First trimester screening for Down syndrome using free β hcg, total hcg and PAPP-A: an exploratory study

Transcription

1 PRENATAL DIAGNOSIS Prenat Diagn 2007; 27: Published online 25 October 2007 in Wiley InterScience ( First trimester screening for Down syndrome using free β hcg, total hcg and PAPP-A: an exploratory study Dave Wright 1 *, Kevin Spencer 2 and Barry Nix 3 1 University of Plymouth, Plymouth, UK 2 Clinical Biochemistry Department, King George Hospital, Goodmayes, UK 3 University of Cardiff, UK Objective To investigate the potential utility of first trimester screening for Down syndrome using Free β-hcg, total hcg and PAPP-A. Materials and Methods Using estimates from the literature, a simulation study was undertaken to estimate the performance of tests incorporating, Free β-hcg, total hcg and PAPP-A at gestations of 8 12 weeks. We used sensitivity analysis to assess the effect of departures from the assumed model. Results We estimate that detection rates in excess of 75% for a false positive rate (FPR) of 3% can be achieved with first trimester measures of PAPP-A, total hcg and Free β-hcg at 8 weeks the addition of total hcg adding 11%. Detection rates of around % for a FPR of 3% can be achieved through the inclusion of nuchal translucency (NT) at 12 weeks to these early first trimester biochemical markers. Our analysis indicates that the marginal benefit of adding total hcg diminishes rapidly with gestational age and that there is little benefit from adding total hcg later than 10 weeks of gestation. Conclusion The performance of first trimester screening using early combinations of total hcg, Free β-hcg and PAPP-A should be assessed in further studies. Copyright 2007 John Wiley & Sons, Ltd. KEY WORDS: Down syndrome; screening; first trimester INTRODUCTION Currently, the most widely used first trimester screening test, the combined test, uses Free β-hcg, PAPP-A and nuchal translucency (NT). Modelling results using SURUSS (Wald et al., 2003, 2004) parameters gives a detection rate of 78% for a 3% false positive rate (FPR) at 10 weeks of gestation. Similar results were obtained from the FASTER study (Malone et al., 2005) which gives a FPR of 3.8% for a 75% detection rate at 11 weeks of gestation. Performance of the combined test has also been demonstrated prospectively, for example in a large scale prospective study in two OSCAR centres over a 6-year period Nicolaides et al. (2005) have shown a detection rate of 92.6% for a 5.2% FPR when screening a population of women in which 325 foetuses had Down syndrome. Similar levels of performance have been reported by other centres with smaller studies. Better performance might be achieved using integrated screening tests combining information from the first and second trimester. In particular, modelling the integrated test from the SURUSS (Wald et al., 2003, 2004) and FASTER (Malone et al., 2005) studies gives FPR below 1% for a detection rate of 85% when the first trimester markers are measured at 10 or 11 completed weeks of gestation. Wright and Bradbury (2005), show, assuming SURUSS parameter estimates, that the inclusion of repeated measures of highly correlated markers across *Correspondence to: Dave Wright, University of Plymouth, Plymouth, UK. dave.wright@plymouth.ac.uk the first and second trimesters, has great potential to improve the performance of integrated tests and there is growing evidence to support the use of repeated measures across trimesters (Palomaki et al., 2006). Previous thinking regarding the value of total hcg in the first trimester has been to dismiss it as a valuable marker since during the period weeks the marker has poor clinical discrimination (Spencer et al., 2000; 2002; Evans et al., 2007). However prior to 10 weeks one study has shown that levels of total hcg in pregnancies with Down syndrome are reduced, rather than increased as is the case in the second trimester (Spencer et al., 2002). In this article we demonstrate the possibility of highly efficient screening tests incorporating total hcg as well as Free β-hcg and PAPP-A early in the first trimester. The results we present should be seen as exploratory but they do show great promise and the need for further studies involving total hcg in combination with Free β-hcg and PAPP-A. MATERIALS AND METHODS Risk assessment requires joint distributions of markers in unaffectedand affectedpregnancies. With an assumption that log multiples of the median (MoM) values follow a Gaussian distribution, this requires specification of mean vectors and covariance matrices for unaffected and Down syndrome distributions. For three markers there are three means for the affected population, six standard deviations (three affected and three unaffected) and six Copyright 2007 John Wiley & Sons, Ltd. Received: 18 March 2007 Revised: 10 July 2007 Accepted: 3 August 2007 Published online: 25 October 2007

2 FIRST TRIMESTER SCREENING FOR DOWN SYNDROME 1119 correlations (three affected and three unaffected) for the different pairings of markers. Moreover, the means in Down syndrome pregnancies depend on gestational age. Estimates of the means and standard deviations of log MoM values for PAPP-A, Free β-hcg and total hcg were obtained from Spencer et al. (2002, 2003), a study based on samples of over 700 Down syndrome pregnancies. The correlations involving total hcg and Free β-hcg were obtained from the study reported in Spencer et al. (2000). For NT, we assumed that log MoM NT values follow a Gaussian distribution in both unaffected and Down syndrome pregnancies with parameters taken from Spencer et al. (2003). We assume that, after standardisation, NT is uncorrelated with the biochemical markers in Down syndrome and in unaffected pregnancies (Spencer et al., 1999). The full set of parameter estimates assumed are given in the Appendix 1 to this article. Detection rates and FPR were obtained using Monte Carlo methods as described in Wright and Bradbury (2005) to sample from the joint distribution of log MoM values in the Down syndrome and in the unaffected populations. The risk cut-off above which pregnancies were screened positive was determined to fix the FPR. This enables different screening tests to be compared in terms of detection rates for a fixed FPR. Samples of observations were drawn from the relevant multivariate Gaussian distributions for Down syndrome and unaffected pregnancies. Likelihood ratios were computed for each simulated patient profile and then used to compute detection rates and FPR for each maternal age. Overall population detection rates for the maternal age distribution of England and Wales (Office for National Statistics, 2002) were obtained by combining maternal age specific rates according to the maternal age distribution for Down syndrome. The use of this reference population enables direct comparisons. Similarly, FPR were obtained using the maternal age distribution of unaffected pregnancies. The precision of the estimated detection rates and FPR was assessed both theoretically and empirically and it was concluded that simulations provide sufficiently precise estimates of detection and FPR. Of course, this accounts only for the imprecision due to the Monte Carlo approach and not for the variability of the estimated means, standard deviations and correlations used in the calculations. Documentation and copies of the S-plus functions used in the estimation of detection rates and FPR are available on request from the first author. RESULTS Table 1 gives a summary of the screening performance of various tests based on maternal serum samples between 8 and 12 weeks of gestation. The most striking feature of Table 1 is the size of the improvement in screening performance obtained from the addition of total hcg to the screening test comprising Free β-hcg and PAPP-A at gestational ages of 8 or 9 weeks. For a 3% FPR, the addition of total hcg increases estimated detection rate by 11% at 8 weeks. The benefit of adding total hcg diminishes rapidly with increasing gestational age and is negligible at 12 weeks. This trend is consistent with the temporal pattern of median total hcg values for affected pregnancies reported by Spencer et al. (2002). To examine the sensitivity of our findings to assumptions, we explored the effect of changes in the mean log MoM total hcg in Down syndrome pregnancies and changes in the correlation between total hcg and Free β-hcg. Figure 1 shows the modelled detection rates at 8 weeks of gestation for a 3% FPR. The vertical line shows the estimated median from Spencer et al. (2002). The horizontal lines show the detection rates obtained without total hcg. It is clear from Figure 1 that the benefit of adding total hcg depends on the median MoM in Down syndrome pregnancies falling below 1. If the median MoM for total hcg is close to 1, then there is very little benefit. Table 1 Comparison of the first trimester tests involving Free β-hcg, total hcg, PAPP-A with and without NT between 8 and 12 weeks. The detection rates and false positive rates relate to the maternal age distribution of England and Wales for the three years 2000 to It is assumed that MoM values are obtained from scan estimates of gestational age and are corrected for weight. NT is assumed to be measured at 12 weeks of gestation Without NT With NT at 12 weeks FPR (%) FPR (%) Weeks Markers PAPP-A + Free β-hcg PAPP-A + Free β-hcg + total hcg PAPP-A + Free β-hcg PAPP-A + Free β-hcg + total hcg PAPP-A + Free β-hcg PAPP-A + Free β-hcg + total hcg PAPP-A + Free β-hcg PAPP-A + Free β-hcg + total hcg PAPP-A + Free β-hcg PAPP-A + Free β-hcg + total hcg

3 1120 D. WRIGHT ET AL Median MoM hcg in Down s syndrome Correlation between total and free beta hcg Figure 1 Detection rates for a 3% false positive rate for screening using total hcg, Free β-hcg and PAPP-A at 8 weeks of gestation with (bold) and without NT. The horizontal line shows detection rates excluding total hcg. The vertical line shows the estimated Median MoM in Down syndrome pregnancies at 8 weeks from Spencer et al. (2002) Correlation between total and free beta hcg (T21) Figure 2 Detection rates for a 3% false positive rate for screening using total hcg, Free β-hcg and PAPP-A at 8 weeks of gestation with (bold) and without NT. Only the correlation in the Down syndrome population is changed The effects of changing the correlation in Down syndrome pregnancies only is shown in Figure 2. The benefit of adding total hcg increases with the correlation. This is generally what happens when two highly correlated markers discriminate in opposite directions. Figure 3 illustrates that when markers discriminate in opposite directions, high correlations in both the affected and unaffected increase discrimination still further. Figure 3 was constructed under the assumption that the the correlation coefficients between total hcg and Free β-hcg are the same for Down syndrome and unaffected pregnancies. As the correlation approaches 1, the discrimination becomes almost perfect. The mechanism for this is illustrated on Figure 4. With a common correlation of 0.7, close to the assumptions in Appendix 1, there is considerable overlap between the bivariate distributions of the log MoM values (top panel). When the correlation is 0.95 (bottom panel) there is very little overlap and hence much improved discrimination. Figure 3 Detection rates for a 3% false positive rate for screening using total hcg, Free β-hcg and PAPP-A at 8 weeks of gestation with (bold) and without NT. The results are obtained under the assumption that the correlation is the same in unaffected and Down syndrome pregnancies log(mom Total hcg) log(mom Total hcg) log(mom Free beta hcg) log(mom Free beta hcg) Figure 4 Bivariate scatter plots showing simulated data on 500 unaffected pregnancies (crosses) and 500 Down syndrome pregnancies (circles). The ellipses represent % contours. Assuming correlations of 0.7 (top panel) and 0.95 correlation (bottom panel) DISCUSSION The results presented here assume multivariate Gaussian models for the distribution of markers in affected and unaffected pregnancies and they fail to take account of

4 FIRST TRIMESTER SCREENING FOR DOWN SYNDROME 1121 uncertainty in the parameter estimates. With this in mind these results should be interpreted with caution. All we would claim at this stage is that the results point to a need for further research on screening tests combining total hcg with Free β-hcg and PAPP-A early in the first trimester. Our findings run counter to established views concerning highly correlated markers and the value of total hcg in the first trimester. The established view is that high correlation is a sign of redundancy. This is not generally the case and in particular, Free β-hcg and total hcg are highly correlated markers that may combine to produce a highly efficient screening test for use early in the first trimester. It is the combination of the high correlation and the difference in the direction of the effects of Down syndrome on the marker that produces the improvement in screening performance. Highly correlated markers which discriminate in opposite directions will indeed have improved discriminatory power. From Figure 1 it is clear that the benefits of adding total hcg depend critically on its median MoM in Down syndrome pregnancies. The evidence published in Spencer et al. (2002) included just 22 Down syndrome pregnancies at 8 weeks or earlier and only 40 at 10 weeks or earlier. This highlights the need for more data at the early gestations to confirm or refute the promising results we have presented. More generally, this work demonstrates that in the search for an improved screening test we need to consider markers jointly and recognise that highly correlated markers have great potential to improve screening performance. ACKNOWLEDGEMENT We are grateful to the referees for valuable suggestions that have improved this article substantially. REFERENCES Evans MI, Krantz DA, Hallahan TW, Galen RS Meta-analysis of first trimester Down syndrome screening studies: free β-human chorionic gonadotropin significantly outperforms intact human chorionic gonadotropin in a multimarkers protocol. Am J Obstet Gynecol 196: Malone FD, Canick JA, Ball RH, et al First trimester or second trimester screening, or both for Down s syndrome. N Engl J Med 353: Nicolaides KHN, Spencer K, Avgidou K, Faiola S, Falcon O Multicenter study of first trimester screening for trisomy 21 in 75,821 pregnancies: results and estimation of the potential impact of individual risk-orientated two-stage first-trimester screening. Ultrasound Obstet Gynecol 25: Office for National Statistics. Birth Statistics, Series FM1, Numbers HMSO, 2000, 2001, London. Palomaki GE, Wright DE, Summers AM, et al Repeated measurement of pregnancy-associated plasma Protein-A (PAPP-A) in Down syndrome screening: A Validation Study. Prenat Diagn 26: Spencer K, Souter V, Tul N, Snijders R, Nicolaides KH A screening program for trisomy 21 at weeks using NT, maternal serum free β-human chorionic gonadotropin and pregnancy associated plasma protein-a. Ultrasound Obstet Gynecol 13: Spencer K, Berry E, Crossley JA, Aitken DA, Nicolaides KH Is maternal serum total hcg a marker of trisomy 21 in the first trimester of pregnancy? Prenat Diagn 20: Spencer K, Crossley JA, Nix ABJ, Dunstan FDJ, Williams K Temporal changes in maternal serum biochemistry markers of trisomy 21 across the first and second trimester of pregnancy. Ann Clin Biochem 39: Spencer K, Crossley JA, Nix ABJ, Dunstan FDJ, Williams K The effect of temporal variation in biochemical markers of trisomy 21 across the first and second trimester of pregnancy on the estimation of individual patient specific risks and detection rates for Down s syndrome. Ann Clin Biochem 40: Spencer K, Bindra R, Nix ABJ, Heath V, Nicolaides KH. 2003c. Delta-NT or NT MoM: which is the most appropriate method for calculating accurate patient specific risks for trisomy 21 in the first trimester? Ultrasound Obstet Gynecol 22: Wright DE, Bradbury I Repeated measures screening for Down s Syndrome. BJOG 112: 83. APPENDIX: ASSUMED MODEL PARAMETERS Table A1 Means log MoM values in Down syndrome pregnancies Weeks Free β-hcg Total hcg PAPP-A The mean log MoM values in unaffected pregnancies are assumed to be 0 in all markers. Table A2 Standard deviations in Down syndrome and unaffected pregnancies, assumed not to be gestation specific Down syndrome Unaffected Free β-hcg Total hcg PAPP-A

5 1122 D. WRIGHT ET AL. Table A3 Correlations in Down syndrome and unaffected pregnancies, assumed not to be gestation specific Down syndrome Unaffected Free β-hcg and Total hcg Free β-hcg and PAPP-A Total hcg and PAPP-A