A high Z-score might increase the positive predictive value of cell-free noninvasive prenatal testing for singleton-pregnant women

Abstract

Objective

To explore the positive predictive value (PPV) in noninvasive prenatal testing (NIPT)-positive cases and analyze the effect of the Z-score intervals on PPV performance.

Methods

In this retrospective study, 26,667 pregnant women underwent NIPT from November 2014 to August 2022, of which 169 were NIPT-positive cases. NIPT-positive cases were divided into three groups according to the Z-score: 3 ≤ Z < 6, 6 ≤ Z < 10, and Z ≥ 10.

Results

The PPVs of NIPT were 91.26% (94/103) for trisomy (T) 21, 80.65% (25/31) for T18, and 36.84% (7/19) for T13. The PPVs for the 3 ≤ Z < 6, 6 ≤ Z < 10, and Z ≥ 10 groups were 50%, 84.62%, and 87.95%, respectively. A higher PPV was found in the NIPT results when the Z-score was larger, with significant differences. The PPVs for T21/T18/T13 were 71.43%/42.86%/25% for 3 ≤ Z < 6, 90.32%/85.71%/57.14% for 6 ≤ Z < 10, and 93.85%/100%/25% for Z ≥ 10. For T21, T18, and T13, the correlations between the Z-score and fetal fraction concentration in true positives were r = 0.85, r = 0.59, and r = 0.71 (all p < .001), respectively.

Conclusion

Z-score is associated with the PPV performance of NIPT in fetal T13, T18, and T21. The possibility of false positives caused by placental chimerism should be considered when determining whether high Z-values lead to high PPVs.

Keywords:

• Noninvasive prenatal testing
• trisomy
• positive predictive value
• false positive reactions
• Z-score

Introduction

Noninvasive prenatal testing (NIPT) is widely applied in clinical practice because it uses massively parallel sequencing analysis of cell-free fetal DNA (cffDNA) in maternal plasma that originates from trophoblasts and is mixed with cell-free maternal DNA and is a noninvasive method of prenatal screening for fetal chromosomal abnormalities [1,2]. Compared with conventional noninvasive prenatal screening, such as serological tests or ultrasonography, NIPT offers high sensitivity and specificity with a relatively low false-positive rate and high positive predictive values (PPVs) to detect three common autosomal aneuploidies: trisomy 21 (T21), trisomy 13 (T13), and trisomy 18 (T18) [3–5]. The PPV refers to the proportion of positive test results that are truly positive, indicating the probability that a positive test result represents a true fetal abnormality. As a vital indicator to properly interpret NIPT results, PPV is meaningful to clinicians and pregnant women. PPV information helps counsel patients to reinforce the limitations of abnormal cffDNA results and guide patients when choosing invasive diagnostic testing to confirm positive results before making irreversible decisions about pregnancy [4,6].

A previous study reported NIPT results as positive by determining whether the Z-score exceeds a predefined threshold [7]. The Z-values of normal and trisomy samples have an obvious interval distribution [8]. When assessing the fetal risk of T21, T18, and T13, an absolute Z-score ≥3 for chromosome aneuploidies is classified as NIPT-positive [9,10]. Still, false-positive or false-negative NIPT results for Z-score-based fetal aneuploidy can be the product of biological factors such as maternal somatic mosaicism, fetal or (confined) placental mosaicism, vanished twins, maternal malignancy, and copy number variations in the mother [5,11–13]. Confirmation of high-risk results by invasive testing should be performed before making irreversible decisions concerning the pregnancy. Hence, NIPT results should be interpreted comprehensively based on all available information about the pregnancy and the parents. Strom et al. [11] reported complete discrimination between autosomal trisomy (Z-scores >8) and unaffected (Z-scores <4) singleton pregnancies in an assay verification study using 2085 known samples and confirmed this discrimination in a validation study using 552 known samples. Sikkema-Raddatz et al. [14] analyzed the relationship between NIPT results (i.e. Z-scores) and clinical outcomes in 209 pregnant women who wanted to know their true chances of carrying a fetus with aneuploidy after testing positive (Z-scores ≥3). This chance might be much lower than the Z-score percentile would suggest (e.g. 99%), which might otherwise be the reason for them to undergo confirmatory invasive amniocentesis. Knowing the real risks can help avoid hasty and sometimes unnecessary terminations or the false comfort of a negative NIPT result.

Therefore, this study aimed to explore the PPV in NIPT-positive cases on a semiconductor platform and analyze the effect of the Z-score intervals on PPV performance. The results could help improve the management of pregnancies. The relationship between Z-scores and PPV was analyzed, which is a great supplement to the clinical diagnosis and treatment in this region.

Materials and methods

Subjects and sample collection

In this retrospective study, pregnant women underwent NIPT at Yantai Yuhuangding Hospital, China, from November 2014 to August 2022. The study included singleton pregnant women with a gestational age >12 weeks and a body mass index <40 kg/m² who underwent NIPT. The patients who had undergone allogeneic blood transfusion, transplantation, or allogeneic cell therapy within 1 year or the NIPT tests failed were excluded. Ethical approval of the study was obtained from the Institutional Review Board of Yantai Yuhuangding Medical Center (No. [2015]131). All study participants provided written informed consent.

All NIPT-positive cases were validated by karyotype analysis using amniotic fluid puncture. All information on the cases is listed in the Supplementary Reference Table. The clinical data in the Supplementary Reference Table are arranged in ascending order according to the Z-score of NIPT.

DNA sequencing

DNA sequencing of fetal-free DNA fragments was performed using the Jingxin BioelectronSeq 4000 semiconductor sequencer System (CFDA registration permit No. 20153400309) [15]. After sequencing, the reads were filtered and aligned to the human reference genome (hg19) to obtain the unique reads. Here, an absolute value of the Z-score greater ≥3 was marked with a NIPT-positive result for the fetal risks of T21, T18, and T13. Simultaneously, the fetal DNA concentration was calculated as quality control, as described by Yin et al. [16]. NIPT-positive cases were divided into the following three groups according to the Z-score: 3 ≤ Z < 6, 6 ≤ Z < 10, and Z ≥ 10. This classification was based on the current sample quantity and sample selection range and the combination of the results of this study and previous relevant studies [8,17,18]. The performance in detecting T21, T18, and T13 in these three groups was compared using karyotyping or follow-up results as the gold standard [5].

Validation by karyotype analysis using amniotic fluid puncture

The NIPT-positive results were validated against karyotype analysis via amniotic fluid puncture, performed routinely. Karyotype analysis was performed according to the International System for Human Cytogenomic Nomenclature (ISCN, 2011) guidelines [19].

Statistical analysis

Stata 17.0 (StataCorp LP, College Station, TX, USA) was used for data analysis. The PPVs were calculated for the three groups of Z-scores. Fisher’s exact test was used to analyze the differences in categorical variables. Logistic regression analysis was used to correlate the Z-scores with PPV for positive NIPT results. Two-sided p < .05 was considered statistically significant. The Spearman test was used to observe the strength of linear correlations between two continuous variables. Structural equation analyses (SEMs) were performed to determine how the different factors interact with the PPV. The SEMs were performed based on three hypotheses: 1) age has a direct effect on the FF, 2) age has a direct effect on the Z-score, and 3) the Z-score has a direct effect on the FF. The SEMs were performed for each trisomy and in true and false positives.

Results

Characteristics of the patients

The flowchart of NIPT results and clinical outcomes of single pregnancies is presented in Supplementary Figure 1. From November 2014 to August 2022 at Yantai Yuhuangding Hospital (China), the maternal blood samples from 26,667 singleton-pregnant women were collected and tested using NIPT. There were 169 NIPT-positive cases. One woman did not return, and 15 women with spontaneous or induced labor could not be validated by amniocentesis and were excluded (Supplementary Figure S1). Therefore, 153 NIPT-positive cases were validated by karyotype analysis using amniotic fluid puncture and included in the study. As shown in the Supplementary Reference Table, the 153 NIPT-positive pregnant women had a median maternal age of 34 years (ranging from 25 to 47 years) and a median gestational age of 15 weeks (ranging from 12 to 21 weeks).

Relationship between the Z-scores and PPV

There were 126 true positive cases, among which nine were in the 3 ≤ Z < 6 group, 44 in the 6 ≤ Z < 10 group, and 73 in the Z ≥ 10 group. In detail, the 3 ≤ Z < 6 group had 18 NIPT-positive cases, among which nine were true-positive, for a PPV of 50.00%. The 6 ≤ Z < 10 group had 52 NIPT-positive cases, among which 44 were true positive, for a PPV of 84.62%. The Z ≥ 10 group had 83 NIPT-positive cases, among which 73 were true positive, for a PPV of 87.95% (Table 1).

Table 1. Performance of NIPT-positive results for T21/T18/T13 among different groups.

	TP			FP			Total TP	Total FP	PPV
	T21	T18	T13	T21	T18	T13
Group 3 ≤ Z < 6	5	3	1	2	4	3	9	9	50.00%
Group 6 ≤ Z < 10	28	12	4	3	2	3	44	8	84.62%
Group Z ≥ 10	61	10	2	4	0	6	73	10	87.95%
Total	94	25	7	9	6	12	126	27	82.35%

TP: true positive; FP: false positive; PPV: positive predictive value; T: trisomy.

Relationships between the Z-scores and PPV for T21, T18, and T13

There were 126 true-positive cases (94 with T21, 25 with T18, and seven with T13), 27 false-positive cases (nine for T21, six for T18, and 12 for T13), and no false-negative cases, resulting in PPVs of 91.26% for T21, 80.65% for T18, and 36.84% for T13 (Table 1). In the 3 ≤ Z < 6 groups, there were seven NIPT-positive cases for T21, seven for T18, and four for T13, among which the true positives were five for T21, three for T18, and one for T13. In the 3 ≤ Z < 6 group, the PPVs for T21, T18, and T13 were 71.43%, 42.86%, and 25%, respectively. The PPVs for T21, T18, and T13 were 90.32%, 85.71%, and 57.14% in the 6 ≤ Z < 10 groups, and 93.85%, 100%, and 25% in the Z ≥ 10 groups.

Multivariable analyses were conducted for T21, T18, and T13 to explore the correlation between the Z-score and PPV (Table 2). Logistic regression was used with the amniocentesis result as the outcome (true positive/false positive). No variables were corrected, and only one group was analyzed at a time. The results showed that the Z-score showed a significant positive correlation with PPV for T18 (p = .018).

Table 2. Multivariable analysis between Z-score and PPV for T21, T18, and T13.

Group	OR	95% CI	p
T21	1.051	0.935, 1.181	.407
T18	1.773	1.102, 2.854	.018
T13	0.891	0.692, 1.147	.371

OR: odds ratio; CI: confidence interval.

Correlation between FF and Z-score on T21, T18, T13

In Supplementary Figure S2, the vertical axis represents the Z-score, and the horizontal axis represents the fetal fraction concentration (FF%). For T21, there was a positive correlation between the Z-score and FF in true positives (r = 0.85, p < .001), indicating a high correlation. Among false positives for T21, the correlation between Z-score and FF was negative (r = –0.33, p < .001) (weak correlation). For T18, there was a positive correlation between Z-score and FF in true positives (r = 0.59, p < .001), indicating a moderate correlation. In false positives of T18, the correlation between the Z-score and FF was positive (r = 0.03, p < .001), indicating a very weak correlation. For T13, there was a positive correlation between Z-score and FF in true positives (r = 0.71, p < .001), indicating a high correlation. In false positives of T13, the correlation between the Z-score and FF was negative (r = –0.16, p < .001), suggesting a weak correlation.

Structural equation model

In true positive T21, age influenced the Z-scores (β = −0.46, 95%CI: −0.71, −2.12, p < .001), and the Z-scores influenced the FF (β = 0.93, 95%CI: 0.78, 1.08, p < .001), while age did not affect the FF (p = .862) (Figure 1(A)). In false positive T21, there were no significant relationships among age, Z-scores, and FF (all p > .05) (Figure 1(B)). In T18 true positive, the Z-scores influenced the FF (β = 0.56, 95%CI: 0.24, 0.89, p = .001), while age did not influence the Z-scores or FF (both p > .05) (Figure 1(C)). In T18 false positive, the age influenced the FF (β = 1.52, 95%CI: 0.38, 2.66, p = .009) but not the Z-scores (p = .474), and the Z-scores did not influence FF (p = .811) (Figure 1(D)). In T13 true positive, the Z-scores influenced the FF (β = 1.16, 95%CI: 0.50, 1.82, p = .001), while age did not influence the Z-scores or FF (both p > .05) (Figure 1(E)). In false positive T13, there were no significant relationships among age, Z-scores, and FF (all p > .05) (Figure 1(F)).

Figure 1. Structural equation analyses. (A) True positive trisomy (T) 21. (B) False positive T21. (C) True positive T18. (D) False positive T18. (E) True positive T13. (F) False positive T13.

Typical cases

No. 19YHD030112 was identified as trisomy 18 (Z-score = 7.827) by NIPT, but the Z-score distribution curve suggested that it might be caused by maternal repetition (Supplementary Figure S3). Hence, the result was confirmed by amniocentesis and karyotype analysis using the copy number variation (CNV)-seq method. The karyotype was 46, XN, t (6;18) (q27; q12.2), and seq [(GRCh37)] dup [18] (q12.2q12.3) was found by CNV-seq. In order to determine the origin of the chromosome duplication, karyotype analysis, and CNV-seq tests were performed on the parents’ peripheral blood. The father’s karyotype was 46, XY, and the mother’s karyotype was 47, XX, +mar [33]/46, XX [22], and seq[(GRCh37)] dup [18] (q12.2q12.3). Thus, in the 19YHD030112 case, the repeat segment of chromosome 18 of the fetus was inherited from the mother. After birth, no abnormality was observed in facial appearance or body reaction to 1 year of age.

No. 20YHD031101 case had 21 trisomeric mosaicism at the placental level. The karyotype analysis of the amniotic fluid showed 47, XN +16. Five cases of T13 (Z = 7.736 for 19YHD011003, Z = 4.861 for 18YHD041213, Z = 6.274 for 19YHD123007, Z = 10.526 for 19YHD091811, and Z = 13.195 for 18YHD111717) had possible confined placental mosaicism (CPM) when analyzed by bioinformatics. One case of T13 (Z-score = 17.904 for 17YHD072312) was confirmed as CPM by invasive diagnosis.

Discussion

The results suggest that the PPVs of NIPT were 91.26% for T21, 80.65% for T18, and 36.84% for T13. The PPVs for the 3 ≤ Z < 6, 6 ≤ Z < 10, and Z ≥ 10 groups were 50%, 84.62%, and 87.95%, respectively. Hence, this study provided useful information concerning the Z-score and PPV of NIPT for T21, T18, and T13, which might help with result interpretation, patient counseling, and clinical decisions. Some biological factors must be considered when interpreting NIPT results.

NIPT is an effective and reliable method for prenatal screening of common aneuploidies using cffDNA in maternal plasma, but the possibility of false-positive or false-negative results for the targeted abnormalities should always be kept in mind [12,20]. From the cases shown above, false positives (i.e. low PPV) can be caused by copy number variations (CNVs) in the mother and CPM in the placenta. Of course, besides the cases observed in this study, false positives can also be caused by maternal malignancies [21], organ transplants [22], and vanishing twins [23].

CPM is a major factor causing false-positive results of NIPT. It refers to a chromosomal abnormality occurring only in the placenta but not in the fetus because of two or more cell lines from a single zygote [24]. False-positive results of NIPT are observed when abnormal cell lines are present in the cytotrophoblast layer, and normal amniotic fluid results were obtained. Hartwig et al. [12] reported that 67% of the cases of discordant results had no obvious biological or technical explanation; among the false-positive cases with a biological technical explanation (33%), CPM represented 32% of the cases. In the T21 false-positive cases in this study, five false-positive cases in the Z ≥ 10 group and two false-positive cases in the 6 ≤ Z < 10 groups were confirmed as CPM by invasive prenatal diagnosis. In the Z ≥ 10 groups, CPM accounted for 100% (4/4) of T21 false-positive cases. For T13, the number of false positive cases was the largest, and the PPV was the lowest (36.84%). The possible reasons for that phenomenon might involve the size of chromosome 13 or sequencing deviation with a relatively low GC content on chromosome 13 (25), and a large proportion of mosaic samples was observed for chromosome 13 [13,24], suggesting that false positives for T13 might be caused by CPM.

In addition to CPM, maternal CNVs can produce false-positive results [5,25,26]. Two patients who had tested positive for T18 by NIPT were observed to carry duplicate regions on their chromosome 18. It was speculated that these repeats interfered with bioinformatics analysis because most of the cffDNA in the plasma comes from the mother. The mother’s repetition increases the number of reads and the chance of false positives [27]. Therefore, to ensure that a positive result for whole-chromosome duplication was not due to maternal duplication, a karyotype should be performed in the presence of an elevated Z score. Therefore, each case with CNV or mosaicism identified by NIPT should be confirmed using maternal white blood cells, amniotic fluid, and multiple placental samples [5].

The PPV for T21 (91.26%) was significantly higher than for T18 (80.65%) and T13 (36.84%). A comparative analysis of the PPV for T21, T18, and T13 within the three groups showed that the PPV for T13, T18, and T21 gradually increased in the 3 ≤ Z < 6 and 6 ≤ Z < 10 groups, but the PPV for T18 in the Z ≥ 10 group was 100%, higher than that of T21 and T13. Combined with the karyotype results, it can be seen that CPM has a great influence on the PPV.

Previous studies and the present one showed that the performance of the PPV is closely related to the Z-score. According to Tian et al. [28], NIPT-positive results at Z ≥ 9 showed higher accuracy than NIPT-positive results at 5 ≤ Z < 9 and 3 ≤ Z < 5. Junhui et al. [8] classified the Z-scores as 3 ≤ Z < 5, 5 ≤ Z < 10, and Z ≥ 10; the PPV performance was different among the three groups, particularly in the 3 ≤ Z < 5 and Z ≥ 10 groups for T21. Zhou et al. [29] also reported that especially for T12 and T18, the PPV in the high Z-value group was higher than in the medium Z-value group, and applying the cutoff value reduced the false discovery rate. These data underscore the importance of evaluating the correlation between the Z-score and PPV and determining the optimal combination of cutoff values when assessing NIPT results, suggesting that each laboratory should define its correlation and optimal combination of cutoff values since those are study-specific.

The FF parameter in NIPT is regarded as a quality control indicator. As reported, NIPT results should not be reported for samples with FF below 4% [30]. In the present study, among true positives, the Z-score increased with the increasing FF. There were high correlations between the Z-score and FF. On the other hand, among false negatives, the correlations were weak. According to previous NIPT studies [31,32]. The FF will also increase as the gestational age increases [31,32]. Therefore, if the influence of CPM and other factors cannot be eliminated for false positive cases, there will still be false positives even if the pregnant women are tested again several weeks later.

Even if the present study and several previous studies focused on the PPV [4,6,8,9,11,25,29,30] and the risk of positive results leading to pregnancy termination, the risk of a false-negative result should not be underestimated. In case of a false positive test leading to pregnancy termination, a novel pregnancy can be attempted shortly after. On the other hand, a false-negative result discovered at birth can have serious consequences and a source of distress for many parents that will endure for life. The present study could not examine the false-negative rates since only the patients with a positive NIPT underwent karyotype. Many physicians will use cutoffs of −3/3, but the cutoff value should not be the only factor considered when counselling the patients. Indeed, physicians should consider the age, FF, and Z-scores comprehensively, as suggested by the American College of Gynecologists [33]. Indeed, the preliminary SEMs analyses suggested relationships among age, Z-scores, and FF.

This study has some limitations. Due to the short time and the small number of people included in the retrospective analysis, this study cannot be used to determine the NIPT thresholds for clinical use. In addition, all cases were from a single center in the Yantai area. Additional observations from more medical institutions are needed to shed more light on the issue of NIPT threshold selection. The false-positive rate should be decreased to improve the performance of PPV in NIPT for aneuploidies. In the present study, the biological factors of the selected groups (e.g. CPM, maternal CNV, or fetal or maternal mosaicism), the sample size, and platform performance might contribute to the performance of the PPV.

Conclusion

The Z-score is associated with the PPV performance of NIPT in fetal T13, T18, and T21. Due to the characteristics of chromosome 13, the possibility of false positives caused by placental chimerism should be considered when determining whether a high Z-value would lead to a high PPV value. Some biological factors must be considered when interpreting NIPT results. Hence, this study provided useful information concerning the Z-score and PPV of NIPT for T21, T18, and T13, which might help result in interpretation, patient counseling and clinical decisions.

Author contributions

Li Ma and Yulan Li carried out the studies and drafted the manuscript. Lei Li and Hong Wu participated in collecting data. Yongming Liu performed the statistical analysis and participated in its design. Xin Yang and Aimin Lin participated in the acquisition, and interpretation of data. All authors read and approved the final manuscript.

Acknowledgments

We thank all those who participated in this study. We are also grateful to the staff who collected information and blood from pregnant women, without whose efforts we could not have conducted the study successfully.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data availability statement

This article includes all the relevant data generated or analyzed during this study.

Additional information

Funding

The author(s) reported there is no funding associated with the work featured in this article.