Cell-free fetal DNA screening for detection of microdeletion syndromes: a cost-effectiveness analysis

Abstract

Objective:

Fetuses with genetic copy number variants are poorly detected through traditional prenatal screening. Microdeletions and duplications are clearly identified with diagnostic testing through chromosomal microarray, and screening of a select number of microdeletions has become available with cell-free DNA (cfDNA). Our study compares the costs and outcomes of cfDNA for five pathogenic microdeletions and aneuploidy to cfDNA for aneuploidy alone in conjunction with ultrasound.

Methods:

A decision-analytic model was constructed using TreeAge software to compare cfDNA microdeletions versus traditional cfDNA in a theoretical cohort of 4,000,000 pregnancies that would also be screened with ultrasound. Probabilities, costs, and utilities were derived from literature. The primary outcomes were the incremental cost per quality-adjusted life-year (QALY), terminations, and procedure-related losses. Because the microdeletion results are available, but not reported, on all cfDNA testing we set the incremental cost of the cfDNA microdeletion screening test to zero at baseline and varied the cost in sensitivity analysis.

Results:

Screening with cfDNA for microdeletions among all pregnant women would result in83 fewer anomalous neonates compared to traditional cfDNA with ultrasound. This reduction is due to increased diagnosis and termination of fetuses with microdeletions in this group. Routine use of cfDNA with microdeletions resulted in more procedure-related losses. cfDNA with microdeletions would improve effectiveness by 977 QALYs and decrease costs by $90,991,784. When we varied the specificity of the screening test, we found that it remained cost-effective down to a specificity of 91%. With a threshold of $100,000/QALY, microdeletion screening is cost-effective to an incremental increase in cost over cfDNA for aneuploidy alone of $47.10.

Conclusion:

For detection of fetal subchromosomal abnormalities, use of cfDNA with microdeletions is a cost-effective strategy compared to cfDNA for aneuploidy alone in conjunction with ultrasound. Cell-free DNA for microdeletions is not currently recommended as routine screening for low-risk obstetric populations by the American College of Obstetrics and Gynecologists or the Society for Maternal–Fetal Medicine. The test characteristics of cfDNA with microdeletions require greater examination before being routinely recommended.

Introduction

Clinically significant copy number variants (CNVs) such as microdeletions arise spontaneously or through familial inheritance, independent of maternal age [1,2]. Although the incidence of each specific microdeletion (or microduplication) syndrome is overall low, such syndromes collectively are common, affecting 1–2% of pregnancies with a structurally normal fetus and 6% of pregnancies with a fetal structural anomaly [3,4]. Because neonates affected by microdeletion syndromes may benefit from early therapeutic intervention, prenatal detection is critical for optimal care and outcomes [5]. Traditional aneuploidy screening is not designed to detect microdeletions, and the prenatal phenotypes associated with microdeletions can be subtle, thereby limiting ultrasound as a primary screening method [3,6–9].

Pathogenic CNVs may result in a spectrum of phenotypes associated with physical impairments, structural abnormalities, and intellectual disability. Accounting for nearly 24% of congenital disease, CNVs comprise the second-largest contribution to congenital disease after structural anomalies [10]. Classic examples of microdeletion syndromes include 22q11.2 (DiGeorge/Velocardiofacial), 15q11–q13 (Prader–Willi and Angelman), 5p15.3 (Cri-du-chat), and 1p36 deletion syndrome. These conditions can be diagnosed prenatally with invasive testing by chromosomal microarray analysis of fetal or placental cells in those who opt for diagnostic testing [11].

Over the past decade, commercial technology using cell-free DNA (cfDNA) has emerged as a prenatal screening option for fetal aneuploidy [12]. More recently, cfDNA screening for chromosomal microdeletions has been introduced. This test performs with high sensitivity on internal validation studies; however, the positive predictive value is low, owing to the low baseline prevalence of microdeletion syndromes [12–15]. In the absence of rigorous clinical validation, cfDNA for microdeletions is not currently recommended as routine screening for low-risk obstetric populations by the American College of Obstetrics and Gynecologists (ACOG) or the Society for Maternal–Fetal Medicine (SMFM) [16,17].

Given this background, we aimed to investigate the costs and outcomes associated with cfDNA detection of five microdeletion syndromes, comparing cfDNA to screen for microdeletions and aneuploidy to cfDNA for aneuploidy alone but in conjunction with ultrasound. We hypothesized that cfDNA screening for fetal micro-deletion syndromes would be a cost-effective strategy.

Materials and methods

We constructed a decision-analytic model using TreeAge Pro software (2019, Williamstown, MA) to simulate a theoretical cohort of four million pregnant women undergoing prenatal genetic screening in the USA (Figure 1). This approximates the average number of continuing pregnancies in the USA each year [18]. As this theoretical model did not involve human subjects, it was exempt from Institutional Review Board approval.

Figure 1.

Cost-effectiveness model. [+] indicates collapsed branches that lead to identical branches as the ones that are displayed.

In our model, the initial decision node stratified pregnant women with noninvasive prenatal testing into two pathways: (1) cfDNA screen for aneuploidy and five microdeletion syndromes including 22q11.2, Prader–Willi, Angelman, Cri-du-chat, and 1p36 deletion syndrome and (2) cfDNA screen for aneuploidy alone plus routine ultrasound (Figure 1). We chose these microdeletions because they are clinically available on cfDNA tests in addition to aneuploidies trisomy 13, 18 and 21. The model assumed that cfDNA microdeletion screening results were characterized as positive or negative. Women who screened positive for microdeletions were offered diagnostic testing with chromosomal microarray analysis through amniocentesis or could elect to forego further testing. Women who screened negative for microdeletions were followed-up with ultrasound; if the ultrasound was positive for a structural abnormality, women could elect to pursue amniocentesis for chromosomal microarray analysis or forego prenatal diagnostic testing. Amniocentesis was also offered in the second strategy of our model to women with structural abnormalities present on routine ultrasound. Pregnancy loss due to amniocentesis and elective termination after identification of a micro-deletion syndrome at amniocentesis were outcomes included in our model. We also factored the increased rate of intrauterine fetal demise in the second and third trimester as well as estimates of neonatal death associated with microdeletion syndromes. Because the cfDNA for aneuploidy screening was identical in both arms, aneuploidy outcomes were not included in the model. All screening pathways were extended to the same potential endpoints.

The baseline model inputs were derived from the literature (Table 1). We used large population studies to estimate the incidence of microdeletion syndromes in the USA [19–23]. We combined the prevalence of the 15q11–q13 deletion syndromes, Prader–Willi and Angelman Syndromes (PWA) and considered only the cases caused by deletion. The vast majority of Prader–Willi and Angelman Syndromes are caused by deletions [21,23]. The sensitivity of cfDNA screening for microdeletion syndromes including 22q11.2, PWA, 1p36, and Cri-du-chat was derived from two studies performing single-nucleotide polymorphism (SNP) analysis for screening of microdeletions using cfDNA [1,24]. The test performance characteristics are similar to others reported in literature [2,12]. Given small sample sizes for individual microdeletion syndromes, we calculated a combined sensitivity of 98.6% and specificity of 99.1% to account for the microdeletions collectively. We assumed positive tests were high-risk calls and negative tests were low risk and unchanged risk calls. We did not account for test failure. Furthermore, any test positive result without diagnostic follow-up was conservatively assumed to be false positive. Probabilities regarding abnormal structural sonographic findings specific for each microdeletion syndrome were derived from retrospective cohort studies and literature reviews [6,7,9,25,26].

Table 1.

Probabilities, costs, and utilities used in the model.

Variable	Value	Range considered in sensitivity analysis	Reference
Sensitivity of cfDNA for microdeletion syndromes 22q11.2, PWA, 1p36 and Cri-du-chat.	0.986	0.1–1	[1,24]
Specificity of cfDNA microdeletions	0.991	0.1–1	[1,24]
Population prevalence of 22q11.2	0.000168	0.0001–0.001	[19]
Population prevalence of PWA	0.0000378	0.00001–0.0001	[21,23]
Population prevalence of 1p36	0.0001	0.00001–0.001	[22]
Population prevalence of Cri-du-chat	0.0000728	0.00001–0.0001	[20]
Probability of amniocentesis uptake	0.7	0.1–1.0	[27]
Probability of amniocentesis-related loss	0.006	0.001–0.01	[28]
Probability of abnormal US in 22q11.2	0.868	0.6–0.95	[6]
Probability of abnormal US in PWA	0.273	0.1–0.5	[7]
Probability of abnormal US in 1p36	0.279	0.1 −0.5	[25,26]
Probability of abnormal US in Cri-du-chat	0.482	0.2–0.7	[9]
Termination rate in 22q11.2	0.689	0.3–1.0	[6]
Termination rate in PWA	0.812	0.3–1.0	[29]
Termination rate in 1p36	0.836	0.3–1.0	[32]
Termination rate in Cri-du-chat	0.836	0.3–1.0	[32]
Probability of fetal loss in 22q11.2	0.361	0.1–0.5	[31]
Probability of fetal loss in PWA	0.361	0.1–0.5	[31]
Probability of fetal loss in 1p36	0.64	0.5–0.85	[31]
Probability of fetal loss in Cri-du-chat	0.64	0.5–0.85	[31]
Probability of normal fetal loss	0.005	0.001–0.01	[41]
Probability of neonatal death in 22q11.2	0.164	0.01–0.3	[6]
Probability of neonatal death in PWA	0.164	0.1–0.4	[6]
Probability of neonatal death in 1p36	0.7	0.5–0.9	[30]
Probability of neonatal death in Cri-du-chat	0.7	0.3–0.9	[30]
Probability of neonatal death in normal fetus	0.00218	0.001–0.005	[42]
Cost of cfDNA screen for aneuploidies	$1,726	$0–10,000	[33]
Cost of cfDNA screen for aneuploidies with added microdeletions	$0	$0–5,000	See text
Cost of genetic counseling	$95	$50–150	[27]
Cost of 20-week ultrasound	$229	$100–500	[34]
Cost of amniocentesis	$584	$400–1,000	[27,36]
Cost of postnatal diagnosis	$282	$100–700	[36]
Cost of termination or fetal loss	$3,165	$2,500–4,000	[27]
Cost of neonatal death	$111,940	$90,000–115,000	[35]
Cost of lifetime 22q11.2	$1,249,429	$900,000–1,500,000	See text
Cost of lifetime PWA	$1,252,029	$900,000–1,500,000	See text
Cost of lifetime 1p36	$1,252,029	$900,000–1,500,000	See text
Cost of lifetime Cri-du-chat	$1,295,654	$900,000–1,500,000	See text
Cost of first 2 years of DS	$89,799		See text
Cost of lifetime DS	$1,072,352		[27]
Utility of amniocentesis-related or spontaneous fetal loss during 2nd/3rd trimester	0.93	0.9–0.95	[38]
Utility of pregnancy termination	0.91	0.85–0.95	[38]
Utility of neonatal death	0.76	0.7–0.8	[39]
Utility of neonate with 22q11.2	0.81	0.75–0.86	[38]
Utility of neonate with PWA	0.81	0.75–0.86	[38]
Utility of neonate with 1p36	0.65	0.6–0.7	[39]
Utility of neonate with Cri-du-chat	0.65	0.6–0.7	[39]
Utility of normal neonate	1	Assumed	[38]

PWA: Prader–Willi and Angelman syndromes; US: ultrasound; DS: Down syndrome.

In the model, amniocentesis was offered as a diagnostic test following a positive cfDNA screen for microdeletions or structural ultrasound anomaly. In these cases, we estimated that 70% of women would elect to proceed with invasive testing and varied this estimate in sensitivity analysis [27]. The rate of amniocentesis-related fetal loss of 0.6% was determined by a meta-analysis of controlled studies that relied on concurrent ultrasound needle guidance [28]. Elective termination rates for 22q11.2 were derived from a retrospective cohort study, whereas termination for PWA was determined through a questionnaire regarding attitudes of pregnancy termination in parents of PWS individuals [6,29]. Due to minimal data, we assumed the rates of termination, pregnancy loss in the second and third trimester, and neonatal death associated with 1p36 and Cri-du-chat to be the same as those observed in T13/18 [30–32]. Pregnancy loss in the second and third trimester in 22q11.2 and PWA were assumed to be similar to the rates associated with Down syndrome (DS) [31]. Lastly, the probability of neonatal death for 22q11.2 was derived from a retrospective cohort study; the probability of neonatal death in PWA was assumed to be the same as neonatal death in 22q11.2 based on the similarity of phenotypic severity [5].

A majority of costs were obtained from literature and standardized to 2019 US dollars following the medical component of the Consumer Price Index (Table 1). The costs assumed a societal perspective. The cost of cfDNA screen for aneuploidies was derived from a review of commercial noninvasive prenatal testing in the USA [33]. Commercial noninvasive prenatal tests for chromosomal aneuploidy also may report sex chromosomal abnormalities and microdeletions if requested by a provider [12]. From a laboratory perspective, microdeletion data is processed alongside aneuploidy information and does not require additional steps for sample processing. This background underlies our assumption that the incremental cost associated with reporting microdeletions in addition to aneuploidy information was zero. However, to account for labs attributing differing amounts of the incremental cost for microdeletion reporting, this incremental cost was varied in sensitivity analysis to determine at which value microdeletion screening was cost-effective above the baseline cost of cfDNA for aneuploidies (Figure 2). We estimated the costs for genetic counseling, amniocentesis, termination, and fetal loss based on an outcome analysis of prenatal screening for Down syndrome in women under 35 years of age [27]. The costs for a 20-week anatomic ultrasound and neonatal death were derived from Medicare provider payment data and population-based studies, respectively [34,35]. The cost of genetic counseling is applied to all women with a positive cfDNA screen result, including true positives and false positives. The cost of US is included for all women in the US strategy and women in the microdeletions strategy, except those who experience an amniocentesis-related loss or undergo elective termination of pregnancy. Furthermore, we included the cost of postnatal diagnosis using chromosomal microarray in those with a positive cfDNA test or abnormal ultrasound findings who declined prenatal diagnostic testing [36].

Figure 2.

Univariate sensitivity analysis. The horizontal axis displays the cost of cfDNA for microdeletions in US dollars. The vertical axis displays the incremental cost–effectiveness ratio ($/QALY).

The lifetime medical costs associated with 15q11–q13, 5p15.3, and 1p36 deletion syndromes were calculated by dividing costs of microdeletion syndromes during the first 2 years of life by the cost of care for those with DS during the first 2 years of life, the total of which was multiplied by the lifetime cost of DS [27,37]. The medical cost for 1p36 deletion syndrome was assumed to be the same as PWA. The cost associated with 22q11.2 was calculated by subtracting the medical costs of DS during the first 2 years of life from the cost of 22q11.2 during the first 2 years, the total of which was added to the lifetime cost of DS. The medical costs associated with the individual microdeletion syndromes during the first 2 years of life were calculated based on medical care claims associated with ICD-9 codes in a retrospective cohort of individuals born between 2008–2011 who had Medicaid insurance in Pennsylvania [37].

We considered only maternal quality-adjusted life years (QALYs) in our analysis. QALYs were calculated by applying utilities – quality-of-life measurements – to maternal life expectancy and discounted at a 3% rate per year. The utilities for in-utero fetal loss and termination were 0.93 and 0.91, respectively, and were applied for 2 years with the remainder of the maternal lifetime unaffected [38]. Utilities associated with the microdeletion syndromes were not available in the literature. In these cases, utilities were estimated using the most comparable conditions. For instance, the maternal utilities associated with having a neonate affected by 22q11.2 or 15q11–13 deletions was assumed to the same as having a neonate affected by DS at 0.81 [38]. Similarly, the utility for severe DS phenotype (0.65) was used as a proxy for neonates affected by 5p15.3 or 1p36 deletion syndrome [39].

Lastly, the utility for neonatal death was 0.76 and the utility for having healthy neonate was assumed to be 1 [38,39].

We calculated the clinical outcomes, including the number of cases affected by microdeletion syndromes, in-utero fetal losses, terminations and neonatal deaths for both screening strategies. We then calculated total costs and QALYs to determine the incremental cost-effectiveness ratio (ICER) of cfDNA screening for microdeletion syndromes. This analysis was conducted from a societal perspective. We considered an ICER of $100,000/QALY or less to be cost-effective.

Univariate sensitivity analysis was performed on probabilities, costs, and utilities to evaluate the robustness of our results. We also created a tornado diagram to determine which variables had the greatest effect on model outcomes and performed threshold analysis to calculate the value over which the outcome remained cost-effective. We performed a Monte Carlo simulation with 10,000 trials to analyze multivariable changes in probability and cost distributions. For probability inputs, a beta distribution was used to approximate the normal distribution. Costs were assumed to have a gamma distribution, which has a left skew to account for outliers in upper medical cost ranges. When standard deviations of probabilities were not available in the literature, we calculated conservative estimates and considered ranges that were larger than expected to see the minimum and maximum extremes.

Results

In our theoretical cohort of four million women who underwent prenatal screening with cfDNA for microdeletions, there would be 83 fewer births affected by microdeletion syndromes than a strategy using standard cfDNA for aneuploidy plus ultrasound (Table 2). Routine use of cfDNA screening with microdeletions resulted in 148 more procedure-related losses of normal fetuses. We also observed 200 fewer second and third-trimester losses and 75 fewer neonatal deaths in this cohort. Overall, cfDNA for aneuploidy and micro-deletions would improve effectiveness by 977 QALYs and decrease cost by $90.9 million in comparison to cfDNA for aneuploidy in conjunction with ultrasound.

Table 2.

Outcomes associated with prenatal testing using cfDNA with aneuploidy and microdeletions versus cfDNA for aneuploidy alone with ultrasound in a theoretical population of 4,000,000 women.

	cfDNA for aneuploidy and microdeletions	cfDNA for aneuploidy only, plus US	Differencea
Neonates affected by microdeletion syndromes	252	335	−83
Amniocentesis-related losses	152	4	148
Terminations	805	450	355
Spontaneous abortions in 2nd/3rd trimester	20,327	20,527	−200
Neonatal deaths	8,783	8,858	−75
Cost	$9,207,462,943	$9,298,454,727	−$90,991,784
QALYs	107,950,761	107,949,784	977
Incremental Cost-Effectiveness Ratio	Dominant	Dominated	–

^acfDNA for aneuploidy and microdeletions minus cfDNA for aneuploidy plus ultrasound.

Univariate sensitivity analyses were performed using a wide range of model inputs. We found that the largest driver of the model included the cost of cfDNA, which was incorporated in both screening strategies. As such, we varied the incremental cost of microdeletions over cfDNA for aneuploidy alone. With the cost-effectiveness threshold set at $100,000 per QALY, microdeletion screening is cost-effective until an incremental cost of $47.10 (Figure 2). When we varied the specificity of the screening test, we found that it remained cost-effective down to a specificity of 91%. Microdeletion screening was cost-effective for elective termination rates all the way down to 27% for 1p36, 51% for Cri-du-chat, 15% for 22q11.2, and 29% for PWA. Multivariate sensitivity analysis was also conducted to evaluate the robustness of the assumptions in our model (Figure 3). We simulated 10,000 trials in a Monte Carlo analysis to simultaneously vary probabilities, costs, and utilities. Cell-free fetal DNA screening with microdeletions was found to be cost-effective in 92.8% of the trials.

Figure 3.

Multivariate sensitivity analysis. This Monte Carlo simulation displays the outcomes of the 10,000 trials. Each dot indicates the result of a single trial. The dashed line represents a willingness-to-pay threshold equal to $100,000. The ellipse displays the 95% confidence ellipse. cfDNA screening for microdeletion syndromes was the cost-effective strategy in 92.8% of the trials.

Discussion

Our model suggests that routinely screening pregnant women for microdeletion syndromes using cfDNA is cost-effective as compared to standard cfDNA for aneuploidy plus ultrasound. That is, if a patient has decided to undergo cfDNA testing, according to our model, there are both outcome and cost benefits to screening for common microdeletion syndromes in addition to aneuploidy. This strategy resulted in 83 fewer cases of microdeletion syndromes due to increased diagnosis and termination of affected fetuses following diagnostic confirmation, although procedure-related losses were also significantly greater for this strategy. Screening with cfDNA for microdeletions led to more positive screens and, subsequently, more diagnostic amniocenteses compared to cfDNA for aneuploidy alone and ultrasound.

The outcome of greater procedure-related losses in the strategy including microdeletion screening can be explained, in part, by the false-positive rate (FPR) and positive predictive (PPV) value that results from screening for conditions with such low prevalence. Clinical data on PPV is limited because cfDNA screening for microdeletions has not been validated in large populations. One study following-up on positive cfDNA screens reported that the PPV for microdeletion syndromes ranged from 0–21% [15]. Screening for microdeletions with cfDNA may be most meaningful in those who are at high risk for chromosomal anomalies due to a previously affected pregnancy or abnormal serum and ultrasound findings. Although cfDNA screening for microdeletions was cost-effective and increased prenatal detection in our model, it is important to counsel women on the clinical utility of such screening, including limitations of false-positive results leading to unnecessary diagnostic procedures and potential fetal loss. Additionally, it is important to note that it has only been shown to be cost-effective in women who are already undergoing cfDNA screening for aneuploidy.

Laboratories differ in sample processing methodology and bioinformatic analysis of cfDNA for aneuploidy, sex chromosome abnormalities, and microdeletions. Furthermore, detection of copy number variants depends on a variety of factors, including fetal fraction, size of copy number variant, and depth of coverage [12]. Massively parallel shotgun sequencing (MPSS) is a technique that amplifies and sequences cfDNA fragments [10,12]. Although MPSS can lead to greater sequencing depth, it comes at the expense of increased false-positive rate, false-negative rate, and possibly increased costs. An alternative technique involves sequencing select single-nucleotide polymorphisms on genes of interest, such as the cfDNA screen used in our model [24]. SNP-based analyses limit amplification to certain genetic regions, therefore it can only identify a proportion of individuals associated microdeletion syndromes. Validation studies of cfDNA for microdeletions have multiple methodological limitations, the most significant of which is the lack of follow-up of clinical outcomes [1,2,24]. As such, the high sensitivities reported in the internal validation studies may be overshadowed by the low PPV in the general obstetric population and a positive screening result should be evaluated with caution. For these reasons, it is important to note that the results of microdeletion screening with cfDNA require confirmation with diagnostic testing in all cases and counseling should be individualized for all women.

The currently available literature on the cost-effectiveness of cfDNA screening focuses only on aneuploidies, specifically excluding microdeletions. A prior decision analysis comparing traditional multiple marker screening versus cfDNA found the traditional multiple marker approach as first-line screening to be cost-effective and to lead to the highest level of detection of significant chromosomal abnormalities including CNVs [40]. This study, however, did not include microdeletions as part of cfDNA screening.

Given the rapid expansion of cfDNA screening beyond common autosomal aneuploidies to include sex-chromosomal differences, rare aneuploidies, and CNVs, there is an immediate need to weigh the costs and benefits of cfDNA as a screening tool for the detection of microdeletions. Our model is the first to evaluate the cost-effectiveness of cfDNA screening in the setting of microdeletion syndromes.

The reliability of any cost-effective model relies on the strength of probability, cost, and utility estimates available in the literature. Although there was limited data supporting some of these estimates for microdeletions, the univariate analysis indicated our model was robust even when model inputs were varied significantly. The cost associated with standard cfDNA was the main input driving our model. Since the current cost for this new technology remains high and applied to all women in our model, it was expected that this cost would have a large impact. When we varied the cost of microdeletion screening in sensitivity analysis, cfDNA screening for microdeletions was cost-effective at a standard of less than $100,000 per QALY up until an incremental cost of $47.10 over the cost of cfDNA for aneuploidy alone. This cost represents the incremental laboratory fee of processing microdeletion information in addition to aneuploidy on cfDNA testing; we acknowledge that there is variability in the costs of noninvasive prenatal testing from the laboratory, payer, and out-of-pocket patient perspectives, and these costs may not be transparent [33]. Because microdeletions can be added-on to most cfDNA tests for aneuploidy, there is the minimal incremental cost of screening for microdeletions from a societal perspective. In multivariate analysis, our model was robust even when inputs were varied significantly across a plausible range. However, additional studies with more specific data on microdeletion syndrome characteristics are needed to strengthen our confidence in this analysis.

As with any theoretical model, we are limited in the ability to capture the spectrum of complexities in clinical medicine. Due to the lack of clinical data on probabilities, costs, and utilities associated with specific microdeletion syndromes, we had to extrapolate some of our assumptions using available estimates of the most comparable conditions. Hence, it difficult to predict the true impact of microdeletion screening on prenatal decision making. The current research on cfDNA screening for microdeletions is limited by small sample size, explained by the low baseline prevalence of microdeletion syndromes. To overcome this limitation, we derived a combined sensitivity for the five microdeletion syndromes from two studies using SNP based analysis for screening of microdeletions with cfDNA. While one internal validation study used maternal plasma and DNA mixture samples, the majority of the test performance data was derived from a retrospective cohort study on maternal serum samples [1,24]. The cfDNA screen for microdeletions was based on SNP analysis, meaning that amplification was limited to certain genomic regions and it does not cover all mutations as compared to sequencing; thus, the cfDNA for microdeletions screen used in our model can only identify a proportion of individuals with microdeletion syndromes. Furthermore, pregnancy outcomes were only available for a proportion of those screened by cfDNA for microdeletions, which may have led to an overestimation of the test sensitivity and specificity. To account for this limitation, we assumed that all positive tests with unknown outcomes were false positives. Also, we set the incremental cost of cfDNA microdeletion screening to zero at baseline based on the assumption that microdeletion results are available on all cfDNA testing. As observed by minimal transparency, commercial laboratories may vary the price for this additional information, but the true cost may only be the additional counseling and testing for those who screen positive. Lastly, we acknowledge that our model accounts only for the microdeletion syndromes considered, and does not include other genetic abnormalities such as microduplications, uniparental disomy, or singlegene disorders.

While our findings demonstrate that cfDNA screening for microdeletions is cost-effective when compared to cfDNA for aneuploidy alone in conjunction with ultrasound, it is still not currently recommended for routine screening by ACOG and SMFM due to the relatively low prevalence of these conditions and therefore poor PPV. The test characteristics of cfDNA with microdeletions require greater examination and validation studies in large clinical trials before routine use is recommended.