CALCULATING CORRECT DOWN'S SYNDROME RISKHoward Cuckle Professor (Reproductive Epidemiology). (h.s.cuckle@leeds.ac.uk) Indera Seluni Research Assistant Centre for Reproduction, Crowrh and Development. University of Leeds Br. J. Obstet Gynecol 1999;106:371-372
Women having both first trimester nuchal translucency and second trimester serum screening tests are likely to receive two different Down's syndrome risks. Neither will be correct, and we describe how to calculate a valid combined risk. This uses the reported serum-based risk and a likelihood ratio derived from the nuchal translucency report. Tables, figures and examples are provided to aid the calculation of the likelihood ratio from either the nuchal translucency in multiples of the normal median, the nuchal translucency and crown-rump length in millimetres, or the reported prior and nuchal translucency-based risks. IntroductionFollowing the publication of impressive results from the Fetal Medicine Foundation Study, an increasing number of women are being screened for Down's syndrome using first trimester nuchal translucency (NT) measurement. The predicted detection rate for a 5% false positive rate is 73%, after excluding nonviable pregnancies2 3. This is higher than can be achieved in the second trimester using two to three maternal serum markers : overall detection was 57% in 17 large prospective studies. However, some women are having both screening tests and this can lead to error. It is standard practice to interpret a Down's syndrome screening test by calculating the risk of an affected birth from the maternal age, family history and marker profile. Consequently, women having separate NT and serum screens are usually given two risks, and if they differ this can lead to confusion and anxiety. In these circumstances women want to know which risk they should trust, but neither is correct and a combined risk is needed. It is not valid to simply take the average of the two risks. We describe and illustrate the correct method here. Methods and ResultsThe combined risk is estimated by taking the serum based risk expressed as an odds and multiplying it by an NT-based likelihood ratio. For example, suppose there is a serum risk of 1 in 500 and a likelihood ratio of 10. Expressed as an odds the serum risk is 1:499 which when multiplied by 10 becomes 10:499 or 1 in 51. The NT-based likelihood ratio can be derived from the ultrasound scan report and applied to any subsequent serum based risk. The best method for working out the likelihood ratio will depend on what information is available from the report. If the NT was reponed in multiples of the normal median (MoM) for gestation or crown-rump length (CRL) a precise estimate can be made using Table 1.
For example, an NT of 1-8 MoM (1-5 + 0-3 MoM) yields a likelihood ratio of 4-8. If the NT was only reponed in mm, it can becon verted into a MoM after dividing by the expected NT for the CRL using Table 2.
For example, if the CRL is 66 mm (65 + 1 mm) the expected NT is 1-68 mm; if the actual NT is 2 0 mm then the result is 1 2 MoM, giving a likelihood ratio of 0-40 from Table 1. An alternative approach that does not require any calculations is to use the normogram in Fig.1 as a rough guide :
In this example, the CRL and NT give a likelihood ratio of between l/3 and 1/2. These calculations assume that the NT measurement was made using the same protocol as the Fetal Medicine Foundation. Sometimes the actual NT is not on the report, but the risk prior to the scan (based on age and family history) and the NT-based risk are reported. An approximate estimate of the likelihood ratio can be derived by comparing the two. For example, suppose the prior risk is 1 in 500 and the NT-based risk is 1 in 1O. Expressed as odds these risks are respectively 1:499 and 1:9 or 55:499 so the NT result has increased the odds 55-fold implying a likelihood ratio of 55. However, likelihood ratios derived in this way are subject to inaccuracy as the reported risks may have been rounded. In our example, if the rounding had been to one significant figure, the unrounded prior risk will have been between 1 in 451 and 1 in 549 and the NT-based risk between 1 in 9-5 and 1 in 10-5. Thus the likelihood ratio could range from 47 to 64. DiscussionThe calculations we have described assume that first trimester NT measurements and second trimester maternal serum markers are independent determinants of Down's syndrome risk. In two large studies no material correlation was found between NT and first trimester serum markers. Therefore any correlation with second trimester markers is extremely unlikely. The best screening results would be achieved by offering all women a test which combines both NT and serum markers in the first trimester. The simplest combined protocol is to obtain a blood sample at 9-11 weeks and schedule an ultrasound for 12-13 weeks. In Leeds we have shown this to be feasible: the NT measurement is telephoned to the laboratory and a combined Down's syndrome risk is immediately faxed to the sonographer. Using serum parameters based on a meta-analys is of 44 studies, the predicted detection rate for NT and serum combined in this way is 88% for a 5% false positive rate, a 15% increase in detection over NT alone. If adopted elsewhere this method would obviate the need for second trimester Down's syndrome screening except for those booking late, and avoid errors of risk calculation. ReferencesSnijders RJM. Noble P, Sebire N, Sou~a A, Nicolaides KH. UK Nicolaides KH, Snijders RJM, Cuclcle HS. Comct estimation of Cuckle H. Antenatal screening for Down's syndrome. Lancet 1998; Cuckle H. Established markers in second trimester maternal serum. Spencer K, Noble P, Snijders RJM, Nicolaides KH. First trimester Orlandi F, Damiani G, Hallahan TW, Krantz DA, Macri JN. First Cuckle HS, van Lith JM. Appropriate biochemical parameters in
Mise en page par le Dr JM BRIDERON le 12/10/2000 (jmbr@wanadoo.fr) |
