An increase in cost-effectiveness of first trimester maternalscreening programmes for fetal chromosome anomalies isobtained by contingent testing

Michael Christiansen* and Severin Olesen Larsen

Department of Clinical Biochemistry, Statens Serum Institut, Copenhagen, Denmark

We assessed the discriminatory efficiency and cost-effectiveness of a novel way of organising first trimesterscreening for Down syndrome (DS), contingent testing, where a serological test (PAPP-A and b-hCG: thedouble test) is made in early first trimester and followed by nuchal translucency testing (NT) only in womenwith an intermediate risk, e.g. <1 : 65 and >1 : 1000, and not in all women as in normal first trimesterscreening (NFTS). Using Monte Carlo simulation contingent testing had a detection rate (DR) of 78.9%and a false-positive rate (FPR) of 4.0% for DS with 19.4% of women offered NT testing. The DR of NFTSwas 85.5% and the FPR 4.4%. The decrease in NT screening was associated with an increase from 23% to29% in the proportion of DS cases born. The cost of the contingent testing programme was £53 000 per DScase not born and £91 000 in NFTS. The number of aborted fetuses per DS case were 0.35 and 0.36,respectively. Thus, contingent testing is an organisation of first trimester screening where costs can bereduced with a marginal decrease in performance. Contingent testing is attractive in areas where NTscreening is the bottleneck preventing the introduction of first trimester screening. Copyright # 2002 JohnWiley & Sons, Ltd.

KEY WORDS: prenatal screening; Down syndrome; mathematical modelling; fetal medicine

INTRODUCTION

Maternal serum screening for fetal chromosomeabnormalities or malformations in the second tri-mester is widely used to assess whether the risk forsuch conditions is so high that it justifies the 1% risk offetal death associated with a diagnostic amniocentesis(AC) (Tabor et al., 1986) and the cost of karyotyping.Depending on the age distribution of pregnant women,two-thirds of Down syndrome (DS) cases and >85%of severe malformations can be detected, for a false-positive rate (FPR) of 5%, by measuring the serumconcentrations of alpha-fetoprotein (AFP), humanchorionic gonadotrophin (hCG), and unconjugatedestriol (uE3) and combining these results with the riskassociated with the mother’s age (Cuckle, 1996, 2000).

It is now possible to perform serum screening forDS and chromosome anomalies as early as in gesta-tional weeks 8–14 using pregnancy-associated plasmaprotein-A (PAPP-A) and the free b-form of hCG(b-hCG) as serological markers with a detection rate(DR) of 70% for a FPR of 5% (Cuckle and van Lith,1999). In weeks 10–14 the thickness of the nuchaltranslucency (NT) of the fetus is a very effectivediscriminator between normal pregnancies and preg-nancies with a fetus with a chromosome anomaly orother severe syndrome (Snijders et al., 1998; Spenceret al., 1999). The combined use of serological testingusing PAPP-A and b-hCG and ultrasound screeningusing NT in weeks 11–13 has been demonstrated to

have a DR for DS of >90% for a FPR of 7% in aprospective study (Spencer et al., 2000). It is, onaccount of the risk of causing fetal death (Tabor et al.,1986), important that screening programmes aredesigned so as to reduce the number of invasivediagnostic procedures. The combination of firsttrimester serological testing and NT followed bysecond trimester biochemical screening, the so-called‘integrated screening’ can, at the price of delaying therisk calculation and reporting until the secondtrimester, reduce the FPR to 0.9% for a DR of 85%(Wald et al., 1999). However, the late reporting of riskresults is a disadvantage from a psychological as wellas a medical point of view.

With the availability of a highly efficient firsttrimester serological test (Cuckle and van Lith, 1999)and an even better first trimester ultrasonographic test,the NT measurement, it is rational to examine whetherit is cost-effective to offer both tests to all pregnantwomen requesting prenatal screening, as many womenwill be strongly positive or strongly negative in bothtests. An alternative would be contingent testing, i.e.so that only women with a risk, as determined by theserological test (which is simple to do and which isnormally centralised), above a certain risk cut-off, e.g.1 : 1000, is offered a NT measurement (which requireslocal training, expensive equipment and extensivequality control and therefore is more difficult toimplement). At the same time women with a veryhigh risk (e.g. >1 : 65), as determined by serologicaltesting, could be offered a chorionic villous sampling(CVS) directly, without the NT measurement. Thecontingent testing makes it possible to reduce theredundancy in testing, and reduce the cost of first

*Correspondence to: M. Christiansen, Department of ClinicalBiochemistry, Statens Serum Institut, 5 Artillerivej, DK 2300 S,Denmark. E-mail: mic@ssi.dk

PRENATAL DIAGNOSIS

Prenat Diagn 2002; 22: 482–486.Published online in Wiley InterScience (www.interscience.wiley.com). DOI: 10.1002 /pd.336

Copyright # 2002 John Wiley & Sons, Ltd. Received: 14 September 2001Revised: 19 December 2001

Accepted: 21 December 2001

trimester screening dramatically. Furthermore, it faci-litates an early introduction of first trimester screeningin areas where the capacity of NT screening does notmake it possible to offer it to all women. Such areascould benefit from offering a double test to everyoneand referring all those individuals with intermediaterisk to NT testing at a centralised facility. As thedouble test seems to perform better earlier than week11 (Cuckle and van Lith, 1999), such an arrangement,where the serological test is performed prior to the NTmeasurement, may represent better use of the availabletests – and make it possible to give a proportion ofwomen a final risk estimate prior to week 11.

We examined the efficiency of contingent testing inthe first trimester, i.e the use of the double test(PAPP-A and b-hCG and maternal age), followed byNT screening, if the risk is not high enough to justify aCVS/AC directly, and not small enough to makefurther testing ineffective. We compared the diagnosticefficiency of contingent testing with other screeningpossibilities in the first and second trimesters. Further-more, we compared the cost-effectiveness of thesestrategies and calculated the cost of detecting one caseof DS in several prenatal screening strategies.

PATIENTS AND METHODS

Statistical calculations

The distribution of serological markers was obtainedfrom Cuckle and van Lith (1999) for the first trimesterand Wald et al. (1994) for the second trimester. TheNT distribution was from a large UK multicentrestudy (Nicolaides et al., 1998). The maternal agedistribution used was standardised (van de Veen et al.,1997). A priori risk values were based on the formulafrom Cuckle et al. (1987) and refer to the probabilityof giving birth to a child with DS. Using a MonteCarlo procedure (Larsen et al., 1998) we evaluated theperformance of different screening modalities. Theperformance of contingent testing was assessed byfirst establishing the distribution of risk values, a,as a result of the serological screening using theMonte Carlo simulation (Larsen et al., 1998). The NTmeasurement results in a likelihood ratio (DS fetus/normal fetus), r, and the final risk for a particularpregnancy is then: arr. This gives a relation betweena, r and final risk, e.g. if the final risk has to be>1 : 400 it follows that r>0.0025/a. Using thepublished NT distribution (Cuckle and van Lith,1999) and the formulae for the distribution of NTlog MoM (multiples of the median of normal values)in normal and DS pregnancies it is found that fora certain value of the likelihood ratio, r, we have:log10 MoM NT=x0.1076+0.2995rd(0.7863+log10 r)(see the Appendix). It is seen that we cannot havelog10 r<x0.7863, a reflection of the larger standarddeviation for DS cases as compared to normalpregnancies. It has the consequence that if theserologically defined risk, a, is >0.0153 then no NTmeasurement can reduce the final risk to a value

<1 : 400. Such calculations are performed for differentdouble test cut-offs.

Cost estimates

Recent cost estimates of prenatal diagnostic proce-dures from a Dutch Survey on Prenatal Diagnostics(Health Council of The Netherlands, 2001) are used.The exchange rate 1 Euro=£0.60 is used. The cost of aCVS or AC, including counselling and karyotyping, is£386.82. Triple test, quadruple test and double testcosts, including counselling, are £34.49, £44.02 and£34.49, respectively. NT measurement and counsellingcosts £48.10. The combined NT, double test andcounselling costs £75.33. The integrated test includingcounselling costs £95.99. Ultrasonography for gesta-tional dating costs £20.42. The level of cost reportedhere is in agreement with experience from Denmarkand the UK (Wald et al., 1997) and the US (Vintzileoset al., 2000). Even though an attempt has been madepreviously to calculate the detailed lifetime costs of aDS patient (Gill et al., 1987), we have chosen to usethe conservative estimate of £35 000 for the treatmentand care costs per case of live-born DS (Vintzileoset al., 2000).

RESULTS

The contingent screening procedure is depicted inFigure 1 for final risk cut-off of 1 : 400. All pregnantwomen are offered an initial double test (PAPP-A,b-hCG and maternal age) in the early first trimester(<week 11) and are only offered a NT measurement ifthe risk is >1 : 1000. Individuals with a risk below1 : 1000 after the double test are informed that the riskis too low to warrant further investigations. In theexample in Figure 1 this group will comprise the largemajority of normal pregnancies (79.2%) and a mino-rity, (12.6%) of DS pregnancies that without interven-tion would have led to the birth of a DS child. Theremaining pregnant women are either offered CVS/ACdirectly, if their risk is so high that NT measurementcannot bring it down below the risk cut-off (1 : 400)necessary for the recommendation af an invasivediagnostic test, or they are offered a NT measurementfor a more precise definition of risk. If the cut-off 1 : 65is used to separate the latter two groups, the groupoffered an invasive test directly will comprise 1.4% ofall normal pregnancies and 45.7% of DS pregnancies,and the group offered NT screening will comprise19.4% of normal pregnancies and 41.7% of DS pre-gnancies. Of the women offered NT screening, 2.6% ofnormal pregnancies and 33.2% of DS pregnancies willhave a final risk >1 : 400 and be offered CVS or AC.In total, the contingent first trimester screening willlead to a FPR of 4.0% and a DR of 78.9% for a finalcut-off risk of 1 : 400, and a FPR of 2.3% and a DR of67.6% for a final risk cut-off of 1 : 250.

Table 1 shows the DR and FPR for different cut-offvalues of the initial double test, and the final cut-offrisk of 1 : 400. It is noteworthy that the FPR is very

EFFECTIVENESS OF DOWN SYNDROME TESTING 483

Copyright # 2002 John Wiley & Sons, Ltd. Prenat Diagn 2002; 22: 482–486.

low and is only slightly dependent on the cut-offvalues of the double test. It is also seen that the DR isonly marginally reduced for a great reduction in the

NT frequency. However, if the final risk cut-offchanges then the changes in FPR and DR will becomparable to the changes in the total performance ofthe involved tests.

Using the data in Table 1 and knowledge of thedistribution of serological and ultrasound markers, weused a previously described Monte Carlo simulationprocedure to compare the performance of contingentscreening with that of normal first trimester (doubletest and NT), double test alone, and NT screeningalone (Table 2). We estimated that only 80% of thepopulation would accept the offer of screening, andthat only 90% of women with a final risk >1 : 400would have a diagnostic examination performed.

The cost-effectiveness of different screening strate-gies was evaluated by calculating the price spent perDS child not born, and, as a second expression of cost-effectiveness, the number of normal fetuses expected tobe aborted per DS child not born (Table 2). Therelative number of DS cases expected to be born is

Figure 1—Flow diagram of contingent testing. A final risk cut-off of 1 : 400. All pregnant women are offered a double test, prior to week 11,and if the risk is very high – in this case >1 : 65 – no result in the NT screening can bring the final risk to <1 : 400, which is the cut-off risklevel for CVS/AC chosen here. In consequence, these high-risk women are offered invasive testing without NT screening. If the risk at thedouble test is sufficiently low – in this case <1 : 1000 – the likelihood that the final risk will be >1 : 400 is so small that NT screening is notoffered to these low-risk women. Women with a risk >1 : 1000 and <1 : 65 are offered NT, and the size of this group depends on the doubletest cut-off level as shown in Table 1. The number of pregnant women and fetal DS cases that will be born are given for each category

Table 1—The frequency of nuchal translucency (NT)screening and overall screening performance in contingenttesting with different risk cut-off values for the initialdouble testa

Double testrisk cut-off

NT frequency(%)

DR(%)

FPR(%)

1 : 1000 19.4 78.9 4.01 : 800 17.6 77.3 3.81 : 533 12.5 73.9 3.61 : 400 9.8 70.9 3.3

aIf the risk cut-off is 1 : 1000, women with a double test resulting in arisk of >1 : 1000 will be offered NT screening. The final risk cut-offused was 1 : 400.DR, Detection rate, FPR, false-positive rate.

Table 2—Performance and cost-effectiveness of different screening strategies for the detection of Down syndrome (DS)a

StrategyDR(%)

FPR(%)

Total cost(£ millions)

Cost/DS not born(£ thousands)

Aborted fetuses/DSnot born

Reduction inDS born (%)

Contingent testing 78.9 4.0 19.6 53 0.35 71Double test 74.2 9.6 21.5 62 0.90 67Nuchal translucency (NT) 74.0 5.7 27.1 78 0.59 67Double test+NT (NFTS) 85.5 4.4 36.3 91 0.36 77

aUptake of screening: 80%. CVS/AC rate: 90%. Prevalence of DS at birth: 1.3 : 1000. Calculated for 500 000 births per year. Risk cut-off 1 : 400.See Patients and methods section for details of the cost estimates.DR, Detection rate; FPR, false-positive rate; NFTS, normal first trimester testing.

M. CHRISTIANSEN AND S. O. LARSEN484

Copyright # 2002 John Wiley & Sons, Ltd. Prenat Diagn 2002; 22: 482–486.

also calculated for each strategy, but a cost–benefitcalculation, based on placing an average cost tosociety of the birth of a DS child was not performed.Partly because the pricing is difficult, and perhapsimpossible, and partly because even low estimates ofthe lifelong costs of a DS child (i.e. £35 000; seePatients and methods section) will have the conse-quence that the mere elimination of 10% of DS caseswill finance any of the possible screening modalities.

DISCUSSION

We have shown that a first trimester serum/NTscreening programme with only 20% of the screenedwomen offered an NT measurement is a highlyefficient form of DS screening as the majority of NTmeasurements, in screening programmes where this isoffered to all women, is expended on women eitherwith such a high risk that they will end up above thecut-off of 1 : 400 for invasive testing irrespective of theNT measurement, or with such a low risk that they cannever reach that cut-off (Figure 1). If the cut-off ischanged to 1 : 250 then the frequency of NT measure-ments as a function of the lower cut-off for the doubletest will nearly be the same, but the DR will decreasewith the FPR. As the performance of a screening pro-gramme using NT is dependent on the strict adherenceto quality control regimens, e.g. as provided by thecertification programme of the Fetal Medicine Foun-dation, and the availability of trained staff and ultra-sound apparatuses, the availability of high-quality NTscreening must be considered a potential bottleneckfor the introduction of first trimester screening.

The principle of reducing redundancy in multiplemarker testing can be applied to other screeningprogrammes and may allow for a more effective,financially as well as medically, use of resources.

A further perspective would be the application ofother ultrasound markers to reduce the frequency ofpregnancies subjected to CVS even further. A promis-ing example is the use of Doppler ultrasound assess-ment of ductus venosus blood flow in reducing thenumber of invasive tests – this should reduce thenumber of invasive test by 90%, with a marginaldecrease in DR (Matias et al., 1998). The cost-effectiveness of restricting the number of pregnantwomen offered a certain test increases dramaticallywith the cost of the test.

The present findings are corroborated by theindependent finding that pregnancies in whichthe risk is either very high or very low do not benefitfrom repeat testing in second trimester screening. Onlyfor samples close to the risk cut-off is it of anyconsequence to repeat a test result (Cuckle et al., 1994;Hackshaw et al., 1995).

The use of contingent screening could be a waytowards the rapid introduction of first trimester screen-ing in parts of the world where NT measurementscannot at present be offered to all with the necessarylevel of quality control. One could also use con-tingent testing to select women that should be offered

a triple test or another examination, but if the secondtest is cheap and easily available it may be moreadvantageous to offer it to all women withoutrestriction. In the case of other tests being offeredin the second trimester it should be remembered thatthe increased test performance is achieved at theexpense of later diagnosis and risk reporting, and itshould be carefully evaluated whether a marginalincrease in performance is worth the longer period ofwaiting for the final risk evaluation.

A first trimester ultrasound examination may revealother pathological conditions (Nicolaides et al., 1998)that are not identified by serological screening, e.g.malformations of the heart (Hyett et al., 1999),however screening for such conditions has not reacheda stage where it can be universally recommended, andit is necessary to look particularly for specific signs ofmalformations, making the malformation scan differ-ent from the NT measurement in requirements oftraining of staff, etc (Johnson et al., 1997). The vastpublished clinical experience with second trimesterultrasound screening has been so disappointing(Jorgensen et al., 1999), particularly when comparedwith serological testing, that the introduction ofultrasonographic screening methods based on para-meters other than NT per se must be based oncarefully controlled clinical studies, and a decision onclinical implementation must depend on a favorablemedical technology assessment. Thus, for years tocome the major application of first trimester ultra-sound will be measurement of NT thickness. We haveshown here that this technique need not to be extendedto the whole population in order to reach asatisfactory first trimester screening performance.

APPENDIX

The logMoM value of a variable, x, correspondingto a particular log likelihood ratio (DS fetus/normalfetus) value, r, is obtainable by calculating in thenormal distribution. If m is the mean logMoM in DScases, s is the standard deviation (SD) of logMoMvalues in DS cases and w is the SD of logMoM valuesin normal pregnancies, this gives:

r ¼ 0:4343� x2=2s2 � x�mð Þ2=2w2h i

þ log10 s=wð Þ

with the solution

x ¼ �0:8686ms2 þpD

� �= 0:8686 w2 � s2

� �� �

where

D ¼ 0:7544m2s2w2�

3:4744 s2w2 w2 � s2� �

log s=wð Þ � log r½ �:

If the variable is NT we have m=0.305, s=0.12 andw=0.235, and (3) becomes

x ¼ �0:1076þ 0:2995p

0:7863þ log rð Þ:

(1)

(2)

(3)

EFFECTIVENESS OF DOWN SYNDROME TESTING 485

Copyright # 2002 John Wiley & Sons, Ltd. Prenat Diagn 2002; 22: 482–486.

REFERENCES

Cuckle H. 1996. Established markers in second trimester maternalserum. Early Hum Dev 47 (Suppl.): S27–S29.

Cuckle H. 2000. Biochemical screening for Down syndrome. EurJ Obstet Gynecol Reprod Biol 92: 97–101.

Cuckle HS, van Lith JM. 1999. Appropriate biochemical parametersin first-trimester screening for Down syndrome. Prenat Diagn19: 505–512.

Cuckle HS, Wald NJ, Thompson SG. 1987. Estimating a woman’srisk of having a pregnancy associated with Down’s syndromeusing her age and serum alpha-fetoprotein level. Br J ObstetGynaecol 94: 387–402.

Cuckle H, Densem J, Wald N. 1994. Repeat maternal serum testingin multiple marker Down’s syndrome screening programmes.Prenat Diagn 14: 603–607.

Gill M, Murday V, Slack J. 1987. An economic appraisal ofscreening for Down’s syndrome in pregnancy using maternalage and serum alpha fetoprotein concentration. Soc Sci Med24: 725–731.

Hackshaw AK, Densem J, Wald NJ. 1995. Repeat maternal serumtesting for Down’s syndrome screening using multiple markerswith special reference to free alpha and free beta-hCG. PrenatDiagn 15: 1125–1130.

Health Council of the Netherlands. 2001. Prenatal Screening:Down’s Syndrome, Neural Tube Defects, Routine Ultrasonography.2001/11. Health Council of the Netherlands: The Hague.

Hyett JA, Perdu M, Sharland GK, Snijders RJM, Nicolaides KH.1999. Using fetal nuchal translucency to screen for majorcongenital cardiac defects at 10–14 weeks of gestation: populationbased cohort study. BMJ 318: 81–85.

Johnson SP, Sebire NJ, Snijders RJM, Tunkel S, Nicolaides KH.1997. Ultrasound screening for anencephaly at 10–14 weeks ofgestation. Ultrasound Obstet Gynecol 9: 14–16.

Jorgensen FS, Valentin L, Salvesen KA, et al. 1999. MULTISCAN –a Scandinavian multicenter second trimester obstetric ultrasoundand serum screening study. Acta Obstet Gynecol Scand 6: 501–510.

Larsen SO, Christiansen M, Norgaard-Pedersen B. 1998. Calcula-tion of roc curves in multidimensional likelihood ratio basedscreening with Down’s syndrome as a special case. J Med Screen5: 57–62.

Matias A, Gomes C, Flack N, Montenegro N, Nicolaides KH. 1998.Screening for chromosomal abnormalities at 10–14 weeks: therole of ductus venosus blood flow. Ultrasound Obstet Gynecol12: 380–384.

Nicolaides KH, Snijders RJ, Cuckle HS. 1998. Correct estimation ofparameters for ultrasound nuchal translucency screening. PrenatDiagn 18: 519–523.

Snijders RJ, Noble P, Sebire N, Souka A, Nicolaides KH. 1998. UKmulticentre project on assessment of risk of trisomy 21 bymaternal age and fetal nuchal-translucency thickness at 10–14weeks of gestation. Fetal Medicine Foundation First TrimesterScreening Group. Lancet 352: 343–346.

Spencer K, Souter V, Tul N, Snijders R, Nicolaides KH. 1999. Ascreening program for trisomy 21 at 10–14 weeks using fetalnuchal translucency, maternal serum free beta-human chorionicgonadotropin and pregnancy-associated plasma protein-A. Ultra-sound Obstet Gynecol 13: 231–237.

Spencer K, Spencer CE, Power M, Moakes A, Nicolaides KH. 2000.One stop clinic for assessment of risk for fetal anomalies: areport of the first year of prospective screening for chromo-somal anomalies in the first trimester. Br J Obstet Gynaecol107: 1271–1275.

Tabor A, Philip J, Madsen M, Bang J, Obel EB, Norgaard-PedersenB. 1986. Randomised controlled trial of genetic amniocentesis in4606 low-risk women. Lancet 1: 1287–1293.

van der Veen WJ, Beekhuis JR, Cornel MC, Mantingh A, de WalleHE, de Wolf BT. 1997. A demographic approach to theassessment of Down syndrome screening performance. PrenatDiagn 17: 717–724.

Vintzileos AM, Ananth CV, Smulian JC, Day-Salvatore DL,Beazoglou T, Knuppel RA. 2000. Cost-benefit analysis of prenataldiagnosis for Down syndrome using the British or the Americanapproach. Obstet Gynecol 95: 577–583.

Wald NJ, Densem JW, Smith D, Klee GG. 1994. Four-markerserum screening for Down’s syndrome. Prenat Diagn 14: 707–716.

Wald NJ, Kennard A, Hackshaw A, McGuire A. 1997. Antenatalscreening for Down’s syndrome. J Med Screen 4: 181–246.

Wald NJ, Watt HC, Hackshaw AK. 1999. Integrated screening forDown’s syndrome on the basis of tests performed during the firstand second trimesters. N Engl J Med 341: 461–467.

M. CHRISTIANSEN AND S. O. LARSEN486

Copyright # 2002 John Wiley & Sons, Ltd. Prenat Diagn 2002; 22: 482–486.