Screen‐positive rate in cell‐free DNA screening for microdeletion 22q11.2

To examine the impact of the fetal fraction (FF) on the screen‐positive rate in screening for microdeletion 22q11.2.

from 1 per 3000 to 1 per 6000 live births. 8 However, in more recent studies, the prevalence was suspected to be up to 1 in 1000. 9 Schmid et al. examined the test performance of cfDNA screening for microdeletion 22q11.2 in 1953 plasma samples, including 122 samples from affected pregnancies. 10 The detection and the falsepositive rate were 75.2% and 0.4%, respectively. In a prospective clinical trial, Kagan et al. focused on the false-positive and test failure rate in cfDNA screening for common trisomies and microdeletion 22q11.2. 11 The study included 1127 singleton pregnancies at 11-13 weeks' gestation. The test failure rate after two potential blood analyses was 0.9% and the false-positive rate 0.27%, respectively.
Bevilacqua et al. examined the test performance in a prospectively collected cohort of fetuses with suspected cardiac defects. 12 In 735 pregnancies, including 46 with a 22q11.2 deletion, the cfDNA test identified correctly 69.6% of affected cases with no false-positive results.
Interestingly, there is a wide range of the positive predictive values (PPVs) between the studies-even between those that were carried out in the general population. Martin 13,14 In a more recent series of Chen et al. including almost 40,000 samples, the PPV in screening microdeletions and duplications was 42%. 15 As most of the studies were done in screening populations, it is unclear why the PPVs are so heterogenous. However, in screening for trisomies, the fetal fraction (FF) is considered as a quality parameter with a strong impact on the test performance and as such also on the PPV. 16,17 Several work groups have shown that a decrease of FF rate results in a reduction of the discriminatory power of the cfDNA test. 17 In a theoretical model, Benn and Cuckle observed a similar effect of the FF on the test performance in screening for microdeletions. 18 In this study, we set out to examine the impact of the FF on the test performance in screening for microdeletion 22q11.2. For this purpose, we examined the test results of a large, prospectively collected cohort with more than 50,000 pregnancies.

| METHODS
This study is based on samples that were analyzed by Cenata, a private laboratory for non-invasive prenatal testing in Tuebingen, Germany. The samples were examined using the Harmony® Prenatal Test (Roche Inc) as previously described. 10,[19][20][21] In brief, this is a targeted cfDNA analysis performed using microarray quantitation of DANSR (Digital ANalysis of Selected Regions) assays of nonpolymorphic (chromosomes 13, 18, 21, X and Y) and polymorphic loci (chromosomes 1-12) to estimate chromosome proportion and FF, respectively. The FORTE (Fetal fraction Optimized Risk of Trisomy Evaluation) algorithm is used to provide patient-specific risk assessments for trisomy. Approximately 500 additional DANSR assays were designed against targets uniformly distributed within a 3.0 Mb region of 22q11.2 between LCR-A and D. Custom DNA microarrays were manufactured by Affymetrix Inc. to specifically quantify all DANSR assay products. Each sample was analyzed on a single custom microarray.
The measurement of the FF is performed using an SNP-based approach as described by Sparks et al. and Juneau et al. 19,20 In short, the measurement is based on the DANSR assays designed against polymorphic loci on chromosomes 1-12. The assays are unbiased, cover each chromosome uniformly, and target single nucleotide polymorphisms (SNPs) with high minor-allele frequencies in the HapMap dataset to maximize the number of informative loci for fetal-fraction measurement. 10 Relative allele frequencies are used to compute FF with the minor percentage of the fraction assumed to be fetal.
Our study cohort comprised samples from women of at least 16 years of age with singleton pregnancies. The gestational age at the time of blood sampling was at least 11 weeks' gestation. Blood was collected in two 10 mL blood collection tubes containing a stabilizer for cell-free DNA and sent to Cenata within 7 days after collection at room temperature without prior processing. All women were counseled prior to the cfDNA analysis and after testing according to national regulations.
We decided to focus on the screen-positive rate rather than on the detection and the false-positive rate. This is due to the fact that many pregnant women opted against an amniocentesis, a CVS, or postnatal testing after having received a high-risk test result for microdeletion 22q11.2. Furthermore, it would have been necessary to test all neonates, including the asymptomatic ones for a reliable assessment of the detection and the false-positive rate.
The data analysis was done as part of the continuous quality assurance program of Cenata.
For the evaluation, an anonymized dataset was used. Thus, a statement of the ethical committee of the university hospital of Tübingen is not necessary.

| Statistical analysis
Statistical analysis was performed using SPSS 26 and Microsoft Excel.
Data are presented as median (25-75th interquartile range) and percentage whatever was appropriate. We used uni-and multivariate logistic regression analyses to detect significant covariates that are associated with a high-risk result for microdeletion 22q11.2. The difference between the screen-positive rate in the group of cfDNA tests with a FF rate of <11% and those with a higher FF rate was investigated using the chi square test. Confidence intervals were calculated according to the Clopper-Pearson method. A p-value of ≤0.05 was used as a significance level.

| RESULTS
Between July 2018 and August 2021, 55,042 blood samples were sent to Cenata for cfDNA screening for common trisomies and microdeletion 22q11.2. Of those, in 393 (0.7%) cases, the analysis LÜTHGENS ET AL.
-289 was not possible mainly because the transport to the laboratory took too long. In 2630 (4.8%) cases, the cfDNA analysis for microdeletion 22q11.2 remained inconclusive. These cases were excluded from the further analysis. Thus, the study population consisted of 52,019 pregnancies. 51,710 women received a low-risk result for microdeletion 22q11.2 and in 309 pregnancies, the cfDNA test indicated a high risk for microdeletion 22q11.2. Thus, the overall screen-positive rate was 0.59%. The detailed pregnancy characteristics of the cohort with lowrisk and high-risk results are given in Table 1. In short, median maternal and gestational age in the low-risk group were 34 years and 12.1 weeks, and in the high-risk cohort, it was 33.0 and 12 + 3 weeks.
Maternal weights in both groups were 65.7 and 69.5 kg, respectively.
In the low-risk population, the FF was 10.1% and in the high-risk group, it was 7.3%. Uni-and multivariate logistic regression analyses were used to detect significant covariates that were associated with a high-risk result in screening for microdeletion 22q11.2. There was a positive correlation between gestational age and a negative correlation between maternal age and the screen-positive rate ( Table 2). The screen-positive rate according to the gestational and maternal age is shown in Figures 1 and 2. However, the correlation between the FF and the screenpositive rate was much stronger. Figure 3 demonstrates the screen-positive rate stratified by the FF. In the 30,541 cases with a FF of <11.0%, there were 282 (0.92%) screen-positive cases. In contrast, in the group of 21,478 cases with a FF of 11.0% or more, there were only 0.13% screen-positive cases (n = 27; chi-square test: p < 0.001). With other words, the screen-positive rate was 67% (+0.36% points) higher than the average if the FF was below 11% and 77% (−0.43% points) lower if the FF was above the threshold of 11%.

| DISCUSSION
In this study, we have observed an overall screen-positive rate in cfDNA screening for microdeletion 22q11.2 of 0.59%. There was a significant correlation between the screen-positive rate and the maternal and gestational age. However, the FF carried the stron-  -291 lower FF, 282 positive test results were observed, which correspond to 0.92%.

Univariate analysis
The prevalence of microdeletion 22q11.2 is estimated to be around 1 per 3000 to 1 per 6000. Some research groups suspect the prevalence to be as high as 1 in 1000. 9 Based on this prevalence, our study population of 52,019 pregnancies should include 52 (95% prediction interval 36-66) affected fetuses. However, we observed 309 positive test results, which is 6 times more than expected.
Interestingly, if only the sub-cohort of pregnancies with a FF of 11% The importance of the FF as a quality marker in cfDNA screening for common trisomies is well known and results should only be reported if the FF is above a certain threshold. 16 Many maternal and fetal covariates alter the FF, such as the maternal weight, the gestational age, or the fetal karyotype. 16 23 Person and Cuckle modeled the effect of such a variability of the measurements and found that an increase in the FF cut-off led to a slight increase in the detection rate, when no-calls were excluded, and a larger decrease in the detection rate, when no-calls were included. 24 These effects were smaller for tests with higher discriminatory power and larger as maternal weight increased.
In contrast, Schmid et al. examined the reproducibility of an SNPbased assay and compared it with the measurements of a Y-sequence quantification as direct representation of the fetal DNA in pregnancies with a male fetus. 25 The SNP-based approach was also used in this study. The authors found a strong correlation between both ways of measuring the FF. They highlighted that in a case with a real FF of 7%, the variation in the measurement is only +/− 0.26% FF.  27 Clinically, many of them are benign, 28 but they could nevertheless lead to a positive result in cfDNA testing, especially in cases with a low FF.
We acknowledge that our study has some limitations. Most importantly, we can only report on the screen-positive rate as the follow-up rate is too low to directly address the false-positive rate.
However, given the known prevalence of the disease and the substantial difference in the screen-positive rate between the low and high FF groups, one can still draw general conclusions in screening for microdeletion 22q11.2.
In conclusion, in our study, we have shown that the screenpositive rate depends on the FF. We have speculated that this is most likely caused by an increase in the false-positive rate with a decreasing FF.