To assess whether there is a correlation between nuchal translucency (NT) and nuchal skin-fold (NF) measurements, in Down syndrome and in normal pregnancies.
To assess whether there is a correlation between nuchal translucency (NT) and nuchal skin-fold (NF) measurements, in Down syndrome and in normal pregnancies.
Nineteen Down syndrome and 224 normal fetuses underwent ultrasound sequential examinations at 11–13 weeks and 14–28 weeks' gestation. NT was measured at the earlier examination and NF at the later one. Both markers were expressed in multiples of the normal gestation-specific median (MoM). The affected cases had been referred to us for termination of pregnancy; NT had been measured locally and NF was measured at our center prior to the procedure. All unaffected pregnancies were scanned routinely at our center.
There was no statistically significant correlation between NT and NF, in either the Down syndrome (r = 0.076, P = 0.76) or the unaffected (r = − 0.021, P = 0.76) pregnancies. The median NF value in Down syndrome fetuses was 1.538 MoM, compared with 0.990 in unaffected fetuses (P < 0.0001). Gaussian modeling with parameters from a published meta-analysis, updated to include the current series, predicted a 91% detection rate of Down syndrome for a 5% false-positive rate when NF replaced second-trimester biochemistry in a sequential contingent screening strategy.
While this study cannot exclude a small correlation between NT and NF, and the Down syndrome karyotype was known at the time of the NF scan, the markers can be considered as independent determinants of Down syndrome risk. Modeling suggests that sequential contingent screening incorporating NF is an effective screening strategy, although this needs to be confirmed in a prospective study. Copyright © 2008 ISUOG. Published by John Wiley & Sons, Ltd.
The excess skin in the fetal neck region which is characteristic of Down syndrome individuals can be observed by ultrasound as either increased nuchal translucency (NT) in the first trimester1 or increased nuchal skin-fold (NF) in the second trimester2. Routine measurement of NT thickness combined with first-trimester maternal serum pregnancy-associated plasma protein-A (PAPP-A) and free β-human chorionic gonadotropin (β-hCG) measurement is an effective method of antenatal screening for Down syndrome, but performance can be improved by testing additional markers in the second trimester3. The most widely promoted sequential screening strategies of this type are based on second-trimester serum markers, tested on most (‘step-wise’) or only a small proportion (‘contingent’) of those screened in the first trimester4. In practice, however, many women also receive sequential ultrasound screening, having a second-trimester anomaly scan5. The Down syndrome risk estimated from their first-trimester results is modified by a likelihood ratio derived from the anomaly scan findings, which include the NF thickness6. This calculation is critically dependent on the assumption that the first-trimester markers and the anomaly scan results are independent predictors of risk. Yet, despite the widespread practice of routine NT and NF measurement, very little is known about the correlation between these two sonographic markers.
Salomon et al.7 reported that there was no significant association between NT and NF among 529 normal pregnancies, but did not give a correlation coefficient. Pandya et al.8 published NT and NF measurements in six Down syndrome fetuses but did not provide sufficient information with which to assess correlation. This study, therefore, aimed to assemble a series in which the correlation coefficients for NT and NF could be estimated in both Down syndrome and unaffected fetuses.
Between October 2005 and May 2007, a consecutive series of women with normal fetuses had an ultrasound examination performed by a single operator (R.M.) on two occasions. At the first visit, between 11 and 13 + 6 weeks' gestation, NT was measured according to the published criteria of The Fetal Medicine Foundation1. At the second visit, NF was determined according to published criteria2, 9. A total of 224 singleton pregnancies had normal fetal anatomy and adequate amniotic fluid, and fulfilled the same exclusion criteria as in our previous study10. The women were being scanned routinely to determine fetal viability and biometry and to rule out fetal anomalies. Informed consent was obtained for both ultrasound examinations.
During the same period, 19 pregnant women who had previously had an NT scan and who were carrying a Down syndrome fetus were referred to our antenatal sonographic unit in the second trimester for termination of the pregnancy. The previous image was used to check the NT measurement by R.M. using published criteria11. All ultrasound images were considered to be either excellent (score, 8–9) or reasonable (score, 4–7) according to Herman's criteria11. Prenatal diagnosis had been made using amniocentesis in 17 cases and chorionic villus sampling in two. The indication for the procedure was increased NT alone in seven cases, and increased risk on the basis of age and/or biochemical markers in the others.
Immediately before the termination each woman was scanned by R.M. to determine placental location, fetal viability and biometry, including NF. Views and measurements were recorded by printing as a thermal hard copy. Only satisfactory images were included for data processing. Ultrasound scans were performed with different ultrasound machines from various manufacturers, all equipped with a 5–7.5-MHz transvaginal probe with a focal range of 6 cm from the transducer tip and/or a 3.5–5-MHz transabdominal probe.
Both NT and NF measurements were expressed as multiples of the gestation-specific median of unaffected pregnancies (MoM), based on regression. To compute the normal NF medians, the series was extended to include an additional 125 women who only had NF measurement during the study period. Gestational age was calculated from ultrasound fetal biometry. The gestation-specific medians for NT and NF are shown in Figure 1. Correlation coefficients were calculated on the log-transformed values after excluding outliers exceeding 3 SDs from the median.
The performance of various Down syndrome screening policies using NT, PAPP-A and free β-hCG combined with NF was estimated by Gaussian modeling12. The model parameters were the mean and SD for each marker and the correlation between markers in normal and in Down syndrome pregnancies. Published meta-analysis estimates were used for NT at 11 weeks and the biochemical markers at 10 weeks13 and a meta-analysis of NF14 was updated to include the current series, except that the Down syndrome SD was estimated by the method of tailoring15. A Gaussian maternal age distribution was used, with mean 27 and SD 5.5 years16. For each screening policy, the model was used to predict the detection rate i.e. proportion of Down syndrome pregnancies with risk above the cut-off, for a fixed false-positive rate i.e. the proportion of unaffected pregnancies with high risk.
Table 1 shows the NT and NF distribution parameters in Down syndrome and unaffected pregnancies. NF values were statistically highly significantly increased among affected compared with unaffected pregnancies, with a median of 1.538 MoM (P < 0.0001, Wilcoxon rank sum test, two-tailed). Although this difference in NF values between affected and unaffected pregnancies was smaller than was the difference in NT values (median MoM NT for affected pregnancies, 1.844, P < 0.0001) the SDs were lower and the discriminatory power of the two markers was similar. This can be measured by the Mahalanobis distance—the difference in log medians divided by the average SD—which was about 2 for both markers.
|GA (weeks, median)||12 + 3||12 + 3|
|GA (weeks, median)||19 + 4||16 + 5|
|Correlation NT–NF||0. 076 (P = 0.76)||− 0.021 (P = 0.76)|
There was no statistically significant correlation between NT and NF in either the Down syndrome (r = 0.076, P = 0.76) or unaffected (r = − 0.021, P = 0.76) fetuses (Figure 2). Individual NT and NF values in Down syndrome fetuses are given in Table 2. For the 10 cases with NT < 2 MoM, the median NF value was 1.65 MoM, not significantly different from that for the nine with NT > 2 MoM, whose median was 1.51 MoM (P = 0.97, Wilcoxon rank sum test, two-tailed).
|GA (weeks)||NT (MoM)||GA (weeks)||NF (MoM)|
|11 + 3||2.016||19 + 4||1.348|
|11 + 5||1.295||20 + 1||1.279|
|12 + 1||1.289||19 + 6||1.896|
|12 + 1||2.846||18 + 3||1.512|
|12 + 1||3.084||15 + 0||1.538|
|12 + 2||1.613||18 + 4||1.864|
|12 + 2||1.844||18 + 6||1.274|
|12 + 2||1.844||20 + 0||1.596|
|12 + 3||1.493||21 + 4||1.886|
|12 + 3||3.508||15 + 0||1.961|
|12 + 4||1.306||20 + 5||1.072|
|12 + 4||1.523||18 + 5||1.706|
|12 + 4||2.467||15 + 1||2.083|
|12 + 4||2.902||15 + 6||1.128|
|12 + 5||1.199||21 + 2||1.125|
|12 + 5||2.257||21 + 5||1.614|
|12 + 6||2.195||20 + 0||1.390|
|13 + 0||1.201||20 + 0||2.034|
|13 + 4||2.275||16 + 2||1.355|
The gestational age at which NF was determined was almost 3 weeks later, on average, in affected cases than in controls. The extended lag-time between the two tests might have attenuated any potential correlation between NT and NF in affected pregnancies. However, there was no evidence for an association between NF levels and lag-time. For the 11 cases examined fewer than 50 days apart, the median NF value was 1.54 MoM, while for the eight cases with greater lag-times it was 1.47 MoM (P = 0.35, Wilcoxon rank sum test, two-tailed).
Extending the published meta-analysis of NF parameters to include the current results yielded a weighted median value in a total of 259 Down syndrome pregnancies of 1.46 MoM. The weighted average difference in the variance of log10 MoM between Down syndrome and unaffected pregnancies was 0.00583, so the tailored Down syndrome SD in the current series was 0.114 (square root of 0.0842 + 0.00583).
Table 3 shows the results of statistical modeling assuming that there is no correlation between NF and first-trimester markers. Sequential strategies based on the addition of second-trimester NF measurement to first-trimester NT, PAPP-A and free β-hCG were predicted to yield a substantial increase in detection rate. Routine NF measurement in all pregnancies in combination with first-trimester screening, rather than calculating risk from the first-trimester results alone (‘non-disclosure’) yielded a 7–11% increase in detection rate depending on the false-positive rate. Step-wise screening with a 1 in 50 first-trimester cut-off yielded almost identical results to combined routine NF and first-trimester screening and had the advantage of a 66% early detection rate. Contingent screening with NF restricted to the 11% of women with borderline first-trimester risk in the range 1 in 50–1500 yielded a 4–10% increase in detection rate compared with first-trimester screening alone. Modeling with the observed NT–NF correlation coefficients made little difference to the predicted rates: at most, a 1% increase in the detection rate for a given false-positive rate.
|Policy||Detection rate (%) for a fixed false-positive rate of:||False-positive rate (%) for a fixed detection rate of:|
|NT, PAPP-A and free β-hCG||74||83||87||1.2||3.8||18|
|NT, PAPP-A, free β-hCG and NF||85||92||94||0.2||0.9||6.4|
Our study indicates that first-trimester NT and second-trimester NF ultrasound measurements are independent determinants of Down syndrome risk. The only other published series of paired NT and NF measurements in Down syndrome fetuses comprised six cases, diagnosed prenatally because of increased NT, whose parents chose to continue the pregnancy8. The NT and NF measurements were reported in mm and the completed week of gestation for each examination was given. Using the gestation-specific medians in our study to calculate MoMs and adding the results to our own yielded a correlation coefficient of − 0.10 (P = 0.63). While Salomon et al.7 reported that there was no significant association among 529 normal pregnancies when levels were adjusted for gestation (P = 0.52) without giving a correlation coefficient, the scatter plot of values was similar in appearance to our own (Figure 2).
As the findings presented here were based mainly on a relatively small series of affected pregnancies, the possibility of selection bias needs to be addressed. The Down syndrome cases were necessarily selective because of the nature of our unit as a referral center. Nevertheless, we believe that these pregnancies are representative of Down syndrome in the catchment area. In our center, termination of pregnancy is conducted by means of dilatation and evacuation (D&E) until 22 weeks' gestation17 and therefore the majority of women with positive first-trimester screening results prefer to wait a few weeks for additional information before considering invasive prenatal diagnosis. Since a substantial proportion of women with Down syndrome pregnancies will have positive first-trimester results, we thus had the opportunity to obtain mid-gestation fetal biometry, including NF results, on an unbiased series.
Another limitation of the study was that the Down syndrome karyotype was known at the time of the NF scan. This could have biased the NF results. However, the median NF level (1.538 MoM) is close to the average value from all previously published series combined (1.46 MoM). Additionally, the operator was not blinded to the NT result, but this is unlikely to have biased the results since a bias would have artificially created a positive correlation and none was found. The Down syndrome fetuses in this series were from pregnancies referred for second-trimester termination of pregnancy. Some cases with extremely high NT values or even generalized edema are likely to have been offered termination earlier in pregnancy and would not have been referred. It is possible that these fetuses would also have had extremely high NF values, but such pregnancies will not generally be candidates for a second-trimester NF scan. Therefore, the fact that we did not include such cases will have had no practical effect on our predictions regarding detection rates of the various sequential screening policies. A similar point could be made about the fetuses with Down syndrome that miscarried between the time of the NT scan and the scheduled date of amniocentesis.
The fact that there was no correlation between NT and NF implies that the mechanism behind these raised values in Down syndrome pregnancies differs for the two markers. The majority of pregnancies affected by Down syndrome, as well as those with other types of aneuploidy, have NT levels above the 95th centile of unaffected pregnancies1, 18, 19. Ultrasonographic and postmortem morphological studies suggest three broad reasons for increased NT thickness, whether in aneuploid or euploid fetuses: (1) cardiac failure; (2) various types of abnormality found in the extracellular matrix of the nuchal skin; (3) abnormal lymphatic development20. There is a brief opportunity between 10 and 14 weeks' gestation, when the fetal lymphatic system is developing and the peripheral resistance of the placenta is high, to detect fluid collection. After 14 weeks, the lymphatic system is likely to have developed sufficiently to drain any excess fluid, and changes to the placental circulation will result in a drop in peripheral resistance. Thus, after this time, many abnormalities causing fluid accumulation may apparently correct themselves and can thus go undetected when measuring NT21.
In mid-gestation, about one-third of Down syndrome fetuses have NF thickness > 6 mm9, but pathophysiology and anatomical studies of fetuses with increased NF have been less informative than have those of fetuses with raised NT. In one study, 11% of fetuses with increased NF or non-septate cystic hygroma had additional structural anomalies, mostly cardiovascular ones22. Johnson et al.23 proposed that first-trimester nuchal edema and mid-gestation nuchal skin thickening are part of a continuum finding, leading to neck webbing in Down syndrome individuals. However, whilst both prenatal features may persist throughout gestation, spontaneous resolution of NT8 and NF24 have been reported in some Down syndrome fetuses. Our own speculation is that, although a webbed neck is a feature of Down syndrome, NT and NF in many cases may represent a different underlying etiology, both in normal and in Down syndrome fetuses.
The practical implication of our findings is that between-trimester sequential screening using measurement of NT, with or without serum markers, followed by NF measurement can be simplified. Since there is no correlation between NT and NF and there is no association between the first-trimester serum markers and NF, risk calculation is straightforward. The risk based on the first-trimester screen (1 in n), expressed as an odds (1 : (n − 1)) can be multiplied by a likelihood ratio (LR) from the NF result to yield a final odds (LR : (n − 1)) or risk (1 in (n − 1)/LR + 1). The LR for an individual NF-MoM is derived from the Gaussian model: 0.74ey, y = 32.39x2 + 12.62x − 1.03 and x = log10MoM. For example, when NF is 1.5 MoM, x = 0.176, y = 2.194 and LR = 6.6. So, if the risk based on first-trimester screening is 1 in 562, or 1 : 561, the final odds will be 1 : 85 (6.6 : 561), a risk of 1 in 86.
In this paper we have considered the effect of various Down syndrome screening strategies involving NF measurement. However, in many localities, ultrasound anomaly examination is routine at 18–20 weeks' gestation and in these circumstances an increased NF thickness, say above 6 mm, is often regarded in itself as a risk factor for aneuploidy6. Using our findings and applying the above formula to the age-specific risk or a screening-derived risk will provide a better interpretation of the sonographic result.