- Top of page
Three series have been published demonstrating that the ultrasound determination of prenasal thickness (PT), a measure of skin elasticity, is a useful second-trimester marker of Down syndrome1–3. An initial study from our group found a median level of 1.32 multiples of the gestational age-specific median (MoM) in 21 affected pregnancies1. A second series of pregnancies was published, in which fetal digit measurements were also obtained2; this included 25 Down syndrome cases with a median PT of 1.35 MoM. Recently, another center has provided independent confirmation of these findings in a series of 26 Down syndrome cases in which 19 (73%) had raised PT (above the normal 95th centile) compared with 15 of 135 unaffected controls (11%)3.
Although there is now compelling evidence that PT can be used routinely to improve screening performance, there is an important practical constraint on its widespread adoption. To our knowledge it has not been incorporated into screening software. Therefore, there is a need for tables that individual sonographers can use to estimate Down syndrome risk from the PT measurement.
In this paper we use all available published data, supplemented by additional cases from our own center, to derive the parameters of a statistical model of PT distribution. From this, we develop simple tables for ‘bedside’ risk calculation.
- Top of page
Prenasal thickness was measured as described previously1, 2. The image of the fetal profile was taken in a mid-sagittal plane of the fetal head, identifying the nose bone, lips, maxilla and mandible. Care was taken to maintain an angle between the insonation beam and the fetal nose axis close to 45° to avoid any false absence or shortening of the nasal bone. The transducer was tilted in order to ensure that the skin and the bone were visually separated. The thin echogenic line of the nasal bone was measured using electronic calipers from the base of the nose closest to the frontal bones to the farthest extent of ossification on the nose. An anechogenic cartilaginous area between the midline of the frontal bone and the nasal bone could be recognized, which identifies the nasal bridge. The prenasal thickness was then measured between the upper edge of the nasal bone as close as possible to the frontal bone (i.e. frontonasal angle) and the outer part of the skin edge. Calipers were placed between the frontonasal angle and the outer part of the nasal bone, and from the same angle to the outer part of the closest nasal skin edge, for measuring the nasal bone and the PT, respectively. In cases of absent nasal bone, calipers were placed at the lower edge of the frontal bone. The images were magnified so that two-thirds of the screen still included the head and neck. All PT measurements were performed without knowledge of the gestational age-specific reference intervals based on the normal population. The views and measurements were printed as a thermal hard copy for the record. Measurements were conducted without a time limitation and only satisfactory images were included for data processing2. Two-dimensional (2D) scans were performed with either a 2–5-MHz curvilinear abdominal transducer or a 5–9-MHz transvaginal probe from different manufacturers.
Three-dimensional (3D) ultrasound imaging was used in the series of Persico et al.3. A mid-sagittal volume was acquired with the transducer parallel to or within 30° of the long axis of the nose and the shortest distance measured to the lowest part of the frontal bone, rather than the frontonasal angle. They used a similar technique to that previously reported for measuring the PT1, 2. However, the lowest part of the frontal bone was selected in all pregnancies because the nasal bone was absent in about 30% of their Down syndrome fetuses.
Information on PT measurements and gestational age were obtained from our previous publications: 36 Down syndrome pregnancies (nine were in both series and one was measured in the third trimester) and 731 unaffected pregnancies. In addition we have subsequently measured 18 further Down syndrome cases and 119 controls using a RAB 4–8 L probe (Voluson 730 Professional, GE Medical Systems, Milwaukee, WI, USA) with a similar protocol. Pregnant women carrying mid-gestation Down syndrome fetuses that had been confirmed by karyotyping following amniocentesis were referred to our antenatal sonographic unit before termination of pregnancy. After giving their informed consent, they were scanned to determine fetal viability and biometry. Each fetus was scanned once and PT was measured. The indications for amniocentesis in the 18 additional cases were: abnormal nuchal translucency with or without serum biochemistry results in six, abnormal triple test results in five, advanced maternal age in five, and mid-gestation sonographic markers in two.
The additional controls were a consecutive series of pregnant women with singleton pregnancies recruited between January 2007 and January 2008 specifically for this study. None had any evidence of a chromosomal abnormality either prenatally or at birth. Similar to our previous studies1, 4, they all had normal fetal anatomy and adequate amniotic fluid. Exclusion criteria included cases with uncertain gestational age, those missing a first-trimester dating scan, multiple gestations, fetal malformations or various medical complications of pregnancy. Fetuses with a difference of 10 days or more between gestational age by last menstrual period and sonographic biometry were also excluded1, 4. All of these women were being scanned routinely to determine fetal viability, biometry and for ruling out fetal anomalies either at the antenatal sonographic unit or in the private facilities of the participating authors. Informed consent was obtained for the biometric measurements from all participants.
Each fetus was scanned during one visit. The data on maternal age, gestational age and fetal biometry at the time of the sonographic examination were stored in a computer database for later analysis2. We did not record ethnicity as there is now a high rate of intermarriage among individuals of widely different geographic origins in Israel2.
In addition, PT MoM values for 26 Down syndrome fetuses in the series of Persico et al. were extracted from Figure 2 in their paper3. In total the data from 80 cases and 850 controls were available for this study.
The median PT among unaffected pregnancies was calculated for each completed week of gestation and regressed against the median gestational age in days, weighted by the number of women. All values in both cases and controls were expressed as gestational age-specific MoMs using this regression equation. Goodness of fit for a log Gaussian model was assessed using probability plots of log MoM and deviations from the distribution were assessed using the Shapiro–Wilk test, after excluding outliers exceeding 3 SD from the mean. The model parameters, i.e. means and SD, were estimated as the median and the 10th–90th centile range divided by 2.563, respectively.
Likelihood ratios (LRs) were computed from the model for a wide range of PT values as the ratio of the heights of the Gaussian distributions in Down syndrome compared with unaffected pregnancies with the same MoM. Risks were calculated by multiplying the maternal age-specific risk, before the scan, expressed as odds by the LR. The prior risks were obtained from a published meta-analysis of birth prevalence studies5 and are shown in Table 1.
Table 1. Risk of Down syndrome (one in x) at term according to maternal age in years and months at estimated date of delivery5
|Maternal age(years)||Maternal age (months)|
- Top of page
There was a steady increase in the median PT with increasing gestational age among all unaffected pregnancies in the combined series (Table 2), which was fitted by a cubic regression curve with the equation:
where d is gestational age in days. At each week of gestation the median PT was higher in Down syndrome than in unaffected pregnancies. Using the regression equation to calculate MoMs the median PT in Down syndrome pregnancies for the combined series was 1.311 MoM; in unaffected pregnancies the median was 1.009 MoM. The SD of log10 MoM was 0.0754 in Down's syndrome and 0.0819 in unaffected pregnancies. There were 35% of Down syndrome cases with MoMs above the 95th centile of unaffected pregnancies and 59% above the 90th centile.
Table 2. Median prenasal thickness (PT) in unaffected pregnancies by completed weeks of gestation, together with regressed values, and in Down syndrome fetuses in our series
| ||Unaffected||Down syndrome|
|Gestational age (weeks)||n||Median PT (mm)||Regressed PT* (mm)||n||Median PT (mm)|
Figure 1 is a probability plot of MoM values in Down syndrome and unaffected pregnancies, in which a straight line indicates a log Gaussian distribution. There was no obvious deviation from fit for either the Down syndrome or the unaffected pregnancies over a wide range of MoMs. There was an apparent deviation in Down syndrome values above the 95th centile (1.62 MoM) with a maximum PT of 1.80 MoM. However, there was no statistically significant deviation using the Shapiro–Wilk test in either Down syndrome (P = 0.64) or unaffected (P = 0.54) pregnancies.
Figure 1. Log-normal probability plot of prenasal thickness (PT) in multiples of the median (MoM) for 80 Down syndrome (●) and 850 unaffected pregnancies (▪). The Shapiro–Wilk test showed no statistically significant deviation from the log-normal distribution in either Down syndrome (P = 0.64) or unaffected (P = 0.54) fetuses. Two values for unaffected pregnancies are not shown on this plot (0.48 MoM at − 2.9, 0.46 MoM at − 3.2).
Download figure to PowerPoint
Table 3 shows the median PT for every day of gestation between 14 + 0 weeks and 27 + 6 weeks. Table 4 shows the LR for every PT in 0.01-MoM increments between 0.80 and 1.79 MoM.
Table 3. Normal median prenasal thickness in mm according to gestational age in weeks and days, as calculated using our regression equation
| ||Gestational age (days)|
|Gestational age (weeks)||0||1||2||3||4||5||6|
Table 4. Likelihood ratios for Down syndrome according to prenasal thickness in multiples of the median (MoM)
| ||MoM increment|
|MoM||+ 0.00||+ 0.01||+ 0.02||+ 0.03||+ 0.04||+ 0.05||+ 0.06||+ 0.07||+ 0.08||+ 0.09|
To demonstrate how the tables are used to calculate risk, consider an example in which the PT was 3.60 mm at 16 weeks and 4 days in a woman aged 27 years and 7 months. From Table 3 the normal median is 2.81 mm so PT is 1.28 MoM (3.60/2.81); from Table 4 the LR is 2.38. From Table 1 the prior risk is one in 1200 or 1 : 1199 so the posterior odds are 2.38 : 1199, a risk of one in 506.
- Top of page
The tables presented here for maternal age-specific risks, normal median PT levels and LRs provide the sonographer with all the tools for a ‘bedside’ estimation of the Down syndrome risk in a pregnancy given the maternal age, gestational age and PT measurement.
The discriminatory power of using PT to modify the maternal age-specific risk can be assessed by modeling using the method of numerical integration6 with a standard maternal age distribution7. This yields a predicted detection rate of 50% for a fixed 5% false-positive rate. When PT is determined at the same time as other markers, such as nuchal skinfold and nasal bone length, the predicted detection rate is further increased. Modeling with the meta-analysis-based parameters for nuchal fold and nasal bone in Cuckle and Benn8 predicts a 79% detection rate for a 5% false-positive rate when all three ultrasound markers are used, taking account of the observed correlation between PT and nasal bone in unaffected pregnancies1.
PT cannot be used as an independent screening test in women who have already had first- or second-trimester Down syndrome screening. The LRs in Table 4 might be used to modify the risk from these tests; however, more data will be needed on any correlations between PT and other screening markers before this can be considered.
Neither our group nor that of Persico et al. has examined any possible relationship between ethnicity and PT. We did not record information on this factor as there is now a high degree of intermarriage between Israelis whose parents or grandparents emigrated from different geographical regions. Although nasal bone measurements are associated with ethnicity, there is no particular reason to expect this of PT. Hence, until information to the contrary is published, it is reasonable to assume that the tables we have calculated apply universally.
The tables of LR assume that the distribution parameters do not change with gestational age. There are no clear trends with gestational age in the present combined series. Although the Down syndrome median PT MoMs at 14–18 weeks (23 cases), 19–21 weeks (36 cases) and 22–27 weeks (21 cases) were 1.26, 1.28 and 1.38, respectively, the correlation coefficient between PT MoM and week of gestation was only 0.09 (P = 0.42). The corresponding SDs of log10 MoM for Down syndrome fetuses were 0.059, 0.082 and 0.066, respectively; values were 0.091, 0.077 and 0.080 in the unaffected pregnancies.
Unlike our series, in the study of Persico et al. 3D ultrasound imaging was used and PT was measured from the lowest part of the frontal bone, rather than the frontonasal angle, because nasal bone is absent in some cases3. Another slight difference was that Persico and colleagues captured the 3D volume of the fetal head and examined it offline using the multiplanar mode to confirm the exact mid-sagittal plane, whereas we always use real-time 2D sonography and the images were captured and measured at the ‘bedside’. However, these differences had no effect on the results. When the regression equation derived by Persico et al. is applied to the unaffected controls from our studies the median is 1.01 MoM. Furthermore, in our 54 Down syndrome pregnancies the median using this equation is 1.30 MoM, compared with a value of 1.33 MoM in the 26 cases of Persico et al.
One limitation of the present study is that the affected karyotypes were known at the time of the scan. This could have biased the results, although actual measurements are unlikely to be biased to the same extent as subjective determinations (such as absent nasal bone)2 and there was no obvious effect on the results. The distribution of PT values is log Gaussian over much of the range in both Down syndrome and unaffected pregnancies, whereas a bias would have distorted the distribution in Down syndrome fetuses.
Persico et al. did not tabulate the 26 individual PT measurements in Down syndrome cases in their series and we extracted them from a figure in the paper. This will have introduced some imprecision but it does not appear to have been great. Their results were expressed as delta values—deviations from the regressed median—rather than in MoMs, with a reported mean of 1.1 mm and standard error of 0.16 for the Down syndrome cases; using the extracted PT measurements we obtained a mean of 1.2 mm and standard error of 0.17. Moreover, any imprecision must be set against a greater accuracy of risk estimation from model parameters based on almost half as many more cases.
In summary, PT is a powerful second-trimester sonographic marker for Down syndrome. This may enhance the discriminatory power of other sonographic determinations, especially other facial landmarks. The tables provided in this paper allow ‘bedside’ estimation of Down syndrome risk without the need for computerized software or complicated calculations.