A statistical investigation of parameters associated with low cell-free fetal DNA fraction in maternal plasma for noninvasive prenatal testing

Abstract

Background

Noninvasive prenatal testing (NIPT) is the most common method for prenatal aneuploidy screening. Low fetal fraction (LFF) is the primary reason for NIPT failure. Consequently, factors associated with LFF should be elucidated for optimal clinical implementation of NIPT.

Methods

In this study, NIPT data from January 2019 to December 2022 from the laboratory records and obstetrical and neonatal data from the electronic medical records were collected and analyzed. Subjects with FF >3.50% were assigned to the control group, subjects with FF <3.50% once were assigned to the LFF group, and subjects with FF <3.50% twice were assigned to the repetitive low fetal fraction (RLFF) group. Factors, including body mass index (BMI), gestational age, maternal age, twin pregnancy, and in vitro fertilization (IVF) known to be associated with LFF were assessed by Kruskal–Wallis H test and logistic regression. Clinical data on first trimester pregnancy-associated plasma protein-A (PAPP-A), beta-human chorionic gonadotropin (β-hCG), gestational age at delivery, birth weight at delivery, and maternal diseases were obtained from the hospital’s prenatal and neonatal screening systems (twin pregnancy was not included in the data on gestational age at delivery and the control group did not include data on maternal diseases.), and were analyzed using Kruskal–Wallis H test and Chi-square test.

Results

Among the total of 63,883 subjects, 63,605 subjects were assigned to the control group, 197 subjects were assigned to the LFF group, and 81 subjects were assigned to the RLFF group. The median of BMI in the three groups was 22.43 kg/m² (control), 25.71 kg/m² (LFF), and 24.54 kg/m² (RLFF). The median gestational age in the three groups was 130 days (control), 126 days (LFF), and 122/133 days (RLFF). The median maternal age in the three groups was 29 (control), 29 (LFF), and 33-years-old (RLFF). The proportion of twin pregnancies in the three groups was 3.3% (control), 10.7% (LFF), and 11.7% (RLFF). The proportion of IVF in the three groups was 4.7% (control), 11.7% (LFF), and 21.3% (RLFF). The factors significantly associated with LFF included BMI [2.18, (1.94, 2.45), p < 0.0001], gestational age [0.76, (0.67, 0.87), p < 0.0001], twin pregnancy [1.62, (1.02, 2.52), p = 0.0353], and IVF [2.68, (1.82, 3.86), p < 0.0001]. The factors associated with RLFF included maternal age [1.54, (1.17, 2.05), p = 0.0023] and IVF [2.55, (1.19, 5.54), p = 0.016]. Multiples of the median (MOM) value of β-hCG and pregnant persons’ gestational age at delivery were significantly decreased in the LFF and RLFF groups compared to the control group.

Conclusion

According to our findings based on the OR value, factors associated strongly with LFF include a high BMI and the use of IVF. Factors associated less strongly with LFF include early gestational age and twin pregnancy, while advanced maternal age and IVF were independent risk factors for a second LFF result.

HIGHLIGHTS

Body mass index, gestational age, maternal age, twin pregnancy, and in vitro fertilization are associated with fetal fraction. We added the repetitive low fetal fraction population and used a large normal population as a control to identify the main factors associated with low fetal fraction.

Keywords:

• Noninvasive prenatal testing
• low fetal fraction
• body mass index
• gestational age
• in vitro fertilization
• maternal age

Introduction

Since its discovery in 1948 in human blood plasma, circulating cell-free DNA (cfDNA) has become a crucial biomarker of many diseases, including autoimmune diseases and malignancies [1]. In 1997, the discovery of cell-free fetal DNA (cffDNA) in maternal plasma marked the beginning of its application in prenatal medicine [2]. CfDNA in gravidae’s blood plasma is composed of maternal and fetal DNA, the latter primarily derived from the apoptosis of placental trophoblast cells consisting of 50–300 bp fragments.

The primary application of noninvasive prenatal testing (NIPT) has been to detect fetal aneuploidies. Fetal fraction (FF) is the percentage value of the ratio of cffDNA to the total cfDNA in maternal plasma. The estimation of FF is crucial in NIPT to determine fetal aneuploidies, and the value of FF at 10–20 weeks of gestation (common time for NIPT) is approximately 10–15% [3,4]. FF evaluation is a crucial component of quality control that ensures that sufficient cffDNA is present in maternal plasma to provide accurate results. Typically, an FF of 2–4% is specified as the lower limit for accurate NIPT results [5,6]. Setting a suitable threshold achieves an equilibrium between maximal statistical reliability of NIPT and minimal failure rate. Our laboratory extensively tested and established the FF threshold at 3.50%. If one sample falls below the threshold, it was deemed a “test failure” and the patient is instructed to resubmit another sample of peripheral blood for a second NIPT.

Hitherto, there has not been a comparative analysis of the factors that affect FF. Several studies have reported that body mass index (BMI), in vitro fertilization (IVF), gestational age, and some other factors are associated with FF [7–9]. The longer the mother is pregnant, the more cffDNA is present in her blood, and a statistically significant positive correlation between gestational age and FF [10] has been established. In contrast, the higher the BMI, the greater the amount of maternal DNA excreted into the bloodstream, decreasing the proportion of cffDNA. The cffDNA in maternal plasma for women with twin pregnancies is somewhat elevated compared to women with singleton pregnancies, but it is still less than double [11–14], which means that the individual cfDNA contributed by each fetus is generally lower than that for a singleton. Additional reasons for why twins might present with a low fetal fraction remain unclear but it is likely that twin pregnancy synergizes with other factors (BMI, IVF) to collectively contribute to a low fetal fraction fraction [15]. Theoretically, any factor that influences the ratio of maternal DNA and cffDNA to cfDNA in the mother’s blood plasma affects the FF. Variations in placental serum markers, such as β-hCG and PAPP-A appear related to why the FF of IVF-facilitated pregnancies may be lower [16,17]. PAPP-A and β-hCG promote placental development, thereby increasing FF.

In our study, the proportion of the low FF (<3.50%) accounted for approximately 0.5% of all samples, indicative of a large sample size being required to analyze the interfering factors contributing toward low FF (LFF). The majority of patients with a first LFF have their blood collected for a second NIPT, and approximately one-third of the samples have a LFF after a second NIPT. These samples we refer to as repetitive low FF (RLFF). The RLFF samples indicate the existence of long-lasting factors that contribute to the LFF despite increasing gestational age. For the present study, we retrospectively collected the data from 2019–2022 to obtain sufficient samples to analyze various interfering factors that have been associated with LFF, and to, furthermore, distinguish those that may differ between LFF and RLFF in order to identify those which we have termed “long-lasting factors.” Reportedly, LFF not only indicates the failure of NIPT but is also a risk factor for premature birth in pregnant women [18], prompting us to follow-up on these LFF cases to determine their gestational age at delivery.

Methods

Study population and data collection

This observational study was conducted in Women and Children’s Hospital of Ningbo University, Zhejiang Province, China, between January 2019 and December 2022. Pregnant patients who underwent NIPT at our hospital represented the overall study population. The protocol was approved by the ethics committee of the hospital (EC2020-048), and informed consent was obtained from each participant. Women were ineligible for inclusion in the study if any of the following conditions were fulfilled: (1) Incomplete medical records or obvious errors in height and weight measurements. (2) Failure of NIPT due to causes other than LFF. The final study population from which the following subsets were designated: those with an initial low FF (<3.50%, designated as low FF) and others with an initial FF (>3.50%) subjects comprised the control group. Blood samples were then collected again for NIPT from low FF (<3.50%) subjects. Those subjects whose initial FF was low but whose second FF was normal were classified as the LFF group. One-third of the LFF samples were still LFF after a second NIPT, and were termed the repetitive low FF (RLFF) group (Figure  ). Clinical data, such as first trimester PAPP-A, β-hCG, and gestational age at delivery, were obtained from the hospital’s prenatal and neonatal screening systems (twin pregnancy is not included in the data on gestational age at delivery). The influence of gestational age on RLFF was compared by dividing RLFF into two groups: RLFF1 (gestational age at first NIPT) and RLFF2 (gestational age at second NIPT). RLFF1 and RLFF2 represent the same population at different gestational stages for NIPT. Maternal health and obstetrical history on LFF and RLFF subjects were obtained from the electronic medical record system of pregnant women. The types of maternal diseases that we included were scarred uterus, diabetes, hypertension, thrombophilia, autoimmune diseases (Hashimoto’s Thyroiditis, ankylosing spondylitis, Systemic Lupus Erythematosus, connective tissue disorders), and inflammation (active hepatitis, mastitis, Fungous Vaginitis, acute pharyngitis). We did not include health history information for the control group, because we did not have the capability to achieve electronic integration of perinatal information with the hospital’s laboratory’s database whereas, for the purpose of this study, we were able to search the medical records for the smaller subset of patients with LFF. The subjects in the control group with gestational age at delivery and birth weight at delivery were selected from the integer multiples of the serial number for every 200 cases, and the cases without information were skipped. Among the individuals in the LFF and RLFF groups who had a history of a scarred uterus, we analyzed their median maternal age with quartile intervals. Prenatal diagnosis (karyotyping and chromosomal microarray analysis) for genetic testing was performed by patient request on 1 of 197 LFF cases and 12 of 81 RLFF cases.

Figure 1. Flow chart of study population. It is necessary to point out that the influence of gestational age on RLFF was compared by dividing RLFF into two groups: RLFF1 (gestational age at first NIPT) and RLFF2 (gestational age at second NIPT). RLFF1 and RLFF2 represent the same population at different gestational stages for NIPT.

Sequencing analysis of maternal plasma DNA and FF calculation

From each patient a 10 ml sample of maternal peripheral blood was collected, separated by double centrifugation (1600 g, 4 °C 10 min) and stored at −20 °C until further processing. Cell-free DNA was extracted from 500 μL of maternal plasma using the Nucleic Acid Extraction Kit (BGI Genomics) on MGISP-960 High-throughput automation system following the blood and body fluid protocol. Then, cell-free DNA was enriched, purified, and subjected to end repair. DNA connectors were linked, and magnetic beads (BGI Genomics) were utilized to purify the connector products. DNA library amplification and purification were followed by pooling. After quantification on the Veriti^TM Dx 96 Well Thermal Cycler PCR platform (Life Technologies), the libraries from 48 different samples (47 samples + 1 blank) were pooled and sequenced on the MGISEQ-2000 platform (MGI) according to the standard operating procedure.

The FF was calculated based on its accurate quantification using Shallow-Coverage sequencing (FF-QuantSC) method based on the neural network model that follows the procedural data preparation: (1) sequencing data alignment with Hg19; (2) filter and count mapped results; (3) normalization; (4) construction of the feature matrix; (5) split into training and test set →Model training → Model evaluation [19]. The advantage of this method was that it evaluates the FF of twin pregnancies without considering the gender of the fetus.

Statistical analysis

Statistical analyses were performed using SPSS 26.0. All measurement data are represented by median and quartile intervals (M (Q₁–Q₃)). Inter-group and pairwise comparisons were carried out using the Kruskal–Wallis H test and Nemenyi test, respectively. Differences were considered statistically significant at p < 0.05 and marked by different symbols in the figures. Multivariate statistical analysis was carried out using multiple logistic regression analysis. Measurement data (BMI, gestational age, and maternal age) was transformed into hierarchical data for regression analysis. BMI was classified into six incremental levels: (1) <18.50 kg/m², (2) 18.50–24.99 kg/m², (3) 25.00–29.99 kg/m², (4) 30.00–34.99 kg/m², (5) 35.00–40.00 kg/m², and (6) >40.00 kg/m². Gestational age was also classified into six levels: (1) <80 days, (2) 80–99 days, (3) 100–119 days, (4) 120–139 days, (5) 140–160 days, and (6) >160 days. The maternal age was classified into six levels as follows: (1) <20-years-old, (2) 20–25-years-old, (3) 26–30-years-old, (4) 31–35-years-old, (5) 36–40-years-old, (6) >40-years-old. Considering the known association between twin pregnancy and IVF, we analyzed whether IVF-twin pregnancy was independent of IVF in affecting FF and the association with IVF could be mediated through twin pregnancies. We removed 859 twins of IVF from the three groups and the factor (twin pregnancy) from the multiple logistic regression. Odds ratio (OR) represents the risk coefficient of each factor relative to the outcome. OR > 1 indicates a positive correlation between the factor and outcome, OR < 1 indicates a negative correlation between the factor and outcome, and OR = 1 indicates that this factor is not related to the outcome. Clinical data on first trimester PAPP-A, β-hCG, gestational age at delivery, and birth weight at delivery were analyzed using Kruskal–Wallis H test; as well, the incidence of maternal diseases in the LFF and RLFF groups was analyzed using Chi-square test.

The figures are illustrated using Prism 9 software. The measurement data are presented in the column tables, while the proportionable data are presented in the contingency tables.

Results

Dataset summary of the studied population

A total of 63,883 subjects undergoing NIPT at Women and Children’s Hospital of Ningbo University during the period of January 2019 to December 2022 were enrolled in the present study, among them, 63,605 subjects were assigned to the control group, 197 subjects were assigned to the LFF group, and 81 subjects were assigned to the RLFF group. The pregnancy characteristics for all three subsets (LFF, RLFF, and Controls) of these cases associated with low fetal fraction are summarized in Table 1. Specific inter-group statistical significance is shown in Figure 2.

Figure 2. Boxplots and stacked bars of factors associated with LFF and RLFF. (A) Distribution of BMI across all 63,883 subjects in the three groups. (B) Distribution of gestational age across all 63,883 subjects in the four groups. RLFF1 (gestational age at first NIPT) and RLFF2 (gestational age at second NIPT). (C) Distribution of maternal age across all 63,883 subjects in the three groups. (D) Ratio of twin pregnancy cross all 63,883 subjects in the three groups. (E) Ratio of IVF across all 63,883 subjects in the three groups. All measurement data (BMI, gestational age, and maternal age) are represented by median and quartile intervals (M (Q1–Q3)) (Table 1). **p < 0.01; ****p < 0.0001; NS, no statistical significance.

Table 1. Dataset summary of factors associated with LFF in 63,883 subjects undergoing NIPT in the study population.

	Control (n = 63,605)	LFF (n = 197)	RLFF (n = 81)	Z/X^2	p
BMI at NIPT (kg/m^2)^†	22.43 (20.55–24.76)	25.71 (22.96-28.66)	24.54 (22.23–28.26)	139.20	<0.0001
Gestational age at NIPT (days)	130 (119–140)	126 (111-134)	122 (108.5-130.5)/ 133 (120–142)	38.72	<0.0001
Maternal age at NIPT (years)	29 (26–33)	29 (26–33)	33 (29–35.5)	23.44	<0.0001
Twin pregnancy (%)	2093 / (3.3%)	21 / (10.7%)	9 / (11.1%)	48.51	<0.0001
IVF (%)	2977 / (4.7%)	23 / (11.7%)	20 / (21.3%)	93.10	<0.0001

Note: ^†All measurement data (BMI, gestational age, and maternal age) are represented by median and quartile intervals (M (Q1–Q3)).

The BMI was higher in the LFF and RLFF groups than in the control group, while that between the LFF and RLFF groups showed no statistical significance (Figure 2(A)). The gestational age of the LFF and RLFF1 (gestational age at first NIPT) groups was less than that of the control group, while that between the LFF group and RLFF1 (gestational age at first NIPT) group did not show any statistical significance (Figure 2(B)). Also, the gestational age between the control and the RLFF2 (gestational age at second NIPT) groups showed no statistical significance, and that of the RLFF2 (gestational age at second NIPT) group was higher than the LFF and RLFF1 (gestational age at first NIPT) groups (Figure 2(B)). In addition, the median maternal age of the RLFF group was higher than the control and LFF groups, but that between the control and LFF groups showed no statistical significance (Figure 2(C)). Furthermore, the proportion of twin pregnancies was lower in the control group than in the LFF and the RLFF groups, but did not differ significantly between the LFF and the RLFF groups (Figure 2(D)). Finally, the proportion of IVF-conceived pregnancies was lower in the control group than in the LFF and RLFF groups, while that in the LFF group was lower than the RLFF group (Figure 2(E)).

Multiple logistic regression analysis of factors associated with LFF

Firstly, we transformed the measurement data (BMI, gestational age, and maternal age) into hierarchical data (Table 2).

Table 2. Transformed measurement data (BMI, gestational age, maternal age) into hierarchical data.

BMI (kg/m^2)	<18.50	18.50–24.99	25.00–29.99	30.00–34.99	35.00–40.00	>40.00
Control (n = 63,605)	3815 (1)	45,264 (2)	12,213 (3)	1902 (4)	284 (5)	127 (6)
LFF (n = 197)	3 (1)	89 (2)	73 (3)	24(4)	7 (5)	1 (6)
RLFF (n = 81)	2 (1)	41 (2)	26 (3)	9 (4)	1 (5)	2 (6)
Gestational age (days)	<80	80–99	100–119	120–139	140–160	>160
Control (n = 63,605)	1 (1)	5844 (2)	10,669 (3)	30,695 (4)	10,634 (5)	5762 (6)
LFF (n = 197)	0 (1)	27 (2)	42(3)	94 (4)	29 (5)	5 (6)
RLFF (n = 81)	0 (1)	10 (2)	28 (3)	37 (4)	5 (5)	1 (6)
Maternal age (years)	<20	20–25	26–30	31–35	36–40	>40
Control (n = 63,605)	1020 (1)	11,616 (2)	25,645 (3)	17,559 (4)	7082 (5)	683 (6)
LFF (n = 197)	4 (1)	34 (2)	80 (3)	52 (4)	24 (5)	3 (6)
RLFF (n = 81)	1 (1)	4 (2)	26 (3)	30 (4)	17 (5)	3 (6)

Secondly, for the five factors in the analysis we performed multiple logistic regression between the control group and the LFF group combined with the RLFF group, as shown in Table 3. Factors including BMI [2.18, (1.94, 2.45), p < 0.0001], gestational age [0.76, (0.67, 0.87), p < 0.0001], twin pregnancy [1.62, (1.02, 2.52), p = 0.0353], and IVF [2.68, (1.82, 3.86), p < 0.0001] all showed statistical significance with LFF, while only maternal age showed no statistical significance. On the other hand, maternal age [1.54, (1.17, 2.05), p = 0.0023] and IVF [2.55, (1.19, 5.54), p = 0.016] did show statistical significance with RLFF, while other factors did not (Table 3). The association between twin pregnancy and IVF is shown in the footnote of Table 3. After removing the 859 twins of IVF from the three groups, twin pregnancy (OR = 1.95, p = 0.026) and IVF (OR = 2.94, p < 0.0001) both showed an increase of OR with statistical significance for LFF, while other factors remained similar (Table 3, footnote^†). After removing the factor (twin pregnancy) from the multiple logistic regression, IVF (OR = 3.13, p < 0.0001) showed an increase of OR with statistical significance for LFF, while other factors remained similar (Table 3, footnote^‡). Hosmer–Lemeshow test did not detect any statistically significant difference, thereby indicating that an appropriate model was selected.

Table 3. Multiple logistic regression analysis of factors associated with LFF and RLFF.

Outcome: 0 vs. 1	Control (n = 63,605) vs. (LFF + RLFF, n = 278)			LFF (n = 197) vs. RLFF (n = 81)
	OR	p-Value	95% CI	OR	p-Value	95% CI
BMI^#	2.18	<0.0001	(1.94–2.45)	0.97	NS	(0.70–1.33)
Gestational age	0.76	<0.0001	(0.67–0.87)	0.89	NS	(0.65–1.22)
Maternal age	1.01	NS	(0.89–1.15)	1.54	0.0023	(1.17–2.05)
Twin pregnancy (0, 1)^†	1.62	0.0353	(1.02–2.52)	0.58	NS	(0.21–1.48)
IVF (0, 1)^‡	2.68	<0.0001	(1.82–3.86)	2.55	0.016	(1.19–5.54)

Note: ^#BMI, gestational age, and maternal age data were transferred into six incremental level (see Table 2).

^† After removing the 859 twins of IVF from the three groups, for control vs. (LFF + RLFF): twin pregnancy (OR = 1.95, p = 0.026) and IVF (OR = 2.94, p < 0.0001), for LFF vs. RLFF: twin pregnancy (OR = 1.50, P = NS) and IVF (OR = 4.07, p = 0.0016). OR of other factors remain similar.

^‡ After removing the factor (twin pregnancy) from the multiple logistic regression, for control vs. (LFF + RLFF): IVF (OR = 3.13, p < 0.0001), for LFF vs. RLFF: IVF (OR = 2.16, p = 0.03). OR of other factors remain similar.

β-hCG and PAPP-a, delivery timing, chromosomal abnormalities and maternal characteristics in regard to LFF

The MOM values of first trimester β-hCG and PAPP-A in the three groups is shown in Table 4. The MOM value of β-hCG was lower in the LFF and RLFF groups than in the control group (Table 4, Figure 3(A)). However, the MOM value of PAPP-A did not differ significantly between the three groups (Table 4, Figure3(B)).

Figure 3. Boxplots of serological markers and pregnancy outcomes associated with LFF and RLFF. (A) Distribution of β-hCG in the three groups. (B) Distribution of PAPP-A in the three groups. (C) Distribution of gestational age at delivery in the three groups. (D) Distribution of birth weight at delivery in the three groups. All measurement data (β-hCG, PAPP-A, Gestational age at delivery, Birth weight at delivery) are represented by median and quartile intervals (M (Q1–Q3)) (Table 4). *p < 0.05; **p < 0.01; ****p < 0.0001; NS, no statistical significance.

Table 4. Serological markers and pregnancy outcome of the LFF and RLFF cases.

	Control	LFF	RLFF	Z	p
β-hCG (MOM)^†	1.19 (0.60–2.12) (n = 314)	0.59 (0.39–1.09) (n = 127)	0.79 (0.46–1.63) (n = 51)	48.41	<0.0001
PAPP-A (MOM)^‡	1.01 (0.68–1.37) (n = 234)	0.87 (0.55–1.21) (n = 83)	0.94 (0.66–1.38) (n = 25)	5.44	NS
Gestational age at delivery (weeks)	39.3 (38.4–39.9) (n = 241)	38.9 (38.1–39.7) (n = 137)	38.6 (37.9–39.1) (n = 56)	16.60	<0.0001
Birth weight at delivery (kg)	3.33 (2.99–3.60) (n = 241)	3.36 (3.04–3.72) (n = 137)	3.28 (2.95–3.65) (n = 56)	2.10	NS

Note: ^†MOM means multiples of the median of this gestational age. ^‡All measurement data (β-hCG, PAPP-A, gestational age at delivery, birth weight at delivery) are represented by median and quartile intervals (M (Q1–Q3)).

In regard to pregnancy outcome, the delivery time and birth weight in the three groups are shown in Table 4. The delivery time was earlier in both the LFF and RLFF groups than in the control group (Table 4, Figure 3(C)), although the birthweight did not differ significantly between the three groups (Table 4, Figure 3(D)). The results of prenatal diagnosis identified 1 case of trisomy 18 in the LFF group and 3 cases of chromosomal abnormalities in the RLFF group (1 case of trisomy 21, 1 case of del 22q11.21, and 1 case of circular Y chromosome) (Table 5). Maternal diseases of the LFF cases during pregnancy are listed in Table 5. Their median maternal age with quartile intervals of the individuals in the LFF and RLFF groups who had a history of a scarred uterus was 34 (30–36) years (Table 5, footnote^#). The median maternal age with quartile intervals of the individuals in the RLFF group was 33 (29–35.5) years (Table 1). Incidence of different maternal diseases in the LFF subjects was shown in Table 5. RLFF subjects showed no statistically higher incidence of any of these maternal disorders than LFF subjects and the proportion of no diseases in RLFF was statistically lower than that of LFF (Table 5).

Table 5. Outcome of prenatal diagnosis and maternal diseases of the LFF and RLFF cases during pregnancy.

	Cytogenetic abnormalities^‡		Scarred uterus^†, #	Diabetes^†	Hypertens-ion^†	Thrombo-philia^†	Autoimmune diseases^†	Inflammation^†	No diseases^†
LFF (n = 181,16 lost to follow-up)		1 trisomy18	31/181 (17.1%)	40/181 (22.1%)	8/181 (4.4%)	13/181 (7.2%)	7/181 (3.9%)	17/181 (9.4%)	90/181 (49.7%)
RLFF (n = 78, 3 lost to follow-up)		1 trisomy21 1 del22q11.21 1 46, r(Y) (p11.31q11.2)	18/78 (23.1%)	18/78 (23.1%)	8/78 (10.3%)	4/78 (5.1%)	7/78 (9.0%)	12/78 (15.4%)	22/78 (28.2%)
p-Value			0.26	0.86	0.07	0.54	0.1	0.16	0.0013

Note: ^†The data in the column is represented as: the number of subjects with this disease in the group/total number of this group (percentage of the disease in this group). All the diseases listed in Table 5 are associate with LFF. Subjects with thrombophilia were undergoing treatment, but we are not sure about the medication and whether it was used when the NIPT was performed. The types of diseases in autoimmune diseases include Hashimoto’s thyroiditis, ankylosing spondylitis, systemic lupus erythematosus, and connective tissue disorders. The types of diseases in inflammation include active hepatitis, mastitis, fungous vaginitis, and acute pharyngitis, which may increase the number of white blood cells in blood.

^‡ Prenatal diagnosis for genetic testing (karyotyping and chromosomal microarray analysis) was performed by patient request on 1 of 197 LFF cases and 12 of 81 RLFF cases.

^# Among the individuals in the LFF and RLFF groups who had a history of a scarred uterus, their median maternal age and quartile intervals were 34 (30–36) years.

Discussion

Prenatal testing via chorionic villus sampling and amniocentesis is highly accurate in diagnosing chromosome aneuploidy, however, it has the risk of fetal loss and is generally reserved for fetuses at increased risk for genetic conditions rather than being used for large-scale detection. NIPT with high throughput is valuable in rapid and accurate prenatal screening for fetal chromosomal aneuploidy, and may be an especially attractive option for pregnant persons who undergo assisted reproduction, and/or are prone to miscarriage. The failure rate of NIPT ranges from 1.58–3.55% based on different methods [20–22]. The failure rate in our laboratory is approximately 1.5%, of which LFF accounts for one-third of NIPT failures. Identifying factors associated with LFF is essential in the clinical application of NIPT. Although many studies have reported on factors associated with FF levels [7–9], large-scale sample sizes are required to substantiate their conclusions. To obtain a large clinical data set to further investigate the factors associated with LFF, the present study was conducted on 63,883 subjects undergoing NIPT at Women and Children’s Hospital of Ningbo University (Table 1).

The results suggested that BMI was higher in the LFF and RLFF groups compared to the control group (Figure 2(A)), indicating that each incremental increase in BMI level is associated with 2.18 times the risk of LFF compared to the previous level (Table 3). Obesity is a well-known disease associated with alterations in the maternal plasma, including increased maternal total blood volume, white blood cell count, and stromal vascular apoptosis [23,24]. The increasing total blood volume decreases the FF. The increasing white blood cell count and stromal vascular apoptosis may increase the concentration of maternal-derived circulating cfDNA. Obesity is also associated with some diseases, such as diabetes and hypertension. Whereas the 5–10% general prevalence of hypertensive disorders in pregnancy may be commensurate with that observed in our LFF and RLFF groups, our observations regarding the incidence of maternal diabetes in these two study groups (22–23%) is much higher than the 2–10% incidence figures for pregnancies in general cited in the literature [25,26]. We speculate that high blood viscosity due to diabetes may lead to a decrease in circulating cfDNA (could attach to the vessel wall), as a result, cffDNA may also decrease accordingly.

The gestational age was lower in the LFF and RLFF1 (gestational age at first NIPT) groups than in the control group (Figure 2(B)). This result was consistent with many previous studies [9,27–29] and indicated that the risk of LFF decreased by 24% for every 20 days of increased gestational age (Table 3). However, the gestational age between the control and the RLFF2 (gestational age at second NIPT) groups did not differ significantly (Figure 2(B)), indicating that at least one-third of the LFF cases are caused by what we have termed as “long-lasting factors” rather than early gestational age.

One such long-lasting factor may be IVF. The utilization rate of IVF was higher in the LFF group than in the control group and even higher in the RLFF group compared to the LFF group (Figure 2(E)). The risk of LFF with the use of IVF was 2.68 times higher than that of spontaneous pregnancy, further increasing the risk of RLFF by 155% (Table 3). IVF is an independent predictor of a low FF and is independently associated with test failure [30]. Some studies suggest lower FF in IVF-mediated pregnancies to be associated with variations in placental serum markers, such as PAPP-A and β-hCG [16,17]. These two markers reflect the development of the placenta, which would thereby reflect the FF. To further investigate the correlation between PAPP-A, β-hCG, and the LFF, we calculated the MOM values of first trimester PAPP-A and β-hCG in the three groups, although these two markers were not assessed in all the subjects. The results showed lower β-hCG in both the LFF and RLFF groups compared to the control group (Figure 3(A)), while PAPP-A was not significantly different between the three groups (Figure 3(B)). Moreover, a significant correlation has been established between low β-hCG and trisomy 18 in serological screening of fetal aneuploidy. Our follow-up identified a case of trisomy 18 in the LFF group (Table 5), in keeping with studies that have shown LFF to have an increased risk for common fetal aneuploidies [31].

Another long-lasting factor might be maternal age. Several studies have demonstrated that FF is negatively correlated with maternal age [9,28,29];no correlation was detected between FF and maternal age in another study [32]. The current study’s results found the median maternal age was higher in the RLFF group than in the control and LFF groups (Figure 2(C)), and the probability of RLFF increased by 54% for every 5 years of age in the LFF population (Table 3).

Our analysis revealed that the associated between LFF and a scarred uterus could be influenced by the association between maternal age and LFF. Many advanced age pregnant patients have experienced a cesarean section or uterine surgery. Among the individuals in the LFF and RLFF groups, 49 had a history of a scarred uterus, and we analyzed their median maternal age and quartile intervals to be 34 (30–36) years (Table 5, footnote^#), which is consistent with that of the RLFF group (33 (29–35.5) years). We speculated that a scarred uterus may affect the implantation of the placenta, and perhaps presents a barrier to the release of apoptotic trophoblast across to the maternal circulation, thus reducing the source of cffDNA and decreasing the FF.

In addition to these several long-lasting factors, our method evaluated FF based on neural network model and calculated the FF of twin pregnancies. The results showed that the proportion of twin pregnancies in both the LFF and RLFF groups was higher than that in the control group (Figure 2(D)). The risk of twin pregnancy leads to 1.62 times higher chance for LFF than singleton pregnancy (Table 3) but the underlying reason for this is not entirely clear. The FF per twin has found to be lower in some studies [15,33] and higher in others [34]. In the case of IVF, most of the twins in the LFF and RLFF groups are dizygotic (DZ). It can be deduced that the FF of DZ twins is lower than the monozygotic twins and singleton pregnancy and IVF may be responsible in part. In considering the known association between twin pregnancy and IVF, we also analyzed whether IVF-twin pregnancy was independent of IVF in affecting FF. After removing the 859 twins of IVF from the three groups, the increase of OR implicates twin pregnancy as an independent factor causing a low fetal fraction and IVF-singleton had more of a chance than IVF-twin pregnancy to result in a low fetal fraction (Table 3, footnote^†). After removing the factor (twin pregnancy) from the multiple logistic regression, there was an increase in the OR for IVF, indicating that IVF affects FF independently of twins (Table 3, footnote^‡).

In addition to the main factors mentioned above, other factors, such as some diseases during pregnancy, may also reduce FF. Thrombophilia increases blood viscosity, reducing FF. Autoimmune diseases, such as Hashimoto’s thyroiditis, systemic lupus erythematosus, and Sjögren’s syndrome, may lead to the death of the patient’s own cells which then release DNA, thus increasing the maternal cfDNA at the expense of cffDNA. Inflammation can elevate the white blood cell count, increasing maternal cfDNA. These maternal health factors appeared to account for 20.5% of the LFF cases and 29.5% of the RLFF cases (Table 5). RLFF subjects showed no statistically higher incidence of any of these maternal disorders than LFF subjects (Table 5). One possibility is that the number of cases is insufficient and another possibility is that these maternal diseases do not increase the chance for a second LFF result. In terms of the number of persons without maternal diseases, the proportion of RLFF was statistically lower than that of LFF (Table 5), the clinical significance for which is unclear.

LFF not only signifies the failure of NIPT but is also a risk factor for premature birth in pregnant women [18] as well as a risk factor for aneuploidy (trisomy 13, trisomy 18, digenic triploidy) as shown by many studies [31,35]. The current results showed that the median gestational age at delivery of the LFF and RLFF groups was earlier than that of the control group (Figure 3(C)), which while statistically significant was not a clinically significant difference, and the birth weight of the newborns was not statistically significant (Figure 3(D)). Thus while we speculate that a correlation between LFF and preterm birth might be associated with placental function, large-scale retrospective studies to substantiate this are lacking.

That pregnant persons with LFF, and especially LFF twice should be offered prenatal diagnostic testing, is borne out by the current study. Our fetal cytogenetic abnormalities, furthermore, were not confined to the three common chromosomal aberrations known to have small placentas: trisomy 13, trisomy 18, maternal triploidy [31]. We identified three cytogenetic abnormalities in the RLFF group (1 case of trisomy 21, 1 case of del 22q11.21, and 1 case of circular Y chromosome) (Table 5). In keeping with the concept that there are not just a few specific chromosomal abnormalities associated with LFF, we recommend that RLFF subjects be offered prenatal diagnostic testing.

Taken together, this study provides clinically valuable data for utilizing and interpreting results with LFF. Nevertheless, the present study has some limitations. For example, we did not consider the impact of race nor of different laboratory methods in evaluating FF. Additionally, the information on β-hCG, PPAP-A, gestational age at delivery, birth weight at delivery and maternal diseases during pregnancy could not be fully obtained from the control group due to technical and quantitative reasons; hence, only a sufficient number (more than twice as much as that of the LFF group and the RLFF group), selected randomly (from the integer multiples of the serial number every 200) comprised the control group for these factors of serologic markers, gestational age and birth weight shown in Table 4, and the data for the control group is lacking in Table 5.

Conclusion

High BMI and IVF are two major factors associated with LFF while earlier gestational age and twin pregnancy represent factors less strongly associated with LFF. Advanced maternal age and IVF conception are factors, which appear to be independently associated with increased risk of having a second LFF result. MOM of β-hCG appears to be lower in LFF cases. Also, there are certain maternal diseases during pregnancy that appear to be associated with LFF, which likely contributes to there being some risk for preterm delivery and are deserving of further study.

Ethical approval

The current study was approved by the IRB of Women and Children’s Hospital of Ningbo University (No. EC2020-048). Signed informed consent was obtained from all participants in this study.

Acknowledgments

We thank BGI Genomics Co., Ltd for their technical guidance and all participants in this work.

Disclosure statement

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

Data availability statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.

Additional information

Funding

This study was supported by grants from the Social Development Public Welfare Foundation of Ningbo [No. 202002N3150, 2022S035], Innovation Project of Distinguished Medical Team in Ningbo [No. 2022020405], First Municipal Medical and Health Brand Foundation of Ningbo [No. PPXK2018-06], and Ningbo Key Research and Development Program [No. 2023Z178].