Screening of placenta accreta spectrum disorder using maternal serum biomarkers and clinical indicators: a case–control study

Placenta accreta spectrum (PAS) disorder is defined as abnormal trophoblastic adhesion or invasion of the placenta into the myometrium of the uterine wall, and it is a major cause of postpartum hemorrhage-associated maternal and fetal death [1, 2]. A recent systematic review has reported that the prevalence rates of PAS range from 0.01% to 1.1%, with an overall pooled prevalence of 0.17%; there has been a rapid increase in the PAS prevalence in most areas of the world [3]. For instance, in the United States, the incidence of PAS was 0.19% in 2005, which was eight-fold higher than that in the 1970s and five-fold higher than that in the 1980s [4]. In Europe, the incidence of PAS between 2003 and 2010 increased from 0.165% to 0.237% in women with a prior caesarean delivery (CD) [5]. In mainland China, the reported prevalence of PAS ranges from 0.26% to 0.80% [6].

In clinical practice, optimal management of PAS involves a standardized approach with a comprehensive multidisciplinary care in a center of excellence, and delivery at 34 to 36 weeks of gestation is strongly recommended [7]. Because there are no particular symptoms before onset of labor or bleeding in women with PAS, the primary antenatal diagnostic methods in clinical practice include obstetric ultrasonography and magnetic resonance imaging (MRI) [1]. However, due to the heterogeneous nature of PAS and the expertise and expense required for an imaging diagnosis, at least a half to two thirds of PAS cases remain undiagnosed before delivery [8]. The resulting delay in treatment may cause serious adverse pregnancy outcomes (APO) and increased maternal mortality [7].

Currently, there are no clinical serum biomarkers for prenatal PAS screening. Recent studies have reported potential biomarkers, such as maternal serum α-fetoprotein [9], pregnancy-associated plasma protein A [10], pro B-type natriuretic peptide [11], free β-human chorionic gonadotropin mRNA [12], and total placental cell-free mRNA [13]. The placental tumor necrosis factor-related apoptosis-inducing ligand receptor 2 [14] and circulating levels of VEGF, soluble fms-like tyrosine kinase 1 (sFIT-1) [15], median antithrombin III, median plasminogen activator inhibitor 1, soluble Tie2, and soluble VEGF receptor 2 were also found to be associated with PAS [16]. In addition, maternal VEGF levels were shown to inversely correlate with the clinical degree of invasive placenta [17]. However, due to limited sample sizes and variable reliability among different studies, few of these biomarkers have entered further investigation or clinical application [18]. In this study, we used high-throughput immunoassay to screen serum biomarkers, built screening models which were validated in two case–control studies, and the aim of current study was to develop new methods for prenatal PAS screening using serum biomarkers and clinical indicators.

Clinical studies on PAS are much less common than those on other pregnancy complications such as PE and GD. In limited PAS studies reported so far, there is a large degree of heterogeneity between the studies because of inconsistencies in diagnostic criteria and lack of histopathological confirmation [3]. In this study, we established screening models for PAS using maternal serum biomarkers and clinical indicators in a case–control study, and further validated these models in an independent group of women with high risk for adverse pregnancy outcomes. The results demonstrated that they could effectively screen PAS in both cohorts, and the changed serum proteins were highly associated with abnormal placentation in PAS.

Our study firstly demonstrated that the combined model was more suitable for PAS screening. As PAS is a complex disease without easily distinguished symptoms, the discovery of a single marker that can diagnose PAS before delivery is highly unlikely [7]. In this study, in order to enhance the efficiency of screening, we calculated the risk score for PAS using models with multiple factors rather than using any single biomarker. The results demonstrated that the model combining serum biomarkers and clinical indicators highly improved specificity and sensitivity (Table 3). For instance, in Cohort one, the percentage of positive screening using clinical indicators was 74.1%, while those of single serum biomarkers were 61.4, 65.9, 60.6, and 64.4%; however, the combination of clinical indicators and serum biomarkers yielded the total percentage of 86.2% (unpublished data). In all subjects, with combined markers in a logistic regression model, the AUC increased from 0.85 to 0.93, suggesting a good sensitivity and specificity for PAS screening (Table 3 and Fig. 3). In addition, the results demonstrated that this non-invasive method had higher positive rates of PAS screening compared with current methods such as prenatal hemorrhage and obstetric ultrasonography, which may help to improve the detection rates of PAS in clinic (Table 4).

Although the etiology of PAS remains largely unknown, this study has proven the critical roles of cell invasion, angiogenesis, and coagulability regulation in the onset and development of PAS. Previous studies have suggested that trophoblastic cells constitutively produce MMPs, and the regulation of MMP activity at the maternal–fetal interface is critical for successful implantation and placentation [20]. As a result, aberrant MMPs expression has been found in the gestational diseases fetal growth restriction (FGR) and PE, and MMPs expression increased in PAS patients compared with those with normal pregnancies [21]. Our studies showed that MMP-1 increased in the placenta of PAS patients (Fig. 2), suggesting that the invasion of placental villi into the uterine wall was unique to PAS patients. In addition, invasive extravillous cytotrophoblasts (EVCTs) can excrete angiogenic factors such as angiogenesis factors, including EGF and VEGF; the maternal decidua, defects of which are a major contributing factor to PAS formation, can also express such factors [22]. Indeed, our studies showed that EGF and VEGF-A were highly increased in PAS placenta (Fig. 2), proving that angiogenesis is important to placental invasion, especially in the uterine muscle at the location of villus invasion area and adjacent placental villi. Moreover, normal pregnancy is a state of hypercoagulability with diminished fibrinolytic activity, which is associated with an increase in plasminogen activator inhibitor type 1 and decrease in tPA [23]. Our data also showed that free tPA decreased in circulating plasma in PAS cases (Fig. 2), which might be linked with hemorrhage throughout pregnancy of PAS.

Our study also suggested that inconsistency largely existed in PAS biomarker discovery. For instance, a proteomic study reported that median antithrombin III, median plasminogen activator inhibitor 1, soluble Tie2, and soluble VEGF receptor 2 were significantly dysregulated in PAS compared with controls [16]. In this study, we did not find exactly the same proteins among the discovered biomarkers, but tPA and VEGF-A (Fig. 2) belong to the same protein family of plasminogen activator inhibitor 1 and VEGF receptor 2. In addition, it was reported that maternal serum VEGF can help in predicting abnormally invasive placenta and hint at the degree of invasion [17]. We found that a similar protein, VEGF-A, significantly increased in serum of PAS women compared with normal pregnancy (Fig. 2). We assumed that genetic backgrounds may have contributed to the observed inconsistence. For instance, East Asian women usually have small stature and the pre-pregnancy BMI ranged from 19 to 22 in our study (Tables 1 and 2), which is lower compared with 23 to 28 in White women [18]. In addition, in a number of previous studies, control groups enrolled pregnant women who had a preterm labor [16, 17], while we used the normal term controls in this study. Although the placentation may be normal in the preterm women, it is likely that they also had altered immunity and were different from normal term controls. Finally, screening time may affect the study results. As the structure of human placenta is established around the third week of gestation [24], the inner third of the myometrium is not fully invaded until at least 24 weeks [25]; therefore, the serum biomarkers may be different during different time stages given that PAS may occur at different time points. To diminish such effects, we added blood sampling weeks as a parameter in our model (Table 3).

For prenatal PAS screening, the ultrasound imaging is particularly important, especially for women with previous cesarean scar and at high risk of altered placentation [26]. However, ultrasound screening for placentation is not routinely performed in low-income area, and ultrasound imaging sometimes have limited detection on posterior placenta and twin pregnancies [27, 28]. The serum biomarkers will be of greatly helpful on these women, as its low price, high operability and capability for high-throughput screening make it easily to apply in a large population or low-income area. In future, these serum biomarkers might be developed to use with ultrasound imaging for PAS screening prenatally in a model similar to that used for aneuploidy screening in clinic. However, as this is a study in a single center in Southern China, selection biases are unavoidable. In addition, its applicability is limited because only Asian women were enrolled with small sample size. In future, large clinical trials of these biomarkers in different medical centers are warranted, and the screening of serum biomarkers in the first trimester, the conjunction of ultrasound markers, clinical markers and serum markers, and the application of biomarker screening in twin-pregnancies might be especially important for future clinical utility.

Based on two case–control studies, we discovered that the combination of serum cytokines and clinical indicators could be a good model for PAS screening. Compared with the current prenatal PAS screening techniques, this method is convenient and inexpensive, with high sensitivity and specificity. Therefore, our research might help to develop a convenient, fast, and economic method for prenatal PAS screening, which would facilitate clinical PAS management and decrease the rates of maternal mortality.