Methods
This study was approved by the Human Subjects Division (Application no. 47,683) and the Institutional Review Board (IRB) at the University of Washington Medical Center (UWMC). The requirement to obtain written consent and the requirement for Health Insurance Portability and Accountability Act (HIPAA) authorization were both waived by the IRB. Medical records were retrospectively reviewed for 632 consecutive patients who had NIPS and were seen between March 2012 and December 2013 in the Prenatal Genetics and Fetal Therapy (PGFT) Program, a tertiary referral center for prenatal genetic counseling and prenatal diagnosis located at UWMC. Patient age at delivery, gestational age at screening, indication for screening, and maternal serum screen results were extracted from outpatient clinic notes. Pregnancy outcomes were extracted from delivery summaries. Fetal ultrasound results were extracted from radiology records; diagnostic genetic test results were extracted from cytogenetics records; and NIPS results were extracted from laboratory records. Of the 632 patients undergoing NIPS, 92 % of the tests were performed in one of the four major commercial laboratories offering testing during this timeframe.
Patients with normal NIPS results typically declined prenatal diagnostic genetic testing. If not already completed, maternal serum alpha-fetoprotein (AFP) and fetal anatomy ultrasound were recommended, and the patient was referred back to her primary obstetrical provider for routine care. Patients with abnormal NIPS results were offered high-resolution fetal ultrasound interpreted by either a UWMC perinatologist or a UWMC radiologist specializing in fetal anatomical imaging, genetic counseling, and prenatal diagnostic genetic testing via chorionic villus sampling (CVS) or amniocentesis. When the latter was declined, postnatal testing was recommended. Maternal chromosome analysis was added to our recommendations several months after we started to offer NIPS. For all patients, follow-up care recommendations were communicated verbally to the patient at the time of the NIPS results disclosure and/or genetic counseling consult, and in writing to the referring provider.
Normal NIPS results were defined as concordant when either diagnostic genetic test results matched the screen results or, for patients declining prenatal diagnostic genetic testing and delivering at UWMC, when normal pregnancy outcome was confirmed by review of delivery records. Newborn exams were performed by pediatricians. For patients with normal NIPS results who declined prenatal diagnostic genetic testing and delivered elsewhere, pregnancy outcomes were not confirmed. Abnormal NIPS results were defined as concordant when either diagnostic genetic test results matched the screen results or when multiple other clinical findings (such as fetal ultrasound abnormalities) corroborated the screen results. Nuchal translucency (NT) was defined as abnormal when the DS likelihood ratio was ≥ 2 [
24]. Patients who declined diagnostic genetic testing and were lost to follow-up with incomplete clinical information were not included in calculations of test performance. Diagnostic genetic test methods included interphase fluorescence in situ hybridization (IFISH), karyotyping, and cytogenomic microarray analysis (CMA). Diagnostic genetic testing was performed by the University of Washington Cytogenetics and Genomics Laboratory unless otherwise stated.
The R statistical software package (version R 2.12.0) [
25] and Excel were used for all statistical analyses. For categorical data, frequencies or percentages with 95 % confidence intervals were derived. For quantitative data, means with standard deviations, medians with minimum and maximum values, and frequencies were calculated. Wilcoxon’s rank sum test was used to evaluate statistical significance between groups, and
P-values of ≤0.05 were considered statistically significant. The PPV was calculated with standard methods [
26]. The reported performance of NIPS is remarkably similar across all test platforms in studies sponsored by the commercial laboratories [
6,
9,
10,
12,
14,
19], so all NIPS results for this cohort were lumped, such that all performing commercial laboratories are represented therein.
Discussion
Publications regarding the performance of NIPS have mostly been sponsored by the commercial laboratories performing the test [
10,
14,
18,
19,
28]. To our knowledge, this is the largest independent report of clinical experience using NIPS in a tertiary referral center in the United States. During the first two years of offering NIPS, we witnessed the powerful benefits of this technology. First, multiple studies have shown that NIPS has a very high NPV, including a recently published prospective multicenter study that found an NPV of 100 % for DS [
19]. Indeed, we are not aware of any normal discordant results among this patient cohort during this timeframe (Fig.
3). Normal NIPS results thus do greatly reduce the probability of a fetus affected with the conditions included in the screen. Whether NIPS is done as a primary screen at 10 weeks gestation or at 22 weeks after abnormal fetal ultrasound findings, normal results provide reassurance. Consequently, our rate of amniocentesis dropped significantly, a trend reported by others [
20,
29‐
31]. In this series, 2 % of patients with normal NIPS results chose prenatal diagnostic genetic testing, as compared to 42 % of patients with abnormal results (Figs.
4 and
5).
Second, NIPS proved to be an invaluable tool for patients wanting to learn as much as possible about fetal health prior to delivery without incurring the miscarriage risk associated with CVS and amniocentesis. These patients commonly expressed that they would not consider TOP, regardless of the diagnosis. Even if NIPS results were abnormal, the family would benefit from being able to prepare for delivery with greater certainty about what lay ahead. In our cohort, this use of NIPS was especially obvious among patients confronted with abnormal findings on their mid-trimester fetal anatomy ultrasound, typically done at 18-22 weeks. Several of our patients in this situation voiced understanding that NIPS results are not diagnostic and still chose NIPS instead of diagnostic genetic testing (Fig.
1). Abnormal fetal ultrasound was the most frequent indication for NIPS (66 %) among patients drawn after 20 weeks.
Our experience also confirmed the limitations of NIPS. First, only a handful of conditions are included in the NIPS panels, so normal results only lower the likelihood of the fetus being affected with those disorders. In this cohort, patient 58 was a 35-year-old who presented at 13 weeks for routine screening. The NT measurement was markedly abnormal. The couple opted for NIPS, which returned with normal results. Fetal ultrasound at 16.3 weeks showed echogenic bowel, a two-vessel cord, unilateral clubfoot, unilateral renal agenesis, and an atrial septal defect. CMA and karyotyping of amniocytes showed mosaicism for trisomy 22, a condition not included in NIPS at that time. Thus, patients with normal NIPS results should still be offered a fetal anatomy ultrasound to screen for congenital abnormalities.
Second, although normal discordant NIPS results are rare with such a high NPV, they do occur. Reports of “false negatives” have started to appear [
32,
33], as the number of patients having NIPS has increased. Since data collection ended for this cohort, we have had one normal discordant result in a 34-year-old who presented at 13 weeks for routine screening. The NT measurement was enlarged. NIPS results were normal, but karyotyping of fetal tissue revealed 47,XY,+21 in all cells.
Third, abnormal NIPS results confer a high risk, but not a diagnosis, of a fetal abnormality. The PPV was 77.4 % (95 % CI 63.4 - 87.3) in this cohort for all conditions included in the NIPS panels. We also observed that the PPVs differed in this cohort for each condition included in NIPS. Abnormal results for DS were the most frequent, and these results had the highest PPV. T18 was the next most frequent condition, and results had the second highest PPV. Abnormal results for the combined SCAs and T13 were the least frequent and had the lowest PPVs. This matches trends in the published literature [
10,
14,
18,
28,
34]. There may be multiple reasons for this. T13 and T18 are less prevalent than DS, which would adversely impact the PPV. Confined placental mosaicism (CPM) for a trisomic cell line was observed more frequently at CVS for chromosomes 13 and 18 than for chromosome 21 [
35,
36]. Postzygotic loss (“trisomy rescue”) in a trophectoderm progenitor cell, leading to placental mosaicism for euploidy, was hypothesized to facilitate the intrauterine survival of T13 and T18 conceptuses [
37]. The percentage of placental DNA in maternal circulation is generally lower when the fetus has T18, T13, or 45,X and higher when the fetus has DS [
4,
28,
38].
There are several causes of discordant results. Discordant results may be caused by statistical limitations of the analysis algorithms and/or uneven sequencing coverage secondary to guanine and cytosine (GC) content differences between chromosomes [
39,
40]. PPV and NPV are influenced by the prevalence of the condition in the screened population. As expected in this cohort, maternal age at delivery was significantly younger among patients with abnormal
discordant results, compared to patients with abnormal
concordant results (Fig.
2a). Other studies have also shown a drop in PPV when women of all levels of risk are included, as compared to a solely high-risk population [
14,
28]. Normal discordant results have been reported to be more likely at an early gestational age because of a lower placental cfDNA fraction [
6,
7]. Gestational age at time of screening was not associated with concordance of results in our series (Fig.
2b), but we do not have data on body mass index and placental cfDNA fraction. Maternal mosaicism was detected in several instances of discordant NIPS results abnormal for an SCA [
34,
41] and has also been reported for T18 [
21]. Patient 51 (Table
2) was an example of this. Maternal chimera due to prior organ transplant was the likely cause of gender discordance for patient 60 (described above). CPM can cause discordant NIPS results, since the source of non-maternal cfDNA during gestation is the placental cytotrophoblastic cell line. Aneuploidy in the cytotrophoblastic cell line with a diploid fetus may cause an
abnormal discordant NIPS result, whereas diploidy in the cytotrophoblastic cell line with a trisomic fetus may cause a
normal discordant NIPS result [
32,
42‐
44]. There are several case reports of discordant NIPS results in pregnancies with proven CPM [
33,
45‐
49]. Placental testing was not done in this cohort, because management is not influenced by results, but CPM may explain the discordant results in patient 46 (Table
2). Other biological reasons for discordant results include a vanishing twin with aneuploidy and maternal metastatic disease [
50].
Apparent sex discordance can have a variety of causes, including inaccurate sex assessment on ultrasound, a co-twin demise, the statistical limitations of NIPS, and a fetus affected with one of the various disorders of sexual development. NIPS did not include the option of fetal sex assessment during the entire timeframe reported here, and not all patients chose to learn the predicted fetal sex from NIPS. Nevertheless, we are aware of only one instance of confirmed fetal sex discrepancy, in a patient who had a personal history of a renal transplant from a male donor.
Thus, it is imperative that providers make every effort to confirm NIPS results that are abnormal or appear discordant for fetal sex. Ideally, confirmation would be with diagnostic genetic testing done prior to making any irrevocable decisions about pregnancy management. Twenty-two families (42 %) in this cohort with abnormal NIPS results chose to have prenatal diagnostic genetic testing (Fig.
5a,
b,
d,
e and Fig.
6). Nine patients in this group continued to term and took advantage of all opportunities to gain information about fetal health during their pregnancies. However, patients may decline prenatal diagnostic testing after an abnormal NIPS result, as did 31 (58 %) patients in this series. Patients incorporate other clinical information about fetal status into their decision-making. If serum screen results, NIPS results, and fetal anatomy ultrasound results all suggest the same diagnosis, the patient may not need further confirmation. Or a patient’s concern about the suspected condition may not be sufficient to warrant the risk of the invasive procedures needed for prenatal diagnosis. And many patients pursue prenatal genetic screening to be better prepared for the birth, and would never incur the risk of a prenatal invasive procedure, regardless of NIPS results. Twenty-two patients in this cohort with abnormal NIPS results (41 %) declined prenatal diagnostic genetic testing with the intention of continuing to term (Fig.
5c,
f,
h, and
i).
Counseling patients about abnormal NIPS results is complex. The likelihood of an affected fetus depends on a priori risk and is influenced by what else is known clinically. For example, in this cohort, abnormal NIPS results were more likely to be discordant when fetal anatomy ultrasound was normal (67 %) than when abnormalities were seen (10 %). Determining appropriate follow-up is also complicated. In some cases, doing a CVS to obtain diagnostic genetic test results may be the best approach, but it is problematic for several reasons. Since the non-maternal component of cfDNA during pregnancy is the placental cytotrophoblast, genetic testing via CVS is a repeat analysis of the same tissue type. If CVS reveals mosaicism, amniocentesis is recommended, and the pregnancy is subjected to two invasive procedures. Cytotrophoblasts are used for karyotyping in direct and short-term culture after CVS. Results from long-term CVS culture are generally based on analysis of the villi mesenchymal cores. Recommendations are to analyze both direct and long-term culture for the most accurate results [
35,
36]. Even so, CVS may not reveal CPM, as only a small portion of the placenta is biopsied during the procedure. Select portions of the placenta may preferentially release cells into maternal circulation [
48], and these may not be the same regions sampled by CVS. Patients should be counseled carefully about these limitations.
Multiple methods for diagnostic genetic testing are available. Karyotyping provides both numerical and structural information and distinguishes free trisomy from an unbalanced Robertsonian translocation. The latter can be inherited, with significantly increased recurrence risks when one parent is a carrier. However, chromosomal microdeletions have been added to NIPS panels, and these are typically undetectable with karyotyping. IFISH and CMA can detect both submicroscopic changes and aneuploidy, and these methods are more successful at obtaining a diagnosis from non-viable, frozen, or formalin-fixed, paraffin-embedded (FFPE) tissue [
51]. But IFISH and CMA do not reveal structural rearrangements, such as unbalanced Robertsonian translocations causing DS.
Maternal karyotyping should be offered after abnormal NIPS results to rule out maternal mosaicism, especially when an SCA is suspected. Beyond possibly providing an explanation for abnormal NIPS results in a current pregnancy, patients with mosaicism are not good candidates for using NIPS in subsequent pregnancies.
For the 632 patients undergoing NIPS, 92 % of the tests were performed in one of the four major commercial laboratories offering testing during this timeframe. Although the results mainly reflected one commercial lab doing NIPS, all four of those laboratories were represented in both the normal and the abnormal results groups of the cohort. In addition, the published performance of NIPS is remarkably similar across all test platforms in studies sponsored by the commercial laboratories [
6,
9,
10,
12,
14,
19]. Thus, the main emphases of this report, that the PPV of NIPS is less than 100 % and abnormal results should be confirmed by diagnostic testing, apply to all NIPS platforms and methods currently in use.
A limitation of this report is that normal pregnancy outcome was not confirmed for 73 % of patients with normal NIPS results. This precluded us from calculating sensitivity, specificity, or NPV for this cohort. However, the PPV of NIPS, the main focus of this report, is unaffected by outcomes of patients with normal NIPS results. Our center does not routinely collect outcome data for patients referred to us who do not deliver at our hospital, which is not unusual in a busy clinical practice. Based on our past experience with referral providers, we feel it likely that we would have been informed of any normal discordant result uncovered by the birth of a child affected with T13, T18, or DS. However, a normal discordant result could have remained undetected if a patient experienced a spontaneous pregnancy loss, and no postnatal genetic testing was done.
Competing interests
The authors declare that they have no competing financial or non-financial interests.
Authors’ contributions
WNK and YL had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. YL conceived of the study and performed the statistical analysis. WNK and YL participated in the study design and coordination, and drafted the manuscript. WNK, EC, and YL participated in acquisition, analysis, and interpretation of the data. All authors read and approved the final manuscript.