Performance characteristics of Elecsys free βhCG and PAPP-A for first trimester trisomy 21 risk assessment in gestational weeks 8+0 to 14+0

Multicenter studies

The data in this publication are obtained using prospective as well as cross sectional collected samples from two international multicenter studies. The sites that participated in these studies are from Belgium (Antwerp), Denmark (Aarhus), Germany (Essen, Ingelheim and Munich), Spain (Barcelona and Girona), Switzerland (Liebefeld), and UK (London). Frozen cross sectional collected repository serum samples as well as fresh left over routine serum samples from healthy pregnant women were used. If samples could not be measured immediately at the recruiting sites, samples were shipped frozen (≤–70 °C at any time) to the analyzing sites.

Technical performance

Technical performance such as functional sensitivity, linearity, endogenous interference and high-dose hook effects was assessed prior to this study by the manufacturer and are summarized in the package insert. Method comparison between Roche Elecsys (Roche Diagnostics Deutschland GmbH, Mannheim, Germany) and Siemens IMMULITE (Siemens Healthcare GmbH, Erlangen, Germany) was performed with 347 samples (free βhCG) and 334 samples (PAPP-A) collected in Munich and Girona. Method comparison between Roche Elecsys and PerkinElmer AutoDELFIA (PerkinElmer, Turku, Finland) was performed with 129 samples (free βhCG) and 124 (PAPP-A) samples, collected in Girona. Method comparison of Roche Elecsys and Brahms Kryptor (Brahms GmbH, Hennigsdorf, Germany) was performed with 3424 (free βhCG) samples and 3416 (PAPP-A) samples, derived from two independent international multicenter studies, the same studies that were also used to evaluate the clinical performance.

Clinical performance

To evaluate the potential clinical performance, day- and week-specific medians for gestational week 8+0 to 14+0 for the Roche Elecsys system were established by merging results of two independent international multicenter studies with samples collected between 2003 and 2010. Median values from unaffected pregnant women for each gestational day of the first trimester were generated. Each individual free βhCG and PAPP-A value is set in correlation to the overall median of unaffected women to generate an individual multiple of medium (MoM) value, which is subsequently corrected for weight, smoking and ethnicity. The median of Elecsys free βhCG values for each gestational age between 56 and 98 days was fitted by a five parametric polynomial function and the median of Elecsys PAPP-A values for each gestational age between 56 and 98 days by a three parametric polynomial function. Based on the age of the mother the individual MoMs of blood marker values and the NT was used to estimate an individual risk for T21.

The principle of multiple marker risk assessment in screening for T21 was established by Wald and Co-workers [5] and the statistical methodology has been explained by Reynolds and Penney [6]. Biochemistry was included from gestational week 8+0 to 13+6, and measurement of the NT from week 11+0 to 13+6. For patients presenting for biochemistry prior to gestational week 11+0, measurement of NT was performed at a subsequent appointment.

For the risk analysis with both Elecsys and Kryptor values, CE IVD approved, software SsdwLab Version 5 developed by SBP SOFT 2007 SL Spain was used. The reliability and precision of the algorithms and the internal mathematical calculations of SsdwLab5 have been proved with the prenatal screening software: Prenatal Screening Decision Support QA Tools version 2.0 from Media Innovation Ltd, partnership of University of Leeds and UK National Screening Committee. In the software, the method described by Palomaki and Haddow in 1987 is used to calculate the likelihood ratio [7], by means of the mathematical calculations for Gaussian multivariate distribution published by Reynolds and Penney in 1990 [6]. The software uses a MoM model for biochemical markers and nuchal translucency (NT). The equation that correlates the NT with the crown-rump length (CRL) is a polynomial function with exponential transformation (10ˆ) applied to the formula [8]: NT=10ˆ(–0.3599+0.0127 CRL–0.000058 CRL²).

The cases of trisomy 21 were verified by quantitative fluorescent PCR. If the test was positive, the samples were cultured to obtain a karyotype.

Ethical approval

Wherever applicable local IRB approvals were received or already in place.

Technical performance

Table 1 shows the within run as well as the total imprecision for the Elecsys free βhCG and Elecsys PAPP-A assay for three different control levels as well as three different human sample pool levels. Imprecision data were obtained using pooled sera and controls: six times daily for 10 days (n=60).

Figure 1 shows method comparison analysis for PAPP-A (Figure 1A) and free βhCG (Figure 1B) performed for the Roche Elecsys vs. Brahms Kryptor, Siemens IMMULITE and PerkinElmer AutoDELFIA with samples derived from two independent international multicenter studies. Comparison (Passing/Bablok regression) of free βhCG and PAPP-A concentrations measured on the Roche Elecsys and the Kryptor from Brahms resulted in slopes of 0.944 (n=3424) and 0.945 (=3416) and Pearson’s correlation of r=0.981 and r=0.987, respectively. Comparison of concentrations measured on the Roche Elecsys and Siemens IMMULITE resulted in slopes of 0.838 (n=347) and 1.39 (n=334) and Pearson’s correlation of r=0.983 and r=0.983, respectively. Comparison of concentrations measured on the Roche Elecsys and PerkinElmer AutoDELFIA resulted in slopes of 0.824 (n=129) and 1.20 (n=124) and Pearson’s correlation of r=0.995 and r=0.979, respectively.

Figure 1:

Method comparison analysis for (A) PAPP-A and (B) free βhCG performed for Roche Elecsys vs. Brahms Kryptor, Siemens IMMULITE and PerkinElmer AutoDELFIA.

X-axis values Roche Elecsys. Y-axis values Brahms Kryptor, Siemens IMMULITE and PerkinElmer AutoDELFIA. Displayed are the regression lines and corresponding equations for the individual comparisons.

Clinical performance

Study population

The median age of the pregnant women included in the total study population (n=5281) was 31.4 years (with a range of 16.3–55.5). The median weight was 65 kg (range: 35–152.5 kg). In total, 417 pregnant women (7.9%) were smokers, 3690 pregnant women (69.9%) were non-smokers and for 1174 pregnant women (22.2%) data on smoking is unknown or missing. The ethnical background was reported for 4051 pregnant women (unknown or missing: n=1230 (23.3%). The majority was Caucasian/White (n=3578, 67.8%), 50 pregnant women (0.95%) were Asian, 198 pregnant women (3.75%) African Black, two pregnant women (0.04%) Hispanic and 223 pregnant women (4.2%) Other. The total of 5281 pregnant women consists of 3847 pregnant women from an international multicenter study conducted between 27 November 2003 and 28 October 2008 and 1550 pregnant women from an international multicenter study conducted between 16 March 2006 and 29 June 2010. From the total number in these studies (n=5397) 116 women were excluded because they did not fulfill the study criteria (gestational age ≥56 and ≤98 days, mother is at legal age, at least one measurement available).

Figure 2 shows the number of pregnant women that were selected for: establishing the algorithm for Roche-specific risk estimation (free βhCG, n=4842; PAPP-A, n=4841) and the method comparison Kryptor vs. Elecsys (free βhCG, n=3424, PAPP-A, n=3416). For the establishment of the algorithm for Roche-specific risk estimation, 439 and 440 women were excluded from the median population parameter calculations for free βhCG and PAPP-A, respectively. The reason for exclusion was one of the following: missing ultrasound measurement; confirmed T21 outcome; gestational age not available or not within specified range (≥56 and ≤98 days; CRL >14 mm). Furthermore, just one single blood marker value needed to be available: free βhCG and/or PAPP-A for Elecsys. For the method comparison Kryptor vs. Elecsys, a further 1418 and 1425 women for comparison of the free βhCG and PAPP-A assay, respectively were excluded, as these women’s samples were not measured on both, Kryptor and Elecsys.

Figure 2:

Selection of pregnant women.

For the estimation of risk and clinical test performance 2652 women were excluded in agreement with GCP principles because of missing analyzer measurement raw data information (2651 unaffected pregnancies and one affected pregnancy). Excluded women were from all gestational weeks and did not result in a bias for the included population. From the included unaffected pregnancies (n=2522) and T21 pregnancies (n=107; in total 2629) the age of the mother as well as gestational age was available and within specified range (≥56 and ≤98 days; CRL >14 mm). Clinical outcome was confirmed for all 107 T21 pregnancies whereas clinical outcome could not be obtained for all unaffected pregnancies. Furthermore, only women were included of which all blood marker values were available: free βhCG and PAPP-A each for Elecsys and Kryptor. For risk calculation with NT further four women were excluded because of missing NT measurements resulting in a total of 2625 women (103 T21 pregnancies with NT). The gestational age characteristics of affected pregnancies included in the risk calculation are given in Table 2.

Table 2

Weight-adjusted and unadjusted Elecsys PAPP-A and free βhCG median MoMs (and inter-quartile range) for T21 pregnancies.

	Quantiles 25	Median	Quantiles 75
Week 8+0 to 9+6 (n=39)
MoM free βhCG	0.84	1.24	2.52
MoM PAPP-A	0.17	0.29	0.51
Adjusted MoM free βhCG	0.83	1.35	2.09
Adjusted MoM PAPP-A	0.21	0.30	0.50
Week 10+0 to 14+0 (n=68)
MoM free βhCG	1.02	1.69	2.58
MoM PAPP-A	0.23	0.33	0.55
Adjusted MoM free βhCG	1.07	1.54	2.39
Adjusted MoM PAPP-A	0.23	0.34	0.52

The Table shows weight-adjusted and unadjusted Elecsys PAPP-A and free βhCG median log10 MoMs (and inter-quartile range) for pregnancies affected by Down syndrome based on blood sampling during early weeks (8+0 to 9+6) and late weeks (10+0 to 14+0) of gestation. n indicates the number of subjects.

Medians from the unaffected pregnant population to be used for risk estimation

Supplemental Table 1 and Figure 3 show the Elecsys medians and regression-based medians from the unaffected pregnant population (8+0 to 14+0 gestational weeks) by using Elecsys free βhCG and PAPP-A assays, respectively.

Figure 3:

Relationship between concentration (free βhCG and PAPP-A) and gestational age in days.

Lines show the regressed medians of Elecsys free βhCG and Elecsys PAPP-A. Scattered dots represent the day-specific Elecsys medians. Blue shadows illustrate the 95% CI of the regression line. The gray bars illustrate the 95% CI of the predictions of the individual gestational day.

Figure 2 illustrates that the concentration of free βhCG steeply increases until day 63, after that it decreases moderately.

Algorithm for risk estimation of gestational age 56–98

Supplemental Tables 2, 3 and 4 shows the Roche-specific coefficients for Elecsys PAPP-A, free βhCG assays and analyzer independent coeficients for NT as well as the published population parameters for T21 pregnancies [9–12]. Weight-adjusted and unadjusted PAPP-A and free βhCG median log10 MoMs and (inter-quartile range) for pregnancies affected by Down syndrome are given in Table 2 based on blood sampling during early weeks (8+0 to 9+6) and late weeks (10+0 to 14+0) of gestation.

As can be seen in Table 2, this weight correction did not have a major influence on the PAPP-A Median MoMs, but it may have an essential impact on the individual risk estimate, in particular on the false-positive rate [13]. Figure 4 shows the scatter plots and regression lines of log10 MoM values of free βhCG and PAPP-A of the Roche Elecsys assays in the 107 cases of T21.

Figure 4:

Scatter plots and regression lines of log10 MoM values of free βhCG and PAPP-A (solid lines) in the 107 cases of T21.

The blue shaded areas represent the 95% CI for the regression lines. Unaffected pregnancies are represented by the zero line.

Technical performance

This large multicenter study describes the validation of the Roche Elecsys free βhCG and PAPP-A assays and the performance of employing the assays for first trimester screening for fetal T21 in gestational weeks 8–14. Imprecision and bias of assays employed for first trimester serum markers for prediction of prenatal aneuploidy directly affects the clinical performance of the screening [14], and it is therefore important to optimize the assays for routine laboratory use. The total imprecision (CV in %) of the two Roche Elecsys assays ranged from 1.1 to 2.9% over the measuring range for both assays, which comply with the demands set by Fetal Medicine Foundation for biochemical assays for prenatal markers. By method comparison the free βhCG and PAPP-A assays both showed good agreement with the Brahms Kryptor assays. Because of this overall good correlation to the Brahms Kryptor methods, the gestational age-specific medians for both the Elecsys free βhCG and PAPP-A assays show values and shape similar to what has been published earlier for the Brahms Kryptor assays [5, 15].

The Roche Elecsys free βhCG assay shows approximately 20% negative bias to the PerkinElmer AutoDELFIA assay and the Roche Elecsys PAPP-A assay shows a positive bias of 20% to the PerkinElmer AutoDELFIA assay. This bias is in accordance with the data reported from the Down Syndrome First trimester External Quality Assessment (EQA) program 2011 [16]. The Elecsys free βhCG assay shows a negative bias of approximately 20% to the Siemens IMMULITE assay, and the Roche Elecsys PAPP-A-assay shows approximately 40% positive bias to the IMMULITE assay. The bias for free βhCG is previously described between the Brahms Kryptor and the Siemens IMMULITE assays [14] and therefore also expected between the Roche and Siemens IMMULITE assays, and in accordance to the Down Syndrome First trimester EQA program 2011 [16]. The positive bias between the Roche Elecsys and the Siemens IMMULITE PAPP-A assays is not in accordance to the Down Syndrome First trimester EQA program 2011 [16], which indicates a negative bias to the Siemens assay. However, prospective EQA data are needed in order to determine the correct correlation between the assays.

Clinical performance

The performance of the Roche Elecsys assays in the first trimester screening risk assessment for fetal T21 was validated by data from two large multicenter studies employing samples from 2522 controls and 107 cases of fetal T21. The results confirmed previously published results performed with the Brahms Kryptor assay [3, 4, 15] that PAPP-A has increased screening performance in early weeks as compared to weeks 11–14, and that free βhCG has a better performance as screening marker in weeks 11–14 as compared to earlier weeks [15]. In the early weeks (gestational weeks 8–10) the median MoM values for Elecsys assays were approximately 1.35 (free βhCG) and 0.30 (PAPP-A) which is similar to the Brahms Kryptor assays 1.30–1.35 (free βhCG) and 0.35 (PAPP-A) [15]. For the late weeks (11–14) the Roche Elecsys assays showed median MoM values of 1.54 (free βhCG) and 0.34 (PAPP-A) which is lower than the Brahms Kryptor assays: 2.0 (free βhCG) and 0.50–0.60 (PAPP-A) [15]. The lower median MoM values of the 64 T21 cases for the Roche Elecsys assays in the late weeks, are lower than expected, considering the high degree of similarity between the Roche and the Brahms Kryptor assays. However, the results are not directly comparable since the MoM values of the assays were generated in different software systems and in different study populations, and further analyses comparing the assays in similar software are needed in order to determine the exact values.

The performance in the risk assessment for fetal T21 employing the Elecsys free βhCG and PAPP-A assays in combination with NT measurement, showed an overall sensitivity of 94–95% at a cut off at 1:300, and a false positive rate of 2–3% with equally good performance from samples drawn early or late in the first trimester. These results are superior to results obtained from prospective studies of first trimester combined screening with a performance closer to 85–90% for sensitivity and 95–97% for specificity [17, 18]. Because of the relative small number of case samples in the validation of potential clinical performance, the confidence intervals of both sensitivity and specificity are wide. It is also important to emphasize that the prevalence of trisomy cases in the case-control study population is higher than in a normal population, reaching 1:12 (cases:controls) in early samples, and 1:32 for late cases, and that results of the clinical performance of the assays must be assessed in a prospective study preferably from multiple centers. For risk analysis a variety of software solutions with multiple algorithms and coefficients is available. In this study risk analysis was done using the CE IVD approved software SsdwLab. Although there was only a trend towards a higher sensitivity in the early as compared to late weeks, the specificity was increased in the early weeks.

The present results indicates that the Roche Elecsys assays perform well as serum screening markers for first trimester screening, and comply with the standards set by the Fetal Medicine Foundation for biochemical screening markers. Although the study includes more than 100 cases of T21, the long term clinical performance of the markers must be determined based on large prospective multicenter studies.