Comparison of first-trimester contingent screening strategies for Down syndrome

Authors


Abstract

Objective

To assess the relative performance of a multi-stage first-trimester screening protocol for fetal Down syndrome.

Methods

Data from 10 767 women who underwent combined ultrasound and biochemistry (BC) screening in the first trimester were reanalyzed using a contingent model approach. Amongst the 10 854 fetuses with known outcome, 32 had Down syndrome, 232 had other abnormalities and 10 590 were unaffected. Nuchal transluceny (NT), BC and combined (NT-BC) gestational age-specific risks were calculated for each individual using The Fetal Medicine Foundation risk calculation algorithms (Mixture Model and Biochemistry). Individual patients were categorized as at low, high or intermediate risk according to one of the following three strategies. In ‘Strategy-NT-BC’ initial screening was performed using both NT and BC. In ‘Strategy-BC’ initial screening was undertaken using maternal serum markers followed by NT assessment in those with an intermediate risk (1 : 51 < risk equation image) while in ‘Strategy-NT’ initial screening was undertaken using NT followed by serum marker assessment in those with an intermediate risk (1 : 51 < risk equation image). The nasal bone was assessed in those with an intermediate risk as the final stage in each of the three strategies. Those with an adjusted risk of 1 in 100 or higher after nasal bone assessment were reclassified as high risk. Detection and false-positive rates were compared between differing strategies in our local population, and this analysis was also performed with the maternal age for our population standardized to the distribution found in England and Wales.

Results

In our local population the detection rate for a 5% false-positive rate using a combined screening policy (NT-BC) was 88% (95% CI, 75.3–98.9%), and 2.3% had an absent nasal bone. The respective detection rate and false-positive rate of the three multi-stage screening strategies were: Strategy-NT-BC: 87.5 and 2.5%; Strategy-BC: 87.5 and 5%; Strategy-NT: 84.4 and 2.9%. In the contingent Strategy-BC only 29% of those initially screened using serum markers required an NT scan. If the model were applied to a hypothetical obstetric population standardized to the maternal age distribution in England and Wales, the detection and false-positive rates of the same three screening strategies would be: Strategy-NT-BC: 86.2 and 1.9%; Strategy-BC: 82.8 and 4%; Strategy-NT: 75.8 and 2.3%, respectively.

Conclusion

First-trimester contingent screening provides detection and false-positive rates comparable to those achieved using combined screening, but could be used to significantly reduce the number of scans performed. Copyright © 2010 ISUOG. Published by John Wiley & Sons, Ltd.

Introduction

First-trimester screening for trisomy 21 using a combination of fetal nuchal translucency (NT), maternal free beta-human chorionic gonadotropin (β-hCG) and pregnancy-associated plasma protein-A (PAPP-A) at 11–14 weeks' gestation is now established as an effective screening program. First-trimester screening has a quoted detection rate of approximately 90% for a 5% false-positive rate1, 2. The effectiveness of a combined first-trimester screening program is, however, strongly dependent on the training and continued audit of sonographers performing the ultrasound assessment3. While intersonographer measurement variance has been minimized by standardizing the protocols by which NT and other ultrasound parameters are measured, the necessity to perform a scan on all women booking for obstetric care requires considerable ultrasound resources if a combined screening approach is to be universally adopted.

‘Integrated’, ‘sequential’ and ‘contingent’ screening strategies have been proposed in an attempt to improve the detection rate, reduce the false-positive rate and improve cost-effectiveness4–7. Under contingent screening the individual risk for Down syndrome is classified as high, low or intermediate, with only those in the intermediate group requiring additional screening5. Classification of individuals into one of these three groups could therefore be based on their NT, biochemistry or combined risk for Down syndrome. One model is the first-trimester contingent screening proposed by Christiansen and Olesen Larsen8. In a recent study Vadiveloo et al.9 showed that, had they adopted this model, then only 29% of Scottish women screened between 2000 and 2005 would have required an NT scan. Their retrospective analysis demonstrated that a within-trimester contingent approach increased the overall cost-effectiveness of the screening program while at the same time achieving a similar detection rate (88.6 vs. 90.9%) and false-positive rate (5.8 vs. 6.4%) to those of combined screening. The authors did not, however, include additional ultrasound markers such as the presence or absence of the nasal bone to assess whether its inclusion would have further increased the detection rate of their within-trimester contingent strategy. Also the model proposed by Christiansen and Olesen Larsen8 and assessed by Vadiveloo et al.9 uses biochemical markers as the initial screening test rather than NT, even though screening using NT alone has a higher detection rate for Down syndrome in the first trimester than screening with biochemistry.

The objective of this study was to retrospectively compare the detection and false-positive rates of alternative within-first-trimester contingent screening strategies in both our local population and a maternal age-standardized population using the updated and recently published Fetal Medicine Foundation (FMF) risk determination models10, 11.

Methods

Study population

This was a retrospective study of 10 767 pregnant women identified within our clinical databases who attended the one-stop first-trimester combined screening program for Down syndrome in a university hospital during the period from January 2005 to May 2008 inclusive. Fetal NT, presence or absence of the nasal bone and crown–rump length (CRL) were measured using standardized techniques (www.fetalmedicine.com) on an HDI 5000 ultrasound machine (Philips Medical System, Seattle, WA, USA). Maternal blood samples were taken at the same visit for the determination of free β-hCG and PAPP-A levels. Maternal blood samples were measured immediately in our laboratory using a Kryptor analyser (Brahms Diagnostica GmbH, Berlin, Germany). Our laboratory is accredited by the FMF and is assessed on a monthly basis by the United Kingdom National External Quality Assessment Service (UKNEQAS www.ukneqas.org.uk). All scans were performed by sonographers accredited by the FMF to assess the fetal NT and nasal bone. Maternal demographic characteristics and ultrasound findings were recorded at the time of screening in both our local perinatal database as well as in a database provided within the FMF risk-calculation software. Fetal karyotype results were also entered into our perinatal database when they became available. Data on pregnancy outcomes were obtained either from our local maternity delivery database for those who delivered in our unit or via telephone calls to the patients themselves.

The maternal age- and gestational-specific background risks for trisomy 21 were calculated for each individual based on maternal age at expected date of delivery and adjusted for the gestational age at which combined screening was carried out12, 13. Gestational age at the time of examination was determined from the CRL measurement14. The likelihood ratio to determine adjusted risks for NT, biochemistry (BC) and combined NT and BC (NT-BC) were determined according to the description of the NT Mixture Model and conversion of free β-hCG and PAPP-A values to their equivalent multiples of the median (MoM) for gestational age10, 11. All BC-MoMs were corrected for weight, ethnicity, parity, smoking, method of conception and multiple pregnancy and chorionicity10, 15. Likelihood ratios for the presence or absence of the nasal bone were determined from models developed specifically for Chinese people, which take into account the interrelationship between nasal bone status and fetal CRL and NT16. Truncation limits were applied to the calculated likelihood ratios and the maximum risk was set as 1 in 2. Minimum likelihood ratios for NT, BC and combined NT-BC were 0.2, 0.14 and 0.05, respectively. The maximum permitted likelihood ratio for NT, BC and combined NT-BC was 50017. Bayes' theorem was applied throughout to produce an estimated risk for trisomy 21.

Contingent screening strategies

Contingent multistage strategies were assessed and their relative performance was compared to that of combined screening. In each strategy risks were stratified as being low, intermediate or high. Three strategies were assessed. In ‘Strategy-NT-BC’ all the women underwent combined screening and only those with an intermediate risk underwent a nasal bone assessment. In ‘Strategy-BC’ all pregnancies were initially screened using serum biochemistry markers, and those with an intermediate risk then had their NT assessed. Those whose combined risk still remained intermediate then had their nasal bone assessed. ‘Strategy-NT’ was similar to Strategy-BC except that the order of the NT and BC screening assessments was reversed.

Low-risk cut-off was defined as an estimated risk lower than 1 in 1000 for NT, BC and the combined NT-BC, and high-risk cut-off was defined as an estimated risk greater than or equal to 1 in 50 for NT, BC and combined NT-BC5, 18. Risks were deemed intermediate if they were between the high- and low-risk levels. Cases were classified as high risk if their risk after nasal bone assessment was greater than 1 in 100.

Statistical analysis

The detection rate and false-positive rate were determined for each multistage strategy as well as the proportion of our screened population requiring an additional assessment if their estimated risk was in the intermediate-risk category.

The performance of the multistage screening strategy in the UK general population was assessed as previously described by Cicero et al.19. A database of maternal ages at expected date of delivery was generated with the maternal age distribution matching that of live births in England and Wales between 2000 and 200220. This database was used to replace the maternal age in our own screened population, and all patient-specific risks for trisomy 21 were recalculated. The number of Down syndrome cases equivalent to that which would have been expected according to age and gestation in the England and Wales general population at which screening had been carried out were randomly selected from among the available cases of Down syndrome. The detection rate and false-positive rate were calculated for comparative assessment as well as the proportion requiring additional screening if their risk was intermediate.

Results

The database search identified 10 955 fetuses in 10 767 pregnancies screened during the selected time interval, of which 101 fetuses (0.9%) were lost to follow up. Of the 10 854 (99.1%) fetuses with known outcome, 32 (0.3%) had Down syndrome, 45 (0.4%) had other chromosomal abnormalities, 187 (1.7%) had non-chromosomal abnormalities and 10 590 (97.6%) were unaffected. The median maternal age at the expected date of delivery was 33.1 years, and 30.1% were 35 years of age or older. The percentage of pregnancies screened at 11, 12 and 13 weeks' gestation were 6.3%, 76.5% and 17.2%, respectively.

Of the pregnancies with known outcome 99.3% of the women were Chinese; 67.9% were nulliparous; 1.1% were smokers; 0.2% were in-vitro fertilization pregnancies; 1.8% were multiple pregnancies; and 0.2% had a history of previous Down syndrome. The nasal bone was absent in 2.3% of screened fetuses. The number of Down syndrome fetuses expected according to the gestation at which screening had been performed and maternal age at expected date of delivery was 36. The mean of the log10 transformed free β-hCG-MoM, PAPP-A-MoM and NT-MoM distributions in the unaffected pregnancies were 0.02, 0.02 and 0.0, respectively.

For a 5% false-positive rate, the detection rates by NT alone, BC alone and combined NT-BC were 69% (95% CI, 52.7–84.8%), 62% (95% CI, 45.7–79.3%) and 88% (95% CI, 75.3–98.9%), respectively, in our local population. This would be equivalent to a gestation-specific risk cut-off for NT alone, BC alone and combined NT-BC of 1 in 205, 1 in 65 and 1 in 165, respectively. Table 1 summarizes the performance of each of the three multistage strategies in our locally screened population.

Table 1. Distribution of risks and effectiveness of differing contingent screening strategies in our local population of 32 Down syndrome cases and 10 590 unaffected cases
 Risk estimated after first-stage screeningRisk estimated after second-stage screening  
Fetal karyotype according to adopted strategyHigh risk*Intermediate riskLow riskHigh risk*Intermediate riskLow riskNasal bone assessment; high risk§Total false-positive rate and detection rate
  • Data are given as %. Patients with a risk of 1 in 50 or higher were considered screen positive. Those with a risk less than 1 in 1000 were considered screen negative. The nasal bone was used to modify the risk of patients with an intermediate risk only. Those whose risk was 1 in 100 or higher after assessment of the nasal bone were considered screen positive.

  • *

    Risk ≥ 1 in 50.

  • †, ‡

    Risk 1 in 51 to 1 in 1000.

  • Risk ⩽ 1 in 1000.

  • §

    Risk ≥ 1 in 100. BC, biochemistry; NT, nuchal translucency.

Strategy-NT-BC
 Euploid2.213.983.9Not applicable0.42.5
 Trisomy 2184.43.112.5 3.187.5
Strategy-BC
 Euploid4.129.466.54.69.785.70.45.0
 Trisomy 2153.140.66.387.50.012.50.087.5
Strategy-NT
 Euploid1.627.471.02.47.390.30.52.9
 Trisomy 2159.431.39.481.33.115.63.184.4

Under Strategy-NT-BC, 27 (84.4%) of the 32 fetuses with Down syndrome would have been classified as high risk, four as low risk and one as intermediate risk. The case with an intermediate risk had an absent nasal bone. The overall detection and false-positive rates would have been 87.5% (95% CI, 76.0–99%) and 2.5% (95% CI, 2.2–2.9%), respectively, with 1481 (13.8%) of all the screened women requiring an assessment of the nasal bone.

Under Strategy-BC, 17 of the 32 (53.1%) Down syndrome cases would initially have been classified as high risk, two as low risk and 13 as intermediate risk. Had these intermediate-risk cases undergone combined screening then 11 of the 13 Down syndrome cases would have been reclassified as high and two as low risk. The overall detection and false-positive rates for Strategy-BC would have been 87.5% (95% CI, 76–99%) and 5% (95% CI, 4.6–5.4%), respectively, with 3125 (29.0%) of those initially screened requiring an NT scan, and a third of these would have required that their nasal bone be assessed.

Under Strategy-NT, 19 of the 32 (59.4%) Down syndrome cases would initially have been classified as high risk, three as low risk and 10 as intermediate risk. Had these intermediate-risk cases undergone combined screening then seven of the 10 Down syndrome cases would have been reclassified as high and two as low risk, and one would have remained intermediate. This remaining intermediate-risk case had an absent nasal bone and would have been reclassified as high risk on assessment of the nasal bone. The overall detection and false-positive rates for Strategy-NT would have been 84.4% (95% CI, 71.2–97%) and 2.9% (95% CI, 2.6–3.3%), respectively, with 2908 (27.0%) of those initially screened requiring maternal serum marker assessment, and 26.6% of these would have required that their nasal bone be assessed.

The median age at expected date of delivery of the hypothetical age-standardized population of England and Wales used to assess screening performance was 29.8 years, and 19.1% were 35 years of age or older. The expected number of cases of Down syndrome based on generated maternal age at expected date of delivery and gestational age at which screening was carried out was 29, with an expected incidence of Down syndrome of 0.3%. For a 5% false-positive rate, the detection rates by NT alone, BC alone and combined NT-BC would have been 66%, 62% and 86%, respectively, for our age-standardized population. This would be equivalent to a gestation-specific risk cut-off for NT alone, BC alone and combined NT-BC of 1 in 238, 1 in 85 and 1 in 200, respectively. Table 2 summarizes the estimated performance of each of the three multistage strategies in a population with the maternal age distribution of live births in England and Wales between 2000 and 2002.

Table 2. Distribution of risks and effectiveness of differing contingent screening strategies applied to our population using a maternal age distribution matching that found in a general population of live births in England and Wales between 2000 and 200220, with the estimated performance modeled using 29 Down syndrome cases and 10 590 unaffected pregnancies
 Risk estimated after first-stage screeningRisk estimated after second-stage screening  
Fetal karyotype according to adopted strategyHigh risk*Intermediate riskLow riskHigh risk*Intermediate riskLow riskNasal bone assessment; high risk§Total false-positive rate and detection rate
  • Data are given as %. Patients with a risk of 1 in 50 or higher were considered screen positive. Those with a risk less than 1 in 1000 were considered screen negative. The nasal bone was used to modify the risk of patients with an intermediate risk only. Those whose risk was 1 in 100 or higher after assessment of the nasal bone were considered screen positive.

  • *

    Risk ≥ 1 in 50.

  • Risk 1 in 51 to 1 in 1000.

  • Risk ⩽ 1 in 1000.

  • §

    Risk ≥ 1 in 100. BC, biochemistry; NT, nuchal translucency.

Strategy-NT-BC
 Euploid1.611.986.5Not applicable0.41.9
 Trisomy 2182.83.413.8 3.486.2
Strategy-BC
 Euploid3.324.072.73.77.988.40.34.0
 Trisomy 2144.8544.8510.379.33.417.23.482.8
Strategy-NT
 Euploid1.219.679.21.95.892.30.42.3
 Trisomy 2158.624.117.272.43.424.13.475.8

Discussion

The retrospective reanalysis of our existing screened pregnancies using the revised and updated FMF screening algorithms by a combined NT-BC screening policy had an estimated detection rate of 88% for a false-positive rate of 5%. However, to achieve this, the high-risk cut-off used within our locally screened population would need to be revised from the present 1 in 300 to 1 in 16521, 22. If the high-risk cut-off level were to remain unchanged, the false-positive rate would increase to over 8% simply as a result of changing from one version of the FMF risk-calculation software to another.

If Strategy-NT or Strategy-BC had been adopted, in our local or the England and Wales population the estimated detection rates would have been within the 95% CI limits of that achieved by either combined NT-BC screening or contingent screening using combined NT-BC in conjunction with nasal bone assessment in those with an intermediate risk. In addition, both these strategies offer the potential for considerable cost reduction. In centers where unit costs for performing a blood test are lower than those of performing a scan, our analysis would indicate that only 29% of all screened cases would have required an NT scan if Strategy-BC had been adopted. Vadiveloo et al.9 reported a similar detection rate and reduction in the number of NT scans required for a 5% false-positive rate using a two-stage screening strategy, while Christiansen and Olesen Larsen8 reported even greater reductions in the number requiring an NT scan (19%) but at the expense of a reduced detection rate of 80%.

Inclusion of nasal bone assessment within a contingent screening strategy in our local as well as the England and Wales general population gave similar reductions in the false-positive rate to that reported in series published by the FMF Group18, 19. Of the 1025 unaffected cases in our locally screened population with an intermediate risk after combined screening using Strategy-BC, 39 (3.8%) had an absent nasal bone and only 11.7% of the initially screened population would have needed to undergo both an NT and a nasal bone scan. No additional visit would be required if the nasal bone assessment could be performed immediately after NT assessment in those whose combined risk still remained intermediate. Our findings would further support the argument for the inclusion of nasal bone assessment as an additional marker in either a two-stage or three-stage contingent strategy provided that sonographers assessing the nasal bone are adequately trained and regularly audited both internally and externally.

If a within-trimester contingent screening approach, such as Strategy-BC, were to be adopted then firstly women seeking screening for Down syndrome will need to have their pregnancy accurately dated to provide a reliable and accurate estimation of gestational age. If estimates of gestational age are unreliable then the calculated risk for Down syndrome would be equally unreliable, and screening performance would thus be adversely affected. Secondly, women may need to be screened earlier in their pregnancy in order to facilitate the performance of an NT scan in those requiring one. Should BC screening be performed earlier, for example at 8 weeks, then existing detection and false-positive rates could potentially be improved owing to the larger differences in PAPP-A levels between affected and unaffected pregnancies23. Additional or alternative markers such as ADAM-12 and Inhibin A could also be considered24, 25. Whether detection rates can be improved or false-positive rates reduced will also depend on whether laboratory and ultrasound standards are maintained. It is estimated that a deviation of 10% on a marker-MoM is sufficient to change the false-positive rate by approximately 2%26. Within our own center all sonographers are audited and re-certified by the FMF on an annual basis, and in addition our laboratory is assessed by UKNEQAS on a monthly basis.

Our analysis was a simulation study to assess the likely effects of changes in the screening algorithms as well as to assess the potential efficacy of a within-trimester contingent screening policy using these new algorithms. While both the SURUSS and FASTER studies compared clinical efficacy between the first- and second-trimester screening27, 28, assessment of contingent approaches has been based on modeling and simulation of existing data5–9, 29. A prospective study or randomized controlled trial comparing within-trimester contingent strategies vs. classical combined screening would need to be performed to assess their efficacy in actual clinical practice.

There are some limitations to our study. Firstly our study population was a homogeneous group consisting predominantly of Chinese women. While NT measurements have been reported not to depend on ethnicity, serum markers have been shown to be dependent on ethnicity10, 15. The adjustment factors for ethnicity employed in our study were derived from those recently reported by the FMF, and should therefore have resulted in a corrected MoM centered on 16, 10. Secondly, there were relatively few affected cases available for analysis, which is reflected in the relatively wide 95% confidence intervals for the detection rate. The incidence of Down syndrome in our study population of 0.3% was as expected and in accordance with our population characteristics. Thirdly, the incidence of absent nasal bone has been shown to be dependent on ethnicity. The incidence of absent nasal bone in our study was 2.2%, higher than the 1.2% previously reported by Cicero et al.19, but comparable with the 2.6% recently reported by Kagan et al.18. Both these studies were performed in predominantly Caucasian populations containing few East Asians. The negative and positive likelihood ratios in Chinese people have been shown to be different from those of Caucasians for the corresponding CRL and NT measurements16.

In conclusion, our study has demonstrated that within-first-trimester contingent screening for Down syndrome provides overall detection and false-positive rates comparable with those achieved using combined screening, while significantly reducing the number of NT scans performed.

Acknowledgements

We wish to thank the Fetal Medicine team for their dedication and professionalism in facilitating the performance of this study.

Ancillary