Effect of Cell-Free DNA Screening vs Direct Invasive Diagnosis on Miscarriage Rates in Women With Pregnancies at High Risk of Trisomy 21: A Randomized Clinical Trial

Key Points

Question  Does performing cell-free DNA testing on women with pregnancies at high risk of trisomy 21, followed by invasive testing only if cell-free DNA test results are positive, reduce the rate of miscarriage compared with immediate invasive testing?

Findings  In this randomized clinical trial that included 2111 women with pregnancies with a risk of trisomy 21 greater than 1 in 250 following combined first-trimester screening, the miscarriage rate was 0.8% in the cell-free DNA group and 0.8% in the invasive procedures group, a difference that was not statistically significant.

Meaning  Using cell-free DNA to stratify individuals for invasive testing compared with direct invasive testing was not associated with a significant reduction in the rate of miscarriage in women with pregnancies at high risk of trisomy 21.

Importance  Cell-free DNA (cfDNA) tests are increasingly being offered to women in the first trimester of pregnancies at a high risk of trisomy 21 to decrease the number of required invasive fetal karyotyping procedures and their associated miscarriages. The effect of this strategy has not been evaluated.

Objective  To compare the rates of miscarriage following invasive procedures only in the case of positive cfDNA test results vs immediate invasive testing procedures (amniocentesis or chorionic villus sampling) in women with pregnancies at high risk of trisomy 21 as identified by first-trimester combined screening.

Design, Setting, and Participants  Randomized clinical trial conducted from April 8, 2014, to April 7, 2016, in 57 centers in France among 2111 women with pregnancies with a risk of trisomy 21 between 1 in 5 and 1 in 250 following combined first-trimester screening.

Interventions  Patients were randomized to receive either cfDNA testing followed by invasive testing procedures only when cfDNA tests results were positive (n = 1034) or to receive immediate invasive testing procedures (n = 1017). The cfDNA testing was performed using an in-house validated method based on next-generation sequencing.

Main Outcomes and Measures  The primary outcome was number of miscarriages before 24 weeks’ gestation. Secondary outcomes included cfDNA testing detection rate for trisomy 21. The primary outcome underwent 1-sided testing; secondary outcomes underwent 2-sided testing.

Results  Among 2051 women who were randomized and analyzed (mean age, 36.3 [SD, 5.0] years), 1997 (97.4%) completed the trial. The miscarriage rate was not significantly different between groups at 8 (0.8%) vs 8 (0.8%), for a risk difference of −0.03% (1-sided 95% CI, −0.68% to ∞; P = .47). The cfDNA detection rate for trisomy 21 was 100% (95% CI, 87.2%-100%).

Conclusions and Relevance  Among women with pregnancies at high risk of trisomy 21, offering cfDNA screening, followed by invasive testing if cfDNA test results were positive, compared with invasive testing procedures alone, did not result in a significant reduction in miscarriage before 24 weeks. The study may have been underpowered to detect clinically important differences in miscarriage rates.

Trial Registration  ClinicalTrials.gov Identifier: NCT02127515

Quiz Ref IDPrenatal screening for Down syndrome has evolved considerably over the last 2 decades. Combined first-trimester screening, which is based on maternal age, fetal nuchal translucency measurement, and maternal serum screening, has been adopted in most countries, yielding a 90% detection rate at a 5% false-positive rate.¹ Fetal karyotyping by amniocentesis or chorionic villus sampling (CVS) is offered when risk of trisomy 21 surpasses a predefined cutoff value, thus defining a patient as at high risk.

Cell-free DNA (cfDNA)–based analyses of maternal plasma have a sensitivity of more than 99% for Down syndrome in pregnancies at high risk of this syndrome^2,3 without the associated risks of an invasive procedure.^3-8 Safety has been the most widely used argument for implementing cfDNA testing in high-risk patients without evidence of its superiority over immediate use of invasive testing.⁹ This theoretical advantage of cfDNA testing prior to any invasive testing over immediate use of amniocentesis or CVS for karyotyping has not been tested in a randomized trial and should be balanced against the lower sensitivity and specificity of cfDNA in the detection of trisomy 21.² In addition, cfDNA testing may overlook other common aneuploidies and structural chromosomal anomalies that are detected by conventional karyotyping following invasive testing.¹⁰

This randomized clinical trial compared rates of miscarriage after either cfDNA testing with invasive procedures only when cfDNA test results were positive vs immediate invasive testing procedures in pregnancies at high risk of trisomy 21 as identified by first-trimester combined screening.

The study protocol and the statistical analysis plan are provided in Supplement 1 and Supplement 2. Quiz Ref IDThis study was a 2-year, nationwide, open-label randomized clinical trial conducted in 64 centers throughout France. It compared the miscarriage rates following either cfDNA testing (followed by invasive testing if cfDNA test results were positive) or direct fetal karyotyping by invasive testing procedures in women with pregnancies with a risk of trisomy 21 between 1 in 5 and 1 in 250 following combined first-trimester screening. Ethics committee approval was obtained and funding was received from the French Ministry of Health. Eligible women who provided consent to take part were registered in a secured database (https://www.bionuqual.org/echo.php).

As part of routine practice in France, all women were informed about Down syndrome screening and were able to undergo first-trimester combined screening or sequential combined screening as part of their routine antenatal care.¹¹ Women with a risk above 1 in 250 were considered screen positive.^12,13 Practitioners performing nuchal translucency measurement as part of prenatal screening for trisomy 21 were licensed for nuchal translucency measurement and were registered in a perinatal network, which ensured continuous quality control monitoring. Biochemical assays and risk calculations were performed in 82 accredited laboratories that adhere to a nationwide quality policy.^12,13

All women with a pregnancy with a risk of Down syndrome between 1 in 5 and 1 in 250 were offered to enroll in this study at one of the participating prenatal diagnostic centers. Additional inclusion criteria were maternal age at least 18 years, singleton pregnancy, health care coverage by the national health care insurance system (Assurance Maladie), and gestational age between 11 and 18 weeks. Eligible women agreed a priori to fetal karyotyping by CVS or amniocentesis and provided written informed consent. Exclusion criteria were nuchal translucency greater than 3 mm and maternal pregnancy-associated plasma protein A or β human chorionic gonadotropin concentrations less than 0.3 or greater than 5 multiples of the median, as these characteristics have been associated with other chromosomal abnormalities. Cases with a fetal malformation found at the first ultrasound examination, a vanishing twin, or prior knowledge of any parental balanced translocation were also excluded.

Prior to patient inclusion in the study, face-to-face standardized information, counseling sessions, and printed information leaflets were provided to patients by health care professionals in each participating center. This content presented the benefits and limitations of both tests. This included the potential of the invasive test to find chromosomal abnormalities that can be overlooked by cfDNA testing, as well as the 1% additional risk of miscarriage associated with invasive procedures. Women’s preferences regarding cfDNA and invasive testing were recorded on eligibility (details will be described in a subsequent study).

Randomization to either cfDNA testing followed by invasive testing procedures only when cfDNA test results were positive or to direct use of invasive testing procedures was performed with a 1:1 ratio using the same centralized, secure web system (https://www.bionuqual.org/echo.php). The randomization sequence was created by an independent data manager using a list of computer-generated random numbers (random block sizes of 2, 4, and 6) and was stratified according to each center. Patients, investigators, and all staff involved in the trial were aware of group allocation.

Services associated with this study were provided at no cost to participants. For women in the cfDNA group, blood samples were drawn in each participating center, identified by a study code, and shipped to the referral cytogenetic laboratory at Necker–Enfants Malades Hospital in Paris. The cfDNA technique as well as the interpretation and reporting of the results are described in the eAppendix in Supplement 3. Positive cfDNA test results were always confirmed with conventional invasive testing before management decisions were undertaken. Women in the invasive prenatal diagnosis group had karyotyping performed by one of the local accredited cytogenetics laboratories.¹³

Genomic DNA was extracted from 5 mL of plasma and used for library preparation with end-repair, A-tailing, and adaptor ligation with a specific tag. Thereafter, sequencing of a pool of 11 patients with 1 case of trisomy 21 (control sample) was performed with a HiSeq 1500 (Illumina) using a rapid run for single reads of 50 base lengths. After demultiplexing, read counts per chromosome and z score computations were performed using the R RAPIDR package, version 0.1.117.¹⁴ The z score is the difference in the number of reads of chromosome 21 in the test and reference set divided by the standard deviation of the number of reads. A result was considered positive when the z score was above +1.645.

The technique including sequencing and bioinformatics tools as well as interpretation and rendering of cfDNA results are detailed in the eAppendix in Supplement 3.

Quiz Ref IDCell-free DNA test results and cytogenetic results from invasive testing were collected for participants according to randomized group. Pregnancy outcomes were collected and categorized as follows: miscarriage before 24 weeks’ gestation, intrauterine fetal death (ie, spontaneous death of a fetus after 24 weeks’ gestation and before delivery), termination of pregnancy, and live birth and perinatal death (restricted to intrapartum stillbirth and neonatal death before 6 days). The primary outcome was miscarriage before 24 weeks’ gestation. This is the outcome commonly used in articles on which the sample size calculation was based, as detailed in the trial protocol^15,16 and in recent review articles on this issue.^17,18 In the initial protocol, a less well-defined term of “pregnancy losses” was used. However, 24 weeks’ gestation is the current definition for viability.¹⁹ Thus, at this threshold, “extremely premature birth” is more appropriate than “pregnancy loss.” Although the French College of Gynecologists and Obstetricians recently attempted to better define miscarriage and issued practical guidelines,²⁰ there is a lack of an internationally accepted set of definitions for many terms used to describe pregnancy losses and also a lack of a standardized French-English reciprocal terminology or glossary.

Secondary outcomes included number and percentage of invasive procedures performed in both treatment groups, performance characteristics of cfDNA testing (including failed tests and evaluation using sensitivity and specificity compared with the reference standard karyotype or phenotype at birth), time interval between blood sample receipt and result availability, and diagnosis of chromosomal abnormalities in each group. The false-positive rate of cfDNA testing was estimated post hoc if a conventional cutoff z-score value of 3.0 had been used.² Other secondary outcomes not reported in this article include the relationship between women’s clinical characteristics and cfDNA results, women’s preferences regarding cfDNA and invasive testing, and anxiety. The protocol also prespecified an economic analysis. At the time the protocol was written in 2012, there was a scarcity of data on this topic and the cfDNA test was much more expensive. However, we used an in-house validated method based on next-generation sequencing to remain independent from private test providers. In a cost assessment, this method would compare very unfavorably with the more recent, lower-cost tests (with the introduction of automation technologies). More importantly, various strategies for the implementation of cfDNA testing, including contingent use, have now been evaluated and seem more appropriate in terms of cost effectiveness,^21,22 making the planned economic analysis less relevant.

Other pregnancy outcomes (intrauterine fetal death, termination of pregnancy, and live birth and perinatal death) were analyzed as post hoc outcomes.

The sample size was calculated hypothesizing a reduction in the number of miscarriages using cfDNA testing. To observe a decrease in miscarriage from 1.5% to 0.5% between the invasive and cfDNA testing groups, respectively, with a 1-sided α=.05 and 80% power, 1250 patients per group were required, yielding a total sample size of 2500 patients. These assumptions were derived from the only existing randomized clinical trial, which found an additional 1% miscarriage rate following invasive testing.¹⁶

All statistical analyses were performed using R software version 2.11.1, with the statistical analysis plan prespecified before locking the database. The primary outcome underwent 1-sided testing; secondary outcomes underwent 2-sided testing. For all tests used, P<.05 was considered statistically significant. The baseline characteristics of the 2 treatment groups were described as means and standard deviations or medians and interquartile ranges (IQRs) for quantitative variables and frequencies and percentages for qualitative variables.

For all analyses, patients without appropriate consent forms were excluded. Regarding the analysis of the primary outcome, the primary analysis set was defined as all patients randomized excluding those with a missing outcome. The per-protocol population was defined as patients from the primary analysis set analyzed according to the first procedure actually performed (cfDNA or invasive testing) and excluded those who underwent neither cfDNA nor invasive testing. A post hoc analysis was performed analyzing patients according to whether they underwent an invasive testing procedure, whatever their allocation group (Figure).

Miscarriage rates (primary outcome) and intrauterine fetal death rates (post hoc outcome) were compared between groups using a 1-sided χ² test on the primary analysis set.

Secondary end points were mainly descriptive; if not, they were compared between groups using 2-sided tests: the χ² test (or the Fisher exact test as appropriate) was used for comparison of qualitative outcomes and the t test (or the Wilcoxon test for nonnormally distributed variables) was used for comparison of quantitative outcomes. Analyses of secondary end points were not adjusted for multiple comparisons and should be interpreted as exploratory.

A binomial generalized linear model with identity link was used to calculate the confidence intervals of the risk differences (cfDNA testing group minus invasive testing group). For the performance of cfDNA testing, 95% confidence intervals were calculated using binomial distribution.

Post hoc sensitivity analyses were performed to investigate the robustness of the results for the primary end point. A generalized linear model with random center effect was computed to handle the multiple-sites design.²³ Multiple imputations by chained equations based on 2 different methods (random sample from observed values and logistic regression including randomization group, age, body mass index, and risk of trisomy 21 as covariates) were tested to handle missing data.²⁴ In addition, a tipping-point analysis was carried out to evaluate the robustness of the results for the primary outcome by examining how it would be modified by different scenarios of missing data replacement.²⁵

A total of 2592 women were assessed for eligibility at 57 centers between April 2014 and April 2016 (7 centers recruited no participants for the study). Among these women, 481 (18.6%) declined to participate prior to randomization, of whom 448 stated a preference for either cfDNA or invasive testing. A total of 2111 women were therefore randomized, 1049 to the cfDNA testing group and 1062 to the direct invasive testing group. After exclusion of patients without properly completed consent forms, 1034 patients were eligible for cfDNA testing and 1017 for invasive testing. Fifty-four women (2.6%) were lost to follow-up (19 [1.8%] in the cfDNA testing group and 35 [3.4%] in the invasive testing group). Participant flow through the trial is shown in the Figure. Both groups had similar demographics and screening results for trisomy 21 (Table 1).

There was no significant difference in miscarriage rates between the cfDNA and invasive testing groups (8 [0.8%] vs 8 [0.8%] miscarriages; risk difference, −0.03%; 1-sided 95% CI, −0.68% to ∞; P = .47) (Table 2). Following cfDNA and invasive testing, miscarriages occurred at a median of 19.9 (IQR, 16.9-21.1) weeks’ gestation and 19.9 (IQR, 18.8-22) weeks’ gestation, respectively.

One woman randomized to the invasive testing group who experienced a miscarriage had refused to undergo invasive testing. When analyzing the data per protocol, there was no significant difference in miscarriage rates between cfDNA testing (n = 1103) and invasive testing (n = 733) (8 [0.7%] vs 7 [1%] miscarriages; risk difference, −0.23%; 1-sided 95% CI, −0.95% to ∞; P = .30). No patient had a miscarriage following invasive testing performed after a positive cfDNA test result.

There were 84 (8.3%) and 751 (76.5%) invasive procedures performed in the cfDNA and invasive testing groups, respectively.

Cell-free DNA testing was successful in 984 of the 1028 women (95.7%) who were randomized to the cfDNA testing group and received cfDNA testing as randomized. Forty-four samples (4.3%) failed quality checks (hemolysis of plasma samples; specimen received after 5 days). Three women with failed first tests declined to provide a second blood sample and 38 of 41 repeat samples were satisfactory, leading to successful cfDNA testing, including repeat testing, in 1022 (99.4%) of 1028 women.

Performance of the cfDNA test was estimated for the 984 samples with satisfactory quality on first check. The detection rate (sensitivity) of cfDNA testing for trisomy 21 was 100% (95% CI, 87.2%-100%) (27/27) with a false-positive rate of 5.6% (95% CI, 4.2%-7.2%) (55/984) for the chosen z-score threshold (z = 1.645). The mean z score was 12.9 (SD, 5.6) in true-positive cases. In false-positives cases, all z scores were below 3 except for 2 cases with z scores of 3.5 and 14. In the latter case, this high z-score value was confirmed with another sample taken 3 months later, while amniocentesis ruled out trisomy 21. This indicated a possible confined placental mosaicism, although this could not be confirmed at birth. Had a z-score cutoff value of 3.0 been used, the false-positive rate would have been 0.2% (95% CI, 0.03%-0.7%).

The mean time interval between blood sample receipt and result availability was 13.0 (SD, 5.0) days and was less than 3 weeks for 903 women (88%) who underwent cfDNA testing.

Chromosomal anomalies identified in both groups are summarized in Table 2 and Table 3. Overall, there were 28 (2.8%) and 49 (5%) chromosomal abnormalities diagnosed in the cfDNA and invasive testing groups, respectively (risk difference, −2.23%; 95% CI, −3.93% to −0.54%; P = .01). These included 1 (0.1%) and 11 (1.1%) anomalies other than trisomy 21, respectively (risk difference, −1.02%; 95% CI, −1.71% to −0.34%; P = .003) (Table 2). In the cfDNA group, the anomaly was a tetrasomy 12p found on amniocentesis after ultrasound revealed multiple malformations at mid trimester. The anomalies in the invasive testing group are summarized in Table 3 and described in detail in the eTable in Supplement 3.

The post hoc analysis among patients analyzed according to whether they underwent an invasive procedure also showed no significant difference in miscarriage rates (9/1178 [0.8%] vs 7/819 [0.9%]; risk difference, −0.09%; 1-sided 95% CI, −0.76% to ∞; P = .41).

A binomial generalized linear model with identity link including site as a random effect did not fit. When modeling logit link, the standard error of the β coefficient including center as a random effect was very close to the one estimated by the model without center as a random effect (SE[β] = 0.50205 vs 0.50199 without site effect). Multiple imputations of missing data based on random sampling from observed values and multiple logistic regression led to comparable results (risk differences, −0.02% [1-sided 95% CI, −0.68% to ∞; P = .47] and −0.03% [1-sided 95% CI, −0.67% to ∞; P = .48], respectively). Results from a tipping-point analysis are presented in the eFigure in Supplement 3.

Following cfDNA and invasive testing, respectively, there were 9 (0.9%) and 9 (0.9%) intrauterine fetal deaths (risk difference, −0.03%; 1-sided 95% CI, −0.72% to ∞; P = .47) and 3 (0.3%) and 2 (0.2%) perinatal deaths (risk difference, 0.09%; 95% CI, −0.35% to 0.53%; P>.99), and 30 (3%) and 39 (4%) of women underwent termination of pregnancy (risk difference, −1.02%; 95% CI, −2.62% to 0.59%; P = .21) (Table 2).

There was no significant difference in overall pregnancy loss rates between the cfDNA and invasive testing groups (17 [1.7%] vs 17 [1.7%]; risk difference, −0.06%; 1-sided 95% CI, −1.01% to ∞; P = .46) (Table 2). Pregnancy losses occurred at a median gestational age of 21.3 (IQR, 18-23.1) weeks and 20 (IQR, 17.1-22) weeks, respectively.

When analyzing the data per protocol, there was no significant difference in pregnancy loss rates between cfDNA testing (n = 1103) and invasive testing (n = 733) (17 [1.5%] vs 14 [1.9%], respectively; risk difference, −0.37%; 1-sided 95% CI, −1.40% to ∞; P = .27). This remained true in the post hoc analysis between the cfDNA testing (n = 1178) and invasive testing (n = 819) (20 [1.7%] vs 14 [1.7%], respectively; risk difference, −0.01%; 1-sided 95% CI, −0.98% to ∞; P = .49).

Quiz Ref IDIn this randomized clinical trial, there was no significant decrease in the risk of miscarriage before 24 weeks in high-risk pregnancies after cfDNA testing followed by invasive testing only in women with a positive cfDNA test result vs immediate invasive testing. The multicenter design of this study included invasive procedures performed in 57 different centers; this makes the result more generally applicable. To our knowledge, this is the first randomized clinical trial comparing miscarriage rates following cfDNA and invasive testing in women with pregnancies at high risk of trisomy 21 by combined screening in the first trimester of pregnancy.

The total miscarriage rate in the invasive procedure group was 0.8%, significantly lower than the commonly expected total rate of 1.5%, which includes a 1% rate of procedure-related miscarriage. This rate, which was also used for the sample size estimation, originated from a randomized clinical trial performed more than 30 years ago.¹⁶ These data have been challenged by a recent study and a meta-analysis that concluded that the actual risks of miscarriage following CVS or amniocentesis are approximately 0.2% and 0.1%, respectively,^17,18 and neither CVS nor amniocentesis was associated with increased risk of miscarriage in a cohort of more than 147 000 pregnancies.¹⁸ In contrast, the miscarriage rate in the cfDNA group was higher than the anticipated rate of 0.5%. Although not statistically significant, this could be related to the high-risk status of the pregnancies.

An important secondary outcome was the detection of a significant number of non–trisomy 21 chromosomal abnormalities in the invasive testing group. It is well established that combined first-trimester screening yields chromosomal abnormalities other than trisomy 21 even after excluding increased nuchal translucency, as in this trial.^26,27 In the present study, among 751 women, there were 11 (1.5%) chromosomal abnormalities identified by invasive testing that would not have been found by cfDNA, including several pathological cases or cases requiring dedicated follow-up and management (Table 3). Given the randomized nature of the study, a similar incidence of abnormalities in the cfDNA group would be expected, although only 1 was detected following a finding of fetal anomaly on ultrasound.

Quiz Ref IDThis study did not show a reduction of miscarriage or intrauterine fetal death associated with use of cfDNA in high-risk cases. However, there were far fewer invasive procedures in the cfDNA group (81 vs 751 in the immediate invasive testing group), and invasive procedures have been associated with extremely rare but severe outcomes, such as maternal septicemia and death.^28-30 None of these occurred in this randomized clinical trial. If safety is no longer an argument for cfDNA testing over karyotyping, then the risk of overlooking rare chromosomal anomalies in high-risk pregnancies should be carefully evaluated. In this study, standard karyotyping identified other chromosomal abnormalities in 1.5% of high-risk pregnancies. The use of chromosomal microarray analysis, which has now become the first-line genetic test on amniotic fluid or chorionic villi in many laboratories, could have increased this rate of discovery of abnormalities even further.^10,31,32 Indeed, the additional value of chromosomal microarray analysis compared with karyotyping was reported to be as high as 6% when there are fetal anomalies on ultrasound and in 0.5% to 2.5% of cases in pregnancies with advanced maternal age or at high risk following first-trimester combined screening.^27,31,33,34

This study has several limitations. First, although the invasive testing rate was considerably less following cfDNA testing, the observed 5.6% false-positive rate was higher than the average 0.1% reported.^2,35 This higher false-positive rate was deemed acceptable to improve cfDNA sensitivity since this was the first randomized trial evaluating this test. Although this induced more secondary invasive procedures and could therefore have increased the miscarriage rate in this group, no patient undergoing invasive testing following a positive cfDNA test result had a miscarriage. The exploratory post hoc analysis also suggested no additional miscarriage or pregnancy loss risks in women who underwent invasive procedures. Second, in this study, 17% of women declined randomization and up to 24% of women initially randomized to invasive testing refused their allocated test. This is a potential limitation because this participant attrition occurred after counseling about a 0.5% to 1% risk of procedure-related pregnancy loss. However, higher dropout rates in women assigned to invasive testing may have reflected an attempt to benefit from cfDNA testing free of charge through the study. Indeed, cfDNA has become widely available through private laboratories at a substantial cost. This prompted national health insurances to consider various risk cutoffs for offering second-line cfDNA testing following combined first-trimester screening.^9,36 This dropout rate could also reflect a better acceptance of cfDNA testing vs invasive procedures. Third, cfDNA testing was only performed for trisomy 21, while the potential for cfDNA to rule out other chromosomal anomalies is likely to improve in the near future. Fourth, the assessment of clinical value of some of the chromosomal anomalies detected in the invasive group could be controversial. Nevertheless, it is difficult to discuss what chromosomal anomaly may modify pregnancy management, as this may also vary according to prenatal counseling, ultrasound findings, parental preferences, and local laws. Fifth, this randomized clinical trial showed no statistical difference in the risk of miscarriage before 24 weeks with more than 2000 women with pregnancies at high risk of Down syndrome randomized. However, the study may have been underpowered to identify potentially clinically important reductions in miscarriage with cfDNA testing. Sixth, it was not possible to conduct the planned economic analysis. Because of this, the cost implications of the study findings could not be estimated. Seventh, because of the potential for type 1 error due to multiple comparisons, the analyses of the secondary end points should be considered exploratory.

Among women with pregnancies at high risk of trisomy 21, offering cfDNA screening followed by invasive testing if cfDNA test results were positive, compared with direct invasive testing, did not result in a significant reduction in miscarriage before 24 weeks. The study may have been underpowered to detect clinically important differences in miscarriage rates.

Corresponding Author: Laurent J. Salomon, MD, PhD, Hôpital Necker–Enfants Malades, Assistance Publique-Hôpitaux de Paris (AP-HP), Université Paris Descartes, 149 rue de Sèvres, 75743 Paris, Cedex 15, France (laurentsalomon@gmail.com).

Accepted for Publication: June 13, 2018.

Author Contributions: Dr Salomon had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Concept and design: Malan, Bussières, Elie, Fries, Vekemans, Ville, Salomon.

Acquisition, analysis, or interpretation of data: Malan, Bussières, Winer, Jais, Baptiste, Le Lorc’h, Elie, O’Gorman, Fries, Houfflin-Debarge, Sentilhes, Ville, Salomon.

Drafting of the manuscript: Malan, Bussières, Jais, Le Lorc’h, Ville, Salomon.

Critical revision of the manuscript for important intellectual content: Malan, Bussières, Winer, Jais, Baptiste, Elie, O’Gorman, Fries, Houfflin-Debarge, Sentilhes, Vekemans, Ville.

Statistical analysis: Baptiste, Elie.

Obtained funding: Malan, Bussières, Winer, Vekemans, Ville, Salomon.

Administrative, technical, or material support: Malan, Bussières, Le Lorc’h, Sentilhes, Vekemans, Ville.

Supervision: Malan, Bussières, Fries, Vekemans, Ville, Salomon.

Conflict of Interest Disclosures: All authors have completed and submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest and none were reported.

Group Information: Members of the Safe 21 Study Group are Emmanuelle Haquet (CHU Montpellier), Romain Favre (CMCO Schiltigheim), Gwenaelle Le Bouar (CHU Rennes), Pierre Mares (CHU Nimes), Edwin Quarello (St Joseph Hospital, Marseille), Hélène Laurichesse (CHU Clermont-Ferrand), Florence Bretelle (La Timone and Nord Hospitals, Marseille), Bérénice Depont (Mathilde Clinic, Rouen), Christophe Vayssière (CHU Toulouse), Laëtitia Trestard (Belvedere Polyclinic, Mont St Aignan), Thibaud Quibel (CHI Poissy-St Germain), Elise Machevin (CH Euroseine, Evreux), Alain Diguet (CHU Rouen), Edouard Kauffmann (CH Sud Réunion, St Pierre), Danièle Vauthier-Brouzes (Pitie-Salpetriere Hospital, Paris), Raphaele Mangione (Bordeaux Nord Hospital), Marianne Fontanges (Jean Villars Clinic, Brugges), Alain Liquier (Bioffice Laboratory, Bordeaux), Martine Marechaud (CHU, Poitiers), Franck Mauviel (CH Verlomme, Toulon), Anne-Sylvie Valat (CH Lens), Dominique D’Hervé (La Sagesse Clinic, Rennes), Florence Chevallier-Helas (CH Le Havre, Montivilliers), Sandrine Reviron (CH Lons le Saulnier), Stéphanie Hardeman (CH d’Elbeuf), Emile Paul Mereb (Manchester Hospital, Charleville-Mézières), Claire Dazel-Salonne (CH Le Mans), Alphonse Kimpamboudi (CH Troyes), André Bongain (CHU Nice), Cyril Huissoud (CHU Croix Rousse, Lyon), Marc Bucher (Tertre Rouge Clinic, Le Mans), Sylvie Capella (Estuaire Hospital, Le Havre), Thierry Rousseau (CH Dijon), Jean Paul Bory (CHU Reims), Jerôme Massardier (CHU Civils Hospital, Bron), Georges Haddad (CH Blois), Thomas Schmitz (Robert Debré Hospital, Paris), Cécile Jacquet (Mutualiste Clinic, Grenoble), Felix Faggianelli (deceased) and Ibrahim Makke (CH Bernay), Robert Mocquard (Champagne Clinic, Troyes), Franck Perrotin (CHU Tours), Michel Aumersier (CH Chalons en Champagne), Véronique Equy (CHU Grenoble), Michel Cingotti (CH Dieppe), Hélène Vervaet (Courlancy Polyclinic, Reims), Joël Le Long (Aubepines Clinic, St Aubin), Abdelhaq Benhaddou (CH Vitry le François), Jean Lambert (CH Chaumont), Didier Fabre (CH Sedan), Marie Gaillard (CH Voiron), Cécile Bakkar (CH Vernon), Elie Azria (St Joseph Hospital, Paris), Renaud Garnier (St Andre Polyclinic, Reims), Romain Molignier (Montaigut Laboratory, Toulouse), François Jacquemard (American Hospital of Paris, Neuilly/Seine), Albert Sultan (CH Moet, Epernay), Jean Marie Delbosc (Espic Hospital, Talence), Marc Schneider (Belledonne Clinic, St Martin d’Hères), Valérie Séror (Inserm, Marseille), Isabelle Durand-Zaleski (Hôtel Dieu Hospital, Paris).

Funding/Support: This work was supported by a grant from the French Ministry of Health.

Role of the Funder/Sponsor: The funder had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; or decision to submit the manuscript for publication.

Additional Contributions: We thank Inès Belaroussi, MD, Hôpital Necker–Enfants Malades, Sonia Bouquillon, MD, CHU de Lille, Bérénice Doray, MD, PhD, CHU de St Denis de la Réunion, Agnès Guichet, MD, CHU d’Angers, Sylvain Goupil, BSc, Hôpital Necker–Enfants Malades, Isabelle Lacroix, MD, CERBA, Laurence Lecomte, MSc, Hôpital Necker–Enfants Malades, Kamran Moradkhani, MD, CHU de Nantes, Sameh Nazih, MSc, Hôpital Necker–Enfants Malades, and Marc Nouchy, MD, Biomnis, for help collecting detailed pregnancy outcomes; Aris Papageorghiou, MD, PhD, Nuffield Department of Women’s and Reproductive Health, University of Oxford, for help editing the manuscript; and Helen Pickersgill, PhD, Life Science Editors, who received compensation for editing a revised version of the manuscript. No other listed contributors received compensation for their role in this study.