Upon implantation, fetal trophoblast cells start invasion and migration into the maternal decidua and these trophoblasts form masses of cytotrophoblasts (CTBs) and primitive syncytiotrophoblasts (STBs). The distal CTBs located at the tips of anchoring villi form columns that merge together, creating shells that encapsulate the developing gestational sac (GS). The CTBs located on the outer surface of the shell change their proliferative to invasive potential. This invasive potential helps these cells to migrate beyond the decidual stroma. These trophoblasts are called “extravillous trophoblasts” (EVTs). EVTs share some key similarities with carcinomas in their ability to invade healthy tissues, form new vessels, and promote immunotolerant environment [11, 12, 13].

EVTs differentiates into 3 subpopulations: (1) endovascular subpopulation, that replaces endothelial cells of uterine vessels; (2) endoglandular subpopulation, which invades endometrial glands and opens their lumen towards the intervillous space; and/or (3) interstitial subpopulation, that invades into the decidual interstitium, thereby anchoring the placenta to the uterus [14], as shown in Fig. 1.

Fig. 1.

Feto-maternal unit. (1) Floating villous, (2) anchoring villous, (3) extravillous trophoblasts (EVTs), (4) endovascular EVTs making trophoblastic plugs, (5) spiral artery, (6) giant cells, (7) interstitial EVTs, (8) endoglandular EVTs, (9) decidual cells, (10) decidual natural killer cells.

Invasion of EVTs to the endometrium presents a lot of benefits. This invasion helps in placental fixation, remodeling of spiral arteries to maintain good blood supply, and contributes to immunological defenses (tolerogenic responses). The interstitial EVTs migrate into the endometrial stroma until they reach the inner third of the myometrium, and at this level they fuse together forming multinucleated trophoblast giant cells [14], as shown in Fig. 1.

Spiral arteries remodeling passes into 2 waves. The first wave (6–12 weeks) involves trophoblastic invasion into spiral arteries to form EVTs plugs (Fig. 1). These plugs prevent the maternal blood flow into the intervillous space at this early gestation. These plugs make a low oxygen environment with O${}_2$ levels of 1–5%. During the second wave (16–18 weeks), the EVTs plugs gradually dissolve and endovascular EVTs start to migrate proximally along the spiral arteries, leading to spiral artery dilation. These changes result in the development of a well-established maternal–placental circulation. Furthermore, the O${}_2$ levels rapidly increase to 8–10% to support fetal growth [15].

There are some factors that promote trophoblastic invasion, while other factors limit this invasion. Factors promoting invasion include factors facilitating degradation of extracellular matrix (serine proteases, proteinases, and collagenases), immune system suppression, angiogenic factors and hypoxic states. Plasminogen activators convert plasminogen into plasmin that enhances extracellular matrix degradation. Trophoblastic invasion is limited by the inhibiting factors. These inhibitory factors are produced by decidual cells and include plasminogen activator inhibitor-1 (PAI-1) that inhibits the activity of urokinase-type of plasminogen activator (uPA). Another inhibitory factor is the tissue inhibitor of metalloproteinases (TIMPs) which regulate the activities of metalloproteinases (MMPs). Decidual transforming growth factor-$\beta$ (TGF-$\beta$) has a major role in limiting trophoblastic invasion. TGF-$\beta$ acts by upregulation of both PAI-1 and TIMPs [16].

In the current study we started by identifying the pathological features that occur in the great majority of PAS disorders. The pathological findings in PAS disorders were myometrial scarification and thinning (100%), myofiber disarray (100%), tissue edema (100%), deeply implanted villi with no villi crossing the entire thickness of the uterine wall (88.1%), dense fibrinoid deposits with a varying thickness from 0.5 to 2 mm at the utero-placental interface (74.4%), and basal plate myofibers [17, 18].

The clinical situations met during management of PAS disorders arise from understanding pathological issues. Serosal invasion was attributed to myometrial defects, extensive scarring, and failure of normal decidualization, which bring the anchoring villi close to the surface of the uterus. The damaged endometrium at the site of implantation allows the EVTs to reach the radial and/or arcuate arteries, and remodeling of these large vessels results in major hemorrhage at attempt of placental separation. The extensive adhesion between placenta and uterine wall was attributed to the presence of excessive fibrinoid deposition at the uteroplacental interface [18].

The clinical risk factors evaluated include maternal characteristics, such as maternal age, body mass index (BMI), parity, previous uterine surgery, previous cesarean delivery, number of cesarean sections, pregnancy after in vitro fertilization (IVF), and history of medical disorders, such as diabetes mellitus and hypertension [19, 20].

Kyozuka et al. [21] conducted a nationwide cohort study in Japan on 90,554 participants, treated in 15 regional centers, during the period of 2011 to 2014. They found that the most risk factors for occurrence of PAS were placenta previa, conception by assisted reproductive technologies (ART), and smoking during pregnancy [21].

Imafuku et al. [22] conducted a prospective cohort study on 4146 pregnant women whom experienced PAS. They found that history of prior cesarean delivery, dilatation and curettage, hysteroscopic surgery, uterine artery embolization, current pregnancy via ART, and the presence of placenta previa in the current pregnancy are high risk for PAS [22].

Ogawa et al. [23] conducted a multicenter cross-sectional study on 472,301 women with singleton deliveries, in the period from 2013 and 2015. They reported that the most powerful predictor for PAS was the number of previous cesarean sections. On the other hand, in absence placenta previa or previous cesarean delivery, conception through ART was a strong predictor of PAS [23].

Smoking is another major risk factor studied for the pathogenesis of PAS. However, the relation between maternal smoking and PAS was not proven [24]. Furthermore, other study attributed this association to placental inflammation from smoke pollutants [25].

D’Antonio et al. [26] conducted a meta-analysis to predict PAS disorders in the first-trimester of pregnancy. They used three signs suggestive of PAS in the first-trimester. The signs were implantation of the GS in lower uterine segment, reduced myometrial thickness, and lacunae. They found that when at least one sign of PAS was found, the presence of PAS was confirmed in 91.4% of patients [26].

Calìet al. [27] investigated the role of early first-trimester ultrasound (5–7 weeks) in prediction of PAS in third-trimester. They evaluated three signs: cross-over sign (COS-1) [27] and site of implantation of GS, either on the scar or in the scar niche [28], and position of the GS below or above the midline of the uterus [29]. They found that early ultrasound could predict the stage of PAS later on, where implantation of the GS in the niche and below the uterine midline occurs, with high predictive accuracy for PAS-3. Moreover, adverse surgical outcomes were linked to implantation in the niche and below the midline of the uterus, with 95.5% and 90.9% reliability, respectively. They concluded that early first-trimester ultrasound in women with previous cesarean delivery can predict reliably PAS disorders [30].

Accordingly, Doulaveris et al. [31] assessed and graded the suspicion of PAS disorders in the first-trimester (11 to 14 weeks) in patients with one or more cesarean delivery. They found that low implantation inside scar niche had a positive predictive value of 100%, and a negative predictive value of 99.4%. They concluded that early transvaginal (TV) ultrasound (11–14 weeks) in patients delivered by cesarean section can predict PAS disorders [31].

Another predictor of PAS and its surgical outcomes was intracervical lakes (ICL), which is a new ultrasound sign presented by di Pasquo et al. [32]. They examined risky patients with placenta previa or low-lying placenta at $\ge$26 weeks, in 332 patients. They found that ICL could predict major postpartum hemorrhage, cesarean hysterectomy, and placenta percreta. They concluded that ICL could be used as a marker of deep villus invasion. It was linked to poor maternal outcomes and severe morbidity [32].

Yule et al. [33] conducted a retrospective study to evaluate the role of TV ultrasound color cines obtained in the first-trimester in prediction of PAS and its related morbidity. They found that increased color pixel area in the interface of bladder and uterus was noticed in women requiring peripartum hysterectomy [33].

The scoring systems identified combine different characters and then calculate a score that could predict PAS disorders. The characters included were demographic information, clinical characteristics, and sonographic findings. One of the suggested scores included 67 cases and found that their scoring model could predict severe postpartum hemorrhage, and hysterectomy [34].

Another risk model was introduced by Sargent et al. [35], where only four ultrasound markers (loss of clear zone, placental bulge, abnormal placental lacunae, and bladder wall interruption) were used at 17–36 weeks to predict PAS. They reported that their new model has a high accuracy [35].

The basis of biomarkers detection is the understanding of promoting and limiting factors in placental invasion. It was noted that a normal pregnancy is the balance between promoting and limiting factors of placental invasion. In cases of excessive invasion (which occurs in PAS), factors promoting invasion are increased while decrease in the limiting factors. On the other hand, decreased placental invasion (which occurs in preeclampsia), factors limiting invasion are more predominant [39, 40].

4.4.3 Angiogenic Factors

It is well known that PAS disorders are associated with excessive neovascularization on the surface of lower uterine segment and at bladder-uterine interface. These new vessels are upregulated by angiogenic factors like vascular endothelial growth factor (VEGF), antithrombin III (ATIII), and other angiogenic factors. Moreover, it is downregulated by soluble Fms-like tyrosine kinase 1 (sFlt-1), TGF-$\beta$, and tumour necrosis factor-$\alpha$ (TNF-$\alpha$) [52]. These factors are shown in Fig. 2.

Fig. 2.

Upregulation and downregulation of angiogenic factors. VEGF, vascular endothelial growth factor; PlGF, placental growth factor; sFlt-1, soluble Fms-like tyrosine kinase 1; TGF-$\beta$, transformation growth factor–$\beta$; TNF-$\alpha$, tumour necrosis factor-$\alpha$.

Biberoglu et al. [52] studied angiogenic factors in their case–control study. They reported no statistical difference in the levels of sFlt-1, placental growth factor (PlGF), sFlt-1/PlGF ratio, and VEGF in PAS, compared to controls [52]. On the other hand, Zhang et al. [53] found that sFlt-1 levels and the sFlt-1/PlGF ratio were lower in PAS group than in the controls, while PlGF levels in the PAS were higher than in the controls. They concluded that sFlt-1, PlGF, and sFlt-1/PLGF ratio could differentiate between patients with PAS from those with placenta previa [53].

Machine learning models are electronic models designed to reduce the workload and mental stress of obstetricians managing PAS disorders. These electronic models could detect the amount of bleeding during PAS surgery, improve decision-making, and help in choosing the best lines of management. Pictures of ultrasound or MRI and plasma biomarkers are translated accurately by these electronic models and could yield accurate results [62].

Liu et al. [62] designed models to predict the amount of blood loss during cesarean delivery, using MRI images translated by DeepLab-V3 network and Visual Geometry Group Network-16. They reported an accuracy of 75.61% [62]. Accordingly, Akazawa et al. [63] designed two machine learning models using MRI images to predict the amount of blood loss in PAS surgery.

Andreasen et al. [64] developed a model using ultrasound images translated by convolutional neural networks (CNN), with an accuracy of 81%. Yaşar et al. [65], used XGBoost electronic algorism and Adaboost models with 10 folds accuracy. Shazly et al. [66] used two machine learning models in the prediction of peripartum outcomes related to PAS disorders. Lastly, combined models using ultrasound and MRI images were used by An et al. [67].