Clinical experience with a single‐nucleotide polymorphism‐based non‐invasive prenatal test for five clinically significant microdeletions

Single‐nucleotide polymorphism (SNP)‐based non‐invasive prenatal testing (NIPT) can currently predict a subset of submicroscopic abnormalities associated with severe clinical manifestations. We retrospectively analyzed the performance of SNP‐based NIPT in 80 449 referrals for 22q11.2 deletion syndrome and 42 326 referrals for 1p36, cri‐du‐chat, Prader‐Willi, and Angelman microdeletion syndromes over a 1‐year period, and compared the original screening protocol with a revision that reflexively sequenced high‐risk calls at a higher depth of read. The prevalence of these microdeletion syndromes was also estimated in the referral population. The positive predictive value of the original test was 15.7% for 22q11.2 deletion syndrome, and 5.2% for the other 4 disorders combined. With the revised protocol, these values increased to 44.2% for 22q11.2 and 31.7% for the others. The 0.33% false‐positive rate (FPR) for 22q11.2 deletion syndrome decreased to 0.07% with the revised protocol. Similarly, the FPR for the other 4 disorders combined decreased from 0.56% to 0.07%. Minimal prevalences were estimated to be 1 in 1255 for 22q11.2 deletion syndrome and 1 in 1464 for 1p36, cri‐du‐chat, and Angelman syndromes combined. Our results show that these microdeletions are relatively common in the referral population, and that the performance of SNP‐based NIPT is improved with high‐depth resequencing.

of fetal loss to obtain fetal cells for analysis, 13 or because it may identify CNVs of uncertain clinical significance, some women may decline CMA. Analysis of cell-free DNA in maternal plasma potentially offers an alternative, non-invasive prenatal-testing path toward the identification of targeted microdeletions; only women found to be at highrisk need to be offered invasive testing. 4,[14][15][16][17] Currently, single-nucleotide polymorphism (SNP)-based noninvasive prenatal testing (NIPT) can screen for 5 clinically significant microdeletions 1 -22q11.2, 18 1p36, 19 cri-du-chat (5p15.3), 20 Prader-Willi (paternal 15q11-q13), 21, 22 and Angelman (maternal 15q11-q13). 21 Previous analysis of a cohort of nearly 22 000 women referred for screening for the 22q11.2 deletion demonstrated a 0.5% screen-positive rate (SPR) and an 18% positive predictive value (PPV). 23 Here, we report for the first time the screening performance of this SNP-based NIPT for 1p36, cri-du-chat, Prader-Willi, and Angelman microdeletion syndromes. We also update our clinical experience in screening for 22q11.2 deletion syndrome to now include over 80 000 referrals. 23 Earlier studies indicated that reflex sequencing of high-risk microdeletion test results at a higher depth of read (HDOR) may substantially reduce the false-positive rate (FPR). 17,24 In this study, we assess the screening performance of this modification, combined with a higher quality-control confidence threshold. We compare the performance of the original and revised protocols, and use these results to estimate the prevalence of the 5 microdeletions in the test population.  (Table S1, Supporting Information). 23 For each maternal blood sample, the maternal date of birth, maternal weight, gestational age, reason for testing, and informed consent were collected. Test-interpretation information, guidance on confirmatory testing options, genetic counseling, and disease-specific literature were available to test providers and patients.

| SNP-based analysis
Samples were analyzed at a Clinical Laboratory Improvement Actcertified, College of American Pathologists-accredited laboratory using previously described methodology. 4

| Definitions of risk
For the 22q11.2 deletion, samples with ≥90% confidence for the deletion at either allele were reported as "high-risk"; those with ≥90% confidence for no 22q11.2 deletion at both alleles were reported as "low-risk"; those with lower confidence values at the maternal allele were analyzed for paternal allele only; and those with lower confidence values at both alleles were reported as "risk unchanged." For the other microdeletion syndromes, samples with ≥80% confidence for a deletion were reported as "high-risk"; those with lower confidence values at the maternal allele were analyzed for paternal allele only; and those with lower confidence values at both alleles were reported as "risk unchanged." The reported risk status for each microdeletion syndrome was accompanied by a numerical estimate of fetal risk based on the population birth prevalence of the syndrome and the SNP analysis, and whether 1 or both alleles were successfully analyzed (Table S2). 4 Cases in which a maternal deletion was suspected were assigned a fetal risk score of 50%. For cases with fetal fractions <6% (22q11.2 deletion) or <7% For 22q, any contiguous loss of heterozygosity involving >80% of the region was reported as "high-risk." For other genomic regions, the full region was required to show loss of heterozygosity to be reported as "high-risk."

| Pregnancy follow-up
Follow-up information, including results of ultrasound examinations performed either prior to, or following NIPT, was requested from providers for all high-risk cases via phone or email. If the initial outreach while the pregnancy was ongoing was uninformative, at least 2 more attempts were made after the estimated delivery date. A deidentified copy of the genetic testing report was requested whenever diagnostic testing was performed. Providers were also requested to report false-negative results.
Cases identified as high-risk by NIPT were categorized as true positive (TP) or false positive (FP) based on the results of confirmatory diagnostic testing (CMA, fluorescence in situ hybridization, or methylation analysis for the 15q region). Samples without follow-up information, either because the patient did not have any additional testing or because there was no follow-up information available, were classified as "unknown" outcome.

| Performance metrics
The PPV, SPR, and FPR were calculated for each deletion syndrome

| Post-hoc analysis
Post-hoc analysis of test performance was performed in an internally blinded manner using a revised protocol with a higher (95%) confidence threshold for reporting a case as high-risk for a microdeletion, and reflex sequencing of high-risk cases at HDOR (≥6 million reads/ sample). 23,24 Because the presence or absence of deletions on the paternally inherited chromosome are readily discerned at normal depth of read, only cases with a suspected deletion on the maternally inherited chromosome were reflexed to HDOR. To allow for cases without post-hoc analysis results (due to samples being unavailable for re-sequencing) in the estimation of PPV, SPR, and FPR for the revised protocol, a proportionate adjustment was made to the total number of cases tested.

| Estimation of disease prevalence in the study population
The prevalence of each microdeletion syndrome in the population was calculated as: number of affected pregnancies/(number of cases with a test result × percentage of syndromic deletions expected to be captured by test), where the numerator is the sum of the number of known TPs, the expected number of affected cases in the group with an unconfirmed outcome, and the number of known false negative cases. Prevalence was calculated using only cases for which both maternally and paternally inherited alleles could be analyzed. For these estimates, it was assumed that the proportion of affected cases among samples with unknown outcome was equal to that of cases with known outcome.

| Institutional approval
The study was exempted from institutional review board approval (Ethical & Independent Review Services, Corte Madera, CA; Study ID 14064-01).

| Referrals
A total of 80 449 referrals were received for microdeletion screening during the study period, of which 42 326 were for the full panel of microdeletion syndromes and the remainder were limited to testing for the 22q11.2 deletion ( Figure 1 and Table S1). 23 Figure 1). An additional 6 cases received a high-risk result due to a suspected maternal deletion. Of the 283 cases identified as being at high-risk for a fetal microdeletion, follow-up information on copy-number truth was available for 153 (54.1%) cases (via invasive diagnostic testing, n = 117; postnatal diagnostic testing, n = 32; or post-miscarriage products-of-conception testing, n = 4).
Twenty-four (8.5%) of the high-risk results were TP (Table S4) and 129 (45.6%) were FP (Table 3).  Table 3). Truth was established on the basis of invasive diagnostic testing (118 cases), postnatal diagnostic testing (5 cases), or postmiscarriage products-of-conception testing (1 case). Additional details of the 7 TPs are in Table S4. One false-negative case was reported (for cri-du-chat syndrome).

| Maternal deletions
Of the 74 938 total cases screened (whether for 22q11.2 deletion alone or for all 5 microdeletions), a fetal risk score of 50% was assigned in 6 cases due to suspected deletions in the 22q11.2 region in the mother; no case had suspected maternal deletions in any of the other interrogated regions. Follow-up information was available for 3 of these 6 cases. In 2 cases, a maternal deletion of the 22q11.2 region was confirmed, but fetal copy number was not provided. In the third case, a fetal deletion was confirmed, and although the mother's copy number for the 22q11.2 region was not assessed, she had tetralogy of Fallot and learning disabilities, both of which are associated with 22q11.2 deletion syndrome.

| Performance based on presence or absence of ultrasound abnormalities
Test performance was compared for high-risk calls with, and without, major fetal structural abnormalities detected by ultrasound prior to NIPT (    Table 3 for complete listing of reasons for exclusion. b These cases are a subset of the 80 449 cases screened for the 22q11.2 deletion. c One case reported as low-risk for cri-du-chat microdeletion syndrome was a false negative. d Eight cases reported as false positive (per original protocol) were unavailable for reflex sequencing. e Seven cases with unknown outcomes reported as high-risk (per original protocol) were unavailable for reflex sequencing. f Three cases reported as false positive (per original protocol) were unavailable for reflex sequencing. g Three cases with unknown outcomes reported as high-risk (per original protocol) were unavailable for reflex sequencing. AS, Angelman syndrome; CdC, cri-du-chat; FP, false positive; PWS, Prader-Willi syndrome; TP, true positive.

| Prevalence of microdeletions
In this study, the estimated prevalence of 22q11.2 deletion syndrome was 1 in 1255, and those for the 1p36 deletion, cri-du-chat, and Angelman syndromes ranged from 1 in 3624 to 5820 (Figure 2). The prevalence of Prader-Willi syndrome was not estimated due to lack of a TP case. The combined estimated prevalence of 1p36, cri-du-chat, and Angelman microdeletions in this cohort was 1 in 1464, and for all 5 disorders combined it was 1 in 676. Details of the calculations for prevalence are presented in Appendix S1. Because there may have been additional false negative cases that did not come to attention, these prevalence rates are minimal estimates.

| Post-hoc analysis
The revised protocol was applied to 268 (24 TP, 121 FP, and 123 unknown) of the 283 cases (94.7%) that were determined to have a high-risk call for 22q11.2 deletion using the original protocol (the remaining cases were unavailable for analysis; Figure 1). Of the 121 FP cases, 92 (76.0%) were reclassified as low-risk with the revised protocol.
Conversely, for the known TP, 23 of 24 remained high-risk, a drop in detection rate of 4.2%. Of the 123 cases with unknown outcome, 88 (71.5%) were reclassified as low-risk, consistent with the assumption that these cases were a proportionate mixture of TP and FP cases (Appendix S2). Overall, there was a 67.2% (180/268) reduction in highrisk calls. Based on these findings, the SPR for the revised protocol was 0.13% (vs 0.39% with the original protocol), the PPV was 44.2% (vs 15.7% with the original protocol) and the FPR was 0.07% (vs 0.33% with the original protocol; Table 5).

| 1p36, cri-du-chat, Prader-Willi and Angelman microdeletions
Similar improvements were seen for the other microdeletions. Considering the 1p36, cri-du-chat, Prader-Will and Angelman results together (n = 215), the revised protocol was applied to 209 (7 TP, 114 FP, and 88 unknown) cases (97.2%) determined to be high-risk   c Upper and lower boundaries assuming all unknown cases were TP or FP, respectively. with the original protocol (the remaining cases were unavailable for analysis; Figure 1). Known FP results were reduced by 86.8% with the original protocol) and an FPR of 0.07% (vs 0.56% with the original protocol) for the combination of disorders (Table 5).

| DISCUSSION
This study extends our initial report 23  We also show that applying a stricter quality-control confidence threshold and reflexively sequencing high-risk samples to a HDOR with the revised screening protocol reduced FPRs and increased PPVs substantially. As a result of these performance improvements, the revised protocol has been implemented commercially.
Our initial study indicated that screening for 22q11.2 deletion syndrome using the original protocol would be associated with a 0.38% FPR, an 18.0% PPV, and a prevalence of 1 in 962 in the test population. 23 In this study, with the original protocol, the FPR was 0.33%, the PPV was 15.7%, and the estimated prevalence was 1 in 1255. With a revised protocol, performance improved to a projected 0.07% FPR and 44.2% PPV. The revised protocol also exhibited relatively high PPVs for the other 4 microdeletion syndromes, although these had larger confidence intervals (Table 5). Because microdeletions are less prevalent than aneuploidies in NIPT cohorts, the PPVs for screening microdeletions are expected to be lower than those observed for aneuploidy. 15,17 In fact, observed PPVs for microdeletions using the updated protocol are lower than reported NIPT PPVs for trisomy 18 and trisomy 21, and similar to those for trisomy 13 and monosomy X. 15,27,28 Moreover, the PPVs we observed are similar to, or better than, those observed with conventional maternal serum screening for trisomies-a bar that has been considered sufficient to justify routine use.
Prior publications have questioned the use of NIPT as a screening test for microdeletions, citing concerns about high FPRs, low sensitivities, and challenges associated with variants of unknown significance. 29,30 However, these reports focused on whole-genome sequencing approaches that employ counting-based methodologies.
By concentrating on specific genomic regions with clinically significant deletions, the targeted nature of the SNP-based method overcomes many of the limitations discussed in these publications. 31 Furthermore, comparison of detection rates for confirmed microdeletions in clinical cohorts and ratios of maternally inherited vs de novo deletions to published ratios suggests that SNP-based methods have substantially higher sensitivity than counting-based methods. [32][33][34] In this study, the observed FPR for Angelman syndrome (conveyed by the maternally inherited chromosome 15q11-13 region) was substantially higher than that for Prader-Willi syndrome (conveyed by the same and open neural-tube defects. 39 The observed prevalences and test performance therefore exceed that of other prenatal screening tests long considered standard of care. Consistent with recent American College of Medical Genetics and Genomics guidelines, 40 our findings support offering microdeletion screening as an adjunct to existing NIPT to refine risks for these 5 clinically significant, well-characterized genetic disorders.