Screening for trisomy 21 in twin pregnancies in the ®rsttrimester using free b-hCG and PAPP-A, combined withfetal nuchal translucency thickness

Kevin Spencer*

Endocrine Unit, Clinical Biochemistry Department, Harold Wood Hospital, Gubbins Lane, Romford, Essex RM3 0BE, UK

In the ®rst trimester of pregnancy the biochemical markers free b-hCG and pregnancy associated plasmaprotein-A (PAPP-A) are used for the prenatal screening of trisomy 21, either alone or in combination withnuchal translucency (NT) thickness. In this study, I have analysed the distribution of these biochemicalmarkers in 159 twin pregnancies and compared this with 3466 singleton pregnancies. On average free b-hCGvalues are 2.099 times greater in twins than in singletons and PAPP-A some 1.86 times greater. The width ofthe analyte distribution in twins is very similar to that in singleton pregnancies. Using statistical modellingtechniques I have predicted that at a 5% false positive rate the detection rate in twins discordant for trisomy21 will be 52% and in twins concordant for trisomy 21 will be 55%, if correction for twin pregnancy iscarried out using the `pseduo risk' approach. The detection rate using biochemical parameters is less thanthat achievable for twins using NT (75%). However, the combination of NT and maternal serumbiochemistry will give detection rates approaching 80%. These rates are some 10% less than in singletonpregnancies, but nevertheless combining NT and biochemistry will allow high rates of detection of affectedtwins with the bene®t of ultrasound and NT being able to speci®cally locate the affected twin. Twinscreening using both modalities should be considered when introducing ®rst trimester screening. Copyright# 2000 John Wiley & Sons, Ltd.

KEY WORDS: Biochemical screening; trisomy 21; twins; PAPP-A; free b-hCG; ®rst trimester; nuchaltranslucency

INTRODUCTION

In the ®rst trimester of pregnancy it has been shown(Spencer et al., 1999) that combining the rapidmeasurement of the biochemical markers free b-hCGand pregnancy associated plasma protein A (PAPP-A)with the measurement of fetal nuchal translucencythickness (NT) at 10 to 14 weeks, will allow theidenti®cation of 89±90% of cases of trisomy 21 for a5% false positive rate in singleton pregnancies. The useof rapid analytical techniques for the measurement ofbiochemical markers has allowed the introduction of aone stop clinic for assessment of risk for fetalanomalies (OSCAR) in which, during a one hourvisit, the patient undergoes pre-test counselling,biochemical assessment, ultrasound assessment andcombined risk estimation prior to receiving post-testcounselling (Spencer, 1999a).

Twin pregnancies occur with a frequency of 1.327per 100 births (Of®ce for National Statistics, 1994±96),and, with the increased use of assisted reproduction,the last decade has seen an increase in maternities withmultiple births from 10.8 per 1000 to 14.1 per 1000. Inaddition, the rates of twinning are maternal agerelated, such that women over 35 are three times

more likely to conceive twins than are women underthe age of 20.

In the second trimester of pregnancy, biochemicalscreening using either the dual marker (alpha-fetoprotein [AFP] and free b-hCG) or triple marker(AFP, total hCG, unconjugated oestriol) approaches,using multi-marker risk assessment can identify60±70% of affected pregnancies for a 5% false positiverate (Macri and Spencer, 1996; Spencer, 1999b). Waldet al. (1991) proposed a method of modifying the riskalgorithm to take into account twin pregnancies byproducing a `pseudo-risk', which was designed to yielda similar false positive rate in twins and singletonpregnancies using the triple marker approach. Spenceret al. (1994) further showed that such `correction' fortwin pregnancies was equally applicable to the dualmarker approach and showed that detection rates intwins discordant for trisomy 21 were lower than insingleton pregnancies (51% at a 5% false positive rate).This was later con®rmed for the triple markerapproach by simulations carried out by Neveux et al.(1996). In prospective practice, the success of thisapproach in identifying twins concordant for trisomy21 has been shown (Verdin et al., 1997).

Despite the ability to produce such `pseudo-risks' intwin pregnancies, screening in the second trimesterunder such circumstances is still considered by some tobe problematical because of the signi®cant clinical,technical and ethical challenges posed for the diagnosisand clinical management of such pregnancies (Spencer

*Correspondence to: K. Spencer, Endocrine Unit, Clinical Biochem-istry Department, Harold Wood Hospital, Gubbins Lane, Romford,Essex RM3 0BE, UK. E-mail: Kevin-Spencer@compuserve.com

PRENATAL DIAGNOSIS

Prenat Diagn 2000; 20: 91±95.

CCC 0197-3851/2000/020091±05$17.50Copyright # 2000 John Wiley & Sons, Ltd.

Received: 18 June 1999Revised: 11 October 1999

Accepted: 19 October 1999

et al., 1994; Reynolds, 1995). Indeed, some have alsochallenged the scienti®c validity of the `pseudo-risk'approach (O'Brien et al., 1997).

In the ®rst trimester the use of individual fetalnuchal translucency thickness has allowed the calcula-tion of speci®c risks for each fetus and this physicalmarker can speci®cally identify the fetus at increasedrisk (Sebire et al., 1996b). In addition it has beensuggested that the risk based on nuchal translucencyand maternal age can be used as the basis for makingdecisions regarding the appropriate diagnostic proce-dure to be followed in such circumstances (Sebire et al.,1996a).

Whilst maternal serum biochemistry alone cannotspeci®cally identify the fetus at risk in the presence oftwins discordant for an anomaly, it may be possiblefor the combination of nuchal translucency thicknessand maternal serum biochemistry to speci®callyidentify those pregnancies at increased risk and forthe speci®c affected twin to be identi®ed by theincreased nuchal translucency. In this study I haveset out to identify the biochemical patterns observed intwin pregnancies in the ®rst trimester and to predictthe likely detection rates when combined with nuchaltranslucency in situations when both twins are affectedand in situations when only one twin is affected.

MATERIALS AND METHODS

Maternal serum PAPP-A and free b-hCG weremeasured in all pregnant women attending our ®rsttrimester OSCAR clinic using the CIS Kryptor rapidrandom access immunoassay analyser and the time-resolved ampli®ed cryptate emission technology (CIS(UK) Ltd, High Wycombe, Bucks, UK). The perfor-mance of these assays has been previously described(Spencer et al., 1999). All samples were analysed freshand results obtained within 30 min of blood collection.During the analysis time all women had theirpregnancy dated by crown±rump length (CRL) andnuchal translucency thickness was measured. Allscreening was performed between 10 weeks and 3days and 13 weeks and 6 days (CRL 38.0±84.0 mm).During the period June 1998 to April 1999, 3523women were screened of which 3466 had singletonpregnancies and 57 had twin pregnancies. Thisrepresented an incidence of twins of 1.62 per 100maternities.

All analytes levels were converted to multiple of thenormal median (MoM) for the gestational age of thepregnancy and corrected for maternal weight asdescribed from data in a previous study (Spenceret al., 1999).

To supplement the estimation of analyte values intwin pregnancies, a further set of samples from 102twin pregnancies collected during the previous twoyear period (1996±98), as part of a research study,were also analysed in the same manner. These sampleshad been kept frozen at x20uC since collection.Table 1 outlines the general characteristics of thestudy populations.

Statistical analysis

Statistical analysis of data was performed using Exceland Analyse-It, a statistical software add-in for Excel 7(Smart Software, Leeds, UK). In order to arrive atan estimate of the trisomy 21 detection rate andfalse positive rate that might be achieved in masspopulation screening of twin pregnancies using bio-chemical, ultrasound or combined modalities, I usedstandard statistical modelling techniques (Royston andThompson, 1992).

To simulate a population of normal twin preg-nancies, two series of 15 000 random NT MoMvalues were selected at random from within thegaussian distribution of log10 MoM NT using thenormal parameter set of Nicolaides et al. (1998). Theresults for series 1 represented the MoM NT for twin1 and the result for series 2 the MoM NT for twin 2.To simulate a population of twin pregnanciesaffected by trisomy 21, a third and fourth series of15 000 random NT MoM values were selected fromwithin the gaussian distribution of the log10 MoMNT using the trisomy 21 parameter set of Nicolaideset al. (1998). For twins discordant for trisomy 21,the results for series 1 represents the MoM NT fortwin 1 (normal) and the results for series 3 the MoMNT for twin 2 (trisomy 21). For twins concordantfor trisomy 21, the results for series 3 represents theMoM NT for twin 1 (trisomy 21) and the results forseries 4 the MoM NT for twin 2 (trisomy 21).Similarly for the biochemical markers PAPP-A andfree b-hCG, to simulate a population of normal twinpregnancies, two series of 15 000 random biochem-ical marker MoM values were selected at randomfrom within the gaussian distribution of log10 MoMof the biochemical marker using the normal para-meter set of Spencer et al. (1999). To simulate apopulation of twin pregnancies affected by trisomy21, a third and fourth series of 15 000 randombiochemical marker MoM values were selected fromwithin the gaussian distribution of the log10 MoM ofthe biochemical marker using the trisomy 21 para-meter set of Spencer et al. (1999). These series werethen combined together as for the NT data sets. TheNT and biochemical marker values in the normaltwin data set and the trisomy 21 affected data setwere then used to calculate likelihood ratios for thevarious marker combinations after correction of thebiochemical marker MoMs for the presence of twinpregnancy using the method described. The likeli-hood ratios were then used together with the agerelated risk for trisomy 21 derived previously fromCuckle et al. (1987) as described in Spencer et al.(1994), in order to calculate the expected detectionrate of affected pregnancies at a ®xed false positiverate. The maternal age distribution of twin pregnan-cies in England and Wales (Of®ce for NationalStatistics, 1994±96) was derived for each single yearfrom linear regression of the half decade percentageof women with twin pregnancies between the ages 15and 39 years (slope=0.0694; intercept=x0.6056;

K. SPENCER92

Copyright # 2000 John Wiley & Sons, Ltd. Prenat Diagn 2000; 20: 91±95.

r=0.9996) and using the point estimates of 1.55%and 2.74% for women 40±44 years, and 45 years andover.

RESULTS

Table 2 shows the marker values and their distribu-tions in both singleton and twin pregnancies. Themedian MoM free b-hCG in twins was signi®cantlyhigher than that in singleton pregnancies (MannWhitney, p<0.001) as was that for PAPP-A. Thewidth (SD) of the distribution of log10 MoM for bothanalytes was similar in twin and singleton pregnancies.This was further emphasized by analysis of the MoMratios at various centiles of the distribution. Table 3shows that the ratio of the MoMs in twins to that insingletons was fairly constant across a wide range.

Table 4 shows the modelled expected detection ratein twin pregnancies when the biochemical analyteMoM values are adjusted for the presence of twins bydividing by the factor 2.099 for free b-hCG and 1.86for PAPP-A.

DISCUSSION

In the second trimester, biochemical markers onaverage appear to be twice as high in the presence ofa twin pregnancy compared to that of a singletonpregnancy of the same gestational age (Spencer et al.,1994; Neveux et al., 1996; Wald et al., 1991), withperhaps levels of unconjugated oestriol being lower at1.7 times that in singletons. Using the `pseudo-risk'correction procedure outlined by Wald et al. (1991),the provision of risks in twin pregnancies can lead todetection rates in twins which are some 15% lowerthan in singleton pregnancies (Spencer et al., 1994;

Neveux et al., 1996). Although such approaches can beshown to identify affected cases in routine screeningpractice (Verdin et al., 1997), many centres stillconsider the ethical and technical dif®culties toogreat to handle in routine obstetric units. As a result,many do not offer screening in twin pregnancies.

In the ®rst trimester, few data are available on thebehaviour of biochemical markers in twin pregnancies.What data there are suggests that free b-hCG may beelevated to about twice normal. Berry et al. (1995)showed median levels of 1.97 MoM in 50 sets of twinsand Noble et al. (1997) similarly found median levelsof 1.94 MoM in 136 sets of twins. Brambati et al.(1997) also showed a median of 1.94 in 35 dichorionicand 4 monochorionic twin pregnancies. For PAPP-Athe median was 1.50. The results from the presentlarger study are somewhat higher than found byBrambati for PAPP-A and similar to all other seriesfor free b-hCG.

Other than the 12 affected twin pregnancies (10discordant and 2 concordant for trisomy 21) publishedby Noble et al. (1997) and the three cases discordantfor trisomy 21 published by Brambati et al. (1997),there are no data on the marker distributions in casesof twins in which one or both twins are affected. Thedistribution of the data in the study by Noble et al.(1997) clearly shows that cases with an affected twindo have free b-hCG values which are elevatedcompared with the normal distribution in unaffectedtwins (3.104 MoM affected versus 1.94 MoM unaf-fected). The conclusion from their study was that sinceonly one case was above the 95th centile free b-hCGwould be unlikely to be useful in the prediction of fetaltrisomy 21 in twins at 10±14 weeks' gestation. Theevidence for this statement is perhaps rather weak.Consider that in singleton pregnancies cases of trisomy21 free b-hCG levels are elevated to 2.15 MoM(Spencer et al., 1999), and on the basis of a 50 : 50contribution from an affected twin and an unaffectedtwin, the expected observed median in affected twinsmight be expected to be (2.15+1.00)/2 (i.e. 1.575MoM), which is very close to the 1.60 observed in thestudy by Noble et al. (1997). In the second trimestersimilar levels were observed (1.54 MoM corrected fortwins) in eight cases of twins discordant for trisomy 21(Spencer et al., 1994) and in this study modellingshowed that a detection rate of 51% for a 5% falsepositive rate would be achievable.

In this present study I have simulated the impact ofscreening twin pregnancies discordant and concordant

Table 1ÐGeneral characteristics of the study population (range)

Singelton Twins Retrospective twins

Number 3466 57 102Median maternal age (years) 29 (15±45.4) 31.3 (19.8±40.2) 31.9 (20.6±41.4)Maternal weight (kg) 65.2 (40±120) 66.5 (45±115) 67.6 (47.2±118)Gestational age (days) 84 (73±97) 84 (74±96) 84 (73±95)Crown±rump length (mm) 57 (38±84) 57 (39±82) 57 (38±83)NT (mm) 1.3 (0.50±4.8) 1.5 (0.5±5.4) 1.4 (0.6±4.1)Storage (days) 0 0 310

Table 2ÐPopulation parameters for free b-hCG andPAPP-A in twins and singleton pregnancies

Singleton Twins

Median PAPP-A MoM 1.0065 1.8723Mean Log10 PAPP-A MoM x0.0050 0.2611Mean Log10 SD PAPP-A MoM 0.2282 0.2116Median free b-hCG MoM 1.0037 2.1073Mean Log10 free b-hCG MoM 0.0067 0.3212Mean Log10 SD free b-hCG MoM 0.2675 0.2745

FIRST TRIMESTER SCREENING FOR TRISOMY 21 IN TWIN PREGNANCIES 93

Copyright # 2000 John Wiley & Sons, Ltd. Prenat Diagn 2000; 20: 91±95.

for trisomy 21 using both biochemical parameters,nuchal translucency and the two combined together.The underlying assumption in cases of twins is thateach fetus contributes 50% of the analyte level foundin the maternal serum. This assumption may not true,as suggested by O'Brien et al. (1997), but the evidencepresented is weak. Other data however would suggestthat maternal serum analyte levels are proportional tothe number of fetoplacental units (Johnson et al.,1994). Using simulated data from the ®rst trimestergaussian distributions of analytes in both the affectedand unaffected populations it has been possible toshow that in a series of twins discordant for trisomy 21in which the median free b-hCG MoM was 1.608(corrected for twins) and the PAPP-A MoM was 0.860(corrected for twins), detection rates should be similarto that predicted using biochemical markers (and twincorrection) in the second trimester. In the case of twinsconcordant for trisomy 21 detection rates may be afew per cent higher. The ®rst trimester twin detectionrate by biochemical markers of 52±55% is less thanthat estimated by NT alone in twins (75.2%), althoughthe NT estimation did not take into account the issueof chorionicity and the known fact that the incidenceof increased NT is 1.5 times greater in monochorionicthan in dichorionic twins (Sebire et al., 1996b).

The use of NT has additional advantages in that thespeci®c affected twin can be identi®ed and thatappropriate diagnostic procedures (CVS or amniocent-esis in both twins) can be instigated based on theseverity of the NT (Sebire et al., 1996a,b). However,the addition of maternal serum biochemistry to NTcould be expected to increase the detection rate by afurther 5±6% which may be a worthwhile addition.

I conclude that the use of maternal serum biochem-ical markers in addition to NT would bring aboutsome small bene®t when screening twins withoutlosing any of the bene®ts of ultrasound screening of

twin pregnancies. Further data are required in normaland affected twin pregnancies in order to make thecurrent algorithm more robust.

REFERENCES

Berry E, Aitken DA, Crossley JA, Macri JN, Connor JM. 1995.Analysis of maternal serum alpha-fetoprotein and free betahuman chorionic gonadotrophin in the ®rst trimester: implica-tions for Down's Syndrome screening. Prenat Diagn 15: 555±565.

Brambati B, Macri JN, Tului L, Hallahan TW, Krantz DA, AlbertiE. 1997. First trimester fetal aneuploidy screening: maternalserum PAPP-A and free beta hCG. In Screening for DownSyndrome in the First Trimester, Grudzinskas JG, Ward RHT(eds). RCOG Press: London; 135±147.

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.

Johnson MR, Abbas A, Nicoladies KH. 1994. Maternal plasmalevels of human chorionic gonadotrophin, oestradiol and proges-terone in multifetal pregnancies before and after reduction.J Endocrinol 143: 309±312.

Macri JN, Spencer K. 1996. Towards the optimal protocol forDown's syndrome screening. Am J Obstet Gynecol 174:1668±1669.

Neveux LM, Palomaki GE, Knight GJ, Haddow JE. 1996. Multiplemarker screening for Down Syndrome in twin pregnancies. PrenatDiagn 16: 29±34.

Nicolaides KH, Snijders RJM, Cuckle HS. 1998. Correct estimationof parameters for ultrasound nuchal translucency screening.Prenat Diagn 18: 519±521.

Noble PL, Snijders RJM, Abraha HD, Sherwood RA, NicolaidesKH. 1997. Maternal serum free beta hCG at 10 to 14 weeks ofgestation in trisomic twin pregnancies. Br J Obstet Gynaecol 104:741±743.

O'Brien JE, Dvorin E, Yaron Y, Ayoub M, Johnson MP, Hume RF,Evans MI. 1997. Differential increases in AFP, hCG and uE3 intwin pregnancies: impact on attempts to quantify Down syndromescreening calculations. Am J Med Genet 73: 109±112.

Of®ce for National Statistics. 1994±96. Birth Statistics. Series FM1,no. 23±25. HMSO: London.

Reynolds TM. 1995. Down's syndrome screening in twin pregnan-cies. Prenat Diagn 15: 386±387.

Table 3ÐDistribution of PAPP-A and free b-hCG twin: singleton MoM ratio's

Singleton Twins

PAPP-A Free b-hCG PAPP-A Free b-hCG Twin : Singleton Twin : SingletonCentile MoM MoM MoM MoM PAPP-A ratio Free b-hCG ratio

10th 0.5017 0.4734 0.9712 0.9329 1.936 1.97020th 0.6454 0.6087 1.1998 1.2230 1.859 2.00950th 1.0066 1.0038 1.8723 2.1073 1.860 2.09980th 1.5264 1.6958 2.8538 3.5590 1.870 2.09990th 1.8968 2.1527 3.6295 4.4077 1.913 2.047

Table 4ÐModelled detection rates in twin pregnancies for various marker combinations after correction of the biochemicalparameters for the presence of a twin pregnancy. Detection rates are quoted at a ®xed 5% false positive rate

Discordant for trisomy 21 Concordant for trisomy 21Marker combination Detection rate (%) Detection rate (%)

NT and maternal age 75.2 75.2Free b-hCG, PAPP-A and maternal age 51.5 55.4NT, free b-hCG, PAPP-A and maternal age 79.7 81.3

K. SPENCER94

Copyright # 2000 John Wiley & Sons, Ltd. Prenat Diagn 2000; 20: 91±95.

Royston P, Thompson SG. 1992. Model based screening forrisk with application to Down's syndrome. Stats Med 11:257±268.

Sebire NJ, Noble PL, Psarra A, Papapanagiotou G, Nicoladies KH.1996a. Fetal karyotyping in twin pregnancies: selection oftechnique by measurement of fetal nuchal translucency. Br JObstet Gynaecol 103: 887±890.

Sebire NJ, Snijders RJM, Hughes K, Sepulveda W, Nicolaides KH.1996b. Screening for trisomy 21 in twin pregnancies by maternalage and fetal nuchal translucency thickness at 10±14 weeks ofgestation. Br J Obstet Gynaecol 103: 999±1003.

Spencer K. 1999a. One stop clinic for assessment of risk for fetalabnormalities. Downs Screening News 6: 10.

Spencer K. 1999b. A review of 7 years of second trimester prenatalscreening for Down's syndrome using the biochemical markersAFP and free Beta hCG. Br J Obstet Gynaecol (in press).

Spencer K, Salonen R, Muller F. 1994. Down's syndrome screeningin multiple pregnancies using alpha fetoprotein and free betahCG. Prenat Diagn 14: 537±542.

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 b human chorionicgonadotropin and pregnancy-associated plasma protein-A. Ultra-sound Obstet Gynecol 13: 231±237.

Verdin SM, Braithwaite JM, Spencer K, Economides DL. 1997.Prenatal diagnosis of trisomy 21 in monozygotic twins withincreased nuchal translucency and abnormal serum biochemistry.Fetal Diagn Ther 12: 153±155.

Wald N, Cuckle H, Hu T, George L. 1991. Maternal serumunconjugated oestriol and human chorionic gonadotropin levelsin twin pregnancies: implications for screening for Down'ssyndrome. Br J Obstet Gynaecol 98: 9008±9050.

FIRST TRIMESTER SCREENING FOR TRISOMY 21 IN TWIN PREGNANCIES 95

Copyright # 2000 John Wiley & Sons, Ltd. Prenat Diagn 2000; 20: 91±95.