatively common and their association with increased NT has been studied extensively. A recent meta-analysis of 20 studies estimated the detection rate for major cardiac defects using NT alone as 44% for a 5.5% false-positive rate36. In five studies combined, the risk of a cardiac defect increased progressively with increasing NT, the risk being 3% for NT 3.5–4.4 mm, 7% for NT 4.5–5.4 mm, 20% for NT 5.5–6.4 mm and 30% for NT ≥ 6.5 mm34. Doppler evaluation of the tricuspid valve at first-trimester ultrasound can improve the detection of cardiac defects because of their association with tricuspid regurgitation. In a first-trimester screening study of 40 990 euploid fetuses, Pereira et al.37 found that 32.9% (28/85) of cases with major cardiac defects had tricuspid regurgitation, compared with 1.3% of those without cardiac defects37. Moreover, the technique for obtaining the Doppler waveform across the tricuspid valve requires magnification of the fetal heart and identification of important cardiac landmarks, such as the atrioventricular valves and the four-chamber view. This requirement in itself has the potential to further increase detection rates of cardiac defects. Doppler evaluation of the DV may also lead to the detection of cardiac defects. Absence or reversal of blood flow in the DV during atrial contraction (a-wave), detected using pulsed-wave Doppler, is associated with an increased risk of a major cardiac defect. In the same screening study, Pereira et al.37 found that 28% of affected fetuses had an abnormal a-wave in the DV, compared with 2.1% of unaffected fetuses37, 38. In a meta-analysis of seven studies, the estimated detection and false-positive rates for major cardiac defects using DV Doppler were 50% and 7%, respectively39. Performance differed according to NT: the detection rate was 19% for a false-positive rate of 4% if the NT was normal, and 83% for a false-positive rate of 20% if the NT was increased. Combining markers is likely to increase detection further. One study estimated that a combination of NT > 99th centile and abnormal DV would have a detection rate of 47% for a false-positive rate of 2.7%40. In their screening study, Pereira et al.37 reported that the same combination of markers would have detection and false-positive rates of 33% and 2.7%, respectively37. The addition of tricuspid valve Doppler assessment would increase the detection rate to 52% for a 4.1% false-positive rate. A first-trimester anomaly scan will detect a proportion of cardiac defects similar to that when a combination of markers is used. In their meta-analysis, Rossi and Prefumo35 reported on 418 fetuses with cardiac abnormalities, of which 48% were detected. A large NT or abnormal Doppler findings serves to focus the sonographer on careful examination of the heart, which leads to a further increase in detection. This should happen not only at the first-trimester scan, but also at the time of the mid-second-trimester anomaly scan. Second-trimester screening for neural tube defects (NTDs) using maternal serum AFP at 16–18 weeks' gestation and ultrasound cranial and cerebellar markers of OSB is well established in clinical practice. However, following the adoption of the first-trimester combined test, there has been increasing interest in first-trimester ultrasound screening for OSB. Several studies have shown that changes in the posterior fossa, probably related to Arnold–Chiari type-II malformation, which is associated with OSB, can be used as first-trimester markers. These reported markers relate mainly to the fourth ventricle, which is compressed, and the brainstem (BS), which is thickened, in affected pregnancies. Examined in the mid-sagittal plane, it has been observed that in OSB the intracranial translucency of the fourth ventricle can be obliterated41, reduced42-45 or replaced by fluid elsewhere in the posterior brain43. In these cases, the BS thickness can be measured as a single marker, but superior results have been reported by calculating the ratio of the BS thickness to the distance between the posterior edge of the BS and the occipital bone (BSOB)43-47. An increased BS/BSOB ratio is associated with an increased risk of OSB. Absence or narrowing of the cisterna magna in the mid-sagittal plane has also been reported as a good marker44, 45. Another first-trimester consequence of Arnold–Chiari malformation is caudal displacement of the forehead, resulting in reduction of the frontomaxillary facial angle, which can also be used as a marker48. Three-dimensional ultrasound imaging of the entire ventricular system has revealed additional markers49. Whilst considerable skill and experience are to use these posterior or facial markers of OSB, a screening approach is available for the It has been shown recently that, in OSB, the is to a similar in the as it is in the In one first-trimester study of fetuses with OSB, of the affected fetuses had a measurement the of a distribution derived from unaffected In study that included cases of OSB, were by calculating a ratio between the and the transabdominal of the affected fetuses were to have a low diagnosis can be achieved in many of the pregnancies that are found to be at high risk for OSB, based on the of these markers. A ultrasound examination of the fetal using a transvaginal will even relatively small However, if a be at this the second-trimester anomaly scan can be than but at a time when the diagnosis can be made with more The meta-analysis by Rossi and Prefumo35 of first-trimester studies included cases described as and abnormalities, of which were detected. The number of in this and their detection rate were not but it was that, for and the detection rate was is as the first-trimester of the which on the of in the due to the of the to It should be that first-trimester examination of the posterior was not carried out in the time period during which the studies included in this meta-analysis were NT has also been in case and studies, with a large number of other structural abnormalities and genetic syndromes, including: fetal syndrome, syndrome and NT is also associated with fetal in many cases, results from between the prevalence of fetal structural defects and NT measurement was in a recent study of with aneuploidy and cardiac defects were from this analysis. with NT > 95th centile had a increase in the risk of and defects. anomalies with an increased risk of or more and of the and of the small and As evaluation of the posterior has become more an association has between a BS/BSOB ratio centile in to the increased ratio associated with and abnormalities. These include malformation, and In Rossi and meta-analysis of first-trimester a of abnormalities, including reduction and was reported to have a detection rate of In the first-trimester measurement is the most accurate method of gestational The of calculating the gestational age be of it into account to some We three relatively common to this the diagnosis of growth abnormalities is based most on ultrasound The estimated fetal is compared with a distribution which is so an accurate gestational age is distribution can only be in a in which accurate gestational age has been first-trimester measurement serves to two important of of are based on gestational age. In patients gestational age is there is the of an early and its can also lead to the use of fetal testing. On the other if gestational age is the fetus may be to the of being an accurate determination of gestational age is for at the of the gestational age at which are made to a being and in an of a can lead to an for a fetus that has no of is an important of risk in multiple and its determination is to the of an during pregnancy. In the of the at the at which it the is an accurate method of However, the of this method as as will not the that need in order to and as will have an number of ultrasound Doppler and fetal testing, as well as an early The incidence of has increased over the an accurate diagnosis all the more An scan generally on the of the However, it is to the uterus itself and the as since both can most and The incidence in of of the uterus is about and that of is about The and for the first-trimester examination include a to the maternal in of the fact that this is the best time in to do The of a or is only to However, even when this is not the case, it is important to that such a as it may become to in of the of the uterus do not cause in however, can cause and, in the case of The of are This is in the when are very These can be from even if are and in based on their on ultrasound and Doppler examination, as well as on their to in as The most common are and These may not during but should be at some However, if become their may be required during pregnancy. during is not an If the clinician is of the of an or of this can be included more in the diagnosis for the The incidence of during is very per and many of these cases are of low An of are only may these at some but also the diagnosis of a diagnosis of a during for the option of at the time of a if the risk and of an additional maternal–fetal complications of are much more common than are and structural fetal abnormalities. Whilst these complications generally in many have been shown to from in the and, as are for early screening. The one most studied is a common cause of fetal and maternal and A meta-analysis of published clinical has shown that is weeks' this was confirmed recently in a more of gestation is now recommended in pregnancies as high-risk due to including maternal and and family these alone are both and However, the addition of first-trimester Doppler and blood to improve the first-trimester screening performance. have been published by The for the risk of using prior with Doppler and with or without maternal serum markers, such as plasma and with the predicted that, without maternal serum there would be a detection rate of for early weeks' and for all with a false-positive rate of These detection rates increased to and inclusion of serum markers. An without PlGF was recently in an in an of women, prior to the of the study, there were developed early In the women were of were as being high-risk by The and were to at a single case in the developed early which a The cost of women pregnancies are by gestational and of their is This cost was estimated for the of in using and rates to The total cost for gestational over that of unaffected pregnancies was estimated to be a value that not even include the costs of complications secondary to such as and other It is clear that a screening for sufficiently early in such that an can be would likely result in cost weeks' gestation can also growth in the absence of A first-trimester has been developed screening to those at risk of the same markers as for screening, estimated the detection rates for and to be and with an overall false-positive rate of In the of public-health when the value of a new test, it is not only the cost and of the test, but also the potentially on other that have to be In the case of routine cfDNA screening, cost is likely to fall in the near future, and in the detection of Down syndrome is greater than is that of existing approaches such as the combined test. However, we are that some when the cfDNA test, will to the NT scan and, if it is already the first-trimester anomaly scan. In this we have the consequences of such a with to three cfDNA screening prenatal diagnosis of other abnormalities and in There is a need for first-trimester ultrasound examination in the of cfDNA results positive for aneuploidy other than Down syndrome, discordance between cfDNA and CVS and cfDNA This could be by a than as a routine although it would need to be for in a cost–benefit analysis of cfDNA screening, which has not been done. screening based on NT with or without Doppler tricuspid valve and DV or a anomaly scan, would detect a large proportion of cardiac abnormalities and as well as many other structural abnormalities. Whilst in many localities, or has to a in major cardiac abnormalities a considerable public-health are much more at birth than are all combined, and To our there are no published cost–benefit for first-trimester cardiac abnormality screening, but this is likely to It could be that the structural abnormalities detected in the would also have been found on the mid-second-trimester anomaly scan that is carried out routinely in most However, diagnosis has a number of the is of a fetal problem with sufficient time to including of at a when this is less and Moreover, in some including the in the latter of the is either not available or will become less Furthermore, and of the in that such are the more there is that the will be For it is important to carry out a first-trimester ultrasound scan in order to estimate gestational age, and in multiple may many would to only these following and USA some also the uterus and However, if this is as a scan, there is a that it will be performed in a much more less reliable maternal–fetal complications of are common and first-trimester screening can a large proportion of and Whilst screening could also be at the time of the mid-second-trimester anomaly scan, this is for prevention by or in the of not only but also the rate of and the cost associated with of the very In a cost–benefit analysis of one first-trimester screening protocol, the cost per was If ultrasound were already being for other it would be even lower. if an individual were to the first-trimester ultrasound examination as being in each of these three it would not necessarily that the scan should be in of cfDNA screening as a a first-trimester ultrasound examination, performed by a a comprehensive evaluation of the and the which to improved and, on is likely to at It is expected that, in the near future, the cost of cfDNA testing will fall sufficiently to make routine screening that may abandonment of the first-trimester ultrasound However, a strong argument can be made for retaining this scan. It can cfDNA detect fetal structural abnormalities, such as cardiac defects and and is for including accurate determination of prevention of and evaluation of the uterus and We have here on ultrasound examination, but similar could be for retaining testing. of between cfDNA screening and existing screening should as being should on protocols that the of all Cuckle is to and a of and and number on use of and in screening for