Cell‐free DNA screening for rare autosomal trisomies and segmental chromosome imbalances

Abstract Objective To assess the outcomes of pregnancies at high‐risk for rare autosomal trisomies (RATs) and segmental imbalances (SIs) on cell‐free DNA (cfDNA) screening. Method A retrospective study of women who underwent cfDNA screening between September 2019 and July 2021 at three ultrasound services in Australia. Positive predictive values (PPVs) were calculated using fetal chromosomal analysis. Results Among 23,857 women screened, there were 93 high‐risk results for RATs (0.39%) and 82 for SIs (0.34%). The PPVs were 3.8% (3/78, 95% CI 0.8%–10.8%) for RATs and 19.1% (13/68, 95% CI 10.6%–30.5%) for SIs. If fetuses with structural anomalies were also counted as true‐positive cases, the PPV for RATS increased to 8.5% (7/82, 95% CI 3.5%–16.8%). Among 85 discordant cases with birth outcomes available (65.4%), discordant positive RATs had a significantly higher proportion of infants born below the 10th and 3rd birthweight percentiles than expected (19.6% (p = 0.022) and 9.8% (p = 0.004), respectively), which was not observed in the SI group (2.9% < 10th (p = 0.168) and 0.0% <3rd (p = 0.305)). Conclusion The PPVs for SI and RAT results are low, except when a structural abnormality is also present. Discordant positive RATs are associated with growth restriction.


| INTRODUCTION
Cell-free DNA (cfDNA) screening, commonly referred to as noninvasive prenatal testing, is a method of prenatal aneuploidy screening. CfDNA released by cytotrophoblast and nucleated maternal cells is extracted from a maternal plasma sample after 10 weeks' gestation, and bioinformatic analysis of the sequences can indicate chromosomal anomalies. 1 Importantly, the screening performance of cfDNA for fetal aneuploidy relies on concordance between the fetal and placental genomes. 2,3 Compared to alternatives, cfDNA screening has demonstrated impressive detection rates and accuracy in predicting trisomies 21, 18 and 13. The sensitivity of cfDNA for trisomy 21, 18 and 13 is 99.7%, 97.9% and 99.0% respectively, with a false-positive rate of 0.04% each. 4 The PPVs for trisomy 21, 18 and 13 are 92.4%, 84.6% and 43.9%. 5 More recently, cfDNA screening has been expanded to report anomalies for all chromosomes, as opposed to targeted panels which assess only chromosomes 21, 18, 13 and the sex chromosomes. This expanded test additionally screens for rare autosomal trisomies (RATs) and segmental imbalances (SIs). Rare autosomal trisomies may result from either mitotic (most common for chromosomes 2, 3, 7, 8), or meiotic errors (commonly chromosomes 14,15,16,22). [5][6][7] Nonmosaic RATs are a rare finding in fetal tissues beyond early pregnancy, as most of these anomalies are life-limiting and miscarry early. 6,[8][9][10] Conversely, many SIs do not result in miscarriage, with chromosomal deletions (excluding microdeletions) estimated to be present in 0.019% of live births, and duplications in 0.007%. 11 The clinical implications for many of these anomalies are poorly mapped, as phenotypic manifestations are highly variable and depend on the gene region involved as well as the size of the aberration, but range from benign to profoundly disabling. [12][13][14] As the use of whole-genome prenatal screening increases globally, attention has focused on the accuracy in identifying these rare anomalies, given that predictive value of screening is related to disease prevalence. Given their rarity in fetal tissues, it is theorized that the majority of RATs identified by cfDNA are confined to the placenta, in full or mosaic form. 15,16 This is a biologically plausible explanation for higher rates of adverse pregnancy outcomes, including fetal growth restriction (FGR), amongst women who screen high-risk for RATs independent of fetal karyotype, although there is debate regarding the strength of this association. 6,[17][18][19][20][21][22] Evidence for the performance of cfDNA for fetal SIs is also limited. 18,[23][24][25] The aims of this study are to assess the predictive accuracy of cfDNA in screening for RATs and SIs, and to investigate the association between discordant positive results and placental insufficiency.

| Study design and governance
This was a retrospective study performed at three fetal medicine practices in Australia (two in Sydney and one in Melbourne). Women underwent cfDNA screening as part of routine clinical care governed by their respective obstetric care providers, as either a primary screening method or subsequent to a high-risk result from other methods of screening or ultrasound findings. Screening was patientfunded. Ethical approval for this study was obtained from the Monash University Human Research Ethics Committee (project ID 26175).

| Inclusion and exclusion criteria
We studied women who received a high-risk cfDNA result for either a RAT or SI between September 2019 and July 2021. Women aged over 18 years with high-risk cfDNA results and available postscreening outcomes were included. Women aged under 18 years were excluded, due to capacity to consent, as were those women who opted for screening of common aneuploidies only, rather than genome-wide cfDNA screening. Women with a multiple pregnancy were not excluded.
Some participants between November 2019 and December 2020 have been included in another publication from our group, 26 however such publication focused on the impact of fibroids on test accuracy rather than on overall test accuracy and pregnancy outcomes, and were therefore not considered duplicated data.

| Procedures
Maternal plasma samples were collected at or after 10 weeks' gestation. All providers offered screening with next-generation massively parallel sequencing technology using the Illumina VeriSeq

| Statistical analysis
Categorical variables are expressed as numbers and proportions and continuous variables in means and standard deviations (SD) or medians and interquartile ranges (IQR) depending on the frequency distribution. Differences in means and medians were analysed using independent-samples t-tests and Wilcoxon rank-sum test, as appropriate. The difference in proportions of pregnancies featuring ultrasound anomalies depending on screening concordance was assessed with the Chi-square test.
Positive predictive value was calculated with exact binomial 95% confidence intervals (CI). Calculations were based on fetal genetic confirmation, thus unconfirmed karyotypes, where only phenotypes were available, were not included in the primary analysis. Women in whom diagnostic testing revealed uniparental disomy (UPD) were classified as 'true-positive' RATs, as trisomy was once most likely present in the embryo prior to trisomic rescue. 26 Accordingly, results which were confirmed to be attributable to CPM (mosaic trisomy detected on chorionic villus sampling (CVS) followed by normal amniocentesis) were considered 'discordant positive', as the analysis pertains to true fetal anomalies. For SIs, screening results were considered concordant to diagnostic investigations providing the aberration was located on the same chromosome. Where women received high-risk results for multiple SIs, variants located on different chromosomes were considered individually.
Birthweight percentiles were calculated using the Australian national birthweight by sex and gestational age chart by Dobbins et al. 27 The Wilcoxon signed-rank test was used to assess for differences between the distribution of birthweight percentiles in each anomaly group and the expected birthweight percentile distribution in the general population, centered at the 50th percentile. For RATs, the median birthweight percentiles were compared between the chromosome trisomies most likely of mitotic origin (2,3,7,8), versus meiotic (14,15,16,22), with grouping guided by the current literature. 5
Maternal characteristics are summarized in Table 1    Outside of the RAT and SI cohorts, monosomy 2 and monosomy 14 were also observed. The fetus screened high-risk for monosomy 14 was found to have no abnormalities on amniocentesis and resulted in a healthy neonate, while the woman with a high-risk monosomy 2 result declined diagnostic testing in favor of immediate termination of pregnancy.

| Main findings and interpretation
In this large multicenter study, there were confirmed fetal genetic anomalies in almost 4% of women at high-risk for a RAT on cfDNA screening where complete diagnostic information was available. This rose to 8.5% when also considering pregnancies with incomplete genetic information where fetal structural or infant anomalies were identified, most of which terminated without diagnostic investigations. The PPV of a high-risk result for a SI was 19.1%. The odds of fetal confirmation of a high-risk RAT or SI cfDNA result were 19.6 times greater in the presence of ultrasound anomalies than those without. 28 The low PPV observed in the RAT group can be attributed to our specification of true-positive status only applying to fetal, not placental, confirmation of aneuploidy. Fetal confirmation of a nonmosaic RAT beyond 10 weeks' gestation is exceedingly rare as these anomalies typically cause early miscarriage. 6,8 Therefore, when a RAT is detected by cfDNA screening it is probable that the aneuploidy is confined to the placenta, or less commonly, that the fetus is mosaic. 15,16 The lower PPV of cfDNA screening for SIs observed in this study compared to common aneuploidies may also be attributable to confined placental involvement, in addition to test inaccuracies arising from sporadic technical issues.

| Clinical implications
At best, approximately 1 in 13 pregnancies that screen positive for RAT will actually have fetal aneuploidy. The rate of fetal confirmation increases in the presence of ultrasound anomalies, although fetal anomalies detected before cfDNA screening may be more appropriately investigated by a diagnostic procedure rather than by cfDNA screening. 33 The association between high-risk RAT results and FGR warrants more rigorous growth monitoring in these pregnancies. A discordant positive RAT result in combination with low PAPP-A levels indicates greater risk of FGR than with normal PAPP-A, although a relationship between low PAPP-A and FGR exists independent of cfDNA screening results. 34 Another potential indicator of placental insufficiency is the origin of aneuploidy (mitotic or meiotic) suggested by the implicated chromosome, but this is not routinely considered in current cfDNA screening. 35,36 The clinical utility of cfDNA screening for SIs is complex. Although suggested improvements for the provision of genome-wide cfDNA screening include restricting referrals for invasive diagnostic testing to high-risk results accompanied by ultrasound anomalies due to the lower PPV for rare anomalies, this approach risks misclassifying clinically significant true-positive results as many SIs may not present with abnormal ultrasound findings, as demonstrated in the current study. 37 However, even when successfully identified, the variable phenotypic consequences are a challenge for genetic counselling. 38,39 Another important consideration is that the frequency of inaccurate cfDNA results increases with every condition added to the 1354screening panel. Higher screen-positive rates will likely lead more women to be referred for invasive diagnostic testing with its low but non-negligible procedure-related risks, as demonstrated by the woman with septic miscarriage in this study. 40

| Limitations
The main limitations of this study are attributable to incomplete or insufficient pregnancy and infant outcome information, which introduces risk of ascertainment bias. Without complete diagnostic confirmation for the entire high-risk group, it is possible that our PPVs are biased by confounding clinical factors which influence a woman's decision to pursue additional investigations and diagnostic testing. Similarly, our findings pertaining to birth outcomes may have been affected by ascertainment bias due to incomplete follow-up of the entire population, as pregnancies screened high-risk for anomalies with evidenced adverse outcomes, such as trisomy 16, would likely demand closer clinical monitoring.
Calculation of test sensitivity and specificity would have only been possible with complete genetic follow-up of women with high and low-risk results, which is often not feasible in clinical practice.
Therefore, we limited the analysis to the predictive value of a positive test, which is arguably the most important accuracy measure from an individual patient's perspective. 44 Another limitation of this study arises when drawing conclusions regarding the risk of FGR associated with discordant positive cfDNA results. While we hypothesize that this risk is attributable to placental aneuploidy, particularly in the presence of meiotic RATs, it is impossible to ascertain whether discordant positive results actually arise from the placenta or are sporadic inaccuracies without conducting detailed placental cytogenetic analyses, which we were unable to undertake in this study.

| CONCLUSION
Decisions regarding screening with expanded cfDNA panels at both the population and individual patient level should take into account the harms of increased discordant positive results along with the benefits of detection of significant rare chromosomal abnormalities that would otherwise be missed prenatally. Closer third trimester monitoring is indicated for high-risk RAT results even following exclusion of fetal aneuploidy due to the risk of placental insufficiency.