Aim: In the algorithm developed by the Fetal Medicine Foundation (FMF) Germany designed to evaluate the findings of routine first-trimester screening, the false-positive rate (FPR) was determined for the entire study group without stratification by maternal weight. Based on the data received from the continuous audit we were able to identify an increase in the FPR for the weight-related subgroups of patients, particularly for patients with extremely high body weights. The aim of this study was to demonstrate that the variability of the FPR can be reduced through adjusting the concentrations of free β-HCG and PAPP-A measured in the maternal serum by means of a nonlinear regression function modeling the dependence of these values on maternal weight.
Material and methods: The database used to establish a version of the algorithm enabling control of the FPR over the whole range of maternal weight consisted of n = 123 546 pregnancies resulting in the birth of a child without chromosomal anomalies. The group with positive outcomes covered n = 500 cases of trisomy 21 and n = 159 trisomies 13 or 18. The dependency of the serum parameters free β-HCG and PAPP-A on maternal weight was analyzed in the sample of negative outcomes by means of nonlinear regression. The fitted regression curve was of exponential form with negative slope. Using this model, all individual measurements were corrected through multiplication with a factor obtained as the ratio of the concentration level predicted by the model to belong to the average maternal body weight of 68.2 kg, over the ordinate of that point on the regression curve which belongs to the weight actually measured. Subsequently, the totality of all values of free β-HCG and PAPP-A corrected for deviation from average weight were used as input data for carrying out the construction of diagnostic discrimination rules described in our recent paper for a database to which no corrections for over- or under-weight had been applied. This entailed in particular the construction of new reference bands for the corrected biochemical values as the basis for calculating the degree of extremeness (DOE) measures to replace the more traditional MOMs. In the final and most crucial step, stratified FPRs were computed and compared over a set of intervals partitioning the whole range of maternal weight into 18 classes.
Results: For the posterior risks of both trisomy 21 and 13 / 18 computed from the weight-corrected database, the use of a cutoff value of 1:150 turned out to be an appropriate choice. For T 21, the overall FPR obtained through comparing the individual risks with this cutoff was found to be 3.51 %. The corresponding proportion of ascertained cases of trisomy 21 detected by means of the new algorithm was 86.2 %. For the trisomy 13 / 18 group, the analogous results were a FPR of 2.07 % and a detection rate (DTR) of 83.0 %, respectively. A comparison between the FPRs obtained for the 18 intervals into which the range of maternal weight had been partitioned, showed the deviation of the strata-specific from the overall FPR to be fairly small: for T 21, the FPR ranged from 2.72 to 4.86 %, and the maximum was found in the group of 87.5 - 95.0 kg. For women with a weight of more than 120 kg, the FPR was only slightly above the FPR for the total sample (3.69 as compared to 3.51 %). Similar results were obtained for the discrimination rule constructed for diagnosing T 13 / 18: here, the minimum FPR (1.17 %) was found for patients weighing more than 120 kg, whereas the maximum (2.66 %) occurred in the interval 75.0 - 77.5 kg.
Conclusion: In this study we demonstrated that the new algorithm developed by the FMF Germany to estimate risks for fetal trisomies 21 and 13 / 18 combines very good misclassification rates with a far-reaching stability of the false-positive rate against even extreme deviations from the average maternal weight.
© Georg Thieme Verlag KG Stuttgart · New York.